Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/165023| Title: | Perancangan Model Deep Learning untuk Deteksi Fenotipe dan Kematangan Buah Melon (Cucumis melo L.) pada Greenhouse |
| Other Titles: | Design of a Deep Learning Model for Phenotype and Ripeness Detection of Melon Fruit (Cucumis melo L.) |
| Authors: | Supriyanto Arham, Muhammad Syauqi |
| Issue Date: | 2025 |
| Publisher: | IPB University |
| Abstract: | Melon (Cucumis melo L.) merupakan tanaman hortikultura bernilai ekonomi tinggi yang memerlukan deteksi akurat pada fase generatif untuk optimalisasi budidaya. Metode deteksi manual memerlukan banyak waktu dan tenaga, sementara metode deteksi otomatis yang ada masih menghadapi tantangan dalam mengenali fase pertumbuhan dan membedakan tingkat kematangan buah. Penelitian ini bertujuan merancang model deteksi berbasis deep learning menggunakan algoritma YOLOv8 yang mampu mengidentifikasi buah melon pada dua tingkat kematangan (belum matang dan matang) serta bunga melon pada tiga tahap perkembangan (kuncup, mekar, dan layu). Metode penelitian meliputi pengumpulan dan penganotasian dataset, augmentasi data, pelatihan model, dan evaluasi kinerja menggunakan precision dan recall. Hasil penelitian menunjukkan bahwa model deteksi buah memiliki performa terbaik dengan precision 0,89, recall 0,89 mAP50 0,93, mAP50-95 0,79 diikuti model deteksi bunga dengan precision 0,78, recall 0,72, mAP50 0,81, mAP50-95 0,44, dan model gabungan dengan precision 0,80, recall 0,85, mAP50 0,87, mAP50-95 0,59. Berdasarkan penelitian ini dapat disimpulkan bahwa model cukup stabil dalam mendeteksi parameter pertumbuhan fase generatif tanaman dengan tepat sehingga model deep learning deteksi bunga dan buah pada tanaman melon bisa digunakan. Melon (Cucumis melo L.) is a high-value horticultural crop that requires accurate detection during the generative phase for cultivation optimization. Manual detection methods require significant time and effort, while existing automated detection methods still face challenges in recognizing growth phases and differentiating fruit maturity levels. This research aims to design a deep learning-based detection model using the YOLOv8 algorithm capable of identifying melon fruits at two maturity levels (unripe and ripe) and melon flowers at three developmental stages (bud, bloom, and wilted). The research method includes dataset collection, data annotation, data augmentation, model training, and performance evaluation using precision and recall. Research results show that the fruit detection model has the best performance with precision 0,80, recall 0,89, mAP50 0,93, and mAP50-95 0,79, followed by the flower detection model (precision 0,78, recall 0,72, mAP50 0,81, mAP50-95 0,44), and the combined model (precision 0,80, recall 0,85, mAP50 0,87, mAP50-95 0,59). Based on this research, it can be concluded that the model is quite reliable in detecting melon fruits and flowers, and the deep learning-based detection system for melon fruits and flowers can be used. |
| URI: | http://repository.ipb.ac.id/handle/123456789/165023 |
| Appears in Collections: | UT - Agricultural and Biosystem Engineering |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_F1401211085_b267ca120c87426681460cba1f21be0b.pdf | Cover | 398.25 kB | Adobe PDF | View/Open |
| fulltext_F1401211085_f0542d1331fd4bd986e434536cdea108.pdf Restricted Access | Fulltext | 3.2 MB | Adobe PDF | View/Open |
| lampiran_F1401211085_b8a293913d6f4bacbe2d5b453e604b2c.pdf Restricted Access | Lampiran | 219.67 kB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.