Mixture texture model with weighted generalized inverse Gaussian distribution for target detection

With the improvement of radar resolution, the amplitude distribution of sea clutter has begun to exhibit significant heavy-tailed characteristics. Existing models for sea clutter amplitude distribution do not sufficiently solve this problem, which results in a diminished target detection probability in two-step generalized likelihood ratio test (GLRT) based target detectors. To address this issue and to enhance the radar target detection capabilities, we propose a new compound-Gaussian clutter model based on a weighted mixture of generalized inverse Gaussian distributions to model the texture. The CG-WGIG model demonstrates greater flexibility and effectively captures the characteristics of sea clutter amplitude distributions with heavy tails, thus affording a better fit. We combine this model with the two-step GLRT criterion to propose a GLRT detector endowed with a weighted generalized inverse Gaussian texture (GLRT-WGIG), specifically tailored for detecting range-spread targets (RSTs). We have conducted a rigorous mathematical proof to demonstrate that the constant false alarm rate (CFAR) characteristic of the proposed GLRT-WGIG detector is independent of the covariance matrix, thereby ensuring its robustness in various operational scenarios. Using measured data, we demonstrate that the proposed model and detector are flexible and outperform the existing methods in terms of fitting and detection performance, where we obtain a mean 12.02% MSE of fitting performance promotion.