Research On Crow Abnormal Dection Method Based On Improved Vibe Algorithm

With the improvement of people's public safety consciousness,the crowd management in public places has important practical significance.At present,the crowd density estimation and abnormal behavior detection are tow important aspects of crowd management,and they have great practical significance. As to detecting the abnormalities of the crowd, such as the escape and riots of the crowd, the crowd motion information and the crowd density information are combined with to detect crowd abnormal behavior. The research work involves the foreground extraction algorithm,feature point extraction and tracking algorithm, the crowd motion features extraction algorithm, the crowd density features extraction algorithm and the crowd anomaly detection algorithm. The main work of this paper is as follows:(1) Based on the analysis of moving targets detection algorithms that commonly used, this paper studies and designs a improved vibe algorithm. The improved algorithm can solve the ghosting and shadow problems in the vibe algorithm and moving targets can be effectively detected under the light mutation. The test results show that the proposed algorithm can accurately detect the moving objects in the scene.(2) Based on the extraction of moving objects, this paper proposes to use the motion characteristic change rate to characterize the crowd motion information. Firstly, the feature points on the foreground image are extracted by pyramid LK optical flow to get the crowd motion vector, then the number of feature points in each interval of the video frame are counted to characterized the distribution of the crowd,finally these two characteristics are combined with to characterize the crowd motion information. Then the crowd density is classified according to the foreground pixel number of the current frame, if the crowd density is sparse,the number of features,the foreground pixel number, the edge feature and the perimeter area ratio are extracted to describe the crowd density information,else characteristics of gray level co-occurrence matrix of gray image and characteristics of gray level co - occurrence matrix of foreground image are used to describe the crowd density information.(3) Based on the study of the existing crowd anomaly detection methods, this paper combines the crowd motion information with the crowd density information to detect abnormal behavior. Using the standard video data set UMN data set to carried out the experiment compare with the method of population movement, the method of crowd movement and direction, and the method of the change rate of crowd motion.The experimental results verify the effectiveness of the proposed method.