Defects Automatic Identification Technology Based On3-D Shapes Registration

This paper describes a method for automatic identification of local physical defects or fatigue in mold, basing on3-D shapes registration.As a key procedure of mold repair and remanufacturing technology, mold defects automatic identification technology not only provides basic theoretical data for surface reconstruction, but also affects the accuracy of digital repair and remanufacturing directly.In order to locate the defects, first, point clouds should be obtained by using reverse engineer technology. Then base on3-D shape registration technology, the point cloud of mold with defects should be matched with standard one. After setting the threshold value, the defects in mold could be screened. So far the most popular algorithm for3-D shapes registration is iterative closest point (ICP) algorithm. Based on the features of ICP algorithm, several efficient variants of ICP algorithms are proposed to improve efficiency and accuracy of program. Before the ICP registration, initial registration is adopted to improve the accuracy. And instead of primary points searching method, k-d tree algorithm is involved to accelerate the program during the ICP algorithm.The paper adopts VS2008and OpenGL to construct a point clouds platform to display and operate point clouds. Through the experiment of3-D shapes registration between perfect model and defect model point clouds, we have identified the local physical defects in mold clouds points.