A Review on Traffic Monitoring System

Techniques

Neeraj Kumar Jain, R. K. Saini and Preeti Mittal

Abstract Increase in traffic density along with the population growth in the world
has resulted in more and more congested roads, air pollution, and accidents. Growth
of total number of vehicles around the world has increased exponentially during past
decade. Traffic monitoring in this scenario is certainly a big challenge in many cities
of India. In most of the cities, we are still dependent on human and accomplishing
this herculean task manually. In controlling enormous volume of traffic with manual
methods, more and more issues like availability, efficiency, and accuracy of manage-
ment staff are always encountered. In this study, various traffic monitoring schemes
striving to cop up and handle enormous traffic flow with optimized human interven-
tion are studied. Applications of these schemes include identification of vehicles in
traffic, sense traffic congestion on a road, measuring speed of vehicle, traffic density
on intersections, the presence of VIP vehicles or ambulances, accidents on roads,
path for pedestrians, and many more. Usually, intrusive and non-intrusive in situ
techniques and in-vehicle technologies are used for traffic monitoring but image
processing techniques win over these traditional techniques. The aim of this paper
is to discuss the previous research work done to monitor the traffic using image and
video processing techniques and future perspective.

Keywords Traffic monitoring · Traffic density · In situ techniques

In-vehicle technologies · Image processing

N. K. Jain (B) · R. K. Saini · P. Mittal

Department of Mathematical Sciences and Computer Applications,
Bundelkhand University, Jhansi, India
e-mail: neerajjain15@gmail.com
R. K. Saini
e-mail: rksaini.bu@gmail.com
P. Mittal
e-mail: preetimittal1980@yahoo.co.in

© Springer Nature Singapore Pte Ltd. 2019 569

K. Ray et al. (eds.), Soft Computing: Theories and Applications,
Advances in Intelligent Systems and Computing 742,
https://doi.org/10.1007/978-981-13-0589-4_53
570 N. K. Jain et al.

Fig. 1 Year-wise vehicle

registered in Delhi

1 Introduction

An exponential growth of running vehicles on roads along with the population growth
in the world has resulted in more and more congested roads, air pollution and acci-
dents. Growth of total number of vehicles around the world has increased exponen-
tially during past decade.
In Indian metro cities, traffic density on road is increasing four times faster than the
population. Figure 1 shows the year-wise vehicles registration status in our national
capital Delhi which depicts the enormity of vehicles and challenges in front of Delhi
traffic police [1].
There are basically two types of traffic congestions—Recurring traffic congestion
which appears at the same place during the same time regularly and nonrecurring
traffic congestion which occurs randomly like an unplanned event [2]. This nonre-
curring effect can cause a sudden traffic volume increase. Detection of nonrecurring
traffic congestion is critical compared to the recurring type, because it requires real-
time traffic information and evaluation thereof with appropriate traffic management
decisions. Due to traffic congestion, there is obvious wastage of many resources.
Traffic congestion may take place due to the following reasons [3]:
1. Too many cars for the roadway due to inadequate mass transit options or other
reasons.
2. Obstacles on the road cause a blockage and merger. These can be double parking,
road work, lane closure due to utility work, road narrowing down, and an accident.
3. Traffic signals out of sync many times on purpose or occasionally when the
computers are malfunctioning.
4. Inadequate green/red time.
5. Too many pedestrians crossing not permitting cars to turn.
6. Too many trucks on the road due to inadequate rail freight opportunities.
A Review on Traffic Monitoring System Techniques 571

7. Overdevelopment in areas where the mass transit system is already overcrowded

and the road system is inadequate.
8. Delay in removal of accidental vehicles on roads.
9. Poor weather conditions.
The rest of the paper is organized as follows: Sect. 2 describes traffic monitor-
ing systems. Section 3 discusses literature review in detail. Finally, Sect. 4 is for
conclusion and future scope.

2 Traffic Monitoring System

A traffic monitoring system must tackle issues like traffic congestion, accident detec-
tion, vehicle identification/detection, automatic vehicle guidance, smart signaling,
forensics, traffic density, safe pedestrian movement, emergency vehicles transit, etc.
An ideal traffic monitoring system can be designed using either of the following:
1. In situ traffic detector technologies.
2. Vehicular sensor networks (VSN) or probe vehicles (PVs), such as taxis and
buses, and floating cars (FCs), such as patrol cars for surveillance.
3. Image or video processing.

