A multi-attention deep learning network for intelligent identification of rock mass fracture in mines

Recognition of rock mass fractures holds significant application value in geological engineering, rock mechanics, and predicting geological hazards. This research introduces an intelligent segmentation method for rock mass fractures based on an enhanced U-Net network to address the limitations in th...

Full description

Saved in:
Bibliographic Details
Main Authors: Ning Li, Zihao Xiong, Liguan Wang, Bibo Dai, Shugang Zhao, Haiwang Ye, Qizhou Wang
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:Results in Engineering
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2590123025010989
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Recognition of rock mass fractures holds significant application value in geological engineering, rock mechanics, and predicting geological hazards. This research introduces an intelligent segmentation method for rock mass fractures based on an enhanced U-Net network to address the limitations in the accuracy of rock mass fracture image recognition observed in prior studies. This method integrates spatial attention, channel attention, and self-attention mechanisms into the encoder structure of the U-Net network by employing visual geometry group 16 (VGG16) as the backbone network, which improves its anti-interference capabilities and feature extraction efficiency during the rock mass fracture identification process. Experimental results from a self-compiled mine rock mass fracture dataset indicated that the proposed approach achieves superior recognition quality compared to other advanced networks, validating its effectiveness. In addition, the mean pixel accuracy (mPA) and mean intersection over union (mIoU) achieved by the proposed method are 89.94 % and 83.55 % , respectively, reflecting a 7.02 % and 6.26 % improvement over the basic Resnet-Unet network, indicating superior segmentation accuracy. The findings of this research present an innovative approach and methodology for the intelligent identification of rock mass fractures in mining applications.
ISSN:2590-1230