Detection of fire based on multi-sensor fusion

When using a single sensor to detect fire, a high false alarm can be caused owing to the low reliability of the single sensor. Consequently, the multi-sensors fire detection methods are proposed. In this paper, three different type sensors are used to collect temperature, smoke concentration and CO concentration features. The information gain rate of each feature is computed, and then the attribute with maximum information gain rate is chosen as the current attribute node, which is the root node of the decision tree. At the same time, the training sets are divided into subnets according to the possible value of the root node attribute, subsequently execute the above step recursively. Finally a C4.5 classifier is designed to integrate those features for fire detection. Experimental results demonstrate the effectiveness of this approach.