Knowledge-based Adaptive Polarization Detection Of Weak Targetes In Seaclutter

The detection performance and the constant false alarm rate behavior of theconventional adaptive detectors are severely degraded in nonhomogeneous clutter.Using polarization characteristics can effectively improve the detection andidentification performance. In this paper we discussed the design and analysis of aknowledge-based adaptive polarization detector in nonhomogeneous clutter.Firstly, the Rayleigh model cannot be met for radar clutter in high resolutionradars. The real radar data collected by IPIX radar are used for the statisticalanalysis of real clutter. This paper addresses the analysis of amplitude and I/Q datacomponents and points out that the real clutter data can be very well modeled bythe compound-Gaussian model with inverse gamma distributed texture. Thespherically invariant random processes model is introduced to generate simulatedsea clutter data with specified mean vector and covariance matrix.Secondly, based on generalized likelihood ratio test, an adaptive polarizationdetector in compound-Gaussian clutter with inverse gamma distributed texture isderived. Its characteristics of the probability of false alarm are proved. Thesimulation results show the effectiveness of the detector and the constant falsealarm rate behavior.Finally, the knowledge-based adaptive polarization detector which composesof a data selector using polarization knowledge and an adaptive polarizationdetector is introduced. The maximum likelihood estimation polarization dataselector and expectation maximum polarization data selector are proposed toremove outliers from the nonhomogeneous training data. The data selectors removeoutliers effectively when outlier-clutter ratio is low and achieve a satisfactoryperformance level for the estimation of clutter covariance matrix. Thehomogeneous training data from the data selector are used to estimate the cluttercovariance matrix. Then, it is used in the adaptive polarization detector to performthe final decision about the target presence. Both the knowledge-based adaptivepolarization detector and the adaptive polarization detector which does not use thedata selector are assessed using simulated clutter data and IPIX radar data. Theresults show that the knowledge-based adaptive polarization detector outperformsits counterpart.