Previous PageTable Of ContentsNext Page

Measuring rice grain dimensions with an image analyser

B.G. Armstrong1,2,4, G.P. Aldred1,4 , T.A. Armstrong1,2, A.B. Blakeney3,4 and L.G. Lewin4

1 Institute of Food and Crop Science, University of Ballarat, Ballarat, VIC, 3353
2
SeedCount Australasia Pty Ltd, PO Box 236, Creswick, VIC, 3363
3
Cereal Solutions Pty Ltd, PO Box 201, North Ryde, NSW, 1670
4
CRC for Sustainable Rice Production, Yanco Agricultural Institute, Yanco, NSW, 2703

Introduction

Appearance is a critical quality attribute for both milled and brown rice. Rice buyers, Millers and Consumers judge the quality of rice on the uniformity of its size and shape as well as the pleasing appearance of its overall size-shape relationship. The measurement of grain length, width and thickness has traditionally been done manually using a microscope, vernier callipers (Ikehashi and Kush 1979) or by projection of an image of the grain onto a solid screen with an overhead projector and measuring the image with a transparent plastic ruler. These classic methods using manual measurement, even with modern electronic, digital callipers, are slow, tedious and limit the number of samples and sub-samples that can be evaluated.

It is likely that fast, accurate measurement of the physical dimensions of rice would result in breeding lines showing greater uniformity. This may not only lead to rice with better appearance but also increase whole grain milling yields as more uniform grains polish more evenly in rice mills.

Image analysis using a camera system is currently used in the rice-breeding programme at Yanco Agricultural institute (Martin et al 1997). The method requires careful placement of grains sampled with a vacuum based positioning funnel and only makes measurements in two dimensions. In addition the available software uses algorithms that do not measure the real length and breadth but a diagonal approximation.

Rice grain size and shape is critical in breeding new varieties, as each variety must fit an existing market class. In Australia these classes are usually based on an existing variety. Size and shape classification of rice has been reviewed by Bhattacharya and Sowbbagya (1980). We report on the use of a specially developed imaging tray for use with a SeedCount image analyser that allows easier collection of data from large numbers of grains. The tray (patent pending) holds some kernels on their ‘side’, allowing direct observation of kernel thickness (Armstrong et al 2003).

Materials amd methods

Rice samples were from pure seed samples provided through the Rice Appraisals Laboratory of SunRice in Leeton NSW. Samples from the 2003-04 crop were taken from all currently grown varieties from locations throughout the rice growing area. We report here on samples of medium grain (Amaroo, Jarrah, Millin and Quest) and arborio (Illabong) rice; varieties that make up the majority of the Australian rice crop. Both brown and milled (white polished) data is presented.

Kernels were hand measured with “Stainless” digital callipers (accuracy 0.01mm), transferred to the medium grain rice tray, scanned at 300 dpi and analysed with a SeedCount SC3 DIA (Digital Image Analysis) system (Figure 1). The SeedCount software uses algorithms that measure the true kernel length, width and thickness. The results of both methods were then compared.

The DIA system was also used to examine larger samples, as up to 1350 rice kernels can be examined per tray. This data was used to compare the dimensional distributions within and between grain lots and to briefly examine the effects of milling on rice.

Figure 1: Indented Rice Tray on Imaging Cabinet

Results and discussion

A comparison of the calliper and DIA thicknesses for 200 brown rice kernels from ten different samples are shown in Figure 2. Each diagonal series represents one of the sample sets. The kernels were selected for the maximum range for each sample. The accuracy of the length, width and thickness measurements for brown and white rice are detailed in Table 1.

Figure 2: Comparison of Medium/Arborio Brown Rice Thickness measured by callipers (Actual) and DIA (Estimate using Multivariate SeedCount Eqn), sorted by cultivar.

Table 1: Comparison of Manual and DIA Dimensional Measurements

Length

Width

Thickness

Roundness

Brown Std Error (mm)

0.086

0.085

0.063

0.018

Brown Correlation (r)

0.99

0.94

0.97

0.92

White Std Error (mm)

0.089

0.109

0.078

0.021

White Correlation (r)

0.98

0.91

0.95

0.87

Table 2 summarises the DIA measurements for full-tray samples of various varieties. It can be seen that there are distinct differences between varieties. Consistent changes in the grain properties resulting from milling the rice can also be seen, confirming that during milling the projecting portions of the grain (eg the ends) are milled more heavily than other portions (eg the sides as shown by the small reductions in thickness). The usual rice industry Aspect Ratio formula (Length/Width) is used and the SeedCount Roundness equation has been re-written to coordinate with these Aspect Ratios, and is now:

Table 2 confirms that Australian medium grain rice fits the U.S. medium grain rice breeding criteria, while the Arborio rice is too round for the medium class and too large for the short grain class (Webb 1985). U.S. Standards are often used in world trade.

