Bipercentile parameter estimators of bias reduction for generalised Pareto clutter model

Abstract

The generalised Pareto distribution is an efficient model to characterise the high-resolution sea clutter. Robust and precise estimation of the model's parameters is a precondition of effective target detection in sea clutter. It is known that the explicit bipercentile (BiP) estimators are efficient in computation and robust to outliers. In this study, it is proved that the explicit BiP estimators are also consistent with respect to sample size. In practical applications, only limited samples are available and the bias of the BiP estimators degrades the estimation precision. The properties of the biases are analysed and BiP estimators of bias reduction are constructed by the look-up table method. The BiP estimators of bias reduction are verified by simulated data and measured sea clutter data.