煤岩显微图像假边界对显微组分组自动识别的影响

在利用图像分析技术进行煤岩自动化检测过程中，发现煤岩显微图像中各种成分的边界表现为"镶边"状的灰度过渡区，过渡区像素点灰度不能反映其两侧显微成分的真实灰度。为了分析这种假边界对煤岩显微组分组识别检测的影响，通过对全国大量煤岩样品所采集显微图像分析的基础上，对各种相邻成分间假边界的特征、成因进行了分析。假边界一般多呈环状、条带状"镶边"形态，呈现出与相邻成分灰度相衔接的灰度斜坡特征，假边界灰度斜坡宽度与煤岩显微图像中各种成分的组合类型相关。假边界的形成是由于煤中各种成分的硬度及韧性差异引起突起高低不同，成像时在不同成分的边界呈现灰度斜坡。选取10个不同变质阶段代表性煤样，利用Prewitt算子对样本假边界进行边缘检测提取，提取的假边界过渡区像素占比整体煤颗粒图像的10%~27%；并将测定结果与国内资深煤岩专家人工鉴定的标准结果进行比对，结果表明，剔除假边界后测定的镜质组、惰质组、壳质组的极差远低于未剔除假边界的极差，更为接近人工鉴定结果。 

Effect of false boundary of microscopic image on automatic identification of maceral group

In automatic detection of coal petrology by image analysis technology, it is found that the boundary of various components in the microscopic image of coal petrology demonstrates a "rim" of gray transition zone, and the pixel of this rim cannot reflect the true gray level of the components on both sides. In order to analyze its influence on the identification and detection of macerals, the characteristics and causes of the false boundary between various adjacent components were studied with analysis of microscopic images collected from a large number of coal samples in China. Generally, the false boundary is ring-shaped or strip-like rim, showing the gray ramp characteristics in-between the gray levels of adjacent components. The width of gray ramp of the false boundary is related to the combination of various components in microscopic images. False boundary develops by different heights of the relief due to different hardness and toughness of various components in the coal. The gray ramp appears at the boundary of different components during imaging. 10 representative coal samples from different rank in China were selected, and the edge detection of the sample false boundary was extracted by Prewitt operator. The pixel of the false boundary transition zone was 10%-27% of the image of total coal particle. Compared with the standard results of manual identification done by domestic authoritative experts in coal petrology field, the results show that the deviation of the vitrinite, inertinite and liptinite group with the removal of false boundary is much lower than that with false boundary, the former is closer to manual identification result.