A machine vision system was used to collect color, morphological, and textural information from soybean seed surfaces. Soybean samples had been infected by Cerco- spora spp., Alternaria spp., Fusarium spp., Phomopsis spp., Soybean Mosaic Virus, or were green immature soybeans or asymptomatic.

A statistical classifier achieved overall accuracies of 91% using feature sets of color, morphology, and texture. On individual seed levels, asymptomatic seed was detected with 100% accuracy, Cercospora spp. with 90% accuracy. The color feature set provided 81% accuracy for Soybean Mosaic Virus (black) and the textural feature set classified Phomopsis spp. with 93% accuracy.

The degree of whiteness of corn kernels was measured using machine vision. Red, green, and blue images of kernels were converted to YCrCb (luminance, chrominance red, and chrominance blue). Kernel germ regions were deleted from images to obtain comparable results regardless of whether the germ was up or down.

Whiteness index values averaged between 18 to 80 as corn samples varied from deep yellow/orange to bright white, with above 54 indicating white corn. After training, human-inspected samples were ranked in color from yel- low to white. Machine vision inspections of the samples maintained the same order of ranking as the human-inspected samples and were more consistent than the human inspector.