Calibration in non-linear near infrared reflectance spectroscopy: a comparison of several methods

摘要: Abstract Principal component regression (PCR) and partial least-squares (PLSR) are the two calibration procedures most frequently used in quantitative applications of near infrared diffuse reflectance spectroscopy (NIRRS). Some systems, however, exhibit a non-linear relationship that neither methodology can model. Frequently, main culprit such non-linearity is multiplicative effect arising from non-uniform particle sizes or diameters samples. In this work, we tested various approaches to minimizing resulting differences size sample thickness, using determination linear density acrylic fibres as physical The involve prior linearizing data by logarithmic conversion and/or use systems; context, results applying stepwise polynomial PCR (SWP-PCR) PLSR (SWP-PLSR), those provided neural network based on scores model (PC-ANN), were compared. PC-ANN approach was found provide best with data. On other hand, SWP-PLSR performed par previous one when variable linearized its values into decimal logarithms.