The Design And Implement Of Pedestrian Anmoaly Detection System Based On Motion Information

With the city reform in recent years,the number of urban population increased,safety awareness has been enhanced,the government promote the construction of safe city vigorously,but the surveillance system is widely used in the market has been unable to meet the demand of the society for artificial intelligence.At present,the main function of monitoring video surveillance is recording and saving the video,to determine whether there is abnormal behavior and the investigation of abnormal events after the observation by security personnel.If abnormal behavior occurs,the monitoring system can automatically detect and notify security personnel,so that not only can save a lot of time for the staff,but also to prevent the occurrence of accidents.Therefore,in the development of abnormal behavior detection system,there is a high research value and commercial market.This paper focuses on the monitoring system,the extraction and analysis of the principle of motion information,the rule based approach to determine the abnormal behavior.In this paper,the software system framework and hardware structure are designed by constructing intelligent monitoring system.The software of anomaly detection system includes four parts:moving object detection,object classification,pedestrian tracking and abnormal behavior detection.Firstly,the paper introduces the algorithm of moving object detection,compares the advantages and disadvantages of several commonly used algorithms,and selects the detection algorithm with better detection effect and real-time performance.Combining motion detection and pedestrian detection algorithm to achieve a series of detection methods.This Paper used head-shoulder detection method to solve the occlusion problem between pedestrians,at the same time,it also reduces the false detection rate of pedestrian detection,which can provide a basis for accurate tracking database.In order to accurately track pedestrians,this paper uses the Kalman filter to predict the pedestrian.In order to solve the multi-target tracking problem,this paper combines the Kalman tracking algorithm with the Hungarian matching algorithm to realize the multi target tracking.In order to enable the system to track the autonomy,by the combination of the motion detection algorithm and the tracking algorithm,the tracking automatic initialization and the problem of occlusion and target loss can be effectively solved.Finally,the abnormal behavior is defined by the rule based method,and the typical abnormal behavior detection is realized.The explanation for the wandering abnormal behavior,is proposed based on trajectory component test track.Considering the individual behavior and group behavior has the very big difference,this paper introduces the entropy value detection method based on the social forces to detect abnormal behavior of groups.