From drones to geospatial analysis

From Drones to Geospatial Data Dr. Jean A. Doumit Reviewed by Prof. Dr. Anatoly V. Pogorelov Lebanese University Kuban State University Acknowledgements This book is dedicated to all Geomatics Engineers and Earth Scientists, the following organizations are thanked for permission to use photographs and data: The Lebanese army, the Lebanese University, Kuban State University, SOUTH company for surveying Instruments and Aerogeomatica. Three must be singled out for special mention: Professor Pogorelov Anatoly Valerievich as the best scientific advisor, the department of Geoinformatics at Kuban State University, and the department of Geography at the Lebanese University. I would also like to express my gratitude to all who helped me take this book from a vision to an actualization, especially my parents for their constant support. Finally, my Very Special Thanks go to my dear wife, for her forbearance in accepting the disruption of our social life and for her encouragement, feeding and watering me, during the preparation of this book. 2 Preface Unmanned aerial vehicles were once the stuff of rumor and legend, identified as new and Mysterious robots in the sky. Apart from military applications there are many jobs to be performed monitoring and rescue, the unmanned aerial vehicles are used for photogrammetry tasks. This new technology been the case across various disciplines such as mapping, remote sensing, civil engineering, geology, geomorphology, military engineering, land planning, and communications. It is encouraging that more literature is now available in this discipline. This book was written with the intention to add to and benefit this field of study. I am pleased to present you with this 1st edition of this book. It introduces the classes and types of platforms available, and examines their performance, structure, propulsion and systems to be suitable for photogrammetry. Sensor and payloads, methods of launch and recovery, image processing and geospatial analysis. In its nine chapters, this book carries out several tasks: Drones field of uses, classification of drones, technical structures of drones, drones flight planning and software, drone photogrammetry methods, data processing and geospatial analysis. The purpose of this book is to give an idea about drone for researchers and students to apply this new technology in researches and projects, and introduce them to the Unmanned Aerial Vehicles new era leading to a treasure of information. My message to the readers is: “Use new technology”. 3 Table of Contents Chapter 1 Introduction to Photogrammetry ............................................................................... 5 What are a UAV and UAS? ................................................................................................... 7 Description of a drone ............................................................................................................ 8 Drones field of Uses ............................................................................................................... 9 Chapter 2 Advantage and limitation of Drones ........................................................................ 11 Advantages of drones ........................................................................................................... 11 Limitations in the use of drones ........................................................................................... 12 Chapter 3 Classification of drones ........................................................................................... 13 Fixed wings .......................................................................................................................... 13 Rotary systems ..................................................................................................................... 14 Chapter 4 Drones structures ..................................................................................................... 19 Unmanned Aerial Vehicles (UAV) ...................................................................................... 19 Unmanned Aircraft System (UAS) ...................................................................................... 21 Chapter 5 Drone Software’s ..................................................................................................... 26 Drones flight planning software ........................................................................................... 26 Drones mapping software..................................................................................................... 28 Processing workflow ............................................................................................................ 30 Chapter 6 Drones photogrammetry methods ........................................................................... 34 Type of photogrammetry ...................................................................................................... 34 Close range photogrammetry (CRP) .................................................................................... 34 Distortions and Camera calibration ...................................................................................... 38 Bundle adjustment ................................................................................................................ 41 Structure from Motion (SFM) .............................................................................................. 43 Aerotriangulation ................................................................................................................. 46 Chapter 7 Flight Planning Principles and image processing .................................................... 48 Ground sampling distance (GSD) ........................................................................................ 48 Photocontrol points .............................................................................................................. 49 The Global Positioning System (GPS) ................................................................................. 51 Flight planning ..................................................................................................................... 53 Image processing .................................................................................................................. 55 Drone mapping accuracy ...................................................................................................... 57 Ground Sampling Distance (GSD) accuracy ....................................................................... 58 DSM Accuracy Assessment ................................................................................................. 58 Chapter 8 Drones and Geospatial Analysis .............................................................................. 59 Multi-scale Digital Surface Models ..................................................................................... 61 Multi-scale landforms classification .................................................................................... 68 Terrain analysis for parcels evaluation................................................................................. 77 Multi-scale Terrain Roughness ............................................................................................ 83 Digital Surface Models to planar areas. ............................................................................... 89 Chapter 9 Drones regulations and buyers guide....................................................................... 98 Drones Regulations .............................................................................................................. 98 Flight safety .......................................................................................................................... 98 Drones buyers guide ........................................................................................................... 100 Conclusion .............................................................................................................................. 101 References .............................................................................................................................. 102 4 Chapter 1 Introduction to Photogrammetry This book provides a general overview of drone photogrammetry, its theory and general working principles with an emphasis on concepts not on mathematical formulas and equations. Photogrammetry is an engineering discipline and as such heavily influenced by developments in computer science and electronics. Photogrammetry, from the Greek photo (light writing) gram (graphics) metry (measure), it is “the art, science, and technology of obtaining reliable information about physical objects and the environment, through processes of recording, measuring, and interpreting images and patterns of electromagnetic radiant energy and other phenomena”. Photogrammetry produces maps directly from photographic images by identifying, symbolizing, and compiling elevation, cultural, and natural features that are visible on the imagery. Each specialist use photogrammetry for his purposes:  A soils specialist may be looking for erodible soils on a proposed highway route.  A forester may be estimating the volume on a timber tract.  A hydrologist may be comparing the degree of suspended matter in several adjacent lakes.  A wetland specialist may be monitoring the decline of wetland areas within a state.  A glaciologist may be charting the movement of a glacier, etc. Several types of photogrammetry exist: aerial, terrestrial, and close range. Each serves the needs of a distinct category of users. Throughout the mapping community, terrestrial and close-range photogrammetry has limited use. Aerial photogrammetry uses near-vertical photographic images that are exposed to a moving platform at a distant point in the sky (Falkner, Morgan, 2002). Photogrammetry and remote sensing are two related fields, the principle difference between photogrammetry and remote sensing is in the application; while photogrammetrists produce maps and precise three-dimensional objects, remote sensing specialists analyze and interpret images for deriving information about the Earth’s surface. Both disciplines are also related to Geographic Information Systems (GIS) in that they provide GIS with essential information. Photogrammetry is the science of obtaining information about land surfaces without physical contact, for the purposes of measuring and interpreting this information. Figure 1 shows a logic systems simplify understanding the photogrammetry procedure. 5 Fig.1.1: The logic system for understanding photogrammetry. Figure 1.1 shows the workflow of photogrammetry beginning from data acquisition from sensors passing by the processing step or procedures and ending by the data production. The development of photogrammetry clearly depends on the general development of science and technology. It is interesting to note that the four major phases of photogrammetry are directly related to the technological inventions of photography, airplanes, computers, and electronics. Photogrammetry had its beginning with the invention of photography by Daguerre and Niepce in 1839. The first generation, with remarkable achievements in terrestrial and aerial balloons. The developments in photogrammetry, from around 1850, have followed four development cycles. Each of these periods extended about fifty years. These cycles include: (a) Plane table photogrammetry, from about 1850 to 1900, (b) Analog photogrammetry, from about 1900 to 1960, (c) Analytical photogrammetry, from about 1960 to 2010, (d) Digital photogrammetry, which just began to be a presence in the photogrammetric industry. 6 In 1855, Nadar (Gaspard Felix Tournachon) used a balloon at 80-meters to obtain the first aerial Photograph, in 1859 Emperor Napoleon ordered Nadir to obtain reconnaissance photography in preparation of the Battle of Solferino (Gruner, 1977). The English meteorologist E.D. Archibald was among the first to take successful photographs from kites in 1882. In France, M. Arthur Batut took aerial photographs using a kite, over Labruguiere, France, in May 1888. In 1893, Dr. Albrecht Meydenbauer (1834-1921) was the first person to use the term "photogrammetry" (Meyer, 1987). In 1903, Julius Neubranner, photography enthusiast, designed and patented a breast-mounted aerial camera for carrier pigeons. 1903: Airplane invented by Wright brothers. 1909: The Wright brothers take the first photograph from a plane over Centocelli, Italy. Captain Cesare Tardivo (1870-1953) is thought to be the first to use aerial photography from a plane for mapping purposes. He created a 1:4,000 mosaic of Bengasi in Italy that was described in his paper to the 1913 International Society of Photogrammetry meeting in Vienna. First drones investigations started at Carnegie Mellon University (Eisenbeiss, 2011). (Nagai et al., 2004) used LiDAR system mounted on a drone, in 2008 they showed the latest results for multi-sensor integration on drones. After 2010 the evolution and the invention of new drones systems were very fast and put a multidisciplinary powerful scientific engine among civilian hands. What are a UAV and UAS? Unmanned Aerial Vehicles “UAVs are to be understood as uninhabited and reusable motorized aerial vehicles” (Von Blyenburg, 1999). These vehicles are remotely controlled, semi-autonomous, autonomous, or have a combination of these capabilities. The UAV is an acronym for Unmanned Aerial Vehicle, which is an aircraft with no pilot on board. UAVs can be remote controlled aircraft or can fly autonomously based on pre-programmed flight plans or more complex dynamic automation systems. The acronym UAV has been expanded in some cases to UAVs (Unmanned Aircraft Vehicle System). The FAA has adopted the acronym UAS (Unmanned Aircraft System) to reflect the fact that these complex systems include ground stations and other elements besides the actual air vehicles). Officially, the term 'Unmanned Aerial Vehicle' was changed to “Unmanned Aircraft System”. The term UAS, however, is not widely used as the term UAV has become part of the modern lexicon. 7 The use of professional civilian drones is increasing rapidly around the world and is expected to explode in the coming years. In the following chapters of our book, we will use the term "drone" instead of UAV and UAS. Description of a drone A simple drone system has two basic parts, airborne part, aircraft frame carrying the payload of the system and containing (frame, camera, battery, gimbal, etc.) and the ground-based part which constitute the operation base interface of all the system (base station and a radio transmitter) in chapter 6 we will discuss it in details. Some drones have the possibilities of automated take-off and landing without any manual operator, few drones can be functioned without a link to a Remote Control (RC) transmitter or base station. This constraint is implemented mainly as a safety precaution to ensure that manual control of the UAV can be resumed if needed at any given point during a flight. Drones are categorized into fixed wing, and helicopters. These drones are powered by brushless electric motors or by combustion engines. Electric motors cause significantly fewer vibrations in the airframe than internal combustion engines and are preferred for photographic applications. It is not very difficult to learn how to fly RC helicopters or fixed wing in a short time by it needs practice and training. Some vehicles are very easy to control and don’t need any flight experience and training because it contains automated orientation sensors and processing power keeping the platform aloft. A basic flight control system for these drones contains very specific sensors linked to a processor that manages power distribution to the motors to stabilize flight. Most flight control systems are composed of the magnetometer, a barometer, and Global Positioning Systems (GPS) receivers to support threedimensional drones navigation. In this book, we will give an overview of existing classifications, structures, compositions of these drones and their field of applications, especially in geography. The unmanned aircraft system (UAS) is a fully designed system comprises: a) A control station (CS) constituting the operation base interface of all the system. b) Aircraft frame carrying the payload of the system. c) A communication system between the CS and the aircraft and contrary, this system is achieved by radio transmission (Austin 2010). 8 Drones field of Uses Drones photogrammetry opens various new applications at a very lowcost comparing to the classical photogrammetry, it allows to capture highresolution and to process data manually or automatically. Drones are used in a numerous very huge number of fields: Agriculture, drones provide an efficient way to assess plant health, after a flight drone images are integrated into a special software for the generation of index maps showing clearly, where plants are struggling or vegetation monitoring (Sugiura et al., 2005). Lidar systems mounted on drones to measure the height of crops, multi-spectral instruments can count plants, infrared imagery can check the health of plants etc... Archeology, drones used for the detection and mapping of archeological sites and monuments (Bendea et al., 2007; Patias et al., 2007). Architecture, simultaneous as built and scanning of the projects made by drones can help in faster design and inspections, quality control, make measurements and modifications. Civil engineering, this field benefits from inspecting infrastructure, bridges, cell phone towers, dams, powerlines, solar fields and road planning and designs. Crime analysis, drones can help in crime detections and investigations. Customs, drones can surveil coastline, ports, bridges and other access points for the import of illegal substances is the remit of the customs (Surveillance for illegal imports) Electricity Companies, inspections of power-lines, damage of structures and deterioration of insulators. Environmental impact assessment, the use of drones for glacial modeling to animal’s movement study, coastal erosion, etc.… Fire Services, Fire departments can use drones to track and map wildfires or for forest fire monitoring (Zhou et al., 2005). Fisheries, the prevention of illegal fishing aided by patrolling drones Forestry, drones can take a forest control by trees counting, trees health and fire controls. Gas and Oil Companies, use drones for pipes patrolling to look for disruption or leaks in accidents. Geology, drones with specific electromagnetic sensors, can be used to gather geological information to approximate the location and presence of minerals, oil, and natural gas. Humanitarian aid, Drones are being increasingly used by NGOs and governmental organizations to respond to and assess the impact of natural disasters. For Rescue, drones with infrared sensors can be used to detect humans that is helpful in search scenarios. 9 Information Services and Mass Media, television companies, newspaper publishers could have the means of covering events, whether planned or accidental. Sports events could be covered in real-time. Land surveying, Land surveyors use drones to survey inaccessible areas like cliffs, rivers, woods etc… to create geo-referenced maps, 3D models, Digital Surface Models (DSM) and other data products. Land registry, drones have made it possible to count buildings and population growth estimation. Meteorological Services, use drones to sample the atmosphere for forecasting by integrating special sensor. Mining, Drone data can be used for taking measurements, automated workflows create orthophotos as well as 3D models that can be used to quickly calculate quantities for materials such as aggregate stockpiles or cut and all. Photomontage, drones are used for posters making and cinemas filming. Rivers Authorities have used drones successfully in monitoring watercourse flow and water levels. Rivers Authorities Water course and level monitoring (Masahiko, 2007). Traffic monitoring (Haarbrink and Koers, 2006, Puri, 2004), road following (Egbert and Beard, 2007), vehicle detection (Kaaniche et al., 2005), car accident and flare inspection of an industrial flue (Haarbrink and Koers, 2006). Geo-informatics, Drones are increasingly being used in place of satellite and photogrammetry images for the creation of up-to-date, high-resolution base maps Geospatial data collection with high geometric and temporal resolution. Below a small comparison between drones, photogrammetry and satellite images. Table 1.1: A brief comparison between satellite images, photogrammetry, and drones. 1. 2. 3. 4. 5. Space-borne Extensive coverage. Low resolution (30 cm/pixel). Timing controlled by the provider. Cloud cover. Expensive. 1. 2. 3. 4. 5. Airborne Large coverage High resolution (7cm/pixel). Timing controlled by the provider. Susceptible to weather. Expensive. Drones-borne 1. Very high resolution 2. (Fixed wing 2 cm/pixel, rotary sub-millimeter). 3. Imagery acquired on demand. 4. Susceptible to weather. 5. Unaffected by cloud cover. The difference between satellite images, ordinary photogrammetry and drones photogrammetry listed in table 1.1 could be summarized in: Satellite for small scales, photogrammetry for big scales and drones photogrammetry very big scales. 10 Chapter 2 Advantage and limitation of Drones Advantages of drones Main drones advantages compared to manned aircraft systems it can be used in high-risk situations without endangering a human life and inaccessible areas. These regions are for example natural disaster sites, floodplains, earthquake and desert areas and scenes of accidents. Furthermore, in cloudy and drizzly weather conditions, the data acquisition with drones is possible, such weather conditions do not allow to fly into manned aircraft. Moreover, drones have the real-time capability and the ability for fast data acquisition, while transmitting the image, video and orientation data in real time to the ground control station. Most of the (non-)commercially available UAV systems on the market focus on low-cost systems, and thus a major advantage of using UAVs is also the cost factor, as UAVs are less expensive and have lower operating costs than manned aircraft have. But, sometimes depending on the application the cost can be similar to manned systems. Due to the low operation altitude, UAVs achieve a very high resolution in terms of ground sampling distance and can, therefore, compete with airborne large format digital camera system (Irschara et al, 2010). In addition to these advantages, the UAV-images can be also used for the high-resolution texture mapping on existing DSMs and 3D-models, as well as for image rectification. The rectified images and derivate, like image mosaics, maps, and drawings, can be used for image interpretation. The implementation of GNSS systems as well as the stabilization and navigation units allow precise flights with sufficient image coverage and overlap and enabling the user to estimate the expected product accuracy preflight. With the huge amount of divert drone advantages, we listed it in a summarized way as: 1. Autonomous and stabilized: real-time capability. 2. Can fly at low altitude close to the objects where manned systems cannot be flown (natural disaster sites, mountainous and volcanic areas, floodplains, earthquake and desert areas etc.…). 3. Data acquisition in cloudy and drizzly weather conditions. 4. Data acquisition with high temporal and spatial resolution. 5. More economical than human pilots. 6. Providing high-resolution texture mapping on existing DSMs and 3Dmodels. 7. Real-time capability and the ability for fast data acquisition, while transmitting the image. 8. Use in high-risk situations without endangering a human life and inaccessible areas. 9. Flexibility, a drone can be launched on demand. 11 10. Timely, drones produce completely up-to-date imagery. This makes drones suitable for monitoring projects. 11. Efficient, using a drone is fast and requires minimal staff. 12. Cost-effective, the project cost of a professional drone system is typically lower than a manned imaging aircraft. 13. Discrete, electric-powered drones make a little of noises and are rarely disturbing people on the ground if they notice them at all. Limitations in the use of drones The limitations of drones compared to manned aircraft systems, first of all, is related to the sensor payload in weight and dimension, so that often low weight sensors like small or medium format amateur cameras are mounted on drones, in comparison to large format cameras, drones have to acquire a higher number of images in order to obtain the same image coverage and comparable image resolution. The drones payload limitations require the use of low weight navigation units, which implies less accurate results for the orientation of the sensors. For general mapping purposes, payload weights may vary from 200 g for a small digital camera to 3 kg for larger digital single lens reflex or multispectral cameras. Apart from the weight of the actual sensor and battery, a stabilizing gimbal mount may have to be added to the payload weight. The take-off weight of "small" UASs for mapping purposes will typically vary from around 1 kg to 5 kg. Existing commercial software packages applied for photogrammetric data processing are rarely set up to support drone images, as though no standardized workflows and sensor models are being implemented (Eisenbeiss, 2009). Based on the communication and steering unit of drones, we can state that the operation distance depends on the range of the radio link for the rotary and fixed wing, which is equivalent to the length of the rope for kites and balloon systems used in the past. radio frequencies may be subject to interferences caused by other systems (remote controlled cars and model aircraft, as well as band radios), which use the same frequencies or may suffer from signal jamming. Thus, depending on the local situation of the area of interest. The limitations in the drone uses are could be summarized in: 1. Limitations of the payload, the drone could not carry big weights. 2. Regulations and insurance, it is very dangerous to fly above a crowd and to give a child a drone guidance. 3. Use of Low-cost Sensors. 4. Short flight distances (civilian use), low battery life. 12 Chapter 3 Classification of drones Fixed wings A fixed-wing drone is generally composed of a central body, which houses all the drone's electronics, and two wings. The aerodynamic profile of the wings enables the drone, once in flight, to generate lift that compensates for the weight of the aircraft. Like an airplane, fixed-wing drones also feature ailerons, which enable the aircraft to steer. Some drones also feature a rudder and elevators, sometimeseven flaps. Fixed-wing drones typically feature one engine with a propeller attached, either a forward-mounted (tractor) propeller or a backward- facing (pusher) propeller. Most propeller-powered drones include a system that folds these components somehow especially if, an aircraft lands on its belly (http://planner.ardupilot.com). Fig. 3.1: Fixed wing drone, Sky Cruiser from SOUTH Company. Fixed wings launching techniques, the listing of different techniques for launching fixed-wing above is seen as the initial result of a concept generation. These solutions are based on the weight of each fixed wing type. For very heavy drones (military uses) the launcher must be a pneumatic catapult mounted on a truck. Big size civilian fixed wings drones are launched by catapult 4.2a, middle size drones are launched using a rubber wire, and most small and foamy drone could be throw by hands for takeoff. As with its launch, fixed wings presents the more complicated provision for recovery. There are a number of solutions: a) Skid or belly landing, b) Guided flight into a catchment net, c) Deployment in flight of a parachute. Skid or belly landing is applicable only to small drones. Guided flight into the net, this way usually comes with disastrous results as on most occasions the drones were damaged or destroyed. Launch equipment. This will be required for those air vehicles, which do not have a vertical flight capability, nor have access to a runway of suitable sur13 face and length. This usually takes the form of a ramp along which the aircraft is accelerated on a trolley, propelled by a system of rubber bungees, by compressed air or by a rocket, until the aircraft has reached an airspeed at which it can sustain airborne flight. Recovery equipment. This also will usually be required for aircraft without a vertical flight capability, unless they can be brought down onto the terrain, which will allow a wheeled or skid-borne run-on landing. It usually takes the form of a parachute, installed within the aircraft (figure 3.2), and which is deployed at a suitable altitude over the landing zone (Reg Austin, 2010). Fig. 3.2: A) catapult drone launching, b) parachute for the drone landing. Parachute deployment, this is the most usual method; it requires the drone to carry a parachute. One disadvantage of this method is that the parachute is at the mercy of the wind, and its precise point of touchdown may, therefore, be unpredictable figure 3.2b. Rotary systems Rotary drones, also called multi-rotors, are more complex systems than fixed-wing. A rotary system moves through the air by varying the power supplied to the different propellers. This determines their revolutions per minute (RPM) and therefore the thrust these generate. Two key technical challenges must be met to optimize a rotary system's performance. First, to ensure a stable flight they need highly advanced autopilot technology to continually choose the correct RPM for the different propellers; making hundreds of tiny adjustments per second. Secondly, rotary systems require very fine motor control to rapidly vary the power sent to different motors, ensuring the propellers achieve the exact RPM commanded by the autopilot. In the case of a four-propeller system quadcopter model, diagonal pairs of propellers spin in opposite directions. 