Fire detection based on infrared video image

Flame detection and recognition is a kernel issue in the automatic fire surveillance system. An effective algorithm was put forward to recognize flame object in infrared video sequence. Otsu thresholding based on particle swarm optimization algorithm is used for segmentation of the image, after that the fire colored objects will be picked up. Next, a double-deck fuzzy system is employed to analyze the difference of objects' height and gray scale distribution, and then the membership degree is given to the flame. The experimental results prove that the method can effectively extract flame object from complex background.