The Study And Implement Of The Segmentation Of The Flame Images Of The Rotary Kiln

The rotary kiln is a modern advanced device of complex, large-scale, sequential production. In the process of produce alumina in rotary kiln, the detection and control of the temperature of burnt belt is important work, the temperature of burnt belt is the key factor to determine whether the process of alumina is high quality, high yield, low consumption. There are many temperature information of the burnt belt in the flame image of the rotary kiln, the detection of burnt belt temperature by analyzing the flame images is the newest detection technology. The detection of the burnt belt based on flame images depends on the gray character of the materiel and flame section in the images. So we need segment the two sections for computing the burnt belt temperature. Because there is the infection of smog, coal powder, clinker and flame in the flame image, and the boundary of every object sections is blur, for the segmentation of materiel section and flame section there are many difficulties. The purpose of this paper is to resolve the severely noise and blur boundary of object sections, and to exactly segment materiel section and flame section from the flame images.In this paper, I provide the auto-segment materiel section and flame section method of flame image of the rotary kiln. In this method, at first, I switch the color flame image into the gray image of YCbCr color systemic; remove noise and smooth for the image using multi-direction window detection and smooth method among images; auto-segment using improved ostu threshold method, K-average clustering method and dividing ridge method; at last, combine the results of the three methods, compare and value the final combined result images and the manually segmented images. This method is used in 70 practical color flame images of the rotary kiln. So the validity of the image segmentation of these complex images has been proved.