STAP中基于知识的杂波协方差矩阵估计技术研究

发布时间：2018-03-03 03:07

本文选题：空时自适应处理　切入点：协方差矩阵估计　出处：《国防科学技术大学》2014年硕士论文　论文类型：学位论文

【摘要】：空时自适应处理(Space-Time Adaptive Processing,STAP)技术由于能够有效提高机载雷达的杂波抑制性能和目标检测性能而受到了广泛关注。STAP中一个关键的步骤是估计待检测单元的杂波协方差矩阵(Clutter Covariance Matrix,CCM)。CCM的估计方法有两类,一类是利用杂波的统计特性,借助于和待检测单元满足独立同分布的训练样本,根据一定的准则(如最大似然准则)实现估计;另一类是利用杂波的结构特性,借助于杂波模型,通过估计杂波模型中的参数实现估计。第一类方法在均匀样本数目足够多的情况下能实现比较好的估计。然而,实际环境中样本往往是非均匀的,直接利用非均匀的样本数据来估计协方差矩阵会引起估计误差,导致性能的下降。第二类方法在模型与真实协方差矩阵匹配且各参数估计准确的情况下能取得比较好的估计性能,然而计算量较大。研究人员发现发掘并使用先验知识实现智能化信号处理能有效提高杂波抑制的性能。本文在此背景下围绕STAP中基于知识的杂波协方差矩阵估计方法展开研究工作。第二章重点讨论了CCM的特性,包括特征谱、功率谱以及实际因素对CCM的影响。仿真结果表明通道误差、杂波起伏以及载机偏航等实际因素会引起杂波自由度的增加和功率谱的展宽或变形。第三章提出了一种新的基于几何特性(协方差矩阵之间的距离)选择训练样本的方法。文章中分析了多种距离指标(包括欧式距离,黎曼距离,谱距离,物理欧式距离以及物理谱距离),并讨论了三种计算距离的方法(相邻样本协方差矩阵之间的距离,样本协方差矩阵与采样协方差矩阵之间的距离以及样本协方差矩阵与知识辅助的协方差矩阵之间的距离)。仿真结果表明利用样本协方差矩阵与知识辅助的协方差矩阵之间的黎曼距离、物理欧式距离或物理谱距离能更加有效地实现样本的选择。第四章研究了基于先验合成孔径雷达(Synthetic Aperture Radar,SAR)图像的CCM估计方法。忽略散射体方位散射特性的变化,理想情况下,基于SAR图像的协方差矩阵估计误差较小,能获得比较好的检测性能。然而,一些强杂波点(如高压电线,桥,房屋等)的散射特性随方位视角变化很大。此外,由于天气、气候的影响,SAR图像和雷达探测的场景散射特性可能不一致,这些实际因素使得利用SAR图像估计得到的协方差矩阵存在误差。针对这两种情况,文章分别提出了利用子孔径SAR图像获取某个特定角度杂波单元散射特性的方法以及联合使用SAR图像和训练样本进行有色加载的方法,仿真结果验证了方法的有效性。
[Abstract]:Space-Time Adaptive processing (Space-Time Adaptive processing) technology has attracted wide attention for its ability to effectively improve the clutter suppression performance and target detection performance of airborne radar. One of the key steps in STAP is to estimate the clutter covariance moments of the unit to be detected. There are two methods for estimating the Clutter Covariance Matrix. CCM. One is to use the statistical characteristics of clutter, and the other is to use the structural characteristics of clutter to realize the estimation according to certain criteria (such as maximum likelihood criterion), with the help of training samples which satisfy the independent and same distribution with the units to be detected. With the aid of clutter model, the parameters of clutter model can be estimated by estimating the parameters of the clutter model. The first kind of method can achieve better estimation when the number of uniform samples is large enough. However, the samples are often non-uniform in real environment. Direct use of non-uniform sample data to estimate the covariance matrix will lead to estimation errors. The second method can obtain better estimation performance when the model matches the real covariance matrix and the parameters are estimated accurately. The researchers found that intelligent signal processing with prior knowledge can effectively improve the performance of clutter suppression. In this paper, the clutter covariance matrix estimation based on knowledge in STAP is proposed. Methods in the second chapter, the characteristics of CCM are discussed. The effects of characteristic spectrum, power spectrum and actual factors on CCM are included. The simulation results show that the channel error, Clutter fluctuation and carrier yaw will cause the increase of clutter degree of freedom and the broadening or distortion of power spectrum. In chapter 3, a new training sample based on geometric characteristics (distance between covariance matrices) is proposed. This paper analyzes a variety of distance indicators (including Euclidean distance, Euclidean distance, Euclidean distance, Euclidean distance, Riemannian distance, spectral distance, physical Euclidean distance and physical spectral distance are discussed. The distance between the sample covariance matrix and the sample covariance matrix and the distance between the sample covariance matrix and the knowledge-assisted covariance matrix. The simulation results show that the sample covariance matrix and the knowledge-assisted covariance matrix are used. Riemann distance between matrices, Physical Euclidean distance or physical spectral distance can be used to select samples more effectively. Chapter 4th studies the method of CCM estimation based on a priori synthetic Aperture radar (SAR) image. The estimation error of covariance matrix based on SAR image is small and can obtain better detection performance. However, the scattering characteristics of some strong clutter points (such as high-voltage wire, bridge, house, etc.) vary greatly with the azimuth angle of view. The effects of climate on the scattering characteristics of SAR images and radar detection scenes may be inconsistent. These practical factors cause errors in the covariance matrix estimated from SAR images. In this paper, a subaperture SAR image is proposed to obtain the scattering characteristics of a particular angle clutter unit, and a method to combine SAR images and training samples for colored loading is proposed. The simulation results show that the method is effective.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【参考文献】