Characteristics of Spectral Values and Spatial Distribution of Oil Palm Health in Relation to Productivity at PT Kayung Agro Lestari Plantation
Atysatya Prawira Adjas (a), Masita Dwi Mandini Manessa (a), Iqbal Putut Ash Shidiq (a),

a. Department Geography, Faculty Mathematic and Science, Universitas Indonesia, Indonesia

Abstract

Oil palm plantations support the non-oil and gas economy in Indonesia, but in business practice oil palm plantations still face challenges in maintaining the health of oil palm plants. Currently, one of the main problems is the large number of unhealthy trees that can directly affect neighboring trees. If an overview of the plantation area can be identified quickly and in detail, this information can be integrated into advanced planning in oil palm planta-tion management. To address this problem, this research presents a system for detecting the health of oil palm plants by utilizing deep learning tech-niques in collaboration with UAV orthophotos and descriptive analysis to see the relationship to the productivity of the oil palm fruit produced. The appli-cation of semantic segmentation using the U-net algorithm aims to identify the palm tree canopy, eliminate the bias between the tree canopy and the ob-jects below it and facilitate the classification of plant health. In the next stage, the algorithm used is a support vector machine (SVM) for accurate plant health classification with a fairly high value. The evaluation results of the classification of oil palm plant health gave satisfactory results with F1-score values in blocks D52 and D53 of 71.4% and in blocks E58 and E59 reaching 81.3%. It is known that the difference in the value of the F1-score results is influenced by the quality of the multispectral aerial photographs used. In multispectral aerial photographs in blocks D52 and D53 there is some noise that makes the difference in the value obtained. The relationship between oil palm plant health conditions and oil palm fruit productivity is carried out using productivity data in oil palm plant blocks with vulnerable months from January to June in 2024. The relationship between oil palm health conditions and productivity shows a tendency that the number of un-healthy trees correlates with decreased productivity, as seen in blocks D52 and E58. Conversely, blocks with fewer unhealthy trees, such as D53, show increased productivity. Although pests and diseases have been shown to have a significant impact, productivity is also influenced by other factors, such as land characteristics, climate, and harvest management, so a holistic approach is needed for optimal management.

Keywords: Health Mapping- Oil Palm Plants- Productivity- Semantic Segmentation.

Topic: Topic B: Applications of Remote Sensing