Study On Object Tracking Method Based On Binocular Stereo Vision

Detecting and tracking of moving objects in image sequence is one of the most important research directions in computer vision and intelligent human-computer interaction. Usually, most moving object detection methods are based on monocular vision. Stereo vision can obtain 3D information under various conditions flexibly, therefore it has unparallel advantages compared with monocular vision. In this thesis, The binocular stereo vision technology is applied to moving objects detection and tracking. Some important topics in stereo vision are studied and discussed, such as camera calibration, stereo matching and disparity image computing. The main work can be summarized as follows:1. Several international popular calibration methods are deeply studied. By analyzing traditional calibration methods, we obtained ideal calibration results by using the camera calibration toolbox under Matlab.2. By studying a binocular stereo vision product Bumblebee with two Software Development Kits provided by the system, Digiclops SDK and Triclops Stereo Vision SDK, we also do some research on stereo matching and disparity image capturing. The binocular stereo camera is used to capture the left and right images, and the appropriate depth stereo image can be obtained by the matching computation between the two images with corresponding constraint setting and parameter validation.3. Binocular stereo vision technology is used in objects detection, particularly moving objects. Because the object has different depth information from the background, which corresponds to gray value in the depth image, therefore the variation of gray value in depth disparity image can be used to extract and locate the moving objects. Experimental results show that better performance can be obtained when detecting moving objects by binocular stereo vision.