Efficient Algorithms for Automated Inspection of Freeform Surfaces

Abstract

This paper reports the development of efficient algorithms for automated inspection of freeform surfaces using point cloud data collected using an inspection system. The registration process to align the point cloud with the input CAD surface model is carried out in two phases, viz, rough registration followed by an accurate (fine) registration. For rough registration, algorithms based on the bounding box method and the convex hull method have been proposed. Fine registration is carried out using a computational geometry approach using a tessellated representation of the surface model. A random sampling technique has been integrated with the final registration algorithm to speed up the registration process. The algorithms have been compared with those reported in literature. Results show that the proposed algorithms substantially reduce the computational time, without sacrificing the accuracy of registration.