Venkatesh Raghavan
Prof. Venkatesh Raghavan has been involved in advancing Free and Open
Source Solutions fr Geoinformatics (FOSS4G) over the last three decades.He is actively
involved in collaborative projects in Asia. Presently serves as Professor of
Geoinformatics at the School for Science, Osaka Metropolitan University, Japan. His
research interest include terrain analysis, Remote Sensing and machine learning.
Further details about his research contributions are available at https://researchmap.jp/read0048564?lang=en
Sessions
Linear geomorphic features such as valleys and ridges tend to represent geologic lineaments and contribute important information in targeting natural resources, evaluating hazard risk and elucidate surface deformation caused by tectonic forces. Several methods have been proposed for automatic lineaments detection and are mainly implemented using proprietary software. As a result, refinement of algorithm and optimization of parameter selection remains unresolved.
The focus of this research was to investigate the process of extracting lineaments from Digital Elevation Models (DEM). The entire workflow was implemented and tested using the Free and Open Source GRASS GIS framework using existing functions and addons.
As a preliminary step, DEM was used to calculates terrain forms and associated geometry using the GRASS r.geomorphon function. As an alternative approach, application of Convergence Index that depicts the structure of the relief as channels and ridges was also applied. Test data covering parts of the Rokko mountain range in southeastern Hyōgo Prefecture, Japan, was used to hightlight valley systems used as input for the Canny edge. The Canny filter generates one-pixel-wide line(s) representing the most probable edge position and combines several steps to produce an edge and an angle map.
In the final step, the Hough Transform algorithm used to extract significant lines features by elimination the outlying pixels. The Hough transform is optimized by using angle of edges as additional input to search only pixels which are in the direction of a particular line.
The results using the data processing workflow were evaluated for the test site and the efficacy of automatic lineament extraction techniques were successfully demonstrated. Further, the algorithm was also tested on multi-direction relief shades that highlight terrain features which would not in terrain form or Convergence Index maps alone. As an ongoing work, optimizing parameter selection using machine learning is being tested to minimize subjectivity in linear feature extraction. The presentation will highlight on advantages of the approach to extract and validate “geologic” lineaments and demonstrate results obtained using high-resolution DEM.
Keywords: Lineaments, DEM, Terrain Forms, Convergence Index, Edge Extraction, Hough Transform