Predicting Nitrogen Content in Oil Palms through Machine Learning and RGB Aerial Imagery

Abstract

—Nitrogen is a crucial nutrient for the sustainable health and productivity of oil palm plantations. Accurate fertilization for Nitrogencan optimize production while reducing maintenance costs. This study investigates the relationship between various vegetation indices and oil palm Nitrogen content using aerial images. We employ and compare different machine learning algorithms to predict Nitrogen content in oil palms, utilizing RGB aerial images obtained from PT. Perkebunan Nusantara IV (PTPN IV) in North Sumatra. Twelve vegetation indices are assessed, considering the limited spectral information available from the aerial images. Our findings reveal that the random forest algorithm, when applied to Hue, Green Leaf Index, and Coloration Index, yields the highest prediction accuracy of 90.13%. Furthermore, the results demonstrate that machine learning algorithms can effectively overcome the limitations of near-infrared channel availability, allowing for the prediction of Nitrogen content using RGB aerial images as a proxy for chlorophyll absorption.