14 A quadcopter climbs and descends by simultaneously varying the RPM, and therefore the thrust, of all four of its propellers. creates a force allowing the drone to climb (ascend), hover or descend. Flying forwards or backward. To pitch forwards, the RPM of the front two motors is decreased compared to that of the back pair. With the drone pitched forwards, the combined thrust of its four motors is also pitched forwards, pushing the drone in that direction. Flying sideways, or rolling, is a case of increasing the RPM of two motors on the same side; increasing the power to the rotors on the left, while decreasing it to the right, will 'roll'and move the aircraft to the right. While different multi-rotors feature different numbers of propellers, these basic flight concepts remain the same. Table 3.1: Questionnaire for fixed wing and rotary system comparison. Coverage Take-off/landing area Object resolution Mechanics Projects Fixed wing Large Large cm per pixel Relatively simple Mapping Flight times and wind resistance High Rotary systems Small Very small mm per pixel Complex Small area mapping and inspection Low The comparison between fixed wing and rotary drones of table 3.1 gives an idea for the user to choose which type to use depending on project size, a fixed wing for big projects (global) and a rotary system for small projects (local). The rotary system drones come with different types such as tricopter, quadcopter, an octocopter. In this section, we will discuss the description, advantages, and disadvantages of each type. Tricopter. From its name, we can understand that it is constituted from three arms, each connected to one motor figure 3.3. The front of the drone tends to be between two of the arms, the rear motor normally needs to be able to rotate (using a normal RC servomotor). 15 Fig. 3.3: Tricopter unmanned aerial vehicle. Advantages: It flies more like an airplane in forwarding motion. Disadvantages: Since the copter is not symmetric, the rear arm is more complex since a servo needs to be mounted along the axis to assure a motor rotation. Most, though not all flight controllers support this configuration. Quadcopter. A quadcopter with four arms each connected to one motor figure 3.4. The front of the drone tends to be between two arms like an (x), but can also be along an arm like a (+). Fig. 3.4: Quadcopter with four arms and motors. Advantages: Most drones in the market are quadcopters, very simple the arms/motors are symmetric. All flight controllers are suitable for this multirotor design. Disadvantages: There is no redundancy, so if there is a failure anywhere in the system, especially a motor or propeller, the craft is likely going to crash. Hexacopter. Six arms connected to six motors figure 3.5, the front tends to be between two arms, or also in one arm with the form of (+). Advantages: hexacopter can lift more payload than other kinds of copters. If a motor fails, there is still a chance the copter can land rather than crash. All flight controllers support hexacopters configuration. 16 Fig. 3.5: Hexacopter with six arms and motors, Sky Walker X61 from SOUTH Company. Disadvantages: hexacopters have more parts (motors, ESC and propellers) then quadcopter and this makes it more expensive these additional motors and parts add weight to the copter, so in order to get the same flight time as a quadcopter, the battery needs to be larger (higher capacity) as well. Y6 Hexacopter. This type of hexacopter have three arms instead of six, it has three with six motors connected to either side of the armed figure 3.6. Fig. 3.6: Hexacopter Y6 with three arms and six motors. Advantages: this copter can lift more payload as compared to a quadcopter and not have the same issue as a tricopter as it eliminates the gyro effect using counter-rotating propellers. In addition, in the case of a motor fail, there is still a chance the copter can land rather than crash. Disadvantages: Additional motors and parts additional weight, the battery needs to be larger (higher capacity) as well. Octocopter. Comes with eight arms, with eight motors. The front of the UAV tends to be between two arms figure 3.7. 17 Fig. 3.7: Octocopter eight arms and eight motors. Advantages: More motors, more thrust, as well as increased redundancy, can carry a big weight. Disadvantages: More motors, higher price, and larger battery pack. Octocopter X 8. This is a quadcopter on each arm two motors a total of eight motors figure 3.8. Fig. 3.8: Octocopter with four arms and eight motors. Advantages: More motors more thrust, as well as increased redundancy. Rather than using fewer yet more powerful motors, can carry more weight. Disadvantages: higher price and larger battery pack. The takeoff and landing of a copter is easier than a fixed wing because it assures verticality. Drones size begin from Nano to mega as the size of a car found at a variety of prices from cheap (toy) to very expensive (professional). Choose of the drone depend on the user and from the way of use. 18 Chapter 4 Drones structures Unmanned Aerial Vehicles (UAV) An unmanned aircraft system comprises a number of sub-systems that include the drone (often referred to as a UAV or unmanned air vehicle), its payloads, the control station, and the remote stations figure 4.1. A UAV, have some automatic intelligence it will be able to communicate with its controller and to return automatically at home (point of the beginning). UAV’s are manually guided by a radio transmitter connected through a receiver on board of the vehicle. A flight controller as a brain of the UAV giving the command to the Electronic Speed Control (ESC) them to motors. In this chapter, we will know the function of each element constituting the UAV. The main components of a simple Unmanned Aerial Vehicle (UAV) are: 1. Frame. 2. Electronic Speed Control (ESC). 3. Motors. 4. Propellers. 5. Flight controller. 6. Batteries. 7. RC receiver. 8. RC radio transmitter. 9. Global Positioning System (GPS). UAV airframe also includes the flight control surfaces, which could be a combination of either aileron/elevator/rudder, or elevator/rudder, or ailerons, airframes are made from materials such as plastic, metal or carbon fiber and equipped with a receiver. The receiver receives information that tells it what to do from a transmitter, receivers and transmitter have communicated using radio frequencies (RF). To help in flight controls a GPS provides accurate positions; GPS plays an indispensable role in the autonomous control of UAVs because it provides an absolute position measurement. GPS receiver could achieve a three-meter accuracy; if more accuracy is needed, there are differential GPS units, which could achieve centimeter-level accuracy. The disadvantage of GPS is its vulnerability to weather factors and its relatively low updating frequency (commonly 4Hz), which may not be enough for flight control applications (Lafay, 2015). 19 Fig 4.1: Main electronic elements of a quadcopter. Drones should be powered with LiPo batteries, which are much faster than other types of batteries because they output power faster, store a large amount of power, and have a long life. The battery life for a flight in good conditions (no wind or cold weather) is for 15 minutes. Propellers are blades attached to drones spin to create lift and move up the drone, some drones come with special propellers that self-tighten like DJI phantom, which does not need a key to tight. These Propellers are attached to motors and an ESC for each controls the speed of each motor independently and ensure the stability of the drone by varying the speed in this case the drone is able to hover in place, climb or descend and move in all directions. Flying with remote control (RC) transmitter means that you will be limited to flying line of sight and you won't have the benefit of advanced communication that comes with smart devices like phones and tablets. Rc transmitters have the capabilities of controlling a drone for long distances, RC controllers do not communicate any position data or battery charge status. The radio transitions are always present everywhere in the word, most remote control drones use 900 MHZ for transmission and smartphones and tablet controllers don't have the range that RC transmitter does (Lafay, 2015). Flight controller, a flight controller is essentially a normal programmable microcontroller, but has specific sensors onboard; a very simple flight controller will include only a three-axis gyroscope to be able to auto level the drone. More sophisticated flight controller includes more specific sensors like:  Accelerometer, it measures linear acceleration in up to three axes (X, Y, and Z). The units are normally in "gravity" (g) which is 9.81 meters per second. Accelerometers detect gravity, and it can know which direction is “down”. This allowing multirotor aircraft to stay stable (www.robotshop.com). 20  Gyroscope, a gyroscope measures the rate of angular change in up to three angular axes (alpha, beta, and gamma) the units are often degrees per second.  Inertial Measurement Unit (IMU), IMUs are electronic devices that are capable of providing three-dimensional velocity and acceleration information of the vehicle they are installed on at high sampling rates. They consist of three accelerometers and three gyroscopes mounted in a set of three orthogonal axes, the major problems encountered with such developed cheap IMUs, are the need to perform the calibration process for the used sensors (Dissanayake et al. 2001).  Compass/Magnetometer, an electronic magnetic compass is able to measure the earth's magnetic field and used it to determine the drone ‘s compass direction This sensor is almost always present in the system has GPS input and is available in one to three axes.  Pressure/Barometer, since atmospheric pressure changes the farther away you are from sea level, a pressure sensor can be used to give you a pretty accurate reading for the drone’s height. Most flight controllers take input from both the pressure sensor and GPS altitude to calculate a more accurate height above sea level (www.robotshop.com).  GPS, Global Positioning Systems (GPS) use satellites signals to determine the specific geographic location of the drone; it registered the coordinates of the takeoff points to return to it in case of emergency of automated flights. A flight controller can have either onboard GPS or one, which is connected to it via a cable. The accuracy of these GPS is not very high. All the listed above electronic system is standard for all professional UAV’s The comparison of a drone with a human body is that the flight controller is the brain, the wires are the blood vessels and nerves, and the motors are your muscles, limbs, and hands (Issod, 2015). Unmanned Aircraft System (UAS) As we mentioned in the first chapter that UAV is an Unmanned Aerial Vehicle remote control or autonomously guided. Otherwise, the UAS Unmanned Aircraft System is very similar the UAV's one but with a ground station control. A UAS is a term more suitable to professional drones specialized in a specific field as military and geomatics. The structure of a UAS comparing to a UAV is constituted from the above listed electronic devices and sensors:  Long distance radio transmitter (TX)/ receiver (RX) for controlling the drone.  Long distance First Person View (FPV) for flight monitoring. 21  Telemetry and On Screen Display (OSD) receiving information about the flight and display on the screen.  Hi-Resolution cameras to move the real flight image.  Weatherproof design.  Accurate GPS (DJI Naza or RTK).  Flexible payload (body of the drone, motors, etc.…).  Autonomous flight compatible, a route tracing for long distances flights. Besides the electronic elements, the UAS must be equipped with:  Field computer or tablet pc for automatic flight control.  Workstation computer for data processing.  Ground control points marks.  High accuracy professional GNSS receivers. The main parts of a typical UAS are: Autopilot, payload, communication system and a ground control station discussed in details below. An autopilot is a system used to guide drones without assistance from human operators, consisting of both hardware and its supporting software. The autopilot is the base for all the other functions of the UAS platform. Autopilots allow defining the coordinates relative to fixed home location or to takeoff position. An autopilot can take control of different objectives such as: a) Pitch attitude hold. b) Altitude hold. c) Speed hold. d) Automatic take-off and landing. e) Roll-Angle hold. f) Turn coordination. g) Heading hold. The autopilot needs also to communicate with the ground station for control mode switch, to receive the broadcast from GPS satellite for position updates and to send commands to UAS motors. New autopilot systems come with new functions as follow me you can be all time follow by your drone or flying around a waypoint with an adjusted radius, etc… GPS plays an indispensable role in the autonomous control of UAVs because it provides an absolute position measurement. A known bounded error between GPS measurement and the real position can be guaranteed as long as there is a valid 3-D lock. There are also differential GPS units, which could achieve centimeter-level accuracy. The disadvantage of GPS is its vulnerability to weather factors and its relatively low updating frequency of 4Hz, which may not be enough for flight control applications. 22 Payload, the payload of UAV could be a camera, or other emission devices like Lidar mostly for intelligence, surveillance, and reconnaissance uses. The type and performance of the payloads are driven by the needs of the operational task it could vary from 200g to 4 kg (Reg Austin, 2010). A communications system providing the data links between the Control System and the drone through radio frequencies and it provides: Transitioning of the flight path to be stored in the drone flight control sytem Transitioning real-time flight control commands. Transmit control commands to the aircraft-mounted payloads (gimbal or camera mounted). Transmit updated positional information to the drone (Reg Austin, 2010). Most UAV have more than one wireless link supported. For example, RC link for safety pilot, WIFI link for large data sharing. Control Stations (CS). A CS is simply the control center of a UAS system, within which the mission is pre-planned and executed. The launching and recovery of the aircraft may be controlled from the main CS. It is necessary for the operators to know, on demand, where the aircraft is at any moment in time. It may also be necessary for the aircraft to ‘know’ where it is if autonomous flight is required of it at any time during the flight. This may be either as part or all of a pre-programmed mission or as an emergency ‘return to base’ capability after system degradation (Reg Austin, 2010). From a laptop, the operator could easily input waypoints onto a map base (google map), and providing easy access to the key and frequently used features figure 4.2. Fig. 4.2: Control station of a UAS system. 23 A control station can include:  Primary battery, used to power the LCD monitor and/or FPV glasses and possibly the video receiver.  Secondary battery for the transmitter.  Mounting for the LCD monitor.  Mounting for the video receiver  Space for storing the RC transmitter.  Mounting for the long-range antenna. A laptop is not something everyone needs within the field but it can give you some nice features like mapping, missions, telemetry, spectrum analyzer, and disk video recorder. You will not need anything too special for computer hardware, just as long as it is capable of running google earth at a minimum and the battery holds for as long you need it (Glover, 2014). Moreover, control stations are equipped with: On Screen Display (OSD), allows the pilot to see various sensor data sent back from the drone. One of the easier ways to include on-screen data is to use a camera with analog output and place an on Screen display board between the camera output and the video transmitter. First person view (FPV), FPV describes photography where you see what the drone is seeing. The video is beamed back to a small monitor or to a set of special goggles, the operator is wearing. FPV gear for drones will work differently than the purely digital cameras we are used to. These are often analog systems and therefore use either a different second camera. This output is coupled to a transmitter with its own antenna- and often needing its own battery. The video signal is then transmitted from the drone to your ground station and displayed on a small monitor on the side of specially designed goggles. FPV does not require a high-resolution camera (Issod, 2015). Smart Devices, Smartphone, and tablets can be used to display video in real time. The difficulty with using smart devices is that most receivers are not made to receive a video signal from a wireless video receiver. A smartphone currently works best with the video sent via WiFi (WiFi camera) with application to run this camera (www.robotshop.com). With the evolution of the drone and the smart devices, tablets or phones could replace nowadays control stations with an application installed fully functioned controlling the UAS very easily and without any added equipment's, these applications give you advance positioning, the first-person video controls FPV, programmable flight routes, etc. Smart devices during flight can display on screen drone position using GPS, flight status, speed, battery life and flight time. Sensor. Many of the most promising application areas for UAS relate to the gathering of information that can be remotely sensed. This ranges from visual range cameras gathering data for surveillance of various kinds to meteorological instruments, to geologic surveying and crop analysis among a wide variety of other existing and potential applications. 24 The sensors need gimbals to be mounted on UAS, a gimbal is often used to stabilize a camera or a sensor, connecting a sensor directly to a UAS frame means it is always pointing in the same direction as the frame itself. Gimbals are high end mounting systems that reduce shake by stabilizing the camera; some gimbals provide remote control for adjusting and rotating the camera to capture a different perspective. Gimbals tend to reduce flight time due to their weight (Lafay, 2015). UAS remote sensing functions include electromagnetic spectrum sensors, gamma ray sensors, biological sensors, and chemical sensors. Cameras. In UAS applications cameras, are made useful and highly adaptable by the addition of gimbals for pointing and stabilization software for removing distortions caused by aircraft vibration and atmospheric buffeting. These cameras are used as photogrammetry sensors to take aerial photography it could be a professional camera or a sports one depending on the resolution needed. Infrared Detectors, a thermographic camera or infrared camera is a device that forms an image using infrared radiation, similar to a common camera that forms an image using visible light. Instead of the 450-750-nanometer range of the visible light camera, infrared cameras operate in wavelengths as long as 14,000 nm (14 µm). These infrared detectors are used for the Normalized Differential Vegetation Index (NDVI) extractions. Multispectral and Hyperspectral Sensors, Recent advances in remote sensing and geographic information have led the way for the development of hyperspectral sensors. Hyperspectral remote sensing, also known as imaging spectroscopy, is a relatively new technology that is being investigated by researchers and scientists in the detection and identification of minerals, and land use (Hyperspectral Remote Sensing). Radar, many of the most promising applications of radar-based sensing to UAS utilize Synthetic Aperture Radar (SAR). SAR is a form of radar, which uses relative motion between an antenna and a target region to provide distinctive long-term coherent-signal variations, which are exploited to obtain a finer spatial resolution for the production of Digital Elevation Models (DEM) LIDAR (Light Detection and Ranging), is an optical remote sensing technology that can measure the distance to a target by illuminating the target with light, often using laser pulses. LIDAR technology has application in geomatics, archaeology, geography, geology, geomorphology, seismology, forestry, remote sensing, and atmospheric physics (Arko Lucieer et al., 2012). This technology helps in 3D terrain modeling by producing 3D point clouds. Meteorological Sensors, the use of a UAS enables the sensor to be deployed to a location in the atmosphere remote from the user of the sensed data. The National Weather Service and others have used radiosondes and operated aircraft to reach regions of the atmosphere remote from the ground observer. The use of radiosondes, however, is inefficient and costly. 25 Chapter 5 Drone Software’s An unmanned aircraft system (UAS) differ from other unmanned aerial vehicles (UAV) by a software system of control and processing. This software system includes a preflight software and a post-flight one, the first one is the flight planning and the last one is the data processing for mapping purposes. Some of these two categories of software are open source, below we will mention some of them and we will introduce some commercial software. From this chapter, the reader will take an idea about the software's and their structure. Drones flight planning software The trend in UAS photogrammetry is moving towards the use of autonomous systems since manually controlled systems are highly affected by environmental conditions. Because of the automation of the flights and the improvements in the workflow of data processing, UASs can be applied for practical geographical applications. Non-experts, unlike UAVs, can control these systems. The drone's flight planning software's deals with the drones control systems and play the role of the pilot, it can make the adjustment of the control system, design the flight and take the guide of the vehicle. In the market, we can find several flight planners software each of them is compatible with the control system designed by their company manufactured below some examples of flight planners software:  Emotions 2 of sense fly to trace the flight line and manage it by modifying the flight speed and the altitude.  DJI Ground control for Ipads of phantom DJI.  Q-Ground control is open source software compatible with windows.  ArduPilot.  Feiyu tec. DOS.  And others… This software’s have a very fast evolution, firstly a ground control station in a big suitcase with cables and antennas constituting the main drone part, nowadays a smartphone or tablet with a small application can take control of all the flight system through a WiFi or a Bluetooth connection. To plan a flight mission, you must go through some basic, these basics are the same for all software with some little modifications in commands, and each automated flight should follow these steps: 1. Running outdoor the flight path software on a pc or tablet, a window similar to google maps will be displayed. 2. Connecting the PC to the aircraft (Bluetooth or WiFi), all the electronic sensors (GPS, radio, telemetry, etc.…) of the aircraft will be connected locating the position on the screen. 26 3. Location the area of interest and drawing the flight path by defining the border of a polygon bounding the area. 4. Drawing a survey grid automatically inside the drawn polygon, taking into account the overlapping side of the aerial photography by dividing the area of interest into flight lines linked by waypoints figure 6.1. In some old software, you can no find this function so you must draw these flight lines manually by dropping a pin on the map of the software background. Fig. 5.1: Tablet flight planner at Zaarour region (Lebanon) displays of Lichi Company At any time, you can amend or cancel your planned operations. Past flights will also be saved and can be re-used or modified for new upcoming activity. 1. Defining the altitude of each waypoint and the angle of the mission by rotating the survey grid, in case of fixed wings the grid must be stated with the direction of the wind. 2. Selecting the start point which is the taking off and it is usually the place of the aircraft and defining the landing point by a waypoint or simply selecting the back home command that returns the drone to the takeoff position. 3. Adjusting the flight speed and the gimbal orientation with the number of camera frames by second if the camera is found on board the drone. 4. Camera adjustment to take simultaneous photos in case of DSLR cameras (time lapse), in case of the camera is built in adjustment is made directly from the software application with the mission planning. The mission planning software displays the GPS number of satellites tracked, battery life and flight duration. Figure 5.1 shows the waypoints dropped manually forming the flight path, the position of the aircraft expressed by latitude and longitude with the beginning and the end of the mission. We define three scenarios in the flight planning module:  Documentation of a surface with the flat or moderate terrain. 27  Exposure of a rough/ mountainous terrain like hazard areas.  3D modeling of buildings and other objects. Some new tablet personal computers applications with new functions like flips and rolls (Abbeel et al., 2007), collision avoidance (Bellingham et al., 2003 and Pettersson and Doherty, 2004), automated target tracking (Nordberg et al., 2002) and operations like the “follow me” modus. Nowadays, waypoint navigation for UAVs is a standard tool (Niranjan et al., 2007). Thus, the autonomous flight based on defined points in a global coordinate system is possible for most UAV systems. For the autonomous flights of UAVs, a start and a home point have to be defined relative to its coordinate. Moreover, some packages allow in addition to the waypoints, the definition of lines, paths, boundaries and no-go areas (Gonzalez et al., 2006; Wzorek et al., 2006). Since most of the autonomous systems are stabilized, it can be expected that the flight trajectory of the autonomous systems is more stable. Thus, for the automation of the workflow, the orientation values coming from the navigation units can be used as an approximation for the image orientation and for the fast production of overview images (Eisenbeiss, 2009). Drones mapping software After finishing the flight mission successfully, a post-flight image processing is needed to extract data, this part is the photogrammetry episode described in detail in chapter 6. For data processing, a powerful PC station and a professional software can make the job, we listed below a series of open source software’s found on the net it could help students and researchers to do their job: Airphoto SE offers essential features needed for rectification of oblique aerial imagery with geo-referencing. It is capable to make an automatic correction for radial lens distortion. It is designed for beginners or experienced users for combining aerial images with maps, orthophotos, and satellite images (http://www.uni-koeln.de/~al001/airphotose.html). Fiji, Fiji is an image-processing package based on Java3D and many plugins organized into a coherent menu structure (http://fiji.sc/Fiji). ImageJ is a public domain Java image-processing program for Macintosh. It can display, edit, analyze, process, save and print 8-bit, 16-bit and 32-bit images(http://imagej.nih.gov/ij/index.html). MapKnitter, MapKnitter is a free and open source tool for combining and positioning images in geographic space into a composite image map. Known as “orthorectification” or “georectification” to geographers (http://mapknitter.org/). Visual SFM, VisualSFM is an application for 3D reconstruction using structure from motion (SFM). The reconstruction system integrates several other projects: Bundle Adjustment, and Linear-time Incremental Structure from Motion (http://ccwu.me/vsfm/). 