Abnormal Behavior Detection Method Of Video Surveillance In The Pedestrian Fights

The quantity of surveillance cameras increases abruptly in public places and communities with the development of society and the improvement of safety awareness,and a massive surveillance video data is produced therefore.Thus,the conventional manual monitoring costs higher and higher,getting lower and lower efficiency.Therefore,the study and application of intelligent monitoring system is very important.Abnormal behavior detection is the key and hot research in the intelligent monitoring system.Analysis of video content via computer vision technology for rapid detection of abnormal behavior of pedestrians.It has important significance on controlling the situation and curbing risky behavior.Therefore,the research focus of this paper is how to detect abnormal behavior of pedestrians.At present,the common abnormal behavior detection method is modeling of the target trajectory[27][28].This method has made good results in hovering retrograde,cross-border and restricted testing,however,it does not apply to fights scenes.We find that optical flow characteristics of the target can well reflect its campaign details.Therefore,this paper proposes a method for pedestrian fights detection based on optical flow energy.First,using background subtraction method to detect moving objects in video.Second,computing optical flow information in target area by using Lucas-Kanade,and define the intensity and direction of optical flow vector.Finally,the direction of optical flow is divided into eight section evenly.Determining whether or not the abnormal behavior is occurred by calculating the change in optical flow intensity consecutive frames.Experimental results show that the method can accurately detect fight abnormal behavior of pedestrians indoor and outdoor.The algorithm has high robustness and it can accommodate many scenes.