| When thermal noise of a radar receiver and external noise are small enough to be neglected,sea clutter can be modeled by compound Gaussian model and a series of optimal and near-optimal coherent detectors have been developed under compound Gaussian models with different types of textures for sea clutter.However,noise is always present in maritime surveillance radars,and its effect cannot be ignored at least in the following two cases.The first is that in the case of low grazing angle and high spatial resolution,small sea surface scattering coefficient at small grazing angle and small spatial resolution cell at high resolution cause sea clutter sometimes to have a low power level comparable to noise level.In this case,the impact of noise cannot be neglected.The second is that in the case when the radar is operating at long coherent integration time,sea clutter power is non-uniformly distributed in the Doppler domain,and in the noise-dominated region and clutter-noise mixture region the effect of noise cannot be ignored.In the two cases mentioned above,if noise is neglected,the model mismatch results in inaccurate parameter estimation of the sea clutter model and improper detector selection,which will bring significant loss in performance.From the perspective of practical application of radar,this thesis focuses on designing computationally simple and near-optimal coherent detectors for the interference model of K-distributed clutter plus white Gaussian noise.The major contributions of this thesis can be summarized as follows:In the second chapter,the theory and methods of coherent detection in compound Gaussian model are reviewed at first.Then,the existing K-distributed clutter plus noise model and its parameter estimation method are briefly introduced.Furthermore,we prove that in uncorrelated K-distributed clutter plus white Gaussian noise model,the general form of the optimum coherent detector is the comparison between the matched filter output and the data-dependent threshold,which is the same as optimum coherent detector in compound Gaussian model.This structure is beneficial to develop new detectors.In the third chapter,a near-optimal detector,which matches the effective shape parameter of the K-distributed clutter plus noise,is proposed in order to solve the problem that the optimum coherent detector in K-distributed clutter plus noise cannot apply to practical radars due to complex numerical integration in its test statistic.Moreover,the effective shape parameter estimation method that relies on the clutter-to-noise ratio(CNR)is introduced to characterize the non-Gaussianity of a K-distributed clutter plus noise environment.The proposed method combines the effective shape parameter with the structure of the near-optimal detector a-MF detector in the K-distributed clutter,which is computationally efficient and in accordance with the optimum detection structure.The simulation experiment results verify the performance of the proposed detector and its adaptive version.In the fourth chapter,a combined detection method controlled by CNR is proposed to reduce the performance loss when the ?-MF detector under K-distributed clutter is applied to the K-distributed clutter plus noise.The proposed method is near-optimal and with low computation complexity.It is also consistent with the optimum detection structure.Firstly,in uncorrelated K-distributed clutter plus noise model,the optimum detector matched filter detector in Gaussian noise and the near-optimal ?-MF detector in K-distributed clutter are fused to obtain a new detector,where the fusion rule depends on CNR but is independent of Doppler shift.Then,replacing the CNR by one dependent upon Doppler shift,the fusion detector is generalized to the correlated K-distributed clutter plus white Gaussian noise.Finally,the experimental results using simulated data and real sea clutter data show that the proposed detectors and their adaptive versions approximate to the optimal detection performance in correlated K-distributed clutter plus white Gaussian noise. |
