1. Feature Extraction and Clustering of Hyperspectral Drill Core Measurements to Assess Potential Lithological and Alteration Boundaries.
- Author
-
Rodger, Andrew, Fabris, Adrian, Laukamp, Carsten, and Pour, Amin Beiranvand
- Subjects
DRILL cores ,CLAY minerals ,WORKFLOW ,VISUAL aids - Abstract
A workflow incorporating hyperspectral reflectance data, hull corrections, absorption feature extraction and clustering is presented. The workflow is applied to dense hyperspectral datasets, as collected by hyperspectral drill core logging systems. The extracted absorption features of the reflectance spectra collected from drill cores are shown to form assemblage clusters when plotting the wavelength position of the first, second and third deepest absorption features in two and three dimensions. Using an unsupervised clustering method to establish clusters based on the extracted absorption features yields viewable down hole distributions of similar mineral assemblages. The proposed workflow has the potential for the rapid identification of differing lithologies, alteration and/or weathering overprints. Application of the workflow with no a-priori assumptions about the composition of the potential mineral assemblages provides a means of generating an informative overview of the dataset that is not biased or constrained by preconceptions. The workflow can easily be added to the current workflows of geologists whom are working with dense hyperspectral data to provide an overview of the potential down hole mineral assemblages and aid in the visual logging process or assist in quickly identifying areas for more detailed observation. Furthermore, key mineralogical parameters for resource characterisation, such as the presence of clay minerals can be assessed in a cost and time efficient manner. The proposed workflow is applied to spectra collected from four different drill cores collected in the Gawler Craton located in South Australia and demonstrates the potential outlined above. [ABSTRACT FROM AUTHOR]
- Published
- 2021
- Full Text
- View/download PDF