Click on “Download PDF” for the PDF version or on the title for the HTML version.


If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

CLASSIFICATION OF CEREAL GRAINS USING MACHINE VISION: IV. COMBINED MORPHOLOGY, COLOR, AND TEXTURE MODELS

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  Transactions of the ASAE.  VOL. 43(6): 1689-1694 . (doi: 10.13031/2013.3069) @2000
Authors:   S. Majumdar, D. S. Jayas
Keywords:   Machine vision, Digital image processing, Pattern recognition, Morphology, Texture, Color, Automated grain grading, Discrimination, Cereal grain classification, Wheat, Barley, Oats, Rye, Automation, T

Classification models by combining two or three feature sets (morphological, color, textural) were developed to classify individual kernels of Canada Western Red Spring (CWRS) wheat, Canada Western Amber Durum (CWAD) wheat, barley, oats, and rye. The mean accuracies (the average of the classification accuracies of the above mentioned cereal grains) of 98.6 and 99.3% were achieved when the morphology-texture model with the 15 most significant features was used to test on an independent data set (total number of kernels used was 10,500) and on the training data set (total number of kernels used was 31,500), respectively. When the morphology-color model (with the 15 most significant features) was tested on the independent data set and on the training data set, the mean accuracies were 99.4 and 99.6%, respectively. Similarly, using the texture-color model (with the 15 most significant features) the mean accuracies were 98.4 and 98.0% for the independent and the training data sets, respectively. The highest classification accuracies were achieved when the morphology-texture-color model was used. The mean accuracies using the 20 most significant features in the morphology-texture-color model were 99.7 and 99.8% when tested on the independent and the training data sets, respectively. The differences in mean accuracies were not significant when the models were tested with independent and training data sets.

(Download PDF)    (Export to EndNotes)