Analysis of the adaptive matched filter algorithm for cases with mismatched clutter statistics

In the majority of adaptive radar detection algorithms, the covariance matrix for the clutter-plus-noise is estimated using samples taken from range cells surrounding the test cell. In a nonhomogeneous environment, this can lead to a mismatch between the mean of the estimated covariance matrix and the true covariance matrix for the test cell. Closed form expressions are provided, which give the performance for such cases when the popular adaptive matched filter algorithm is used. The expressions are exact in some cases and provide useful approximations in others. To simplify the analysis, the samples from the surrounding range cells are assumed to be independent and identically distributed and these samples are assumed to be independent from the sample taken from the test cell. Performance depends on a small number of important parameters. These parameters describe which types of mismatches are important and which are not. Numerical examples are provided to illustrate how performance varies with each of the important parameters. Monte Carlo simulations are included which closely match the predictions of our equations. An airborne radar example is provided which demonstrates that covariance 1 This research was supported by the Air Force Research Laboratory, Sensors Directorate and the Air Force Office of Scientific Research, Air Force Materiel Command under contracts F30602-97-C-0065-01 and F49620-97-C-0065-01.