An end-to-end three-dimensional reconstruction framework of porous media from a single two-dimensional image based on deep learning

Abstract Stochastically reconstructing a three-dimensional (3D) structure of porous media from a given two-dimensional (2D) image is an outstanding problem. For such problem, despite that the big progress has been made on reconstruction methods such as optimization-based and multi-point statistics-based algorithms, however, the reconstruction of topologically complex or non-stationary materials is still not well addressed. Besides, the reconstruction efficiency is another remarkable challenge, and a 12 8 3 reconstruction using these methods generally requires several hours. In this paper, to overcome these problems, we propose a general end-to-end deep learning-based 3D reconstruction framework. Specially, the mapping (function) between a 2D slice and its 3D structure is first learned by a neural network. Then, the 3D reconstruction of a new 2D image using this mapping is instantaneous. For a 12 8 3 reconstruction, our method only requires 0 . 2 s , thus achieving a 3 . 6 × 1 0 4 speedup factor compared with the classical methods. Besides, to yield diverse 3D structures for the same 2D input, a Gaussian noise is introduced into the network. Our approach is tested on two statistically isotropic materials and a non-stationary porous material, and evaluated in terms of both visual and quantitative comparisons. Experimental results indicate that the proposed method is accurate, fast, and stable. The proposed framework also enables that theoretically an arbitrary number of constraints can be incorporated to further improve the reconstruction accuracy.

[1]  Wei Chen,et al.  Stochastic microstructure characterization and reconstruction via supervised learning , 2016 .

[2]  Hilfer,et al.  Local-porosity theory for flow in porous media. , 1992, Physical review. B, Condensed matter.

[3]  Ahmed H. Elsheikh,et al.  Exemplar-based synthesis of geology using kernel discrepancies and generative neural networks , 2018, ArXiv.

[4]  Stefan Zaefferer,et al.  Anisotropic distribution of the micro residual stresses in lath martensite revealed by FIB ring-core milling technique , 2018 .

[5]  Qizhi Teng,et al.  Improved multipoint statistics method for reconstructing three-dimensional porous media from a two-dimensional image via porosity matching. , 2018, Physical review. E.

[6]  R. M. Srivastava,et al.  Multivariate Geostatistics: Beyond Bivariate Moments , 1993 .

[7]  Marina V Karsanina,et al.  Hierarchical Optimization: Fast and Robust Multiscale Stochastic Reconstructions with Rescaled Correlation Functions. , 2018, Physical review letters.

[8]  Xiaohai He,et al.  Markov prior-based block-matching algorithm for superdimension reconstruction of porous media. , 2018, Physical review. E.

[9]  XiaoHai He,et al.  Reconstruction of three-dimensional porous media from a single two-dimensional image using three-step sampling. , 2015, Physical review. E, Statistical, nonlinear, and soft matter physics.

[10]  Pierre M. Adler,et al.  Real Porous Media: Local Geometry and Macroscopic Properties , 1998 .

[11]  Bo Chen,et al.  MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications , 2017, ArXiv.

[12]  Eric Laloy,et al.  Inversion using a new low-dimensional representation of complex binary geological media based on a deep neural network , 2017, 1710.09196.

[13]  Muhammad Sahimi,et al.  Higher-order correlation functions in disordered media: Computational algorithms and application to two-phase heterogeneous materials , 2018, Physical Review E.

[14]  W Chen,et al.  Characterization and reconstruction of 3D stochastic microstructures via supervised learning , 2016, Journal of microscopy.

[15]  Salvatore Torquato,et al.  Two‐point cluster function for continuum percolation , 1988 .

[16]  Sadok Lamine,et al.  Multidimensional upwind schemes and higher resolution methods for three-component two-phase systems including gravity driven flow in porous media on unstructured grids , 2015 .

[17]  Yang Li,et al.  Accurate and Fast reconstruction of Porous Media from Extremely Limited Information Using Conditional Generative Adversarial Network , 2019, Physical review. E.

[18]  Xiaohai He,et al.  Pattern density function for reconstruction of three-dimensional porous media from a single two-dimensional image. , 2016, Physical review. E.

[19]  Marcel Utz,et al.  Efficient reconstruction of multiphase morphologies from correlation functions. , 2001 .

[20]  Muhammad Sahimi,et al.  Flow and Transport in Porous Media and Fractured Rock: From Classical Methods to Modern Approaches , 1995 .

[21]  G. Mariéthoz,et al.  Reconstruction of Incomplete Data Sets or Images Using Direct Sampling , 2010 .

[22]  S. Torquato,et al.  Reconstructing random media. II. Three-dimensional media from two-dimensional cuts , 1998 .

[23]  Christoph H. Arns,et al.  Three-dimensional porous structure reconstruction based on structural local similarity via sparse representation on micro-computed-tomography images , 2018, Physical Review E.

