Automatic Milled Rice Quality Evaluation by Image Analysis

다국어 초록 (Multilingual Abstract)

This paper proposes an automatic evaluation method for the quality of milled rice. Shape descriptors determine the quantity of headrice, broken kernels, and brewers in milled 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) is then used to classify rice kernels according to rice defectives using this color features. The accuracy of the classifier is 94%. Linear regression model is developed for estimating individual kernel weight given a blob area. We obtained a very promising result with a coefficient of determination, R2 of 0.991 between the measured and estimated weight. The regression model may provide incorrect results when the blob area is less than the minimum threshold value(0.96㎜2).

참고문헌 (Reference)

1 I. Y. Zayas, "Wheat classification using image analysis and crush force parameters" 6 : 2199-2204, 1996

2 "Philippine Grains Standardization Program" National Food Authority 2002

3 B. K. Yadav, "Monitoring Milling Quality of Rice by Image Analysis" 33 : 19-33, 2001

4 R. Mukundan, "Moment Functions in Image Analysis: Theory and Applications: World Scientific"

5 "Microsoft WIA Library"

6 N. S. Visen, "Image analysis of bulk grain samples using neural networks" 46 : 7.11-7.15, 2004

7 N. S. Visen, "Identification and Segmentation of Occluding Groups of Grain Kernels in a Grain Sample Image" 79 : 159-166, 2001

8 B. Ni, "Design of an automated corn kernel inspection system for machine vision" 40 : 491-497, 1997

9 R. Choudhary, "Classification of cereal grains using wavelet, morphological, colour, and textural features of non-touching kernel images" 99 : 330-337, 2008

10 O. C. Agustin, "Automatic Milled Rice Quality Analysis" 11 : 112-115, 2008

11 O. C. Agustin, "Applications of Ward Network and GRNN for Corn Quality Classification" 5 : 218-225, 2007