Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/170984| Title: | Implementasi Algoritma YOLO Untuk Identifikasi Jenis Tanaman Pada Indoor Farming Menggunakan Crazyflie |
| Other Titles: | |
| Authors: | Priandana, Karlisa Hardhienata, Medria Kusuma Dewi Rafif, Muhammad Rangga |
| Issue Date: | 2025 |
| Publisher: | IPB University |
| Abstract: | Pertanian dalam ruangan (indoor farming) merupakan solusi atas keterbatasan lahan, cuaca ekstrem, dan degradasi kesuburan tanah yang dihadapi sistem pertanian konvensional. Namun, identifikasi jenis tanaman secara manual dalam skala besar masih memerlukan waktu dan tenaga yang signifikan. Sampai saat ini belum terdapat penelitian yang menggunakan algoritma YOLO untuk identifikasi jenis tanaman dengan citra yang dihasilkan oleh kamera AI-deck pada UAV jenis Crazyflie. Oleh karena itu, penelitian ini bertujuan menganalisis hasil model deteksi tanaman berbasis YOLOv11 menggunakan citra dari AI-deck Crazyflie. Data diperoleh dari Laboratorium Smart Urban Farming I-Surf IPB University dengan objek cabai, kangkung, dan anggur, kemudian melalui tahapan pelabelan, pembagian data, penanganan ketidakseimbangan kelas dengan undersampling serta kombinasi undersampling–oversampling, pelatihan model, dan evaluasi. Hasil menunjukkan bahwa pendekatan kombinasi menghasilkan performa terbaik dengan presisi sebesar 97%, 96%, dan 94%, recall 97%, 99%, dan 93%, serta akurasi 92% untuk anggur, cabai, dan kangkung secara berturut-turut, sedangkan pendekatan undersampling murni memperoleh presisi 94%, 96%, dan 92%, recall 94%, 98%, dan 92%, serta akurasi 91%. Simpulan penelitian ini adalah kombinasi undersampling–oversampling lebih efektif untuk meningkatkan kinerja deteksi tanaman, dan implementasi lebih lanjut diperlukan untuk integrasi model ke sistem navigasi UAV otonom. Indoor farming is a solution to challenges in conventional agriculture, such as limited land, extreme weather, and soil fertility degradation. However, manual plant species identification in large-scale systems still requires significant time and labor. To date, no research has applied the YOLO algorithm to identify plants species using images captured by the AI-deck camera on a Crazyflie UAV. Therefore, this study aims to analyze the results of a YOLOv11-based plant detection model using images from the Crazyflie AI-deck. Data were collected at the Smart Urban Farming Laboratory I-Surf IPB University with chili, water spinach, and grape plants as objects, followed by annotation, data splitting, handling class imbalance through undersampling and a combination of undersampling–oversampling, model training, and evaluation. The results showed that the combined approach achieved the best performance with precision of 97%, 96%, and 94%, recall of 97%, 99%, and 93%, and accuracy of 92% for grape, chili, and water spinach respectively, while the pure undersampling approach achieved precision of 94%, 96%, and 92%, recall of 94%, 98%, and 92%, and accuracy of 91%. This study concludes that the combined undersampling–oversampling method is more effective in improving plant detection performance, and further implementation is required for integration into autonomous UAV navigation systems. |
| URI: | http://repository.ipb.ac.id/handle/123456789/170984 |
| Appears in Collections: | UT - Computer Science |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_G6401211083_84fa3773e67e4bc7b57e9410cbb23ea3.pdf | Cover | 506.58 kB | Adobe PDF | View/Open |
| fulltext_G6401211083_ed8e7439e0174857843c0dd8b4f0831b.pdf Restricted Access | Fulltext | 1.22 MB | Adobe PDF | View/Open |
| lampiran_G6401211083_2e8311a490f3474aad21778ebe527e24.pdf Restricted Access | Lampiran | 288.82 kB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.