DisplayTitle

Effects of Sampling and Wheat Grade on Precision and Accuracy of Kernel Features Determined by Digital Image Analysis

¹Department of Food Science, University of Manitoba, Winnipeg, MB, R3T 2N2, Canada. ²Corresponding author. E-mail: harry_sapirstein@umanitoba.ca ³Present address: Policy Branch, Agriculture & Agri-Food Canada, 500-303 Main St., Winnipeg, MB, R3C 3G7, Canada.

The effect of sampling on the precision and accuracy of digital image analysis of different commercial sample grades of Canada Western Red Spring (CWRS) wheat was investigated. Kernel perimeter, length, width, and area measurements were used to determine mean and dispersion statistics for composite railcar CWRS samples of No. 1, 2, and 3 grades; the numbers of railcars sampled were 27, 40, and 36, respectively. Sample sizes ranged from 10 to 2,000 kernels. Instrumental measurement precision was routinely better than 0.1 mm for macroview images, with a resolution of 0.0054 cm² per pixel. Computed mean kernel feature measurements and dispersion statistics were highly dependent on sample size and grade. Comparative analysis of wheat samples by digital imaging of individual kernels required a sample of no less than 300–500 kernels, depending on sample grade, for accurate representation of the parent sample. This level of sampling resulted in detection of significant differences (P < 0.05) in mean kernel features that, on average, differed by <1%. Except for some samples containing low numbers of kernels, lower grade wheat had more variable kernel features compared with higher grade samples. In relative terms, for comparably sized samples (≥133 kernels), variance in No. 2 grade wheat was 6–11% higher that for No. 1 grade wheat, depending on kernel feature. Similarly, variance in No. 3 grade wheat was 13–23% higher than for No. 2 grade wheat and 20–37% higher than for No. 1 grade wheat, indicating that wheat grading has a predictable effect on and is influenced by the uniformity of kernel characteristics in a sample. The ability of digital image analysis to detect these effects reflects the potential of this technology for use in objective classification of wheat according to grade.