The Research Of Moving Targets Trajectory Tracking Technology Based On Machine Vision

Machine vision is the use of computer to replace human eyes and the brain, the major means of getting information and understanding the world. Moving target tracking technology is an important research direction of machine vision. Moving target tracking technology is to study, process and explain the image sequence which contain the target movement information. The technology combines computer technology, image processing, pattern recognition, statistics and other multi-disciplinary knowledge. Its applications involve on military navigation, security monitoring, intelligent robot, biomedicine, etc. It has broad prospects and higher practical value.The paper focused on moving object tracking technology after deep study of machine vision theory and and applications. This moving target trajectory tracking technology research is mainly from the two parts of the movement target recognition and target tracking.Moving object detection research, firstly it introduced the the principle, the advantages and disadvantages of the three classical motion target detection methods including optical flow method, frame differential method, background difference method. Focus on study of background modeling method in the background difference method. Compare the background modeling methods of average, median, gaussian mixture and frame difference through the experiment.It mainly studiy on frame difference background modeling method and improve the method of the initial background. The feasibility of the algorithm is verified by the experiments. The background getted from the improved frame difference background modeling method, is used in moving target detection which is based on background difference method. The detection of moving targets is realized.The research on moving target tracking algorithm, the paper introduces principle and the related content of the Meanshift tracking algorithm and kalman filtering algorithm,and compare their effects through simulation. Meanshift algorithm had a disadvantage of manual initialization and sensitive to the color. The Meanshift algorithm is improved. The result of the moving target detection will be used in Meanshift algorithm.Combine the pixel color space information into the algorithm.The algorithm can track the target accurately under complex background. The feasibility and the advance of the algorithm are verified by video simulation.