Intelligent Video Surveillance Target Tracking Algorithm Research And Application

With the continuous development of Computer Vision,the VISS(Vedio Intelligent Surveillance System)has drawn more and more attention.The Target Detection and the Target Tracking are the most basic technologies in the VISS.These two methods to research of VISS is particularly important.Therefore,this article focuses on the detection and tracking of moving targets in video sequences.Specific work arrangements are as follows:At the process of researching target detection,the traditional target detection method is firstly studied and analyzed,including the method of Background Subtraction,the method of Interframe Difference and the method of Optical Flow.After comparing these detection algorithms,an improved target detection method is proposed.Firstly,the background is modeled by using the method of mixed Gaussian model,and obtain the foreground by using the method of Background Subtraction,and the foreground obtained by the five-frame difference method is AND-operated.Finally,the morphology of the foreground was obtained,and the complete moving object was obtained.The simulation experiment was carried out on the improved detection algorithm.The experimental results show that the proposed method effectively solves the problems of voids and edge blurring,and the detection effect is much better.In the target of tracking phase,aiming at the problem that the traditional Meanshift algorithm fails when the target is severely obstructed or completely occluded,an improved algorithm based on Meanshift combined with Kalman filter is proposed.The value of the Bhattacharyya is used to determine if the target is occluded.If the target is not occluded,according to the information of past moving targets,a possible target position is predicted by using the Kalman filter.In order to obtain a relatively accurate target position,the Meanshift algorithm around the predicted target point Iterative search.On the contrary,if the moving target is occluded,the use of Kalman filter algorithm to predict the target path,and the target tracking results will be regarded as the predicted value;Secondly,when the moving object's color is similar to the color of background or other interfering object's color,the Meanshift algorithm cannot realize the accurate tracking.In view of this problem,we propose to use the Edge Direction Histogram to describe the characteristics of the object,Edge Orientation Histogram contains transport Moving target spatial texture and shape information to solve the problem of color interference.