Automatic Radiance Level Data Quality and Signal-to-Noise Assessment for Hyperspectral Imaging
Abstract
Raw hyperspectral imaging (HSI) data require multiple processing steps before physically meaningful reflectance products can be derived. Quality issues at the radiance level – e.g., detector artifacts, low-signal regions, or excessive noise – can propagate through the processing chain and compromise subsequent analyses. HSI workflows typically focus on calibration and correction but often lack standardized integrity checks prior to atmospheric compensation. To address this gap, we developed an automated Data Quality Assessment (DQA) toolbox within the EU Horizon m4mining project. The toolbox is integrated into INTACOR ® , an automated processing pipeline for airborne hyperspectral imagery developed by ReSe Applications LLC, and provides quality assessment of calibrated radiance data prior to subsequent processing. It identifies bad pixels and low-signal regions, derives vegetation information, and estimates spatially resolved apparent signal-to-noise ratio (SNR) using spectral differencing and local residual statistics, from which robust band-wise SNR statistics are derived. The toolbox produces spatially explicit quality layers and an automated report to support rapid assessment of acquired data. Application to airborne hyperspectral mining data demonstrates its ability to characterize spatial and spectral variations in data quality, providing a standardized radiance-level quality assessment stage within an operational HSI processing chain.