Automatic Registration of Building's Point Clouds Based on Planar Primitive Groups

In this paper, we present an automatic registration method based on planar primitive groups for building point clouds. This method distinguishes planar features with similar structures found in urban scenes, and reduces feature matching search complexity. In the method, planar patches with similar normal vectors are defined as a planar primitive group. We extract planes from point clouds as planar primitives. Using a threshold, we cluster the planar primitives with the similar normal vectors into groups. Finally, we match the planar primitives in groups, and calculate transformation parameters with an extended quaternion method. Experimental results show that this method is effective for automatic registration of building point clouds.