Enhanced Tail Modeling of High-Resolution Radar Sea Clutter Using GEMD-Based Signal Preprocessing and the Generalized Pareto Distribution

Radar sea clutter is an important component of the maritime radar environment because it represents the background signal against which maritime targets must be detected. Strong clutter echoes can mask weak targets, such as small boats or floating objects, and may sometimes exhibit characteristics similar to those of actual targets, thereby degrading target detection performance and increasing the probability of false alarms. In addition, the complex and non-stationary statistical behavior of high-resolution sea clutter makes accurate statistical characterization particularly challenging, especially in the tail region, which is of critical importance for target detection and false alarm regulation.

This paper proposes a non-stationary time–frequency preprocessing framework based on Graph Empirical Mode Decomposition (GEMD) to improve the statistical modeling of high-resolution radar sea clutter. The proposed method first applies GEMD to the in-phase (I) and quadrature (Q) components of the radar return signal to suppress noise while preserving the intrinsic characteristics of the clutter through the extraction and appropriate reconstruction of intrinsic mode functions (IMFs). Following the denoising stage, the clutter intensity is computed using a square-law detector. The shape and scale parameters of the Generalized Pareto (GP) distribution are then estimated in the intensity domain using the Maximum Likelihood Estimation (MLE) method. To evaluate the effectiveness of the proposed framework, the GP distribution is fitted to real high-resolution IPIX radar sea clutter data with and without GEMD preprocessing. The quality of the statistical modeling is assessed by comparing the fitted distributions with the empirical data, with particular emphasis on the tail region, where extreme clutter samples can strongly affect radar detection performance and false alarm rates. The experimental results demonstrate that GEMD preprocessing significantly enhances the accuracy of tail modeling by reducing noise-induced distortions and improving the estimation of extreme clutter amplitudes. Consequently, the proposed approach provides a more reliable statistical characterization of sea clutter, leading to improved robustness of Constant False Alarm Rate (CFAR) detectors operating in challenging maritime environments. These results indicate that integrating GEMD with Generalized Pareto modeling offers an effective and practical solution for high-resolution radar sea clutter analysis and can contribute to the development of more reliable maritime surveillance and target detection systems.