Reconnaissance and Characterization of Discontinuities in a Landslide using UAV and Structure from Motion

Abstract:

Photogrammetric methods are promising tools to monitor and characterize complex geomorphological and structural discontinuities in those areas where field mapping and terrestrial photogrammetry is not feasible due to unsafe/near-vertical slope or complex geometry. This study presents investigation, characterization, and 3D point cloud generation of a landslide using 191 photos collected from a DJI Phantom-3 Advanced unmanned aerial vehicle (UAV). We have used the structure-from-motion method to generate a 3D point cloud representing the outcrop from the acquired images and videos.

The discontinuity orientations of a section of the landslide were mapped in the field manually. Moreover, the discontinuity points were manually digitized over the point cloud, and discontinuity planes were generated using the least square fit approach in CloudCompare software. Several measurements were collected for each discontinuity plane, and the average measurement was taken to represent the discontinuity geometry. The automatic discontinuity mapping was performed using the discontinuity set extractor algorithm in MATLAB 2019. The manual (field observed) and semi-automated detection was performed for validation of the automated method. It is possible to observe clusters that represent different rock mass discontinuity sets through geometric analysis of this 3D mesh and plotting of the facet orientations in a polar plot. To cluster discontinuity sets accurately, we have used the density-based spatial clustering of applications with noise (DBSCAN) algorithm. We have performed a stereo-net-based kinematic analysis on the manually and automatically detected discontinuities to analyze the possible differences and inconsistencies. The automatic methods did not correctly detect those discontinuities, which are perpendicular to the slope face. These differences in discontinuity detection can influence the kinematic analysis of potential failure mechanisms. The detected discontinuities can be utilized as input for 3D slope stability analysis using the point cloud and reconstructed surface. From field observations and photogrammetric analysis, we can confirm that UAV-based images, combined with 3D scene reconstruction algorithms, provide a flexible and useful tool to characterize the large landslides.