Stereo Matching And Moving Object Detection And Tracking Method Based On Stereo Vision

Stereo Vision is an important branch of computer vision. One important task of stereo vision is to get 3-D information of objects under various conditions. Stereo vision has unparallel advantages compared to monocular vision and it is a preceding research area of computer vision. We do research on stereo matching and proposes a two-level matching algorithm based on the graph cuts of network.Moving targets detection and tracking are crucial subjects in computer vision, and they have very important practical value and wide developmental prosperity in military visual missile guidance, robot vision navigation, industrial product detection, medical diagnosing, traffic surveillance, etc. Since the traditional method is based on monocular vision, it has difficulties in tracking multiple moving targets, especially when the objects are occluded by each other. We adopt stereo vision method in targets detection and tracking. It mainly researches on the application of method which incorporates disparity background difference algorithm and intensity background difference algorithm, Camshift algorithm and multi-level disparity algorithm in the targets detection and tracking area. Thecontributions are summarized as follows:1. By researching on the stereo matching algorithm, we propose a two-levelmatching algorithm based on the graph cuts of network. First, we get the two-level pyramid data structure for the original image pair and obtain the global optimization matching in the lower resolution image pair by using the graph cuts method. Then under the constraint of the acquired disparity map, the area-based stereo matching algorithm is employed to get the dense disparity map of the original image pair. The algorithm not only reduces the search range of matching, but also ensures the validity of matching.2. We propose a novel detection method based on stereo vision. It detects the region of the target by using the integration of the disparity-based analysis and the intensity-based analysis. Then taking the region of the object as the initial model, the snake algorithm is employed to get the true boundary of the target. This algorithm has three advantages: (1) It can detect target correctly unaffected by lighting conditions. (2) Shadows do not cause problems in the proposed algorithm. (3) The snake algorithm can be used to get the contour of the object correctly even if the shape changes.3. Based on the detected targets, we adopt two methods to track targets. Camshift algorithm is employed to track targets. The Camshift algorithm has advantage to track the single target and targets which have different color distribution. But it is difficult to track targets which have the same color distribution. The multi-level disparity algorithm has the advantage to predict the 3-D information of targets and track targets according to their depth and position. It can track multiple targets under fairly difficult environments.