Target Detection for Heterogeneous Cyclostationary Sea Clutter

Abstract

This paper mainly deals with the problem of target detection under the heterogeneous background of cyclostationary sea clutter. Conventional approaches generally assume the ideal condition requiring the secondary data to be homogeneous with the primary data in order to exactly estimate the clutter covariance matrix and implement the adaptive filters. To the contrary, the realistic clutter environments appear heterogeneous, leading to the performance degradation of these traditional processors. For the sake of alleviating the effect of the heterogeneity, the non-homogeneous detectors, especially with knowledge-aided (KA) method based on the prior knowledge, are presented under the heterogeneous Gaussian condition. However, the experimental data manifest that the compound-Gaussian distribution is successfully applied in modeling the heterogeneous sea clutter, which also presents the cyclostationarity. Accordingly, when lacking prior information as used in the KA method, a new non-homogeneous detector based on mean value (M-NHD) is proposed against the heterogeneous sea clutter with cyclostationarity by operating solely on the primary data, in terms of the generalized likelihood ratio test (GLRT) criterion. The expressions of the probabilities of detection and false alarm are subsequently given. Since the detection performance depends on the steering vector, an adaptive non-homogeneous detector based on the steering vector (SV-NHD) is proposed subject to the design method for the optimal steering vector. Finally, the numerical results evaluate the performance of the two proposed detectors with Monte Carlo method under heterogeneity.