Oversampling MTD Method In Sea Clutter Background

With the development of technology, more and more attention has been paid to the oceans. Sea clutter is widely concerned in military and civil fields. Sea clutter is the radar echo from sea surface, which is the unwanted signals for detection of sea-surface targets and is desired to be mitigated by signal processing methods. Sea clutter is nonstationary stochastic process and forms the main obstacle to find targets in sea surface. Investigation of the characteristics of sea clutter is inevitable for target detection in sea clutter.So far, the statistical modeling is still the main approach to characterize sea clutter. The amplitude distribution models of sea clutter have been through studied. According to different radar grazing angle, resolution, and frequency bands, the amplitudes of sea clutter can be modeled by the Rayleigh distribution, the Log-Normal distribution, the Weibull distribution and the K-distribution. In this thesis, a brief introduction of these amplitude distributions is given and the methods to simulate clutter data of these distributions are described. Besides the amplitude distributions, the Doppler spectral characteristic of sea clutter is rather important for moving target detection that utilizes the difference between target returns and sea clutter in the Doppler domain. A set of real sea clutter is used to calculate the Doppler spectra of sea clutter and the Doppler offset, bandwidth, and power fluctuation at individual Doppler bins are discussed.Next, several non-coherent target detection methods in time domain are reviewed, including the mean level constant false alarm rate(ML-CFAR) detector and the order statistics CFAR(OS-CFAR) detector. The ML-CFAR detector includes the three forms: the cell average CFAR(CA-CFAR) detector, the greatest of CFAR(GO-CFAR) detector, and the smallest of CFAR(SO-CFAR) detector. Real sea clutter datasets are used to examine the performance of these detectors and we conclude that the CA-CFAR detector is better as sea clutter is stationary and there are not strong interferences in the reference cells and the OS-CFAR detector is better as there are strong interferences in reference cells. It is found that targets can be effectively detected by these detectors only when target returns are strong enough because these detectors only use the amplitudes of the received echoes.Finally, the moving target detection(MTD) method in sea clutter is discussed, which is composed of the Discrete Fourier Transform of the received time series followed by the ratio test in individual bins. It is found that when the critical-sampling DFT is used the MTD method suffers from some performance loss because the DFT base in one Doppler bin mismatches the Doppler steering vector of a target. In order to remove this loss, the oversampling DFT is used and the decision threshold at individual Doppler bins depends upon the shape parameter of the amplitude distribution of sea clutter. The experiments using real sea clutter data are made and the results show that the oversampling factor of four is a good trade-off between the performance loss and computational cost.