Image texture quantification from Corescan mineral classifications

Corescan generates short-wave infrared (SWIR) derived mineralogy images from drill core that implicitly contain information on mineral texture. Mineral texture information is useful for understanding ore deposit formative processes, mineral processing monitoring and planning, geotechnical assessments and is important in understanding Acid Mine Drainage (AMD) generation capacity of mined materials. This research assesses texture quantification metrics based on the three general measures of image object (mineral grain) size, shape and spatial configuration for the development of indices representing geologically meaningful textures. Individual object texture metrics are summarised within moving windows of a specified size down the length of core using a weighted mean and a measure of variability similar to the coefficient of variation. The most promising meso-scale mineral texture metrics for the construction of geologically meaningful texture indices are: proportion, homogeneity, fractal dimension, elongation and fragmentation. The application of derived texture metrics to a specific need, i.e. geometallurgy or geoenvironmental applications, is limited by the finite resolution (500 μm pixels) of the Corescan mineralogy images and the limitations of the spectral unmixing algorithms used to classify these images. Several image processing choices that affect texture metrics are explored and discussed. The combination of two or more mineral texture metrics within conditional statements appears to be the most promising approach to the construction of geologically meaningful texture indices.