Research Of Movin Obiect Detection And DSP Implemention

Moving object detection is an important topic in the field of the research on computer vision. Moving target detection is designed based on analyzing the sequences of video or image and segmented the moving target which would be interesting from the background, obtaining the target area and its related parameters, which is the basis of target tracking, behavior recognition,scene understanding and other high-level visual processing. Moving object detection is widely used in all areas, such as the video surveillance system, the interaction of human-computer, the compression of image, intelligent transportation systems, industrial control, biomedicine, weather analysis, weapons and equipment. So The moving object detection is not only very important on the research value and practical significance, but also for other field of computer vision research,which will promot an important postion.The paper which has thoroughly studied the video image sequence in the moving target detection method with the camera fixed case uptake in the video image sequence as the research object. Firstly, this paper explain the general steps of some commonly used methods of moving target detection including the optical flow method, frame difference method, the background difference method, and also which analysis of the principles of these methods and processes, and compares their advantages and disadvantages. Then I studied the view of the background difference method of common background model, and the analysis and comparison of different background modeling algorithms. At last I research the image pretreatment, morphological processing in depthly.This paper studies the method of moving target detection on based on mixture of Gauss background model. It is the key of this paper. This article first introduces the method for moving target detection on the traditional of mixed Gauss background model, and then pointed out that the traditional hybrid Gauss background model of moving target detection method in the presence of the large amount of calculation, error detection and" slur" problem. So the article proposed one new method for moving target detection based on combining the three frame difference method and mixed Gauss background model. The improved algorithm using adaptive median filtering method on the input image pretreatment, throughing the three frame difference with area method to judge whether the existence of moving targets in video image, if there exists a moving object for the next processing, finally we can use the mathematical morphology processing. Experiments shows that the method can overcome these shortcomings of the traditional algorithm based on mixture Gaussian background model, can effectively detect the moving target.Finally, the paper introduces the hardware platform which the chip DM642is the central processor of by TI company. Introduce the module circuit of the system, including the video capture module, the video display module, and extended memory module. Design the system software process, and part of the porting and optimization of the algorithm.