Error Detection of Packing Box Based on Registration between Incomplete Point Cloud and CAD Model

Abstract

Because the scanning point cloud of the packing box is a bottom-free cuboid shape, the traditional Iterative Closest Point (ICP) algorithm cannot realize the registration of the scanning point cloud and the CAD model, thus cannot realize the error detection of the box shape. To solve this problem, this paper presents a registration algorithm of the incomplete point cloud with shape error and the CAD model. Firstly, a method of the quaternion-based surface normal vector alignment is proposed for the initial registration. Then, to achieve fine registration, a new Attribute Surface Labelling (ASL) method is proposed to establish the nearest neighbor set of point cloud to improve ICP algorithm. A registration experiment of packing box point cloud is conducted at last. According to the experiment, the algorithm proposed in this paper can realize fine registration between the incomplete point cloud with shape error and its standard CAD model, as well as detect error of its 3D shape.