A New Registration Algorithm for 3D Reconstruction of Objects based on 3D Laser Scanning

摘要

The task of surface reconstruction from three dimensional (3D) range data covers a variety of techniques. The Iterative closest Point (ICP) algorithm is the well-known algorithm used for registration of multiple range images. However, key to the success of the ICP algorithm is to acquire a proper initial transformation values and look for the correspondence relation between a pair of range images to be registered. In this paper we address a somehow different problem that has only little been addressed in the past. Combined with the match technique in photogrammetry, the paper briefly describes the principle of ICP algorithm, presents a new method to get the initial values and the method of looking for the correspondent points in the two adjoining coordinate system for the registration, and practical range images are applied for the implementation of proposed rule to show the efficiency of our approach. In addition, the whole adjustment of image registration is also employed to multiple range images.