基于深度学习的砂岩样品三维重构研究

朱淳; 徐佳俊; 孙文斌; 何昌迪; 王晓

doi:10.20174/j.JUSE.2025.02.06

地下空间与工程学报 >

2025 , Vol. 21 >Issue 2: 412 - 419

DOI: https://doi.org/10.20174/j.JUSE.2025.02.06

理论与试验研究

基于深度学习的砂岩样品三维重构研究

朱淳 ,
徐佳俊 ,
孙文斌 ,
何昌迪 ,
王晓

展开

1.河海大学地球科学与工程学院,南京 211100;
2.山东科技大学能源与矿业工程学院,山东青岛 266590;
3.犹他大学采矿工程系,犹他州盐湖城美国 84112

朱淳 (1993—),男,江苏徐州人,博士,教授,主要从事深部岩石力学与工程灾害控制等方面的研究工作。E-mail:zhu.chun@hhu.edu.cn

王晓 (1991—),男,山东济宁人,博士,副教授,主要从事岩石动力学方面的教学和研究工作。E-mail:x.wang@sdust.edu.cn

收稿日期: 2024-08-06

网络出版日期: 2025-05-06

基金资助

国家重点研发计划项目(2022YFC3080100);国家自然科学基金(52374119,52404090);江苏省自然科学基金(BK20242059)

收起

Research on 3D Reconstruction of the Sandstone Specimen using Deep Learning

Zhu Chun ,
Xu Jiajun ,
Sun Wenbin ,
He Changdi ,
Wang Xiao

Expand

1. School of Earth Science and Engineering, Hohai University, Nanjing 211100, P.R. China;
2. College of Energy and Mining Engineering, Shandong University of Science and Technology, Qingdao, Shandong 266590, P.R. China;
3. Department of Mining Engineering,University of Utah, Salt Lake City, Utah 84112, USA

Received date: 2024-08-06

Online published: 2025-05-06

Fold

摘要

CT扫描和重构可以提供岩石内部结构的详细视图并对其定量分析,而在此过程中需要对扫描所获得的CT图片中的不同部分进行准确分割。本研究对传统的U-Net、增加残差模块的ResUNet,以及融合了残差模块与迁移学习特性的ResUNet-TL三种深度学习模型进行比较和分析,并借助ImageJ软件中的Weka3D进行图像分割作为对照。通过对比和分析,发现采用预训练的VGG模型以及深度残差网络技术的ResUNet-TL模型在处理复杂的岩石CT图像分割任务上优于其他模型,在准确率和F1得分两种不同的评价指标上都表现出较大的优势。将ResUNet-TL模型应用于二维CT图像中裂隙的识别,随后采用堆叠的方式,将所有被分割的CT图片进行三维重构,获得岩石样本的三维模型并进行定量分析,为岩石科学的研究和应用提供了有效工具。

关键词： 砂岩; CT扫描; 深度学习; 三维重构

本文引用格式

朱淳 , 徐佳俊 , 孙文斌 , 何昌迪 , 王晓 . 基于深度学习的砂岩样品三维重构研究[J]. 地下空间与工程学报, 2025 , 21(2) : 412 -419 . DOI: 10.20174/j.JUSE.2025.02.06

Abstract

CT scanning and reconstruction provide detailed insights into the internal structure of rocks and enable their quantitative analysis. A critical step in this process involves accurately segmenting different components in the CT images. This study compares and analyzes three deep learning models: the conventional U-Net, the ResUNet enhanced with residual modules, and the ResUNet-TL, which incorporates both residual modules and transfer learning features. Image segmentation using the Weka3D plugin in ImageJ is employed as a baseline for comparison. The analysis reveals that the ResUNet-TL model, leveraging a pre-trained VGG model and deep residual network techniques, outperforms the other models in segmenting complex rock CT images, demonstrating advantages in both accuracy and F1 scores. The ResUNet-TL model is applied to identify fractures in 2D CT images, which are then stacked and reconstructed into a 3D model of the rock sample for quantitative analysis. This approach provides an effective tool for advancing research and applications in rock science.

