Automated structural discontinuity mapping in a rock face occluded by vegetation using mobile laser scanning

Mapping of structures, such as discontinuities, in an exposed rock mass is fundamental for slope stability analysis. This study investigates mobile laser scanning technology to identify structural discontinuity in a complex environment where the exposed rock face is partially covered with vegetation. The conventional terrestrial laser scanning and photogrammetry based approaches for structure identification rely solely on coplanarity criteria or point normal vectors and often miss several prominent discontinuity planes in a complex environment with inherent noise. To enhance structural mapping in such environments, this study tests a mobile scanner which has multi-view data collection to reduce blind spots. A new automated and robust algorithm termed clustering on local point descriptors (CLPD) is developed for more accurate discontinuity identification. The algorithm involves a rigorous pre-processing step to remove erroneous and irrelevant points followed by computation of local point descriptors. Five descriptors (i.e. eigenvalue descriptor (EVD), radial surface descriptor (RSD), fast point feature histogram (FPFH), normal and curvature) were generated for each point to capture spatial distribution, geometrical relationships and local surface variations in the point cloud. Finally, a K-Medoids clustering was performed on the computed descriptors using a histogram of normals to identify discontinuity planes. Results indicate that the proposed CLPD algorithm outperforms existing approaches in terms of accuracy in discontinuity orientation estimate (3.50° in dip angle and 4.32° in dip direction) for the study site. The stereonet comparison also validated that the poles distribution from CLPD is on par with the ground truth as well as results from commercially available software. The study presents a comparative evaluation of various approaches for kinematic feasibility of planar, wedge and sliding failures. •Mobile laser scanning captures multi-view data to avoid blind spots.•Conventional algorithms miss several important discontinuity planes.•Local point descriptors can lead to robust discontinuity identification.•K-medoids clustering effectively segregates discontinuity using more descriptors.•The novel CLPD algorithm provides a reliable rock slope kinematic analysis.