A new method for target detection in hyperspectral imagery based on extended morphological profiles

Hyperspectral remote sensing increases the detectability of pixel-and subpixel-sized targets by exploiting the finer detail in the spectral signatures. In this paper, we describe a new unsupervised algorithm for the detection of both full pixel and mixed pixel targets in hyperspectral imagery. The proposed method automatically resolves targets by using extended mathematical morphology operations. The performance of the resulting detector is experimentally evaluated using simulated and real hyperspectral data collected by the NASA/Jet Propulsion Laboratory Airborne Visible/Infrared Imaging Spectrometer and the DLR Reflective Optics System Imaging Spectrometer (ROSIS).