Abstract: With the continuous development of 3D real scene construction, the requirement for the current situation of entities for terrain-level 3D real scene is increasing day by day. In order to solve the timeliness and high efficiency of geographical scene update, this paper uses unmanned aerial vehicle (UAV) to obtain the laser-point cloud data onto the changed pattern monitored by remote sensing satellites. TerraScan point cloud classification filter algorithm is used to extract ground points in batches, and CloudCompare Laplacian smoothing algorithm is used to smooth and eliminate discrete non-ground points. DEM is output, and finally entities for terrain-level 3D real scene update is achieved. The experimental results show that this method is high degree of automation, and reliable in precision. The updated 3D real scene is accurate and natural.

Key words: entities for terrain-level 3D real scene, geographic scene, laser point cloud, DEM data, Laplace smoothing