Study On Algorithm Of Moving Object Detection & Tracking In Video Surveillance System

In the modern time of pursuing for the high automatization, there is very strong need for a smart surveillance system, which can monitor the moving object in some surveillance scape automatically and consecutively. In most cases, this intelligent system has some functions, such as moving object detecting and tracking. Moving object detecting is chief issue in design of surveillance system, which extracts varying region in video stream from background, it affects latter tracking directly. Object tracking achieves moving contrail of monitored object, and provides credible data information for moving analysis. On the basis of referring to algorithms of relating technologies, the main work in this paper is as follows.On the research of the moving object detection, for the image sequences of simple background, this paper presents an algorithm of detecting moving objects based on feedback information. It combines information of last frame and background estimation to reconstruct background, and uses threshold method based on online Otsu method to pick up object. The results show that this algorithm combines the advantages of veracity and of runtime, and fit for fast detection. For the video of multi-mode scene, the author researches two algorithms of detection. One is moving object detecting based on Gaussian mixture model, which enhances speed of the method. the other is detection based on support vector regression, which improves precision of the algorithm.On the research of object tracking, firstly, we constructs Bayesian network model combining correlative information of several object features, and tracks using estimated matching probability. The results show that the algorithm is computationally efficient. Secondly, this paper researches the mean shift algori- thm based on object feature of color histogram, proposes an improved colorhistogram containing information of location and confidence, and achieves tracking by combining mean shift algorithm, which increases reliability of tracking results. Finally, this paper presents a framework assembling detection and tracking, to make sure detected region and tracked location more reliable.