Revealing the Anatomy of Cities through Spectral Mixture Analysis of Multispectral Satellite Imagery: A Case Study of the Greater Cairo Region, Egypt.

Abstract

This paper examines the feasibility of spectral mixture analysis (SMA) in deriving comparable physical measures of urban land cover that describe the morphological characteristics of cities. SMA offers a way of analyzing satellite imagery of urban areas that may be superior to more standard methods of classification. Mixing models are based on the assumption that the remotely measured spectrum of a given pixel can be modeled as a combination of pure spectra, called endmembers. SMA, using four image endmembers (vegetation, impervious surface, soil, and shade), was applied to an IRS‐1C multispectral image in order to extract measures that describe the anatomy of the Greater Cairo region, Egypt, in terms of endmember fractions. The resulting fractions were then used to classify the urban scene into eight classes of natural and human‐built features through a decision tree (DT) classifier. The accuracy of the DT classification was compared to the accuracies of two per‐pixel supervised classifications of the IRS‐1C image employing maximum likelihood (ML) and minimum distance‐to‐means (MDM) classifiers. Overall KAPPA accuracies were 0.88 for the DT classification based on SMA fractions, and 0.60 and 0.45 for the classifications conducted through ML and MDM respectively.