Track-Before-Detect Algorithm for Weak Target in Sea Clutter with Generalized Inverse Gaussian Texture

In weak target detection, Track-Before-Detect (TBD) algorithms have become a major research focus. Unlike traditional “detect-before-track” methods, TBD algorithms avoid threshold detection on individual measurements and instead jointly process multiple frames of data. Under complex sea conditions, where target-like outliers occur frequently, weak targets are often submerged in clutter and conventional tracking methods fail. Existing TBD algorithms for cluttered environments commonly model sea clutter using the K-distribution; however, the K-distribution provides a less refined characterization of clutter compared to the compound Gaussian model with generalized inverse Gaussian (CG-GIG) texture. To date, the application of the CG-GIG model in TBD remains largely unexplored.This paper derives a Dynamic Programming-based TBD (DP-TBD) algorithm under CG-GIG clutter, termed GIG-TBD. In this model, sea clutter is characterized by the CG-GIG distribution, targets are modeled following the Swerling 3 model, and the log-likelihood ratio (LLR) is employed as the merit function. To address the absence of a closed-form LLR expression, an efficient and accurate approximation method is proposed. Moreover, sea clutter parameters are estimated using the method of moments. Simulation results compare the proposed GIG-TBD with the K-distribution-based DP-TBD (K-TBD) and the amplitude-based DP-TBD (A-TBD). Under complex sea conditions, GIG-TBD achieves detection performance gains of approximately 3 dB and 6 dB over K-TBD and A-TBD, respectively, with corresponding track accuracy improvements of about 50% and 75%. Under moderate conditions, GIG-TBD and K-TBD exhibit comparable performance.