Image Tracking Technology Based On Intelligent Video Surveillance Research And Implementation

With the continuous expansion of city in our country, the city safety management has become an important issue. Image tracking technology based on intelligent video surveillance is one of research hot topic in the field of computer vision in recent years, due to its good application prospect, many researchers at home and abroad has also been around its conducted in-depth research, and many achievements have been made. But with the development of image tracking technology and the people more and more high to the requirement of intelligent tracking, image tracking technology is also encountered many problems, for example, when the target tracking, the change of the light, shade, changes in the form of objects are stringent requirements are put forward for target tracking. Research on tracking algorithm based on intelligent video monitoring has also been a hot topic in the field of computer vision, the researchers have proposed many effective algorithms such as: Mean Shift algorithm, kalman filtering algorithm, particle filter algorithm, etc. This paper also studied the problems, and the specific studies several algorithm, in the end, the fusion of multiple information tracking algorithm is put forward the concrete work is as follows:1、mainly studies the algorithm are: Mean Shift Mean Shift algorithm, Cam Shift algorithm, kalman filtering algorithm, particle filter algorithm, and experimental verification of these algorithms, analyzes the advantages and disadvantages of various algorithms through the experimental results. Provides theoretical basis for the proposed algorithm in this paper.2、this paper proposes a fusion feature of particle filter algorithm. Because the particle filter algorithm of background noise is not linear, gaussian limit, provided a prerequisite for the fusion of multiple features. Experiments to extract the target of double color and shape features of target tracking, solved the complex environment of a single appearance suddenly reduce or disappear, tracking lost caused by the algorithm. After the single feature weaken suddenly, through the fuzzy logic rules given another feature higher weights, algorithm can still use the other features continue to track. 3、put forward a kind of fusion particle filter algorithm of Mean Shift thought. Conventional particle filter algorithm to calculate the process, with the passage of time, the invalid particles increases, the invalid or weight low particle will greatly increase the amount of calculation, using the Mean Shift Mean Shift, the resampling particle cluster, make the particles along the direction of the Mean Shift vector convergence, it improves the computational efficiency, thus increasing the real-time tracking.