Human Detection and Tracking Using HOG Feature and Particle Filter in Video Surveillance System

Video surveillance system has recently attracted much attention in various fields for monitoring and ensuring security. One of its promising applications is in crowd control to maintain the general security in public places. However, the problem of video surveillance systems is the required continuous manual monitoring especially for crime deterrence. In order to assist the security monitoring the live surveillance systems, intelligent target detection and tracking techniques can send a warning signal to the monitoring officers automatically. Towards this end, in this paper, we propose an innovative method to detect and track a target person in a crowded area using the individual’s features. In the proposed method, to realize automatic detection and tracking we combine Histogram of Oriented Gradient (HOG) feature detection with a particle filter. The HOG feature is applied for the description of contour detection for the person, while the particle filter is used for tracking the targets using skin and clothes color based features. We have developed the evaluation system implementing our proposed method. From the experimental results, we have achieved high accuracy detection rate and tracked the specific target precisely.