Knowledge-aided generalized likelihood ratio test for weak target detection in sea clutter with lognormal texture

Adaptive coherent detection in compound Gaussian sea clutter with lognormal texture (CG-LN) typically depends on sufficient secondary range cells to reliably estimate the speckle covariance matrix. In practical maritime scenarios, secondary data can be scarce, under which both covariance inference and texture-related estimation become unreliable and the performance of conventional CG-LN detectors degrades notably. To address this limitation, a prior-aided CG-LN detection framework is developed by integrating covariance prior knowledge with structure exploitation. Power-spectrum symmetry (PSS) is incorporated to utilize inherent spectral redundancy and to regularize covariance-related processing, and a spatially correlated texture inference strategy based on local aggregation is introduced to stabilize texture normalization when training cells are limited. The above components are integrated into an iterative knowledge-aided detector, (termed KAGLRT-SDAM-LND), which jointly leverages prior information and available secondary data through a small number of fixed-point updates. Simulated data and measured sea-clutter experiments validate that the proposed detector achieves consistent performance gains over conventional CG-LN coherent detectors and representative knowledge-aided baselines in data-scarce regimes.