Available online at www.sciencedirect.com Procedia Computer Science 00 (2019) 000–000 ScienceDirect www.elsevier.com/locate/procedia Available online at www.sciencedirect.com Procedia Computer Science 00 (2019) 000–000 ScienceDirect www.elsevier.com/locate/procedia Third International Conference on Computing and Network Communications (CoCoNet’19) Procedia Computer Science 171 (2020) 148–157 Automatic detection and characterization of Lunar Wrinkle ridges a* a b Third International Conference on Gandhi Computing and Network Communications Suchit Purohit ,Savita ,Nidhi Dubeya,Prakash Chauhan(CoCoNet’19) Department of Computer Science,Gujarat University,Ahmedabad Automatic detection and characterization of Lunar Wrinkle ridges Indian Institute of Remote Sensing , ISRO,Dehradun a b Suchit Purohita*,Savita Gandhia,Nidhi Dubeya,Prakash Chauhanb Abstract a Department of Computer Science,Gujarat University,Ahmedabad b Indian Institute of Remote Sensing , ISRO,Dehradun Related computational advances, across various geoscience disciplines, have led to development of an upcoming field: Computers & Geosciences. This multidisciplinary research area is attracting interest of both computer science and planetary/space scientists since last two decades. Under this domain, the most significant area which is gaining importance is Abstract application of computational intelligence to automatically extract landforms present on planetary surfaces and determine their morphometric attributes.advances, Wrinkle ridges one ofgeoscience the common features on lunar that need toofbeanexplored more for Related computational acrossare various disciplines, have ledsurface to development upcoming field: further computation and analysis. paper focuses onresearch application processing approaches automatic science detectionand of Computers & Geosciences. ThisThis multidisciplinary areaof isimage attracting interest of bothforcomputer wrinkle ridges and evaluation morphometric parameters like length,the width, area. The issoftware planetary/space scientists sinceoflast two decades. Under this domain, mostorientation significantand area which gainingdeveloped importancewas is tested on DEMs (Data Elevation Model) oftolunar wrinkle ridges utilizing SLDEM (Selene Lunar Data Elevation datatheir set. application of computational intelligence automatically extract landforms present on planetary surfaces and Model) determine The test datasetattributes. comprisesWrinkle of 9 lunar wrinkle 62 segments withona total of that 1414need km. to Theberesults weremore verified morphometric ridges are ridges one ofhaving the common features lunarlength surface explored for by manual calculations QGISThis software. software was foundoftoimage perform with a mean percentage from further computation andusing analysis. paper The focuses on application processing approaches for deviation automaticvarying detection of 3.10 to 5.97 forand different parameters wrinkle ridges evaluation of morphometric parameters like length, width, orientation and area. The software developed was tested on DEMs (Data Elevation Model) of lunar wrinkle ridges utilizing SLDEM (Selene Lunar Data Elevation Model) data set. The test dataset comprises of 9 lunar wrinkle ridges having 62 segments with a total length of 1414 km. The results were verified © 2020 Thecalculations Authors. Published by Elsevier by manual using QGIS software.B.V. The software was found to perform with a mean percentage deviation varying from This to is an open article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) 3.10 5.97 for access different parameters Peer-review under responsibility of the scientific committee of the Third International Conference on Computing and Network © 2020 The Authors. Published by Elsevier B.V. Communications (CoCoNet’19) This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) © 2020 The under Authors. Published by B.V. committee of the Third International Conference on Computing and Network Peer-review responsibility ofElsevier the scientific This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Communications (CoCoNet’19). Peer-review under responsibility of the scientific committee of the Third International Conference on Computing and Network Keywords:Lunar wrinkle ridge; DEM; Automatic detection; morphometric parameters evaluation; Communications (CoCoNet’19) 1. Introduction Keywords:Lunar wrinkle ridge; DEM; Automatic detection; morphometric parameters evaluation; Computational issues are becoming increasingly critical for virtually all fields of geoscience. This includes the development of improved algorithms and models, strategies for implementing high-performance computing 1. Introduction platforms, or the management and visualization of the large datasets provided by an