Self-learning parameter estimation of K-distributed clutter using nonlinear GBDT model

In this letter, a self-learning method using gradient boosting decision tree (GBDT) is proposed to estimate two parameters of K-distributed sea clutter. Different from the traditional methods using limited two moments or percentiles, a feature vector extracted from four moment ratios and nine percentile ratios are fully exploited by a nonlinear GBDT model, as to automatically estimate shape parameter. It is proved that the feature vector is independent of scale parameter. Then, scale parameter is determined by a shape-parameter-dependent percentile. Finally, both simulated data and measured data are used to confirm that the proposed estimator can attain robust and good performance in complicated and various clutter environments.