Curated Spectral Reflectance Dataset to Aid Robotic Exploration Research

Reflectance spectra help understand the mineralogy and play a crucial role in robotic exploration. The reflectance spectra can be multispectral, covering tens of spectral wavelengths or hyperspectral, covering hundreds of spectral wavelengths. Though the availability of multispectral and hyperspectral datasets has increased in the past decade, accessing and using these datasets is hard mainly due to logistic reasons like large file sizes covering large regions. For research that relies on spectral data, it is beneficial to first focus on small regions that are mineralogically diverse and then expand to large areas.
We have curated a dataset that focusses on three mineralogically diverse sites – Cuprite Hills in Nevada, Hopi Buttes in Arizona and Yellow Cat in Utah. We include multispectral data from ASTER and LANDSAT9 instruments and hyperspectral data from AVIRIS instrument. In addition to the spectral data, we also share high resolution (3-5cm/pixel) RGB data for these sites collected using drones and mineral maps that specify mineral at each pixel predicted using the hyperspectral data. This curated dataset will help research using spectral datasets and enable faster algorithm development.