ever-growing sensor technology Computational issues are becoming increasingly critical for virtually all fields of geoscience. This includes the 1877-0509© 2020 The Authors. Published by Elsevier B.V. development of improved algorithms and license models, strategies for implementing high-performance computing This is an open access article under the CC BY-NC-ND (http://creativecommons.org/licenses/by-nc-nd/4.0/) platforms, or the management visualization ofthe theThird large datasets provided an ever-growing sensor technology Peer-review under responsibility of theand scientific committee of International Conference by on Computing and Network Communications (CoCoNet’19) 1877-0509© 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the Third International Conference on Computing and Network Communications (CoCoNet’19) 1877-0509 © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the Third International Conference on Computing and Network Communications (CoCoNet’19). 10.1016/j.procs.2020.04.016 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 Author name / Procedia Computer Science 00 (2019) 000–000 2 149 from increasing planetary missions. Such issues are applicable to scientific fields like geological modelling, geophysics or climatology in general and planetary sciences in particular. Related computational advances, across various geoscience disciplines, have led to development of an upcoming field: Computers & Geosciences. This multidisciplinary research area is attracting interest of both computer science and planetary/space scientists since last two decades. Under this domain, the most significant area which is gaining importance is application of computational intelligence to automatically extract landforms present on planetary surfaces and determine their morphometric attributes. Some of the key landforms present on planetary bodies are impact craters, wrinkle ridges, faults, valleys, domes etc. Of many landforms, wrinkle ridges are the oldest one to be detected on any planetary surface. Geologically, a ridge is a feature that consists of hills or chain of mountains that form a continuous elevated crest for some distance. Ridges are formed by structural changes that have occurred on the lunar surface when the basaltic lava is cooled and contracted on lunar surface [12], [11], [5] and [10].The topography of wrinkle ridge being irregular and disconnected in nature, can be described as gently sloping from one side and drop-off from the other side with a broad platform in between both sides [13]. In an image/DEM, a ridge can be defined as points having an extremum (i.e. the maximum or minimum value of a function) in the direction of the largest surface curvature. Over the years, the measurements and analysis of morphometric parameters of wrinkle ridges are done using visual interpretation and the parameters measured using GIS-based tools with manual intervention. The method is inefficient for large volumes of data and is subjective to individual perceptions. To overcome this problem, we propose an algorithm for automatically detecting the ridges on lunar surface and determining their morphometric parameters of length, width, area and orientation. The method is fully automated in the sense that whole process from detecting it in selected region, evaluating the parameters and exporting results to an external database for generation of reports is completely automatic and free from manual intervention. The rest of this paper is structured as follows: Section 2 summarizes the experimental data sets. Section 3 presents the literature survey, Section 4 presents the proposed methodology; Section 5 elaborates the results followed by conclusion in Section 6. 2. Dataset In this study, SLDEM data set is used. The Lunar Orbiter Laser Altimeter (LOLA) data and Selenological and Engineering Explorer (SELENE or Kaguya, operated by the Japan Aerospace Exploration Agency) data have been combined to generate the SLDEM digital elevation models [14]. This combination produces a near-global lunar DEM with an effective resolution of ~60 m at the equator. The SLDEM has a typical vertical accuracy of ~3 to 4 m. The data are given in both PDS floating-point image (IMG) and JPEG-2000 (JP2) formats. Data in the GLOBAL subdirectory cover +/- 60 degrees in latitude and 360 degrees in longitude. Resolutions of the images of SLDEM data are either of 128 or 256 pixels per degree. Files in the TILES subdirectory contain the 512 pixels per degree product in the form of tiles (http://ode.rsl.wustl.edu). The developed software was tested on 9 wrinkle ridges having a total of 62 segments and encompassing over a length of 1414 km. 3. Literature survey The automatic detection of wrinkle ridges has always been a challenging task due to its