Research On Application On Video-based Moving Targets Detecting And Tracking Algorithm

This thesis was focused on the technology of targets detecting and tracking of the surveillance system. In resent years, the application of intelligent surveillance system was becoming one of the most important research topics in many countries. Moving targets detecting and tracking was considered as the basic and the most important problem in this research, it was also an active field that had attracted many researchers. This paper was do some researches on moving targets detecting and tracking, and made some progresses of it. Some new methods were put forward in this article and some experiments were done to demonstrate the validity. The main contents of the thesis included such aspects as follows:1. Moving targets detection was now based on the methods of background subtraction, we proposed an improved method of background reconstruction, which was based on frame difference. This method was proved to be very simple, and it could be realized on computer easily. As it only added the background pixel, when it came to the region of moving targets, the pixel value was replaced by the average value of the gray image, so it could reduce the interference of foreground. Experiment showed that this method could obtain background image effectively ,so that it resolved the problem which existed in background subtraction.2. Taking into account the background was not static, but the need for real-time update. This article proposed a self-adaptive algorithm for background updating. This algorithm overcame the shortcomings of the update rate value of traditional algorithm could only decide by experience and could not self-adaptive.3. According to the optical properties of shadow, the value of pixels within the shadow was lower than that of background region. So we could get the possible region of shadow when we divided current image with background image. After that, we could estimate the shadow region by the histogram of image divided. Finally, we used gradient operator to detect shadow region. The practice proved that the method could eliminate the shadow of moving targets efficiently.4. Kalman filtering was an efficient and common used method in the filed of moving targets tracking now. It could predict the possible direction of moving targets, so it decreased the region of target matching and achieved tracking in shorter time. This method could also resolve the"hidden"problem. So we used this method in moving targets tracking filed. The result showed that it could complete the tracking mission.