Improved Meanshift Method And Its Application In Intelligent Viedo Surveillance System

As people's safety consciousness is enhancing, more and more places are using monitoring equipment, since the intelligent video surveillance of moving object detection and tracking is an important question in the field of computer vision, and what's more, intelligent video surveillance of moving object detection and tracking algorithm design is the core of the intelligent video surveillance system, the study of key techniques of intelligent video surveillance system and the improvment the system are of vital significance.This paper has mainly studied the video monitoring of intelligent motion detection and tracking algorithm, and has summarised the corresponding solutions. The main researches are as follows:On the part of target detection, firstly it discussed light flow method, background difference method, frame difference method and the motion energy method etc. As light flow method and background difference method are very complex, they can't meet the requirement of real-time processing without strong hardware support; Frame differential method and background difference method is simple, but they could not make comparison on all pixels of two pictures. As the system expenses does not allow so, and basing on multiple matching feature extraction pixel target detection methods. And through comparing frame differential method, background difference method and algorithm of testing effect, it shows that the pixel-based times matching algorithm can complete and extract moving targets quickly, laying a firm foundation for the following study.On the aspect of target tracking, based on computer vision tracking technology research, this essay probes into the expression of the current computer vision in target tracking technology and the criteria of feature selection. Aiming at the Mean Shift algorithm shortcoming that it can't track fast-moving object, it raises the use of mean shift algorithm and the pixel matching method of characteristics of moving target tracking algorithm. Taking feature matching method to extract interesting target area, then it bases on the iteration algorithm to search for mean shift starting point. Furthermore, the two experiments have verified classical algorithm and the improved algorithm, showing that this algorithm can efficiently and rapidly track moving target, and can solve problems in the process of tracking error accumulatively. On the realization part, this essay combines system to realize intelligent video surveillance in target detection and tracking algorithm, using VC++2005 for development platform and OpenCV development library, it designs an intelligent video surveillance system. By testing and tracking moving target in shipment and block, it proves that the prediction of this essay is feasible and practical.Finally, it summarizes the whole research and points out the direction of the future work.