HIDSAG: Hyperspectral Image Database for Supervised Analysis in Geometallurgy

Area of Science:

• Geoscience
• Data Science
• Mineralogy

Background:

• Supervised spectral analysis demands robust response variable characterization and ample data.
• Hyperspectral imaging provides rich data for mineral sample analysis.

Purpose of the Study:

• To present a comprehensive hyperspectral dataset for geometallurgical samples.
• To facilitate general regression and classification research using spectral data.

Main Methods:

• Scanning mineral samples using SPECIM VNIR and SWIR hyperspectral cameras.
• Acquiring hyperspectral reflectance images across VNIR (400-1000 nm) and SWIR (900-2500 nm) ranges.
• Generating RGB images via sensor fusion and providing response variables for machine learning.

Main Results:

• A validated hyperspectral dataset comprising spectral images and mineral characterization data.
• Successful validation of all data subsets using machine learning models.

Conclusions:

• The presented dataset is suitable for diverse regression and classification tasks in spectral data analysis.
• The dataset's validation confirms its utility for machine learning applications in geometallurgy.