Researching on human abnormal behavior recognition algorithm

Abstract

Aiming at the defects in the performance of three-dimensional convolutional residual network in surveillance video data sets, this paper proposes a three-dimensional convolutional residual network structure based on the improvement, redesigns the three-dimensional convolution kernel, and decomposes the three-dimensional convolution into a combination of one-dimensional convolution and two-dimensional convolution to increase the nonlinearity of the network and improve the accuracy of network recognition. At the same time, on the basis of the reconstruction of the convolution, The optical flow features were added for fusion to strengthen the influence of human abnormal behavior recognition results. Experimental results show that the improved network recognition accuracy is effectively improved.