28 CMPMVS, CMPMVS is a multi-view reconstruction software. From images to texture mesh (http://flightriot.com/post-processing-software/cmpmvs/). CloudCompare, Cloud Compare is a 3D point cloud (and triangular mesh) processing software. It was also meant to deal with huge point clouds (typically more than 10 million points, and up to 120 million with 2 Gb of memory)(http://www.cloudcompare.org/). Meshlab, MeshLab is an open source, portable, and extensible system for the processing and editing of unstructured 3D triangular meshes. The system is aimed to help the processing of models arising in 3D scanning, providing a set of tools for editing, cleaning, healing, inspecting, rendering and converting meshes (http://meshlab.sourceforge.net/). Drone image processing is based on a classical photogrammetric model, which is enhanced in turn by a powerful computer vision algorithm. This enables the automatic extraction of numerous key points in the images and optimizes camera parameters such as external orientation and camera model. Other quality processing programs can alternatively be used, such as PhotoScan by Agisoft, Pix4D, SkyPhoto of SOUTH Company and many others. All these softwares have a first processing phase employed by identifies and extracts matching key points in the overlapping sections of the aerial images acquired by the drone. These key points are entered into an equation that determines the precise position and orientation of each image, as well as the internal and external camera parameters. The next step of the process is point cloud densification, which is required to obtain a highly accurate 3D model. This is used to generate a digital surface model (DSM) and orthophotos. Obviously, we won´t tell you which software is “the best”, but we can definitely present you some of the most interesting solutions which we have tested. As a good mapping software, we can only recommend SkyPhoto from SOUTH surveying company, if you need to generate high-resolution georeferenced orthophotos or extraordinarily detailed DEMs. Which we will speak about its structure in details. This software creates a 3D model from multiple digital photos of the area to map. Moreover, drones use airborne GPS data in order to fulfill the geo-referencing task. If you need a better accuracy, into the 3D model can be imported and matched GCP (Ground Control Points) or use a drone with Real Time Kinematic (RTK) mounted on board which can give you directly georeferenced data. The 3D mapping program I would like to present you in detailed is SKY PHOTO from SOUTH surveying company, which we are grateful to give us the permissions and material to introduce you the structures of drone processing software. With this solution, you will be able to handle large volumes of data from drone orthophotos by bringing them into a virtual environment. This software is linking between drone’s data and traditional geodetic surveying; it gives you a chance to explore your project with your UAS with a very high accuracy. 29 All UAV software for data processing has the same interface principle, the workflow is very similar with some advanced function, a basic notion advanced way is as follows figure 5.2. Fig. 5.2: Basic drone data processing steps. After getting photos from the camera the first step in a data processing software begins with the picture insertion, a structure from motion method align the picture to each other see chapter 7, the bundle adjustment of these photos allow the generation of 3D point clouds. A mesh of triangulated irregular network (TIN) interpolated from point clouds for a Digital Surface Model extraction (DSM), some software has the possibility to add Ground Control Points (GCP) for the georeferencing. All the photos joined to form a mosaic of the whole flight area draped on the DSM to produce an orthophotoplan. Some more advanced photogrammetric software’s own more functions than a basic software helpful for specialists in the field of geomatics, as an example of SkyPhoto. Processing workflow All professional drones data processing software designed for transforming low-altitude aerial images into consistent and accurate points cloud, DEM (Digitized Elevation Modeling), DOM (Digitized Orthophoto Modeling) mosaics, etc. These software's features sharply are not only one-key processing for workflow automation but also advanced settings and editable output options. The special functions of this software, begin from indoor camera calibration, dodging process, accuracy quality report, measurement tool, 3D modeling generation and browse, DLG (Digitized Line Graphics) production based on stereo image pair and so on, the sequential processing actions are summarized in figure 5.3. Fig. 5.3: SkyPhoto UAS processing data. In figure 5.3 an example of SkyPhoto software data processing path with the same principle of advanced professional functions in red fonts as Importing 30 GPS and IMU data of each photo these data obtained from the flight control software of the drone as a text file figure 5.4. Fig. 5.4: GPS/IMU file from flight control of a SOUTH drone. This file in figure 5.4 containing the name of the image, coordinates, flight altitude and IMU data Heading, Pitch and Roll. These three rotations are the transformation between the image reference system, and a flat, projected mapping plane, most often in Universal Transverse Mercator (UTM). Omega – Rotation about the X-axis. Phi – Rotation about the Y-axis. Kappa – Rotation about the Z-axis. These angles will produce the same result as all of the above-described transformations (based on Heading, Pitch, Roll), mapping the raw image directly into the UTM mapping plane figure 5.5. Fig. 5.5: Rotation of the aerial photography. The drone in on the flight path taking aerial photography the GPS get the coordinates of the middle of the photo and the IMU calculate its rotation in a Universal Transverse Mercator (UTM) system. 31 In Skyphoto the aero-triangulation is made basing to GPS/IMU data and as we said that the GPS accuracy is not very high about approximately 2 meters, we must refer to Ground control points (GCP) surveyed by accurate geodetic instruments. When these previously surveyed points, GCP's added in the software a reaero-triangulation is needed to reprocess the data after matching the GCPs. This SkyPhoto aero-triangulation editing function allows the user to control the accuracy and adjust it if possible based on a plotted report of errors. The intellectualized aerial triangulation algorithm satisfactorily deals with tough cases like images from unstable flight attitude (Kappa or Omega angle out of tolerance) and sparse textures, for example, deserts, a large water area with just a little land, etc. The overlap percentage and rotating angle of images are very little restricted. High-precision POS data from airborne GNSS-RTK system successfully gets you to minimize the huge efforts in dealing with ground control points fieldwork, and you may go straight to adjustment then mapping without the GCP concern. Similar to all other photogrammetry software’s SkyPhoto can generate a mosaic, a DSM and orthophotos, more than that SkyPhoto has the possibility to DEM edit, this function transforms the DSM to a DTM by subtraction of the trees and buildings heights. Millions of orientation points are attributed and even reach to hundreds of millions after densification. Instead of monochromatic points cloud, the colorful output is more convenient for users to view and analyze the shape and properties of surface features. In addition, you may browse the points cloud in the software like the way that you do with a 3D laser scanner. Many terrain analysis software based on DEM can extract contour lines, ridges, and thalwegs, after aero-triangulations and DEM edit SkyPhoto allow the user the quick extraction of terrain features in DLG format without any additional software. UAS use a simple digital camera, these cameras are not metric to calculate the cameras internal parameters that give us the possibility to use it as metric one, SkyPhoto has a specialized camera calibration program is included for image distortion correction, as similarly required by all other professional aerial photogrammetry software solutions. By manually inputting camera parameters (eg.principal point, principal distance, pixel size, etc.), you may easily finish the procedures indoors with a desktop LCD monitor. It is to either be done before or after photography, but normally, better before the flights (www.southinstrument.com). As an example of photogrammetry processing figure, 5.6 shows the workflow of SkyPhoto software by South surveying instruments company china. This workflow is similar to the workflow of the aerial mapping system with some different, it consists of the definition of the new project, importing 32 the aerial images and their position file for processing and image stitching insertion of GCP’s in case of non-accurate GPS image position files for the generation of dense point clouds, Digital Surface Model (DSM) and Digital Ortho Model (DOM). Fig. 5.6: SkyPhoto photogrammetry processing workflow. 33 Chapter 6 Drones photogrammetry methods Traditional photogrammetry methods begin from analogical, analytical to digital photogrammetry lead to a new drones photogrammetry methods more simple based on fully computerized systems (software), these software open a new era in aerial and terrestrial photogrammetry. Type of photogrammetry There are two main types of photogrammetry that are used today in geomatics, which are close-range and aerial photogrammetry. As the name implies, aerial photogrammetry is by attaching downward facing cameras to an aircraft and taking images from above. According to Matthews (2008), Aerial photogrammetry utilizes large-format imagery and ground control points (GCP) to georeferenced an earth portion in a virtual model. This virtual model allows us to take horizontal and vertical measurements. Aerial photogrammetry is commonly used for mapping surveys, digital terrain models (DTM), and digital orthophotos (Matthews, 2008). The other type of photogrammetry is close-range, which is very similar to aerial photogrammetry as many of the basic principles. The main difference between the two is the distance from the camera to the object. In close-range photogrammetry, the object-to-camera distance is less than 300 m (Matthews 2008). A variety of different camera configurations and platforms are included under close-range, including low-level aerial for light sports aircraft and helium-filled blimps (Matthews 2008). Also included in close-range photogrammetry terrestrial and UAVs systems. Close-range Photogrammetry has not always been a preferred method until the appearance of the new structure from motion software. Another difference from aerial photogrammetry is the ability to use simple non-special cameras due to the advancements in 3D modeling software and digital camera resolution. Close range photogrammetry (CRP) The same basic principles of traditional aerial photogrammetry can be applied to pictures taken from lower altitudes or from the ground. Terrestrial, ground-based, and close-range photogrammetry (CRP) are all descriptive terms that refer to photos taken with an object-to-camera distance of less than 300 m, the case of drones photogrammetry. Since the same basic principles of a terrestrial and suspended from a light sports aircraft (low-level aerial), both types of nontraditional photogrammetry are referred to in close-range photogrammetry (Matthews, 2008). The reduction of a three-dimensional object to a two-dimensional image implies a loss of information such as the object areas which are not visible in the image that cannot be reconstructed from it. For the reconstruction of an object from photographs or images, it is, therefore, necessary to describe all elements which contribute to this process, 34 such as light sources, properties of the surface of the object, the medium through which the light travels, sensor and camera technology, image processing. Close range photogrammetry has significant links with aspects of graphics and photographic science, for example, computer graphics and computer vision, digital image processing, computer-aided design (CAD), geographic information systems (GIS) and cartography. Traditionally, there are also strong associations of close-range photogrammetry with the techniques of surveying, particularly in the areas of adjustment methods and engineering surveying. The application of CRP is very wide we list below some specialist area of uses such as:  Architectural photogrammetry: architecture, heritage conservation, archaeology.  Engineering photogrammetry: general engineering (construction) applications.  Industrial photogrammetry: industrial (manufacturing) applications.  Forensic photogrammetry: applications to diverse legal problems.  Biostereometrics: medical applications.  Motography: recording moving target tracks.  Multi-media photogrammetry: recording through media of different refractive indices.  Shape from stereo: stereo image processing (computer vision). A very important question to ask, how does automated CRP work? 1. Data collection: Multiple overlapping photos from different locations 2. Automated feature matching: Over 2000 matches and nearly 1000 incorrect matches. 3. Derived interpolated surface geometry (mesh). Figure 6.1 shows the close-range photogrammetry sequential processing workflow. Beginning with the image preparation which could be photo enhancement contrast illumination and selection, Keypoint Detection the selection of similar pixels between photos, Keypoint Matching for the two first images than to the remaining ones, Loop Closing image preparation for the Bundle Adjustment step of least square calculation which lead to the Transformation to absolute coordinates. After that comes the Image Rectification based to the absolute coordinates for the Dense Depth Estimation to build a 3D model by Triangle Mesh Generation of the Model Fusion. 35 Fig. 6.1: CRP processing workflow. As for traditional photogrammetry works, in the CRP work some things to avoid such as:  Very dark surfaces.  Reflective surfaces.  Transparent surfaces (including water).  Uniform textures and solid color surfaces.  Moving light sources/shadows.  Capturing your own shadow. In the last years the evolution of the aerial photogrammetry was very fast and lead to new advanced computer technologies based on very powerful software, the difference between traditional aerial photogrammetry and close range photogrammetry in table 6.1 was made by Andrew Marshall in the school of surveying of the University of New South Wales in 1989, in this year as all we know that the drone wasn't used for civilian photogrammetry but the close range photogrammetry still has the same principle and procedure the only change is advanced in technology. 36 Table 6.1: The difference between aerial photogrammetry and close-range photogrammetry (Marshall, 1989). Aerial photogrammetry Close range photogrammetry Relief is small comparing to flying height. Objects may have truly spatial characteristics (large depth). Precision requirement differs for height and Precision in all three coordinates may be planimetry. equally important. The entire format is usable. A restricted format is likely. Vertical photography is used exclusively. Spatial nature of the object necessitates photography with varying position and orientation. Target only for control points if at all. Maybe possible to target all points. A large block may consist of thousands of The total number of photography is usually photographs. too small. Auxiliary data have only limited accuracy. It is possible to determine camera parameters accurately. A fairly standardized approach for all appli- Flexible approach required due to differences cations. from project to project. Table 6.1 of A.R. Marshall (1989) still valid and could be applied as a comparison between drone photogrammetry and plane photogrammetry with some modifications such digital cameras precision, high planimetry accuracy GPS one, point cloud extractions, 3D models, etc… In CRP camera precision plays a very important role, especially in the spatial precision of the output data. A few years ago metric cameras were used in CRP, "A metric camera is a general term applied to a camera which has been designed for surveying and has a well-defined inner orientation. That is a camera with a good lens of small distortion, in which the position of the principal point can be located in the image plane by reference to fiducial marks. All Cameras not possessing these characteristics can be defined as simple or non-metric Cameras" (Adams, 1980). It is around this time (1980) that non-metric cameras were established as a suitable tool for close-range photogrammetry, and that the accuracy of projects using non-metric cameras could equal those using metric cameras (Karara, and Faig, 1980). Nowadays non-metric cameras could be used in close range photogrammetry after camera calibration procedure for obtaining the internal parameter and characteristics of the camera. In his paper, Xie Feifei spoke about The Design of Four-combined Wideangle Camera. There are a number of developed products of the four-combined wideangle camera (Xie Feifei et al., 2014). The structure of the new four-combined wide-angle camera is shown in figure 6.2. 37 Fig. 6.2: Combined oblique cameras system. These oblique aerial images are used in photogrammetry to create threedimensional models because they capture all sides of an object due to the differing angles captured by the various cameras figure 6.2. This type of imagery is used most commonly for three-dimensional urban mapping (Redweik, 2012). Distortions and Camera calibration The model describing the geometric properties of the camera and lens system is known as the inner or interior orientation, sometimes also referred to as camera intrinsic (Luhmann et al., 2006). As stated in Collinearity, the 2D image coordinate system has its origin at the center of the image. The principal point is the orthogonal projection of the projection center on the sensor, and is not necessarily the same as the center of the image, hence the necessary computation of principal point offset values. The interior orientation describes the parameters required to allow the principles of collinearity to be applied to distorted images. The exterior orientation (also referred to as camera extrinsic) describes the position (for example x, y, z) and attitude (roll, pitch, and yaw, or omega, phi, and kappa) of the camera’s projection center when the image was taken. The SfM approach, used, for example, by Bundler, Agisoft Photoscan, Photomodeler and Pix4D mapper software performs an automatic calibration using, in the case of digital images, some of the exchangeable image file format (EXIF) metadata in the image file as a starting point. This defines the camera’s interior orientation and simultaneously calculates the exterior orientations using tie points identified on the input images in a process known as bundle adjustment (a reference to the bundles of light rays converging on the optical center of each camera). One time internal parameters of a camera are known, any point in the space can be fixed by the intersection of two rays of light that are projected from two different photos. This is the principle photogrammetry to align pictures and produce models. However, there are two major factors that can affect the accuracy of the resulting model. 1 – The first is a system error due to lens distortion, this error causes the point to shift from its true position to a skewed position (Dai and Lu, 2010). 38 This is due to a combination of decentering distortion and radial distortion, decentering distortion is caused by the camera lenses not being perfectly centered in relation to each other, and radial distortion is a distortion in each lens. This type of error is easily corrected by the photogrammetry software (Dai and Lu, 2010). 2 – The second type of errors is errors due to human factors. Most human errors are attributed to the imprecise marking of points in two different photos. This can be overcome by marking the points in three or more photos This is why most programs require the points to be marked in at least three photos (Dai and Lu, 2010). To reduce these distortions, we need a camera calibration, camera calibration is the determination of principal distance (c), principal point coordinates (Xp and Yp) and lens distortion parameters (K1, K2, K3, P1, and P2). Without calibration, high accuracy cannot be achieved, for sub-millimeter surveying accuracy, the camera calibration should be carried out in laboratories. The photogrammetric approach in early vision calibration methods, where an object with known characteristics is used to infer the internal and external parameters from images. The advent of structure from motion and projective reconstruction without knowledge of camera parameters has made this unnecessary; scene reconstruction is attempted where no calibration objects are present. In briefing we discussed four categories of camera calibration approaches: 1. Classical calibration: using the geometry of known objects in the world (Leibowitz, and College, 2001). 2. Auto-calibration assuming fixed internal parameters: assuming that the same or an identical camera is used for each view of a scene. 3. Auto-calibration with varying internal parameters: relaxing (partially) the assumption of identical cameras. 4. Special or constrained camera motion: solving the special cases where the motion between cameras are known to be of a particular form (Leibowitz, and College, 2001). Classical calibration using known objects originated with photogrammetry, aimed at extremely accurate camera calibration and making use of specially designed calibration devices. The Direct Linear Transformation (DLT) of Abdel-Aziz attempts to apply the calibration from objects to common types of cameras in a computationally direct manner (Abdel-Aziz, Karara, 1971). The calibration grids of Tsai represent a move towards automated calibration from images. Tsai uses a planar calibration pattern (which has come to be known as the Tsai grid) consisting of a grid of black and white squares. An image of grid points with known world coordinates allows calibration from a single view by computing the projection matrix from the images of points and their known world coordinates. The calibration method of Tsai is the most applied calibration method in most of photogrammetry software (Tsai, 1986). 39 Auto-calibration was first explored by Faugeras et al, using the Kruppa equations. The Kruppa equations embody the constraint that the epipolar planes tangent to the absolute conic project to corresponding epipolar lines tangent to the Image of the Absolute Conic (IAC) in each view (Faugeras et al., 1992). Auto-calibration with varying internal parameters, the constant internal parameter of many auto-calibration techniques is impractical. The assumption that a single camera with identical parameter used for the entire sequence is often invalid, although some of the parameters may remain constant, the focal length and principal point change. This is the approach taken by Heyden and Astrom allowing for the varying, unknown focal length and the principal point of square pixel cameras. An iterative minimization with the square pixel parametrization of the camera is used to compute metric rectified cameras (Heyden, Astrom,1997). Special or constrained camera motion auto-calibration, Certain camera motions, reduce the number of ambiguities in a calibration problem or allow a simple solution. The cases mentioned here include pure translation or rotation of cameras. When a camera is rotating about its optic center and not translating, all points in a pair of images are related by a homography. Furthermore, this homography is the infinite homography between the cameras (Leibowitz, and College, 2001). With the advance of the software technologies camera calibration is made automatically inside the processing workflow. To meet accuracy requirements of three-dimensional (3-D) photogrammetric mapping, the conventional photogrammetric aircraft flying mission is usually exactly designed for some flying parameters, such as overlap, side-lap, flying height, the direction of flight, velocity of flight, etc. For the UAV-based photogrammetric mapping system using a high-resolution non-metric CCD camera, camera calibration has to be performed to provide the interior orientation parameters. A listing of 91 articles on aspects of optical camera calibration for the period 1889 until 1951 is provided by (Roelofs, 1951). A large number of camera calibration approaches have been proposed since 1951. Camera calibration is the main feature in photogrammetric 3D object restitution. Calibration parameters such as principal distance, principal point, and lens distortion are usually determined by a self-calibrating bundle adjustment based on the collinearity equations and additional correction functions. The self-calibrating bundle adjustment is a very flexible and powerful tool for camera calibration and systematic error compensation, and it provides for accurate sensor orientation and object reconstruction while treating all the system unknowns as stochastic variables (Hongxia et al., 2007). 40 Bundle adjustment Bundle Adjustment refines a visual reconstruction to produce jointly optimal 3D structure and viewing parameters. ‘bundle’ refers to the bundle of light rays leaving each 3D feature and converging on each camera center figure 6.3. Bundle adjustment is a mathematical model, which allows the determination of camera position, camera orientation, object point coordinates and camera calibration parameters (SMITH and PARK 2000) (see Figure 6.3 below). Fig 6.3: Typical camera configuration of bundle adjustment in close-range photogrammetry The bundle adjustment method of figure 6.2 is summarized graphically, the first image i, joined by corresponding feature points to image i+1 and so on for other images, the process continues for building the 3D model and transforming all corresponding points into 3D point clouds. The bundle adjustment image matching process relies on a least squares solution with geometric constraints, where matching image points are determined together with the object space coordinates of the corresponding object points figure 6.2. Bundle adjustment is really just a large sparse geometric parameter estimation problem, the parameters being the combined 3D feature coordinates, camera, and orientation parameters. The basic model for photogrammetric bundle adjustment is based on the well-known Collinearity equations (Mikhail et al., 2001; McGlone, 1989): With: O = center of the projection, f = calibrated focal length, 41 a = coefficient of the rotation matrix. Many scientists worked on bundle adjustment methods, all of these methods are based on the same Principe and structure with some modification, these methods are Gradient Descent Method, Newton-Rhapson Method, GaussNewton Method, and Levenberg – Marquardt Method. These four are the most popular techniques for nonlinear least squares optimization, the last two are used to optimize nonlinear least squares in particular. Gradient Descent Method:  Based on the first-order optimization algorithm.  To find a local minimum of a function using gradient descent, one takes steps proportional to the negative of the gradient of the function at the current point.  