A machine vision approach for detecting changes in drill core textures using optical images
Date
2024
Authors
Gu, X.
Cook, N.J.
Metcalfe, A.V.
Aldrich, C.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Journal article
Citation
Mining Technology: Transactions of the Institutions of Mining and Metallurgy, 2024; 133(2):162-175
Statement of Responsibility
Xiaomeng Gu, Nigel J. Cook, Andrew V. Metcalfe and Chris Aldrich
Conference Name
Abstract
Drill core images offer valuable insights into the texture, structure and mineralogy of ores and their host rocks, which can be used to optimise downstream processes in the mining industry. The impact on downstream processes from particles of similar composition and mineralogy but different textures has been examined by several previous researchers through the application of supervised machine-learning techniques. This study proposes a novel approach for detecting changes in drill core textures through the analysis of optical images. This approach compares three widely used image feature extraction techniques (local binary patterns, grey-level co-occurrence matrix and convolutional neural network), followed by calculation of a uniqueness measure, based on the Hotelling statistic, designed to identify anomalous segments of core. The effectiveness of the uniqueness measure is validated on a test core comprising six sections with different textures. Two drill cores, from the Brukunga test site in South Australia, were selected as case studies. Of the three feature extraction methods, local binary patterns were found to give the strongest signals of change. There exist two main regimes that separate halfway along both drill cores, indicating a change in lithology or the presence of mineralisation.
School/Discipline
Dissertation Note
Provenance
Description
First published online April 11, 2024
Access Status
Rights
© The Author(s) 2024