Double-branch abnormity detection method based on crowd behavior priori knowledge

Abstract

The invention provides a double-branch abnormity detection method based on crowd beharivor priori knowledge. The method comprises the steps of extracting interaction information of crowds in a video by utilizing a social force model; learning abnormal scores for different time slices in the video by using a multi-instance learning method; capturing global dependence of the video features by utilizing an attention model; and combining the original video with the crowd interaction information video corresponding to the original video by using a double-branch model. According to the method, priori information of abnormal behavior judgment of human beings is fully considered; a sufficient number of normal and abnormal samples are used for learning normal and abnormal modes of crowd behaviors,so that anomaly detection can recognize the crowd behaviors in a video on a certain semantic level, the problem of performance loss caused by insufficient samples and background interference of crowdsin the video can be well solved and adapted, and the method has higher robustness; the method does not need a data label which is accurate to the fragment level, and even if the training object is the fragment of the video, only the label of the video level is needed.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and specifically relates to a double-branch anomaly detection method based on prior knowledge of crowd behavior, in particular to abnormal detection focusing on crowd abnormal behavior under a monitoring camera. Background technique [0002] Surveillance cameras are increasingly being used in public places, such as streets, intersections, banks and shopping malls, where there is a lot of traffic. However, the relevant administrative law enforcement agency's ability to detect abnormalities in the surveillance video has not kept up, resulting in the inability to fully utilize the resources of the surveillance camera, and there are obvious defects in its use. It is also very unrealistic for people to observe the surveillance video in real time, because the number of surveillance cameras in our country is already very large. It is not only costly to rely on human resources to monitor the surv...

AI Technical Summary

Problems solved by technology

The advantage of C3D over CNN is that it can extract the space-time features of the video to a certain extent, but in terms of anomaly detection tasks, C3D cannot accurately extract the behavior interaction information of the crowd in the video

Existing methods do not make good use of this prior knowledge, which is also likely to cause the model performance to fail to meet expectations

Images