A Method to Distinguish the Fire and Flickering Vehicle Light

The vehicle light in the tunnel or other relatively dark and limited spaces is always mistaken as fire by most of existed video-based fire recognition systems. In this paper we propose a novel method to distinguish the fire and the flickering vehicle light for avoiding the undesirable situation. In addition to ordinary motion and color detection, we make use of several fire features such as layered feature, circumference and area so that we can process and eliminate light affection from original video frames. Our experiment results demonstrate that the method is robust to discriminate flickering vehicle light from fire so it reduces the false alarm rate.