Stereo Matching Algorithms Based On The Binocular Vision

With the improvement of computer technology promotes the development in the area of mobile robots, space exploration, virtual reality technology, etc. And then computer vi-sion is more and more important, the key technology of the computer vision is stereo mat-ching, it is difficulty in the implementation, specific, the process is to finding points corresponding to each other. But because of the occlusion, lower texture, noise and others factors, it is hard to making pixel one-to-one correspondence. So, stereo matching is rather popular in computer research field always, the aim is to balance the accuracy and the speed.In this thesis, we studied the existing algorithms on stereo matching and proposed some improved algorithms, as followed:1. The paper starts from the imaging model and introduces the mathematical models of the binocular vision and relevant theories and some knowledge of the stereo matching. Then summarizes basic steps of the stereo matching, analysis the characteristics and the differences between global algorithm and local algorithm, then the evaluation criterion for the stereo matching are provided.2. Stripes was produced easily in the lower texture according to traditional dynamic programming method, and then put forward a stereo matching optimization algorithm based tree dynamic programming, the smooth function is include pixel-to-pixel informati-on and the edge information of the image in the processing of building cost function. Calculate optimization disparity by dynamic programming. The experiments shows this algorithm is effectively in reduce the error in lower texture matching.3. According to the assuming that in a certain range if pixel intensity is similar, they have the same disparities, and then put forward a stereo matching optimization algorithm based on image segment. Image is divides into several parts, computed cost function based chunks combined with adaptive weight matching results as the initial matching cost, Through the strict constraints choose control points (GCP), guidance dynamic programmi-ng to find the optimal path by GCP, the result shows this algorithms the disparity is accu- racy in discontinuous and occlude areas, more accurate disparity map was obtained.