TLE-UNet-based image segmentation and feature extraction of water-cooling-induced complex cracks in high-temperature rock masses

Authors

FU Tianyu, School of Emergency Management and Safety Engineering, China University of Mining and Technology (Beijing), Beijing 100083, ChinaFollow
HU Mangu, School of Emergency Management and Safety Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China
ZHANG Xiaojun, School of Emergency Management and Safety Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China
YANG Xiaobin, School of Emergency Management and Safety Engineering, China University of Mining and Technology (Beijing), Beijing 100083, ChinaFollow
LYU Fu, Software College, Liaoning Technical University, Huludao 125105, China; Basic Teaching Department, Liaoning Technical University, Huludao 125105, China
LYU Xiangfeng, Hebei Key Laboratory of Collapse Prevention, University of Science and Technology Beijing, Beijing 100083, China
PENG Lei, School of Mechanics and Civil Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China

Abstract

Objective and Method Water cooling-induced macroscopic surface cracks in high-temperature rock masses are characterized by fine scales, pronounced length variations, and severe class imbalance in images, posing significant challenges for image segmentation. To address these challenges, this study proposed TLE-UNet, a semantic segmentation network for rock crack images. First, through thermal treatment and uniaxial compression tests of granite specimens, crack images under varying temperatures were acquired. Then, using a self-developed Lite Edge Fusion module, high-resolution shallow features were finely aligned with deep semantic features at varying scales. Furthermore, edge detection and channel-attention mechanisms were combined to enhance crack boundary perception. Additionally, this study designed an auxiliary decoding head module, EWS Head, to achieve multi-scale fusion of features from shallow- and mid-level encoders through edge hint and lightweight texture enhancement. During training, the Tversky loss function was incorporated as auxiliary supervision to improve discrimination of fine cracks and alleviate the adverse effects of background-dominant class imbalance. Generally, the proposed architecture effectively improved the image segmentation accuracy and boundary continuity of fine cracks while preserving multi-scale semantic representation. Results and Conclusions Compared to the baseline U-Net, TLE-UNet increased the crack-related IoU from 38.32% to 46.32% and pixel accuracy from 45.34% to 65.70%. In contrast to mainstream segmentation models, TLE-UNet outperformed UNet++, Attention U-Net, and DeepLabv3+ models in crack-related IoU, demonstrating a higher capability to identify fine cracks. Ablation experiments verified the effectiveness of various modules in the TLE-UNet algorithm. Heatmap-based visualization analysis further indicates that TLE-UNet can more precisely highlight crack edges. Finally, based on the image segmentation results and relevant calculation methods, this study derived the geometric parameters of cracks, including their lengths, maximum widths, and mean widths. The analysis of these parameters across multiple images reveals a high consistency between TLE-UNet-derived results and actual measurements, further verifying the effectiveness of the TLE-UNet algorithm in crack formation extraction.

Keywords

rock cracks, improved U-Net, crack information extraction, edge enhancement, multi-scale feature fusion, class imbalance

DOI

10.12363/issn.1001-1986.25.07.0546

Recommended Citation

FU Tianyu, HU Mangu, ZHANG Xiaojun, et al. (2026) "TLE-UNet-based image segmentation and feature extraction of water-cooling-induced complex cracks in high-temperature rock masses," Coal Geology & Exploration: Vol. 54: Iss. 2, Article 16.
DOI: 10.12363/issn.1001-1986.25.07.0546
Available at: https://cge.researchcommons.org/journal/vol54/iss2/16

Reference

[1] WANG Xiaoyu,ZHONG Zhen,XIE Xinghua,et al. A nonlinear flow model and criterion for 3D natural fracture intersections:Effects of surface roughness and mechanical aperture[J]. International Journal of Rock Mechanics and Mining Sciences,2025,193:106169.

