Research Of Moving Object Detection’s Algorithm In Video Surveillance

Intelligent monitoring system is an important component of a community to achieve efficient management and moving object detection is an important step in the implementation of this system. This paper mainly research the lack of the two typical technologies named moving object extraction and object matching. Also it proposes some improvement comparing with the effect of the past methods. At last, this paper puts the research results into the practical projects.Vibe(Visual Background Extractor) algorithm runs very quickly and can restrain the impact of noise, but it does has flaws. For example the several classical related papers didn’t propose the efficient method to remove shadows, and couldn’t remove the ‘ghost’ areas quickly. On the issues mentioned above this paper proposes a fusion of Vibe algorithm and frame differential method. Using a new frame differential method, it gives two improvements : 1. a new way to detect and remove the shadow using the brightness information only in gray space is proposed; 2. It can remove the ‘ghost’ area quickly, via using the frame differential feature skillfully. Finally this paper improve the result with morphological method and show the result of experimental verification.Object matching methods can be divided into two methods like matching method based on global features and matching method based on local features. However the way using a single matching method will be relatively limited. This paper presents a fusion of an improved color distribution method based on color features and local SIFT feature and acquire some certain effect. Besides, it makes a corresponding improvement to the case that objects divided to several blocks.Finally, relying on target intrusion detection system in a specific area of a museum, we put this method into a practical application. And we get some achievement.