A real-time action detection system for surveillance videos using template matching

In past two decades, computer vision techniques applied to prevent criminals for surveillance videos receive much attention. Because hands-up actions in surveillance videos likely implicate that someone commits robbery, hands-up becomes a good clue to notify robbery happening. This paper presents a real-time action detection system to recognize hands-up actions in surveillance videos. We develop a template matching method to apply to the foreground extracted by background subtraction. According to the matching results, a multi-frame decision delivers detection results of hands-up actions in a video. The experimental results indicate that our proposed system is effective and efficient to detect hands-up actions, and achieves a good performance of 90% recall and 89% precision rates. © 2015, Ubiquitous International. All rights reserved.