Research On Target Detection Method With Compound Gaussian Distributed Clutter Plus Noise Model

Through analysis of sea clutter data and investigation of the electromagnetic scattering mechanism of sea surface,a series of statistical models and corresponding target detection methods in sea clutter have been developed.In target detection of maritime radars,the influence of noise cannot be ignored in two cases.In the case of high resolution and low ground angle,the power level of sea clutter sometimes reduces to the level comparable with the noise power and at this time the influence of noise cannot be ignored.It is known that the power of sea clutter distributes non-uniformly distributed in the Doppler domain.The influence of noise fails to be ignored in the noise-dominated region of the Doppler domain.If noise is neglected,the corresponding detection methods will suffer from some performance loss from the model mismatch.At present,there are only relevant research results in the K-distribution plus noise model.Aiming at the applications of different maritime radars,this thesis focuses on the parameter estimation of the K distribution,a near-optimal detection method under the compound Gaussian distributed clutter plus noise model based on the effective shape parameter and a computationally-efficient near-optimal coherent detection method under generalized Pareto distributed clutter plus noise.The main work of this thesis is summarized as the follows.In the first chapter,we introduced the research background and content arrangement of the paper.In the second chapter,research on parameter estimation methods of K distribution model is aimng at.The K-distribution model,the generalized Pareto distribution model,and the inverse Gaussian distribution model are firstly reviewed.Secondly,the high-order moments estimation method,the high-order and fractional moments estimation method and the Z-LOG-Z estimation method of the K distribution model are briefly introduced.Then,based on the fact that the parameter estimation performance of the fractional-order moment estimation methods depends upon the order of the fractional-order moment,an order-adaptive K-distribution fractional moment estimation method is proposed.Moreover,simulation and measured data are used to verify the proposed estimation method and the results show that it obtains better and more stable estimation performance.In the third chapter,a near-optimal detection method based on effective shape parameters is proposed to reduce the problem that there is no computationally achievable detection method in the context of generalized Pareto distribution and IG distribution plus noise.The process to derive the near-optimal detector based on the effective shape parameter of K-distributed clutter plus noise model is reviewed firstly.Next,the effective shape parameter formula of the inverse Gaussian distributed clutter plus noise model is derived.Then,a near-optimal detection method based on effective shape parameter of generalized Pareto distribution and inverse Gaussian distribution clutter plus noise models are proposed.And the proposed detection method is extended to the case of correlated high-resolution sea clutter plus noise background.Finally,experiments show that the proposed method has better performance than the existing detectors in the clutter noise mixing area of uncorrelated clutter,and in different Doppler channels of correlated clutter,the performance of the proposed method is greatly improved compared with the existing detection methods.In addition,taking the near-optimal coherent detector in the K-distributed clutter plus noise as an example,we illustrate the limitation of the approach using the effective shape parameter.In the fourth chapter,a detection method based on fusion ideas are proposed to solve the problem that performance loss of the GLRT-LTD detector under generalized Pareto distribution clutter for under generalized Pareto distribution plus noise background.The process to derive the near-optimal detector based on the fusion idea under K-distributed clutter plus noise model is reviewed firstly.Secondly,the test statistics of the near-optimal detector based on the effective shape parameter under generalized Pareto distributed clutter plus noise model are linearly combined with the test statistics of the optimal MF detector under Gaussian white noise to develop a new detector.Thirdly a near-optimal detection method based on fusion idea under white generalized Pareto distribution clutter and noise is derived by finding,the relationship between the fusion factor and the number of pulses.Fourthly,the proposed detection method is extended to the case of correlated generalized Pareto distribution clutter plus noise.Finally,simulation experiments show that the proposed detector has better performance than existing detectors in the case of clutter noise mixing.In the fifth chapter,the work of this thesis is summarized and the further works are discussed.