Crowd abnormity detection and positioning system and method based on time recurrent neural network

Abstract

The invention discloses a crowd abnormity detection and positioning system and a crowd abnormity detection and positioning method based on a time recurrent neural network. The method comprises the steps of carrying out gridding division on a scene on the basis of preprocessing acquired sample data, and dividing a video fragment consisting of n frames into a plurality of time-space blocks; then constructing a multi-scale optical flow histogram and Gabor wavelet texture features in each time-space block; vectoring and combining the features in the multiple time-space blocks, finding out a relation between a space dimension and a time dimension by taking the video fragment as a time sequence through respectively using hidden nodes and feedback nodes of the time recurrent neural network, and training a time recurrent neural network model capable of detecting a long dependency relation; finally detecting and positioning a crowd abnormity by the model. The method is higher in instantaneity and accuracy, and can detect abnormities activated by a small quantity of individuals or a large quantity of individuals.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a system and method for crowd anomaly detection and positioning based on a time recursive neural network. Background technique [0002] Group event detection refers to the detection of events composed of a large number of targets to find abnormalities or specific events of interest to help people make decisions quickly. Anomaly detection and localization is an important aspect in crowd event detection. On the one hand, with the development and progress of society, the population continues to grow rapidly, and with urbanization, crowd phenomena are becoming more and more frequent, which leaves a hidden danger for the safety of public spaces. Especially in contemporary times, terrorist incidents occur frequently, which seriously endanger the lives and property safety of the masses. In order to detect them as soon as possible and take corresponding countermeasures as soon as possible...

AI Technical Summary

Problems solved by technology

[0005] At present, the existing patent technologies for crowd abnormal behavior detection are mainly divided into several categories, among which the China Institute of Metrology is the representative, the patent application number is CN201210223375, and the invention title is "A method for detecting abnormal behavior of group crowds in video surveillance". It is carried out on the basis of target detection and tracking. For crowds, due to the complexity of crowd scenes, there are a lot of occlusions in crowd scenes, and it is difficult to achieve good results in target detection and tracking; Representative, the patent application number is CN201210065523, the invention name is "the detection method of crowd abnormal events based on LBP weighted social force model", which does not take into account the time dependence; the representative of China Metrology Institute, the application number is CN201310494769, the invention The patent titled "Crowd Abnormal Behavior Identification Method Based on SIFT Flow and Hidden Conditional Random Field" takes into account long-term dependencies, but it is different from our method. The hidden conditional random field hopes to describe the probability distribution behind the dynamics of the crowd , while the time recurrent neural network is to directly fit the data, or directly find the discriminant interface that distinguishes different sequence categories. For crowd dynamics, it is difficult to find a better probability distribution; and Sun Yat-sen University is the representative Yes, the application number is CN201410312813, and the patent title is "A Method for Crowd Abnormal Behavior Detection Using a Single-Class Serialization Model", which still does not consider the time dependence, but uses the Hidden Markov Model for smoothing at the end deal with

And most of the methods do not take into account the relationship between different locations, and for the abnormal situation caused by an individual, the relationship between different locations is helpful to detect anomalies; most of the methods also do not This makes it difficult to detect certain anomalies, such as a car driving down a road at a slow speed, or a cow appearing in a crowd, given the shape characteristics

Images