Research On Detection And Tracking Moving Object In Video Monitoring

In order to meet the "smart city" and "safe city" construction needs,the city's video surveillance has been basically achieved a full coverage of the region.How to develop and apply high-precision monitoring video analysis method has become the depth and breadth of the application of video hot and difficult problems.The primary task of monitoring video analysis is to detect and track the moving target.However,due to the existence of perspective and tilt in the imaging process,the robustness of target detection and tracking in complex monitoring scene is difficult to meet the needs of practical application.And the influence of the tilt change on the target detection tracking becomes extremely important.In this paper,we first obtain the size information of the target in the video by camera calibration,and then introduce the noise removal method to improve the detection precision in the process of moving target detection.Finally,we use the target size information to correct the image area to eliminate the perspective and angle changes.Combine the feature and contour description Information to improve the robustness of target tracking.The research contents and achievements of this paper include:(1)Camera calibration method based on optical flow extraction vanishing pointIn this paper,a camera calibration method based on the optical flow extraction vanishing point is proposed to extract the vanishing point based on the moving target extraction method.By using the optical flow generated by the moving process,Calibration based on the calculation of camera parameters.The experimental results show that the method of extracting vanishing point based on optical flow is more accurate and easy to use than pedestrian extraction method.(2)based on the foreground memory of the moving target detection methodIn this paper,a Vibe algorithm with memory is proposed to solve the problem that Vibe motion target detection algorithm is easy to generate noise.According to the difference between noise and real prospect in temporal and spatial characteristics,the algorithm parameters of different pixel regions are dynamically adjusted by using memory layer to record foreground pixels.The experimental results show that the Vibe algorithm with memory is more accurate than the improved Vibe algorithm.(3)Moving target tracking method based on compound descriptionIn this paper,a new moving object tracking algorithm based on composite description is proposed to overcome the problem of target perspective change and tilt change based on the size information of the target parameters of the camera.In this paper,based on the problem of low robustness of single tracking method in complex scene,Characteristics and contour-based target description methods to overcome the impact of target occlusion and deformation.The experimental results show that the tracking accuracy of the moving target tracking algorithm is higher than that of the single description based on the single description.The results show that the method proposed in this paper is suitable for the existence of optical flow monitoring scene,which can extend the application range of the existing camera calibration method based on vanishing point.On the basis of camera calibration,this paper improves the accuracy and robustness of the existing moving target detection and tracking method,and can automatically detect and track the moving target in the complex monitoring scene,effectively overcoming the influence of perspective and tilt change.