Building Segmentation of UAV-based Oblique Photography Point Cloud Using DoPP and DBSCAN

Abstract

The segmentation of building point cloud is the basis of fast three-dimensional city models reconstruction. A building segmentation method of UAV based oblique photography dense matching point cloud is proposed using density of projection points(DoPP) and density based spatial clustering of applications with noise(DBSCAN). First, the building facades are extracted according to the density of projection points by using the rich facade features, based on the analysis of different spatial target features. Then, the density clustering method is introduced to further segment the extracted building facades, so as to realize the monomer segmentation of building facade from UAV tilt photography point clouds. Experimental results show that the proposed method can achieve good results, and provide a new building segmentation method from UAV based oblique photography dense matching point clouds.