SPATIAL-SPECTRAL FEATURE EXTRACTION ON HYPERSPECTRAL IMAGERY

In this work we extend several spatial feature extraction techniques to operate on hyperspectral imagery (HSI), which has several image bands. The HSI image bands provide a high dimensionality spectral image space, which is then compounded by the addition of several spatial features per pixel. We investigate enhancing the spatial resolution of Airborne Real-time Cueing Hyperspectral Enhanced Reconnaissance (ARCHER) HSI imagery with coincident high spatial resolution panchromatic imagery and compare feature extraction performance when both the spatial and spectral attributes are considered simultaneously as opposed to individually. Our experimental results show that spatial-spectral features of pansharpened HSI imagery improve the separability of different target classes, particularly when targets move between looks.