Research And Implementation On Security System Of Intelligent Behavior Analysis

In security scenes such as self-service bank, ticket hall and station square, targets’ abnormal behavior may cause danger to other targets’ safety of life and property. Traditional security systems only record and display events happened in the monitoring area, lack of the ability of dealing with unexpected abnormal behavior. Therefore, this paper focuses on the security scenes in which the surveillance camera is fixed and the vision is broad, and analyses 7 common abnormal behavior in the bank, aiming to design and implement the behavior analysis algorithm which can detect 7 abnormal behavior in the scenes, including region invasion, security-line crossing, wandering, trailing, crowding and ATM sticking.Mainly for the feature of 7 abnormal behavior, this paper regards the targets as a whole and studies the continuous status information which is the basis of behavior analysis. Therefore, this paper studies the application of relevant algorithms of target detection and target tracking in security scenes firstly. According to the characteristics of moving targets in the security scenes, this paper chooses combination of background-subtraction and frame-difference as the method of detecting target. Besides, with reference of the original background image, moving targets, stationary targets and false targets are classified. In addition, this paper selects combination of Kalman filter and Meanshift algorithm to track moving targets. According to the feature matching result in the zone whose center is determined by tracking algorithm, the target’s final location is fixed. Meanwhile, the failed tracking target will be reserved for several frames to solve the problem of target combination and separation.Based on the result of target detection and target tracking, this paper designs and implements the behavior analysis algorithm. The algorithm uses the targets’ continuous status information to match the abnormal behavior models. And then, the abnormal behavior which breaks the pre-set rules in the scenes is detected. Lastly, this paper designs and implements the intelligent system of behavior analysis and validates the above algorithms. The experimental data shows that, the behavior algorithms can detect 7 abnormal behavior in the scenes with high recognition accuracy and rapid calculation speed and have certain practical value.The research on this topic focuses on the need of security with active detection and hopes to promote the development of intelligent security industry.