[Paper Review] Morphology Decoder: A Machine Learning Guided 3D Vision Quantifying Heterogenous Rock Permeability for Planetary Surveillance and Robotic Functions

Omar Alfarisi, Aikifa Raza+4 more|arXiv (Cornell University)|Nov 26, 2021

Enhanced Oil Recovery Techniques59 references19 citations

TL;DR

This paper introduces the Morphology Decoder, a machine learning-guided 3D vision framework that predicts heterogeneous rock permeability from micro-CT and MRI images by integrating geometrical modeling, supervised segmentation, and multi-modal image fusion. It achieves high-accuracy permeability estimation with reduced computational cost compared to traditional lattice Boltzmann simulations, enabling efficient planetary surface analysis and robotic decision-making.

ABSTRACT

Permeability has a dominant influence on the flow properties of a natural fluid. Lattice Boltzmann simulator determines permeability from the nano and micropore network. The simulator holds millions of flow dynamics calculations with its accumulated errors and high consumption of computing power. To efficiently and consistently predict permeability, we propose a morphology decoder, a parallel and serial flow reconstruction of machine learning segmented heterogeneous Cretaceous texture from 3D micro computerized tomography and nuclear magnetic resonance images. For 3D vision, we introduce controllable-measurable-volume as new supervised segmentation, in which a unique set of voxel intensity corresponds to grain and pore throat sizes. The morphology decoder demarks and aggregates the morphologies boundaries in a novel way to produce permeability. Morphology decoder method consists of five novel processes, which describes in this paper, these novel processes are: (1) Geometrical 3D Permeability, (2) Machine Learning guided 3D Properties Recognition of Rock Morphology, (3) 3D Image Properties Integration Model for Permeability, (4) MRI Permeability Imager, and (5) Morphology Decoder (the process that integrates the other four novel processes).

Motivation & Objective

• Address the high computational cost and error accumulation in lattice Boltzmann simulations for permeability prediction in heterogeneous rocks.
• Develop a data-efficient, machine learning-driven method to predict permeability from 3D micro-CT and nuclear magnetic resonance (NMR) images.
• Enable real-time, accurate permeability estimation for planetary surface surveillance and autonomous robotic exploration.
• Introduce a novel supervised segmentation approach using controllable-measurable-volume to map voxel intensity to grain and pore throat sizes.
• Integrate multi-modal 3D image data (micro-CT and MRI) to improve permeability prediction accuracy and robustness.

Proposed method

• Employ a novel supervised segmentation technique called 'controllable-measurable-volume' to map voxel intensity to physical grain and pore throat dimensions.
• Implement a five-stage pipeline: (1) Geometrical 3D Permeability, (2) ML-guided 3D morphology recognition, (3) 3D image properties integration model, (4) MRI Permeability Imager, and (5) Morphology Decoder fusion process.
• Use machine learning to reconstruct parallel and serial flow paths from segmented 3D rock textures derived from micro-CT and NMR data.
• Integrate micro-CT and MRI data into a unified 3D image properties model to enhance permeability estimation accuracy.
• Apply geometric topology and image processing techniques to extract morphological boundaries and aggregate them into permeability-relevant features.
• Train and validate the Morphology Decoder using real Cretaceous rock samples with known permeability, using lattice Boltzmann results as ground truth.

Experimental results

Research questions

• RQ1 Can a machine learning-guided 3D vision system accurately predict rock permeability from micro-CT and MRI images without relying on computationally intensive simulations?
• RQ2 How does the controllable-measurable-volume segmentation method improve the fidelity of pore and grain size representation in 3D rock morphology?
• RQ3 To what extent does fusing micro-CT and MRI data enhance permeability prediction accuracy compared to single-modality approaches?
• RQ4 Can the Morphology Decoder achieve comparable or better accuracy than lattice Boltzmann simulations while reducing computational cost?
• RQ5 How robust is the method in handling heterogeneous rock textures typical of planetary regoliths?

Key findings

• The Morphology Decoder reduces computational cost by orders of magnitude compared to traditional lattice Boltzmann simulations while maintaining high prediction accuracy.
• The controllable-measurable-volume segmentation method successfully maps voxel intensity to physical pore and grain dimensions with improved spatial resolution.
• Fusion of micro-CT and MRI data through the 3D image properties integration model enhances permeability estimation, particularly in low-contrast or heterogeneous regions.
• The method achieves high consistency in permeability prediction across diverse Cretaceous rock textures, demonstrating robustness to morphological heterogeneity.
• The MRI Permeability Imager component effectively captures fluid flow dynamics, contributing to the overall accuracy of the final permeability estimate.
• The Morphology Decoder framework enables real-time permeability inference, making it suitable for deployment in planetary rovers and autonomous exploration systems.