Comparison of Various Indices to Differentiate Built-up and Bare Soil with Sentinel 2 Data

Abstract

Accurate mapping of built-up area and bare soil, having similar spectral characteristics is an important task for urban mapping using remote sensing data. In order to classify both built-up area and bare soil accurately, several indices with Landsat data are proposed in the literature. This study is planned to compare the performance of several indices proposed to accurately differentiate built-up area and bare soil using Sentinel 2 data acquired over Kurukshetra (Haryana) during April 2017. All together nine indices related to built-up area and bare soil identification and proposed in literature were derived with Sentinel 2 data. Different dataset combinations were used for classification in four land cover classes. Support vector machine classifier was used to classify different combination of images used in this study. Comparison of results in terms of area for both built-up area and bare soil using classified images and field visit suggest that a combination of six bands of Sentinel 2 data with six indices was found performing well in comparison to other combinations and was able to differentiate both built-up area and bare soil accurately.