2.1 In Situ Technologies

In situ traffic detector technologies are further divided into two categories—Intrusive
technology and non-intrusive technology. In intrusive technologies, detectors are
physically mounted at or below the road surface, installation of which causes potential
disruption to traffic. These include embed magnetometers, pneumatic tube detectors,
inductive detector loops, and Weigh-in-Motion (WIM) systems.
In nonintrusive technologies, detectors are mounted at or above the road surface,
and their installation causes little or no disruption to traffic. Nonintrusive technologies
include manual methods, video data collection, passive or active infrared detectors,
microwave radar detectors, ultrasonic detectors, passive acoustic detectors, laser
detectors, and aerial photography (Fig. 2).

2.2 Vehicular Sensor Networks

Vehicular sensor networks (VSNs) have been proved as a great solution for traffic
monitoring. Sensing devices attached to the moving vehicles move all over the city
to collect the traffic information. These moving sensing devices are attached to each
other and to the traffic monitoring center as well. The collected information is passed
572 N. K. Jain et al.

Fig. 2 Types of detectors

to city traffic monitoring center on vehicle-to-vehicle or vehicle-to-infra wireless

communications basis. Traffic monitoring center takes appropriate decisions in order
to ensure hassle-free traffic movement.
In-vehicle devices are generically termed as automatic vehicle location (AVL)
systems. AVL devices either provide positional information whenever a suitably
equipped vehicle passes a certain point in the network, or continuous information as
the vehicle travels through a network. The former system typically relies on appro-
priate vehicles being equipped with transponders which transmit and receive infor-
mation from roadside units. The latter system uses vehicles equipped with global
positioning system (GPS) technology.

2.3 Image and Video Processing

Video monitoring and image processing have been widely used in traffic manage-
ment for solving number of traffic issues. The traffic density estimation and vehicle
classification can also be achieved using video monitoring systems. Live video feed
from the cameras at traffic junctions for real-time traffic density calculation using
video and image processing. It can be used efficiently in switching the traffic lights
according to vehicle density on road, resulting in reducing the traffic congestion on
roads, and hence lower the number of accidents. It ensures safe transit to people and
reduce fuel consumption and waiting time. It will also provide significant data which
will help in future road planning and analysis. In further stages, multiple traffic lights
can be synchronized with each other with an aim of even less traffic congestion and
free flow of traffic [4].
Some of the most demanding and widely studied applications of image or
video processing are travelers’ information, ramp metering, traffic monitoring,
A Review on Traffic Monitoring System Techniques 573

automatic vehicle guidance, accident detection, smart signaling, vehicle identifi-

cation/detection, forensics, traffic density, safe pedestrian movement, etc.
Practically, it has been observed although intrusive detectors or sensors have
become important in traffic applications mainly due to their rapid response, cheaper
installation, operation, and maintenance, and their ability to control and monitor
wide areas [5], but they have more disadvantages too like they require to dig up
the road and therefore costlier to install. It also gives limited amount of information
than other methods of traffic monitoring system. Radar gun is another way of getting
information about the speed of vehicle but it gives less traffic information. Similarly,
the pressure tubes are reliable enough but give limited information about traffic [6].
Advantages and disadvantages of these traffic monitoring techniques are tabulated
as follows [7].

Technology Strengths Weaknesses

Inductive loop • Flexible design to satisfy large variety of • Installation requires pavement cut
applications • Decreases pavement life
• Mature, well-understood technology • Installation and maintenance require lane
• Large experience base closure
Provides basic traffic parameters (e g., volume, • Wire loops subject to stresses of traffic and
presence, occupancy speed, headway, and gap) temperature
• Insensitive to inclement weather such as rain, • Multiple detectors usually required of monitor
fog, and snow a location
• Provides best accuracy for count data as • Detection accuracy may decrease when design
compared with other commonly used requires detection of a large variety of vehicle
techniques classes
• Common standard for obtaining accurate
occupancy measurements
• High-frequency excitation models provide
classification data

Magnetometer • Less, susceptible than loops to stresses of • Installation requires pavement cut
(Two-axis traffic • Decreases pavement life
fluxgate • Insensitive to inclement weather such as snow, • Installation and maintenance require lane
magnetometer) rain, and fog closure
• Some models transmit data over wireless RF • Models with small detection zones require
link multiple units for full lane detection
• Installation requires pavement cut or boring
under roadway
• Cannot detect stopped vehicles unless special
sensor layouts and signal processing software
are used

Magnetic • Can be used where loops are not feasible (e.g., • Installation requires pavement cut or boring
(induction or bridge decks) under roadway
search coil • Some models are installed under roadway • Cannot detect stopped vehicles unless special
magnetometer) without the need for pavement cuts. However, sensor layouts and signal processing softwares
boring under roadway is required are used

• Insensitive to inclement weather such as snow,

rain, and fog
• Less susceptible than loops to stresses to traffic
574 N. K. Jain et al.

