INTEGRATING THE ADVANTAGES OF HYPERION IMAGE WITH LINEAR SPECTRAL UNMIXING TO DETECT THE URBAN COMPOSITION

摘要

Studying the urban environment with its highly dynamic nature is a challenging and difficult task. Use of remote sensing images has given potential scientific background to study such environments. Due to the non homogeneous land cover classes, rapid brightness variations of these classes, mixed pixel effects, shadows, and spectral confusions, identifying the urban land cover accurately through remote sensing has faced lot of difficulties. The Vegetation, Impervious surface and Soil (V-I-S) models has been accepted to parameterize the biophysical composition of the urban environment, further V-I-S models can serve as the foundation for characterizing the urbanear-urban environments universally, hens this model was used in the study to define the urban composition. Spectral Mixture Analysis (SMA) techniques with multispectral remote sensing images have been widely applied in the past to study the urban composition. Lots of technical difficulties were faced due to the spectral variation of each of the V-I-S components due to there brightness differences and the lack of the high spectral resolution of the multispectral imageries. The main objective of this study is to integrate the advantageous of Hyperspectral Images (Hyperion) with its high spectral resolution to account for these brightness variations over the multispectral imageries. The study uses EO1 Hyperion data over the Dehradun city of India. The linear unmixing results show the advantageous of atmospheric correction and de-stripping on Hyperion and also the disadvantageous of smiling effect inherent to the Hyperion image. A supervised endmember selection was used for the study to account for the linear unmixing and the result were validated with respect to Maximum Likelihood Classification (MLC) result with the use of IKONOS multispectral image. The high correlation in the range of 0.7 for Vegetation and Impervious classes, between the reference images and the Hyperion Linear Unmixing results shows the ability of Hyperspectral images to account for the urban composition.