It is robust when x is far from optimum but has poor final convergence. Newton – Rhapson Method:  Based on the second order optimization method  Newton's method can often converge remarkably quickly, especially if the iteration begins "sufficiently near" the desired root.  For a quadratic function, it converges in one iteration.  For other general function, its asymptotic convergence is quadratic.  The disadvantage of this method is the high computation complexity of H . Gauss-Newton Method:  The Gauss-Newton algorithm is a method used to solve non-linear least squares problems.  For well-parametrized bundle problems under an outlier-free least squares cost model evaluated near the cost minimum, the Gauss-Newton approximation is usually very accurate. Levenberg – Marquardt Algorithm:  This method interpolates between the Gauss-Newton algorithm and the method of gradient descent.  When far from the minimum it acts as the steepest descent and it performs gauss newton iteration when near to the solution.  It takes into account the best of both gradient descent and gauss newton method. General Facts about optimization methods:  Second order optimization methods like Gauss-Newton and LM requires a few but heavy iterations.  First order optimization methods like Gradient descent requires a lot of light iterations. Bundle adjustment is really just a large sparse geometric parameter estimation problem, the parameters being the combined 3D feature coordinates, 42 camera poses, and calibrations. bundle adjustment and similar adjustment computations are formulated as nonlinear least squares problems (Brown, 1976; Cooper, Cross,1991; Atkinson, 1996). Bundle block adjustment The name bundle block adjustment is based on the fact that the rays from the projection center to the photo points are building a bundle of rays, the bundle block adjustment is using the photo coordinates, based on these photo coordinates the bundle block adjustment is leading to more accurate results than the other methods. Of course, some additional corrections like self-calibration with additional parameters are improving the results as well as additional observations like GPS- positions of the projection centers. The Bundle block adjustment provides three primary functions: 1. It determines the position and orientation of each image as exterior orientation parameters. In order to estimate the exterior orientation parameters, a minimum of three/four GCPs is required for the entire block, regardless of how many images are contained within the project. 2. It determines tie points ground coordinates on the overlap areas of multiple images. The highly precise ground point determination of tie points is useful for generating control points from imagery. 3. It minimizes and distributes the errors associated with the imagery, image measurements, and GCPs. The bundle block adjustment processes use statistical techniques to automatically identify, distribute, and remove error (Rüther et al., 2012). Structure from Motion (SFM) Structure from Motion (SfM), was introduced that allows the extraction of the 3D structure of an object by analyzing motion signals over time (Dellaert et al., 2000). The SfM technique can be applied to large collections of overlapping photographs to obtain sparse point clouds for a wide range of objects, such as buildings and sculptures (Snavely et al 2007; Snavely et al 2006). The power of this technique was demonstrated by (Snavely et al 2007) who developed the Bundler software and used it to construct 3D models. The term Structure-from-Motion has evolved from the machine vision community, specifically for tracking points across sequences of images occupied from different positions (Spetsakis and Aloimonos, 1991; Boufama et al., 1993; Szeliski, Kang, 1994) (figure 6.4). Structure-from-Motion (SfM) operates under the same basic tenets as stereoscopic photogrammetry, the 3D structure can be resolved from a series of overlapping images. However, it differs from traditional photogrammetry by camera positions and orientation that are solved automatically without the need to specify the network of targets with known X, Y, and Z. 43 Fig. 6.4: Structure-from-Motion (SfM) functional schema requires multiple, overlapping photographs of the object. Many SFM software's are found as a commercial and open source especially for terrestrial and aerial Unmanned aerial vehicles (UAV) uses, open sources SfM packages have very limited post-processing functions unlike commercial software for a variety of tools and models, due to their high prices and fast evolution. SfM involves the 3D location is determined through automatic identification of matching features in multiple images, these features called tie points, used to establish both interior and exterior orientation parameters of cameras. In photogrammetry, 3D geometry is obtained by creating images of the same object from different positions. This makes a single point on the object visible as a pixel in multiple images. For each image, a straight line can be drawn from the camera center through the pixel in the image. These lines will intersect at one point (tie point), which is the 3D location of the object point. These tie points are tracked from image to image pixel by pixel estimating camera positions and object coordinates which are then refined iteratively using non-linear least-squares (Snavely et al., 2008). SfM method identifies features in each image that is invariant in scale, rotation and in illumination changes conditions and 3D camera viewpoint (Lowe, 2004). (Micheletti et al., 2014) demonstrated that a big the number of images produces a very high spatial resolution data at a good accuracy because of the big overlapping areas covered by pictures. Transparent, reflective or homogeneous surfaces present difficulties because incorrect features can be linked during the automatic feature-matching process (Autodesk, 2014). Tie points number depends on image texture and resolution; they are automatically identified over all scales and locations in each image. 44 Sufficient tie points allow for the reconstruction of the relative position of all images. Additionally, known points or ground control points(GCPs) with 3D world coordinates should be added to obtain scale and absolute coordinates. The SfM part of the process generates a sparse point cloud comprising tie points identified and matched across the input images. In order to construct the sparse point cloud, several steps are involved such as feature extraction, feature matching, and bundle adjustment. The SfM algorithm needs to estimate the interior and exterior orientations for each image by combining all the relative orientations of the image pairs in the form of their fundamental matrices (Verhoeven et al 2013). Once complete, a technique called image triangulation is used to calculate the relative position and orientation for each image in every pair. The overlapping pairs are then combined to form a single block, achieved by a bundle adjustment, because it necessitates adjusting the bundles of rays between each camera’s projection center and the set of projected 3D points until there is minimal discrepancy between the positions of the observed and re-projected points (the image distance between the initial estimated position of a point and its ‘true’ or measured value) (Verhoeven et al., 2013). A bundle adjustment step comes after the key point's detection to find 3D point positions and camera parameters that minimize the re-projection error, bundle adjustment, and similar adjustment computations are formulated as nonlinear least squares problems (Cooper, Cross, 1988; Granshaw, 1980; Atkinson, 1996; Karara, 1989; Wolf, Ghilani, 1997). Least square problems are the differences between the observed feature location and the projection of the corresponding 3D point on the image plane of the camera, the problem can be cast as a re-weighted non-linear least squares problem (Hartle, Zisserman, 2003). Bundle adjustment is really just a large sparse geometric parameter estimation problem, the parameters being the combined 3D feature coordinates, camera poses, and calibrations. After bundle adjustment in data sets, processing comes the point clouds generations in a relative ‘image-space’ coordinate system, which must be georeferenced to a real world using a 3D transformation by adding a small number of known control points (Doumit, Kiselev, 2016). The technique is based on identifying matching features in images that are taken from different viewpoints. Image features are identified by the scale invariant feature transform(SIFT) algorithm (Lowe, 2004), which is robust in terms of its feature descriptors for image features at different viewing angles. Based on these SIFT matches, the camera positions, orientations, radial lens distortion, and finally the 3D coordinates of each SIFT feature are calculated using a bundle block adjustment. The 3D positions of the SIFT feature essentially form a 3D point cloud that captures the structure of an object. The point cloud is known as a sparse point cloud that can be densified with a more recent technique called multi-view stereopsis (Furukawa, Ponce, 2009). The stereopsis algorithm takes the output from the Bundler algorithm, 45 camera positions, orientations, and radial undistorted images, and applies a match, expand, and filter procedure. It starts with the sparse set of matched key points, and repeatedly expands these points to neighboring pixel correspondences, and finally applies visibility constraints to filter out false matches (Furukawa & Ponce, 2009). The algorithm is implemented in the Patch View Multi-Stereo (PMVS2) software tool. SIFT, Bundler, and PMVS2 work in sequence to generate an extremely dense 3D point cloud just from overlapping photographs. The PMVS2 point cloud contains the following information for each point: XYZ coordinates, point normal (i.e. the direction of the slope through the point), and point RGB color values (i.e. derived from the photographs. The dense point cloud can then be used as the basis of a triangulated irregular network (TIN) or mesh, onto which textures generated from the input images can be projected. SFM requires the uploads of multiple images of an object and the results are produced fully automatically without user interaction. The principal disadvantages of SFM are that the mathematical process is not transparent and many research has been done and published without being confident about metric accuracy and reliability of their results. Aerotriangulation An important and critical phase in photogrammetric mapping is rectifying the aerial images to their appropriate place on the surface of the earth. This is accomplished by collecting horizontal and vertical data, to ascertain the spatial location of a number of features that are visible and measurable on the aerial images. Geometric stability requires a minimum of four points of known positions spaced in the corners of a full stereo model be used to fully rectify On a project involving a few stereo models this may be a conventional ground surveying enterprise. The expense and time required to collect the ground survey data in this manner may render the mapping project impractical (Falkner, Morgan, 2002). Computer processing has played a major role in driving mapping scientists to develop rigorous and efficient mathematical protocols that allow for the densification of stereo model control from a minimal number of strategically positioned ground survey points. This procedure is generally referred to as aerotriangulation. Analytical software available today, with its built-in quality checks, has made aerotriangulation the preferred method of image adjustment to the earth for photogrammetric mapping. In this book will not discuss the theory of these processes, but rather it will give the reader a necessary guidance and explanation of procedures to plan and estimate the efforts required to perform satisfactory aerotriangulation for a photogrammetric mapping project (Falkner, Morgan, 2002). Aerotriangulation Computer Processing After the collection of the field control point, they are imported into the computer and processed through an aerotriangulation software module procedure including: 46  first, each stereo model is processed through a relative orientation routine involving a least squares adjustment of collinearity equations. This solution produces individual model coordinates unrelated to any reference system.  Second, a strip formation procedure joins the independent stereo models through a three- dimensional transformation (X, Y, Z). A series of equations link successive models by common pass points. The coordinates, at this stage, remain in an arbitrary reference scheme.  Then each strip undergoes a polynomial adjustment which produces preliminary ground coordinates for all of the photo control points, notes that the control points are in a coordinate system of the project.  A simultaneous bundle adjustment provides a fully analytical aerotriangulation solution and the entire block of data passes through an iterative weighted least squares adjustment until a convergent “best fit” solution is obtained. An RMSE error is noted so that the observer can judge how far the coordinates of each point were mathematically “stretched” out of position in order to resolve a solution (Falkner, Morgan, 2002). 47 Chapter 7 Flight Planning Principles and image processing Before beginning a flight project, the operator should ask and answer a series of sequential questions of drone design projects, approach, and preparations. What is the size of the area to be mapped? Depending on the equipment you have, larger the area to be mapped more the choice goes to a fixed wing. What is the size of the Ground Sample Distance (GSD)? Small GSD values generally require lower flying heights and slower ground speeds. For high-resolution projects (GSD 1 to 2cm) multi-rotors tend to be the platform of choice. What is the Nature of Terrain? Multi-rotors are often referred to as Vertical Take-off and Landing (VTOL), they are employed in congested areas such as forests and urban environments. What are the Flying Height and Ground Speed? High ground speeds at low altitude require short exposure times to avoid image blur. When GSD and focal length dictate a particularly low flying height, the exposure distance intervals tend to be short, thus requiring the camera to expose at very short time intervals. Ground sampling distance (GSD) The spatial resolution of digital maps is commonly expressed as the ground sampling distance (GSD). This is the dimension of a square on the ground covered by one pixel (p) in the image and is a function of the resolution of the camera sensor, the focal length (f) of the camera and the flying height (H = the distance between camera and ground) (Barnes et al., 2014). The formula is: GSD H p f The pixel size (p) is found in the camera technical specifications, the dimensions of the image are specified in linear units (e.g. 17.3 x 13.0 mm) as well as in a number of pixels (4000 x 3000 pixels). Pixel size is simply determined by dividing the linear units by the number of pixels. In photogrammetry to generate 3D models image capturing should be in stereo modes which express an overlap between captured images. Once a flying height has been determined it is necessary to compute the distance between each exposure position, the spacing between flight lines, and the overlap (figure 7.1). 48 Fig. 7.1: Image footprints and overlaps (Barnes et al., 2014). The exposure distance intervals (s) and the spacing between the flight lines (d) are dependent from the forward and lateral overlaps respectively. For drone mapping projects we have found that a forward overlap of 80% and a lateral overlap of 70% yield good results. If the forward overlap is a% then: s 100 (7.1) Similarly, if the lateral overlap is b%, then: (7.2) 100 Where β = GSD (width of the sensor in a number of pixels) and α = GSD (length of the sensor in a number of pixels). Some drone cameras cannot trigger at given distance intervals. In such cases, the camera is programmed to expose at a fixed time. The time interval t between two successive exposures is then calculated as (7.3), where d is the distance between exposures and v is the estimated ground speed of the drone (Barnes et al., 2014). Photocontrol points Prior to commencing mapping from aerial photos, ground survey information is required on specific terrain features in order to relate the photogrammetric spatial model to its true geographical location. These terrain features may be portrayed in two ways: by identifiable photo image features or ground targets. Acquisition of ground control data on photo image points is a necessary requirement for photogrammetric mapping for two primary reasons:  To georeference the imagery.  To check the accuracy of the spatial data collected. Technology is constantly changing and adding to the tools that can be used to collect ground control. Recent advances in GPS are the underlying reason for many of the advances in ground survey methods, as well as photogrammetry in general (Falkner, Morgan, 2002). A ground target is some kind of a panel point that is placed on the unobstructed ground prior to photography. Figure 7.2 illustrates the effective presen- 49 tation of a ground target (+) on an aerial photo. Ground targets create a discrete image point and can perhaps lead to better map accuracy. Fig. 7.2: Ground target on aerial photography. A target must be of sufficient size to be recognized on the image. Dimensions of a target in the shape of a cross are easily computed (Falkner & Morgan, 2002). Equation 7.4 defines the width of the legs of a typical ground target, as shown in figure 7.3. 0.002 (7.4), where: W = width of the target legs (cm), Sp = scale denominator (cm). Equation 7.5 defines the length of the target legs of a typical ground target, as shown in figure 7.3, for any photo scale. 10 (7.5) Where: l = length of each cross arm (cm), w = width of target legs (cm). 50 Fig 7.3: Typical ground target. To produce the mapping from stereo models, the aerial photo image must be scaled and leveled it means georeferenced to a true geographic ground location. In current mapping procedures, most points contain both horizontal and vertical information and are used for both scaling and leveling. If the final mapping products are to be referenced to a defined coordinate reference frame, then it is necessary to either geo-reference the aerial images or set ground control points (GCPs). Ground control points are point features in the object space that can be positively identified in the images figure 7.2. The targets are surveyed to determine their precise coordinates in a defined spatial reference frame. In most cases, ground control points are surveyed by means of differential GNSS. The Global Positioning System (GPS) The global positioning system (GPS) an electronic receiver measures the distances between the ground point and a minimum of four satellites and the intersection of the divergent rays establishes the spatial coordinates of the observing station. The anticipated ephemerides (positional) information is broadcast by the satellite. Several continuous tracking stations are scattered throughout the world, meticulously charting the paths of the satellites. Both the tracking data and broadcast information are available to the user, with the former providing more accurate data for processing receiver information (Falkner, Morgan, 2002). The determination of the coordinates of a ground station by GPS procedures relies upon intersection geometry. The GPS receiver, a pseudo-range measuring device, accepts carrier signals from multiple satellites. By measuring the carrier waves from a ground station to several satellite positions simultaneously, the XYZ coordinate can be determined. For ascertaining threedimensional coordinates, the receiver must maintain a continuous lock on a minimum of four orbiting platforms simultaneously (Falkner, Morgan, 2002). 51 UAV’s usually use a standalone GPS receiver enabling the drone to navigate safely and it is able to achieve a satisfactory accuracy for safe navigation, the overall accuracy which can be achieved could not reach the survey grade accuracy. Static mode In static traverses, at least two receivers must be used; both must be locked on to the same group of satellites. One receiver resides over a location with known coordinates, which could be a previous point in the circuit. The other receiver is set over a point with unknown coordinates. Observation time spent at each baseline point pair may span a significant period of time. After the observation is complete at the baseline pair, both receivers can be moved to new positions so long as one occupies a point with known coordinates. In this fashion, observations are made at all of the stations on the traverse, in turn (Falkner, Morgan, 2002). Airborne Global Positioning Today many photogrammetric mapping projects employ Airborne Global Positioning technology to minimize ground control collection for the mapping project. By locking on to several navigation satellites this device maintains a constant spatial positioning record of the sensing systems. Relating the film coordinates to a map allows the pilot to pinpoint the location of the aircraft on a specific exposure frame. This coupling of remote sensing and spatial positioning offers unique mapping capabilities. Airborne Global Positioning technology has been developed to a point where standard procedures and repeatable results make it almost standard practice for all medium to large project areas (Falkner, Morgan, 2002). Many private and public organizations currently mount GPS receivers in the airplane during a photo flight mission. This system allows the spatial fixing of the camera at the instant of exposure. Reference to Ground Station allows the Airborne Global Positioning receiver to be interfaced with a camera system. In this procedure, the GPS in the aircraft is correlated with a static GPS ground station so as to relate the onboard receiver with the known ground station. Real-Time Kinematic mode (RTK) kinematic traverses demand the inclusion of at least two receivers. One receiver must remain fixed on a known ground station over the same point, during the entire survey period. The other receiver onboard of the UAV giving corrected coordinates of the exposure frame center. By enabling a drone to become the ‘rover’ of a GNSS/GPS base station or a Continuously Operating Reference Station (CORS) survey grade accuracy can be achieved in real time at a sub-centimeter accuracy. 52 RTK requires a constant communication link between the GNSS/GPS base and the drone. In case this communication link is lost RTK will not work anymore, to solve this problem the correction made by post-processing using correction data coming either from a local base or CORS. In RTK mode the receiver of the base station is dependent from the link with the drone receiver trough UHF or mobile SIM card. Post Processing Kinematic mode (PPK) PPK surveys require data from at least two receivers: a reference base receiver and a moving rover receiver on board of the drone. If you have access to raw data from a reference station (base or CORS). PPK same as RTK requires a base station, in many countries cases the public CORS network is dense enough to provide corrections for the RTK and the PPK. It is preferable to have your own base station. Using CORS comes with the possibility of connection loss, in both RTK and PPK, when the rover loses lock. The advantage of PPK is that the search can proceed from previous and future data relative to that instant. There are two types of survey methods in PPK surveys: topo points and continuous surveys. Topo points are short (about 15 seconds) occupations, e.g. over a survey station. Continuous survey mode allows ongoing data collection at a specified logging interval; continuous mode is used for drone mapping. After drone landing, raw data collected from the mobile receiver (from the drone) and the fixed receiver (from a base station), should be post-processed in a computer software to produce an EXIF file containing images position data. The accuracy requirement (horizontal and vertical) for the final mapping datasets establishes the ground survey accuracy requirements and thus the GPS methods. The advantage of including techniques such as RTK (real-time kinematic) or PPK (post-processing kinematic) will: - Eliminate the need for GCP’s, therefore reducing costs tremendously Reduce possible human errors. - Achieve survey grade accuracy for the mapping/survey project. - Makes it easy to proof accuracy by using a couple of checkpoints (Falkner, Morgan, 2002). Flight planning In all drone projects before beginning the flight user should plan it first following these planning steps: 1. Define the mission area, resolution and overlap. Most professional drones employ waypoint navigation to define their flight path. Modern flight planning software handles this automatically, incorpo- 53 rating the three pillars of aerial mapping: mission area, overlap and ground resolution. 2. Generate a 3D flight plan. When flying, the optimal flight planning approach is for the altitude of each flight line. An ideal flight planning system will have a high-quality elevation dataset built-in, as well as allowing the operator's own elevation data to be imported. 3. Define take-off and landing points. When defining the drone’s take-off trajectory, the software allows you to define the exact direction the drone will follow after being launched. When landing, choose the point and the position of landing, with its ground proximity sensor, to adapt its altitude to the terrain, employing reverse thrust at an altitude of five meters to reduce its speed and ensure a short, precise and soft landing. 4. Monitor and adapt during flight. Even with intelligent, automated mapping drone systems, it is crucial to carefully monitor every flight via line of sight and using the drone's ground station software. The drone's position and important parameters such as its battery's remaining charge, and therefore flight time, should always be visible on the screen display (OSD) figure 7.4. Fig. 7.4: Ground station on Screen Display of the telemetry, landing and post-flight tasks (Barnes et al., 2014). 5. Launching the drone. Drones are typically launched in one of the three ways: by hand, via rolling take off on a runway, or using a launching system, most often a catapult. 6. During flight. After take-off, a drone will quickly gain altitude and travel, either automatically or on command, to take its mapping area and to start the aerial imaging mission. 7. In-flight monitoring. 54 Typically, an operator will ensure the safety of an operation by keeping the drone in visual line of sight. The drone supports this human monitoring by constantly sending status information displayed on the screen. 8. Landing. After a mapping mission is finished, when landing, choose the point and the position of landing, with its ground proximity sensor, to adapt its altitude to the terrain, employing reverse thrust at an altitude of five meters to reduce its speed and ensure a short, precise and soft landing. It is good practice to inspect the UAV immediately after the landing for any signs of overheating of the motors or electronics. The aerial images should also be inspected for quality after the first flight and any required adjustments made to the camera settings before further flights. Flight planning software Most flight planning software or computer tablet applications are automated and based on satellite imagery such as Google maps to help you locating your mission and planning it. For flying heights below 100m, it is important to inspect the terrain for objects, like power lines or cell towers, that are either not visible or do not appear on the satellite imagery. The first step for generation of a flight plan is drawing a polygon marking the outline of the project area, putting the flight height, adjusting longitudinal (forward) and lateral (side) overlap (%). The flight planning software gives you the ability of flight plans modification. One time drawing the flight plans in Google maps it automatically references the flights to the WGS84 datum which is the same spatial reference frame used by drones for navigation. Flight planning software communicates with the autopilot on board of the drone to receive broadcast positioning information from GPS satellite for position updates and to send out control inputs to the motors of the drone. The most common state observer is the inertial measurement unit (IMU) including gyros, accelerometers, and magnetic sensors. Image processing The drone non-metric cameras are lighter and less costly they are generally preferred over conventional metric cameras. However, non-metric cameras are not geometrically stable, which is a basic requirement for classic photogrammetric mapping. To overcome this problem a combination of classic photogrammetry and computer vision techniques, known as Motion Vision Stereopsis (MVS), is used by the most drones image processing software. 