Related Experiment Video
Updated: May 13, 2025
12:08
From Voxels to Knowledge: A Practical Guide to the Segmentation of Complex Electron Microscopy 3D-Data
Published on: August 13, 2014
24.4K
Image segmentation algorithm based on improved YOLOv8 model and its application in underground coal and gangue
Lei Zhu1,2, Wenzhe Gu1, Chengyong Liu1
1China Coal Energy Research Institute Co., L td., Xi'an, Shaanxi, China.
Plos One
|May 9, 2025
Summary
This study introduces an improved YOLOv8 model for accurate coal and gangue image segmentation in intelligent mining. The enhanced model achieves superior accuracy and speed, outperforming previous versions for efficient underground applications.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Mining Engineering
Background:
- Coal and gangue recognition is crucial for intelligent coal mine construction.
- Existing segmentation algorithms suffer from low accuracy, missed detections, and slow speeds.
Purpose of the Study:
- To develop a fast and accurate coal gangue segmentation model.
- To improve upon the YOLOv8 architecture for enhanced performance.
Main Methods:
- Modified the YOLOv8 backbone by replacing standard convolutions with depthwise separable convolutions (DSC).
- Integrated the CBAM module into the neck network to improve feature differentiation.
- Expanded the dataset to 11,265 images and fine-tuned hyperparameters.
Main Results:
- Achieved 95.67% precision, 95.74% recall, 32.11 FPS, and 96.88% mAP.
- Demonstrated significant improvements over baseline YOLOv8 and other models (YOLOv3, YOLOv5, YOLOv7).
- Successfully applied the model to underground coal gangue image segmentation via transfer learning.
Conclusions:
- The improved YOLOv8 model offers a reliable and efficient solution for coal gangue image segmentation.
- The enhancements provide a substantial boost in accuracy and processing speed.
- The model shows practical viability for intelligent mining operations.
