The Research Of The Image Tracking System Based On Embedded Processor

With the embedded technology and automatic control technology's development, image tracking systems are widely appeared in people's life and in the war. Image tracking system is a very complex system. Because it is not only used in the precise detection of tracking object, but also it must complete real-time object tracking. Under the application of special occasions, image tracking system also need to overcome the interference of exterior circumstance. Therefore, the image tracking system need higher requirements for software and hardware.This article mostly researches into the image difference tracking algorithm, Mean Shift tracking algorithm and Kalman filter tracking algorithm. It puts forward two algorithms for the difficult problem of video tracking. One way is the mean shift tracking algorithm based on image difference detection. The algorithm firstly distills the target model of moving objects by image background difference algorithm. Then it completes moving object tracking by meanshift tracking algorithm. Another way is the image difference tracking algorithm based on Kalman filter prediction. The algorithm firstly predicts the area that moving object may appear in the next frame by Kalman filter. Then it completes moving object tracking by image difference algorithm in the divinable area. The former algorithm supplys a gap that setting object model and object centroid coordinates in tracking area by people are required before tracking different object using meanshift tracking algorithm. The latter algorithm does not only supply a gap that computing every pixel information are required in tracking area using image difference algorithm, but it also figures out the instance in effect that moving object can be tracked accurately when it is partially blocked.In this paper, S3C2410 embedded development platform is used for validating the above tracking algorithm. In small granary moving white plastic bottle tracking is implemented by image difference tracking algorithm, each frame takes approximately 0.375s.In the laboratory moving white bubble is automatically tracked by meanshift tracking algorithm based on image difference detection, each frame takes approximately 0.735s. In the laboratory people is tracked by image difference tracking algorithm based on Kalman filter, each frame takes approximately 0.258s. Experimental results show that the mean shift tracking algorithm based on image difference detection makes up the disfigurement of mean shift tracking algorithm and it automatically completes different sorts of moving object tracking. The image difference tracking algorithm based on Kalman filter prediction abbreviates the time of tracking by image difference tracking algorithm and enhances the ability of the real-time tracking using image tracking system.