Related Experiment Video
Updated: Jun 25, 2026
08:27
Image Recognition and Parameter Analysis of Concrete Vibration State Based on Support Vector Machine
Published on: January 5, 2024
Research on Coal Gangue Image Recognition Method Based on EMAM-YOLO
Ying Jia1,2,3,4, Baoshan Li5, Yongxing Du5
1School of Mining and Coal, Inner Mongolia University of Science and Technology, No.7 Arding Street, Baotou 014010, China.
ACS Omega
|May 25, 2026
Summary
A new EMAM-YOLO model improves coal-gangue detection in mines, achieving 81.17% mAP with high speed and fewer parameters. This robust model enhances accuracy in challenging underground conditions.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Mining Engineering
Background:
- Underground coal mines present complex environments with challenges like uneven illumination, coal dust, and varied coal-gangue shapes.
- Existing detection models often suffer from low accuracy and poor robustness in these conditions.
Purpose of the Study:
- To develop a real-time coal-gangue image detection model (EMAM-YOLO) that addresses the limitations of current methods in complex mining environments.
- To enhance detection accuracy, robustness, and real-time performance for intelligent coal washing applications.
Main Methods:
- Utilized YOLOv12n as the baseline and integrated EfficientNetV1 for backbone reconstruction, reducing parameters while maintaining feature representation.
- Introduced a Multiscale Attention Feature Pyramid Network (MAFPN) for improved cross-scale feature interaction.
- Designed an Adaptive Spatial Feature Fusion detection head (Detect_ASFF) for optimized multiscale feature fusion and enhanced localization accuracy.
- Incorporated a multiscale channel attention (MCA) mechanism to focus on critical feature channels.
Main Results:
- The EMAM-YOLO model achieved a mean average precision (mAP50-95) of 81.17%, a 5.04% improvement over the baseline YOLOv12n.
- The model has 2.59 million parameters and a detection speed of 69.89 FPS, demonstrating a balance between accuracy and real-time performance.
- Outperformed Faster R-CNN, SSD, YOLOv8n, and YOLOv10n in robustness and detection accuracy under simulated complex conditions.
Conclusions:
- The proposed EMAM-YOLO model offers significant improvements in coal-gangue detection accuracy and robustness for underground mining environments.
- The novel Detect_ASFF head and the synergistic combination of modules provide effective technical support for intelligent coal washing.
- The model's performance highlights its potential for practical application in enhancing mining automation and efficiency.