55 Fig. 7.5: MVS workflow. Normally drone processing software will attempt to find the approximate camera geometry (focal length and sensor size). Note that in an MVS approach the camera geometry for each and every image is computed separately, in this way the camera exposure coordinates are computed in some arbitrary local spatial reference system. Hence, if the final product does not have to be referenced to some spatial reference frame, the images are all that is required to build a model. However, the process of determining the relative camera exposure positions is significantly accelerated if the relative camera exposure positions are seeded with the coordinates established in the after-flight image geo-referencing step mentioned below. Ground control is required to rectify (georeference) the imagery to its true geographical position on the earth’s surface. The differential rectification method is a phased procedure which uses several XYZ ground control points to georeference an aerial photograph to its true place on the earth. The exact location and number of ground points required are based upon the scale and accuracy of the final orthophoto. Selecting the ground control points is generally not a task for the project manager and should be decided by the photogrammetric technician designing the project based upon his/her experience and equipment (Falkner, Morgan, 2002). The georeferencing is necessary since the photogrammetry software creates its own coordinate system when the models are generated. The selected GCPs were marked in the processing software by placing markers in the photos containing each point. The purpose of marking the points in multiple photos is to allow the software to triangulate the locations of the point using the camera locations that have already been processed and gives a new georeferenced sparse point cloud (Newsome, 2016). Then the software uses the optimized sparse point cloud and improved camera positions to generate a dense point cloud. Several accuracies and depth filtering settings are provided in this process. The generation of the dense point cloud usually takes the longest time of all the processing steps. The dense point cloud is used to generate a surface model in the form of a triangulated irregular network (TIN), a texture is built and applied to the surface features of the model. Various spatial products (Digital Surface Models and Orthophotos) can be generated via the export options offered by the image processing software. 56 Digital surface models can be prepared either in a vector format as triangulated irregular networks (TIN), used extensively in CAD software for engineering applications, or in a raster format for GIS uses. A suitable DSM must be obtained to provide a vertical datum for an orthophoto, DSM for orthophoto rectification does not have to be as dense or as detailed as a terrain model for contour generation (Falkner & Morgan, 2002). Today, many uses for geospatial mapping products require current planimetric feature data. Analysis and design from geospatial data sets generally require a known positional accuracy of features. The collection and updating of planimetric features in a data set can be costly. Many end users are also not accustomed to viewing and analyzing vector-based mapping data sets. They prefer to view planimetric features like a photo image. The final phase of the orthophoto process is the merger of the digital image and the DSM along with corrections in pixel intensity throughout the image. Software, used to merge the digital raster image with the DSM, makes adjustments in the horizontal location of pixels based upon their proximity to DSM points. This process removes the errors due to displacement and produces an image that is orthogonally accurate (Falkner, Morgan, 2002). Suitable imagery and ground control are the basic elemental data that determines the final orthophoto reliability which involves both the accuracy of distances and areas within the orthophoto as well as the relative accuracy of features with respect to their true location on the earth. Distance and area accuracy is based on the pixel size. Relative feature precision is based on the accuracy of the DSM used in the rectification process. The relative accuracy cannot be more precise than the reliability of the DSM (Falkner, Morgan, 2002). Drone mapping accuracy The optimal condition for high accuracy mapping  Work in soft lighting environment.  Avoid transparent or reflective surface.  Flight time around 11 am to 4 pm.  Avoid Fisheye wide angle cameras because of the high distortions and errors.  Optimal coverage 80% overlap and 60% sidelap.  Terrestrial, nadir and oblique images capture, must be processed in different projects and then merge or unify in one project.  Take GCP on transition images.  Use manual stitch when needed.  The use of drones with the highest precision GPS for not doing GCP.  Use an image overlap of >85%.  Set GCPs using a zigzag pattern.  Add checkpoints to verify data quality. 57  Batteries are affected by the cold good idea to pack spare, fullycharged drone batteries.  Remain aware from bird attacks.  Avoid flying in high winds and thermal winds (uplift), rain and snow (Reg Austin, 2010). Ground Sampling Distance (GSD) accuracy The quality of a project is linked to its accuracy, referring to the positional accuracy not to be confused with camera resolution or a single image's Ground Sampling Distance (GSD). The ground sampling distance achieved depends on an imaging sensor's pixel size and lens focal length, and is directly proportional to the drone's flight altitude. However, data output quality depends on much more than a camera's resolution alone. (In fact, the dataset produced by a 12 MP camera will often be hard to differentiate from that produced by a 16 MP model.) DSM Accuracy Assessment Accuracy assessment of the Digital Surface Model (DSM) is an important task for horizontal and vertical assessments by comparing DSM with GCPs measured by precise surveying instruments (GNSS receivers or Total Stations) in terms of Root Square Error (RMSE). More specifically, assessments in Easting (RMSEx), Northing (RMSEy) and vertical (RMSEz), horizontal (RMSExy), and all components (RMSExyz) as suggested Aguera-Vera using equations as follow: RMSEx SQRT ∑ X (7.6) X RMSEy RMSEz RMSExy RMSExy Z SQRT SQRT SQRT SQRT ∑ ∑ ∑ ∑ Y Z Y X X Z (7.7) (7.8) X X Y Y Y Y (7.9) Z (7.10) Where XGCPi and XDSM are the X-coordinate component of GCP and corresponding coordinate in DSM, respectively; YGCPi and YDSM are the Ycoordinate component of GCP and corresponding coordinate in DSM, respectively; ZGCPi and ZDSM are the Z-coordinate component of GCP and corresponding coordinate in DSM, respectively (Agüera-Vega et al., 2016). 58 Chapter 8 Drones and Geospatial Analysis There are a set procedures used for geospatial scientific research methods divided in case studies and field work. Drone Field work is the collection and gathering of information applied on the case studies of Multi-scale Digital Surface Models, Multi-scale landforms classification, Multi-scale Terrain Roughness, Terrain analysis for parcels evaluation, and Digital Surface Models to planar areas. The geospatial analysis based on drone terrain datasets experiments carried out at an altitude of 1700 meter above the sea level, in Zaarour region situated on the western Lebanese mountainous chain. We choose this bare non urbanized mountainous area because of its representative morphological terrain forms. The study area with a slight natural slope, represented by bare lands with elements of anthropogenic relief. The inclusion of anthropogenic micro-relief in the studying area due not only to the requirements of representativeness, but the presence of complicating microform for the experimental modeling of the terrain concave and convex smoothed areas. An autopilot Dji Phantom 3 Unmanned aerial vehicle (UAV), caring a camera of 14 megapixels at a focal length of 3.61 mm flown the study area at different heights. The flight heights are measured from the takeoff point of the quad copter; the experiment constituted from 6 missions: FA-20, FA-40, FA-60, FA-120, FA-240 and FA-360 of 20, 40, 60,120,240 and 360 meters height. The flight path followed by the quad copter was identical for all the flights and it was designed in a mobile application called Litchi. The application shows the flight path and the flight parameters (coordinates, height, time, etc…). Before starting the aerial surveying, well-distinguishable 10 control points were evenly distributed within the area of interest for scaling and georeferencing the resulted data. Ground control points (GCP) were collected with Global Positioning System (GPS) in stereographic coordinate system. The drone took Aerial photography with 80% overlapping and 70% side lapping. SfM-based 3D methods operate on the overlapping images. The drone flight in an autonomous way, defined by waypoints to avoid image coverage gaps, every surface that will be reconstructed needs at the minimum to be covered by at least 2 images taken from different positions. All datasets (photos) of the six missions of different flight heights was processed in Agisoft photoscan software for the generations of Digital Surface Models (DSM) and Digital Ortho Model (DOM), when the camera focal length and the flying height of the UAV are known the scale is determined by this formula: 1 With S =scale, f = focal length of the camera, H= flying height. 59 The result of scale calculations for each flight Height is listed in table 8.1. Table 8.1: The spatial resolution, calculated scale and approximated scale of each DSM type. DSM FA-20 FA-40 FA-60 FA-120 FA-240 FA-360 Spatial resolution (m) 0.37 0.55 0.80 1.73 3.20 4.47 Calculated Scale 1/5540 1/11080 1/16620 1/33241 1/66481 1/99722 Approximated Scale 1/5000 1/10000 1/15000 1/30000 1/60000 1/100000 Category Plans Maps Figure 8.1 shows six DSM of the study area of different spatial resolutions: FA-20 of 20 meters flight Height with a very high resolution data set highlighting all the terrain details even rock textures, passing by FA-60 and ending by FA-360 of 360 meters’ flight Height with a very low spatial resolution and high generalization effect with the «disappear» and «growth» of some terrain morphological features. Fig. 8.1: Multi-scale DSM extracted from UAV photogrammetry methods. 60 These 6 DSM can be classified visually from figure 8.1 by rough and smooth; FA-20, FA-40 and FA-60 for rough and FA-120, FA-240 and FA-360 for smooth, also figure 8.1 constitute an interval of scales and smoothness showing the generalization at different scales. As per table 8.1 different flight Height lead to different spatial resolution (pixel size); the minimum spatial resolution is 0.37 m which is a high resolution showing all terrain details and a maximum resolution of 4.47 m quite good resolution for geomorphological analysis at a local scale. After the scales calculation we categorized our data in two categories plans and maps; the first three DSM of flight Height 20, 40 and 60 are related to the category of plans and the other are related to the category of geographical maps. Multi-scale Digital Surface Models In geoinformatics, scale is predominantly considered a function of the resolution of Digital Elevation Models (DEMs) (Hengl and Evans, 2009; Mac Millan and Shary, 2009). The fast progress in technologies especially the unmanned aerial vehicles (UAV) and new photogrammetry softwares encourages acquisition and processing of Digital Surface Models (DSM) at ever finer resolutions. The scale dependency of land-surface parameters was noted by Evans (1972) as «a basic problem in geomorphometry» (Shary et al., 2002). Meanwhile, the scale dependency of land-surface parameters and land-surface objects has been confirmed by several researches (Chang, Tsai, 1991; Wood, 1996; Florinsky, Kuryakova, 2000; Evans 2003; Hengl, 2006; Arrell et al. 2007; Deng et al., 2007; Pogorelov, Doumit 2009; Wood, 2009). The factor of scale and resolution plays an important role in the uses of digital models in GIS research, the scale may range from micro to macro scale and may be in different levels of measurement, this study will only discuss the spatial scale with the UAV flight Heights. For the selection of an image with appropriate spatial resolution for a study demand the characteristics examination, especially the changing pattern as a function of changes in scale and resolution. As per Li (1993), based on different kinds of philosophy, three types of approach can be identified to the generation of multi-scale representations for a given DTM (Digital Terrain Model), i.e. critical-points-based, smoothing-based, and scale-driven (Li, 1993). Woodcock and Strahler (1987) suggested the use of the local variance method to find images with the optimum scale and resolution. In their experiment, Woodcock and Strahler (1987) used image data were degraded to coarser spatial resolution by resampling the neighbors’ cells and the pixel value of the coarse resolution is an average of a group of finer resolution. In our experiment we have 6 DSM for the same study area with different spatial 61 resolution obtained from UAV different heights survey. The effect of scale (spatial resolution) on these surface models is analyzed according to the morphometric indices by calculating their direct indicators of spatial correlation. Therefore, the experiment implemented the opportunities of multi-scale measurement technology based on UAV. For multi-scale analysis we applied Wood and Strahler (1987) method of Local Variance, the texture method measuring surface properties such as coarseness and smoothness, and as a third index of fractal dimension. Many scientists, quantitively discussed the issue of optimal resolution of digital elevation models. Our study summarized the methods expressing relation between UAV flight Height and spatial resolution, and they are local variance, texture analysis and the fractal method. The first two methods are relative simple and very useful in practical sense, while the third fractal dimension method has a great potential in detecting the resolution effects and is used especially in several geosciences researches. Local variance method proposed by Woodcock and Strahler (Woodcock and Strahler 1987) is such a method, originally developed in image analysis, with potential for dealing with scale in DEM analysis (Li, 2008). Local variance measures the mean of standard deviation within a 3 by 3 pixels in a moving windows, the mean of all local standard deviation values over the entire image were then used as an indicator of local variability contained by the image. The local variance is the degree of similarity between the values of two points depending on the spatial distance between them. The longer the distance is, the higher the degree of similarity is, and the smaller the variance is. According to Schmidt and Andrew (2005), the land surface is hierarchically structured and it can be represented differently across scales for example convex hillslope embedded into a concave hillslope, which in its turn is embedded into a valley. These cases could be detected and seen only on high spatial resolution DSM and are homogeneous relative to scale levels. In figure 8.2 the variance maps of the study area at FA-20 showing all morphological forms in detail even small concave and convex forms with some finesse in FA-40, the same forms became bigger in size and dimensions. In the last stage of plan categories FA-60 the ridges boarding the roads became more highlighted with disappearance of the small morphological forms. Hence in the second category of scales FA-120, FA-240 and FA-360 the degree of smoothness is increasing with the scale. 62 Fig. 8.2: Variance digital models of the six flight Heights. Texture is a measure of coarseness, smoothness and regularity of the surface (Gonzalez, Wintz 1987); texture analysis can measure the spatial variability of image data, and can improve the statistical separation of the otherwise similarity reflecting surfaces (Nellis, Briggs, 1989). The differences in texture for images covering the same area with different scales and resolutions (the case of our study) can indicate the heterogeneity of the scene under observation. In that way we can compare the data between each other, in this case highest texture index indicates the highest variation. Texture is calculated by extracting grid cells that outline the distribution of valleys and ridges. Each grid cell value represents the relative frequency (in percent) of the number of pits and peaks within a radius of ten cells (Iwahashi, Pike, 2007). Similar to local variance methods the texture analysis method is also based on the variability of the geographic data change of scale and resolution, and the scale at which the maximum variability occurs is where most of the relief formation processes operate. By finding the maximum variability of the dataset, we could find the operational scale of the geographic phenomenon and therefore we could decide the observational scale of the study. The texture index in figure 8.3 shows a stability in the morphological forms in the first three scale and an unclear form in the last three level of scales. Consequently, the first three scales allow us to determine the Genesis of the relief-forming processes. 63 Fig. 8.3: Texture digital models of the six flight Heights. Fractal dimension, the concept of fractals fascinates geographers because all the patterns of nature are too irregular and too fragmented to be quantified using the traditional measure of geometric shapes. In real world curves and surfaces are pure fractal that’s why fractal could play an important role in detecting the scale and resolution effects in remote sensing and GIS. Eastman (1985) developed a single pass technique for measuring the fractional dimension of lines. The procedure considers each slope segment, to provide evidence of an underlying angularity that can be considered as the generating angle of the fractal form. The formula is based on calculated slopes as follows: D log(2) 180  slope log(2)  log(sin( )) 2 (8.1) We applied Eastman’s method to calculate the fractal dimension of the six DSMs (Figure 8.4). As it is known, surface fractal dimension values vary inside the interval between 2 and 3. Empirical studies indicate that the fractal dimensions of curves and surfaces change with the scale and resolution, the scale at which the highest fractal dimension is measured may be the scale at which most of the processes operate (Goodchild, Mark, 1987; Lam, Quattrochi, 1992). 64 Fig. 8.4: Fractal dimension digital models of the six flight Heights. The fractal dimensions show the roughness and complexity. Goodchild (1980) found that fractal dimension could be used to predict the effects of cartographic generalization; many scientists used fractals to characterize landscape environmental data, soil, topography (Burrough 1983; Mark, Aronson 1984; Pogorelov, Doumit 2009). The fractal dimension of an image is expected to be lower as the resolution becomes coarser; because coarser resolution is likely to result in low variability in digital surface model, it is argued that the best resolution level for a study is the one that has the highest fractal dimension within a stable scale range, the results of fractal dimensions calculations shown on figure 8.4. If we compare figure 8.2 the local variance index maps with figure 8.4 of fractal dimensions a high degree of similarity is seen, which indicates the usefulness of the application of indices in this analysis. After the application of the three scale indices on the six levels of details a statistical table summarizing in values the generalization between them. 65 Table 8.2: Local variance, texture and fractal dimension statistical values over the six DSM’s. Local Variance DSM Texture Fractal Dimension Min* Max Mean STD Min Max Mean STD Min Max Mean STD FA-20 0 0.129 0.005 0.007 0.189 0.864 0.457 0.170 2.0 2.179 2.008 0.011 FA-40 0 0.136 0.010 0.012 0.000 0.851 0.337 0.182 2.0 2.112 2.008 0.009 FA-60 0 0.143 0.015 0.017 0.000 0.75 0.333 0.157 2.0 2.067 2.006 0.007 FA-120 0 0.458 0.088 0.065 0.009 0.561 0.221 0.125 2.0 2.042 2.007 0.006 FA-240 0 0.811 0.222 0.141 0.160 0.408 0.252 0.059 2.0 2.023 2.006 0.004 FA-360 0 0.985 0.414 0.231 0.211 0.401 0.295 0.050 2.0 2.016 2.005 0.003 *Min – minimal value, Max – maximum value, Mean – mean of the values, STD – standard deviation of the values. The local variance maximum values increase with the flight Height and the scale, the standard deviation and the mean are increasing proportionally in all scales, contrary to local variance values, texture and fractal dimensions maximum, mean and standard deviation are decreasing with the flight Height and the scale as a result of the smoothing effect. To determine the spatial distribution of the analyzed indices at different scales, we applied a simple and effective tool – spatial correlation. Fig. 8.5: Correlation scatterplots of the DSMs values. A correlation scatterplot of all the elevation values of each DSM compared to the main DSM the FA-20, from figure 8.5 we can see a good correlation between FA-20 and FA-40 (coefficient of determination R2 is equal to 78.7%), which decreases between DSM FA-20 and FA-60 (R2=65.4%). 66 The scatterplots of the last three levels show a very poor degree of similarity comparing to FA-20 in fact, it shows a lack of connection between DSM at various levels of scale. It means we are losing accuracy and we have a big generalization lost in the maps category, built using the UAV. Similar results were obtained when comparing the estimated parameters of the texture. Table 8.3: Correlation analysis values between different scales of different indices. Values of R2% DSMs FA-20/FA-40 FA-20/FA-60 FA-20/FA-120 FA-20/FA-240 FA-20/FA-360 DSM Fractal 78.71 63.9 65.39 24.95 40.03 27.04 14.01 20.72 0.78 13.55 Local Variance 71.79 25.09 26.07 18.39 11.89 Texture 78.68 65.28 39.49 13.15 0.79 When interpreting the values of local variance, we consider the following: If the spatial resolution is considerably finer than the objects in the scene, most of the measurements in the image will be higher correlated with their neighbors and the local variance will be low. If the objects to be studied approximate the size of the resolution cells, the value tend to be different from each other and therefore the local variance increases. The resulted correlation values in table 8.3 of DSM and texture are decreasing proportionally with the scale otherwise FA-20/FA-120 at fractal and local variance constitute a barrier of values due to a raised amplitude in R2 values which give a similarity in scale changes between fractal and local variance in a way and in another way between DSM and Texture. The correlation between the values of the local variance has a statistical reliability (significance) only for the DSM with a spatial resolution close to FA20 and FA-40. For cases with large differences in spatial resolution, these indices become essentially independent on the same site. We can conclude that the correlation in local variance decrease with flight Height. Scale and resolutions effects have been and will continue to be an important issue in geographic research. It is essential to have a good understanding of the effects of scale on the analysis results. We tested in this study three morphological indexes (local variance, texture, fractal dimension) that can be used to detect the scale and resolution effects. The results showed a weak relationship of spatial characteristics to compare the images to scale plans and maps. Simultaneously, the survey showed that the characteristics of the local variance and texture are good indicators to determine the optimal spatial resolution of the morphometric analysis of the Earth's surface. 67 Multi-scale landforms classification The factor of scale plays a very important role in Landform classification different levels of measurement (nominal, ordinal, interval and ratio) this study will discuss the terrain analysis with the applications of Terrain Position Index (TPI), Iwahashi and Pike index and the morphometric features and their effects on generalization and spatial resolution at different UAV flights altitudes. Pike et al. (2009) remarked that no digital elevation models derived map is definitive, as the generated parameters differs with algorithms and can vary with resolution and scale. Landform classification stand out with terrain complexity which necessitated specific methods to quantify its shape and subdivide it into more manageable components (Evans 1990) which constitutes a central research topic in geomorphometry (Pike 2002; Rasemann et al., 2004). An Arc Map Jenness module GIS software for landforms terrain computations was applied on a three different spatial resolutions drone based DSM’s for the extractions of Topographic Position Index (TPI), Iwahashi and Pike landforms and the morphometric features at different scales. Throughout the assessment, we comprehensively used this UAV for aerial images acquisition to the generation and interpretation of Digital Surface Models (DSM) by using new photogrammetry technologies. A three DSM of different spatial resolutions, FA-20 of 20 meters’ flight altitude with a high resolution highlighted all the terrain details even rocks texture, passing by FA-120 the terrain is smoothed with some concave and convex areas and ending by FA-360 a very low spatial resolution and a very smoothed terrain of 360 meters’ flight altitude. Topographic Position Index (TPI), the analysis was performed by DSM’s simulation to obtain topographic position index (TPI). The process of formulae (1) calculate the difference between elevation at a specific cell and the average elevation of the neighborhood surrounding cells (Tagil & Jenness 2008); describing higher and lower areas for the classification of the terrain into different morphological forms (Jenness 2005). The simulation required the radius adjustment of neighborhood and its geometric shape based on two different scales or two sizes (Barka et al. 2011). In this study, a radius between 5 m and 25 m was applied to determine the slope positions. TPI Z ∑ (8.2) Where: Z0 = elevation of the model point under evaluation, Zn = elevation of grid within the local window, n = the total number of surrounding points employed in the evaluation. These neighborhood radiuses values were applied for all DSMs spatial resolutions, to be similar in parameters for best comparison analysis. 68 Positive TPI values represent high locations e.g. ridges, negative values of TPI represent low terrain representations e.g. valleys otherwise flat areas have TPI values near zero, high positive values go to high elevations geomorphological structures such as peaks and ridges (Jenness, 2010). The flight altitude FA-20 has a maximum positive value of 1.03, FA-60 of 0.71 and the higher flight altitude FA-360 with 0.48 a decreasing in maximum and minimum values with the increasing of flight altitude. Iwahashi and Pike had developed a Landforms classification unsupervised method based on only three terrain attributes: slope gradient, surface texture and local convexity (Iwahashi and Pike 2007). This method restricts a number of landform classes 8, 12 or 16 with a physical meaning of statistical landscape properties. The unsupervised approach treats topography as a continuous random surface, especially for the three level of details FA-120, FA- 240 and FA-360 independent of any spatial or morphological orderliness imposed by fluvial activity and other geomorphic processes. Morphometric elements, the standard method for the identifcation morphological elements is to establish a mutually position for the central cell in relation to its neighbors (Peucker, Douglas 1974; Evans 1979). The classification algorithm can be done by maintaining the continuity of linear elements, which gives advantages over the method of selection on the basis of logical comparison of neighboring cells (Peucker, Douglas, 1974; Jenson, Domingue 1988; Pogorelov, Doumit, 2009). Morphological elements take the forms of: Planar, pit, channel (thalweg), pass, ridge (division line), and peak. The names of morphological elements may vary in different sources, but they can be uniquely explaining in terms of changes in the three orthogonal components x, y and z (Wood J, 1996; Pogorelov, Doumit 2009). Landform classifications delineated using the TPI method is shown in figure 8.6, TPI values present a powerful way to classify the landscape into morphological classes (Jenness, 2005). Landform Classifications consist of “Canyons, Deeply Incised Streams’’, ‘‘Midslope Drainages, Shallow Valleys’’ and ‘‘Upland Drainages, Headwaters’’ all tended to have strongly negative curvature values of a concave shape, while “Local Ridges or Hills”, ‘‘Midslope Ridges, Small Hills in Plains’’ and ‘‘Mountain Tops, High Ridges’’ all tended to have strongly positive curvature values of a convex shape. Figure 8.6 of the three maps shows land forms classification of all morphological forms listed above at different scale level, a visual analysis of these maps highlight a cartographic generalization between them making a very clear evolution in morphological forms at each stage. 