Discrimination of wheat class and variety by digital image analysis of whole grain samples

Abstract

The application of digital image analysis for objective classification of wheat cultivars according to kernel type and identity was studied. Plan-form spatial shape features and Fourier descriptors of kernel perimeters were automatically determined from computed whole-grain contours. Perfect type classification was obtained for four durum wheat varieties in admixture with ten common wheats representing a broad range of kernel types. Likewise, samples of each of five Canada Western Red Spring (CWRS) wheat varieties were all correctly allocated to the CWRS class. None of 432 kernels of contrasting wheat classes was misclassified as CWRS wheat. Discrimination of Canada Western Red Winter, Canada Western Soft White Spring, Canada Utility and Canada Prairie Spring wheats was less satisfactory. Discrimination of varieties within classes gave inconclusive results with correct classification scores ranging from 15 to 96%.