Machine Learning for Recognition and Mapping of Rare Earths in Brazil using Reflectance Spectroscopy and Hyperspectral Imagery

Rare Earth Elements are fundamental to the global technology industry. Found in various rocks and soils, the elements that compose them are the raw material for the production of components that are incorporated into a wide range of applications, from household appliances to aircraft, and therefore it is relevant to identify and map them in an automated way, through Artificial Intelligence. In this context, reflectance spectroscopy is capable of providing the important data for the eventual identification of soils, rocks and minerals, and, in conjunction with images from hyperspectral sensors onboard satellite or airborne platforms, it is also possible to map the regions where these elements occur. This work measured the reflectance spectra of soil and rock samples containing Rare Earth Elements from the Poços de Caldas region, a municipality in the state of Minas Gerais, Brazil. The measured spectra were used as input data for a classification neural network, which compared them with spectra captured by the hyperspectral sensor of the PRISMA Satellite.