基于小波变换的煤岩惰质组显微图像分类

Microscopic Image Classification of Inertinite of Coal Based on Wavelet Transform

在分析煤岩惰质组各组分显微图像特点的基础上,提出一种基于小波变换的煤岩惰质组显微组分自动分类方法。首先用离散小波变换对煤岩显微图像进行分解,根据分解所得的水平、垂直、对角3个方向小波系数设计描述其纹理属性的特征量;通过对特征量的分析,选取水平方向能偏、对角方向方差特征量以及图像的亮度比等特征量构成特征向量集;采用支持向量机对惰质组各组分进行分类,获得对其7类典型组分的较为理想的分类效果。与其他同类方法的实验结果比较表明,本文特征量选取与分类方案在分类效果上具有明显的优势。

On the basis of analyzing the characteristics of microscopic images of inertinite in coal, an automatic classificationmethod for macerals in inertinite of coal based on wavelet transform was proposed. Firstly, the coal microscopicimage was decomposed with discrete wavelet transform, features from coefficients corresponding to three directions（horizontal, vertical and diagonal） were designed, which were employed to characterize the texture feature. Aftercomprehensively analyzing these features, energy deviation of horizontal, variance of diagonal, as well as brightnessratio of the image were selected to build a feature set. Then, a class of support vector machine （SVM） based classifierswere constructed, and 7 macerals of inertinite were classified. By comparing the results of the proposed method withthose of others, it shows that the proposed method of feature selection and classifier has distinct advantage in classificationaccuracy .