Pengembangan Metode Analisis Kematangan Padi Berdasarkan Indeks Vegetasi dari Citra UAV RGB Menggunakan Algoritma Support Vector Machine (SVM)

Tasidin, Muhammad Hidayat (2025) Pengembangan Metode Analisis Kematangan Padi Berdasarkan Indeks Vegetasi dari Citra UAV RGB Menggunakan Algoritma Support Vector Machine (SVM). Other thesis, Institut Teknologi Sepeuluh Nopember.

Text 5009211002-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Pemantauan kematangan padi penting untuk menentukan waktu panen yang optimal, namun metode konvensional masih bersifat subjektif dan kurang efisien pada lahan skala luas. Penelitian ini bertujuan untuk mengembangkan metode klasifikasi tingkat kematangan padi varietas IR64 menggunakan citra RGB dari UAV dan algoritma Support Vector Machine (SVM). Adapun fitur citra yang digunakan meliputi indeks vegetasi dari kanal RGB dan fitur tekstur menggunakan Gray-Level Co-occurrence Matrix (GLCM). Klasifikasi dibagi menjadi dua kategori yaitu “belum matang” dan “matang”. Hasil dari model SVM mampu mencapai akurasi pelatihan sebesar 97,11%, akurasi pengujian 95,42%, dan skor dari cross validation sebesar 96,94% ± 0,20%. Kelas “belum matang” memperoleh precision 0,99 dan F1-score 0,96, sedangkan kelas “matang” memiliki recall 0,99 dan F1-score 0,93. Adapun Bagan Warna Daun (BWD) konsisten berada pada nilai 4 hingga 80 HST, meskipun mulai terdeteksi penurunan ke nilai 3 di beberapa titik. Citra RGB menunjukkan penurunan indeks vegetasi seperti ExGR secara lebih jelas pada 80 HST, yang mengindikasikan bahwa penginderaan jauh lebih sensitif dalam mendeteksi perubahan awal kematangan dibanding observasi visual. Adapun pengaruh tinggi tanaman menunjukkan kecenderungan stagnan pada 80 HST dan tidak memberikan kontribusi signifikan dalam membedakan fase kematangan.
================================================================================================================================
Monitoring rice maturity is essential for determining the optimal harvest time; however, conventional methods remain subjective and inefficient, particularly over large-scale fields. This study aims to develop an classification method for the maturity stages of IR64 rice using RGB imagery captured by UAVs and the Support Vector Machine (SVM) algorithm. The image features used include vegetation indices derived from RGB channels and texture features extracted using the Gray-Level Co-occurrence Matrix (GLCM). The classification was divided into two categories: “immature” and “mature.” The SVM model achieved a training accuracy of 97.11%, a testing accuracy of 95.42%, and a cross-validation score of 96.94% ± 0.20%. The “immature” class yielded a precision of 0.99 and an F1-score of 0.96, while the “mature” class achieved a recall of 0.99 and an F1-score of 0.93. The Leaf Color Chart (LCC) consistently remained at score 4 up to 80 Days After Transplanting (DAT), although some areas showed a decline to score 3. The RGB imagery revealed a more noticeable decline in vegetation indices such as ExGR at 80 DAT, indicating that remote sensing is more sensitive in detecting early maturity changes compared to visual observation. Plant height showed a stagnating trend at 80 DAT and did not significantly contribute to distinguishing between maturity phases.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Citra RGB, Gray-Level Co-occurrence Matrix, Indeks Vegetasi, Kematangan Padi, Support Vector Machine (SVM), UAV, Gray-Level Co-occurrence Matrix, Rice Maturity, RGB Imagery, Support Vector Machine (SVM), UAV, Vegetation Indices
Subjects:	S Agriculture > S Agriculture (General) T Technology > T Technology (General) T Technology > TR Photography > TR810 Aerial photography
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Hidayat Tasidin
Date Deposited:	25 Jul 2025 07:37
Last Modified:	25 Jul 2025 07:37
URI:	http://repository.its.ac.id/id/eprint/120245

Actions (login required)

View Item