Skip to Main Content
Since the texture is an important feature of smoke, a novel method of texture analysis is proposed for real-time fire smoke detection. The texture analysis is based on gray level co-occurrence matrices (GLCM) and can distinguish smoke features from other none fire disturbances. For the realization of real-time fire detection, block processing technique is adopted and the computation of texture features is done to every block of image. Neural network is used to classify smoke texture features from none-smoke features and the fire alarm trigger is set according to the total smoke blocks in one frame. The accuracy of the method is discussed as a function of frames in the end.