Rotary kiln flame image segmentation based on FCM and gabor wavelet based texture coarseness

Presents an improved segmentation algorithm for flame image of rotary kiln burning zone, based on Gabor wavelet based texture coarseness and Fuzzy C-MEANS (FCM) cluster algorithm. At first, analyses the flame image in detail, divides it into four areas (flame area, material area, illuminated area, background area) by expert experiences and applies threshold segmentation in order to get rid of background area from flame image. Then carries out general FCM algorithm with grey level of grey flame image and gets a preliminary membership matrix, calculates texture coarseness of the flame image by Gabor wavelet texture filters. At last performs Defuzzification operation on the above membership matrix based on the differences of texture coarseness between material area and flame area to exactly separate material area and flame area from the original flame image. Experiments have proven the validity of this new segmentation algorithm for flame image.