Identifying individual rocks within laser scans for a rigorous deformation analysis of water dams

Water dams are an important infrastructure component for energy generation, water supply and flood control. Regular inspections of the structures for damage and deformation are necessary for safe operation and to ensure stability. In addition to the traditional concepts of geodetic network measurement, laser scan data can be used to deliver areal information on deformation. Within this topic, we aim at developing a method that identifies individual rocks of dam walls within the laser scans to introduce them as identical feature points for a rigorous deformation analysis. For this purpose, it is necessary to identify the solid stone surfaces on the water dams from scan data, and separated them from vegetation and joints. In this paper, we investigate method for identification of individual stones on gravity dams made of rubble stones. For the segmentation, the intensity values, RGB color information and local geometric structure from textured point clouds acquired with Terrestrial Laser Scanners are investigated. The classification should be robust against outer measurement conditions and provide sharp object boundaries. Our results show that – although many different methods are available – a reliable classification of single rubble stones is still a challenge task.