Abstract: Aiming at the delayed response and missing alarm problems of traditional fire detection, we propose a fire detection algorithm based on multi- sensor fusion. Firstly, a algorithm combined with motion detection and color detection is adopted to extract the candidate fire regions of visible light images and a threshold segmentation algorithm is used to extract the candidate fire regions of infrared images. Secondly,the algorithm extracts the dynamic and static characteristics of both candidate fire regions. Finally, the fusion fire identification based on SVM is adopted accroding to the acquired features and the real- time temperature and smoke concentration are detected by sensor simultaneously to achieve reliable and rapid fire detection. Experimental results indicate that the proposed algorithm has high accuracy and robustness in an acceptable time range.

Key words: multi-sensor fusion, fire detection, SVM