Spatio-spectro-temporal characterisation and correction of dark current in a snapshot hyperspectral sensor

Abstract

Low-cost uncooled snapshot hyperspectral sensors mounted on UAV platforms offer new opportunities for field-scale remote sensing high-throughput phenotyping, but their reliability is constrained by sensor-intrinsic artefacts, particularly dark current. In this study, we present the first spatio-spectro-temporal characterisation and correction of dark current in the Senop HSC-2 dual-CMOS Fabry-Perot snapshot hyperspectral camera. Six controlled dark experiments (∼6,100 image cubes) revealed that dark current in the sensor is highly structured and reproducible, exhibiting CMOS-specific baseline offsets, monotonic temporal drift, pixel-wise dark signal non-uniformity (DSNU) with wavelength-dependent structure, and persistent hot pixels, and exposure-dependent baseline shifts that do not scale linearly with integration time. These results confirm that conventional single-frame dark subtraction is insufficient for quantitative analysis in uncooled snapshot hyperspectral sensors. Building on this characterisation, a modular correction framework was developed to stabilise the dark signal across spatial, spectral, and temporal domains. Across laboratory datasets, the framework reduced temporal drift by 70-85%, DSNU variance by approximately 37.5%, and suppressed >99.9% of persistent hot pixels, substantially improving radiometric stability. Corrected data exhibited simultaneous spatial uniformity, temporal stability, and spectral integrity, enabling downstream radiometric processing without introducing spectral distortion. Application of Senop HSC-2 to UAV-acquired wheat canopy imagery demonstrated effective transfer to field conditions, improving spectral continuity and robustness of vegetation indices after dark current correction. These results establish a transferable calibration approach for affordable snapshot hyperspectral sensors and demonstrate that rigorous dark current correction is essential for achieving quantitative radiometric performance in UAV-based phenotyping and precision agriculture applications without active thermal control, extending calibration principles traditionally applied in satellite hyperspectral systems to low-cost UAV snapshot sensors.

Plant phenotyping relevance

UAVハイパースペクトルセンサの暗電流を空間・スペクトル・時間的に補正する手法の開発と検証が中心であり、圃場のコムギ群落画像への適用も行っているため、植物フェノタイピング手法として採用する。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.