Street environment change detection from mobile laser scanning point clouds

Mobile laser scanning (MLS) has become a popular technique for road inventory, building modelling, infrastructure management, mobility assessment, etc. Meanwhile, due to the high mobility of MLS systems, it is easy to revisit interested areas. However, change detection using MLS data of street environment has seldom been studied. In this paper, an approach that combines occupancy grids and a distance-based method for change detection from MLS point clouds is proposed. Unlike conventional occupancy grids, our occupancy-based method models space based on scanning rays and local point distributions in 3D without voxelization. A local cylindrical reference frame is presented for the interpolation of occupancy between rays according to the scanning geometry. The Dempster-Shafer theory (DST) is utilized for both intra-data evidence fusion and inter-data consistency assessment. Occupancy of reference point cloud is fused at the location of target points and then the consistency is evaluated directly on the points. A point-to-triangle (PTT) distance-based method is combined to improve the occupancy-based method. Because it is robust to penetrable objects, e.g. vegetation, which cause self-conflicts when modelling occupancy. The combined method tackles irregular point density and occlusion problems, also eliminates false detections on penetrable objects.