Design and development of deep Proximal support vector machine Based filtering and classification Technique for hyper spectral Image datasets

Abstract

The importance of remote sensing applications and the need required newlinefor the study on earth science, geography and related military based newlineapplications, hyperspectral image and its data have become significant all newlineover the world. Due to which, in this thesis work it is intended to develop newlinetechniques for improving the quality of hyperspectral images and perform newlinevarious analysis over them to extract useful information and do necessary newlineoperations. Earlier multispectral remote sensing methods were more common newlineand this involves acquiring the near and short wave infrared images pertaining newlineto broad wavelength patterns. But the limitation of multispectral imaging is newlinethat it lacks in direct identification and hence is the growth of hyperspectral newlineimaging that performs identification and material characterization. These newlinehyperspectral imaging process is capable of attaining images over hundreds of newlineneighbouring spectral bands. newlineHyperspectral sensing devices pose the ability for identifying and newlinequantifying each and every molecule based absorptions. These hyperspectral newlineimages with higher resolution permits detecting, identifying and quantifying newlinethe earth surface materials and minerals, and also are able to infer their newlinebiological and chemical mechanisms. These characteristics of hyperspectral newlineimaging and the importance of earth science study has intended in the newlinedevelopment of this research contribution made on the various stages of newlinehyperspectral imaging process. This thesis is designed for contributions on newlinepre-processing, filtering, feature extract on classification of hyperspectral newlineimages. The research contributions made in this thesis are elucidated below. newline