The Application of Vision Principle in 3D Reconstruction

Abstract

3D reconstruction technology provides digital model of the actual product for the CAD/CAM system of the reverse engineering, which has practical significance for promoting the development of manufacturing automation. Meanwhile, the digital models through 3D reconstruction have the important applications in the fields of virtual reality, video conferencing, criminal investigation and cosmetic surgery. Especially the research for 3D reconstruction of freedom surface is an important research topic in the computer graphics. In this paper, the technology of 3D reconstruction based on stereo vision principle is presented, which can effectively realize reconstruction of 3D digital model of freedom surface. Firstly, the experimental system of stereo vision is established consisting of two CCD cameras in accordance with the principle of binocular stereo vision. Then, the calibration of stereo vision system is completed by selecting the BP neural network for simulating the mapping relation between the images in left and right imaging planes and 3D spatial objects after comparing the two calibration methods of genetic algorithm and BP neural network. Finally, the vision system is experimentally verified taking the face of plaster and human foot as examples, and very good results are achieved.