Moving Object Tracking Method Based On Intelligent Video Surveillance Studies

Video target detection and tracking is the key technology in intelligent video surveillance system, it is a wide range of applications, such as in civilian and military fields including intelligent transportation systems, image coding and compression, the man-machine interface and navigation system and so on. With the continuous advancement of the science and technology, the vision technology of its core computer has been advanced, so that provides good technical support for the study of the target detection and tracking. Simultaneously, people’s living standards has been gradually improving, increasingly concerning about public safety, then this makes a trend of the development of intelligent video surveillance systems. Therefore, the significance of the research moving target detection and tracking is very important.In this paper, it detects and tracks the moving target in the case of the shooting video sequences of a single stationary camera. Then it focuses on studying the target tracking method. First, this paper introduces an overview of intelligent video surveillance system. Then it presents a detailed introduction to the theory and the specific methods of digital image processing used to this article. It also discussed in the target detection methods, and focused on the Gaussian mixture model for background subtraction method.Then, in the field of moving target tracking, it focused on Mean Shift tracking algorithm based on color feature. But when the moving target is rapidly moving, the effect of Mean Shift algorithm is not good. And when the moving target is hidden, that algorithm easily lost the moving tracking target. This paper improved the Mean Shift algorithm for above-mentioned problems. Namely it used the gray Markov model to predict the target in the center position of the current moment, this point served as the starting position of mean shift algorithm for searching the target. Due to using the gray Markov model to predict the target position in the current frame, it makes the starting point position of tracking iteration to closer to the actual position of the target, so that narrowing the search range and reducing the number of iterations, thus improving the speed and efficiency of the tracking.Finally, this paper used its library functions to stimulate and implement the tracking algorithm for the moving target of video sequences in Matlab platform simulation. Experimental results show that compared with Kalman filtering based Mean Shift algorithm, improved Mean Shift algorithm has greatly improved tracking accuracy and speed.