Video Surveillance System Based on 3D Action Recognition

Abstract

Human action recognition using depth-map images from 3D camera for surveillance system is a promising alternative to the conventional 2D video based surveillance. We propose a security-event detection method based on body part classification and human action recognition for more effective video surveillance system. Experimental results show that the body part classification accuracy of 65.0% and security event detection accuracy of 0.878 were achieved for 9 security events.