Multiple Sub-Pixel Target Detection for Hyperspectral Imaging Systems

Hyperspectral target detection is a task of primary importance in remote sensing since it allows identification, location, and discrimination features. To this end, the reflectance maps, which contain spectral signatures related abundances materials observed scene, are often used. However, due to low spatial resolution most hyperspectral sensors, targets occupy fraction pixel and, hence, spectra different sub-pixel (including background spectrum) mixed together within same pixel. solve issue, paper, we adopt generalized replacement model accounting for multiple formulate problem at hand as binary hypothesis test where under alternative modeled terms linear combination endmembers whose coefficients also account presence background. Then, devise architectures based upon likelihood ratio unknown parameters suitably estimated through procedures inspired by maximum approach. The performances proposed decision schemes evaluated means both synthetic well real data compared with an analogous counterpart showing effectiveness procedure.