69 Fig. 8.6: maps of landform elements of the three DSM derived from TPI classification analysis. a) FA-120, b) FA-240, c) FA-360. The results of table 8.4 shows how the area of some morphological elements is increasing against other elements relating to scale variations. In table 8.4 the area percentages of some morphological elements are increasing in values and other are decreasing with the scale variation. streams, plains, open slopes and high ridges are increasing in area and geometrical forms due to the variations in spatial resolution. Some morphological elements such as Upland drainage type are not found in any of the three maps and other like Local ridges are disappearing with scales variation and constituting a basic for generalization processes. Open slopes comprised between 6 and 11% of the total area in all flight altitudes while midslope drainages increasing with the flight heights between 9.75% and 11.46% from the total study area. Landforms show a decreasing in their numbers; the dilution of 647886 pixels of different morphological elements from the flight height FA-120 to the flight height FA-360. All the ten morphological elements are affected by scale generalization. Open slopes comprised between 6 and 11% of the total area in all flight altitudes while midslope drainages increasing with the flight heights between 9.75% and 11.46% from the total study area. 70 Table 8.4: Percentage of Morphological elements and pixels numbers of each Morphological element in the three DSMs levels based on TPI classification. Area (%) Type FA-120 FA-240 Number of pixels FA-360 FA-120 FA-240 FA-360 Canyons, deeply incised 5.87 streams 9.32 14.7 6155 1036 800 Midslope drainages, shal- 9.75 low valleys 10.6 11.46 31823 6684 5984 Upland drainages, head- 12.82 waters 11.58 9.7 79684 19287 14025 U-shaped valleys 17.24 13.02 9.01 112465 29124 18901 Plains 13.1 9.69 6.39 158603 42986 22098 Open slopes 10.94 9.1 6.35 153612 42504 21892 Upper slopes, mesas 10.9 10.77 8.87 110730 30022 18910 Local ridges, hills in val- 7.99 leys 9.49 9.53 76494 19574 14150 Midslope ridges, small 6.4 hills in plains 8.88 10.72 35575 6902 6227 Mountain ridges 7.55 13.28 6818 1663 1086 771959 199782 124073 tops, high 4.99 Total numbers of pixels Landforms show a decreasing in their numbers; the dilution of 647886 pixels of different morphological elements from the flight height FA-120 to the flight height FA-360. All the ten morphological elements are affected by scale generalization. To understand the degree of generalization between the big scale of FA120 and the small scale of FA-360 we provided an ascending classification of the geomorphological forms, ridges (Local, Midslope and high) then drainage areas (upland and midslope), hence all other morphological forms are positively affected by generalization by raising their areas. 71 Fig.8.7: landform maps of unsupervised classification (Iwahashi and Pike method), a) FA-120, b) FA-240, c) FA-360. Table 8.5: Iwahashi and Pike landform percentage of areas at different scales. Area (%) Type 1) very steep slope, fine texture, high convexity 2) very steep slope, coarse texture, high convexity 3) very steep slope, fine texture, low convexity 4) very steep slope, coarse texture, low convexity 5) steep slope, fine texture, high convexity 6) steep slope, coarse texture, high convexity 7) steep slope, fine texture, low convexity 8) steep slope, coarse texture, low convexity 9) moderate slope, fine texture, high convexity 10) moderate slope, coarse texture, high convexity 11) moderate slope, fine texture, low convexity 12) moderate slope, coarse texture, low convexity 13) gentle slope, fine texture, high convexity 14) gentle slope, coarse texture, high convexity 15) gentle slope, fine texture, low convexity 16) gentle slope, coarse texture, low convexity 72 FA-120 FA-240 0.00987 ─ 47.33649 48.5 0.00144 ─ 51.13720 51.2 ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ 0.67723 0.2 ─ ─ 0.83771 0.2 FA-360 ─ 49.9 ─ 49.9 ─ ─ ─ ─ ─ ─ ─ ─ ─ 0.1 ─ 0.1 The concavity and convexity of the very steep slope with fine texture found only in high spatial resolution models (FA-120), the coarse texture of high convexity increasing with the pixel size. Step and moderate slopes are not detected in all three models, gentle slope coarse texture high and low convexity are increasing with the flight altitude. Varying DSM spatial resolution can achieve an elements separation of appropriate scale, without the need of generalization. Fig. 8.8: Morphometric features maps, a) FA-120, b) FA-240, c) FA-360. Table 8.6: Surface specific points area percentages of the study area at different scales. Type Planar Pit Channel Pass (saddle) Ridge Peak FA-120 0.00001 ─ 49.72501 ─ 50.27497 ─ Area (%) FA-240 ─ ─ 48.3 ─ 51.7 ─ FA-360 ─ ─ 47.4 ─ 52.6 ─ As per table 8.6 some morphometric features like pit, pass and peak are not detected in all flight altitudes, otherwise planar areas are detected in FA-120 the lower flight altitude at a very low percentage of area in order of 0.00001 %, we cannot judge on this result because the value of this pixel could be a processing artifact. The area of channels is increasing with the flight altitude and the ridge area is decreasing against the channel one. 73 The dominating land forms of surface specific points channel and ridges of the study area form a comparison models of each flight height with TPI land forms. By splitting channels and Ridges of FA-120, FA-240 and FA-360 and exanimating which TPI land forms are included in each type, table 8.7 shows the area percentage of each landform. Table 8.7: Percentage of TPI landforms containing in Ridges and Channels at each flight altitude. Percentage of area TPI Landforms Ridge120 Channel120 Ridge240 Channel240 Ridge360 Channel360 in- 1.5 10.4 2.4 16.8 4.3 26.3 Midslope drainages, shallow valleys 3.3 16.3 3.9 17.8 4.9 18.6 Upland headwaters 5.7 19.7 5.4 18.0 5.6 14.4 U-shaped valleys 11.9 23.1 9.3 17.1 7.0 11.2 Plains 13.7 12.4 9.6 9.6 6.3 6.5 Open slopes 14.7 7.1 11.4 6.8 7.1 5.4 Upper slopes, mesas 16.5 5.3 15.3 5.7 11.5 6.0 Local ridges, hills in valleys 12.9 3.0 14.9 3.7 13.8 4.9 Midslope ridges, small hills in plains 10.7 1.9 14.6 2.7 16.8 3.9 Mountain ridges 9.1 0.9 13.1 1.6 22.7 2.9 Canyons, deeply cised streams drainages, tops, high Upper slopes areas in ridge-120 and ridge-240 occupied a high percentage of areas, for ridge-360 the higher percentage of area goes to Mountain tops. Upland drainage owns high values in channel FA-120 and FA-240 but for FA-360 the higher area goes to Canyons. From these results we can see a TPI landform transition with scales. 74 Ridge 25 20 15 10 5 0 Ridge-120 Ridge-240 Ridge-360 Логарифмическая (Ridge-120) Логарифмическая (Ridge-240) Логарифмическая (Ridge-360) Fig. 8.8: diagram of area percentage of TPI landforms containing in Ridge at flight altitudes of 120,240 and 360 meter. The diagram of figure 8.8 shows the percentage of TPI land forms area in ridges at different scales, the log curves of 120, 240 and 360 have an intersection point at upper slope this point made a transition of values from low percentage to higher percentage of areas. The correlation value of R2 between land forms of FA-120 is 0.6 for FA240 is 0.9 well correlated because of the proportionality and small percentage interval of areas, for FA-360 the correlation value is 0.6 similar to FA-120. 75 Channel 30 25 20 15 10 5 0 Channel-120 Channel-240 Channel-360 Логарифмическая (Channel-120) Логарифмическая (Channel-240) Логарифмическая (Channel-360) Fig. 8.9: Diagram of area percentage of TPI landforms containing in Channel at flight altitudes of 120, 240 and 360 meter. Channel usually are concave areas, in figure 8.9 we can see dominating the area of canyons in FA-360, the correlation of area percentage between the landforms of FA-360 is very high with R2= 0.97 and a concave logarithmic trend line. Otherwise for FA-240 a less concavity logarithmic trend line with R2= 0.75 due to the proportional percentage of areas between landforms. Fa-120 has a low correlation between landforms R2= 0.35 even less than the average. We can conclude from these values that due to cartographic generalization and the transition from flight altitude to other, the degree of similarity for channels landforms areas rising with the flight altitude. Hence for ridges land types the area of canyons and midslope, upper slope local ridge, midslope ridge and mountain tops are increasing with flight altitude, upland drainage, u-shaped valley, plain and open slope area is decreasing with the flight heights. By using Topographic Position Index and unsupervised classification of Iwahashi and Pike, the study area was classified into landform categories of different scale DSM. The result shows that ridges and drainage forms are more affected to generalization than other forms. 76 The landform classes obtained for the three scales differentiate dynamic terrain characteristics of the study area. Landform classifications extracted form drone DSM and GIS fast the presented results and discussion by integrating the geospatial multiscale approach of terrain analysis. The result shows that TPI provided a powerful tool for describing topographic attributes of a study area and there is a relationship between landform map and spatial resolution. By deep understanding of the terrain characteristics, potential and specific constraints of cartographic generalization. Information and methods discussed in this study are valuable results for cartographic multiscale studies and analysis. Landforms are dissolving with scales against each other’s, some of them gaining areas and some disappeared. This study analyzed the generalization at three different scales (flight altitude), for future researches we are planning to examine and monitor changes of landforms at micro, local and global scales. Terrain analysis for parcels evaluation Terrain criteria means geomorphometrical parameters or land terrain quantitative analysis, elevations interval, slope and terrain curvature are the key for parcel evaluation. Topographic and cadastral maps give an idea about the boundary and the terrain structure, for best reputation some experts took into account terrain forms in parcels evaluation. With the appearance of geoinformation technologies applied to real estate, terrain analysis became very easy. Unmanned aerial vehicles (UAV) combined with Computer Vision and Geomatics knowledge, allowed the generation of horizontal and vertical terrain spatial information with positional quality at the centimeter level. Structure from Motion (SFM) a modern tool for UAV's images processing (Westoby et al., 2012) has been used for data generation. This approach, implemented in many processing photogrammetry softwares (PhotoScan, VisualSfM, and others). Horizontal information expressed by Digital Ortho Model (DOM) or simply aerial images and the vertical terrain spatial information is a Digital Surface Model (DSM). In this study at an experimental land in Zaarour region (Lebanon) an elevation, slope and curvature, derived from the DSM processed and integrated in a GIS spatial terrain weighted analysis to output a parcel evaluation map. The mean values of the resulted evaluation map within the parcels boundaries are used for ranking and real estate arrangement. Datasets were collected at a flight height of 360 m above the ground from the taking off point, resulting a swath of 556×456. The images forward and side overlaps were approximately 90% and 70% respectively. The final products of Photoscan standard workflow, described by manufacturer (Agisoft, 2013) are a points cloud, 3D model, an orthophoto and a digital surface model. 77 Fig. 8.10: A) terrain representation DSM of the study area, b) true color Digital Ortho Model. The DSM generated from Photoscan in figure 8.10 a shows the terrain structure (ridges and channels) and elevations varying from 1689 to 1821 meters above the sea level. Terrain texture and pattern in figure 8.10 b highlight a very smoothed bare land and some manmade traces as roads and ponds. Usually real estate experts evaluate lands by their terrain, nowadays GIS modules and algorithms are a very important tool for evaluation and decision making. Terrain analysis and land assessment based on DSM resolved: local relief, slope and terrain curvature for parcels real estate evaluation. The local relief is defined as the difference between the highest and lowest elevations occurring within that area. Local relief was introduced by Partsh (1911), Evans (1972) compared the values of local relief determined over more than one size of area, and he recommended the use of a fairly large sample area. In our study we calculated the local relief inside each parcel with a calculation of all elevation statistic values (minimum, mean, and maximum). The local relief of a big parcel shouldn’t act same as local relief of a small parcel that’s why we divided the local relief value by the area in square meter and multiply it by 100 to reduce decimals value equation 3. Where: Emax – Maximum elevation, Emin – Minimum elevation, A – Parcel area. 100 78 (8.3) The values of V express the difference in elevations related to parcels area, they are in the range of 1 to 59, high values went to parcels with low local relief and big areas for high local relief values. The Slope is one of the most fundamental assessments of parcels and landscape characteristics, and is reported as a driving variable in many construction studies. A parcel with an extreme slope is bad evaluated in real estate market, very difficult accessibility and not useful for investment. The mountainous location of the study area and the nearest to Zaarour country club for winter games gives an idea about the high slopes of the region. A variation of slope from 0.8 % plain terrain to 96 % rocky cliff structures characterized the evaluated parcels. The variety in parcels slopes gave a variety in real estate prices. Terrain curvatures computation is very complicated because, in general, the surface has different curvatures in different directions. With GIS technology and software, terrain curvature became a very easy parameter to calculate, the parameter describing the concavity and convexity of the surface is called curvature, which give a proper assumption about the nature of the land surface with in the parcel boundary. Based on Digital Elevation Models, the most popular algorithms for deriving terrain curvature are those of Evans (1972), Shary (1995), Zevenbergen and Thorne (1987), as well as the modified Evans-Young (Shary et al., 2002) method Burrough and McDonnell (1998) gave preference to the Zevenbergen-Thorne algorithm. For the extraction of the parcels terrain curvatures, we used the Zevenbergen-Thorne’s method which is based on the idea of representation of a surface by an equation of partial quadratic form (Zevenbergen, Thorne, 1987). The Zevenbergen and Thorne module applied to the DSM generated from the photogrammetric processing shows that the concave surface of the study area is 14%, 53% convex and 33% flat areas. These results can be the prove that must of the parcels are hilly as per the DSM 8.10 a. All listed above DSM extracted maps constitute criteria for real estate land evaluation map, McHarg describes the process of conducting multi-criteria analyses by categorizing and ranking values from a variety of thematic datasets, creating a transparency for each dataset, and then overlaying the transparencies together to create a composite image. This final composite image was then used to evaluate the suitable land uses in the design scenario (McHarg, 1995). The value of this weighted overlay approach was understood and adopted by many researchers, who saw great potential from the early GIS programs. The weighted overlay approach allows researchers to create composite maps, in our project a parcel evaluation map for real-estate assessments. These composite maps made it possible for decision makers to consider multiple attributes in a single map (Hopkins, 1977). 79 Multiple Criteria Decision Making (MCDM) methods designed to help stakeholders (real estate experts) make well-informed decisions based on various attributes (Jankowski, 1995). Each criterion evaluation of local relief, slope and curvature takes a value from: 1 to 5 number of classes, 1 being the worst value and class 1 the worst class, otherwise the class number 5 is acting as the best value. Table 8.8: Weights of Local relief, slope and curvature classes. Local relief Classes Slope Curvature Score Classes Score Classes Score 0.04-0.15 5 0-14 5 -110.37--12.07 1 0.15-0.17 4 15-29 4 -12.07--3.52 2 0.17-0.27 3 30-44 3 -3.52-2.1 5 0.27-0.70 2 45-59 2 2.17-12.14 4 0.70-2.52 1 >60 1 12.14-71.26 3 Local relief and curvature raster datasets were classified using the geometric interval classification of ArcMap method. This classification method was used for visualizing continuous data, the specific benefit of the geometrical intervals classification is that it works reasonably well on data that are not distributed normally. Local relief calculated by formulae 1 gives a range of values from 0.04 to 2.52, the small values with small interval of elevation and high values with big interval the highest score went to the small values and gradually going down to 1 for the interval of big values. Curvature negative values are concave terrain forms and positive values are convex terrain forms the minimum concave extreme values of the first interval -110--12 are found in deep valleys their score is 1, the highest score went to the interval of plane surfaces. Slopes are a very important index in parcel evaluation, real estate expert and contractors inspect the slope as a first stage in their studies and reports, more the slope is high more the terrain is not suitable and the price is low. Slope was calculated with the ArcMap algorithm based on DSM and classified manually in table 8.9, the slope of the study area is moderate 83% of the area fall in the slope interval between 0 and 29 percent, high score went to the lowest interval 0-14 and the lower score of one is for the extreme slopes. 80 Table 8.9: Percent Slope Ranges of DSM Data Sets, Associated with Scores. Score 5 4 3 2 1 Slope Interval % 0-14 15-29 30-44 45-59 > 60 Area % 43 40 12 4 1 Fig. 8.11: Сriteria classified maps a) local relief, b) slope c) curvature, d) resulted real estate. The resulted three classified datasets maps of local relief, slope and curvature in figure 8.11 shows the high score values in red color and the low score pixels in dark green. The method used is weighted overlay analysis, performed by overlaying classified datasets, assigning a weight to each dataset, summing the values of each vertical cell stack, and then evaluating the resulting composite map (Collins et al., 2001). The developed composite map of parcel evaluation for real estate requires the analysis of criteria illustrated in Table 8.9, in our study, criteria are not all of equal importance. In a single criteria map we must prioritize values. Values have been reclassified in the input criteria maps, slope has the big influence on terrain 81 evaluation and assessments that’s why it took a weight of 60% and for Local relief and curvature 20 % for each. Criteria maps have been weighted and aggregated together to produce parcel evaluation composite map figure 8.11d. High scores in composite map figure 8.11d are found in the smoothed plain areas hence low scores are found in extreme valleys. The cadastral vector map of the study area draped on the resulted raster composite map with zonal statistics (maximum, minimum and mean) of the scores inside parcels boundaries, the evaluation real estate map classified manually in 5 classes using the mean values within each parcel. Fig. 8.12: Parcels evaluation real estate map. The parcels evaluation real estate map of the study area constitutes a real estate ranking coefficient, parcels of the first class are not influencing on the square meter price but the class number five is five times better then class one. The results of this study is satisfactory, it would suffice to have all reliable and necessary data relating to the case study if better outcomes were to be expected. These data are easily introduced in the system and can be updated in any time, we can add more criteria such as neighborhood analysis, accessibility analysis etc. Therefore, using the developed system, results should be closer to reality since all criteria and them relative importance is taken into account at the same time. The main aim of this research was to determine whether UAVs DSM can offer a suitable material for real estate parcels evaluation. To achieve that, a UAV flight was done to produce DSM for terrain analysis. Afterwards the ter82 rain analysis results provide parcels evaluation criteria used for the creation of a parcel evaluation real estate map. The photogrammetry processing results in terms of precision are acceptable, since the level of precision only depends on pixel size. The resulted values of figure 8.12 could be a coefficient of parcel pricing and real estate market arrangement, the parcel evaluation real estate map is a basic for decision makers and experts. The issue addressed is land evaluation for real estate. The ideal solution would be to incorporate a module including important classification methods in a GIS as well as appropriate analysis methods independently of data, of the study area. It will represent a spatial decision support system dedicated to developing land evaluation map for real estate parcels assessments. Multi-scale Terrain Roughness Surface roughness could be defined as a value ranging between smooth and complex surfaces, this study specifically focuses upon the broad area at different scales of general geomorphology (Evans, 1729) and, more explicitly, on the quantification of surface-roughness variability using Digital Surface Models (DSMs) generated from UAV. Surface roughness is treated here as a geomorphometric variable influencing at the physiography of the terrain, not as a parameter due to the precision and accuracy of the generated digital surface models. Measurement of terrain roughness is important for a number of disciplines of terrain quantifying characteristics have been evolving within fields such as geomorphology, engineering, biologists and ecologists (Doumit, 2017). In terrains descriptions, roughness parameters should be established that can be used to describe surface irregularities and they should fulfill some requirements. The parameters should be descriptive and give the reader an image of the physical characteristics of the study area and should be easily measurable in the field so that large sites can be quickly sampled. If possible, roughness parameters should be selected that require similar types of field measurements with a minimal amount of equipment. Nowadays with the appearance of the Unmanned aerial vehicles and the advanced of Geographical Information Systems these parameters can be measured and compared at several different scales, and suitable for statistical and numerical analysis. The simplest traditional method of terrain complexity is the profile method, by providing multi-sections on the terrain it is very easy to evaluate roughness of the terrain. Hobson among the first scientists who calculated terrain Roughness using computer technologies, he wrote in Fortran language modules for calculating roughness parameters such as: comparison of the estimated actual surface area with the corresponding planar area; bump elevation frequency distribution; and the distribution of planes (Hobson, 1972). 83 With the fast evolution of GIS and geoinformatics methods, many scientists worked on the development of other methods for calculating terrain roughness such as: the application of Fourier analysis (Stone and Dugundji, 1965) geostatistics (Herzfeld, et.al., 2000), the fractal dimension of a surface (Elliot, 1989; Doumit, Pogorelov, 2017). From the first recognized traditional methods for quantifying roughness was the land surface roughness index (LSRI) developed by (Beasom et al.,1983). This index is a function of the total length of topographic contour lines in a given area. (Riley et al., 1999) developed a terrain roughness index (TRI) that is derived digital elevation models (DEM) implemented in a geographical information system (GIS). TRI uses the sum of changes in elevation within an area as an index of terrain roughness. Based on (Hobson, 1972) method developed for measuring surface roughness in geomorphology, a Vector Roughness Measure (VRM) quantifies terrain roughness by measuring the dispersion of vectors orthogonal to the terrain surface. In this study we tested the regression between VRM and TRI values at the six different levels and we provided a correlation analysis between the raster datasets of VRM, and TRI, to examine their distributions within each scale, we generated scatterplots and calculated descriptive statistics (Min, Max, SD, skewness, kurtosis and r2) to characterize terrain heterogeneity at different level. Our study is independent from DSM accuracy and precision it will test roughness at six different levels expressed by flight height of a drone at 20, 40,60,120,240 and 360 meters. The flight datum was calculated from the same takeoff points of the drone of the six flights. As this study is restricted to evaluating array-based geomorphometric methods for calculating surface Roughness, an input DSM is required for further analysis. DSM selection criteria were based on spatial resolution, with a highspatial-resolution DSM required in order to test the heterogeneity across a range of resolutions and within the study area presenting multiscale Roughness features. The Terrain Roughness Index (TRI) based on an index described by (Riley, et. al. 1999) that calculated the sum change in elevation between a grid cell and its eight neighboring grid cells table 8.10 by squaring the eight differences in elevation, summing the squared differences, and taking the square root of the sum. (Valentine,et.al. 2004) calculated the average of the absolute values of the eight differences in elevation, by using the TRI equation given as: 0,0 1,1 0,0 0,0 1, 1 1, 1 0,0 0,0 0,1 1,1 /8 84 0,0 1,0 0,0 0, 1 0,0 1,0 (8.4) Table 8.10: 3×3 grid of the TRI equation values. 1,-1 0,-1 1,-1 -1,0 0,0 1,0 -1,1 0,1 1,1 Fig. 8.13: TRI maps at different flight altitudes 20, 40, 60,120,240 and 360 above the datum. TRI high values at FH-20 shows details in ridges and water erosion traces, in FH-60 structures are very smoothed, FH-120 shows the pixel's boundaries and at FH-360 the map is totally pixelated. It is very clear in this map the disappearance of the small structures with the loss of spatial resolution, running from coarse to smooth then to pixelated surfaces. Based on a method developed for measuring surface roughness in geomorphology (Hobson, 1972), the surface of elevation values can be divided into planar triangles very similar to Triangulated Irregular Network (TIN models) and normal to these planes represented by unit vectors. Values of vector mean strength (R), and dispersion (k) can be calculated for each square cell. In smooth areas, with similar elevations, the vector strength is expected to be high and the vector dispersion to be low since the vectors will become parallel figure 8.14. In rough areas, the nonsystematic variation in elevation will result in low vector strength and high vector dispersion. The inverse of k can be a better representation of roughness (Mark, 1975). 