Abstract: Plumule ratio is of the most important criterion for evaluating the quality of plumule rice. A dual structure neural network classifier was developed which consisted of two parallel identifiers(one per type)followed by a comparing selector. Images of rice kernels were captured using a machine vision system. The identifiers were individually trained using physical attributes of rice kernel extracted from their images as the input. Then the classifier can be used to classify two types of rice kernel(kernel with or without plumule).And the plumule ratio of rice can be measured automatically. Tests showed that classification accuracy was high and the classifier was robust.