Detecting And Tracking Moving Targets Based On Active Contour Model

Moving object detecting and tracking are the main topics of computer vision and image processing,they concern many subjects such as pattern recongnition,application mathematic, Physics, image processing and so on,and have many applications in robotic navigate,intelligent monitor system,medical image analysis,vedeo image compression and so on.Since Kass and others proposed the active contour model and applied it to track people's lip it had been as an important subject in the field of the active vision,and attracted more and more widespread attention. Different from the visual processing in the traditional methods, the active contour model is a high-level information on the full use of the image top-down approach. Because of its advantages, it was applied in multiple areas, such as medical image processing, image sequence tracking, image segmentation, surveillance monitoring. In recent decades, people have made many improvements on active contour model, proposed a number of different target detection and tracking methods, but because there is a shortage of Snake model itself, making the model in the target detection and tracking process still exist a number of problems.Therefore, it is necessary to continue its more in-depth study.This paper illuminates the fundaments of active contour model, focus on analysis of B-Snake model, dynamic programming and Kalman filter.Based on the work of early researchers, we designed the algorithm which based on Kalman filter and B-spline active contour, and integrated into the global optimization of dynamic programming well. First of all, in the case of B-spline active contour has fewer control points, we can approach the target profile; while using dynamic programming techniques to solve the problem which he background interference caused local minima problems; and within the framework of the Kalman filter, predict target velocity, position and scale size accurately, to form a stable algorithm for tracking.Through the multiple testing of the classic image sequences to verify the stability and accuracy of this algorithm, the results show that in the tracking process the method has the features of high segmentation accuracy and continuously tracking.