85 Based on slope and aspect definitions, normal unit vectors of every grid cell of a digital elevation model (DEM) are decomposed into x, y and z components. Fig. 8.14: Vector dispersion method used to calculate surface roughness at different scales for a topographical surface. Graphic from (Grohmann et al. 2011). DSM resolution dependent from the flight height, in figure 8.14 the topographic surface profile showing the terrain variation, at high spatial resolution vectors are very dense and orientated in several directions otherwise for low spatial DSM resolution as per example FH-360 vectors and far from each other perpendicular to segments expressing geometrical terrain forms. The translation from the vector dispersion traditional method applied on topographic maps to Vector Roughness Measure (VRM) calculated by GIS algorithms, was done by applying the method and formulas used by (Veitinger et al., 2016). Based on slope and aspect definition, the normal unit vector of every grid cell of a Digital Surface Model is decomposed into x, y, and z. A resultant vector R is then obtained for every pixel by summing up the single components of the center pixel and its neighbors using a moving window technique. ∑ ∑ ∑ (8.5) The magnitude of the resultant vector is then normalized by the number of grid cell and subtracted from 1. 1 (8.6) Where VRM is the vector ruggedness measure (Veitinger et al., 2016). Figure 8.15 shows the six VRM maps generated from DSM, by using formulae 6, for the first three high spatial resolution FH-20, FH- 40 and FH-60, terrain structure are very fine highlighted similar to TRI map of figure 8.15. The two indices TRI and VRM of the resulted roughness maps showed a loss in ter86 rain heterogeneity and a trend to terrain homogeneity by a high degree of smoothness especially in the last three DSMs FH-120, FH-240, and FH-360. VRM measures the variation in terrain independent of its overall gradient, VRM is able to differentiate among terrain types. Fig. 8.15: VRM maps of the six DSMs In this work, we have tested two widely used methods: Terrain Roughness Index (TRI), Vector Roughness Measure (VRM), Terrain Roughness Index (TRI) calculates the sum change in elevation between a grid cell and its neighborhood, according to the algorithm by (Valentine, et.al., 2004). Table 8.11: Terrain Ruggedness Index statistical values at each level. Std. – standard deviation; Skew – Skewness; n – number of cell in a raster grid. TRI-20 TRI-40 TRI-60 TRI-120 TRI-240 TRI-360 Mean 0.116 0.171 0.211 0.520 0.822 1.113 Std 0.062 0.085 0.101 0.198 0.302 0.381 Skew. 1.202 0.702 0.602 -0.424 -0.385 -0.723 Kurtosis 2.861 1.233 0.575 -0.045 -0.323 -0.448 n 40436 18522 8891 1901 559 286 Min. 0.005 0.008 0.016 0.037 0.136 0.152 Max. 0.544 0.591 0.636 1.093 1.695 1.874 Median 0.112 0.172 0.208 0.552 0.879 1.234 r2 0.0014 0.00006 0.0059 0.0081 0.0033 0.0292 The statistics of the TRI values at each flight height listed in table 8.11, the values of Min., Max., Mean and Std. showed that the TRI values increased with the flight height hence with the scale. From the values of r2 it is proven that no homogeneity of TRI values with their neighborhoods in each layer, it is nor87 mal especially for the high spatial resolution layer TRI-20, TRI-40, and TRI-60 with high n values. For TRI-20 no symmetric data distribution because of the high skewness value of 1.202, but the evidence is that negative values for the skewness at TRI120, TRI-240 and TRI-360 indicate data that are skewed left and positive values for the skewness indicate that high spatial resolutions layer TRI-20, TRI-40, and TRI-60 skewed right. Table 8.12: Vector Ruggedness Measure statistical values at each level. n Min. Max. Median r2 Mean Std Skew. Kurtosis VRM-20 0.021 0.019 2.219 6.906 40436 0 0.166 0.017 0.013 VRM-40 0.021 0.017 1.753 4.757 18522 0 0.130 0.018 0.007 VRM-60 0.015 0.012 1.512 3.137 8891 0.0001 0.080 0.013 0.026 VRM-120 0.019 0.011 0.384 0.563 1901 0.0001 0.063 0.020 0.0001 VRM-240 0.015 0.008 0.119 -0.561 559 0.0006 0.037 0.015 0.0009 VRM-360 0.014 0.006 -0.289 -0.914 286 0.0009 0.027 0.015 0.0081 The distributions of roughness values (VRM) for the five levels were highly skewed to the right with the highest proportion of VRM values at the mean instead of FH-360 values skewed to the left. Our results showed that TRI and VRM directly measured heterogeneity of terrain more independently of scale, and both indices exhibited a pattern of bias in that the minimum value of roughness increased with increasing spatial resolution. A correlation analysis provided to understand the similarity between TRI and VRM. Fig. 8.16: Scatterplot of TRI and VRM ruggedness values at all levels of details. a) FH-20, 88 b) FH-40, c) FH-60, d) FH-120, e) FH-240, f) FH-360. High correlation recorder at all flight heights, the scattered plot of figure 8.16 shows a high degree of similarity in small values at FH-20, FH-40 and FH60 expressed in dark color elongated areas of figures 8.16 a b and c. At high flight height the concentration of the correlated values is moving from small to mean values with a trend to the right figure 8.16e, otherwise the correlation values of TRI and VRM in figure 8.16 f became more scattered and less dense due to a dilution of similarity resulted from the changing of the spatial resolution (pixel size). We can say from figure 8.16 that the two roughness indices are very similar and have a high correlation and the degree of terrain roughness vary with the spatial resolution. Differences in the distributions of roughness, measured by VRM, and TRI reflected the characteristic terrain physiography of the terrain. Surface Roughness in Earth sciences is used as an explanatory index. It is dependent upon exogenic and endogenic geographical processes. Many methods for surface Roughness measuring such as: area ratio, vector dispersion, the standard deviation of first and second terrain derivative (elevation, slope, and curvature) have been implemented in GIS and based on digital models. The possibility of the production of digital models at different spatial resolution spatially UAV based one, allows fast and inexpensive multiscale analysis of surface Roughness. Two applied indices Topographic Roughness Index (TRI) and Vector Roughness Measure (VRM) at different scale level express a variety in terrain heterogeneity at a UAV flight height of 20, 40, 60, 120 240 and 360. Both indices show a roughness variation with scales and a transition from coarse to smooth between FH-60 and FH-120, a cartographic generalization influenced by flight height is very clear in figure 8.14 and 8.16. Our statistical and correlation analysis of roughness indices prove that multiscale and multilevel UAV flights datasets are: a visual cartographic generalization, a transition scale from level to another, a live roughness monitoring apparatus leads to a detection of fine scale/regional relief, and performance at a variety of scales. Researchers must be aware of potential biases that originate in DSM at multiscale (different spatial resolution) when TRI and VRM values are interpreted. All DSMs contain inherent inaccuracies due to the sources errors in original data. The elevation accuracy of a DSM is greatest in flat terrain and decreases in steep terrain where the roughness incises (Koeln et al., 1996). Terrain roughness is a complicated geomorphometric parameter, it could be calculated in many ways, under many names roughness, micro relief, and others. Digital Surface Models to planar areas. Rugosity is a measure of topographic heterogeneity, it describes and counts discrete structures, from categorical to quantitative (McCormick, 1994). 89 Rugosity is an index of surface roughness that is widely used as a measure of landscape structural complexity. Rugosity is traditionally evaluated in-situ across a two dimensional terrain profile by draping a chain over the surface and comparing the length of the chain with the linear length of the profile figure 8.17. Fig. 8.17: The rugosity of a surface (e.g. yellow profile of a terrain № 1) is the ratio between the contoured distance (dashed line № 2) and the planar distance (or area for three-dimensional data). Two-dimensional rugosity is defined as the ratio between the surface contour distance and the linear distance between two points (Risk, 1972) and is synonymous with the term tortuosity or the arc-chord ratio (Moser et al., 2007). There are multiple measures of structural complexity (McCormick, 1994), for over forty years many scientists have used the rugosity index for topographic (elevation) or bathymetric (depth) datasets; rugosity can be calculated from twoor three-dimensional data at any scale. Drones based Digital Surface Models at different flight heights with different scale could be a good material for rugosity analysis at different scales. The standard surface ratio (SR) method for calculating a planar distance is to project the surface onto a horizontal plane (solid line), this method confounds rugosity with the slope (θ) at the scale of the surface (solid line). Figure 8.17 is a very expressive section, the length of the section number 1 natural terrain with 5.8 km a very complex terrain, section number 2 smoothed terrain owning a similar shape as the natural terrain but with 5.3 km length. Section number three express a slope projection of the natural terrain with a 3.5 km length. The last section a horizontal planar one number 4 with 2.8 km influenced by the slope angle (θ), the big change in length from section 1 to the horizontal section 4 is approximately the half. The idea here is not the changing in length nor the surface, but the rugosity change at different DSM spatial resolution. A six DSM generated from different UAV flight height constitute the base of our study by comparing the transition to planar areas at different levels. 90 In the early 1970s The standard surface ratio (SR) method for measuring rugosity was introduced by (Risk 1972; Dahl, 1973). They calculated rugosity by projecting the surface onto a horizontal plane (Lundblad et al., 2006; Wright, Heyman, 2008; Friedman et al., 2012) thereby coupling rugosity with the slope at the scale of the surface equation 1. (8.7) in the case of equation 7 the rugosity increases with increasing of the slope figure 8.17; the law of cosines, where cos (θ) = adjacent side ÷ hypotenuse side), here lies the fundamental issue presented by the traditional methods for measuring rugosity. A flat surface has a rugosity value of one, while a rougher surface, or a surface with more relief, has a higher rugosity value than one, rugosity encompasses and combines both structural relief and roughness (Moser et al., 2007). A new arc–chord ratio (ACR) rugosity index for quantifying landscape structural complexity was developed by Du Preez (2014) to overcome significant issues presented by traditional rugosity indices. In comparison to other methods for measuring rugosity, ACR rugosity is separated from the slope and easy to execute an ACR rugosity analyses using the GIS software (Du Preez, 2014). Many marine and land studies use rugosity such as environmental risk assessment, species management, distribution and conservation, predictive mapping of vulnerable marine and land ecosystems (Stambaugh, Guyette 2008; Wedding et al., 2008; Galparsoro et al., 2009; Woodby et al., 2009). The ACR three dimensional method is tested on the six generated DSM and compared at different levels of spatial resolution by correlation and statistical analyst. Spatial resolution is a very important factor of scale influence on rugosity, to prove that our study will find the answer to the question: did the transition of the multiscale Digital Surface Models to planar areas have the same results? Many methods of evaluating rugosity on a three dimensional surface have been proposed. These methods measure a ratio of areas rather than lengths, as shown in Equation (7). Surface Area to Planar Area (SAPA) method introduced by Jeff Jenness evaluates rugosity using a 3 x 3 neighborhood, by drawing a line from the center of each cell in the window to the center of the central cell in three dimensions. The result is a network of eight triangles in the central cell which approximates the contoured surface at the cell location. The sum area of these triangles is divided by the two dimensional cell area to obtain a measure of rugosity (Jenness 2004). Du Preez and Tunnicliffe (2012) propose a novel method for measuring rugosity that decouples rugosity from the slope and is consistently independent of data dimensionality and scale. it is a simple adaptation of an Arc Chord Ratio 91 (ACR). The method replaces the horizontal plane with a plane of best fit (POBF), where the POBF is a function of boundary data interpolation (Preez, Tunnicliffe, 2012). The ACR method can be used in multi-scale analyses, an important attribute of a spatial analysis as morphological processes act at a variety of spatial scales (Levin, 1992), and differ in effects and importance with scale (Wu, 2013). Basing on Du Preez (2014), Jeff Jenness developed A new technique, operates on a 3×3 neighborhood, using the triangulated area of each adjacent cell and applying the Pythagorean theorem to compute the surface area. By default, the planar area of each grid cell is corrected by dividing the cell area by the cosine of slope (Jenness, 2004). In our study ACR was calculate in GIS tool installed on ArcGIS® Software available for download in Du Preez (2014), from the six generated DSMs we calculated Arc-Chord ration (ACR). The ACR rugosity index is a measure of three-dimensional structural complexity defined as the contoured area of the surface divided by the area of the surface orthogonally projected onto a plane of best fit. The arc-chord ratio (ACR) method calculates the planar distance by projecting the surface boundary onto a boundary data section 1 figure 8.17 (Red dashed line) plane of best fit section 3 figure 8.17 (POBF; dashed-dotted line; 3) effectively decoupling rugosity from the slope at the scale of the surface. ACR is calculated by creating two TINs a contoured surface and a planar one representing the plane of best fit (POBF) figure 8.18a. The POBF is a function (interpolation) of the boundary data only of the area of interest the area of interest is the boundary of the study area. The surface area of the first TIN within is divided by that of the second TIN to obtain a single ACR value for the area of interest (Du Preez, 2014). Fig. 8.18: A) ACR simultaneous surfaces leading to the horizontal planar one, b) an example of ACR surfaces of FH-20. As a first step, the conversation from raster to TIN for the six DSMs to form contour surfaces at different scales. 92 Fig. 8.19: TIN models of contour surface at the six flight heights. All the six TIN models expressed the terrain morphology with some variations detected on the colored contour lines, and a generalization in triangles quantity. Step two, the contoured surface translated to a plane of best fit (POBF) decoupling rugosity from the slope at the scale of the surface (Du Preez, Tunnicliffe, 2012; Friedman et al., 2012). Figure 8.20 shows six similar planar surfaces owning the same trend of values by simplifying elevation values of the contour surfaces. The innovation of the ACR method lies in the analysis used to generate the POBF: identify and isolate the boundary data (step three) figure 8.18a an illustrated example the boundary data of FH-20 the triangulated irregular network data frame. By using a linear polynomial interpolation of the boundary data at the six levels to generate surface datasets (step four). 93 Fig. 8.20: the POBF of the six flight heights. Some softwares are unable to interpolate the actual planar area, an alternative is to interpolate the angle of the POBF and use the cosine equation (and the horizontal planar area) to extract the planar area of step five (Du Preez 2014). To solve for the ACR rugosity index, by application of formula 7 using the contoured and planar areas (step six). By following Du Preez and Tunnicliffe (2012) arc-chord ratio (ACR) we Computed a ratio between the three-dimensional surface area and the planar area of the surface, this tool uses a novel methodology to develop a surface area dataset. The output values represent ratios between the surface area and planar area, typically ranging from 1 in flat areas to 4 in areas of high variation. The first step of Du Preez methodology of conversation from raster to TIN, figure 8.19 of the similar six TIN models. Same study area at different spatial resolution lead to visual data similarity a statistical comparison was done to test this degree of similarity table 8.13. The statistical values of table 8.13 shows a decreasing in triangles numbers from 321 to 41 due to decreasing in spatial resolution, the difference in average maximum and minimum elevations due to the interpolated predicted values. 94 Table 8.13: Elevation TIN statistics, quantity of triangles, average minimum and maximum elevations and the slope average. Flight height, Meter 20 40 60 120 240 360 Contoured area TIN Quantity of Average Min. Average Max. triangles Elev., Meter Elev., Meter 321 1728.95 1729.94 230 1730.26 1731.53 1734.85 172 1733.58 85 1718.65 1720.76 1729.25 58 1727.14 1748.87 41 1746.67 Average Slope % 15.89 16.70 14.52 17.03 14.48 13.00 The quantity of triangle from elevation area to planar area of table 8.13 and table 8.14 are reduced more than twenty times in high spatial resolution data of 20, 40 and 60 meters’ flight heights, otherwise low spatial resolution data with low triangles quantity in contoured areas are reduced less than ten times in planar areas TIN models. Average of maximum and minimum elevations approximately in all levels are reduced in a range of 2 meters between surface and planar areas, hence a reduction on the average slope in the range of 2 percent. Table 8.14: POBF, planar TIN area statistics quantity of triangles, average minimum and maximum elevations and the slope average. Flight height, Meter 20 40 60 120 240 360 Planar area TIN Quantity of Average Min. triangles Elev., Meter 1725.72 10 1726.72 9 1732.90 9 1715.46 8 1724.62 8 1746.22 7 Average Max. Elev., Meter 1731.67 1732.41 1732.98 1722.23 1729.84 1746.66 Average Slope % 13.23 14.13 10.44 14.87 11.72 10.75 In planar area TIN, the average values of the minimum and maximum elevations in all the six flights is reduced with the number of triangles due to the transition from contoured to planar area. The variation in values between table 8.13 and table 8.14 showed an unstable change in elevations and slope. The first part of the transition from contoured area TIN to planar area TIN (POBF) is very similar to trend analysis simplifying the complexity of values with a conservation of the same datum. Otherwise the second transition part from surface area two planar one record a loss of initial datum elevation down to zero table 8.15. 95 Table 8.15: Surface areas statistics at different flight heights. Flight heights 20 40 60 120 240 360 Surface Area (Meter) Min Max Mean 0.144 0.227 0.141 0.319 0.415 0.312 0.657 0.756 0.645 3.073 3.342 3.007 10.867 11.3 10.7 20.291 20.705 19.976 Std 0.005 0.008 0.013 0.048 0.105 0.172 High spatial resolution data of 20, 40 and 60 have sub meter surface area values, rising with approximately in double values between flight heights. The surface area is designed to determine the amount of similarity between the tested area surface and planar surface. it is hypothesized that the surface area increases with surface irregularity. Because there is a definite interplay between the number and magnitude of terrain irregularities such that similar surface area estimates could arise from different manipulations of these two variables. Table 8.16: Planar area statistics at different flight heights. Flight height 20 40 60 120 240 360 Planar Area (Meter) Min Max Mean 0.99 1.154 1 0.99 1.132 1 0.99 1.045 1 0.99 1.013 1 0.99 1.003 1 0.99 1.003 1 Std 0.005 0.003 0.002 0.001 0.0006 0.0005 The values of planar area at different scales is practically the same with small variations at low flight height, basing the results of table 8.16 especially the similar mean values, we can answer the above question constituting the target of our study by: yes, the transition of the multiscale Digital Surface Models to planar areas have the same results. Visual and statistical results prove the similarity of multiscale planar data, a regression analysis run to test this similarity in planar surface at multiscale. 96 Fig. 8.21: scattered plots of planar area at different scales. The correlation analysis between the highest spatial resolution data of FH20 as a reference and the other datasets, figure 8.21a a test scatterplot with same data set in X and Y axes of FH-20, with hundred percent similarity. The graph FH-40 with FH-20 gives 39.77 % of similarity figure 8.21b, r square values of FH-60, FH-120, FH-240 and FH-360 against FH-20 are less than 11 %, the core of the scatterplots for the high spatial resolution datasets are in the lower left corners, moving positively with the Y axes in FH-120 then falling down negatively for FH-360. The correlation analysis contrary to visual and statistical showed a difference in multiscale planar areas. The present study provides multiscale DSM analysis by adaptation and improvement of ACR geo-processing model tools and step-by-step application of Du Preez 2014 module. Improving standard methods for the detection and investigation of geomorphological patterns at different spatial resolution will lead to better scientific information for generalizations, terrain analysis, management and conservation initiatives. We can conclude from this study that Scale and resolutions effects on terrain data and form an important issue in geographic researches, DSM UAV based at high flights should be tested before use. It is essential to have a good understanding of the effects of scale on the analysis results, each elevation data has its own surface to planar result, terrain rugosity depends with spatial resolution and visual analysis should follow a correlation one. 97 Chapter 9 Drones regulations and buyers guide Drones Regulations When it comes to flying professional drones within the law, the respective rules around the world are varied, till now there is no unique regulations for all countries. Aviation authorities around the world are integrating unmanned aircraft into civilian airspace, and each jurisdiction has its own rules and regulations. As an example in USA, you must acquire a Certificate of Authorization (COA) through the Federal Aviation Administration (FAA). While the requirements of each COA include (Reg Austin, 2010):  Flights below 300 meters. Daytime operation in Visual Flight Rules (VFR).  Range limited to Visual Line of Sight (VLOS).  Greater than 4 km from an airport.  Flight altitude ceiling of 120 m above take-off point is common.  Maximum UAV take-off weight/weight classes (a lighter weight increasingly equates to more flexible usage).  No, fly zones, such as within several km of an airport, dense urban areas and military places.  An operator certificate/license of some description is often required.  A second drone observer is sometimes required. The investigations reveal that UAV regulations are subject to national legislation and focus on three key issues: 1) Targeting the regulated use of airspace by UAVs as they pose a serious danger for manned aircrafts; 2) Setting operational limitations in order to assure appropriate flights. 3) Tackling administrative procedures of flight permissions, pilot licenses and data collection authorization in order to address public safety and privacy issues. Approximately half of all countries do not provide any information regarding the use of UAVs for civil applications, some countries have UAV regulations were in other countries the use of UAVs is prohibited (Stöcker et al., 2017). Flight safety A drone operator is responsible for ensuring the safety of every operation, including the protection of nearby people, animals, property and the environment in general, he should evaluate weather conditions and choose suitable, safe take-off and landing locations (Reg Austin, 2010). For a safe flight drone operator must: 1. Flight within line of sight LOS. 2. Do not interrupt the LOS. 98 3. Do not fly with wind over 30 km/h. 4. Do not fly in fog. 5. Do not fly in rain. 6. Do not fly long periods (the ESC got hot). 7. Stay away from birds. The ICAO is an international actor that serves as a collaboration and communication platform for national civil aviation authorities. They are concerned with fundamental regulatory frameworks at a global scale and provide information material, Standard and Recommended Practices and Procedures for Air Navigation Services (ICAO 2011). One important step taken in Riga 2015 was the publication of the Riga Declaration on Remotely Piloted Aircrafts. The declaration highlights five main principles that should guide the regulatory framework in Europe: 1) Drones need to be treated as new types of aircraft with proportionate rules based on the risk of each operation; 2) EU rules for the safe provision of drone services need to be developed now; 3) Technologies and standards need to be developed for the full integration of drones in the European airspace; 4) Public acceptance is key to the growth of drone services; 5) The operator of a drone is responsible for its use (Riga declaration 2005). In some countries the required distance bounds a vague interpretation and strictly determines the term visual line-of-sight (VLOS). Some cases further include extended visual line-of-sight (EVLOS) operations. Here, the pilot uses an additional observer or remote pilots to keep the visual contact to the UAV figure 9.1. The US, UK, Italy and South Africa particularly mention the possibility of EVLOS operations within their UAV regulations. Furthermore, some countries basically allow beyond visual line-of-sight (BVLOS) flights. BVLOS flights, which are outside the general permission for the commercial utilization of UAVs, require either special flight conditions or exceptional approvals (Stöcker, 2017). 99 Fig. 9.1: Schematic distinction between UAV flight ranges. Besides this, the majority of the cases demand a pilot certification or a license. A certificate is usually granted by intermediaries like authorized training centers or UAV manufacturers and entails a basic practical and theoretical training of the pilot. In contrast, the procedure and requirements to obtain a UAV pilot license usually involve sophisticated aeronautical background knowledge and is issued by national aviation authorities (Stöcker, 2017). Drones buyers guide There are many different types of UAVs on the market today. The solution you choose depends on various factors, such as the type of data that is to be collected, the safety requirements, and the time and distance the drone must be capable of flying in a single mission. Deciding on the type of UAV that is most suited to your project’s needs and requirements is not always an easy task. Below some guidance on things to consider and the questions to ask when you decide to purchase a drone. Questions to ask when choosing a drone and things to consider before purchasing a UAV: The payload, how much weight and volume is needed? Redundancy, for how long and how far is the drone needed to fly? Safety, where and when am, I flying? Are there special regulations in place? Transport (Collapsible, portable, easy-to-carry)? Software Accessories (ground station with hardware and software for mission planning)? User friendliness, How easy is mission planning, start, landing, flight, and data collection? Customer service, is maintenance and support available? Price, What is the total cost of ownership? Software, what are the post processing software capabilities included? 