NIR Technology Australia

                                                                                                                                                                     Suite 103, 56 Kitchener Pde

                                                                                                Bankstown  NSW  2200

Tel: 02 9708 5068

Fax: 02 9708 5537

Application Note 01:Cropscan 2000G Wheat Calibration

February 2000

Introduction
The Cropscan 2000G Whole Grain Analyser has been used to develop a calibration capable of predicting both the protein and moisture content of the four major wheat classes grown in Australia. To the original calibration data, a temperature stabilisation set of samples were added whose NIR spectra were scanned under temperature conditions at the extremes of the normal working range of the instrument. The effects of recalibration with these stabilisation samples are discussed below.

Description
121 samples of Australian wheat were analysed in the laboratory for protein (Leco) and moisture (Oven) content. These samples were scanned on the Cropscan 2000G between 720-1100nm using a 20mm pathlength cell. A calibration model was developed using partial least squares regression (PLS-1 algorithm in the software package Unscrambler®). A further 80 samples were analysed in the laboratory and scanned between 720-1100nm at various temperatures at the extremes of the normal working conditions expected for the instrument. The sampling procedure is as follows: The instrument was equilibrated to 28oC and five (5) wheat samples (equilibrated to 10oC) were scanned on the Cropscan 2000G. The next step involved scanning five samples equilibrated to 45oC on the instrument at 28oC. The procedure was repeated in reverse order (ie. The samples were equilibrated to 28oC measured on the instrument at 10oC and then at 45oC). These samples were added to the original calibration model and regression analysis repeated.

 

Results
The regression statistics for the calibrations are given below

 

Number of elements (n)

Range (%)

Number of Principal Components (PC’s)

Correlation (R)

Standard Error of Determination (SED)

Original Calibration

 

 

 

 

 

Protein

121

7-16

10

0.9831

0.35

Moisture

121

7-15

6

0.9746

0.26

Original + Stabilisation Set

 

 

 

 

 

Protein

201

7-16

10

0.9710

0.46

Moisture

201

7-15

7

0.9611

0.33

 

Conclusion
The addition of samples measured on the Cropscan 2000G at the extremes of the expected working temperature range of the instrument has the effect of reducing the correlation and increasing the SED (>0.1%) of both protein and moisture calibrations. NIR calibrations are sensitive to temperature changes and therefore the incorporation of a temperature stabilisation set of samples provides a means of correction for measurements performed at the extreme temperature ranges.