Technology Strengths Weaknesses

Microwave • Typically insensitive to inclement weather at • CW Doppler sensors cannot detect stopped
radar the relatively short ranges encountered in vehicles
traffic management applications
• Direct measurement of speed
• Multiple lane operations available

Active infrared • Transmits multiple beams for accurate • Operations may be affected by fog when
(Laser radar) measurement of vehicle position, speed, and visibility is less than ≈ 20 ft (6 m) or blowing
class snow is present
• Multiple lane operations available • Installation and maintenance, including
periodic lens cleaning, require lane closure

Passive • Multizone passive sensor’s measure speed • Passive sensor may have reduced sensitivity to
infrared vehicles in heavy rain and snow and dense fog
• Some models not recommended for presence
detection

Ultrasonic • Multiple lane operations available • Environmental conditions such as temperature

• Capable of overheight vehicle detection change and extreme air turbulence can affect
• Large Japanese expedience base performance. Temperature compensation is
built into some models
• Large pulse repetition periods may degrade
occupancy measurement on freeways with
vehicles traveling at moderate to high speeds

Acoustic • Passive detection • Cold temperatures may affect vehicle count

• Insensitive to precipitation accuracy
• Multiple lane operations available in some • Specific models are not recommended with
models slow-moving vehicles in stop-and-go traffic

Video image • Monitors multiple lanes and multiple • Installation and maintenance, including
processor detection zones/lanes periodic lens cleaning, require lane closure
• Easy to add and modify detection zones when camera is mounted over roadway (lane
• Rich array of data available closure may not be required when camera is
• Provides wide-area detection when mounted at side of roadway)
information gathered at one camera location • Performance affected by inclement weather
can be linked to another such as fog, ram. and snow: vehicle shadows:
vehicle projection into adjacent lanes:
occlusion: day-to-night transition: vehicle
road/contrast: and water, salt grime, icicles,
and cobwebs on camera lens
• Requires 30- to 50-ft (9- to 15-in) camera
mounting height (in a side-mounting
configuration) for optimum presence detection
and speed measurement
• Some models susceptible to camera motion
caused by strong winds or vibration of camera
mounting structure
• Generally cost-effective when many detection
zones within the field-of-view of the camera
or specialized data are required
• Reliable nighttime signal actuation requires
street lighting
A Review on Traffic Monitoring System Techniques 575

3 Literature Review

Kamijo et al. (2000) used spatiotemporal Markov random field (MRF), for traffic
images at intersections. It models a tracking problem by determining the state of
each pixel in an image and its transit, and how such states transit along both the
image axes as well as the time axes. It segments and tracks occluded vehicles at a
high success rate [8].
Messelodi et al. (2005) proposed a real-time vision system by analyzing monoc-
ular image sequences from pole-mounted video cameras. Their experimental results
demonstrated robust, real-time vehicle detection, tracking, and classification over
several hours of videos taken under different illumination conditions [9].
Lee and Baik (2006) introduced a video-based vehicle tracking system to provide
information on directional traffic counts at intersections. The extracted counts were
fed to estimate an origin–destination trip table which was necessary information for
traffic impact study and transportation planning [10].
Aycard et al. (2011) proposed an approach for intersection safety developed in the
scope of the European project INTERSAFE-2. A complete solution for the safety
problem including the tasks of perception and risk assessment using on-board Lidar
and stereovision sensors presented and better results were shown [11].
Wang et al. (2015) introduced an approach for real-time multi-vehicle tracking and
counting using fisheye camera based on simple feature points tracking, grouping, and
association. Motion similarity and neighbor-weighted grafting were used to transfer
motion knowledge between long and short point trajectories [12].
Jodoin et al. (2016) proposed a tracking system to track the various road users of
diverse shapes and appearances. Finite state machine handled fragmentation, split-
ting, and merging of the road users to correct and improve the resulting object trajecto-
ries. This tracker was tested on several urban intersection videos and it outperformed
[5].
Liu et al. (2016) used a method that accurately detects vehicles using a proba-
bilistic classification method followed by a refinement based on object segments.
Both classification and segmentation methods made use of coregistered aerial RGB
images and airborne LiDAR data [13].
Huang et al. (2017) introduced vehicle detection in the tunnel which is a chal-
lenging problem due to the usage of heterogeneous cameras, varied camera setup
locations, low-resolution videos, poor tunnel illumination, and reflected lights on the
tunnel wall. The proposed method was based on background subtraction and Deep
Belief Network (DBN) with three-hidden layer architecture [6].
Tang et al. (2017) proposed a technique for vehicle detection and type recognition
based on static images. This technique was highly practical and directly applicable
to various operations in a traffic surveillance system. First, Haar-like features and
AdaBoost algorithms were applied for feature extracting and constructing classifiers,
which were used to locate the vehicle over the input image. Then, the Gabor wavelet
transform and a local binary pattern operator was used to extract multiscale and
multi-orientation vehicle features [14].
576 N. K. Jain et al.