irregular shape that attributed to gradual erosion and degradation over a period of time. Very little work has been reported regarding automation of detection and characterisation. An algorithm proposed in [1], is based on a combination of fractal dimension and morphological operators for automatic detection of wrinkle ridges from Magellan Synthetic Aperture Radar (SAR) imagery at different scales. A technique based on phase symmetry and phase congruence for automatic ridge detection in lunar images is proposed in [6]. This approach provides better results than the plan curvature method, but does not work for ridges having distorted shape. Wrinkle ridges typically consist of a broad arch and a superposed sharper ridge ([8], [9]). This topography enables them to be distinguished from the surrounding terrains by the slope change [10]. This property was exploited in the algorithm proposed in [7] . Their proposed algorithm consists of the slope map generation; conversion to grey scale (0-255) image; phase symmetry generation of the slope map with filter wavelength and filter scales parameters; followed by application of regional threshold to limit the ridge candidates. The method was tested on Lunar Reconnaissance Orbiter Camera (LROC) WAC image and topographic data from LOLA have been used in this research. Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 Author name / Procedia Computer Science 00 (2019) 000–000 150 3 The existing approaches work towards automation of detection process only. Morphometric parameter evaluation is manual or semi-automatic. To facilitate automatic mapping of wrinkle ridges, there is need of methodology to automatically detect ridges as well as extract the individual segments followed by calculation of the parameters of individual segments automatically. More or less all researchers have applied their algorithms on LOLA and/or LRO data sets. In this study, SLDEM data set is used which is superior in terms of resolution and coverage, as compared to LOLA data set. 4. Methodology The flow chart of the adapted methodology is shown in Fig. 1. Global Lunar Image (DEM) Small Ridge Elimination Orientation Computation of each ridge segment Crop Image Contrast Stretching Crater Elimination Computation of width and length of each ridge segment with the help of its topographic profile. Pre-Processing Noise Reduction Edge Detection (using Canny edge detection technique) Computation of area of all the wrinkle ridge segments Converting RGB image into binary image Filtering Process (by using Gaussian filter) Testing and Validation against benchmark /manual measurements Fig. 1. Methodology of the proposed automatic ridge detection and morphometric parameters evaluation in lunar images 4.1. Pre-processing In the pre-processing stage, contrast stretching is applied to improve the contrast in an image by `stretching' the range of intensity values between 0.4127 and 0.4427. As digital images are prone to various types of noises that can be introduced during image acquisition and/or transmission process, a noise removal process was followed by contrast stretching. For noise reduction, linear filtering method was applied using Gaussian filter with sigma value as '0.8'. 4.2. Edge Detection and crater elimination Edges of the enhanced image are delineated using canny edge detector with a threshold ranging from 0.25 to 0.30. The crater is the most important feature present in the planetary image which may also be represent in the areas pertaining to ridges. These features need to be eliminated to avoid misinterpretation of ridge on boundary of crater. In this study Circular Hough Transform (CHT) is used to segregate craters from the terrain in an image. The Hough Transform and its several modified versions are recognized as robust techniques for curve detection. The intent of this CHT is used to find craters; which are circular patterns within an image. In this study, crater detection algorithm proposed in [16] below. The algorithm returns the center and radius of each crater. Accordingly area enclosed by each crater is calculated. If the crater’s locations belong to the ridge location, then it is excluded from elimination else removed using area filter. The image obtained after the crater elimination consists of required ridge 4 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 Author name / Procedia Computer Science 00 (2019) 000–000 151 regions, which is further analysed to retain significant ridges. 4.3. Small ridges elimination The ridges of relatively small area may be the crater rims which may not have been handled in crater elimination stage. In order to retain the significant ridge, the area of each ridge segment is obtained. The trial and error method was used to choose the area threshold value and it is applied to limit the ridge area to be detected. 