Penentuan Umur Tanaman Sawit Menggunakan Metode Artificial Neural Network dan Indeks Vegetasi Berbasis Sentinel-2

Abstract

Kelapa sawit menjadi varietas unggul dalam total produksi dan distribusi minyak nabati global pada tahun 2019/2020, yaitu mencapai volume sebesar 81,54 juta metrik ton (40%). Pengaruh positif kelapa sawit memberikan pertumbuhan yang sangat besar pada sektor sosial dan ekonomi. Penentuan umur tanaman sawit berbasis data penginderaan jauh sangat penting dalam manajemen perkebunan diantaranya perhitungan dosis pupuk dan penentuan waktu untuk peremajaan. Tujuan penelitian ini mengembangkan model yang dapat menentukan umur tanaman kelapa sawit pada lahan mineral dengan metode Artificial neural network MLPClassifier (ANN-MLPC) berbasis indeks vegetasi dari citra satelit Sentinel-2. Penelitian dilakukan pada Kebun Pendidikan dan Penelitian Kelapa Sawit IPB-Cargill Jonggol. Pengambilan data dilakukan dengan mengunduh citra satelit Sentinel-2 yang diolah menjadi nilai reflektansi tingkat grid. Nilai reflektansi tingkat grid digunakan sebagai variabel independen, sedangkan informasi umur aktual digunakan sebagai variabel dependen. Pembagian proporsi dataset dengan jumlah sampel 1500 data, dimana 80% digunakan untuk data training dan 20% untuk data testing. Berdasarkan hasil evaluasi, model yang dibangun dengan algoritma ANN-MLPC diperoleh nilai akurasi sebesar 97,47%, nilai presisi sebesar 93,87%, recall sebesar 93,81%, serta F1-score sebesar 93,84%. Hasil tersebut menunjukkan model machine learning yang digunakan mampu menentukan umur tanaman sawit pada lahan mineral secara akurat dan presisi.

Palm oil is the leading variety in total global vegetable oil production and distribution in 2019/2020, reaching a volume of 81.54 million metric tons (40%). The positive influence of oil palm provides enormous growth in the social and economic sectors. Determining the age of oil palm plants based on remote sensing data is very important in plantation management including calculating fertilizer doses and determining the time for replanting. The purpose of this research is to develop a model that can determine the age of oil palm plants on mineral land with the Artificial neural network MLPC Classifier (ANN-MLPC) method based on vegetation indices from Sentinel-2 satellite images. The research was conducted at the IPB-Cargill Jonggol Oil Palm Education and Research Farm. Data collection was carried out by downloading Sentinel-2 satellite images which were processed into grid-level reflectance values. The grid-level reflectance value was used as the independent variable, while the actual age information was used as the dependent variable. The proportion of datasets with a total sample size of 1500 data, where 80% and 20% of the total data samples are used traing and testing respectively. Based on the evaluation results, the model built with the MLPClassifier Artificial neural network algorithm has accuracy, precision, recall and F1-score values of of 97,47%, 93,87%, 93,81%, and 93,84% respectively. These results show that the machine learning model used is able to determine the age of oil palm plants on mineral land accurately and precisely.