Enhancing Coal and Rock Classification in Adverse Mining Conditions with Transfer Learning

Accurate classification of coal and rock in mining operations is essential for maximizing resource extraction and maintaining safety. However, common image identification systems have significant difficulties in mining operations due to issues such as low visibility, dust, and inconsistent image quality. To address these limitations, this paper proposes an innovative strategy that uses Transfer Learning (TL) to improve coal and rock classification under adverse mining conditions. Image preprocessing techniques such as resizing and contrast enhancement are utilized to improve clarity and uniformity. Adaptive global threshold applied to precise segmentation, ensuring accurate delineation of coal and rock regions. Advanced shape and contour analysis is used to extract features, which capture the important qualities of coal and rock. A pretrained TL model, VGGNet, specially fine-tuned to recognize mining materials with higher precision with less labeled data. The proposed work is developed using Python as the implementation platform for developing and training models, achieving an accuracy of $96 \%$. This work indicates the approach’s capacity to execute proper classification under adverse mining conditions, providing an automated, scalable solution to assist operational efficiency and safety in mining operations.