UAV-Based Digital Imaging System for the Derivation of 3D Point Cloud for Landslide Hazard Analysis
Emergency disaster response and analysis of landslides depend on accurate, rapid detection and extraction of a landslide area. Terrestrial laser scanning systems (TLS) are highly accurate and provide quick 3D point cloud data with high resolution, but suffer from occlusions, truncation, and orientation bias. This dissertation proposes an augmentation of TLS and an image-based point cloud generated from a semi-global matching (SGM) algorithm on an UAV platform outfitted with a low-cost action camera to overcome these limitations. The experimental results provided high quality measurements for the geotechnical discontinuity plane orientation parameters, increased safety, saved cost and time, and provided more accurate results compared to manual field measurements, TLS data only, and SGM data only. This dissertation developed a comprehensive system using UAVs and SGM techniques to accurately identify and extract landslide scarps within centimeter-scale accuracy through three automated approaches. These approaches accurately detected and extracted landslide scarps based on the ratio of the normalized Eigenvalues derived using principal component analysis, surface roughness index, and slope measurements from the 3D image-based point cloud. Experimental results using the fully automated 3D point-based analysis algorithms confirmed that these approaches can effectively distinguish landslide scarps. The developed algorithms are a flexible and effective tool for monitoring landslide scarps and are acceptable for landslide mapping purposes. A robust image-based registration method also was developed for the simultaneous evaluation and temporal monitoring of landslide dynamics from different epochs. This method includes the camera’s IOPs and EOPs of the involved images from all the available observation epochs via a bundle block adjustment with self-calibration. A SGM technique was implemented to generate 3D point clouds for each epoch using the images captured for each epoch separately. The accuracy of the co-registered surfaces was estimated by comparing the non-active patches within the monitored area of interest. Since non-active sub-areas are stationary, the computed normal distances theoretically should be close to zero. The quality control of the registration results showed an average normal distance of approximately 3.7 cm, which is within the noise level of the reconstructed surfaces. Overall, the registration approach proposed in this dissertation is low level.
Remote Sensing, Environmental Sciences, Engineering--Environmental, Geotechnology
Al-Rawabdeh, A. (2016). UAV-Based Digital Imaging System for the Derivation of 3D Point Cloud for Landslide Hazard Analysis (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/28539