Assessment of urbanization of an area with hyperspectral image data

Abstract

This study attempts to apply time series hyperspectral data to detect change in landcover and assess urbanization of a small town in West Bengal, India. The objective is to utilize the potential of hyperspectral data to extract spectral signatures of the urban components of the study area using automated end member extraction algorithm, classify the area using Linear Spectral Unmixing (LSU) and assess the rate of urbanization that has taken place in the region over a period of 2 years. The automated target generation algorithm has successfully identified the pure spectra of 9 urban features after which their individual abundances in the hyperspectral imageries have been estimated. Post classification, the classes have been compared on a pixel by pixel basis and the increase/decrease in pixels noted. The change thus detected indicates a significant depletion in green cover and water bodies in the study area with increase in concrete cover over the years indicating rapid urbanization.