Image Analysis and Recognition in Coal and Gangue Identification

Accurate identification of coal and gangue is pivotal for optimising mining operations, reducing environmental impact and improving resource efficiency. Traditional manual sorting is labour-intensive and prone to error, while conventional separation methods may generate pollution or require costly equipment. Advances in computer vision and machine learning have transformed this field by enabling automated, real-time analysis of material streams. High-resolution cameras, thermal and multispectral imagers, and radiation detectors capture rich data, which are processed through deep learning pipelines comprising segmentation, feature extraction and classification stages. Convolutional neural networks (CNNs) and their variants deliver robust performance under variable lighting, dust and motion conditions. Transfer learning and data augmentation address the challenge of limited annotated datasets, while segmentation models delineate gangue from coal at pixel level. Techniques such as near-infrared spectroscopy and gamma-ray detection complement visual analysis, offering spectral signatures that further distinguish materials. Together, these developments foster intelligent separation systems that can be integrated into shearers, conveyors and robotic handlers, with global significance for energy security and sustainable mining.

Research from Nature Portfolio

A study has demonstrated that natural gamma-ray emissions can be harnessed to distinguish coal from roof-rock in fully mechanised top coal caving. By analysing radiation intensity profiles from multiple mine sites, researchers established a quantitative link between gamma-ray counts and gangue content in real-time coal draws. An experimental detection system recorded radiative signals during caving operations, achieving rapid assessment of refuse content and enabling closed-loop control of extraction machinery. The work highlights the viability of integrating radiation sensors into automated coal-caving workflows, improving yield and reducing waste.

Image Analysis and Recognition in Coal and Gangue Identification publication trend

The graph below shows the total number of articles in image analysis and recognition in coal and gangue identification across all publications each year (not limited to Nature Index journals).

Technical terms

Convolutional Neural Network (CNN): A class of deep learning models designed for grid-like data, employing convolutional layers to automatically extract hierarchical spatial features from images.

SegFormer: A transformer-based semantic segmentation architecture that captures multi-scale context through hierarchical encoders and lightweight decoders, enabling efficient pixel-level classification.

YOLOv4: A one-stage object detector optimised for real-time applications, combining efficient backbone networks, feature pyramid enhancements and anchor box clustering for accurate and fast recognition.

Natural Gamma-Ray Detection: A non-invasive method that measures the spontaneous emission of gamma-rays from geological materials to infer composition differences between coal and rock.

Intersection over Union (IoU): A metric for evaluating segmentation and detection performance, defined as the ratio of the overlap between predicted and ground-truth regions to their union.

Data Augmentation: Techniques that generate new training samples through transformations such as rotation, scaling, noise addition and colour shifts to improve model generalisability under diverse conditions.

References

1. Research on coal-rock identification method and data augmentation algorithm of comprehensive working face based on FL-Segformer. International Journal of Coal Science & Technology (2024).
2. A near-infrared spectroscopy dataset of coal and coal-measure rock under diverse conditions. Scientific Data (2024).
3. Image Recognition of Coal and Coal Gangue Using a Convolutional Neural Network and Transfer Learning. Energies (2019).
4. Radiation characteristics of natural gamma-ray from coal and gangue for recognition in top coal caving. Scientific Reports (2018).
5. Coal/Gangue Recognition Using Convolutional Neural Networks and Thermal Images. IEEE Access (2020).
6. Multispectral Imaging: A New Solution for Identification of Coal and Gangue. IEEE Access (2019).