Efficient detection in hyperspectral imagery

Hyperspectral sensors collect hundreds of narrow and contiguously spaced sp ectral bands of data. Such sensors provide fully registered high resolution spatial and spectral images that are invaluable in discriminating between man-made objects and natural clutter backgrounds. The price paid for this h igh resolution data is extremely large data sets, several hundred of Mbytes for a single scene, that make storage and transmission difficult, thus req uiring fast onboard processing techniques to reduce the data being transmit ted. Attempts to apply traditional maximum likelihood detection techniques for in-flight processing of these massive amounts of hyperspectral data suf fer from two limitations: first, they neglect the spatial correlation of th e clutter by treating it as spatially white noise; second, their computatio nal cost renders them prohibitive without significant data reduction like b y grouping the spectral bands into clusters, with a consequent loss of spec tral resolution. This paper presents a maximum likelihood detector that successfully confron ts both problems: rather than ignoring the spatial and spectral correlation s, our detector exploits them to its advantage; and it is computationally e xpedient, its complexity increasing only linearly with the number of spectr al bands available. Our approach is based on a Gauss-Markov random held (GM RF) modeling of the clutter, which has the advantage of providing a direct parameterization of the inverse of the clutter covariance, the quantity of interest in the test statistic. We discuss in detail two alternative GMRF d etectors: one based on a binary hypothesis approach, and the other on a 'si ngle' hypothesis formulation. We analyze extensively with real hyperspectra l imagery data (HYDICE and SEBASS) the performance of the detectors, compar ing them to a benchmark detector, the RX-algorithm. Our results show that t he GMRF 'single' hypothesis detector outperforms significantly in computati onal cost the RX-algorithm, while delivering noticeable detection performan ce improvement.