[2] 李轶惠,许振浩,潘东东,等. 基于数字图像的隧道岩体裂隙智能识别与参数提取方法[J]. 应用基础与工程科学学报,2023,31(6):1427−1443

LI Yihui,XU Zhenhao,PAN Dongdong,et al. An intelligent identification and parameter extraction method for rockfractures based on digital images[J]. Journal of Basic Science and Engineering,2023,31(6):1427−1443

[3] 张农,袁钰鑫,韩昌良,等. 基于Mask R–CNN的煤矿巷道掘进迎头裂隙检测与定位算法[J]. 采矿与安全工程学报,2023,40(5):925−932

ZHANG Nong,YUAN Yuxin,HAN Changliang,et al. Research on crack detection and localization algorithm for advancing face in coalmine roadways based on Mask R–CNN[J]. Journal of Mining and Safety Engineering,2023,40(5):925−932

[4] LIU Ziqi,ZHENG Lulin,ZUO Yujun,et al. Investigation of three–dimensional model reconstruction and fractal characteristics of crack propagation in jointed sandstone[J]. Geomechanics and Geophysics for Geo–Energy and Geo–Resources,2024,10(1):75.

[5] 雷瑞德,顾清恒,胡超,等. 裂隙砂岩声发射信号特征及破裂前兆识别研究[J]. 岩土力学,2025,46(7):2023−2038

LEI Ruide,GU Qingheng,HU Chao,et al. Acoustic emission signal characteristics and precursory recognition of rock failure in fractured sandstone[J]. Rock and Soil Mechanics,2025,46(7):2023−2038

[6] HE Yuxiang,TAN Yu,YANG Mingshan,et al. Accurate prediction of discontinuous crack paths in random porous media via a generative deep learning model[J]. Proceedings of the National Academy of Science,2024,121(40):e2413462121.

[7] 金解放,廖强强,陈萌,等. 高水压高应力裂隙岩石动态强度特性试验研究[J]. 岩石力学与工程学报,2025,44(5):1133−1145

JIN Jiefang,LIAO Qiangqiang,CHEN Meng,et al. Experimental study on dynamic strength characteristics of fractured rocks under high water pressure and high stress[J]. Chinese Journal of Rock Mechanics and Engineering,2025,44(5):1133−1145

[8] 王登科,房禹,魏建平,等. 基于深度学习的煤岩Micro–CT裂隙智能提取与应用[J]. 煤炭学报,2024,49(8):3439−3452

WANG Dengke,FANG Yu,WEI Jianping,et al. Intelligent extraction of Micro–CT fissures in coal based on deep learning and its application[J]. Journal of China Coal Society,2024,49(8):3439−3452

[9] 王娟,吴禄源,袁超. 不同裂隙几何特征岩石强度智能预测研究[J]. 金属矿山,2024(8):45−52

WANG Juan,WU Luyuan,YUAN Chao. Research on intelligent prediction of rock strength with different crack geometric characteristics[J]. Metal Mine,2024(8):45−52

[10] 江松,章睿,崔智翔,等. 半监督学习模型下的露天矿高陡岩质边坡裂隙识别研究[J]. 安全与环境学报,2025,25(10):3821−3829

JIANG Song,ZHANG Rui,CUI Zhixiang,et al. Intelligent identification of high–steep rock slope fractures in open–pit mines using semi–supervised learning model[J]. Journal of Safety and Environment,2025,25(10):3821−3829

[11] WU Jin,WU Shunchuan,SUN Beibei. An adaptive methodology for rock mass fracture image enhancement with generalized gamma correction[J]. The Visual Computer,2024,40(8):5201−5217.

[12] WANG Shaofeng,YIN Jiangjiang,PI Zizi,et al. Automatic detection and characterization of discontinuity traces and rock fragment size distribution using a digital image processing method[J]. Measurement,2024,228:114343.

[13] HE Changdi,SADEGHPOUR H,SHI Yongxiang,et al. Mapping distribution of fractures and minerals in rock samples using Res–VGG–UNet and threshold segmentation methods[J]. Computers and Geotechnics,2024,175:106675.

[14] 李元海,徐晓华,朱鸿鹄,等. 基于计算机视觉的岩石裂隙识别表征与软件研制[J]. 岩土工程学报,2024,46(3):459−469

LI Yuanhai,XU Xiaohua,ZHU Honghu,et al. Identification and characterization of rock fractures based on computer vision and software development[J]. Chinese Journal of Geotechnical Engineering,2024,46(3):459−469

[15] 肖福坤,刘欢欢,单磊. 基于连通性阈值分割的煤岩裂隙识别方法[J]. 工矿自动化,2024,50(8):127−134

XIAO Fukun,LIU Huanhuan,SHAN Lei. Coal rock crack recognition method based on connectivity threshold segmentation[J]. Journal of Mine Automation,2024,50(8):127−134

[16] MOHAN A,POOBAL S. Crack detection using image processing:A critical review and analysis[J]. Alexandria Engineering Journal,2018,57(2):787−798.

