Performance Report

NIR Technology Australia, Cropscan 2000B™

This report details a study conducted by Agrifood Technology to compare the Cropscan 2000B with the Infratec 1241 and reference analysis conducted according to AWB standard approved methods.

Background

The Cropscan 2000B Grain Analyser (Plate 1) is a Near Infrared Transmission (NIT) Instrument. It is capable of simultaneous determination of several constituents in whole grain samples.

Measurement is based on the fact that grain constituents, such as protein and moisture, absorb electromagnetic radiation in the near infrared region of the light spectrum.

(Plate 1)

NIT analysis is an indirect, predictive method of analysis and therefore must be calibrated, monitored and updated on a regular basis. It is essential to have robust calibrations that include samples from an appropriate range of constituent levels as well as from differing environmental conditions and seasons.

NIR Technology Australia has developed calibrations for Wheat (Hard Wheat) and has requested Agrifood Technology to evaluate the Cropscan 2000B performance and operation against the Infratec 1241 and AWB approved reference methods for protein and moisture.

The following report details the methodology, results and evaluation for the NIR Technology Cropscan 2000B.

Results

Results from the Infratec 1241, Agrifood Reference methods and Cropscan 2000B were statistically analysed using Regression, Hypothesis and One Way ANOVA.

(Hypothesis and One Way ANOVA analysis - Attachment 1).

Wheat Protein Results – Infratec 1241 Global Wheat Application Model compared to Cropscan 2000B Hard Wheat calibration (Table 1, Plate 7).

Table 1:

No. Samples Tested (n)	Average Difference	Range (Infratec)	Standard Deviation of Differences (SDD)	Correlation R^2
50	-0.125	9.3 to 16.7%	0.304	0.9704

(Plate 7)

Wheat Moisture Results– Infratec 1241 Global Wheat Application Model compared to Cropscan 2000B Hard Wheat calibration (Table 2, Plate 8)

Table 2:

No. Samples Tested	Average Difference	Range (By Infratec)	Standard Deviation of Differences (SDD)	Correlation R^2
50	0.111	8.2 to 12.3%	0.248	0.9807

(Plate 8)

Wheat Moisture Results – Oven Reference Method as compared to Cropscan 2000B Hard Wheat Calibration (Table 4, Plate10).

Table 4 :

No. Samples Tested	Average Difference	Range (Reference)	Standard Deviation of Differences (SDD)	Correlation R^2
50	-0.129	8.0 to 12.1%	0.380/0.189*	0.8613

Repeatability

Individual trial samples were analysed in duplicate by each test method and standard deviation between duplicates calculated (Table 5).

Table 5:

Instrument/ Method	No. of Samples	Repeatability (SDD between Duplicates)
Infratec Protein	50	0.02
Cropscan Protein	50	0.04
Infratec Moisture	50	0.00
Cropscan Moisture	50	0.00

Reproducibility

A sub set of ten samples was chosen from the original 50 samples (based on region, grade and protein content) as a means of determining the reproducibility of the Cropscan 2000B.

Analysis was performed by an untrained operator, instructed in the sample preparation procedure and directed to conduct duplicate analysis at the same time of day on three consecutive days.

Results were recorded and standard deviation of differences between the duplicate analyses on the different days calculated (Table 6).

Table 6:

Instrument/ Method	No.of Samples	Reproducibility (SDD between Duplicates)
Cropscan Protein Day 1	10	0.00
Cropscan Moisture Day 1	10	-0.03
Cropscan Protein Day 2	10	-0.02
Cropscan Moisture Day 2	10	0.03
Cropscan Protein Day 3	10	0.02
Cropscan Moisture Day 3	10	0.00

Average SDD between duplicates for total three days for Protein = 0.00

Average SDD between duplicates for total three days for Moisture = 0.00

Conclusions

After initial bias and skew adjustment to Australian Grain Industry Collaborative Trial samples, the Cropscan 2000B Hard Wheat calibration showed significant correlation with the Infratec 1241 Global Wheat Application Model for both Protein (0.9704) and Moisture (0.9807) results.

The Cropscan 2000B showed good repeatability and reproducibility, even after limited operator training (Table 5 & 6).

When compared to the reference method for protein and moisture, both the Cropscan and the Infratec had some difficulty predicting the 02/03 pure variety samples. These samples were specifically chosen for their high protein, low moisture content and may be unique to the previous drought year. (The scans for these samples have been provided to NIR Technology to include in their calibration). (Plate 11 & 12 for comparison)