Research And Implementation Of 3D Reconstruction Algorithm Based On Computer Vision

One of the main components of the computer vision is three-dimensional reconstruction. It is to calculate the depth information of an object by two or more images of two-dimensional. The technology of binocular stereo vision based on two images is a hot spot. The main task of the research on the binocular stereo vision is simulating human eye using technology of binocular vision or multi-vision to understand the objective world .We can get different images of one object from different views and then calculate the position deviation of stereo images pixels based on stereo matching. Finally, we obtain the disparity depth of the object. Theoretically, the reconstruction based on computer vision has the advantages of low-cost, high-reality and high-automatization, so it has a good extensive application prospect. How to recover the three-dimensional information according to the incomplete or noise-corrupted two-dimensional information is the hard problem of this technique and computer vision. The future research of three-dimensional reconstruction based on computer vision can promote understanding those problems and the development of related scientific. Based on the analysis and research of much related technology and literatures, this paper studies the three key technology of binocular stereo vision: the camera calibration, stereo matching and three-dimensional reconstruction. The corresponding work is shown following:(1) Doing the calibration template by ourselves, then getting some images of the calibration template using a common camera from different angles. This paper uses Zhang zhengyou algorithm to calibrate the internal and external parameters.(2) Stereo matching is a key of three-dimensional reconstruction, and it is also research emphasis in this paper. Based on local constraint and global constraint, this paper analyzes advantages and disadvantages of different algorithms. Then we use feature matching and Belief Propagation to calculate the disparity.(3) In algorithm based on global constraint, we use a new method to calculate disparity: first we use LBP/C to analyze images texture, then doing matching by BP algorithm to obtain accurate disparity map.