Robust hyperspectral detection with algorithm fusion

Two simple methods are described for fusing the outputs of hyperspectral rare target detection algorithms to achieve more consistent results across a variety of images and objects of interest. The methods are demonstrated with atmospherically corrected (spectral reflectance) visible/near-infrared/shortwave-infrared and long-wavelength infrared hyperspectral imagery using five different detection algorithms that output generalized likelihood ratio decision statistics. Results are presented for nine test cases.