Prediction of Petrophysical Parameters Based on Digital Video Core Images

Abstract

Summary Core-slab photography is a common way to document geological information from cores. Past practice has been to photograph core slabs with ordinary cameras that produce paper photographs. The presented method retrieves petrophysical properties from high-resolution digital video core images. The procedures described in this work are based on video images (standard RIO/B camera) of cores taken with a digital recording system. The system is able to record in both visible and UV light at different illumination angles, store images, compress/decompress images, and display one or several images as a continuous long core. The seamless core image is marked with depth scale and can be scrolled, scaled, and zoomed. Facilities for correlation with other related data, such as wireline logs, discrete core data, and microscopy images, are also included in the system. We used homogenous dry core plugs from three North Sea oil fields in this work. We recorded images of plug surface, together with conventional core-analysis data (i.e., porosity, gas permeability, average grain size, and mineralogy). The new method is based on processed digital images: light/shadow patterns are obtained by use of asymmetric, low-angle illumination in the green channel. Texture spectra of the rock material are obtained by dedicated image-analytical processing of these gray-scale images and by detecting textural features by use of a unique set of specially designed texture filters. We then calibrate these spectra with respect to measured petrophysical parameters by use of multivariate calibration [partial least squares (PLS)-regression]. Multivariate calibration is based on a set of representative training images, selected to span representative ranges of the intensive petrophysical parameters being modeled. On the basis of this calibration model, similar gray-level video images from new, unknown core sections (with geologically similar fades) are used to estimate properties of the core material by PLS-prediction. In this study it has been possible to model porosity, gas permeability, and average grain size (ORZ) of different formations with a relatively high accuracy and precision. PLS-modeling/-prediction is a strict empirical calibration procedure. The present method is critically dependent upon a thorough, geologically well-documented training data set. Results show that the method is capable of predicting a continuous log of these three petro-physical parameters based on core images calibrated against a set of routine laboratory core-analysis data taken at discrete intervals for a particular formation. The advantages of the new method are rapid and cost-efficient methods for prediction of petrophysical parameters, particularly from slim cores, and improved integration of geological records with wireline data. The method is proposed to be included in future routine laboratory core analysis studies because of its low cost and ability to predict values continuously along the core.