Enhancing mineral exploration using hyperspectral imaging and dual regression reflectance conversion

This paper proposes a novel, automated system for mineral detection using hyperspectral imaging (HSI), with a focus on kaolinite identification. Unlike traditional approaches, our method introduces a dual regression model that learns to convert top-of-atmosphere (TOA) reflectance to bottom-of-atmosphere (BOA) reflectance, treating atmospheric correction as a denoising problem. The model is trained on paired TOA-BOA reflectance data and supports bidirectional learning for improved stability and spectral fidelity. Post-correction, spectral angle mapping is used to identify mineral presence from spectral libraries. The experimental results from EnMAP satellite imagery show both high accuracy and spatial reliability which validates the model's performance. The research combines atmospheric correction with mineral classification into a complete system that supports real-time and scalable mineral exploration operations.