Symbolic Method Of Contour Tracking Based On Ellipse-Specific Fitting

One important task of applied vision research is detecting and tracking moving objects. As a kernel task of Computer Vision, the technology of tracking moving objects comprises image processing, pattern recognition, artificial intelligence, automation and many other advanced technologies. This technology has been used in military control and guide, robot's vision navigation, security surveillance, traffic guidance, medical diagnostics, weather analysis and many other fields. Although the technology is promising, it's not appropriate for real-time application because of its high computational requirements.Herein, the idea, which is using symbol method to track moving objects, is proposed. Here, the symbol is achieved by ellipse fitting. Previous algorithms either fitted general conies or were computationally expensive. By modifying the ellipticity constraint, we can ensure what we get is an ellipse, so that we can get the symbol to track moving objects. The approach is steady and its computation cost is low. The total cost of the method proposed here is the same as that of acquiring the initial estimate of Sampson's fit, but its effect is similar to Sampson's method.Several steps of the work have been done to achieve the system. First is to analysis the camera model and imaging transformation by homogeneous reference frame. Second is to study the moving object detection by difference between images and the problems produced by thresholding, then enhance the image and get its contour. Third is to study the conic fitting's status in quo and propose to use ellipse-specific method to track moving object, so that an ellipse symbol is obtained to represent the moving object. Finally the system is realized by VC++ programming.Ellipse-specific fitting is applied to tracking moving objects in the paper. The system based on the idea works fine.