Longtime Adaptive Coherent Detection in Heterogeneous Compound-Gaussian Sea Clutter

Sea clutter in coastal waters exhibits spatial heterogeneity and fast-varying Doppler spectra along radial distance. Due to the limited number of reference cells, even less than the integrated pulses, longtime adaptive coherent detection fails or has a severe performance degradation. Post-Doppler detection, known as general moving target detection (MTD), can work well in this case, which requires fewer reference cells and lower computational cost. In this paper, MTDs in compound- Gaussian clutter are proved to be of constant false alarm rate (CFAR), and the CFAR thresholds are determined by the number of reference cells, the shape parameter, the false alarm rate, and the clutter power estimator at each Doppler bin. The detection probability is proved to be determined by signal-to-clutter ratio after integration (SCR-AI) and threshold. Moreover, it is revealed that MTDs are superior to adaptive near-optimum coherent detectors in K-, GP-, and CGIG distributed clutter with moderate and large shape parameters when the reference cells are no more than twice the integrated pulses. The boundary conditions of the two kinds of detectors are given. As another contribution, longtime adaptive near-optimum coherent detectors via robust speckle covariance matrix inversion based on firstorder AR adaptive loading are proposed, working well when the reference cells are much less than the integrated pulses and remedying the poor performance of MTDs at small shape parameters. Finally, the longtime adaptive near-optimum detectors are verified by P- and X-band measured radar data, which can cooperate with MTDs effectively to detect low-velocity small targets in heterogeneous sea clutter.