[24]  M. Sahimi,et al.  Cross-correlation function for accurate reconstruction of heterogeneous media. , 2013, Physical Review Letters.

[25]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[26]  Sunetra Sarkar,et al.  An efficient stochastic framework to propagate the effect of the random solid-pore geometry of porous media on the pore-scale flow , 2017 .

[27]  Gregoire Mariethoz,et al.  The Direct Sampling method to perform multiple‐point geostatistical simulations , 2010 .

[28]  Gui-Bin Bian,et al.  Performance Analysis of Google Colaboratory as a Tool for Accelerating Deep Learning Applications , 2018, IEEE Access.

[29]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[30]  Pejman Tahmasebi,et al.  Reconstruction of three-dimensional porous media using a single thin section. , 2012, Physical review. E, Statistical, nonlinear, and soft matter physics.

[31]  Dirk Mallants,et al.  Improving pattern reconstruction using directional correlation functions , 2014 .

[32]  Yang Jiao,et al.  Dynamic reconstruction of heterogeneous materials and microstructure evolution. , 2015, Physical review. E, Statistical, nonlinear, and soft matter physics.

[33]  Pejman Tahmasebi,et al.  MS-CCSIM: Accelerating pattern-based geostatistical simulation of categorical variables using a multi-scale search in Fourier space , 2014, Comput. Geosci..

[34]  Yang Jiao,et al.  Modeling Anisotropic Multiphase Heterogeneous Materials via Directional Correlation Functions: Simulations and Experimental Verification , 2012, Metallurgical and Materials Transactions A.

[35]  Heping Xie,et al.  3-D Reconstruction Method for Complex Pore Structures of Rocks Using a Small Number of 2-D X-Ray Computed Tomography Images , 2019, IEEE Transactions on Geoscience and Remote Sensing.

[36]  Karen Abrinia,et al.  Efficient three‐phase reconstruction of heterogeneous material from 2D cross‐sections via phase‐recovery algorithm , 2016, Journal of microscopy.

[37]  Tao Wu,et al.  Reconstruction of three-dimensional anisotropic media based on analysis of morphological completeness , 2019, Computational Materials Science.

[38]  Martin J. Blunt,et al.  Reconstruction of three-dimensional porous media using generative adversarial neural networks , 2017, Physical review. E.

[39]  Zhibin Pan,et al.  Correlation-driven direct sampling method for geostatistical simulation and training image evaluation. , 2019, Physical review. E.

[40]  Alexei A. Efros,et al.  Toward Multimodal Image-to-Image Translation , 2017, NIPS.

[41]  Yang Ju,et al.  Multi-thread parallel algorithm for reconstructing 3D large-scale porous structures , 2017, Comput. Geosci..

[42]  S. Torquato,et al.  Random Heterogeneous Materials: Microstructure and Macroscopic Properties , 2005 .

[43]  Zhi Xu,et al.  Stable-phase method for hierarchical annealing in the reconstruction of porous media images. , 2014, Physical review. E, Statistical, nonlinear, and soft matter physics.

[44]  F. Stillinger,et al.  Modeling heterogeneous materials via two-point correlation functions. II. Algorithmic details and applications. , 2008, Physical review. E, Statistical, nonlinear, and soft matter physics.

[45]  Forrest N. Iandola,et al.  SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <1MB model size , 2016, ArXiv.

[46]  Simon K. Alexander,et al.  Hierarchical Annealing for Synthesis of Binary Images , 2009 .

[47]  Yang Jiao,et al.  Modeling and predicting microstructure evolution in lead/tin alloy via correlation functions and stochastic material reconstruction , 2013 .

[48]  Yukai Wang,et al.  CT-image Super Resolution Using 3D Convolutional Neural Network , 2018, Comput. Geosci..

[49]  Pejman Tahmasebi,et al.  HYPPS: A hybrid geostatistical modeling algorithm for subsurface modeling , 2017 .

[50]  Salvatore Torquato,et al.  Accurate modeling and reconstruction of three-dimensional percolating filamentary microstructures from two-dimensional micrographs via dilation-erosion method , 2014 .

[51]  Hao Liang,et al.  Accelerating simulation for the multiple-point statistics algorithm using vector quantization. , 2018, Physical review. E.

[52]  Hanan Samet,et al.  Pruning Filters for Efficient ConvNets , 2016, ICLR.

[53]  Marc Secanell,et al.  Stochastic reconstruction using multiple correlation functions with different-phase-neighbor-based pixel selection. , 2014, Physical review. E, Statistical, nonlinear, and soft matter physics.

[54]  Wing Kam Liu,et al.  Computational microstructure characterization and reconstruction: Review of the state-of-the-art techniques , 2018, Progress in Materials Science.

