Land cover classification in Romanian Carpathians and Subcarpathians using multi-date Sentinel-2 remote sensing imagery. Issue 1 (1st January 2017)

• Record Type: Journal Article
• Authors: Rujoiu-Mare, Marina-Ramona; Olariu, Bogdan; Mihai, Bogdan-Andrei; Nistor, Constantin; Săvulescu, Ionuț
• Is Part Of: European journal of remote sensing. Volume 50:Issue 1(2017)
• Page Start: 496
• Page End: 508
• Publication Date: 2017-01-01
• Subjects: Sentinel-2 -- land cover -- multi-date images -- supervised classification -- support vector machine -- maximum likelihood; Remote sensing -- Periodicals; Remote sensing; Electronic journals; Periodicals; 621.3678
• DOI: 10.1080/22797254.2017.1365570
• Languages: English
• ISSNs: 2279-7254

Abstract

ABSTRACT: In this article, we processed Sentinel-2 images in order to obtain high accuracy land cover maps for two complementary study areas. The first is represented by the Romanian Subcarpathians, a hilly highly fragmented area with heterogeneous land cover pattern and the second by Romanian Carpathians, a mountain area with homogenous structure of vegetation cover. The aim of this article is to evaluate the potential of a singledate in comparison with multi-date images for which a complete calibration and an iterative process of supervised classification using Maximum Likelihood (ML) and Support Vector Machine (SVM) algorithms were applied for the both study areas. The results show that in the case of Subcarpathian area, the SVM classification on multi-date images has better accuracy due to high complexity of the land cover pattern and spectral similarities between classes, while in the Carpathians, the ML returns good accuracy, consequence of high spectral separabilities between compact features. The validation process based on ground reference data shows good accuracies, about 92.41% for the Subcarpathians and 98.65% for the Carpathians. It is clearly noticed that the land cover pattern determines the use of different algorithms and the multi-date images enhance the overall accuracy of the classification.