Model retrieval based on point cloud encoding of airborne lidar

Many building models are available in the internet. Based on the concept of data reuse, we propose a system to retrieve models whose shapes are similar to that of point clouds. To consistently encode building models and point clouds, a novel encoding approach is introduced. Our approach can not only encode polygon models but also can handle unorganized, noisy, and incomplete point clouds. In addition, to ease of encoding problem suffering from incomplete data, we propose a resampling technique for point clouds. The experimental results show that our approach can solve the inherent problems of point clouds, i.e., unorganized, noisy and incomplete, and can encode the 3D shape consistently.