[17] LI Xiaobin,LI Bingke,LIU Fangzhou,et al. Advances in the application of deep learning methods to digital rock technology[J]. Advances in Geo–Energy Research,2023,8(1):5−18.

[18] XU Haoran,TANG Shibin,WANG Jia,et al. Rock fracture identification algorithm based on the confidence score and non–maximum suppression[J]. Bulletin of Engineering Geology and the Environment,2024,83(6):213.

[19] 王有志,李缘平,罗甦元,等. 基于UNet语义分割模型的矿山采集钻孔裂隙识别研究[J]. 世界有色金属,2024(23):217−219

WANG Youzhi,LI Yuanping,LUO Suyuan,et al. Research on borehole fracture identification based on UNet semantic segmentation model[J]. World Nonferrous Metals,2024(23):217−219

[20] 胡咤咤,张寻,金毅,等. 基于μCT和深度学习的煤裂隙智能提取方法[J]. 煤田地质与勘探,2025,53(2):55−66

HU Zhazha,ZHANG Xun,JIN Yi,et al. A method for intelligent information extraction of coal fractures based on μCT and deep learning[J]. Coal Geology & Exploration,2025,53(2):55−66

[21] LI Mingze,CHEN Ming,LU Wenbo,et al. Automatic extraction and quantitative analysis of characteristics from complex fractures on rock surfaces via deep learning[J]. International Journal of Rock Mechanics and Mining Sciences,2025,187:106038.

[22] LI Ning,XIONG Zihao,WANG Liguan,et al. A multi–attention deep learning network for intelligent identification of rock mass fracture in mines[J]. Results in Engineering,2025,26:105023.

[23] LEI Jian,FAN Yufei. Rock CT image fracture segmentation based on convolutional neural networks[J]. Rock Mechanics and Rock Engineering,2024,57(8):5883−5898.

[24] TIAN Yajie,WANG Daigang,XIA Jing,et al. Digital rock modeling of deformed multi–scale media in deep hydrocarbon reservoirs based on in– situ stress–loading CT imaging and U–Net deep learning[J]. Marine and Petroleum Geology,2025,171:107177.

[25] 赵伟,张文康,王涛,等. 基于关键裂隙识别的离散裂隙网络骨架提取研究[J]. 中国矿业,2024,33(增刊1):409−413

ZHAO Wei,ZHANG Wenkang,WANG Tao,et al. Study on skeleton extraction of discrete fracture networks based on key fracture identification[J]. China Mining Magazine,2024,33(Sup.1):409−413

[26] LI Wenxi,LI Quangui,QIAN Yanan,et al. Structural properties and failure characteristics of granite after thermal treatment and water cooling[J]. Geomechanics and Geophysics for Geo–Energy and Geo–Resources,2023,9(1):171.

[27] RONNEBERGER O，FISCHER P，BROX T. U–Net：Convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer–Assisted Intervention. Cham：Springer，2015：234–241.

[28] SALEHI S S M，ERDOGMUS D，GHOLIPOUR A. Tversky loss function for image segmentation using 3D fully convolutional deep networks[M]//WANG Q，SHI Y，SUK H I，et al. Machine learning in medical imaging. Cham：Springer，2017.

[29] AL–HUDA Z,PENG Bo,ALGBURI R N A,et al. Asymmetric dual–decoder–U–Net for pavement crack semantic segmentation[J]. Automation in Construction,2023,156:105138.

[30] NGUYEN Q D,THAI H T. Crack segmentation of imbalanced data:The role of loss functions[J]. Engineering Structures,2023,297:116988.

[31] LI Mingyao,SUN Hefeng,PENG Lei,et al. Laboratory investigation on physical and mechanical behaviors of granite after heating and different cooling rates[J]. Energy,2024,302:131718.