Research Of Moving Target Detection And Tracking Technique Under Shifting Background

Video-based moving target detection and tracking technology is an important research direction in intelligent video surveillance and computer vision, it is of great significance for other advanced processing like target identification. Many practical applications require the camera moving to realize continuous tracking and expand monitoring range as well as reduce the costs. But the motion of the camera led to the movement of the entire background, compared to static background, it greatly increases the complexity of moving target detection and tracking, so the research of moving target detection and tracking under moving background becomes a priority.In this paper, the technology of detecting and tracking moving target under moving background is studied. The focus of the research is moving object detection and human based tracking under moving background. The corresponding algorithms and strategies are proposed as following.In moving target detection, this paper uses affine parameters model to describe the global motion of the background. The extraction and matching of corner feature are used to estimate the parameters of the model. As for the invalid matching points, a distance-based principle is utilized to eliminate them, which realize the robust estimation of the motion parameters of the background. After compensation of the background with the global motion parameters, the frame difference with the current frame can effectively detect moving objects.After the detection of the moving targets, a HOG and SVM based method is applied to determinate whether the targets is human body. If it is, then the color histogram of the target is calculated and CamShift combined with Kalman filter are used to track the target. In order to enhance the robustness and reliability of the algorithm, this paper improved the calculation of the color histogram of the target and integrated color and motion information for tracking. Further, the problem the target is completely occluded has been analyzed and an occlusion handler is given.Experimental results show that the algorithm can detect and perform accurate and continuous tracking for human target effectively.