Ground-based multi-view photogrammetry for the monitoring of landslide deformation and erosion

Recent advances in multi-view photogrammetry have resulted in a new class of algorithms and software tools for more automated surface reconstruction. These new techniques have a great potential to provide topographic information for geoscience applications at significantly lower costs than classical topographic and laser scanning surveys. Based on open-source libraries for multi-view stereo-photogrammetry and Structure-from-Motion, this study investigates the accuracy that can be obtained from several processing pipelines for the 3D surface reconstruction of landslides and the detection of changes over time. Two different algorithms for point-cloud comparison are tested and the accuracy of the resulting models is assessed against terrestrial and airborne LiDAR point clouds. Change detection over a period of more than two years allows a detailed assessment of the seasonal dynamics of the landslide; the possibility to estimate sediment volumes and 3D displacement are illustrated for the most active parts of the landslide. Algorithm parameters and the image acquisition protocols are found to have important impacts on the quality of the results and are discussed in detail.