Research And Application On Object Shadow Elimination Algorithm In Complex Scene

With the rapid development of society and social demand, analysis of video surveillance is becoming a hot spot of research. However, object detection and tracking is the basis of video surveillance work, therefore, a stable and robust object detection tracking algorithm is particularly important. In the actual scene, object detection and tracking faces many practical problems, for example moving object shadow, object occlusion and so on. In this paper, we proposed some new methods to solve these problems, specifically, by improving existing detection and tracking algorithm. Finally, the new methods were tested and proven to be effective in an intelligent surveillance system.Firstly, the current popular detection algorithms such as frames subtraction method, background subtraction method and optical flow method, are analyzed in a comprehensive comparative way. Based on the analysis, we proposed a new method which combines ideas of the frames subtraction and the background subtraction methods. The experiment result shows that our method achieves the better performance than the frame subtraction method and the background subtraction method.Secondly, for the problem of target detection affected by moving target shadow in the image difference method, we proposed an algorithm for shadow elimination based on geometric projection. The idea of this algorithm is to utilize the direction difference between the moving targets and the shadow of moving object to identify the moving targets and their shadow. The experiments prove that our method is effective in shadow elimination and also is able to identify the moving targets and their shadow in different situations.Thirdly, a Mean Shift based shadow elimination algorithm is proposed, this algorithm updates the tracking area and tracking position by judging if the tracking area contains shadow or not. The experiment shows that our algorithm is better than the Mean Shift algorithm.Fourthly, an intelligent surveillance system is realized based on the proposed detection and tracking algorithm and shadow elimination algorithm. In the intelligent surveillance system, the shadow elimination algorithm is verified. The system not only has the functions of ordinary surveillance systems but also increases the function of box and dome linkage, which overcomes the disadvantages that large scene is unclear and small scene is incomplete. Meanwhile, the functions of detecting the moving target suddenly changing its behavior and moving target number statistics are added in this system. In addition, auxiliary function modules, e.g., calibration module, alert setting module and playback retrieving module are also developed.