EVALUATION OF SPECTRAL CHARACTERISTICS OF URBAN LAND USE USING HYPERSPECTRAL DATA

Abstract

Hyperspectral sensors are advancement in remote sensing technologies; it produces data at a higher spectral resolution which is a powerful tool in discriminating urban area and in preparation of urban maps. For appropriate urban planning of a given area understanding of urban problems such as temporal changes in physical structures in the area are important. In this study, the applicability of hyperspectral remote sensing techniques is validated to classify urban surface materials commonly used in Dehradun, based on an analysis of spectral reflectance data of various urban materials, which were collected using full range (350 nm to 2500nm) ASD spectro-radiometer. The aim of this study is to compare the spectral characteristics of selected urban area using Hyperion data and ASD spectro-radiometer data and mapping the urban area after atmospheric correction of the Hyperion data using several mapping techniques. Hyperion is a space borne sensor providing hyperspectral imagery with 242 bands within the 400 nm to 2500 nm wavelength range. Although the technical specifications of the sensor are quite high, but the Hyperion image which is available in raw form is badly affected with a few atmospheric effects e.g. scattering, aerosols and water vapor which degrade the signals corresponding to the feature being sensed and cause haziness. The haziness in atmosphere accounts for the reduced radiation from Sun reaching the Earth, causing blurriness in the image. The Hyperion image is corrected by using existing radiative transfer model for better understanding of the earth features. In order to retrieve pure ground radiances from the target materials the radiometric correction of hyperspectral data is required. In this study the FLAASH model of ENVI is used for atmospheric correction of Hyperion data. Before applying atmospheric correction on the Hyperion image, p re-processing for b ad band and bad column removal has been done. After that geometric correction of Hyperion image is done by using image to image registration method. After performing the atmospheric and geometric correction of the Hyperion image, the urban area was mapped using the spectral endmember collected by the procedure which includes minimum noise fraction (MNF), pixel purity index (PPI) and n-dimensional visualization in ENVI software. Endmembers were collected from the study area after understanding and analysis of the Hyperion image. Various mapping II techniques Spectral Angle Mapper (SAM), Mixture Tune Matched Filtering (MTMF) and Linear Spectral Unmixing were applied to generate the final classified urban area image. The Supervised classification Spectral Angle Mapper (SAM) approach has given good results with overall accuracy 89.41% in mapping the study area. The other approaches Linear spectral Unmixing and Mixture Tune Matched Filtering have also given the good results. The extracted spectra from Hyperion image after atmospheric correction and the whole spectral signatures from the field samples collected with the ASD spectro-radiometer are compared in spectral domain. The created spectral library has showed anomalies on 1353-1430 nm, 1803-1956 nm and above 2387 nm regions which is mainly covered by water vapor and aerosols. The results of the spectral analysis suggest that many urban materials have showed relatively similar reflectance patterns in visible part of the spectrum. Those materials, however, displayed noticeable differences from one another in the spectral responses in the infrared range. III