Investigating rock micro-structure of sandstones by pattern recognition on their X-ray images

Abstract

Summary Analysing rock micro-structure from micro-computed tomographic images of porous media is vital to understand fluid-flow and estimating petrophysical properties like permeability. The two main approaches of analysing rock micro-structure are (1) through experiments, a timeconsuming process and (2) using numerical simulations which are a part of the standard digital rock physics (DRP) workflow. The standard DRP workflow requires the micro-computed tomographic (micro-CT) images to be segmented into distinct phases (pores and minerals). Segmentation is a user-biased and manual process. It relies heavily on the user to choose a threshold(s) that distinguishes unique phases present in the rock microstructure. Thus, introducing uncertainty in these petrophysical properties. Our approach to resolving this subjectivity and uncertainty in analysing rock microstructure is to directly apply techniques on the micro-CT images or the greyscale images, rather than using segmented images. For this purpose, we use a pattern recognition technique namely the Grey-Level Cooccurrence Matrix (GLCM). The GLCM technique is used to calculate spatial maps that describe features present in the rock micro-structure. Calculating these spatial maps at varying length-scales by using different displacement vectors aid in analysing the grain-sizes, grain-pore interface and pore-sizes. Unlike the histograms which only preserve the frequency of intensity values that represent different features in micro-CT images, GLCM is a secondorder pattern recognition technique that additionally preserves the spatial variation and occurrence of grey-level intensity values. This method of studying the rock microstructure using greyscale images and pattern-recognition techniques provides an advantage over the conventional segmentation techniques because full-information regarding the rock micro-structure captured during microcomputed tomography is preserved and a threshold-less workflow leads to lesser user subjectivity. Lastly, the GLCM based analysis also provides a pathway for automated investigation of rock-microstructure.