Abstract: Aiming at the problem that is easily affected by adjacent vegetation, a building point-cloud progressive extraction algorithm is proposed in this paper for airborne LiDAR. Firstly, the ground point-clouds and the non-ground point-clouds are distinguished by using the cloth simulation filter. Next, the initial building point-clouds are extracted by using Otsu to get threshold to segment the non-ground point-clouds on the basis. Then, DBSCAN is used to eliminate outlier noise points on the initial building point-clouds through the connectivity of the point-cloud. Finally, the edge of the building is obtained through the Alpha Shape algorithm. The experimental selectes three sets of LiDAR point-cloud dates of different type of urban building provided by the ISPRS official website. The result shows that the algorithm of this paper can achieve the extraction of airborne LiDAR building point-cloud better.

Key words: aerial LiDAR, building point-cloud extraction, Otsu, DBSCAN, Alpha Shape