𝐵𝐵𝑊𝑊𝑎𝑎 >= Ar (1) Where BWa is a segmented image and Ar is an area threshold which limits the minimum ridge feature to be detected. The area threshold (Ar) is image independent and thus the minimum detectable area is kept constant for all the images. The area threshold used in this study varies from 50 to 80 and pixel connectivity is '8'. After area filtering, the image consists of disconnected ridges. These ridges are connected using the morphological operations of dilation and thinning. The image obtained after the Post-processing consists of significant ridge segments. 4.4. Wrinkle ridge orientation. length, width and area The length of wrinkle ridges was calculated by dividing the dorsa (or ridge) in multiple segments on the basis of their orientation along with the horizontal axis(x) in DEM image. The orientation of ridges is depicted through rose diagrams. A rose diagram is defined as a circular histogram plot which displays directional data and the frequency of each class of direction. Fig. 5 shows rose diagram for each segment of Nicol ridge. The rest of orientations are provided in the supplementary file. A bounding box on each segment of wrinkle ridge is created on the basis of their orientation with respect to x axis as shown in Fig. 2.The coordinates of bounding boxes (i.e. xmin, xmax, ymin, ymax) were generated from each bounding box of segment, with the help of those two end points, the automatic lines were generated on each segment of ridge as shown in Fig. 2. The width of ridge was calculated with the help of topographic profile of each ridge segment. In the topographic profile of ridge segment, the distance between the two end points gives automatic width for each segment of ridge (shown in Fig. 3). For calculating area of ridge, sum all the area occupied by each ridge segment with the help of connected components of wrinkle ridge as shown in Fig. 4. 152 Suchit Purohit et al. / Procedia Computer 171 (2020) 148–157 Author name / Procedia Computer Science Science 00 (2019) 000–000 Fig. 2. Calculation of length of ridge segment in an image of dorsa Nicol Fig. 3. Calculation of rim width of segment of Nicol ridge with the help of profile 5 6 Author name / Procedia Computer Science 00 (2019) 000–000 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 153 The schematic image-based framework of the above aforementioned methodology is shown in for Dorsa Nicol is shown in Fig. 4. Fig. 4. Schematic framework of automatic ridge detection and morphometric parameters evaluation (1) Lunar image of Dorsum Nicol (18.0°N and 23.0°E in Mare Serenitatis ); (2) Cropped image; (3) Pre-processing and filtering; (4) Canny edge detection; (5) Post-processing includes crater elimination, small ridge elimination; (6) Calculating ridge length of first segment of ridge on basis of their orientation; (7) Calculating wrinkle ridge segment width and (8) Calculating area occupied by all the segments of the wrinkle ridge. 154 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 Author name / Procedia Computer Science 00 (2019) 000–000 Fig. 5. Rose diagram showing orientation of each segment of Dorsa Nicol 7 8 Author name / Procedia Computer Science 00 (2019) 000–000 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 155 5. Results The algorithms for determination of parameters were implemented in MATLAB (version R2016b) using mapping and image processing toolboxes. The programs were executed on workstation with 12GB internal RAM and processor Intel core i7 CPU with 3.60GHz. The program was tested on wrinkle ridges of different types using developed algorithms. For this purpose a data set was created with 9 lunar wrinkle ridges from mare and highland region as listed in Table 1. Fig. 6 shows the marked test sites on global Lunar map. The summary of the data set prepared number of segments for each ridge and observed outputs of morphometric parameters are provided in the supplementary file. For the validation of output, mean percentage deviation of each measurement was calculated along against the manual observation using QGIS software. The program performed at different levels of accuracy for different parameters as summarized in Table 2 and demonstrated in Fig. 7 through Fig. 10. Fig. 6. Global image of lunar where wrinkle ridges are labeled as (a)Dorsa Nicol, (b)Dorsa Lister, (c)Dorsa Smirnov, (d)Dorsa Ewing, (e)Dorsa Zirkel, (f)Dorsa Mawsom, (g)Dorsa Argand and (h)Dorsa Whiston Table 1. Dataset of wrinkle ridges included in proposed algorithm Name of ridges Number of segments 1 Dorsa Burnet 6 2 Dorsa Ewing 6 3 Dorsa Lister 6 4 Dorsa Smirnov 8 5 Dorsa Whiston 4 6 Dorsa Zirkel 6 7 Dorsa Nicol 11 8 Dorsa Mawson 6 9 Dorsa Argand 9 S.no. 9 Author name / Procedia Computer