An Early Fire Detection Method Based on Smoke Texture Analysis and Discrimination

Abstract

Texture is an important property of fire smoke, which is a significant signal for early fire detection. This paper describes a method of analyzing the texture of fire smoke combining two innovative texture analysis tools, Wavelet Analysis and Gray Level Cooccurrence Matrices (GLCM). Tree-Structured Wavelet transform is used to represent the textural images and GLCM are used to compute the different scales of the wavelet transform and to extract the features of fire-smoke texture. The smoke texture and the non-smoke texture are classified by neural network classifier. The discrimination performance is related to the quantity of input vectors.