Development of Machine Vision System for Size Classification of Potatoes (Solanum tuberosum L.)
Abstract
Machine vision was explored in this study to characterize potatoes that are locally available in the Philippines based on the Philippine National Standards (PNS). A machine vision hardware was fabricated and its software was developed using LabVIEW for estimating the weight of potato and classifying them according to PNS. A weight estimation equation was determined by conducting polynomial regression analysis on the weight and projected area from 155 potato samples. Furthermore, test for measurement consistency was done to evaluate the capability of the developed MVS software to yield consistent readings on thirty sampling angles. Fifty potato samples were used on testing for sorting accuracy. Results of testing yielded a high R2 value of 98.86% with p-value of 2.2 x 10-16. Moreover, results revealed a low coefficient of variation of 0.33% and had no significant difference at α= 5%. Test for accuracy, on the other hand, exhibited a computed individual accuracy of 96.2% on the medium size, and 100% on the rest. Overall accuracy of the developed MVS software was computed at 98%, where the incorrect classification was noted to occur near the boundary of the sizes. The developed MVS projected capacity was computed at 7200 samples per hour, excluding time to load and unload potatoes.
Citation:
QUILLOY, E., SANCHEZ, P. R., MANUEL, L. J., & RENOVALLES, E. (2019). Development of Machine Vision System for Size Classification of Potatoes (Solanum tuberosum L.). Philippine Journal of Agricultural and Biosystems Engineering, 15(2), 23–30. https://doi.org/10.48196/015.02.2019.03
