Study Of The Moving Target Detection And Tracking Based On TMS320DM642

Detection and tracking of moving target is a technology which can detect and track moving target. The technology is the center of computer vision system, which can be widely used in many fields, such as traffic, biology, medicine and so on.Digital Signal Processor (DSP) has advantages in real-time image or video proeessing. It has high computing speed, large data processing capacity, and low cost. It is suitable for the target detection and tracking system.The TI ICETEK-DM642 evaluation board is chosen to accomplish the target detection and tracking system experiment in this thesis. A variety of preprocessing filter methods are compared and the median filtering method is selected to remove the noise. A variety of edge detection methods are compared and the Laplacian algorithm is selected to detect the edge. In target identification and segmentation, by comparing the Otsu segmentation's with the edge detection segmentation's results, the Otsu segmentation algorithm combining with the morphological dilation method is selected to detect the target and improve the segmentation result which can remove most of the noise and separate the target and background clearly. A color-based object recognition and segmentation method is proposed, and is combined with the dilation method to improve the results. The Meanshift tracking method and the improved Meanshift algorithm named Camshift method is shown, and combined with the Kalman filter method to improve the tracking. The combining method is first using the predictive function of Kalman to predict the emergence location of target, then using the predict location as the search location in the Camshift method. Using the two segmentation methods above respectively, combining the Camshift and Kalman algorithm, the detection and thacking experiments are taken, the algorithm with the traditional Meanshift method and the combining method are compared to show the combining algorithm's advantage.Carry out experiments in VC++and on TI ICETEK-DM642 platform separately. The video is input from the camera, the image is filtered and then the Kalman combining with Camshift algorithm is used to track the target based on the color target detection method and the Otsu method. Experimental results show that the algorithm can extract the target rapidly and effectively, and can track the target accurately. When similar color object disturbs the target, the algorithm can also track the target normaly without deviation.