Reconstruction And Rendering Scenes Based On The Binocular Stereo Vision

In recent years the development of The Computer Vision was very fast, and gradually became one of the hottest research areas in computer science. The Binocular Stereo Vision is an important branch of The Computer Vision, it gets images of the same scene from a different view points by two cameras and recovery the three-dimensional coordinates of the scene of each of the target object by using the triangulation principle and the stereo matching disparity map, then rebuild them.The purpose of this paper is to draw the scene, we are going to test the camera calibration, the stereo calibration, the stereo matching method and the three-dimensional reconstruction.In this paper, we have completed the following works:1. We proposed a stereo calibration method by studying the stereo calibration principle in OpenCV. Because of the three-dimensional reconstruction in this paper is based on triangulation, so we discussed the principles of triangulation and the mathematical model of The Binocular Stereo Vision in detail. Considering the actual factors, we used the stereo calibration principle in OpenCV, the device of The Binocular Stereo Vision consists of two high configuration camera composition and it is parallel to the front of camera lens, then we make a checkerboard calibration board, we obtained the internal and external parameters of the camera through our experiments of the calibration plate calibration. We use the OpenCV library based onVS2010, test the Stereo calibration method, then we acquired the re-projection matrix and the two calibrated images.2. Combining the global match and the local match, we proposed a fast stereo matching method based on graph cuts. In this paper, the most complicated part is the stereo match, this paper reference to the idea of the local match, we optimized the node of network diagram of graph cut method combines the idea of partial matching. In this way we improve the efficiency of the algorithm. Meanwhile we optimized the energy function after considered the occlusion factor in realistic environment, we added a penalty term and imposed a penalty value of occluded pixels. After a number of tests about our stereo matching algorithm, we verified the Stereo matching algorithm in this paper is faster than other global matching algorithm, and the disparity map is better than the local matching algorithm.3. We designed a 3D reconstruction system by using the reconstruction algorithm in OpenCV. We’ve discussed the OpenGL library, according to the principle of triangulation, we obtained to restore the three-dimensional coordinates of the objects in the scene by using the disparity map. Then we used OpenGL technology to draw the three-dimensional coordinates of the acquired and displayed them. The result of our tests is in line with our original goal.