Key words： sandstone; CT scan; deep learning; 3D reconstruction

参考文献

[1] Guo P, Zhang P, Bu M, et al. Microcracking behavior and damage mechanism of granite subjected to high temperature based on CT-GBM numerical simulation[J]. Computers and Geotechnics, 2023, 159: 105385.
[2] Ketcham R, Colbert M, Zhou J, et al. Fracture in granite[DB/OL]. www.digitalrocksportal.org, 2018.
[3] Hassan H, Hassan A, Khan U, et al. Review and classification of AI-enabled COVID-19 CT imaging models based on computers in biology and mediance[J]. Computers in Biology and Medicine, 2022, 141: 105123.
[4] Wang F, Zai Y. Image segmentation and flow prediction of digital rock with U-net network[J]. Advances in Water Resources, 2023, 172: 104384.
[5] Zhao J, Fang Q, Guan Z A. Noise Reduction and Enhancement Processing Method of Cement Concrete Pavement Image based on Frequency Domain Filtering and Small World Network[A]// 2022 International Conference on Edge Computing and Applications (ICECAA)[C], 2022: 777-780.
[6] Ge M, Liu Q. Rock Image Contrast Enhancement Method Based on Improved De-sharpening Mask[A]// 3rd International Conference on Information Science, Parallel and Distributed Systems (ISPDS), IEEE[C], 2022: 223-228.
[7] Sezgin M, Sankur B. Survey over image thresholding techniques and quantitative performance evaluation[J]. Journal of Electronic Imagine, 2004, 13(1): 146-168.
[8] Canny J. A computational approach to edge detection[J]. Transactions on Pattern Analysis and Machine Intelligences, 1986, 8(6): 679-698.
[9] Adams R, Bischof L. Seeded region growing[J]. IEEE Transactions on pattern analysis and machine intelligence, 1994, 16(6): 641-647.
[10] Bhardwaj M, Khan N U, Baghel V, et al. Brain tumor image segmentation using K-means and fuzzy C-means clustering[A]// Digital Image Enhancement and Reconstruction[C]. Academic Press, 2023: 293-316.
[11] Pantovic A, Ollivier I, Essert C. Hybrid U-Net for segmentation of SEEG electrodes on post-operative CT scans[J]. Computer Methods in Biomechanics and Biomedical Engineering-Imaging and Visualization, 2023, 11(4): 1106-1112.
[12] Napte K M, Mahajan A. Liver segmentation using marker controlled watershed transform[J]. International Journal of Electical and Computer Engineering, 2023, 13(2): 1541-1549.
[13] Mano A, Anand S. Local Average Based Kinetic Gas Molecular (LA-KGMO) Optimized MR Brain Image Segmentation Using Modified Self Organizing Map (MSOM)[J]. Wireless Personal Communications, 2023, 128(4): 2703-2723.
[14] Saxena N, Tyagi A, Bisht R, et al. Rock properties from micro-CT images: Digital rock transforms for resolution, pore volume, and field of view[J]. Advances in Water Resources, 2019, 134: 103419.
[15] 徐君, 黄昕, 王君朋, 等. 礁灰岩孔隙结构表征及关键孔隙节点识别研究[J]. 岩石力学与工程学报, 2023, 42(增1):3355-3366. (Xu Jun, Huang Xin, Wang Junpeng, et al. Characterization of coral reef limestones pore structure and identification of key pore nodes[J]. Chinese Journal of Rock Mechanics and Engineering, 2023, 42 (Supp.1):3355-3366 (in Chinese))
[16] Wang S, Ruspini L C, Øren P E, et al. Anchoring Multi-Scale Models to Micron-Scale Imaging of Multiphase Flow in Rocks[J]. Water Resources Research, 2022, 58(1): e2021WR030870.
[17] Ramos M J, Espinoza D N, Torres-Verdín C, et al. Stress-Dependent Dynamic-Static Transforms of Anisotropic Mancos Shale[A]// 51st U.S. Rock Mechanics/Geomechanics Symposium, OnePetro[C], 2017: 580-587.
[18] Garcia-Garcia A, Orts-Escolano S, Oprea S, et al. A review on deep learning techniques applied to semantic segmentation[R]. arXir, 2017.
