Abstract:

In the problem of the useful signal detection in sea clutter, the reference data are used for the estimation of the clutter covariance matrix. Different covariant matrix estimators result in different signal detection performances. The several covariant matrix estimators are given, such as sample covariance matrix, normalized sample covariance matrix, maximum likelihood estimator, approximated maximum likelihood estimator. The non-Gaussian and non stationary of sea clutter are analyzed secondly. In the end, the impact of different covariance matrix estimators on the signal detection performance is analyzed using the normalized matched filter as the signal detector. The results show that the detection performance using simulated data is better than that using measured data because of the non-stationary of sea clutter. And the detection performance using the maximum likelihood estimator of clutter covariance matrixes is better than those using other estimators.