Recursive Estimation And Prediction Methods Of Sea Clutter Characteristic Parameters

Sea clutter is the backscattering signals from dynamically evolving sea surface illuminated by the electromagnetic waves transmitted from the radar and is an important factor to affect target detection of maritime radars.The investigation of the Doppler spectral characteristics and amplitude characteristics of sea clutter provides the theoretical foundation of the analysis of ocean current characteristics based on radar data and target detection and tracking in maritime radars.Doppler spectra of sea clutter can be characterized by the Doppler shift and Doppler bandwidth and their high-precision estimation can provide an important help for the above-mentioned two applications.At present,most of the estimation methods are developed for radar operating mode of long dwelling time at one beam position and there exists low-precision problem when these methods are directly used in traditional radars at the scan mode.In large-scene detection of maritime radars,the Doppler characteristics of sea clutter is closely relevant to the operating parameters of a radar and alter with weather condition and sea state.Because the weather condition and sea state alter with spatial position and time,the Doppler characteristics of sea clutter have spatiallyvarying and time-varying properties.Due to varying wind speed and direction and wave speed and direction with spatial position,at a single scan period the number of adjacent spatial cells with comparable Doppler shift and bandwidth is limited,which is referred to as the spatial small sample problem in the Doppler characteristic parameter estimation of sea clutter.In addition,the estimation and prediction of the scale and shape parameters in the amplitude distribution model of sea clutter play an important role in radar target detection in sea clutter.This thesis investigates the recursive estimation method of the Doppler shift and bandwidth of sea clutter using radar data in multiple sucessive scan periods and the prediction method of the shape parameter of sea clutter amplitude dsitribution model from radar parameters,weather parameters,oceanic parameters,and the viewing geometry of the radar.The content of this thesis is organized as follows:The first chapter introduces the research background and significance of the work in this thesis,and briefly describes the development of Doppler spectral characteristics of sea clutter,and reviews the development of sea clutter amplitude distibution models.Based on the precent situation of investigation,we discuss several issues that needs to be solved in this field.At last,the work in this thesis is summarized and its organization is given.In the second chapter,the physical scattering mechanism of sea clutter is reviewed,then the Doppler spectral characteristics of sea clutter are discussed,particularly its spatially-varying characteristics.At last,the compound Gaussian model of sea clutter and the relevant parameter estimation methods are briefly introduced.In the third chapter,in view of the fact that the moment-based estimation of Doppler shift and bandwidth are severely affected by the shape of the Doppler spectrum of sea clutter,a method to estimate the Doppler shift and bandwidth of sea clutter using the normalized Doppler amplitude spectrum(NDAS)is proposed and by the NDAS the influence of the texture on the Doppler shift and bandwidth estimation is removed.Moreover,the spatial small sample problem in estimation of Doppler shift and bandwidth of sea clutter is discussed and a recusive estimation method using data in multiple scan periods is proposed.At last,the effectiveness of the estimation method is verified by measured radar data.In the fourth chapter,aiming at the low precision of the shape parameter prediction of K distribution model using the empirical formula,we investigate the method of the shape parameter of the amplitude distribution models of sea clutter from radar parameters,weather parameters,sea state,and the viewing geometry of the radar.Firstly,the existing empirical formula for the K-distributions is discussed and the nonlinear function formats of the phyiscal quanlities in prediation are summared.Secondly,by using existing data classification method,the datasets in the CSIR database are classified into three groups in terms of the types of the texture distributions and the shape parameters of each dataset are estimated for use.At last,referring to the prediction formala of K-distributions,the empirical formulae for prediction are given for the other types of texture distributions.The empirical formulae are linear combinations of the nonlinear functions of the basic physical quanlities and the coefficients are is transformed into a least square problem with inequality constraints,and finally solved by transforming into a convex optimization.The fifth chapter includes a summary of the work of this thesis and a prospect of future work.