The Template Matching Tracking System Based On Dm642 Is Designed

In recent years, with the rapid development of the image tracking technology and the continuously contacting to it, more and more people begin to realize that the image tracking technology is the mainly and important means of understanding and changing the world. At present, the image tracking technology has been applied to many fields. It has broad application prospects in all walks of life, like medicine, public security, transportation, etc.The template matching and target tracking are the difficulties in the video image real-time processing. For the limited speed of device, it is not able to do large-scale template matching and a wide range of search operation in the previous system design. However, it will lose a large amount of information on the basis of ABS algorithm in the compress template, and bringing about a low matching accuracy and an unstable tracking, multiprocessor mode is the only solution to the real-time video processing system design. But in the previous design, a lot of system resources are used for bus control and system data transmission which will affect the system performance.According to the video’s large data flow and the high processing speed characteristic, and the requirement of real-time complex target tracking, we design a set of image tracking system based on the existing technical reserves and the current advanced devices and DM642which used TMS320DM642chip (DSP) and EP3C25F324C7chip (FPGA) as the main body in this paper. FPGA is used to complete template matching calculation. According to the matching results, the system can update templates adaptively, and meet the demands on video displacement, scaling, rotating and so on. Due to the isolation of the template matching calculation and the video channel, it avoids the phenomenon of image lag in the multi-DSP system. At the same time, the FPGA parallel processing mode makes a great development in the number of enhancing the system data and processing speed.Compared with the ABS algorithm, normalized cross-correlation algorithm and moment matching method, the paper summarizes a matching algorithm based on the normalized cross-correlation algorithm. Making full use of the high accuracy and the adaptability of the normalized cross-correlation algorithm, and its image gray linear distortion invariance, it can also quickly search out the best match between the real-time image and template image, and update the target in the new strategy, achieve rapid and stable tracking finally through the calculation template and to match the image cross-correlation value to determine the degree of match.As shown in the research, it has a flexible structure and a strong application in the template matching real-time tracking system based on DSP+FPGA. The system can be easier to meet the real time requirements of complex algorithm and object extraction than previous single DSP system. Its image data processing capacity can reach or exceed25frames per second for720x625pixel and its biggest matching template can reach128x128pixel.