Automatic registration of multi-view terrestrial laser scanning point clouds in complex urban environments

Abstract

This paper proposes a method to automatically register multi-view Terrestrial Laser Scanning Point Clouds in complex urban environment. Firstly, the point cloud segmentation method is applied to cluster the cross-section by self-adaptive distance cluster algorithm. Then, geometric primitives are extracted from point clouds by fitting lines or cylinders, whose spatial continuity is used to extract feature lines and feature triangles. Next, triangles are matched according to their similarity and mismatches are rejected based on geometric constraint. Finally, the proposed method registers multi-view point clouds by constructing a minimum spanning tree of weighted undirected graph. Experimental results show that our method is feasible and it improves the efficiency and precision of registration in urban environment.