Computers and Electronics in Agriculture
| Hi-index | 0.00 |
This paper proposes an automatic quality evaluation framework for milled rice kernels. Shape descriptors determine the quantity of headrice, broken kernels, and brewers in rice samples using six geometric features. Color histograms of rice kernels in RGB and Cielab color channels are used to extract 24 color features. A probabilistic neural network (PNN) classifier is used to categorize kernels according to rice defectives. The accuracy of the classifier is 94%. Linear regression model is also developed for estimating individual kernel weight given a blob area. Promising result was obtained with a coefficient of determination R2 of 0.991. The linear regression model provided excellent weight estimate when the blob area is greater than 1.0 mm2.