Optical Flow Based Violence Detection in Video Surveillance

摘要

Video Surveillance has become an issue of utmost importance due to rising crime and violence rate in the world. Many algorithms exist to analyze the behavior of the moving objects. In this paper, an approach to detect any violent behavior in videos using optical flow algorithms. The SDHA 2000 dataset is used to analyze foreground behavior. The main task in violence detection is to detect moving objects and classify their behavior using mean of intensity as violent or non-violent.