Cracknell, MJ and Parbhakar-Fox, A and Jackson, LM and Fox, N and Savinova, E, Automated identification of sulphides from drill core imagery, Proceedings of the 2019 Mineral Systems of the Pacific Rim Congress (PACRIM 2019), 3-5 April 2019, Auckland, New Zealand, pp. 79-82. (2019) [Refereed Conference Paper]
Quantification of mineral concentrations is crucial for planning efficient and economical ore extraction, metals processing and mine waste management (Berry et al., 2016). Several analytical methods are available to automatically identify minerals, including sulphides, e.g. Scanning Electron Microscopy (SEM), Laser Raman Spectroscopy and X-ray diffraction (XRD). These methods operate at microscopic scales and require samples to be prepared prior to analysis, hence, they can be time consuming to carry out and problematic when scaled to represent mining ore and waste materials (Goodall et al., 2005; Berry et al., 2016).
The use of visible and near infrared (VNIR), shortwave infrared (SWIR), and more recently thermal infrared (TIR) scanning systems for mineral identification are well established and offer rapid, cheap and non-destructive methods for characterising rock mineralogy drill core scales (Schodlok et al., 2016). Despite their advantages, VNIR (450-1100 nm), SWIR (1100–2500 nm) and TIR (1.1-14.5 μm) systems are only useful for the identification of minerals that are active in these portions of the electromagnetic spectrum. Sulphides, which are economically and environmentally important minerals, typically do not have characteristic absorption features in VNIR, SWIR and TIR wavelengths (Bolin and Moon, 2003; Merrill et al., 2016). Nevertheless, recent research suggests that TIR wavelengths of around 7.6 μm can be used to identify sulphide minerals (Merrill et al., 2016). Furthermore, iron-sulphides have been identified from hyperspectral drill core images (across VNIR wavelengths) using supervised classification (Bolin and Moon, 2003). This approach exploited ironsulphide mineral colour and albedo to distinguish them from other minerals.
In this study, red-green-blue (RGB) images of drill core in combination with hyperspectral data are used as input into a Random Forests supervised classification algorithm in order to discriminate ironsulphides from other minerals.
|Item Type:||Refereed Conference Paper|
|Keywords:||drill core, sulphides, prediction, supervised classiﬁcation|
|Research Division:||Earth Sciences|
|Research Field:||Geology not elsewhere classified|
|Objective Division:||Mineral Resources (excl. Energy Resources)|
|Objective Group:||Primary Mining and Extraction Processes of Mineral Resources|
|Objective Field:||Primary Mining and Extraction of Mineral Resources not elsewhere classified|
|UTAS Author:||Cracknell, MJ (Dr Matthew Cracknell)|
|UTAS Author:||Parbhakar-Fox, A (Dr Anita Parbhakar-Fox)|
|UTAS Author:||Jackson, LM (Miss Laura Jackson)|
|UTAS Author:||Fox, N (Dr Nathan Fox)|
|Deposited By:||CODES ARC|
Repository Staff Only: item control page