[55]  Pejman Tahmasebi,et al.  Enhancing multiple‐point geostatistical modeling: 2. Iterative simulation and multiple distance function , 2016 .

[56]  Xiaohai He,et al.  Super-resolution of real-world rock microcomputed tomography images using cycle-consistent generative adversarial networks. , 2020, Physical review. E.

[57]  Qizhi Teng,et al.  Accelerating multi-point statistics reconstruction method for porous media via deep learning , 2018, Acta Materialia.

[58]  Simon Osindero,et al.  Conditional Generative Adversarial Nets , 2014, ArXiv.

[59]  Chao Ren,et al.  Dictionary optimization and constraint neighbor embedding-based dictionary mapping for superdimension reconstruction of porous media. , 2019, Physical review. E.

[60]  F. Stillinger,et al.  A superior descriptor of random textures and its predictive capacity , 2009, Proceedings of the National Academy of Sciences.

[61]  R. A. Leibler,et al.  On Information and Sufficiency , 1951 .

[62]  K. Gerke,et al.  Calculation of tensorial flow properties on pore level: Exploring the influence of boundary conditions on the permeability of three-dimensional stochastic reconstructions. , 2019, Physical review. E.

[63]  Xiang Li,et al.  Designing phononic crystal with anticipated band gap through a deep learning based data-driven method , 2020 .

[64]  Ran El-Yaniv,et al.  Binarized Neural Networks , 2016, NIPS.

[65]  Yang Li,et al.  Reconstruction of three-dimensional heterogeneous media from a single two-dimensional section via co-occurrence correlation function , 2018 .

[66]  Kirill M. Gerke,et al.  Improving stochastic reconstructions by weighting correlation functions in an objective function , 2015 .

[67]  Christoph H. Arns,et al.  Porous Structure Reconstruction Using Convolutional Neural Networks , 2018, Mathematical Geosciences.

[68]  P. Stroeven,et al.  Local porosity analysis of pore structure in cement paste , 2005 .

[69]  Martin J. Blunt,et al.  Stochastic Reconstruction of an Oolitic Limestone by Generative Adversarial Networks , 2017, Transport in Porous Media.

[70]  Mian Lin,et al.  An Improved Method for Reconstructing the Digital Core Model of Heterogeneous Porous Media , 2017, Transport in Porous Media.

[71]  S. Torquato,et al.  Reconstructing random media , 1998 .

[72]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[73]  Ahmed H. Elsheikh,et al.  Parametric generation of conditional geological realizations using generative neural networks , 2018, Computational Geosciences.

[74]  Xueqing Zhou,et al.  Challenges and Prospects of Digital Core-Reconstruction Research , 2019, Geofluids.

[75]  M. Sahimi,et al.  Three-Dimensional Stochastic Characterization of Shale SEM Images , 2015, Transport in Porous Media.

[76]  S. Torquato,et al.  Lineal-path function for random heterogeneous materials. , 1992, Physical review. A, Atomic, molecular, and optical physics.

[77]  Xiang Li,et al.  Predicting the effective mechanical property of heterogeneous materials by image based modeling and deep learning , 2019, Computer Methods in Applied Mechanics and Engineering.

[78]  Pejman Tahmasebi,et al.  Enhancing multiple‐point geostatistical modeling: 1. Graph theory and pattern adjustment , 2016 .

[79]  Lei Song,et al.  Efficient 3D reconstruction of random heterogeneous media via random process theory and stochastic reconstruction procedure , 2019, Computer Methods in Applied Mechanics and Engineering.

[80]  Yang Jiao,et al.  Direct extraction of spatial correlation functions from limited x-ray tomography data for microstructural quantification , 2018, Materials Characterization.

[81]  Tao Huang,et al.  GPU-based SNESIM implementation for multiple-point statistical simulation , 2013, Comput. Geosci..

[82]  Yang Ju,et al.  3D numerical reconstruction of well-connected porous structure of rock using fractal algorithms , 2014 .

[83]  M. Blunt,et al.  Pore space reconstruction using multiple-point statistics , 2005 .

[84]  Eric Laloy,et al.  Training‐Image Based Geostatistical Inversion Using a Spatial Generative Adversarial Neural Network , 2017, ArXiv.

[85]  Xiaohai He,et al.  Super-dimension-based three-dimensional nonstationary porous medium reconstruction from single two-dimensional image , 2019, Journal of Petroleum Science and Engineering.

[86]  A. Safekordi,et al.  A multiple-point statistics algorithm for 3D pore space reconstruction from 2D images , 2011 .

[87]  Martin J. Blunt,et al.  Conditioning of three-dimensional generative adversarial networks for pore and reservoir-scale models , 2018, ArXiv.

[88]  T. Tang,et al.  A pixel selection rule based on the number of different‐phase neighbours for the simulated annealing reconstruction of sandstone microstructure , 2009, Journal of microscopy.