[19] 周中, 张俊杰, 龚琛杰, 等. 基于深度语义分割的隧道渗漏水智能识别[J]. 岩石力学与工程学报, 2022, 41(10): 2082-2093. (Zhou Zhong, Zhang Junjie, Gong Chenjie, et al. Automatic identification of tunnel leakage based on deep semantic segmentation[J]. Chinese Journal of Rock Mechanics and Engineering, 2022, 41(10): 2082-2093. (in Chinese))
[20] 谢雄耀, 王皓正, 周彪, 等. 复杂环境公路隧道裂缝快速识别与分割算法研究[J]. 地下空间与工程学报, 2022, 18(3): 1025-1033. (Xie Xiongyao, Wang Haozheng, Zhou Biao, et al. Research on fast identification and segmentation algorithms for cracks of highway tunnels in complex environment[J]. Chinese Journal of Underground Space and Engineering, 2022, 18(3): 1025-1033. (in Chinese))
[21] Liu Y, Yao J, Lu X, et al. DeepCrack: A deep hierarchical feature learning architecture for crack segmentation[J]. Neurocomputing, 2019, 338: 139-153.
[22] Roslin A, Marsh M, Piché N, et al. Processing of micro-CT images of granodiorite rock samples using convolutional neural networks (CNN), Part I: Super-resolution enhancement using a 3D CNN[J]. Minerals Engineering, 2022, 188: 107748.
[23] Lorensen W E, Cline H E. Marching cubes: A high resolution 3D surface construction algorithm[J].Computer Graphics (ACM), 1987, 21(4): 163-169.
[24] Drebin R A, Carpenter L, Hanrahan P. Volume rendering[J]. Computer Graphics (ACM) 1988, 22(4): 65-74.
[25] Hohne K H, Bomans M, Tiede U L F, et al. Display of multiple 3D-objects using the generalized voxel-model[A]// Proceeding of SPIE-The International Society for Optical Engineering[C]. 1988, 914: 850-854.
[26] Fishman E K, Ney D R, Heath D G, et al. Volume rendering versus maximum intensity projection in CT angiography: what works best, when, and why[J]. Radiographics, 2006, 26(3): 905-922.
[27] Zhou X P, Jiang D C, Zhao Z. Digital Evaluation of Micro-Pore Water Effects on Mechanical and Damage Characteristics of Sandstone Subjected to Uniaxial, Cyclic Loading-Unloading Compression by 3D Reconstruction Technique[J]. Rock Mechanics and Rock Engineering, 2022, 55(1): 147-167.
[28] Hou P, Liang X, Gao F, et al. Quantitative visualization and characteristics of gas flow in 3D pore-fracture system of tight rock based on Lattice Boltzmann simulation[J]. Journal of Natural Gas Science and Engineering, 2021, 89: 103867.
[29] Wang G, Chen X, Wang S, et al. Influence of fracture connectivity and shape on water seepage of low-rank coal based on CT 3D reconstruction[J]. Journal of Natural Gas Science and Engineering, 2022, 102: 104584.
[30] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[A]// 18th International Conference on Medical Image Computing and Computer-Assisted Intervention-(MICCAI)[C]. 2015: 234-241.
[31] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[A]// 2016 IEEE Conference on Computer Vision and Pattern Recognition[C]. 2016: 770-778.
[32] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[A]// 3rd International Conference Learning Representations[C]. 2015.
[33] Yang B, Wang B. A full-size fracture of tight sandstone induced by water, N₂ and CO₂[DB/OL]. www.digitalrocksportal.org, 2020.
[34] Milletari F, Navab N, Ahmadi S A. V-net: Fully convolutional neural networks for volumetric medical image segmentation[A]// 4th International Conference on 3D Vision (3DV), IEEE[C]. 2016: 565-571.
[35] Arganda-Carreras I, Kaynig V, Rueden C, et al. Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification [J]. Bioinformatics, 2017, 33(15): 2424-2426.
[36] Yang B, Lu L, Zhang C, et al. Digital quantification of fracture in full-scale rock using micro-CT images: A fracturing experiment with N₂ and CO₂[J]. Journal of Petroleum Science and Engineering, 2021, 196: 107682.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献