Study On Algorithm Of Moving Object Tracking In Video Surveillance System

Video surveillance system is a crucial issue in the field of computer vision,pattern recognition and artificial intelligence, and it has broad application foreground for safety monitoring, intelligent transportation and military navigation. Moving object detection technique is an important component of video surveillance system. The detection results directly affect the following components, like target location, identification and tracking, as well as comprehension and description of movement behavior. Large numbers of researchers devoted themselves in the area and have already achieved many progresses. The dissertation studies moving object detection algorithm and object tracking of video surveillance system based on the current conclusion.This paper mainly discusses the fundamental theories and key technologies of object detection and tracking for Intelligent Video Surveillance. The following topics are researched in details, such as detection and extraction of object with static background, the image segmentation measurement, tracking of moving objects and so on.In moving object detection aspects, background subtraction is studied. After analyses and studies on usual background modeling method, the novel gaussian mixture model based the traditional one is proposed. Traditional mixed-Gaussian model takes the same rate of study for the variance and the mean, which did not take into account the variance and mean of the different features and is easy to make into a small variance. To solve the problem, the new method takes a different update strategy to the mean value and the variance according to the different characteristics of them. With the use of dynamic methods on the update of variance study, and collection of historical data on the impact of the current frame, the variance study improves convergence speed and accuracy. Simulation results show that the method can not only improve the accuracy on background modeling to identify the target, and at the same time effectively suppress noise.As for target tracking, a new tracking algorithm is proposed based on the analysis and discussion on the traditional target tracking algorithm. It is of great importance to develop a robust and fast tracking algorithm in tracking system because of its inherent disadvantages such as weak observability and large initial errors. An improved algorithm referred to as the iterated divided difference filter is proposed based on the analysis and comparison of conventional nonlinear tracking algorithms. The algorithm predicts states by an iterated observation instead of a simple approximation, which reduces the error effects brought by observation linearization. And based on the concept of sequential importance sampling (SIS) and the use of Bayesian theory, particle filter is particularly useful in dealing with nonlinear and non-Gaussian problems. How to select an important density function to reduce the affection of the particle degeneration and improve the accuracy of the particle filter is one of the major problems. In this paper, a new particle filter is proposed that uses a iterated divided difference filter to generate the importance proposal distribution is proposed to decrease the posterior probability distribution estimation error, enhance tracking effect. The proposal distribution integrates the latest measurements into system state transition density so it can match the posterior density well.In practical application aspects, a moving object tracking system is designed to detect and track indoor moving people. System framework and the equipment composition are given, and a modular design is presented according to actual condition, including working principle of module and system design flow.