A real-time white ginseng quality evaluation system based on a machine vision
technique and artificial neural networks was developed to replace the current manual grading and
its efficiency was tested. The system consisted of conveyor, image acquisition system
synchronized with a sample-detecting sensor, and image processing and decision-making system.
Software running under Windows system was developed. The algorithm included three
consecutive stages of (a) image acquisition and preprocessing, (b) mathematical feature extraction,
and (c) grade decision using artificial neural networks. Mathematical features such as area ratio,
mean and standard deviation of gray level, skewness of gray level histogram, and the number of
run segment, were extracted from five equally divided parts of a specimen. An artificial neural
network model was used to classify samples into three grading categories. The grading error of the
system was about 26%, which is comparable to the 30% in case of manual grading. The grading
rate was one sample per a second.