Multiplicative scatter correction during on-line measurement with near infrared spectroscopy

In order to operate an automatic near infrared (NIR) sensor for on-line measurement of some material properties, the multiplicative scatter correction (MSC) is commonly used as one of the important pre-processing steps of spectra. This study was undertaken to demonstrate the development of a modelling procedure to carry out on-line MSC. The procedure was conducted on spectra of soil samples but could be extended to various agricultural materials. Using Unscrambler (R) software, the partial least-squares (PLS) regression with full cross-validation was performed on 300 soil spectra to establish models for calculating MSC coefficients. The procedure was carried out for all MSC options including full, common offset and amplification MSC. These MSC models developed were incorporated into a custom-built LabVIEW program to enable calculating MSC coefficients during on-line measurement and prediction. The validation of the models developed on 78 separate sample set for all MSC options showed large correlation (coefficient of determination R-2 > 0.99) between calculations made on-line using the method developed and corresponding calculations obtained with Unscrambler (R) software. This suggests that the procedure developed for the calculation of MSC coefficients is robust to be adopted for on-line prediction of material properties when MSC is to be used for data pre-processing. (c) 2006 IAgrE. All rights reserved Published by Elsevier Ltd.