Science 00 (2019) 000–000 Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 156 Table 2, Mean percent deviation of morphological parameters Parameter Mean percent deviation (in %) Ridge Length 5.97 Ridge Width 4.62 Ridge Orientation 3.10 Ridge Area 5.57 wrinkle ridge width 30.00 300 250 Observed ridge width (in kms) Observed ridge length(in kms) Wrinkle ridge length 25.00 200 y = 0.9856x R² = 0.9876 150 100 20.00 y = 0.947x R² = 0.9753 15.00 10.00 50 0 0 100 200 300 5.00 0.00 Expected ridge length using QGIS (in kms) 0 10 Wrinkle ridges area Observed ridge orientation (in kms) 0 50 100 150 Expected ridge orientation using QGIS (in kms) Fig. 9. Observed orientation vs. manual observation Observed ridge area (in kms) Wrinkle ridges orientation y = 0.9892x R² = 0.9581 30 Fig. 8. Observed width vs. manual observation Fig. 7. Observed length vs. manual observation 160 140 120 100 80 60 40 20 0 20 Expected ridge width using QGIS (in kms) 30000 25000 20000 y = 0.9807x R² = 0.9962 15000 10000 5000 0 0 10000 20000 30000 Expected ridge area using QGIS (in kms) Fig. 10. Observed area vs. manual observation Suchit Purohit et al. / Procedia Computer Science 171 (2020) 148–157 Author name / Procedia Computer Science 00 (2019) 000–000 10 157 6. Conclusions An automated method for detection and determination of morphometric parameters of lunar wrinkle ridges has been proposed in this research paper. Our proposed automated methodology was tested on 9 lunar wrinkle ridges with a total of 63 segments on different regions of moon, i.e. Mare and Highland region. The proposed methodology gives mean percentage deviation varying from 3.10 to 5.67 for different morphometric parameters. The proposed methodology and its implementation through program have been found to give quite accurate results and thus it is regarded as fast and robust. Acknowledgements This work received support from the Indian Space Research Organization under the grant ISRO/SSPO/Ch1/2016-17. We would like to thank Space applications centre, ISRO for provision of expertise, technical and logistical support throughout the development of work depicted in this paper References [1] Barata, M. T. et al. 2015. “Automatic Detection of Wrinkle Ridges in Venusian Magellan Imagery.” Geological Society Special Publication 401: 357–76. [2] Duda, Richard O., and Peter E. Hart. 1972. “Use of the Hough Transformation to Detect Lines and Curves in Pictures.” Communications of the ACM 15(1): 11–15. [3] Ioannou, Dimitrios, Walter Huda, and Andrew F. Laine. 1999. “Circle Recognition through a 2D Hough Transform and Radius Histogramming.” Image and Vision Computing 17(1): 15–26. [4] Kimme, Carolyn, Dana Ballard, and Jack Sklansky. 1975. “Finding Circles by an Array of Accumulators.” Communications of the ACM 18(2): 120–22. [5] Lucchitta et. al. 1977. “Topography, Structure and Mare Ridges in Southern Mare Imbrium and Northern Oceanus Procellraum.” The British Journal of Psychiatry 111(479): 1009–10. [6] Michael, G. G., K. Di, J. Liu, and Z. Yue. 2014. “Global Survey of Lunar Wrinkle Ridge Formation Times.” Earth and planatery science 477: 14–20. [7] Peng, M., Y. Wang, Z. Yue, and K. Di. 2019. “Automated Detection of Lunar Ridges Based on Dem Data.” International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences - ISPRS Archives 42(2/W13): 1431–35. [8] Sharpton and Head et. al. 1988. “Lunar Mare Ridges: Analysis of Ridge-Crater Intersection and Implications on Techtonic Origin of Mare Ridges.” LPSC 112(483): 211–12. [9] Strom, Robert G. 1972. “Lunar Mare Ridges, Rings and Volcanic Ring Complexes.” Symposium - International Astronomical Union 47(1): 187–215. [10] Tanaka, L et al. 1991. “Reconciliation of Stress and Structural Histories of the Tharsis Region of Mars.” 96: 617–33. [11] Watters, T. R. 1988. “Wrinkle Ridge Assemblages.” Journal of Geophysical Research 93(7): 236–54. [12] Watters, Thomas R., and Mark S. Robinson. 1997. “Radar and Photoclinometric Studies of Wrinkle Ridges on Mars.” Journal of Geophysical Research E: Planets 102(E5): 10889–903. [13] Yue, Z. et al. 2015. “Global Mapping and Analysis of Lunar Wrinkle Ridges.” Journal of Geophysical Research: Planets 120(5): 978– 94. [14] Barker, M. K. et al. 2016. “A New Lunar Digital Elevation Model from the Lunar Orbiter Laser Altimeter and SELENE Terrain Camera.” Icarus 273: 346–55. http://dx.doi.org/10.1016/j.icarus.2015.07.039. [15] Ioannou, Dimitrios, Walter Huda, and Andrew F. Laine. 1999. “Circle Recognition through a 2D Hough Transform and Radius Histogramming.” Image and Vision Computing 17(1): 15–26. [16] Gandhi, Savita R., and Suchit Purohit. "Sparse Coded SIFT Feature-Based Classification for Crater Detection." In Information Systems Design and Intelligent Applications, pp. 863-876. Springer, Singapore, 2018..