Automatic estimation of coal rank and maceral composition from reflectogram

Article Type

Research Article

Publication Title

International Journal of Coal Preparation and Utilization

Abstract

The physical and chemical properties of coal depend majorly on maceral composition and rank. Existing methodology to estimate coal rank and maceral composition is tedious and time-consuming and fully dependent on human expertise. We propose an image processing-based method that automatically determines the coal rank and its maceral composition quickly. The geological age of the coal samples used in our experiment is Lower to Upper Permian and Carboniferous. Grey values of different reflectance standards are used for calibrating the images. A statistical technique approximates the gray values of pixels representing maceral groups using a Gaussian distribution. Then inflection points are detected mathematically on the cumulative distribution function derived from the approximated Gaussian distribution of pixel values. These inflection points separate pixels representing vitrinite, inertinite, mineral, etc. About 2471 images of eight different coal samples from Permo-Carboniferous age are assessed. The result shows almost 94% accuracy in rank estimation. For maceral composition, about 95%, 85% and 76% accuracies in terms of the area under the curve measure, are reported for vitrinite, inertinite and mineral matters, respectively. Our method is unsupervised, precise, and repeatable, and has better industrial applicability compared to expert-driven maceral analysis.

DOI

10.1080/19392699.2025.2538200

Publication Date

1-1-2025

Link to Full Text

Share

COinS