Violence Detection Algorithm Based on Local Spatio-temporal Features and Optical Flow

Abstract

A new algorithm of detecting the violence behavior based on local spatio-temporal features and optical flow is introduced in this paper. It differs from the existing works of violence detection in what concern the data representation. Based on the physical contact detection algorithm which retains the frame containing possible violence, this paper combines Harris 3D spatio-temporal interest point detector and optical flow method to make up for defects of optical flow method in huge computation and slow processing. Experiments show that the proposed algorithm is of high to detect aggressive actions regardless of the context and number of human involved.