100 Conclusion Today, UAVs can be used as a precise, automated and computer controlled data acquisition and measurement platform, this work deals with the challenging task: The use of UAV systems as photogrammetric data acquisition platform. The main goal of this book is the identification of a generic workflow using a UAV as a photogrammetric data acquisition platform investigated on a real application, focusing on the precision and resolution of the generated photogrammetric products, like elevation models, orthophoto and textured 3D models. The main motivation of this book is to generate high-resolution information of inaccessible areas, such as maps, orthoimages, general topography, detailed elevation models, extraction of dangerous obstacles. The extraction of the terrain, orthoimages and textured 3D models from UAV-images or other sensor data can be applied to all kinds of hazards, catastrophic or environmental disasters, including 3D documentation of the environment, cultural heritage sites, etc… Drones technologies allow cartographers to produce better mapping, coupled with multispectral, hyperspectral and LiDAR systems. The evaluation of low-cost UAV systems showed that these systems are easily controllable and usable for scientific researches. Many challenges still remain as UAV are adopted in the geospatial industry, this book showed that flying the photography did not present as many challenges as the postprocessing of the resulting spatial data. This book demonstrated that unmanned aerial vehicles (UAV) can be used to produce geospatial data in the geographical analysis. The modular nature of UAS and the ability to produce current spatial information (orthophotos, 3D models) mean that we no longer need to turn to large mapping contracts because of the economies of scale offered by the latter. Today, UAVs can be used as a precise, automated and computer controlled data acquisition and measurement platform, this work deals with the challenging task: The use of UAV systems as photogrammetric data acquisition platform. The key with using all photography or imagery, whether aerial or satellitederived, is to only use the data in a sensible and meaningful way, understand the error levels and their potential impact. Geographic correction can be a complicated process, full of pitfalls it should transform a better image into a useful GIS information. The acquired data has great potential in GIS projects. Moreover, our study showed also its usability for teaching Students to use new technologies.in future work UAVs will be integrated as a fixed part in the remote sensing and GIS courses. 101 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. Abbeel P., Coates, A., Quigley, M. and Ng, A. Y. (2007). An application of reinforcement learning to aerobatic helicopter flight, In: NIPS, 19. Abdel-Aziz Y. I. and Karara H.M. (1971). Direct linear transformation into object space coordinates in close-range photogrammetry. In Symposium on Close-Range Photogrammetry. Adams L.P. (1980). The Use of Non Metric Cameras in Short Range Photogrammetry. 14th Congress of the International Society for Photogrammetry, Commission V, Hamburg, Germany. Agisoft, (2013). Agisoft PhotoScan User Manual: Professional Edition. Version 1.0.0 edn. Agüera-Vega F., Carvajal-Ramírez, F., Martínez-Carricondo, P. (2016).: Accuracy of digital surface models and orthophotos derived from unmanned aerial vehicle photogrammetry. J. Surveying Eng. 04016025. and Pattern Recognition, 1997. archaeological areas using a low-cost UAV the Augusta Bagiennorum Test site, In: XXI Arrell K, Fisher P.F, Tate N.J, Bastin L (2007) A fuzzy c-means classification of elevation derivatives to extract the morphometric classification of landforms in Snowdonia, Wales. Computers & Geosciences 33, pp.1366-1381. Atkinson K. B. (1996). Close Range Photogrammetry and Machine Vision. Whittles Publishing, Roseleigh House, Latheronwheel, Caithness, Scotland. Austin Reg. (2010) Unmanned aircraft systems UAVs design development and deployment, John Wiley & Sons Ltd, p.332 Autodesk. (2014). 123D Catch.www.123dapp.com/catch. Barka I., Vladovic J., Malis F. (2011) Landform classification and its application in predictive mapping of soil and forest units. Proceedings: GIS Ostrava 2011. Barnes G., Volkman W., Sherko R., Kelm K. (2014). Design and Testing of a UAVbased Cadastral Surveying and Mapping Methodology in Albania. World bank conf. Land and Poverty, The World Bank - Washington DC, March 24-27. Beasom S.L., Wiggers E.P., Giordono R. J. (1983). A technique for assessing land surface ruggedness. Journal of Wildlife Management. 47: pp. 1163–1166. Bellingham J.S., Tillerson M., Alighanbari M. and How J. P. (2003). Cooperative Path Planning for Multiple UAVs in Dynamic and Uncertain Environments, In: 42nd IEEE Conference on Decision and Control, Maui, Hawaii (USA). Bendea H. F., Chiabrando F., Tonolo F.G. and Marenchino D. (2007). Mapping of archaeological areas using a low-cost UAV the Augusta Bagiennorum Test site, In: XXI International Symposium, Athens, Greece. Boufama B., Mohr R., Veillon F. (1993) Euclidean constraints on uncalibrated reconstruction. Proceeding of the Fourth International Conference on Computer Vision, Berlin, Germany, 466-470. Brown D.C. (1976). The bundle adjustment—progress and prospects. Int. Archives Photogrammetry, 21(3), Paper number 3–03 (33 pages). Burrough P. A., McDonell R.A. (1998). Principles of Geographic Information Systems. Spatial Information and Geostatistics, Chapter 8: Spatial Analysis Using Continuous Fields, ISBN 0-19-823365-5. Burrough P.A. (1983). Multiscale sources of spatial variation in soil- the application of fractal concepts to nested levels of soil variation, J. Soil Sci., 34, 577. Chang K., Tsai B (1991) The effect of DEM resolution on slope and aspect mapping. Cartography and Geographic Information Science 18, pp. 69–77 scheduling, Whistler, British Columbia, Canada. 102 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. Collins M. G., Steiner, F. R., & Rushman, M. J. (2001). Land-Use Suitability Analysis in the United States: Historical Development and Promising Technological Achievements. Environmental Management, 611-621.decision-making methods. International Journal of Geographic Information. Cooper M.A. R. and Cross P.A. (1991). Statistical concepts and their application in photogrammetry and surveying. Photogrammetric Record, 13(77):645–678. Dahl A.L. (1973). Surface area in ecological analysis: quantification of benthic coralreef algae. Mar Biol. 23, pp.239–249. Dai F., & Lu M. (2010). Assessing the accuracy of applying photogrammetry to take geometric measurements on building products. Journal of Construction Engineering and Management, 136(2), 242–250. Reston, VA: ASCE. Dellaert F., Seitz S.M., Thorpe C.E. & Thrun, S. (2000) Structure from motion without correspondence. Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662). pp. 557-564. IEEE Comput. Soc, Hilton Head Island, SC, USA. Deng Y., Wilson J.P., Bauer B.O. (2007) DEM resolution dependencies of terrain attributes across a landscape. International Journal of Geographical Information Science 21, pp. 187–213. Dissanayake G., Sukarieh S., Nebto E. and Durrant Whyte H. (2001) “The Aiding of a Low-Cost Strap Down Inertial Measurement Unit Using Vehicle Model Constraints for Land Vehicle Applications”, IEEE transactions on robotics and automation, Vol. 17. Doumit J.A., Kiselev E.N. (2016). Structure from motion technology for macro scale objects cartography// Breakthrough scientific research as the modern engine of sciences, St. Petersburg.: Publishers “Cult Inform Press”. pp 42-47. ISBN 978-5-8392-0627-4. Doumit J.A., Pogorelov A.V. (2017). Multi-scale Analysis of Digital Surface Models Based on UAV Datasets. Modern Environmental Science and Engineering (ISSN 23332581), Volume 3, No. 7, pp. 460-468. Du Preez, C. (2014). "A new arc-chord ratio (ACR) rugosity index for quantifying three-dimensional landscape structural complexity." Landscape Ecology. 30, pp. 181– 192. Du Preez C., Tunnicliffe V. Shortspine thornyhead and rockfish (Scorpaenidae). (2011). distribution in response to substratum, biogenic structures and trawling. Mar. Ecol. Prog. Ser. 425, pp. 217-231. [doi:10.3354/meps09005]. dynamic detection of moving vehicles with a UAV, In: IEEE International Conference on Robotics and Automation, pp. 1878-1883. Eastman J.R., (1985). Single-Pass Measurement of the Fractional Dimensionality of Digitized Cartographic Lines. paper presented to the Canadian Cartographic Association, Annual Meeting, June 1985. Egbert J. and Beard, R. W. (2007). Road following control constraints for low altitude miniature air vehicles, In: American Control Conference, New York, U.S., 353-358. Eisenbeiss H. (2009). UAV photogrammetry. Dissertation ETH No. 18515, Institute of geodes and Photogrammetry, ETH Zurich, Switzerland, doi:10.3929/ethz-a-005939264. Eisenbeiss H. (2011). The Potential of Unmanned Aerial Vehicles for Mapping, Zurich. Photogrammetric Week '11 Dieter Fritsch (Ed.) Wichmann/VDE Verlag, Belin & Offenbach.pp135-144. Elliot J. K. (1989). An investigation of the change in surface roughness through time on the foreland of Austre Okstindbreen, North Norway,” Comput. Geosci., vol. 15, no. 2, pp. 209–217. 103 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. Evans I. (2003) Scale-specific landforms and aspects of the land surface. In: Evans, I.S., Dikau R., Tokunaga E, Ohmori H, Hirano M. (Eds.), Concepts and Modelling in Geomorphology: International Perspectives. Terrapub, Tokyo, pp. 61–84. Evans S. (1990) “General Geomorphometry”. In: Goudie, A.S., Anderson M., Burt T., Lewin J., Richards, Whalley K., Worsley B., Geomorphological Techniques. 2nd edition. Unwin Hyman, London, pp.44–56. Evans I.S. (1972) General geomorphometry, derivatives of altitude, and descriptive statistics. In: Chorley, R.J. (Ed.), Spatial Analysis in Geomorphology. Methuen, London, pp. 17–90. Falkner E., Morgan D. (2002) Aerial Mapping: Methods and Applications, (Mapping Science) 2nd Edition, by CRC Press LLC, pp.195. ISBN 1-56670-557-6. Faugeras O. D., Luong Q., and Maybank S. 1992 Camera self-calibration: Theory and experiments. In Proc. European Conference on Computer Vision, LNCS 588, pages 321–334. Florinsky I.V., Kuryakova G.A. (2000) Determination of grid size for digital terrain modelling in landscape investigations-exemplified by soil moisture distribution at a micro-scale. International Journal of Geographical Information Science 14, pp. 815–832. Friedman A., Pizarro O., Williams S.B. (2012). Johnson-Roberson M. Multi-scale measures of rugosity, slope and aspect from benthic stereo image reconstructions. PLoS One 7, 12, pp. 1–14. Furukawa Y. & Ponce, J. (2009) Accurate, Dense, and Robust Multi-View Stereopsis. IEEE Transactions on Pattern Analysis and Machine Intelligence. p. IEEE Computer Society Press. Galparsoro I., Borja A., Bald J., Liria P. (2009). Chust G. Predicting suitable habitat for the European lobster (Homarus gammarus), on the Basque continental shelf (Bay of Biscay), using Ecological-Niche Factor analysis. Ecol Model 220, pp. 556–567. Glover J.M. (2014). Drone University, USA, Middletown, DE. P.134. Gonzalez, J. P., Nagy, B. and Stentz, A., (2006). The Geometric Path Planner for Navigating Unmanned Vehicles in Dynamic Environments, In: Proceedings ANS 1st Joint Emergency Preparedness and Response and Robotic and Remote Systems, Salt Lake City (Utah), USA. Gonzalez M.F. and Wintz, P. (1987). Digital Image Processing, Addison Wesley, Menlo Park, CA, p.414. Goodchild M.F. and Mark D.M. (1987). The fractal nature of geographic phenomena. Ann. Assoc. Am. Geogr., 77, p.265. Granshaw S. (1980). Bundle adjustment methods in engineering photogrammetry. Photogrammetric Record, 10(56), pp.181–207. Grohmann C., Smith M., Riccomini C. (2011). Multiscale analysis of topographic surface roughness in the Midland Valley, Scotland, IEEE T. Geosci. Remote, 49, pp. 1200–1213. Gruner H. (1977). “Photogrammetry: 1776-1976”, Photogrammetric Engineering and Remote Sensing, 43(5), pp.569-574. Haarbrink, R. B. and Koers, E. (2006). Helicopter UAV for photogrammetry and rapid response, In: International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, ISPRS Workshop of Inter-Commission WG I/V, Autonomous Navigation, Antwerp, Belgium. Hartley R.I., Zisserman, A. (2003): Multiple View Geometry in Computer Vision. Cambridge University Press 104 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. Hengl T., Evans, I.S. (2009). Mathematical and digital models of the land surface. In: Hengl T., Reuter, H.I. (Eds.), Geomorphometry — Concepts, Software, Applications. Developments in Soil Science, vol. 33. Elsevier, Amsterdam, pp. 31–63. Hengl T. (2006) Finding the right pixel size. Computers & Geosciences 32, pp. 1283– 1298. Herzfeld U.C., Mayer H., Feller W., Mimler M. (2000) Geostatistical analysis of glacier-roughness data, Ann. Glaciol., vol. 30, no. 1, pp. 235–242. Heyden A. and Astrom K. (1997). Euclidean reconstruction from image sequences with varying and unknown focal length and principal point. In Proc. IEEE Conference on Computer Vision and Pattern Recognition. Hobson R. D. (1972). Surface roughness in topography: quantitative approach. in R. J. Chorley, editor. Spatial analysis in geomorphology. Harper and Row, New York, New York, USA, pp. 221–245. Hongxia C., Zongjian L., Guozhong S. (2007). Non-metric CCD Camera Calibration for Low Attitude Photogrammetric Mapping. The Eighth International Conference on Electronic Measurement and Instruments ICEMI’. Hopkins, L. (1977). Methods for Generating Land Suitability Maps: A Comparative Evaluation. Journal of the American Institute of Planners, 380-400. ICAO, (2011). Unmanned Aircraft Systems (UAS). Published in separate Arabic, Chinese, English, French, Russian and Spanish editions by the International civil aviation organization. University Street, Montréal, Quebec, Canada H3C 5H7. P. 38. ISBN 97892-9231-751-5 Irschara a., Kaufmann V., Klopschitz M., Bischof H., Leberl F. (2010). Towards fully automatic photogrammetric reconstruction using digital images taken from UAVs, ISPRS TC VII Symposium – 100 Years ISPRS, Vienna, Austria, IAPRS, Vol. XXXVIII, Part 7A pp.65-70. Issod, C.S. (2015). Getting Started with Hobby Quadcopters and Drones, USA, Middletown, DE. P.97. Iwahashi J & Pike R. J (2007) “Automated classifications of topography from DEMs by an unsupervised nested-means algorithm and a three-zone geometric signature,” Geomorphology, vol. 86, no. 3/4, pp. 409–440. Jankowski, P. (1995). Integrating geographical information systems and multiple criteria Decision-Making Methods. Geographical Information Systems 9(3) :251-273. DOI:10.1080/02693799508902036 Jenness J. (2005) Topographic Position Index. Extension for ArcView 3.x. http://jennessent.com. Jenness J. (2010) “Topographic Position Index (tpi_jen.avx) extension for ArcView 3.x”, v. 1.3a. Jenness Enterprises, http://www.jennessent.com/arcview/tpi.htm. Jenness J. (2004). "Calculating landscape surface area from digital elevation models." Wildlife Society Bulletin 32.3, pp. 829-839. Jenson S.K., Domingue, J.O., 1988. Extracting topographic structure from digital elevation model data for geographic information system analysis. Photogrammetric Engineering and Remote Sensing. 54, pp. 1593–1600. Kaaniche K., Champion, B., Pegard, C. and Vasseur, P. (2005). A vision algorithm for dynamic detection of moving vehicles with a UAV, In: IEEE International Conference on Robotics and Automation, pp. 1878-1883. Karara H.M. (1989). Non-Topographic Photogrammetry. Americal Society for Photogrammetry and Remote Sensing. 105 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. Karara H.M., and W. Faig. (1980). An Expose on Photographic Data Aquisition Systems in Close-Range Photogrammetry. 14th Congress of the International Society for Photogrammetry, Commission V, Hamburg, Germany. Koeln G.T., Cowardin L.M., Strong L.L. (1996). Geographical information systems. in T.A. Bookhout, ed. Research and management techniques for wildlife and habitats. Fifth ed., rev. The Wildlife Society, Bethesda MD., pp. 540-566. Lafay M. (2015). Drones for Dummies, John Wiley and sons, Inc. Hobokon, New Jersy. P.272. Lam, N. and Quattrochi, D.A. (1992). On the issues of scale, resolution, and fractal analysis in the mapping sciences, Prof. Geogr., pp.44-88. Leibowitz D., College M. (2001). Camera Calibration and Reconstruction of Geometry from Images. Robotics Research Group Department of Engineering Science University of Oxford Trinity Term, p. 209. Levin S.A. (1992). The problem of pattern and scale in ecology. Ecology. 73, pp. 1943– 1967. Li Z., (1993). Mathematical models of the accuracy of digital terrain model surfaces linearly constructed from gridded data, Photogrammetric Record, (82), pp. 661-674. Lowe, D. (2004). Distinctive image features from scale-invariant key points. International Journal of Computer Vision 60, pp. 91–110. Lucieer A., Watson C. and Turner D. (2012) “Development of a UAV-LiDAR System with Application to Forest Inventory Luke Wallace”. School of Geography and Environmental Studies, University of Tasmania. http://www.mdpi.com/2072-4292/4/6/1519. Luhmann, T, Robson, S, Kyle, S, and Harley, I. (2006). Close Range Photogrammetry: Principles, Methods, and Applications. Caithness: Whittles Publishing. Lundblad ER., Wright DJ., Miller J., Larkin EM., Rinehart R., Naar DF., Donahue BT., Anderson SM., Battista T A. (2006). benthic terrain classification scheme for American Samoa. Mar Geod., 29, pp. 89–111. MacMillan R.A, Shary P.A. (2009). Landforms and landform elements in geomorphometry. In: Hengl, T., Reuter, H.I. (Eds.), Geomorphometry—Concepts, Software, Applications. Developments in Soil Science, vol. 33. Elsevier, Amsterdam, pp. 227– 254. Mark D.M. (1975). Geomorphometric parameters: A review and evaluation, Geografiska Annaler. Ser. A, Phys. Geography, vol. 57, no. 3/4, pp. 165– 177. Mark, D.M. and Aronson, P.B. (1984). Scale-dependent fractal dimensions of topographic surfaces: an empirical investigation, with applications in geomorphology and computer mapping, Math. Geol., 11,671. Marshall A. R. (1989). Network design and optimization in close ranges photogrammetry.UNISURV S-36, School of Surveying, University of new South Wales, Australia., p. 249. Masahiko N. (2007). UAV borne mapping system for river environment, In: 28th Asian Association of Remote Sensing Conference, Kuala Lumpur, Malaysia Matthews N. A. (2008). Aerial and Close-Range Photogrammetric Technology: Providing Resource Documentation, Interpretation, and Preservation. Technical Note 428. U.S. Department of the Interior, Bureau of Land Management, National Operations Center, Denver, Colorado. 42 pp. McCormick M.I. (1994). Comparison of field methods for measuring surface topography and their associations with a tropical reef fish assemblage. Marine Ecology Progress Series 112, pp. 87-96. 106 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. McGlone J.C. (1989). Analytic data-reduction schemes in non-topographic photogrammetry, Non-Topographic Photogrammetry, 2nd ed., ed. H.M. Karara, ASPRS, Falls Church, VA, Chap. 4, 37-55. McHarg I. L. (1995). Design with Nature. New York: J. Wiley. (Reprinted from Design) Meyer R. (1987). “100 Years of Architectural Photogrammetry”, Kompendium Photogrametrie, Vol. XIX, Leipzig: Akademische Verlagsgesellschaft, pp. 183-200. Micheletti N, Chandler JH, Lane SN. (2014). Investigating the geomorphological potential of freely available and accessible structure-from-motion photogrammetry using a smartphone. Earth Surface Processes and Landforms, DOI: 10.1002/esp.3648. Mikhail E., Bethel J. and McGlone J. (2001). Introduction to modern photogrammetry. John Wiley & Sons, New York, p.479. Moser K., Ahn, C., Noe, G. (2007) Characterization of micro topography and its influence on vegetation patterns in created wetlands. Wetlands 27, pp. 1081-1097. Nagai M., Shibasaki R., Manandhar D. and Zhao H. (2004): Development of digital surface and feature extraction by integrating laser scanner and CCD sensor with IMU, In: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XX ISPRS Congress, Istanbul, Turkey, XXXV-B5, pp.655-659. Nellis M.D. and Briggs, J.M. (1989). the effect of spatial scale on konza landscape classification using textural analysis, landscape ecol.,2,93. Newsome, Sam R. Jr. (2016). "Discrepancy Analysis Between Close-Range Photogrammetry and Terrestrial LiDAR" Electronic Theses & Dissertations. Paper 1423. Nordberg K., Farnebäck, G., Forssén, P.-E., Granlund, G., Moe, A., Wiklund, J. and Doherty, P. (2002). Vision for a UAV helicopter, In: Workshop on aerial robotics, Lausanne, Switzerland. Partsh J. (1911). Schlesien, eine Landeskunde fur das deutshe Volk.II, S.586, Breslau. Patias, P., Saatsoglou-Paliadeli , C., Georgoula, O., Pateraki, M., Stamnas, A. and Kyriakou, N.(2007). Photogrammetric documentation and digital representation of the macedonian palace in Vergina-Aegeae, In: CIPA, XXI International CIPA Symposium, Athens, Greece. Pettersson, P.O. and Doherty, P. (2004). Probabilistic roadmap based path planning for autonomous unmanned aerial vehicles, In: 14th Int’l Conf. on Automated Planning. Peucker T.K., Douglas D.H., 1974. Detection of surface specific points by local parallel processing of discrete terrain elevation data. Computer Graphics and Image Processing, 4, pp. 375-387. Pike R. J. (2002). A bibliography of terrain modeling (geomorphometry), the quantitative representation of topography - supplement 4.0., Open-File Rep. No. 02-465. U.S. Geological Survey, Denver, p. 116. Pogorelov A.V & Doumit J.A. (2009). Relief of Kuban river basin: Morphometric analysis. M.: Geoc, 208 p. (In Russian). Puri A. (2004). A Survey of Unmanned Aerial Vehicles (UAV) for Traffic Surveillance, Internal Report, Department of Computer Science and Engineering, University of South Florida, Tampa, FL, USA, p. 29. Rasemann S., Schmidt J., Schrott L., Dikau R. (2004). “Geomorphometry in mountain terrain”. In: Bishop, M.P., Shroder, J.F. eds. GIS & Mountain Geomorphology. Springer, Berlin, pp.101-145. Redweik P. (2012). Photogrammetry Springer. Retrieved from http://link.springer.com/chapter/10.1007/978-3- 642-28000-9_4/fulltext.html. Reg Austin (2010). Unmanned aircraft systems UAVs design development and deployment, John Wiley & Sons Ltd, p.332 107 111. Riga Declaration on remotely piloted aircraft (drones) " Farming the future of aviation" Riga - 6 March 2015. 112. Riley S.J., De Gloria S. D., Elliot R. A. (1999). Terrain ruggedness index that quantifies topographic heterogeneity. Intermountain journal of sciences, Vol.5, No. 1-4. 113. Risk, M.J. (1972). Fish diversity on a coral reef in the Virgin Islands. Atoll Research Bulletin 193, pp. 1-6. 114. Roelofs R. (1951). Distortion, principal point, point of symmetry and calibrated principal point, Photogrammetria.vol.7, pp. 49-66. 115. Rüther H., Smit J., Kamamba D. (2012). A Comparison of Close-Range Photogrammetry to Terrestrial Laser Scanning for Heritage Documentation// South African Journal of Geomatics, Vol. 1, No. 2, pp. 149-169. 116. Shary P.A. (1995). Land surface in gravity points classification by a complete system of curvatures. Mathematical Geology, 27(3), pp.373-390. 117. Shary P.A., Sharaya L.S., Mitusov A.V. (2002). Fundamental quantitative methods of land surface analysis. Geoderma, 107(1-2), pp. 1-32. 118. SMITH, M. J. and PARK, D. W. G. (2000). Absolute and exterior orientation using linear features. International Archives of Photogrammetry and Remote Sensing, 33(B3): 850857. 119. Snavely N., Seitz, S.M. & Szeliski, R. (2006). Photo Tourism: Exploring image collections in 3D. ACM Transactions on Graphics, 25, p.835. 120. Snavely, N., Seitz, S.M. & Szeliski, R. (2007). Modeling the World from Internet Photo Collections. International Journal of Computer Vision, 80, pp. 189-210. 121. Snavely, N., Seitz, S.N., Szeliski, R. (2008). Modeling the world from internet photo collections. International Journal of Computer Vision 80, pp.189-210. 122. Spetsakis M.E., Aloimonos Y. (1991). A multi-frame approach to visual motion perception. International Journal of Computer Vision 6.pp. 245-255. 123. Stambaugh M.C., Guyette R.P. (2008). Predicting spatio-temporal variability in fire return intervals using a topographic roughness index. For Ecol Manag, 254(3), pp. 463– 473. 124. Stöcker, C., Bennett, R., Nex, F., Gerke, M., Schmidt, C., Zein, T. (2017). Guide to regulatory practice on UAVs for land tenure recording. H2020 its4land 687828 D4.1 innovation for Land Tenure.p.48. 125. Stone R. O., Dugundji J. (1965). A study of micro relief-Its mapping, classification and quantification by means of a Fourier analysis, Eng. Geol., vol. 1, no. 2, pp. 89-187. 126. Sugiura R., Noguchi, N. and Ishii, K. (2005). Remote-sensing Technology for Vegetation Monitoring using an Unmanned Helicopter, In: Biosystems Engineering, 90, 4, 369–379. 127. Szeliski R., Kang S.B. (1994). Recovering 3-D shape and motion from image streams using nonlinear least squares. Journal of Visual Communication and Image Representation 5. pp.10-28. 128. Tagil S. & Jenness J (2008) GIS- based automated landform classification and topographic, Landcover, and geologic attributes of landforms around the Yazoren Polje, Turkey. Journal of Applied Sciences 8(6): 910-921. 129. Tsai Y. R. (1986). An efficient and accurate camera calibration technique for 3D machine vision. In unknown focal length and principal point. In Proc. IEEE Conference on Computer Vision. 130. Valentine P.C., Scully L.A., Fuller S.J. (2004). Terrain Ruggedness Analysis and Distribution of Boulder Ridges and Bedrock Outcrops in the Stellwagen Bank National Marine Sanctuary Region —Posters presented at the Fifth International Symposium of the Geological and Biological Habitat Mapping Group, Galway, Ireland. 108 131. Veitinger J., Purves R. S., Sovilla B. (2016). Potential slab avalanche release area identification from estimated winter terrain: a multi-scale, fuzzy logic approach/Natural Hazards and Earth System Sciences, 16, pp. 2211–2225. 132. Verhoeven, G, Sevara, C, Karel, W, Ressl, C, Doneus, M and Briese, C. (2013) ‘Undistorting the past: new techniques for orthorectification of archaeological aerial frame imagery’, in Corsi, C et al (eds) Good Practice in Archaeological Diagnostics, Natural Science in Archaeology. Switzerland: Springer International Publishing. 133. Von Blyenburg P (1999) UAVs-Current Situation and Considerations for the Way Forward. RTO-AVT Course on Development and Operation of UAVs for Military and Civil Applications. 134. Wedding L.M., Friedlander A.M., McGranaghan M., Yost R.S., Monaco M. E. (2008). Using bathymetric LIDAR to define nearshore benthic habitat complexity: implications for management of reef fish assemblages in Hawaii. Remote Sens Environ, 112(11), pp. 4159–4165. 135. Westoby M., Brasington J., Glasser N.F., Hambrey M.J., Reyonds M.J. (2012). Structure from Motion photogrammetry: a low-cost, effective tool for geoscience applications. Geomorphology 179, pp. 300-314. 136. Wolf P. R. and Ghilani C.D. (1997). Adjustment Computations: Statistics and Least Squares in Surveying and GIS. John Wiley & Sons. 137. Wood J. (1996). The geomorphological characterization of digital elevation models, PhD Thesis. University of Leicester. 138. Wood J. (2009). Geomorphometry in LandSerf. In: Hengl, T., Reuter, H.I. (Eds.), Geomorphometry — Concepts, Software, Applications. Developments in Soil Science, vol. 33. Elsevier, Amsterdam, pp. 333–349. 139. Woodby D., Carlile D., Hulbert L. (2009). Predictive modeling of coral distribution in the Central Aleutian Islands, USA. Mar Ecol Prog Ser. 397, pp. 227–240. 140. Woodcock C. E., Strahler. A H. (1987). “The Factor of Scale in Remote Sensing,” Remote Sensing of Environment, vol.21, pp. 311-332. 141. Wright D.J., Heyman W.D. (2008). Introduction to the special issue: marine and coastal GIS for geomorphology, habitat mapping, and marine reserves. Mar Geod., 31, pp. 223– 230. 142. Wu J. (2004). Effects of changing scale on landscape pattern analysis: scaling relations Landscape Ecol. 19, pp. 125–138. 143. Wzorek M., Land´en D. and Doherty P. (2006). GSM Technology as a Communication Media for an Autonomous Unmanned Aerial Vehicle, In: 21th UAV Systems Conference, Bristol. 144. Xie Feifei et al. (2014). 1:500 Scale Aerial Triangulation Test with Unmanned Airship in Hubei Province IOP Conf. Ser.: Earth Environ. Sci. 17 012177. 145. Zevenbergen L., Thorne C. (1987). Quantitative Analysis of Land Surface Topography, Earth Surface Processes and Landforms, Vol. 12, pp. 47-56. 146. Zhou G., Li C. and Cheng P. (2005). Unmanned aerial vehicle (UAV) realtime video registration for forest fire monitoring, In: IEEE International, Geoscience and Remote Sensing Symposium, pp. 1803-1806. 109 Scientific edition FROM DRONES TO GEOSPATIAL DATA Печатается в авторской редакции Подписано в печать 26.12.2018. Формат 60×84 1/16. Печать цифровая. Уч-изд. л.6,7. Тираж 60 экз. Заказ № 3609. Издательско-полиграфический центр КубГУ. 350040, г. Краснодар, ул. Ставропольская, 149.