Research On Video Moving Object Detection And Tracking

Moving object detection and tracking is an important issue in the field of computer vision, combining image processing, mathematical analysis, probability density analysis and other fields, used in security, health care, video surveillance, intelligent transportation and other fields. It has realized using computer to replace human eyes to the outside world sports awareness and understanding of the analysis object, providing convenience for human life. The technology of moving object detection and tracking has matured, But environmental complex and severe occlusion, multi-objective situation have become the main obstacle to its process. To solve these problems, this paper puts forward the improved method, to break through the obstacles provides a research direction. However, there are many problems.This article main research content is as follows:Moving object detection and tracking are introduced the purpose and significance and research status at home and abroad mainly. Several common algorithms in visual object detection are background subtraction, inter-frame difference method, optical flow. But in the process of the observed, due to the cameras, the weather or light shade and environmental impact, night illumination inadequate machine collected video image is not the most perfect, contains a lot of noise. So you have to in front of the processing and analysis of video image data, smoothing filter for video image frames. Put forward on the basis of the background subtraction, to get the binary image for filtering and morphology processing, then fusion target characteristics analysis of the new detection method to detect the moving targets. In the actual video monitoring, infrared monitoring or visible light video monitoring, through to the background subtraction is used to improve to achieve for the moving target detection.Moving target tracking is according to the information before learning to predict target information to forecast the current frame of target information estimation process. At present there are particle filter target tracking algorithm, the Mean shift algorithm. In this paper, based on Mean shift algorithm to track moving targets. In general, through the establishment of the target model and candidate model and the analysis of the similarity measure correctly to track the target. The specific method is to calculate the target the probability of each value of the color features in the feature space to establish a target model, and then in the subsequent frames may exist target area each eigenvalue computing feature space, thus establishing the candidate target model. and use Bhattacharyya similarity measure function of initial frame model and the current frame the candidate model similarity, by calculating the maximum value of similar function about Mean shift vector of the target, constant iterative calculation Mean shift vector. The ultimate goal will converge to the true location of the target, so as to achieve the purpose of tracking.At the same time in this paper, through the prediction of kalman filter to realize the serious event of even appeared in the process of target tracking. Because rely on color information to feature tracking, causes in the serious barrier tracking is not accurate, the method of using Mean shift algorithm and kalman filter to track and predicted respectively, And the tracking results observed as kalman value correction, using occlusion coefficient to keep out of judgment. Once the serious occlusion, will use the kalman forecast, forecast as a result of the track. Experiments show that the algorithm can accurately to keep out in the process of target tracking.