Research On Algorithm Of Moving Object Tracking Methods In Intelligent Visual Surveillance System

With the rapid development of computer technology, networking technology and image processing technology, intelligent visual surveillance system has been more widely used. Intelligent visual surveillance system can track the abnormal target and analysis int monitor area by video image processing and computer vision. Moving target tracking algorithm is an important technology of intelligent visual surveillance technology and the basis of real-time analysis and understanding of the target. Over the last decade, a large number of researchers have proposed a number of high efficiency and high success rate of moving target tracking algorithm, but there are still many difficulties, such as time-consuming of tracking is still large, the success rate is more difficult meet the actual requirements and occlusion.This paper introduces the history of intelligent visual surveillance system, and analyzed the current situation at home and abroad, analyzes the difficult issues of tracking algorithm that is important technical of monitoring system. Research on the image pre-processing theory of the moving target tracking algorithm, including color images turn gray, mean filter, median filter, morphological image processing, presentation and analysis of the PSO(Particle Swarm Optimization). Moving target tracking algorithm carried out a detailed analysis, and two high-efficiency and high success rate of tracking algorithm was proposed.Firstly, this paper introduces the basic principle of tracking algorithm based on template matching and analyzes strategy of search the candidate by traversing pixels, thus the time of tracking is very high. The anti-blocking ability of the tracking algorithm is poor. To solve the above problem, combined with PSO and template matching tracking algorithm was proposed, using the particle swarm optimization alternatives instead of the search strategies of traversing pixels to improve the efficiency of tracking algorithm. Using PSO global search capability when the target is covered, and the tracking algorithm can quickly track the target when the target appear again.Secondly, introduces the basic principles of compressed sensing, feature extraction based on compressed sensing and target detection based on Bayesian classifier, this paper proposes a tracking algorithm that combines PSO and compressed sensing, which use compressed sensing theory to reduce the dimensional of feature, using Belize classifier for target detection,and using PSO to search the candidate template. Make full use of the ability of global search of PSO to search candidate target, effectively improve the anti-blocking performance of the algorithm.Both tracking algorithms are compared to a variety of the same type of algorithm by experiment on multiple groups with different types of interference of video sequences, Experiments show that the tracking algorithms proposed by this paper has increased significantly on the efficiency and tracking the success rate, and the ability of anti-occluding are all improved significantly. Finally, after analyzing the two algorithms, two algorithms can be appropriately applied in different scenarios.