Vision-based fire detection algorithm using optical flow

Abstract

Recently automatic video surveillance in the digital video recording CCTV system is rapidly becoming one of the most accepted security system. The dangerous situations such as forest fire, flood, and terrorism are increasing, and cause serious casualty and property loss. In this paper, we particularly focus on the fire detection system in video. The proposed block-based fire detection algorithm consists of three basic steps. In the first step, we find motion vector utilizing optical flow and distinguish the suspect as fire region. In the second step, we conduct chromatic detection based on Lab color space. In the final step, we employ motion information for detecting the correct fire block by using the characteristics that fire goes almost upward. Experimental results show that the proposed method yields good performance for fire detection.