Machine Learning Guided 3D Image Recognition for Carbonate Pore and Mineral Volumes Determination

• DOI: 10.36227/techrxiv.16961551
• Publication Date: 2021-11-11T13:24:08Z
• Authors: Omar Alfarisi; Aikifa Raza; Hongtao Zhang; Mohamed Sassi; TieJun Zhang
• Keywords: Petrophysics

<p>Automated image processing algorithms can improve the quality, efficiency, and consistency of classifying morphology heterogeneous carbonate rock deal with a massive amount data images seamlessly. Geoscientists petroleum engineers face difficulties in setting direction optimum method for determining petrophysical properties from core plug optical thin-sections, Micro-Computed Tomography (μCT), or Magnetic Resonance Imaging (MRI). Most successful work is homogeneous clastic rocks focusing on 2D less focus 3D requiring numerical simulation. Currently, analysis methods converge to three approaches: processing, artificial intelligence, combined intelligence. In this work, we propose two determine porosity μCT MRI images: an Image Resolution Optimized Gaussian Algorithm (IROGA); advanced recognition enabled by Machine Learning Difference Random Forest (MLDGRF).</p><p>Meanwhile, have built reference micro models collected calibration IROGA MLDGRF methods. To evaluate predictive capability these calibrated approaches, ran them natural rock. We also measured lithology using industry-standard ways, respectively, as values. Notably, produced results accuracy 96.2% 97.1% training set 91.7% 94.4% blind test validation, comparison experimental measurements. limestone pyrite values methods, X-ray powder diffraction, grain density has (limestone pyrite) volume fractions 97.7% measurements.</p>