Neural network-based quality evaluation of germinated Hang rice

Abstract

Germinated Hang rice is found widely in the northeast of Thailand. It is produced by the traditional folklore wisdom and high nutritional value to the human body. Hence, its quality is very crucial. Traditionally, quality of rice grains has been assessed manually. Apparently, this method is very time consuming and it highly relies on human skills and experience. Thus, the purpose of this research is to develop an image processing algorithm incorporated with a neural network classifier that can detect the following defects of geminated Hang rice grains: broken grains, discolored grains, un-husked paddy grains, deformed grains and withered grains. These defects do not exist in general milled rice grains. Thus, twenty-four features composed of nine grain color components, five grain shape parameters, and ten grain textural features are extracted from images. Then, these features are fed into the neural network classifier. As shown in the results, percent accuracy of our proposed method was 98.0%.