Research Of Pedestrian Detection And Object Tracking Of Intelligence Visual Surveillance System

Intelligent visual monitoring is a new industry of the commercial application. It has become a very active research direction in the fields of computer vision, and has shown great research value in science&technology and business. Except that, it also has a broad prospect application in many other areas such as medical diagnosis, virtual reality, video compression, and etc. Intelligent visual monitoring makes track identification by analysis and processing the video streaming recorded by camera. On the basis of it, the related behavior of the target can be analysis and judged, a warning would be send out and relevant information would be provided when abnormal situation occurs. Based on the previous studies, this paper classified and compared the current detection and tracking methods, and then puts forward a new algorithm for motion target detection and tracking.1)In the aspects of human motion target detection, four kinds of common motion target detection algorithms are deeply analyzed and compared, and realized by programming. Moreover, its prototype system data was analysis. This paper emphasizes on discussing the main content and the thought of the three background modeling algorithm (i.e., the single Gaussian model, Gaussian mixture model and nonparametric model), and then estimating the performance of the detection through experiment:At last, analysis the respective advantages and disadvantages and test results of them.2)In the aspects of human motion target tracking, the particle filter algorithm based on the color feature combined with geometric feature was put forward. Firstly, the basic principle of target tracking of video monitoring was discussed, and then introduces several typical tracking algorithms and their advantages and disadvantages, then mainly discusses the basic principle of particle filter algorithm and its application in target tracking. On the basis of it, to avoid the insufficient of present particle filter algorithm based on color features, the particle filter tracking algorithm based on target color characteristics combined with geometric feature was presented. The experimental results shows that the algorithm based on target color characteristics combined with geometric feature has better tracking performance, and even has relatively ideal tracking efficiency.