基于圆形LBP均匀模式的煤镜质组显微组分纹理分析

Texture Analysis of Vitrinite Macerals Based on the Uniform Pattern of Round LBP

依据煤不同显微组分的纹理特点,分别采用不同尺度下圆形LBP及LBP均匀模式获取煤显微图像的LBP信息值,构建包含能量、对比度、局部平稳性等特征量的特征集对煤显微组分进行表征,并以此对镜质体中各类别纹理数据进行分析。结果表明:构建的纹理特征量对于镜质体中不同类别具有较好的可分性;圆形LBP中尺度的选择对有效特征量的提取有较大影响;LBP均匀模式在保持特征数据有效性的同时,可明显减少LBP模式数,从而减少计算量。文中结果可为煤显微组分的自动分类与识别提供依据。

According to texture characteristics of different coal macerals, round Local Binary Pattern （LBP） and LBP uniform pattern with different scales were employed to obtain LBP information of coal microscopic images. Features such as energy, contrast, local stationarity based on LBP were selected to characterize the coal macerals, and thus texture features in each subclass of vitrinite were analyzed. Results show that, these selected features are separable for different subclasses of vitrinite; scales of round LBP have an impact on effective feature extraction; LBP uniform pattern can reduce the LBP numbers and simplify the calculation while maintaining the effectiveness of feature data. These results can provide basis for automatic classification and recognition of coal macerals.