Urban area classification with quad-pol L-band ALOS-2 SAR data: A case of Chennai city, India

Abstract

Globally, 55% of the population lives in urban areas in 2018, and this number is expected to hit 68% by 2050. Earth Observation (EO) images based mapping of the urban regions is a critical parameter in the sustainable urban planning process. In recent years, rapid urban growth is experienced in the coastal metropolitan city of India-Chennai. The two land regions, having heterogeneous land uses, as high-rise high-density and medium-rise low-density of the Chennai city are taken as study area. The fully-polarimetric L-band ALOS-2 Synthetic Aperture Radar (SAR) data is used for rapid identification of the urban regions. With respect to this, a comparative assessment of the two supervised classification algorithms such as Wishart and Support Vector Machine (SVM) is presented. The same training data set is used for both algorithms, and a confusion matrix is created algorithm wise. The results of classification with the two classes as urban and non urban indicate that the SVM outperformed the Wishart supervised classification algorithm.