Abstract

Hyper-spectral optical data has been the key for accurate mapping in various field of scientific research to get results in different dimension. Based on this the present study involves two different images classify by different technique to improve the spectral resolution classification for the LULC areas using their unique spectral reflectance. The Optimum bands for the urban, vegetation, agriculture and water features are found using the spectral library is created for different invariant LULC features. The performance evaluation of the Hyperion image is carried out in terms of spatial, spectral and feature based and the results shows Spectral Angle Mapper with n-D visualizer produces a better classification output compared to the Spectral Angle Mapper and Support Vector Machine method for a heterogeneous LULC area. Accuracy assessment also revealed choosing reference pixels for classification using MNF scatterplots and then refining them use n-D visualizer increases classification accuracy.