Abstract
In high-resolution maritime surveillance radars, sea clutter exhibits highly spatial heterogeneity due to modulation of long waves with wavelengths longer than the width of one range cell. Compound-Gaussian model (CGM) fails to characterize the heterogeneous high-resolution sea clutter in both amplitude distribution and Doppler spectrum. In this article, a bi-phase compound-Gaussian mixture model (BP-CGMM) is proposed to characterize the heterogeneous sea clutter. In the BP-CGMM, spatial resolution cells are grouped into two disjoint sets, and the sea clutter in each set is represented by one CGM with inverse Gamma-distributed texture. The spectral heterogeneity indicates that sea clutter vectors at spatially adjacent resolution cells in one set share the same speckle covariance matrix, while that at two adjacent spatial cells separated in the two sets often have different speckle covariance matrices. The BP-CGMM is validated by a mass of measured high-resolution sea clutter data. Moreover, under the BP-CGMM, a detection method based on batch test is given to detect sea-surface small targets, which is composed of scene segmentation, by the aid of Bayesian threshold and morphological filtering, and adaptive generalized likelihood ratio test linear-threshold detector (GLRT-LTD) separately in each set. The detection method is verified by measured data with small targets under test. The experimental results show that it attains better detection performance than the adaptive GLRT-LTD under the CGM of sea clutter.
Highlights
A IRBORNE/SPACEBORNE synthetic aperture radar (SAR) systems [1], [2] and traditional maritime surface surveillance radars operating at scan or dwelling mode [3], [4], are two types of complementary sensors in ocean and ship surveillance systems
To examine the characteristics of long-wave modulated texture at dwelling mode, texture time series are estimated from time series of sea clutter at each range cell by a sliding averaging method [17], where the length of the sliding window is determined by texture coherent length (TCL) and the pulse repetition interval (PRI)
In the Compound-Gaussian model (CGM), the amplitude distribution of sea clutter is determined by the texture statistics, and the Doppler spectral shape of sea clutter is mainly determined by the speckle temporal correlation
Summary
A IRBORNE/SPACEBORNE synthetic aperture radar (SAR) systems [1], [2] and traditional maritime surface surveillance radars operating at scan or dwelling mode [3], [4], are two types of complementary sensors in ocean and ship surveillance systems. Several bimodal coherent sea clutter models are proposed to characterize varying Doppler spectra with time and range [20] These mixture models exhibit satisfactory goodness-of-fit on measured data by integrating the contribution of regular Bragg resonance and fast scatterers related to sea spikes in backscattering echoes from sea surface. Various methods have been developed to estimate the clutter characteristics at the CUT, including the cell-clustering method for selection of RCs [27], the knowledge-aided methods [28], the Bayesian methods using the prior structural information to compensate inefficient RCs [29], and outliers-censoring method [30], which all attain better performance owing to their subtle clutter models These methods all suffer from highcomputation complexity due to their statistic-based strategies to deal with heterogeneous environments.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
