Spectral Pre-Treatment for Diffuse Transmittance Linearity Improvement

Abstract

The linearity of diffuse transmittance spectra was considered on the basis of the Kubelka-Munk theory. The sub-linear dependence of -logT on the component concentration was determined and a linearisation method is suggested. The main point of linearisation is solving a cubic equation with absorbance to be linearised as an unknown variable and with the equation coefficients being power functions of the sample scattering capacity. The performance of the method is tested by employing the spectra of adsorbed water in the 1796-2004 nm range. The cellulose sheet was a matrix for adsorbed water molecules. Seventeen water content values from 0.6% to 22.7% (w/w) were prepared for calibration-prediction procedures. Raw spectra and linearised spectra were analysed by principal component analysis and partial least squares (PLS) regression. PLS cross-validation shows the 2.85-fold decrease of the prediction error after linearisation.

Spectral Pre-Treatments of Hyperspectral near Infrared Images: Analysis of Diffuse Reflectance Scattering

“…Much has been written about various numerical preprocessing treatments proposed to linearise spectral response and reduce or eliminate additive or multiplicative effects in near infrared (NIR) spectra. [1][2][3][4] The most basic approach is to simply convert the diffuse reflectance R measurements into more linear units such as absorbance which can then be directly related to concentration. Alternatively, the Kubelka-Munk transform can be applied.…”