Research Of Registration Algorithm In Three-Dimension Optical Imaging System

Registration of multiple range images is always an important research field in optical three dimension (3D) profilometry because it is the essential for complex objects modeling and computer aided design (CAD). In this dissertation, the original iterative closest point (ICP) algorithm is improved , and programmed with Matlab computer language. We successfully completed the registration of two partially overlapping range images in different viewpoint which were obtained by the optical 3D measurement system which was based on the digital phase-shifting and fringe projection technique.The dissertation mainly includes followings:1. The meanings of range images registration, research actuality, interrelated registration algorithm and the optical 3D measurement system we used were introduced in the following chapter.2. By introducing the theory and lots of algorithms, an ICP-based registration algorithm for two partially overlapping range images was presented in this dissertation. Main components of ICP algorithms were considered, such as initial transform parameters, corresponding points pair selection, control point sets rebuilt and so on. The algorithm has been programmed by Matlab computer language.3. Through some experiments of measuring different objects, two partially overlapping range images in different viewpoint for the same object were obtained. Then some experiments for testing the registration result and computing the measuring precision of registration-RMS were made. Those preliminary results show that both of computing speed and accuracy are improved.