Research On Omni-directional Vision Tracker Based On DSP

Auto Guidance Vehicle (AGV) is an important domain for the mobile robot research and engineering application. As the development of the computer science and robotics technology, AGV has been utilized in several aspects, such as manufacturing industry, military and civil application and scientific research. The most critical technology for an AGV which is navigated by vision is beacon recognition and tracking.General beacon tracker which is based on mobile robot usally utilizes an ordinary camera which has a small angle of view. To observe a big angle of view, AGV has to adjust the pan-tilt of the camera, so the real-time of the tracker is affected and the beacon is losable. With an Omni-directional vision system, a hemisphere image can be observerd. It is convenient for recognition, tracking and localization. Fisheye lens is an effective method to construct an Omni-directional vision system which has superwide-angle up to 180°, a whole spatial image can be obtained by using double fisheye lens.Vision tracking system usually needs high computation, but most of tracker consists of PC and video capture board which has a big size and redundant resource. It is not suitable for an on-board environment. Embedded image processing system is a special-purpose system designed to perform one or a few dedicated function often with real-time computing constraints. DSP is powerful processor which has been widely used for audio and video processing.According to the characteristic of the robot vision and the DSP's successful application in image processing domain, an embedded solution which is based on DSP is proposed. An Omni-directional vision system is constructed which consists of 1/2 inch fisheye lens and wat-221s color industrial camera. Beacon recognition algorithm and particle filter has been implanted in TMS320DM642 DSP. Improvement of robust and real-time performace can be obtained after several refining method.