Rockfall source detection and volume measurement from autonomous UAV-acquired photogrammetry: A case study from a transportation corridor in northwestern Ontario, Canada

摘要

Land-based and traditional aerial photogrammetry techniques have been applied widely and successfully to develop precise three-dimensional (3D) terrain models for many applications, including geological or structural analyses of simple rock cuts. However, whereas the geometry and scale of many hazardous rock slopes adjacent to highways or railways are not really conducive to either terrestrial or downward-looking aerial cameras and laser scanners, the newest generation of autonomous, unmanned aerial vehicles (UAV) equipped with high-resolution side-looking digital cameras present a solution to this vantage-point issue. In this study we generated detailed 3D photogrammetric models of a slope along a railway adjacent to Lake Superior near Marathon, Ontario, Canada, and compared them quantitatively for changes. This slope has a similar configuration to local highway slopes. The models were created using hundreds of digital photographs collected by an autonomous UAV in August 2012 and October 2013, between which a number of rockfall events had occurred and affected railway operations. We calibrated the change-detection routine using a large rockfall event that occurred at another site in western Canada, for which we also had detailed LiDAR as ground-truth. At the Lake Superior site we were able to identify both the source location and volume (approximately 9m~3) of a single rockfall event, using only the UAV-acquired photogrammetric data. In the paper we discuss the data collection and model development, the change-detection and volume calculation methodology for the single 9m~3 rockfall, and we explore the best practices and major limitations in the analysis and applications of these methods to highway problems.