Part of the rockmass assessment and its application in numerical modelling, within the geotechnical engineering field, is acquiring information such as discontinuity number, density, intensity, size etc., which can be obtained by mapping fracture traces on exposed rockmass surfaces and processing of the recorded field data. Moving past from traditional mapping techniques in the field, fracture traces can be extracted from terrestrial light detection and ranging (LiDAR) point-clouds or LiDAR-derived surface models. However, similarly to field mapping, the extraction of fracture-traces is often done manually. This is an arduous and timely task in most cases. The automatic-detection of such traces is an emerging topic in geotechnical engineering; however, existing methods focus solely on the spatial domain. Space-frequency representations are ideal for detecting singularities due to their localization in space and frequency. Furthermore, they allow multiscale analysis, which is important for isolating LiDAR-data noise and weak traces in lower scales. In this study, three space-frequency transforms are evaluated, namely, (1) wavelet, (2) contourlet, and (3) shearlet. In addition, the well-known methods of Sobel, Prewitt, and Canny for edge detection are used for comparison purposes. The performance of the different edge-detection methods is tested using data collected from the Brockville Tunnel in Ontario, Canada. Numerical and visual assessment show that contourlets and shearlets achieve the highest agreement with manually-extracted traces that are used for validation. The two methods, along with minimal user interaction, can be used in order to increase the efficiency of rockmass mapping and geometric modelling in stability assessment of tunnels, mines, slopes, and related applications.
All Science Journal Classification (ASJC) codes
- Geotechnical Engineering and Engineering Geology
- Soil Science