Research And Implement On Video Traking Technology Based On Kalman Filter

The technology of object detection and tracking in video sequences is one of the hotspots in the field of computer vision. In the civilian uses and military it has a wide range of applications, such as robot vision, visual pre-warning, machine navigation, traffic manage, medical diagnosis and intelligent monitoring. They have great research and application values.This topic is still a problem especially when the background is complex. Such as pose variations of the object, significant illumination changes, background interference and occlusion problem, etc. By reflecting these problems, the main achievements are obtained as follows:Firstly, the research is focused on the two major components of tracker: target predicted and tracking in this dissertation. Analyzed the usual algorithm and explains some simple theoretical.Secondly, introduced several main object features and matching algorithms. This paper mainly studies the color image histogram. And based on that, 2D histogram are proposed. Analyzed the Mean Shift and Cam Shift algorithm. 2D histogram used in the CamShift. The experiment proved that the improved of the CamShift algorithm in the target rotation, illumination changes, partial occlusion and color feature of the background similar to color feature of the target conditions than the traditional algorithm have been improved.Finally to solve this problem of completely occlusion, this paper proposes the use of Kalman filtering to predict the target. This kind of algorithm use the Kalman filtering predict the possible position target occurs in the current frame and then use the improved of the CamShift algorithm to tracking. The results proved that the combination algorithm has excellent tracking result in target totally occluded and. The algorithm for moving object to stay a long time and speed of tracking the rapidly changing circumstances have very good practical. To further improve the video tracking robustness and accuracy and stability under the complex background.