A Video Abnormal Behavior Recognition Algorithm Based on Deep Learning

Abstract

Aiming at the problems of low detection accuracy and slow processing speed of crowd abnormal behavior in public places, this paper proposes a method of crowd abnormal behavior recognition based on three-dimensional spatiotemporal convolution neural network. Three dimensional convolution neural network is used to model the temporal and spatial characteristics of the captured video based on deep learning. Multiple convolution cores with variable temporal depth are used, and the temporal transition layer is redesigned combined with depth separable convolution layer. The temporal information in the input signal is used to classify abnormal behavior and normal behavior, and then the abnormal behavior is identified. Experimental results show that the accuracy of the proposed method can reach more than 91%, and the accuracy and generalization performance of the improved model are further improved than the existing algorithms.