e-ksper: A Convolutional Neural Network Based System for Seedless Raisin Quality Grading

No Thumbnail Available

Date

2023

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Seedless raisins are graded according to their quality which is determined based on several features such as color, size, texture, and humidity. Currently, most of the raisin grading process is performed by human experts manually, which is laborious and subjective work. Therefore, an automated system that enables objective evaluation of the raisins would be helpful for both producers and experts during this process. In this study, we propose a simple machinery prototype that takes images of raisins under standard background and illumination conditions and an automated system that performs quality grading of raisins using convolutional neural networks. The proposed model not only targets color classes but also aims to identify foreign matters and defected kernels. The model achieves about 88.2% average classification accuracy on five classes including four color classes and a defected kernels class; whereas the model's accuracy becomes 98.6% if defected kernels are excluded. Hence, the proposed model is very successful in differentiating colour classes and has also considerable success in detecting foreign matters and defected raisins. We provide a comprehensive analysis and discussion on these results.

Description

Keywords

Citation