Automated Video Surveillance System for Snatching Detection Using Heuristic Analysis of Moving Objects

Abstract

In this paper, we present an automated video surveillance system for snatching detection using a heuristic algorithm. Localization of moving targets in a video stream and human behavior estimation are key techniques in the proposed system. Some motion characteristics are determined from video streams. Using metric system of a feature vector, the system classifies video streams into criminal and non-criminal scenes. We use two kinds of action sequences for the training data set. After constructing the classifier, we use test sequences that are continuous video streams of human behavior consisting of several successive actions. The experimental results show that the system is useful in detecting criminal scenes of snatching by the discrimination of human behavior as criminal or non-criminal.