Sentinel 1 Classification for Garlic Land Identification using Support Vector Machine

Abstract

The high demand for garlic is not comparable with the results of domestic garlic production. Indonesian garlic needs fulfilled by imports up to 95% of national needs. The Ministry of Agriculture has a program of the cultivation of garlic in Sembalun, East Lombok, West Nusa Tenggara in order to realize garlic self-sufficiency. This study aims to identify the garlic land in Sembalun using the Sentinel 1A satellite image. The image consists of dual-polarization VV and VH values. Images were acquired in July and November 2019 for the area of Sembalun, East Lombok, West Nusa Tenggara Indonesia. Preprocessing data steps involve applying orbits, calibrations, speckle filters, terrain corrections, and linear to dB. Support vector machine algorithm is used to classify Sentinel 1A images. Hyper parameter tuning was done to get the best parameters which are regularization parameter (C) 10, gamma 1, and the RBF kernel. The classification model has accuracy of 76%, precision of 71% and recall of 89%.