摘要
Abstract
The frequent occurrence of incidents endangering social and public safety makes it significant to study the abnormal behavior of crowds in video surveillance for restoring public order and ensuring public safety.Due to the diverse scenarios involved in video surveillance,the complex environment affects the accurate detection of abnormal behavior of the crowds.Therefore,in order to improve the effect of crowd abnormal behavior detection in video surveillance,a crowd abnormal behavior detection algorithm based on improved optical flow method is proposed.An improved single Gaussian model is used to extract corner points within the crowd video frame in video surveillance and take them as the feature points.On the basis of the improved optical flow method,the motion speed and direction of the feature points are calculated and the effective feature points are extracted,so as to obtain the crowd moving object images.The product of the direction entropy,amplitude entropy and average velocity of the optical flow points in the object image of crowd movement is calculated to determine the degree of motion chaos.The degrees of motion chaos are contrasted and analyzed,and the thresholds are set,so as to complete the detection of crowd abnormal behavior.The experimental results show that the algorithm can effectively extract corner points and moving object images within crowd video frames,detect crowd abnormal behavior accurately,and has good performance in crowd abnormal behavior detection in video surveillance.关键词
改进光流法/视频监控/人群异常行为检测/单高斯模型/特征点/方向熵/幅值熵/运动混乱度Key words
improved optical flow method/video surveillance/crowd abnormal behavior detection/single Gaussian model/feature point/direction entropy/amplitude entropy/degree of motion chaos分类
电子信息工程
