Study Of 3-Dimensional Stereo Matching Based On Edge Effect

It is necessary to obtain depth map in the stereo matching. There has been a significant development in this field over the past few years. Depth map is mainly a kind of object position information that can be yielded by processing 2D images based on image attributes—such as edge and pixel intensity.In this paper, we present a research about binocular stereo vision algorithm. We focus on camera geometrical model, stereo matching algorithm and depth map extraction.The first part provides a state of the art introduction to Epipolar Geometry Model in the stereo vision system, and we summarize the related knowledge in this field. In the first part we also set the camera intrinsic parameters for constructing epipolar geometry, which will help people understand stereo matching and provide theory proof for real environment experiment. In the second part we present a survey of local matching and global matching algorithms. And then two algorithms based on edge effect are presented.In the first algorithm accumulation of edge difference and energy minimum are adopted in order to get the optimal disparity space. The second algorithm focuses on algorithm speed. For the given disparity, only edge constraint areas are matched by the feature correspondences. So it could improve the algorithm speed greatly in this way.Finally we use Bumblebee2 camera to capture real environment images and achieve the algorithms in the VC++6.0 and Matlab. Experiment results show that those algorithms are effective in the real environment experiment. We could obtain the depth data and depth map model quickly and efficiently.