CLASSIFICATION OF ASTER DATA USING HYPERSPECTRAL ALGORITHMS

Abstract

Remote sensing means measuring ' Remotely' without any physical contact. Common categories of remote sensors include panchromatic, multispectral, hyperspectral, and ultraspectral sensors. Multispectral sensors cover two or more spectral bands simultaneously typically from 0.3 m to 14 m wide. Hyperspectral sensors have dozens to hundreds of narrow contiguous bands. Image data from several hundred bands are recorded at the same time offering much greater spectral resolution than a sensor covering wider bands. Satellite imagery can be analyzed through various approaches. The most common image analysis approach is `Image Classification'. Classification approach is based on the fact that different earth objects always have different spectral information content. Utilizing the spectral information that is stored in the images, the image is classified. Multi-spectral classification is based on the image statistics while in case of hyper-spectral classification, it is based on predefined spectral library, but each method differs in using these spectral libraries for classification In order to evaluate the utility of the ASTER, a multi-spectral data in the discrimination and mapping of forests and related land cover types (broad classification) and to develop an effective remote sensing methodology for the utilization of ASTER data for mapping various features by applying hyper-spectral analysis techniques. The hyperspectral algorithms used for Aster Data classification are 'Spectral Angle Mapper' (SAM), spectral feature fitting (SFF), matched filtering, linear spectral unmixing. All the 14 bands of ASTER data have been used in classification. The software used for classification is ENVI 4.1.