Research Of Detecting And Tracking Moving Objects Methods On Vedio Surveillance System

In recently, the intellectualized video monitoring system is a concern frontier topic in computer vision areas. it has been widely used in industry, medicine, navigation, etc. moving object detection and tracking is the key of video monitoring technology to realize automation and real-time application in video monitoring system, but also is the basis of video scene analysis, behavior understanding and so on many subsequent processing. At present, moving object detection and tracking exists various difficulties because of target motion of complexity in actual environment and particularity of video data. In this paper, we mainly study the algorithm of moving object detecting and tracking under Complex background, the reality based on intelligent visual monitor system, the moving Multi-Object tracking under exceptional circumstances of moving targets shade each other.Firstly, in this paper, we present an approach for the scene of the monocular vision monitoring system that camera bracket fixing and camera can rotate around. This approach is combination of two-way difference multiplication algorithm and matched difference algorithm based on affine motion model, which adopts two-way difference multiplication algorithm when the camera stationary and adopts matched difference algorithm based on affine motion model when the camera is rotating around .Secondly, for the laboratory intelligent visual monitor platform,we propose an approach that the edge information of the outline of the campaign target is used to track a single moving object. This approach is tested on the Surveillance platform with complex background target detection algorithms. Experiment results show that moving object detection approach eliminates the relative movement of the background to achieve accurately detect the moving object under the complex background. and moving object tracking algorithm avoids inaccurate target localization and target lost owing to the target and background color closing or target color in great changes. And so it achieves accurately track single moving object.Finally, for the target motion can be approximated as the linear model in stationary scene, Multi-Object Tracking approach which combines predicting based on Kalman filter and Feature matching is presented this paper. This method uses Kalman filter to predict the current frame moving target location in the next frame, and then uses centroid and area matching equations to calculate the similarity between adjacent frames target, establishes matching matrix. Through the process introduced above, we can find out the best matching result. Simulation results show that this approach can only track Multi-Object accurately under ideal state of motion, and can not fully effective when moving target abnormality occurred, such as targets occlusion each other, a target disappears in the video or a new target appears, etc. So, this paper introduce matching matrix algorithm to Multi-Object Tracking approach which combines predicting based on Kalman filter and Feature matching to solve the problem of target tracking under abnormal situations. Experiment results show that this approach can achieve basically matching and tracking of Multi-Object under abnormal situations.