Adaptive target detection across a clutter boundary: GLR and maximally invariant detectors

We present and compare adaptive detection algorithms developed for synthetic aperture radar (SAR) targets in structured clutter, utilizing both generalized likelihood ratio (GLR) tests and maximal invariant (MI) tests. We consider the problem of detecting a target straddling a known boundary between two independent clutter regions inducing a clutter covariance matrix with block diagonal structure. GLR and MI tests are presented for various clutter scenarios: two totally unknown clutter types, one of the clutter types known except for its variance, and one of the clutter types completely known. Numerical comparisons illustrate that GLR tests and MI tests are complementary-neither test strategy uniformly outperforms the other-suggesting that it may be worthwhile to hybridize these tests for overall optimal performance.