Research On Target Tracking And System Development In Intelligent Video Surveillance System

With the development of science and technology, the requirement of the societysystem is rising. Especially, the slogan of "Building a safe city" has been proposed inmajor cities across the country, which brings a strong vitality to the video surveillanceindustry. What’s more, the intelligent video surveillance, as a new stage of developmentof video surveillance, has more broad prospects for development and more importantsignificance for theoretical study.The paper aims at the research for target tracking and relevant technologies inintelligent video surveillance system. The paper mainly explores and solves thedifficulties, including the complex background and the light mutation, in the targettracking field, and improves our lab’s achievement in moving target detection. Finally,the paper realizes a series of functions in the system, including perimeter intrusiondetection, regional retrograde detection, tripwire detection, cooperation tracking and soon. The main contents and work in the paper are as follows:①The paper focuses on the research of target tracking algorithm, and a novelapproach has been proposed. The new algorithm is characterized by that it introducesthe conception of non-uniform quantization in HSV color space, which is more inaccordance with human vision perception, establishes the color-space histogram of thetarget, which takes the spatial distribution of target’s color, also establishes thecolor-space histogram of the background, which the paper uses of to eliminate thosepseudo foreground pixels, and is embedded in a particle filter framework toperformance target tracking, adapts the second order dynamic regression model in orderto estimate the target’s position from coarse to fine. Experimental results demonstratethat the new algorithm can not only solve the problem of the light mutation, but alsoenhance the robustness in target tracking.②The paper implements some improvement in moving target detection. The paperestablishes the background update model, uses self-define Morphological Operators,combines with the perimeter technology, and realizes the perimeter surveillance. Thenthe paper makes it become a reality that four videos were processed by the intelligentalgorithm simultaneously. After that, the paper proposes the regional retrogradedetection method and implemented it based on the moving target detection technology.Finally, the paper overcomes the bad real time performance with the application of target cross line model having been proposed in the paper.③The paper establishes a control model based on the PTZ camera, and applies thenovel algorithm in the system, then achieves the PTZ camera’s initiative tracking. Afterthat, the paper designs a target handover scheme in multiple cameras, and adapts thethought of parallel processing to combine target tracking with moving target detection,then successfully makes the cooperative tracking among the multiple cameras.④The paper achieves a returnable program module of intelligent video analysis,including perimeter intrusion detection, regional retrograde detection, tripwire detection,cooperation tracking and perimeter setting. Both systems have been achieved during theprocess of the research, which are running at Niu jiao tuo station on the3th line ofChongqing rail transit and Ji an garden in Jiangbei district of Chongqing respectivelynow.