1Queensland Department of Primary Industries, Farming Systems Institute, PO Box 2282, Toowoomba Qld 4350
2BRI Australia Ltd PO Box 7, North Ryde, NSW, 1670
Introduction
Colour evaluation has been used for many years as a marketing tool for agricultural and food products. Consumers have a general perception that products with a clean, bright appearance are of a higher quality than products that may appear dull. Similar perceptions seem to apply when marketing grain. Grains such as wheat and barley are traded on many quality parameters, eg. protein content but also on visual characteristics, including the percentage of black point grains and overall sample brightness. At most grain receival depots within Australia, grain samples are visually assessed for colour. Most of the grain discolouration is attributed to fungal infection (Keen, 1963 and Pepper, 1960 in Etchevers et al., 1976; Buitendag et al., 1994).
A number of types of instruments and technologies are available to quantitatively measure barley grain colour (Etchevers et al., 1976; Svensson et al., 1994). Colour measurements for barley in the past have been used for both breeding programs and industry evaluation (Blakeney et al. 1994; ASBC, 1992). While it is possible to measure grain colour rapidly with colour instruments and image analysis, to do so at grain receival depots would require significant capital investment to purchase the many instruments to station at each depot. An alternative is to use existing technology such as Near Infrared Reflectance (NIR) spectroscopy. NIR technology is a rapid and non-destructive technique and has been applied to predict barley grain colour in a Visible/NIR spectrophotometer (Osborne and Jackson, 1997), as well as Near Infrared Reflectance and Transmission (NIT) instruments (Blakeney and Stuart, 1998; K. Ayton pers comm).
The objectives of this study were to standardise a reference method for NIR calibrations and to develop calibrations within a number of NIR and NIT instruments specifically for grain receival in Northern Australia (northern New South Wales and Queensland). These calibrations would also be available to grain handlers in other states of Australia.
Materials and Methods
Barley Samples
Commercial barley samples were collected from the various bulk handling companies within Australia. Additional samples from the Barley Improvement in Northern Australia program were included.
Reference Method
All samples for NIR calibrations were measured with the Minolta CR310 50mm diameter measuring head (Blakeney et al., 1994) calibrated to the white tile supplied with the instrument. A second tile was used to establish the difference in colour readings using a different calibration tile. In addition, alternative methods for measurement of colour were assessed. The Light Projection Tube (LPT) and the Granular Materials Attachment (GMA) of the Minolta CR310 were compared to assess the variation in colour readings between these two methods. The Light Projection Tube was attached to the colorimeter and placed on top of the sample inside the sample container. The Granular Materials Attachment assessed samples in isolation from the sample container. The comparison of the two tiles and assessment methods was performed on a subset of the NIR calibration set.
Colour readings were recorded after harvest (January 1999) using the Granular Materials Attachment. To assess if changes in colour occurred during storage, colour readings were repeated in June 1999.
NIR Calibrations
The instruments used for the development of calibrations were two used by Australian grain handlers (which are only NIT instruments) together with the two scanning instruments used by Australian barley breeding programs’ quality laboratories. The list and specifications of each instrument is given in Table 1.
Table 1. Details of the NIR/NIT instruments
Instrument |
Type |
Light source |
Wavelength Range |
Grainspec |
NIT |
33 filters |
808-1075 nm |
Infratec 1229 |
NIT |
33 filters |
850-1050 nm |
NIRSystems 6500 |
Vis/NIR |
Monochrometer |
400-2500 nm |
Perten DA7000 |
Vis/NIR |
Diode Array |
400-1700 nm |
Results and Discussion
Precision of Reference method
Any NIR calibration data is only as good as the precision of the reference method. For measuring grain colour a number of options are available. The Granular Materials Attachment gave slightly better precision than the Light Projection Tube for both Calibration tiles (Table 2). These values were similar to those reported by Osborne and Jackson (1997). While the precision value for the LPT and the GMA were similar there was a slight increased average “L” value using the GMA. For NIR calibrations it would be necessary to use the same method each season when determining laboratory values. Additionally, for a standard method that would measure laboratory values for calibrations to be utilised by a number of users, the calibration tiles should be identically matched.
Table 2. Precision data for Minolta calibrated against two different white tiles1
Tile 1 |
Tile 2 | |||
Mean |
SE |
Mean |
SE | |
Light Projection Tube |
60.12 |
0.52 |
59.86 |
0.53 |
Granular Materials Attachment |
60.36 |
0.48 |
59.71 |
0.46 |
1 Precision values determined on subset (10) of calibration set. Measurements were conducted in triplicate
Colour stability
The calibration set contained “L” values determined shortly after harvest. The colour measurement was repeated six months later. It would appear that there was an average increase over six months (Table 3). The L value increased by 1.39 L units using the LPT and 0.27 using the GMA. The largest change was 2.90 units and the smallest change was 0.10 units. However, the samples that had the greatest change were those with “L” values less than 58.
