Object Detection And Tracking Based On Video

Visual information, especially dynamic visual information, takes the most part of environment information that is perceived by human beings. Although human beings can see both moving and static objects, people usually have strong interest in the moving objects in many spots, such as safety supervises and controlling in important sites, automation and assistant drive of motorcar, the control of transportation discharge, making guided missile in aviation and military aircraft and so on. Thus, there are great realistic significance and practical value in researching the detection and track of moving objects.This paper mainly studied the fundamental theories and key technologies of intelligent video surveillance. The following topics were researched in details, such as detection and extraction of moving objects with static background, tracking of moving objects and so on. The contributions were summarized as follows:1,Concerning the object detection with static background, current theories of object detection and extraction algorithms were summarized. What's more, the paper proposes a method that base on the gray difference and edge difference, and update the background real-time to guarantee the accurate and integrated moving detection. First of all, the frame subtraction result between background and foreground were obtained by tacking subtraction operation. After that, the edge information for the two frames was obtained by means of Sobel operator. Finally, we obtained the results by logic or processing for the gray difference and edge difference, and then we got the location of moving object. The experimental results show that the improved scheme can satisfy the need of system to vast extent and improve the accuracy of detected result.2,About the tracking of moving objects, for the existence lack of mean-shift algorithm, we used continuous adaptive mean shift(CamShift) algorithm for target tracking. CamShift is a color-based tracking algorithm. In the process of object tracking, CamShift operates on a back-projection image produced from object histogram model. Computed information is used previously to adjust current search window size and location, and then the location and size of the color object were found. The experimental results indicate that color object is efficiently tracked even with complex background and irregular motion by using this algorithm.