LAUSR

A mobile mapping system (MMS) is effective for capturing dense point-clouds of roads and roadside objects. In order to create 3D models from huge point-clouds, it is necessary to efficiently extract objects from point-clouds. However, since points captured using the MMS are highly anisotropic, it is difficult to detect local connectivity between points using a constant threshold. In this paper, we discuss the method to define adaptive thresholds for local connectivity of highly anisotropic point-clouds captured using the MMS. In our method, point-clouds are mapped on the 2D lattice and they are connected on the lattice. Then we introduce adaptive thresholds by simulating laser scanning with the MMS and comparing the simulated point intervals with actual ones. By using the adaptive thresholds, continuous surfaces can be stably extracted from large-scale point-clouds.