基于纹理特征的回转窑熟料烧结状态分类

采用灰度共生矩阵方法,利用Fisher系数提取出最佳分类位置算子和纹理特征参数,通过对实际回转窑窑头熟料图像分析,发现位置算子为(5,-5)即距离为5、方向为45°下的灰度共生矩阵对应的和平均、逆差距、差异熵、对比度、差方差和熵这6个参数具有较好的区分度,其表面纹理特征能客观地反映其烧结程度,并通过基于C4.5算法实现了过烧、欠烧和正常烧结3种不同状态下的熟料纹理分类,其精度达到了95.65%.同时结合实际工况对熟料纹理进行了分析,给出了各自的变化特点.

Classification of Sintered Clinker in Rotary Kiln Based on Texture Features

The texture analysis of the clinker image based on the grey-level co-occurrence matrix was proposed to predict the clinker''s sintered quality. The best position operator and feature sets of the grey-level co-occurrence matrix were extracted with Fisher coefficient. Then, these reduced features were applied by C4.5 to classify these clinker images into three categories, over-sintered, less-sintered and normal-sintered. The experiment results have shown that six texture features, which are SA, IDM, DE, Contrast, DV and Entropy, of the grey-level co-occurrence matrix under the position operator (5,-5) have the highest degree of discriminability, and the classification accuracy reaches 95.65% with C4.5 classifier. Finally, the difference between these three kinds of clinker textures was summarized.