Band selection and decision fusion for target detection in Hyperspectral imagery

A band clustering and selection approach based on standard deviation (STD) and orthogonal projection divergence (OPD) is introduced in this paper. STD of Hyperspectral image data is calculated. Hyperspectral image data is analyzed for multiple target detection. Spectral signatures of required target are used to measure OPD. Optimal number of bands preserving maximum information is calculated by using a new developed technique, virtual dimensionality (VD). For endmember extraction, vertex component analysis (VCA) is used. A new approach for decision fusion is also introduced by using spectral discriminatory entropy (SDE) and spectral angle mapper (SAM). A comparative study is conducted to show the effectiveness of new approaches of band clustering and selection and decision fusion.