De-husked coconut quality evaluation using image processing and machine learning techniques

Abstract

Qualitative evaluation provides the basis for determining if the quality of products meets the target specifications. Manual evaluation of de-husked coconuts is still being performed by coconut farmers, however, it is time consuming and costly. Ergo this study aiming to replace the manual inspection, a prototype was developed for objective and automated quality evaluation of de-husked coconuts through the application of computer vision and machine learning, identifying good-quality de-husked coconuts from defective ones with respect to its RGB color space. JavaFX platform was utilized to create the system performing K-Nearest Neighbor and Arduino technology played a significant role in the hardware control of the device. The image samples were captured by a CMOS camera in an imaging chamber with invariant illumination on top of a conveyor belt. Image processing is done to get the required features of the sample and by comparing the average RGB value from the custom dataset, then the maturity level of the coconut is determined. With the accuracy of 86.67%, the system is able to evaluate de-husked coconuts which are good for further processing used in export and premature coconuts that are to be rejected. © 2019 ACM.