Discrimination of formal and informal settlements based on multipolarized TSX imagery in the Megacity Mumbai, India

Abstract

The significant growth of urban areas, which was observable in the last decades, is expected to continue. By the year 2050 almost three out of four people live in cities. Most parts of this development will take place in developing countries. This favors the increase of numbers of mega cities, huge urban agglomerations of more than 10 million inhabitants. Since this is an unsolvable challenge for the administrative institutions informal settlements – commonly called slums – will develop rapidly. Often it is impossible to map these settlements accurately. Remote sensing can support mapping techniques with information about the physical appearance of urban areas. Therefore the knowledge based classification of satellite images can provide thematic maps with data deduced from diverse features. In this study second order texture features of multi-polarized synthetic aperture radar (SAR) data are used for urban land cover classification. To enhance the feature spaces of the SAR data Kennaugh-elements are utilized. The classification is accomplished with a linear discriminant analysis, which also allows to test the influence of the individual features on the classification result. The results provide an overview of the classification accuracy with different combinations of Kennaugh-elements. The most valuable classification reaches an overall Kappa of 0.76 and a quality of 69.95.