Land cover analysis based on descriptive statistics of Sentinel-2 time series data

Abstract

In our paper we examined the opportunities of a classification based on descriptive statistics of NDVI throughout a year’s time series dataset. We used NDVI layers derived from cloud-free Sentinel-2 images in 2018. The NDVI layers were processed by object-based image analysis and classified into 5 classes, in accordance with Corine Land Cover (CLC) nomenclature. The result of classification had a 76.2% overall accuracy. We described the reasons for the disagreement in case of the most remarkable errors.