Please use this identifier to cite or link to this item:
https://rsuir-library.rsu.ac.th/handle/123456789/3164Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Rong Phoophuangpairoj | - |
| dc.contributor.author | Hong Chen | - |
| dc.date.accessioned | 2026-01-28T02:51:25Z | - |
| dc.date.available | 2026-01-28T02:51:25Z | - |
| dc.date.issued | 2025 | - |
| dc.identifier.uri | https://rsuir-library.rsu.ac.th/handle/123456789/3164 | - |
| dc.description | Thesis (M. Eng. (Electrical and Computer Engineering)) -- Rangsit University, 2024 | en_US |
| dc.language.iso | en | en_US |
| dc.publisher | Rangsit University. Library | en_US |
| dc.subject | Machine learning | en_US |
| dc.subject | Data structures (computer science) | en_US |
| dc.subject | Bananas -- Postharvest technology | en_US |
| dc.title | Determining banana ripeness using machine learning classifiers | en_US |
| dc.type | Thesis | en_US |
| dc.description.other-abstract | This study explored the application of machine learning models for classifying banana ripeness and predicting internal fruit qualities such as Brix (sweetness) and pH. Recognizing the inefficiency and subjectivity of traditional fruit quality assessment methods, the research aimed to develop accurate, scalable systems using advanced classification and prediction techniques. The study comprised two main parts. In the first part, four classifiers—MobileNet, ResNet50, a simple CNN, and VGG16—were evaluated for banana ripeness classification. MobileNet achieved the highest accuracy (98.45%), surpassing VGG16 (96.82%), CNN (95.79%), and ResNet50 (92.43%), demonstrating its superior performance in ripeness classification tasks. The second part investigated various prediction models for Brix and pH values, including linear regression, support vector regression (SVR), and k-nearest neighbors (KNN). Softmax features extracted via MobileNet were utilized for predictions. KNN demonstrated superior performance, attaining R² values of 0.984 for Brix and 0.972 for pH, surpassing linear regression and SVR, which yielded R² values between 0.925 and 0.958. Additional experiments using RGB, L*a*b*, and combined RGB and L*a*b* color values showed KNN’s superiority, with R² values of 0.947 for Brix and 0.896 for pH using RGB and L*a*b* color values | en_US |
| dc.description.degree-name | Master of Engineering | en_US |
| dc.description.degree-level | Master's Degree | en_US |
| dc.contributor.degree-discipline | Electrical and Computer Engineering | en_US |
| Appears in Collections: | Eng-ECE-M-Thesis | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| HONG CHEN.pdf | 1.57 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.