Research On Recursive Bayesian Estimation Method Of Characteristic Parameters Of Sea Clutter

The K-distribution amplitude model with Gamma distributed texture can well describe the amplitude characteristics medium-and low-resolution of sea clutter in maritime radars.It has been widely used in sea clutter modeling of practical radars.K-distribution model is determined by two parameters and their estimation accuracy severely affects the performance and constant false alarm rate(CFAR)property of radar target detection.The parameter estimation methods of K-distribution model include various moment-based estimators,the maximum likelihood estimator and the percentile-based estimators.When sea clutter comes from locally stationary region,it can be modeled by traditional K-distribution model.In a large-scene detection of a maritime radar,power and non-Gaussianity of sea clutter alter with space position because radar viewing geometry and sea state are different in individual local regions,which is referred to as spatially-varying property of large-scene sea clutter characteristics.In this case,for each local region with similar parameters,there are not enough data samples for parameter estimation in a scan,and the parameter estimation becomes into a small sample problem in a single scan.In addition,radar returns from sea surface inevitably contain some outliers that are probably from returns of islands,ships,etc.These outliers are of much higher amplitude than sea clutter,and it is difficult to eliminate them completely in target detection stage.These high amplitude abnormal samples severely influence on parameter estimation.In order to solve these problems,this thesis focuses upon parameter estimation of spatially-varying K-distribution(SV-KD)model.A recursive Bayesian estimation method that jointly uses radar returns at multiple successive scans is presented to realize outlier-robust parameter estimation of spatially-varying K-distribution model.The major contributions of this thesis can be summarized as follows:In the second chapter,the physical scattering mechanism to generate sea clutter is reviewed.Then,several kinds of commonly-used sea clutter amplitude distribution models are briefly introduced.At last,several commonly-used parameter estimation methods of K-distributions are reviewed.In the third chapter,we first discuss the spatially-varying K-distribution model resulted from spatially-varying power and non-Gaussianity of large-scene sea clutter and the key problems to estimate the spatial distributions of its scale and shape parameters.Because the characteristic parameters of sea clutter vary slowly with scans,the sea clutter data of several successive scans in a local region can share the same shape and scale parameters,so the radar data of multiple scans can be available to estimate the characteristic parameters of sea clutter on the local region.Based upon this situation,a multiscan recursive Bayesian(MSRB)estimation scheme is given for parameter estimation of spatially-varying K-distribution model.Through ingenious utilization of data at previous scans and current scan data,the radar implements the perception mode of "memorizing information rather than data".In the fourth chapter,we first review the Bipercentile(Bi P)parameter estimator of K-distribution model that is robust to outliers of high amplitude.It is known that outliers from returns of islands or ships seriously affect parameter estimation of sea clutter.A multiscan recursive Bayesian Bipercentile(MSRB-Bi P)estimator is presented by combining the multiscan recursive Bayesian scheme with Bipercentile estimator.In the recursive process,updating of parameters with scans is realized by mixture of simulated data from prior information and current data.Finally,the proposed MSRB-Bi P estimator is verified by using simulation and real radar data.