Study of a Point Cloud Segmentation with Part Type Recognition for Tolerance Inspection of Plastic Components via Reverse Engineering

This paper presents a point cloud segmentation based on a spatial multiresolution discretisation that is derived from hierarchical space partitioning. Through part type recognition it aims to simplify Computer Aided Tolerance Inspection of electromechanical components avoiding cloud-CAD model registration. A voxel structure subdivides the point cloud. Then, through a suitable surface partitioning, it is linked to component volumes by means of the morphological components of the binary image that is derived from voxel attributes (‘true state’ if points are included in a specific cluster or ‘false state’ if they are not). The proposed approach is then applied on a din-rail clip of a breaker, made by injection moulding. This case study points out the suitability of the approach on box-shaped components or with normal protrusions, and its limits concerning the assumptions of the implementation.