An Empirical Model-based Method for Signal Restoration of SWIR in ASD Field Spectroradiometry

Abstract

ASD spectroradiometer field measurements in the SWIR water absorption region are sometimes problematic due to atmospheric effects such as moisture in the air. The reduced signal-to-noise ratio (SNR) in these wavebands makes it difficult to diagnose water stress in tree foliage usingspectroscopy. This paper investigates the SNR issue in the 1,350 to 1,410nm waveband using laboratory-based experiments and practical field measurements of mountainous tree foliage spectra. With laboratory spectra data, three empirical signal models along with a Gaussian bias model were testedand validated using noisy field spectra data. Results demonstrate that a combination of either a logistic or sigmoid signal model coupled with a Gaussian bias (residual) model (LOGGM and SIGGM complex signal models) can effectively describe reflectance behaviors in the spectral region of 1,350to 1,410 nm of red cypress (Chaemacyparis formosensis) foliage and further show that our proposed approach is promising in restoration of field spectra measurements.