Investigation of Classification Algorithm for Identification of Oil Palm Plantation Using Multiscatter and Multiresolution SAR Data

Abstrak

Oil palms are plants with green leaves, it is very difficult to differentiate between oil palms and forests only by color grouping. It is therefore important to recognize and distinguish species of oil palm trees specifically from other tree and forest species for proper planting and management, one of them is by usage of the classification algorithm. The aim of this research is to identify the best classification algorithm for the determination of oil palm plantations using radar satellite imagery. The methodology including primary and secondary data collection, primary data are using Synthetic Aperture Radar (SAR) images including Sentinel-1A, ALOS PALSAR-2 and TerraSAR-X, while secondary data is the oil palm plantation training field. The processing of SAR data was involved radiometric and geometric correction and scattering calibration. The classification technique were using the algorithm of parallelepiped, support vector machine (SVM), and maximum likelihood. From three algorithm of classification. the results showed that the support vector machine (SVM) algorithm was the best accuracy with 86.18 % in Sentinel-1A, 91.17 % in ALOS PALSAR-2, and 77.57 % in TerraSAR-X.