An Algorithm For Crowd Abnormal Behavior Recognition In High Definition Videos

Crowd abnormal behavior recognition based on computer vision is an important research topic in the filed of public safely monitoring,anti terrorism,and early warning of mass incident.Aiming at the problem of slow recognition speed,this paper proposes an algorithm for detecting abnormal behavior based on motion vector.In this method,the social force model is used to calculate the interaction force based on the motion vector information of the crowd;then,the direction and intensity of the interaction force are extracted,and the direction and intensity of the movement of the crowd are classified according to the histogram;finally,the simulation experiments show that the method can be used to calculate the crowd interaction,effective area divided into motion,accurate to judge the abnormal changes of population movement,good robustness,fast detection speed,compared with the traditional optical flow method,frame processing speed increase by about 30%.Aiming at the problem of non regular local abnormal target extraction,this paper proposes an algorithm for detecting abnormal behavior based on double circular scan statistics.In this method,the compensation factor of the optical flow vector field is proposed,and the optical flow vector is compensated,and the feature of the moving object is extracted;and improved doubly periodic scanning method and scanning region in the likelihood ratio test statistic as abnormal behavior discriminant feature values by the first round of the suspect region based scanning can of different crowd motion scene,irregular abnormal region realize identification and extraction,the simplified calculation parameters,reduce the amount of computation;finally,simulation results show that the method can extract the abnormal behavior of the crowd.