Table 2: Dimensional Relationships of Australian Medium and Arborio Rice Varieties

Sample ID

Length Mean

Length StdDev

Width Mean

Width StdDev

Thickns Mean

Thickns StdDev

Aspect Ratio

Round- ness

Amaroo Brown

5.74

0.30

2.78

0.13

1.74

0.12

2.07

2.32

Amaroo White

5.42

0.26

2.70

0.13

1.67

0.09

2.01

2.29

Amaroo Diff

0.32

0.04

0.08

0.00

0.07

0.02

0.06

0.03

Illabong Brown

6.08

0.37

3.16

0.22

2.02

0.18

1.92

2.17

Illabong White

5.72

0.32

3.07

0.21

1.97

0.15

1.86

2.11

Illabong Diff

0.36

0.05

0.09

0.01

0.05

0.03

0.06

0.05

Jarrah Brown

5.89

0.31

2.80

0.13

1.80

0.12

2.10

2.31

Jarrah White

5.45

0.25

2.69

0.12

1.71

0.11

2.03

2.26

Jarrah Diff

0.44

0.06

0.11

0.01

0.09

0.01

0.08

0.05

Millin Brown

5.74

0.36

2.78

0.16

1.71

0.12

2.07

2.35

Millin White

5.45

0.31

2.69

0.15

1.64

0.09

2.03

2.33

Millin Diff

0.29

0.05

0.09

0.02

0.07

0.03

0.04

0.03

Quest Brown

6.22

0.38

2.86

0.17

1.87

0.14

2.17

2.35

Quest White

5.79

0.29

2.73

0.15

1.76

0.10

2.12

2.33

Quest Diff

0.43

0.10

0.13

0.03

0.11

0.04

0.05

0.02

Overall Diff

0.37

0.06

0.10

0.01

0.08

0.03

0.06

0.04

Additional data can also be retrieved from SeedCount that pertains to the classification of rice and highlights the milling changes in rice, as shown in Table 3. TKW As-is is the Thousand Kernel Weight of the rice on an As-Is moisture content basis. The Mini Test Weight is assessed on a sample cup of 26 ml, while the Average Seed Area uses only whole seeds in the ‘wide’ area of the tray.

Table 3: Further Relationships of Australian Medium and Arborio Rice Varieties

SampleID

TKW As-is

Mini Test Weight

Avg Seed Area

Amaroo Brown

21.7

85.5

11.83

Amaroo White

19.9

85.6

10.33

Amaroo Diff

1.8

-0.1

1.50

Illabong Brown

28.9

85.3

14.36

Illabong White

26.6

87.0

12.62

Illabong Diff

2.3

-1.7

1.74

Jarrah Brown

22.4

79.3

12.30

Jarrah White

20.3

87.1

10.40

Jarrah Diff

2.1

-7.8

1.90

Millin Brown

20.9

80.4

11.90

Millin White

19.0

84.1

10.30

Millin Diff

1.9

-3.7

1.60

Quest Brown

24.1

80.9

13.30

Quest White

22.5

85.1

11.35

Quest Diff

1.7

-4.2

1.95

Overall Diff

2.0

-3.5

1.74

On a TKW basis, once again the medium grain rices comply with the US standards, while the Arborio rice is too heavy. Perhaps one interesting result of the milling process is the tendency for milled rice to have a higher test weight than brown rice, probably due to the milled rice being smoother and more uniformly shaped, allowing denser packing. However, the effect is patchy, while changes in the kernel weight and average seed area are quite consistent and could be used as check on the degree of milling achieved.

Conclusions

The DIA system trialled was able to measure the three cardinal dimensions of brown and white rice with reasonable accuracy. Reducing the pixel size (eg 600dpi) when imaging should increase this accuracy. The strong correlation between Aspect Ratio and Roundness (r = 0.73) suggests that Roundness could be a useful parameter when defining rice standards. Changes in rice cardinal dimensions, TKW and area could be helpful in determining the degree of milling achieved and form the basis of a degree of milling index.

Acknowledgments

Thanks to the staff of the Rice Appraisals Laboratory for providing the rice samples.

References

Armstrong, B., Weiss, M., Greig, R. I., Dines, J., Gooden, J., & Aldred, G. P. (2003). In Black, C.K. and Panozzo, J.F. Eds. Proceedings of the53rd Australian Cereal Chemistry Conf, Glenelg. 246-249.

Bhattacharya, K. R. and Sowbhagya, C.M. (1980) Il Riso, 29 181-185.

Ikehashi, H. and Kush, G.S. (1979) Methodology of assessing appearance of rice grain, including chalkness and whiteness. In Proceedings of the Workshop on Chemical Aspects of Rice Grain Quality. International Rice Research Institute, Los Banos. 223-229.

Martin, M. Hart, K.R. Blakeney, A.B. and Lewin L.G. (1997) In Tarr, A.W., Ross, A.S. and Wrigley, C.W. Proceedings of the 47th Australian Cereal Chemistry Conference, Perth. 332-335.

Webb, B.D. (1985) Criteria of Rice Quality in the United States. In Julino B.O. Ed. Rice Chemistry and Technology, AACC, St. Paul. Minn. 403-442.

Previous PageTop Of PageNext Page