Fire smoke recognition based on image entropy
Pan Chaofeng;Yang Shusen;Chen Ning;School of Energy and Power Engineering,Jiangsu University of Science and Technology;
Fire smoke recognition is of great significance for monitoring fire disasters. This paper proposes a new smoke movement recognition method which is based on image entropy to realize the detection of fire smoke.Through the theoretical analysis of free diffusion model of smoke,the conclusion is drawn that the diffusion of smoke is a kind of entropy-increasing movement. By introducing the concept of "image entropy",which is normally used in signature analysis,to capture the features of entropy changing,we can realize the detection of fire smoke. Through the experiment and analysis and by using Fourier transform on the changes of image entropy,a conclusion is drawn that interference movement and smoke movement are two kinds of deferent movement pattern. In this way,we can realize the recognition of smoke. Unlike researchers who study the characteristics of smoke image to detect smoke,we analyze the thermodynamic properties and fluid mechanic properties of smoke.We conclude that the smoke movement pattern is different from other rigid movement pattern. By using the image entropy as the core parameter,we get a simple and high efficient algorithm to detect the smoke. Experiment shows that the detection algorithm has much less interference from the surroundings. It is especially proper to be used in complex environment. It can also be used as an evaluation algorithm of fire disaster environment.