Table 3. Change in L values over six months
December |
June |
||
Mean |
Mean |
Difference | |
Light Projection Tube |
59.17 |
60.56 |
1.39 |
Granular Materials Attachment |
60.55 |
60.82 |
0.27 |
NIR Calibrations
The range in L values for the calibration set was 52 – 65. Initially all calibrations were derived using all samples. The accuracy on some calibrations was well below those reported previously (Osborne and Jackson, 1997 and Blakeney and Stuart, 1998). The calibration accuracy improved when samples with “L” values less than 58 were removed (Table 4). The NIR with greatest accuracy was the NIRSystems 6500 with a SECV 0.55 “L” units. This results was similar to that reported by Osborne and Jackson (1997). For the NIT instruments commonly used by grain handlers, the Grainspec provided the best calibrations. Figure 1 shows the scatter plot for actual vs predicted L values from the Grainspec.
Table 4. Calibrations data for barley grain colour
Instrument |
Software |
NIR/NIT Region |
Visible |
Combined Region | |||
r |
SECV |
r |
SECV |
r |
SECV | ||
Grainspec |
Tracker 3 (PLS) |
0.85 |
0.69 |
- |
- |
- |
- |
Unscrambler (PLS) |
0.93 |
0.78 |
- |
- |
- |
- | |
Infratec 1229 |
Unscrambler 6.1 (PLS) |
0.88 |
0.88 |
- |
- |
- |
- |
NIRS 6500 |
Winisi 2 (PLS) |
0.86 |
0.68 |
0.87 |
0.67 |
0.91 |
0.55 |
DA7000 |
Grams/32 (PLS) |
0.89 |
1.12 |
0.68 |
1.98 |
0.78 |
1.58 |
Figure 1. Scatter plot Actual L vs Predicted L from Grainspec NIR
Barley kernal colour may be measured objectively using NIR and NIT instruments. The overall results for the NIR calibrations were acceptable when compared to the precision of the Minolta method. Grain colour can be predicted using NIR as demonstrated in this study and confirms systems already in place (K Ayton pers comm). However, the results for the precision of the reference methods show that for NIR calibrations to be shared by a number of users a number of points should be confirmed. Firstly, the method with which the colour instrument observes the sample should be set. Secondly, the white tile for calibrating the Minolta should provide identical “L” values. Finally, the time when colour readings were recorded in relation to the NIR calibration scans is critical.
Acknowledgments
The following people are thanked for their assistance and advice Mr Tony Blakeney, Ms Rachael Jackson, Ms Hayfa Salman from BRI Australia, Mr Pat Wilson, Grainco, and Ms Jodi Orman, NSW GrainCorp Grower Services, and Mr John Ronalds from CSIRO Grain Quality Laboratory. Ms Jodi Orman and Mr Ken Saint, Australian Barley Board, are thanked for supplying samples. The Australian Grain Research and Development Corporation is acknowledged for funding this research through the Grain Industries Centre for NIR.
References
1. American Society of Brewing Chemists (1992) Methods of Analysis.
2. Baker, D.A. (Chairman ASBC Analysis Subcommittee) (1979) Journal of the American Society of Brewing Chemists, 37, 121.
3. Blakeney, A.B., Glennie-Holmes, M.R. and Taylor, H.R. (1994) Proceedings of 44th Australian Cereal Chemistry Conference, Ballarat, 319.
4. Blakeney, A.B. and Stuart, J. (1998) Cereal Foods World, 43, 532.
5. Buitendag, C., Lubben, A., Rabie, C., Rundall, P., Vogler, H. and Yhema, Z. (1994) Proceedings of the 23rd Institute of Brewing Convention – Australia and New Zealand Section, Sydney, 202.
6. Etchevers G.G., Banasik, O.J. and Watson, C.A. (1976) Cereal Chemistry 53, 846.
7. Keen, E. (1963) Proceedings of the Irish Maltsters Conference, 51.
8. Osborne, B.G. and Jackson, R. (1997) BRI Australia report.
9. Pepper, E.H. (1960) Ph.D thesis Michigan State University.
10. Svensson, E., Egelberg, P., Peterson, C. and Oste, R., (1994) Proceeding of the 26th Nordic Cereal Congress 74.