Cereal Chem 68:357-361   |  VIEW ARTICLE

Classification of Wheat Kernels Using Three-Dimensional Image Analysis.

W. H. Thomson and Y. Pomeranz. Copyright 1991 by the American Association of Cereal Chemists, Inc. 

A computer-controlled laser scanning system has been developed to acquire three-dimensional images of the surface of cereal grains. Each picture element represents the elevation of the kernel surface above the reference plane. In addition, a separate image is acquired in which each picture element represents the intensity of the light reflected from the kernel surface. The dual images are then used for feature extraction. The system was used to distinguish between kernels of the wheat cultivars Daws (soft white winter) and Tyee (club). A combination of 14 features based on nine topographic images and five intensity images permitted discriminate analysis to correctly classify 92-94% of the kernels. The system was also used to identify sprout damage in harvested wheat kernels. A topographical image of the kernel surface helps to distinguish between sprouted and unsprouted kernels. In particular, features that measure the deformation of the germ end of the kernel are crucial to the discrimination process. A discriminant model based on four features correctly identified 89% of the sprouted kernels and 83% of the unsprouted kernels when applied to an independent set of test kernels.