查看文章

Abstract Analysis of combustion quality from flame images in thermal and gas turbine power plants is of great importance in the domain of image processing and is the primary objective in detection, recognition and understanding of combustion condition. In this work, soft sensors using feed forward neural network trained with Back Propagation Algorithm (BPA) is used for flame image classification. The basic idea behind this work uses the information from the color of the flame images which dependent on the combustion quality. The first step is to define a feature vector for each flame image including 7 feature elements, which are the brightness of flame, the area of the high temperature flame, the brightness of high temperature flame, the rate of area of the high temperature flame, the flame centroid respectively. The quality of the captured images is enhanced using curvelet transform. The concept of object (flame feature) recognition and classification of the flame image is carried out to measure the temperature from the flame color and the flue gas emissions from the flame color. The samples including 51 flame images, parts of which are used to train and test the model. Finally, the entire samples are recognized and classified. Experiments prove this method to be effective for classification of flame images.