Hyperspectral target detection using a Bayesian likelihood ratio test

A new approach to multivariate detection has been devised, which replaces the generalized likelihood ratio (GLR) with another, the Bayesian likelihood ratio (BLR). The new test is partly based on selectable prior distributions of the parameters appearing in the GLR. Through these distributions, the method facilitates the incorporation of prior knowledge generated by simple physics and experience in measurement programs. This paper explores these potentialities by applying the new formalism to the problem of matched subspace detection in hyperspectral data sets.