Crowd exceptional event detection method, electronic equipment and storage medium

Abstract

The invention discloses a crowd exceptional event detection method. The method comprises the steps that a first image and a second image are acquired; multiple optical flow points of the first image are calculated according to the first image and the second image; an optical flow point adjacency matrix is calculated according to a horizontal coordinate relation and a vertical coordinate relation among all the optical flow points; the optical flow points are divided into corresponding region sets according to the optical flow point adjacency matrix and clustering conditions; each region set isdivided into at least one region subset according to a distance relation among the optical flow points; all the region subsets are corrected according to historical grouping information, and correctedsubsets are obtained; and whether a crowd exceptional event occurs is judged according to an average motion distance of the optical flow points in each corrected subset. According to the method, optical flow is adopted to represent motion features of a scene, cluster grouping is performed on the optical flow points, then optical flow information in each group is calculated to perform statisticalanalysis on motion conditions of the corresponding group, therefore, an exceptional event occurring in a crowd can be detected, and the method can be suitable for different scenes and has good instantaneity.

Description

technical field [0001] The invention relates to the field of video applications, in particular to a method for detecting crowd abnormal events, electronic equipment and storage media. Background technique [0002] With the increase of population and the diversity of human activities, crowded scenes have become more and more frequent, and the frequency of accidents and injuries caused by large-scale crowds has also increased, which has brought huge challenges to public management and public safety. Therefore, it is necessary to monitor the crowd, detect and alarm abnormal behaviors, so as to avoid the loss of people and property. [0003] The prior art provides the following three abnormal behavior detection methods for crowds: [0004] 1) Model-based crowd abnormal behavior detection methods: such as fluid dynamic models and hybrid dynamic texture models. Since the modeling is related to the scene, this model will fail in some scenes and lacks wide applicability. [0005]...

AI Technical Summary

Problems solved by technology

This method is not effective for noisy data, and there is a problem that the clustering effect is poor when the density difference is large.

Images