Research On Model Selection And Multiresolution Characteristics Of Sea Clutter

Searching,surveilling and identifying sea-surface targets are one of the main tasks of maritime radars.It is necessarily interfered by sea clutter,the backscattering signal from the sea surface,to carry out these tasks.Therefore,it becomes into an important foundation of maritime radars to work to deeply analyze the characteristics of sea clutter,to model sea clutter,and estimate the parameters of sea clutter models.In target adaptive detection technique in sea clutter,sea clutter is often modelled by compound-Gaussian models(CGM).In the CGMs,sea clutter is modelled into the product of slowly-varying texture component and fast-varying speckle component,where the two components are mutually independent.The speckle reflects the inter-pulse phase change of sea clutter and the texture component follows positive non-Gaussian distributions,reflecting the amplitude/intensity change.At present,there are four families of commonly-used texture distributions:Gamma distributions,the inverse Gamma distributions,the inverse Gaussian distributions,and the lognormal distributions.The four families of texture distributions correspond to the Kdistribution model of sea clutter,the generalized Pareto model,the IGCG amplitude model,and the lognormal-texture amplitude model.However,in practical oceanic environments of radars to work,the model selection and parameter estimation of sea clutter is an important precondition of target detection in sea clutter.A mass of data analysis shows that the type of parameters of the CGM of sea clutter are affected by many factors,including radar parameters,oceanic parameters,the geographical location,and the viewing geometry of the radar.At the small grazing angles,sea state and spatial resolution are major factors to affect the sea clutter model and its parameters.This thesis investigates the selection methods of the optimal model of sea clutter and by a mass data analysis the empirical relation between spatial resolution of the radar and the optimal model of sea clutter is given.Moreover,from the two aspects of statistical analysis and data analysis,the influence of the spatial resolution on the model type and model parameters is investigated and the empirical relation between the type and parameters of the optimal model of sea clutter and spatial resolution is described by tables,which forms the foundation of multiresolution detection of targets of different sizes in high-resolution maritime radars.The main content of this thesis is as follows:In the first chapter,the background and significance of this paper are introduced,and the present researches of the compound-Gaussian models of sea clutter,the optimal model selection and prediction of sea clutter characteristic parameters are reviewed.Finally,the structure and content of the thesis are briefly described.In the second chapter,the basic theory of the CGMs of sea clutter is introduced,the amplitude distributions with the four families of texture distributions are reviewed in detail,and the corresponding parameter estimation methods are introduced.In the third chapter,the main factors to influence the amplitude distribution of sea clutter model are analyzed,the traditional model selection based on the K-S distance is reviewed,the logarithmic K-S distance selection method is presented to improve the fitting of tails,and the model selection method based on the combination of the two distances is proposed.Based on the analysis of the IPIX and CSIR databases,it shows that the optimal models are the IGCG model and the generalized Pareto model at most cases and the IGCG model is suitable for low and moderate resolution sea clutter data and the generalized Pareto model is suitable for high-resolution sea clutter data.The fourth chapter investigates the model type and shape parameter prediction problem in the multiresolution sea clutter model.Besides small targets,high-resolution maritime radars are required to find ships of large or moderate sizes and the multiresolution layered detection scheme is used to find targets of different sizes at different range resolutions.Therefore,it is meaningful to predict the model types and shape parameters of sea clutter at moderate and low resolutions when the shape parameter of sea clutter at the highest range resolution is estimated.Here,the resolution conversion method of sea clutter is given and the prediction method of the model type and shape parameter of moderate and low resolution sea clutter is presented,which is made by establishing tables.The fifth chapter summarizes the work of this thesis and looks forward to the direction and content of the future researches.