The Research Of Abnormal Behavior Detection Based On Computer Vision In The Security Monitoring

The technology of testing abnormal behavior in computer vision is an important constituent part of realizing the security monitoring system intelligence. There are two ways achieving intelligent monitoring-the one based on DSP hardware architecture and the other one based on the Backend Database Server of computer. The former has been practically used in transportation management, automation of industrial production and military, but it is heard to be widely used in the security monitoring because of its high cost, poor compatibility, hard update and single testing function. This essay mainly does research in the latter one, which is based on the Backend Database Server of computer. The detecting function of abnormal behaviors can be directly applied to general monitoring computer, which cuts down the cost of hardware sharply, sets up the movement c haracteristic rules base of abnormal behavior and general matching algorithm of abnormal behavior detection by analyzing the common abnormal behaviors in security monitoring, and gives an alarm in time when discovering abnormal behaviors. It not only break s through the single and inflexible limits of present abnormal behavior detecting pattern, but also helps to set up the rules base of abnormal behavior in central data base, from which the detection and standard of abnormal behaviors can be updated, expanded and studied. This paper contributes to the following work.Firstly, the essay aims to set up the characteristic model for the common abnormal behaviors in security monitoring. It includes doing detailed research in different scenes of four abnormal behaviors- invasion, robbery, unusual stop and staff gather, analyzing the movement characteristic of the four abnormal behaviors, and finally setting up relevant characteristic model according to separate movement characters.Secondly, to extract and track the movement object in monitoring. It compares the advantages and disadvantages of four movement objects extraction algorithm, and designs the extracting ways of goals’ location coordinate and shape and center matching algorit hm.Thirdly, according to the security monitoring of abnormal behaviors, set up the relevant rules base and achieve the rules automatically matching. The essay separately analyzes the extraction algorithm of characteristic parameter about different movement object, sets up rules form of abnormal behaviors and finally achieves the process of how to match the abnormal behaviors.Finally, by means of using the designed Experiment Software, it will test the detection algorithm of abnormal behaviors- invasion, robbery, unusual stop and staff gather. The experiments results show that the detection algorithm can carry out the valid dist inguishing testing in experimental environment.