MACHINE LEARNING-BASED RICE PHENOLOGY MONITORING WITH SATELLITE-DERIVED VEGETATION INDICES

Sudana, Komang Harry (2025) MACHINE LEARNING-BASED RICE PHENOLOGY MONITORING WITH SATELLITE-DERIVED VEGETATION INDICES. Undergraduate thesis, Universitas Pendidikan Ganesha.

[img] Text (2115101053-COVER)
2115101053-COVER.pdf
Download (634kB)
[img] Text (2115101053-ABSTRAK)
2115101053 - ABSTRAK.pdf
Download (208kB)
[img] Text (2115101053-BAB 1 PENDAHULUAN)
2115101053-BAB 1 PENDAHULUAN.pdf
Download (156kB)
[img] Text (2115101053-BAB 2 KAJIAN TEORI)
2115101053-BAB 2 KAJIAN TEORI.pdf
Restricted to Repository staff only
Download (291kB) | Request a copy
[img] Text (2115101053-BAB 3 METODOLOGI PENELITIAN)
2115101053-BAB 3 METODOLOGI PENELITIAN.pdf
Restricted to Repository staff only
Download (193kB) | Request a copy
[img] Text (2115101053-BAB 4 HASIL DAN PEMBAHASAN)
2115101053-BAB 4 HASIL DAN PEMBAHASAN.pdf
Restricted to Repository staff only
Download (872kB) | Request a copy
[img] Text (2115101053-BAB 5 PENUTUP)
2115101053-BAB 5 PENUTUP.pdf
Restricted to Repository staff only
Download (120kB) | Request a copy
[img] Text (2115101053-DAFTAR PUSTAKA)
2115101053-DAFTAR PUSTAKA.pdf
Download (191kB)
[img] Text (2115101053-LAMPIRAN)
2115101053-LAMPIRAN.pdf
Download (239kB)

Abstract

Thailand's rice sector faces significant challenges in maintaining productivity, including low yields, labour shortages, and limited water resources. Thailand needs accurate data to guide effective farmland management and address emerging issues as a major global food producer. A comprehensive understanding of crop phenology, especially in rice, is essential to increase production and ensure sustainable management of agricultural resources. This research aims to develop an integrated approach to monitor rice phenology in Chachoengsao Province, Thailand, using multiple vegetation indices, including NDVI, EVI, LSWI, and BSI, combined with machine learning techniques. This study utilized three machine learning models: Support Vector Machine (SVM), Random Forest, and XGBoost. Sentinel-2 satellite imagery with a spatial resolution of 10 meters was used as the dataset, which incorporated 50 position points of rice plants for vegetation index extraction. The analysis showed characteristic patterns in vegetation indices (NDVI, EVI, LSWI, and BSI) corresponding to key phenological stages, where Start of Season (SOS) is characterized by increasing NDVI, EVI, and LSWI with decreasing BSI, Peak of Season (POS) shows peak values except the lowest BSI and End of Season (EOS) features decreasing NDVI, EVI, and LSWI with increasing BSI, reflecting the ageing and drying of rice plants. Among the three models developed, SVM showed the highest performance, achieving 99.4% accuracy, with macro average precision, macro average recall, and macro average f-1 score of 99.3%, 99.4%, and 99.3%, respectively. XGBoost ranked second with 98.8% on all evaluation metrics, while Random Forest ranked last with 97.6% accuracy, 97.9% for macro average precision, macro average recall, and macro average f-1 score. These findings demonstrate the effectiveness of machine learning in phenology analysis and highlight SVM as the most reliable model compared to Random Forest and XGBoost.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Machine Learning, Vegetation Index, Crop Phenology, Satellite Imagery
Subjects: T Technology > T Technology (General)
Divisions: Fakultas Teknik dan Kejuruan > Jurusan Teknik Informatika > Program Studi Ilmu Komputer (S1)
Depositing User: Komang Harry Sudana
Date Deposited: 11 Feb 2025 07:45
Last Modified: 11 Feb 2025 07:45
URI: http://repo.undiksha.ac.id/id/eprint/23107

Actions (login required)

View Item View Item