Fire smoke detection in train carriages based on multiple features fusion

Owing to complex environment of the train carriage, in order to improve the accuracy of the video based on smoke detection algorithms, eliminate the similar object of smoke movement and reduce the false detection rate caused by environment light changes, a modified smoke detection algorithm based on motion, color and attenuation is presented. This algorithm can effectively reduce the interference of background motion and the changing scene illumination, and eliminate the shadow phenomenon. The algorithm consists of three parts: the detection of smoke movement, extracting the characteristics of smoke color and multiple features fusion. First, the movement of the pixels is detected by the motion segmentation. Then the model of normalized RGB color space is introduced for eliminating the interference of similar smoke area and reducing the impacts of changes in illumination effectively.