Imaging carbonate minerals and ore exploration using hyperspectral scanner

Carbonate minerals such as magnesite, dolomite, and calcite have a high economic value, and each of those has different industrial usage due to the variation in chemical composition. Because the geologic appearance of the minerals varies case by cases, the exploration of carbonate ore requires labour intensive field survey. Remote sensing approaches using spectroscopic analysis may provide an alternative solution saving time and labour efforts significantly. We scanned various carbonate minerals including magnesite, dolomite, calcite group, and talc with a hyperspectral scanner, and classified them with imaging techniques based on spectroscopic analysis. We applied a random-forest algorithm to hypespectral images of minerals for optimal band selection. Then logistic regression analysis for detection of each mineral group (magnesite, dolomite, calcite group, and talc) were developed. The logistic regression applied in a stepwise manner classified the minerals successfully with over 90% accuracy. Application of this technique can be effectively used for field exploration of carbonate ores.