Progress in detector fusion

Introduced to the hyperspectral imaging community in the WHISPERS 2010 Reykjavik meeting, the “clairvoyant fusion” methodology for creating detection algorithms has since been applied to a variety of binary decision problems. These are summarized here. For any detection problem amenable to a fusion approach, an infinity of “flavors” can be devised. This paper describes what problems have been formulated in ways that insure closed form solutions. Such solutions are necessary in all applications that fuse a continuum of tests, because otherwise an infinity of logical operations would need to be implemented. The applications have featured a variety of subspace detection problems of varying dimensionality, along with several types of elliptically contoured distributions. Model subtleties not recognized by classical methods are discussed, along with a recent theoretical extension enabling the design of unbeatable methods.