Considering the unstability and high erroneous recognition with traditional methods for fire detection, this paper analyzes the characteristics of fire image and the familiar interference light source. Four criteria have been mentioned: the area increases gradually, the centroid changes random, the taper angles and the circularity. Fuzzy neural network(NN) is used in information fusion to distinguish the fire. Experiments show that Fuzzy neural network fire detection algorithm can distinguish the fire flame effectively and improve accuracy.