Comparison between image- and surface-derived displacement fields for landslide monitoring using an unmanned aerial vehicle

• Published in: International journal of applied earth observation and geoinformation Vol. 116; p. 103164
• Main Authors: Teo, Tee-Ann, Fu, Yu-Ju, Li, Kuo-Wei, Weng, Meng-Chia, Yang, Che-Ming
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.02.2023; Elsevier
• Subjects: Displacement field; Iterative closest point (ICP); Landslide; Particle image velocimetry (PIV); Unmanned aerial vehicle (UAV); Particle image velocimetry (PIV); Displacement field; Iterative closest point (ICP); Unmanned aerial vehicle (UAV); Landslide
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2022.103164

•This study determined the benefits of using UAV-image-derived 3D point clouds to extract 3D displacement fields for landslide monitoring.•This study compared the landslide displacement fields from three approaches (i.e., image-based PIV, DSM-based PIV and point-based ICP).•This study demonstrated the extraction of angular changes using point-based method for landslide monitoring. The traditional particle image velocimetry technique generates a 2D displacement field for landslide monitoring using multi-temporal unmanned aerial vehicle (UAV) orthoimages. As UAV photogrammetry can produce a 2.5D digital surface model (DSM) and 3D point clouds, two different surface-based approaches—DSM- and point-based methods—were developed to provide 3D displacement fields for landslide monitoring. The DSM-based approach utilized the image matching technique via an interpolated surface model, while the point-based approach used the windowed iterative closest point technique via irregular points. Several in-situ real-time kinematics measurements were used to analyze the quality of the different approaches. The experimental results showed that the performance of the point-based method was better than the image- and DSM-based approaches and attained 0.1 m accuracy for horizontal and vertical displacement. In the qualitative analysis, the results of the point-based method were similar to the actual surface movement, demonstrating uniform behavior in the landslide region. In summary, the use of point clouds from dense image matching proved beneficial for providing 3D displacement fields for landslide monitoring.