Zhang, Yinping and Chuah, Joon Huang (2025) An intelligent grading system for mangosteen based on improved convolutional neural network. Knowledge-Based Systems, 309. p. 112904. ISSN 0950-7051, DOI https://doi.org/10.1016/j.knosys.2024.112904.
Full text not available from this repository.Abstract
This study addresses cognitive challenges in mangosteen production and processing by developing an efficient grading mechanism. A specialized hardware unit was constructed to evaluate mangosteen based on external characteristics, utilizing image preprocessing techniques for color attribute extraction using the RGB model, contour extraction through image segmentation, and diameter measurement via ellipse fitting. We propose a novel convolutional neural network (CNN) model, MobileSeNet, which outperforms MobileNetV2 in average Precision, Recall, and F1-score metrics, achieving a grading accuracy of 98.13 %. Furthermore, MobileSeNet processes images in an average of 76 ms, which is 48 ms faster than MobileNetV2, underscoring its efficiency. Additionally, we developed a user-friendly human-computer interaction interface to enhance usability and minimize cognitive challenges, optimizing user experience and contributing to improved task execution and reduced errors. This comprehensive approach establishes a robust foundation for future advancements in automated grading systems, thereby enhancing productivity in the agricultural sector.
| Item Type: | Article |
|---|---|
| Funders: | UNSPECIFIED |
| Uncontrolled Keywords: | Mangosteen; Intelligent grading; Deep learning; Convolutional neural network |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Faculty of Engineering > Department of Electrical Engineering |
| Depositing User: | Ms. Juhaida Abd Rahim |
| Date Deposited: | 12 Mar 2025 04:22 |
| Last Modified: | 12 Mar 2025 04:22 |
| URI: | http://eprints.um.edu.my/id/eprint/47787 |
Actions (login required)
 |
View Item |
