The Study Of Segmentation Methods For Flame Images Of Alumina Rotary Kiln

Alumina rotary kiln is a large, complex and continuous thermal production equipment，currentlythe temperature of calcining zone is the manipulated variable of most sintering control strategies.Flame image contains a great amount of information related to temperature, the calcining zonetemperature detection by using image processing methods is a technology based on the analysis offlame zone, material zone and the overall gray-scale features, thus it is particularly necessary to realizeaccurate segmentation for the above-mentioned feature zones. The flame image’s segmentation is achallenging task, due to the factors such as smoke, materials and flame. In this paper, we carried outin-depth study and the research achievements are mainly as follows:(1) We explored the application of Contourlet in the segmentation of flame zone and material zone,due to the distinct texture feature differences between those two zones. First to carry out blockdecomposition for flame image using Countourlet, then calculated the variance of sub-band coefficientsas the feature vector which was the representation of image texture features, finally the segmentationresults were obtained by using FCM clustering method whose input was the feature vector. Thesimulation proved that this segmentation method met the precision requirements, but the algorithm wascomplex and took long processing time.(2) We presented a method based on local entropy for the segmentation of flame zone and materialzone combined with the concept of information entropy. Based on the texture differences betweenflame zone and material zone, this method firstly calculated the local entropy of image to obtain thetexture image, then the edge of material zone was extracted through thresholding segmentation, andfinally the segmentation was completed by morphology operations, which included edge linking andimage filling. Simulation results showed that the method was of validity and timeliness.