A Support Vector Machine Based Approach for Accuracy Enhancement in End Member Estimation from Hyper Spectral Images

Abstract

This study affords a help vector system-based totally (SVMB) technique to beautify the accuracy of endmember estimation from hyperspectral snapshots. The research applies the SVMB method to the hyperspectral photograph band selection and endmember extraction using the guide vector regression (SVR) method to extract endmembers correctly from hyperspectral pix. The proposed model is examined with an artificial dataset composed of four substances and a real-international dataset of 4 forms of soil to illustrate the model's effectiveness as it should be estimating endmembers from the pix. The version is located to have significant improvement in phrases of accuracy compared with existing techniques. The proposed technique merges the characteristic extraction procedure, the selection of suitable bands, and the endmember extraction system into an unmarried level to decorate accuracy and reduce the time needed for hyperspectral photograph evaluation. The studies also propose a method to select appropriate bands with the assistance of SVR to sharpen the spectral data. The proposed version's outcomes show the proposed method's effectiveness for correctly estimating end members from hyperspectral snapshots.