Combined Bipercentile Parameter Estimation of Generalized Pareto Distributed Sea Clutter Model

The generalized Pareto distributed sea clutter model, known as one of the compound-Gaussian models, is able to describe heavy-tailed characteristic of sea clutter under high-resolution and low grazing angle detection scene efficiently, and the accuracy of parameter estimation under this condition heavily impacts radar’s detection property. In this paper, Combined BiPercentile (CBiP) estimator is proposed to estimate the parameters. The CBiP estimator is realized based on the explicit roots of low-order polynomial equations and full application of sample information in returns, which provides a highly-accurate parameter estimation process. Besides, the CBiP estimator can maintain the robustness of estimation performance when outliers with extremely large power are existing in samples, while other estimators, including moment-based and Maximum Likelihood (ML) estimators, degrade extremely in estimation accuracy. Without outliers in samples, the combined bipercentile estimator shows similar accuracy with the ML estimator. With outliers, the combined percentile estimator is the only method with robustness in performance, compared with other estimators aforementioned. Moreover, the ability of the new estimator is verified by measured clutter data.