Burning state recognition using CW-SSIM index evaluation of color flame images

Recently the burning state recognition of the rotary kiln based on flame images has attracted much attention. Most methods involve the image segmentation technique, usually requiring various complex algorithm. In this paper, a new method is proposed to identify the burning state by comparing the structural similarity of each wavelet subband of the flame image. Instead of the grayscale image, RGB model is adopted not only to achieve better recognition performance but also provide deeper understanding about the relation of burning state with different color components. Experiments show that the proposed method is robust to the small displacement and noise, and can effectively reduce the error in recognition when the flame images are contaminated by spatial translation or blurring.