Recognition of Furnace Flame Combustion Condition Based on Stochastic Model

The recognition of the furnace flame combustion condition is an important domain in the flame monitoring system. In recent years, the image processing technology is widely applied to detection of the flame combustion condition. The combustion in furnace, such as the combustion of the pulverized coal, is the complex, stochastic and unstable burning process. The flame images are static and include a lot of noise signals from different reasons; so the method based on the processing of the single image does not reflect the combustion in furnace exactly. In this paper, the stochastic model, that is, hidden Markov model (HMM) is introduced to achieve modeling and recognition of the flame combustion condition in furnace. It makes use of a hidden Markov process to characterize the image frames correlation in the image sequences and transition of image states where the model parameters are determined by the feature vectors of image frames that form the observation sequences. Experiments demonstrate that the HMM can better describe the flame combustion condition in the furnace so as to improve recognition performance