Analysis Of Flame Detection Technique Based On Image Processing

Furnace flame consists of lots of burning information, including flame position, flame shape, flame brightness, flame flicker frequency and temperature field distribution, which can fully reflect the combustion of the fuel. If parameters such as temperature field distribution and flame status can be detected rapidly and accurately, the power utility can be operated economicly and prevented from some accidents. In recent years, according to the upgrading of industrial image technique and computer technology, digital image processing also has developed gradually in the furnace flame detection, which is especially important for efficiecy and safety.This paper studies the theory and hardware system of flame detection, describing the flame temperature field testing method in detail, especially focus on monochrome and two-color measuring principle. Because of interference and noise, denoising, image enhancement, edge detection and image segmentation should be done as pretreatment. We discuss why and how to pretreat in the paper. Some commonly used flame charateristics including average gray level, variance, entropy, energy and so on will be then extracted, and artificial neural network method is used to identify the combusiton state. All the methods mentioned above are simulated by MATLAB toolbox.