Stereo Matching Key Technology Research Based On Region Image

Based on the mobile robot navigation for independent research background, to realize no one car or mobile robot identify obstacles and autonomous obstacle-avoiding and the purpose of the path planning, the critical technique of stereo matching is researched deeply in this paper, and a stereo matching algorithm based on region is realized.The primary coverage is organized as follows. At first, the basic principles of computer vision and stereo matching, imaging camera model, triangulation principle, Marr vision theory and stereo matching primary coverage are briefly described, which of these theories are the foundation of the stereo matching algorithms. Then some basic camera calibration methods of the camera calibration technique are discussed. Computer simulations have been conducted to demonstrate their advantages,disadvantages and application area, and the results are given on MATLAB. A summary and research of several typical stereo algorithms based on test pictures have been given to present their general results in every section. Some representative algorithms are researched in details. This part ends with the concluding experiment results and analysis of these algorithms.For the marriage problems, which meet both demands for high precision and real time, a real time stereo matching algorithm based on region is proposed in this paper. The proposed algorithm based on region growing due to its consistent performance in high precision, however, the conventional region growing does not work effectively at real time. Some improvements of the proposed algorithm overcome this shortcoming and are able to reduce computing time. In this paper, the developed algorithm includes several steps, such as computing Error Energy with SAD operator, line growing based on region, disparity optimization. By suitably combining absolute value computer of SAD operator with disparity growing based on the new rules, a better equalization effect can be obtained with a marginal increase in the algorithm efficiency. The average error threshold mechanism makes our algorithm more robust. In order to attain the optimum real time, some techniques are using in the algorithm. Simulation results are given to demonstrate the superiority of the proposed algorithm. These techniques include computational optimization, OpenCV, IPP, which is given with some samples in this paper.