Abstract：The airborne LiDAR technology provides a new and rapid way for detecting 3D building. In order to separate building points from mass LiDAR point clouds, a progressive densifying TIN method is first used to filter terrain points, and then top surface points of building are classified with principal component and normal vector. After initial building facets are segmented with region growing method, some fence regions are removed and small holes are filled by using morphological closing operator. Ultimately, small vegetation regions are erased on the basis of area criterion of building. Three sampled complex urban areas are selected to test extraction effects of building. The results show that the efficiency and quality of building defection with this method are better than 92%.
