Fitting Cuboids from the Unstructured 3D Point Cloud

Abstract

The structural of 3D point cloud data is very important. This paper proposes an algorithm for fitting the cuboid from the unstructured point cloud data directly. In this paper, a 3D point cloud is fitted into a combination of multiple cuboids. Initializing several seeds randomly, and then segment the point cloud and fitting cuboids from the segmented point cloud through iteration. The method in this paper has obvious advantages than traditional methods, with better robustness and more accurate fitting results.