Video smoke detection based on semitransparent properties

Video smoke detection has been widely researched for its advantages on fire alarm. But false alarm is still an outstanding issue. In this paper, a novel semitransparent properties based on algorithm of video smoke detection is proposed. The original is that the marginal area of smoke is semitransparent. Firstly, in order to get semitransparent region, the background image is matched with the current image in the video which is restored using the haze image optical model. Then, region of interest (ROI) is found by a region growing method. Finally, fuzzy clustering analysis of boundary in ROI is counted to reduce false alarms. The experimental results indicate that the proposed approach can work effectively with non-false alarm.