The potential of spectral unmixing method applied to PRISMA hyperspectral images in the identification of Li minerals: an evaluation for prospecting purposes

Abstract

Different remote sensing methods already applied have proven efficient in identifying pegmatites, but the high number of the false positives and the size of the study areas involved, make the location of new points of interest for exploration a difficult task. In order to develop and evaluate more autonomous tools for localization of new points of interest, this study aims to apply the Envi Spectral Hourglass Wizard (SHW) algorithm and spectral analysis, both applied on PRISMA hyperspectral images, to determine mineral distribution in St. Austell greisen deposit, a Li exploration target located at Cornwall, UK. The SHW finds endmembers within the dataset to map their location and sub-pixel abundance. This processing workflow is composed of several steps: (i) MNF (Minimum Noise Fraction) Transform; (ii) PPI (Pixel Purity Index); (iii) n-D space visualizer, allowing the extraction of the endmembers and; (iv) the SAM (Spectral Angle Mapper) classification algorithm, which classifies the image creating a class for each collected endmember. The classification results show the potential of the method to indicate the presence of Li minerals being able to identify Kaolinite and map the distribution and abundance of Topaz, Tourmaline and Biotite. This approach is highly valuable for the Li mining industry.