Ukani et al. (2017) used video surveillance systems for real-time vehicle detection
and classification. Background subtraction and from each detected vehicles Scale-
Invariant Feature Transform (SIFT) features were extracted. Vehicles were classified
using the neural network and Support Vector Machine (SVM). SVM showed better
generalization than artificial neural networks [15].

4 Conclusion and Future Scope

In this review paper, various existing traffic monitoring systems have been studied.
Presently, these systems do provide a cost-effective solution, but the rate of success-
ful operation is not good. Inductive loop detectors installed below the road fail in
the case of poor road condition. IR detectors along the side of the lane for density
calculation operate less efficiently where they are not applicable in the real-time
process. On the other hand, several other image processing based methods are doing
better and helping the present traffic control system to be more efficient. Computer
vision being one of the most researched fields is for the future technologies. There is
more and more scope of video and image processing techniques in traffic monitoring
and analysis. Intelligent solutions to traffic issues like monitoring, automatic vehi-
cle guidance, accident detection, smart signaling, vehicle identification/detection,
forensics, traffic density, and vehicle theft can be found and used effectively using
enormous capabilities of image processing and other techniques.

References

1. https://delhitrafficpolice.nic.in/about-us/statistics/
2. Nellore, K., Hancke, G.P.: A survey on urban traffic management system using wireless sensor
networks. Sensors 16(2), 157 (2016)
3. http://bklyner.com/what-really-causes-traffic-congestion-sheepshead-bay/
4. Kanungo, A., Sharma, A., Singla, C.: Smart traffic lights switching and traffic density calcu-
lation using video processing. In: 2014 Recent Advances in Engineering and Computational
Sciences (RAECS), pp. 1–6. IEEE (2014)
5. Jodoin, J.P., Bilodeau, G.A., Saunier, N.: Tracking all road users at multimodal urban traffic
intersections. IEEE Trans. Intell. Transp. Syst. 17(11), 3241–3251 (2016)
6. Huang, B.J., Hsieh, J.W., Tsai, C.M.: Vehicle detection in Hsuehshan tunnel using background
subtraction and deep belief network. In: Asian Conference on Intelligent Information and
Database Systems, pp. 217–226. Springer (2017)
7. Leduc, G.: Road traffic data: collection methods and applications. Working papers on energy,
transport and climate change, vol. 1, p. 55 (2008)
8. Kamijo, S., Matsushita, Y., Ikeuchi, K., Sakauchi, M.: Traffic monitoring and accident detection
at intersections. IEEE Trans. Intell. Transp. Syst. 1(2), 108–118 (2000)
9. Messelodi, S., Modena, C.M., Zanin, M.: A computer vision system for the detection and
classification of vehicles at urban road intersections. Pattern Anal. Appl. 8(1–2), 17–31 (2005)
10. Lee, S.M., Baik, H.: Origin-destination (od) trip table estimation using traffic movement counts
from vehicle tracking system at intersection. In: IECON 2006-32nd Annual Conference on
IEEE Industrial Electronics, pp. 3332–3337. IEEE (2006)
A Review on Traffic Monitoring System Techniques 577

11. Aycard, O., Baig, Q., Bota, S., Nashashibi, F., Nedevschi, S., Pantilie, C., Parent, M., Resende,
P., Vu, T.D.: Intersection safety using lidar and stereo vision sensors. In: 2011 IEEE Intelligent
Vehicles Symposium (IV), pp. 863–869. IEEE (2011)
12. Wang, W., Gee, T., Price, J., Qi, H.: Real time multi-vehicle tracking and counting at intersec-
tions from a fisheye camera. In: 2015 IEEE Winter Conference on Applications of Computer
Vision (WACV), pp. 17–24. IEEE (2015)
13. Liu, Y., Monteiro, S.T., Saber, E.: Vehicle detection from aerial color imagery and airborne
lidar data. In: 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS),
pp. 1384–1387. IEEE (2016)
14. Tang, Y., Zhang, C., Gu, R., Li, P., Yang, B.: Vehicle detection and recognition for intelligent
traffic surveillance system. Multimedia Tools Appl. 76(4), 5817–5832 (2017)
15. Ukani, V., Garg, S., Patel, C., Tank, H.: Efficient vehicle detection and classification for traf-
fic surveillance system. In: International Conference on Advances in Computing and Data
Sciences, pp. 495–503. Springer (2016)