The Algorithms Reserach Of Multi-target Tarcking System Based On DSP

As an important branch in the area of Computer Vision,the technology of the detection and tracking of the moving targets has been widely used in many fields,such as Artificial Intelligence,Security Technologies and Automatic Control.The DSP system has high clock and can process arithmetic statement quickly.It has powerful processing function and can do a great deal of calculation in a short time.The DSP is low in price and easy to be integrated.As a result,it is undoubtedly worthy researching that build a system about the detection and tracking of the moving targets on the hardware platform of the DSP processor.Based on the research of the hardware platform of TMS320DM6437 processor,we have designed a system about the detection and tracking of the moving targets.The design can be divided into three modules: image preprocessing,target detection and target tracking.In order to obtain high-quality images for subsequent processing,we use median filter for image preprocessing in the module of image preprocessing.In order to improve the shortcomings of three-frame-difference method and background subtraction,this paper presents a new moving object detection algorithm based on edge information of the three-frame-difference method combined with the background subtraction.In target tracking module,this paper improve the Kalman filter.This paper presents a new algorithm based on Mean shift and Kalman filter to track the multi objects.In order to verify the functionality of the improved algorithm,the target detection algorithm and tracking algorithm are implemented on the Matlab simulation platform,compared and analyzed with other traditional algorithms.The algorithm described in this article is implemented in DSP.To improve the real-time performance of the algorithm system and reduce each module takes time.This paper optimizes the software which is transplanted into the embedded system.After getting the image,we store the image in Cache to reduce CPU capture data time.In the aspect of code,we through the loop in the code and eliminate redundant loops to improve the algorithm efficiency.In the end,to verify the functionality of this algorithm,we use the oscilloscope to calculate the running time of each module algorithm.The experiment proves that this paper improves the correctness of the algorithm.At the end of the article,the realization of the hardware and software of multi-target tracking system is summarized and prospected.