Crowd Anomaly Detection Method for Weighted Convolutional Autoencoded Long Short-Term Memory Networks

Abstract

The invention discloses a weighted convolutional autoencoder-long short-term memory network (Weighted convolutional autoencoder-long short-term memory network, WCAE-LSTM network) method for anomaly detection, dedicated to learning the generation model of moving pedestrians for anomaly detection and Positioning to ensure public safety. The present invention proposes a novel dual-channel framework, uses WCAE-LSTM network to learn the original data channel and the generation mode of the corresponding optical flow channel respectively, reconstructs the data, and performs anomaly detection based on the reconstruction error. In addition, for the complex background problem, the present invention proposes to separate the sparse foreground from the low-rank background by using block robust principal component analysis, and design a weighted Euclidean loss function according to the obtained background information, thereby suppressing the background noise. The WCAE-LSTM network designed by the present invention can not only detect anomalies from a global perspective, but also roughly locate abnormal regions from a local perspective, and finally achieve robustness to abnormal events by jointly considering the results of global-local anomaly analysis and optical flow anomaly analysis , Accurate detection.

Description

technical field [0001] The invention belongs to the field of computer vision, specifically relates to the field of intelligent monitoring, in particular to a method for detecting crowd abnormalities using a weighted convolutional self-encoded long-short-term memory network. Background technique [0002] With the development of the field of video surveillance, more and more surveillance cameras are used in public places to ensure public safety. However, a large amount of surveillance video data not only has too much redundant information, it is also a severe challenge for surveillance personnel, and video analysis is time-consuming and tedious. Therefore, developing a system that can automatically detect video anomalies plays a vital role in alleviating human and financial resources. Public safety will be further assured if continuous or potential anomalies can be accurately detected. [0003] Most of the traditional video anomaly detection methods detect anomalies through ...

AI Technical Summary

Problems solved by technology

However, these generated models are susceptible to complex backgrounds and cannot reconstruct normal behavior well

Meanwhile, methods that detect anomalies based on the reconstruction error of the entire frame may ignore anomalous behavior in a small region

In addition, Generative Adversarial Network (GAN) can also detect anomalies, however, GAN models are difficult to train and cannot capture the differences between moving objects well

In addition, the model is still unable to resolve the effects of complex backgrounds and small area anomalies

Images