Computer Vision Based Real-time Fire Detection Method

摘要

The video based detection of fire hazards is one of the significant developments in recent years. However, the conventional vision based flame detection algorithms suffer with the issues of low detection rate and high false-alarm rate. A novel real time video processing method is proposed in this paper that detects the flames by combining the flame motion detection technique with color clues and flame flicker to reach the final detection. The proposed fire detection framework builds an efficient background model by optimizing the selective background update to extract the fire-like moving regions in the video frames. Furthermore, a YCbCr color space based analysis technique is applied to improve the fire-pixel classification. Finally, a flame flicker identification algorithm based on the statistical frequencies is used to confirm whether it is fire region. The experimental results show that the proposed algorithm has high detection rate and low false-alarm rate, it is accurate, robust and effective compared to the existing methods.