Anomaly detection method for sudden crowd based on instantaneous energy

Abstract

The invention discloses an anomaly detection method of a sudden crowd based on instantaneous energy, the method of extracting motion feature points based on pixel statistical analysis is used to extract the feature points of the image motion region, and the kinetic energy is used as the basic energy feature of the crowd, and the instantaneous energy feature of the crowd motion state is extracted based on the kinetic energy difference between two consecutive frames of the energy block, so as to detect the human sudden crowd anomaly. The anomaly detection method of the sudden crowd can not onlyeffectively reduce the number of obtaining motion feature points so as to avoid subsequent processing of some meaningless feature points, but also can reflect the texture feature of the foreground motion region and the distribution of the crowd density, which makes the extracted motion feature points more representative and uniform. Whether in the abnormal occurrence of the response speed or the description of the movement of the crowd state changes, the above-mentioned outburst crowd anomaly detection methods have shown outstanding advantages.

Description

technical field [0001] The invention relates to the field of intelligent video monitoring, in particular to a sudden crowd abnormality detection method based on instantaneous energy. Background technique [0002] With the increasingly tense public security situation, automatic detection of abnormal events in crowd scenes will be of great significance to public security. Therefore, intelligent video surveillance has become an important research direction in the field of computer vision. At present, the difficulty of this research direction mainly focuses on how to automatically obtain the effective characteristic information of individuals or groups, and make correct understanding and judgment on the characteristic information. In order to try to solve the problem of automatic identification of crowd abnormalities in surveillance video, many methods based on computer vision have been proposed at home and abroad. [0003] At present, the methods commonly used to obtain the fe...

AI Technical Summary

Problems solved by technology

The corner detection method can obtain the feature points of the foreground motion area more comprehensively, but the feature points obtained for some local areas are too concentrated, which affects the extraction of subsequent motion features

The uniform distribution method is to artificially mark the particle points evenly in the plane of the foreground motion area, and consider these particle points as the feature points of the crowd motion image. Although this method can obtain the feature points of the foreground motion area evenly, it cannot reflect the foreground motion area. The texture changes in the medium make the extraction of feature points rough

[0004] Anomaly detection algorithm based on kinetic energy features, but the response time of kinetic energy features is slow, and the description of sudden anomaly features is not significant

The sudden anomaly detection algorithm based on the acceleration feature, the calculation amount of the acceleration feature is too large, it cannot meet the real-time requirements, and the robustness is poor

When the crowd is abnormal, the most obvious feature of the crowd state change is the sudden change of the crowd's kinetic energy, which makes the energy of the crowd jump from one energy level to another. Research

Images