The Research Of Obstacle Detection Algorithm Based On Binocular Stereo Vision Technique

Binocular stereo vision, which is an important branch of computer vision research, direct simulation human eyes with the way of scenery to realize the three-dimensional information of the perception. Binocular stereo vision use two cameras installed in the different positions, and then recover the depth information. Binocular stereo vision has unparalleled advantages relative to monocular vision. With the development of the computer vision technology, the application of stereo vision is widely spreaded. Especially, with the merits of simple organization and easy-operating, the binocular vision is successfully used in the domains of industry examination, object recognition, work piece localization, robot self-leading, spaceflight etc. The study of binocular stereo vision not only has important theoretical significance, but also has important practical value.The obstacle detection based on stereo vision in the intelligent driving is the background of application. The stereo images are acquired from binocular sensors. Obstacle detection algorithm based on disparity is used to acquire the location and depth of the obstacle. The information of underlying obstacle will be supplied for the drivers or non-driver’s car to achieve the purpose of early warning. This research is significative for intelligent driving and navigation of robots. The main work of this paper is as follows:1. The epipolar rectification of stereo image pairs. The stereo images are hard for stereo matching, because they can’t satisfy the epipolar lines condition. This paper introduces a rectification method to rectify the images in order to improve the computational efficiency and narrow the search range of stereo matching.2. Stereo matching algorithm. The traditional pixel-based stereo matching algorithms have larger amount of calculation, and lower matching accuracy. To solve this problem, this paper presents a stereo matching algorithm based on image segmentation. The number of image segmentation regions are significantly less than the number of pixels in the image, which greatly reduce the algorithm computational complexity. The partial occlusion pixels of the segmented regions calculated by the disparity plane meet the disparity smoothness constraints, and split the region within the constraints of the same model and eliminate the influence of distortion of pixels on the result image.3. Obstacle detection. The V-disparity image is calculated according to the stereo disparity image. The straight line information extracted from the V-disparity map, and through these straight lines information we can get the approximate region of the obstacle. Then we can detect the obstacles.Experiments have shown that this method is correct and feasible.