【作者】 王菁；

【导师】 周建江；

【作者基本信息】 南京航空航天大学 ， 通信与信息系统， 2010， 博士

【摘要】 近年来,随着现代信号处理技术以及高分辨雷达技术的发展,雷达目标回波特性的研究与分析受到人们广泛的关注。其中,研究的重点之一是通过目标回波数据分析雷达目标强散射中心的物理位置、散射类型、散射强度以及极化信息,借以判断目标的状态、特性与类型。本文主要研究了光学区基于几何绕射理论(GTD)的雷达目标散射中心提取及其应用。主要研究内容如下:1基于GTD的雷达目标一维散射中心参数估计为了快速准确的估计基于一维GTD模型的散射中心参数,本文研究了两种方法:(1)将改进TLS-ESPRIT算法与特征分析法结合用于精确快速提取基于GTD模型的一维散射中心参数。其中,利用特征分析法的信号与噪声子空间正交特性估计散射中心的类型参数,提高了密集散射点的类型判断正确率。仿真实验中,通过比较信噪比、雷达发射带宽以及散射点疏密程度等对散射中心参数估计性能的影响,验证了改进TLS-ESPRIT方法结合正交类型判断法具有较好的参数估计性能和较高的分辨率。(2)将基于传播算子的多重信号特征方法(PM-MUSIC)用于散射中心参数估计。其核心思想是利用传播算子法取代MUSIC方法中的特征值分解,以快速计算出噪声子空间矩阵。仿真实验验证了较高信噪比条件下PM-MUSIC算法在保证良好的散射中心估计性能的同时也提高了运算效率。2基于GTD的多维目标散射中心参数估计多维散射中心提取能够提供比一维散射中心更多有价值的信息。本文在前人对多维散射中心参数的理论及提取方法的研究基础上做了进一步研究。具体内容如下:(1)分析并推导了基于几何绕射理论的2D-GTD模型,并将其近似为二维谱估计模型。同时,采用重采样技术推导了3D-GTD模型与三维谱估计模型的关系。(2)将二维修正矩阵束法(2D-MMEMP)及2D-ESPRIT算法用于二维GTD模型的散射中心参数提取。修正方法选用置换矩阵简化了第二维参数计算过程,并利用自相关矩阵的特征值分解代替Hankel矩阵的奇异值分解,一定程度上减少了经典二维矩阵束法(2D-MEMP)的运算量;其中,2D-MMEMP方法是依据一维参数有无重复或近似值选择第二维参数的计算方法, 2D-ESPRIT算法则是通过参数调节避免了特征值重根造成的特征向量不唯一问题,这两种修正方法均保证了二维散射中心参数估计的准确性且无需额外的参数配对过程。(3)提出一种修正的三维增广矩阵束算法(3D-MMEMP) ,并将该算法及修正3D-ESPRIT(3D-MESPRIT)方法用于三维GTD模型的散射中心参数提取。这两种修正方法利用空间平滑三种扫描顺序之间的关系,建立置换矩阵,简化了目标三维散射中心参数提取的计算量。3D-MMEMP和3D-MESPRIT方法提取三维参数均无需额外配对过程。即使在一维参数中出现重复或近似值的情况下,仍然能够准确估计出另外两维参数。3基于GTD的全极化目标散射中心参数提取本文结合全极化信息与高分辨技术研究了基于全极化GTD理论的散射中心一维、二维参数提取问题。(1)提出一种极化线性变换(Polarization Linear Transform)的回波数据处理方法。这种方法既可以避免分别对单极化通道进行参数提取,也可以一定程度上简化并行极化处理的运算量。文中采用极化线性变换和基于特征空间的空间谱估计(PL-ES)相结合的方法提取出目标全极化一维散射中心的参数信息,取得了较好的估计效果。(2)提出基于极化线性变换的2D-ESPRIT方法(2D-PL-ESPRIT)。并采用该算法对全极化二维GTD模型的散射中心参数进行估计,克服了单极化通道运算复杂以及参数估计不准确的问题,同时避免了二维并行极化处理方法的大运算量问题,估计效果较好。4基于散射中心提取技术的雷达目标识别将五类战斗机的主散射中心作为特征进行目标识别。利用五种散射中心参数估计方法及两种识别方法对五类战斗机进行识别,并对其识别性能进行分析,论证了散射中心作为目标特征的可行性。具体内容有:(1)将PM-MUSIC方法与支持向量机(SVM)联合用于目标识别。首先,用PM-MUSIC方法提取五类战斗机的一维散射中心特征;其次,依据飞机尺寸剔除由噪声产生的虚假散射中心;然后,利用中心矩标准化各散射中心参数,划分训练样本以及测试样本;最后通过支持向量机(SVM)进行目标识别。(2)采用南京航空航天大学雷达目标特性研究中心开发的雷达目标后向散射仿真软件,计算五类战斗机F16、J6、M2000、Su27、J8II的回波。分别使用传统的MUSIC、PM-MUSIC、改进TLS-ESPRIT、IFFT以及矩阵束(MP)等五种不同的散射中心提取方法结合支持向量机(SVM)以及自适应高斯分类器(AGC)两种分类方法对五类战斗机进行识别。仿真比较了各类算法的识别性能,并分析了实验数据压缩比、信噪比以及雷达带宽对识别性能的影响。通过比较说明在一般情况下,PM-MUSIC和支持向量机(SVM)联合的散射中心目标识别算法效果较好。5基于散射中心提取技术的RCS数据压缩拟合本文以GTD散射中心模型为基础,分析散射中心提取技术在RCS数据处理上的应用。首先,利用MP、改进TLS-ESPRIT以及MUSIC等算法获得散射中心各参数,然后通过GTD模型拟合重建RCS数据。其目的是研究用散射中心参数压缩回波数据,节省存储空间。(1)分别通过IFFT方法以及MP、改进TLS-ESPRIT、MUSIC三种基于GTD模型的散射中心提取方法进行回波数据压缩。(2)分析对比了几种压缩方法的RCS拟合效果,讨论了点目标与体目标的拟合结果、目标在不同方位角条件下的拟合性能以及频率域与角度域压缩拟合的效果。说明了散射中心技术在RCS数据压缩拟合领域具有良好的应用效果。更多还原

【Abstract】 The research on radar target echo properties has aroused great attention among researchers due to the development of modern signal processing and high resolution radar techniques in recent years. Especially, most researchers focus on making use of the information obtained from the radar target echo data, such as scattering centers’locations, types, amplitudes, and polarization to judge the targets’states, characteristics and classes. The thesis studies several key problems of radar target scattering centers extraction and its application based on the geometrical theory of diffraction (GTD) model in optical region. The main works are summarized as follows:1 1D scattering centers parameters estimation of radar target based on GTD model In order to increase accuracy and efficiency of the parameters extraction based on 1D GTD model, two improved algorithms are studied in the thesis. The two methods are summarized as follows:(1) The improved total least squares-rotational invariance technique (TLS-ESPRIT) combined with the orthogonality of signal and noise subspace is applied to extract the scattering center based on 1D-GTD model efficiently. Where, the orthogonality of signal and noise subspace is used to determine the scattering center’s type. In this way the dense points’types can be judged more accurately. Simulations demonstrate that the improved TLS-ESPRIT combined with the orthogonality method can achieve good accuracy by comparing the results of different conditions, such as signal to noise ratio(SNR), radar bandwidth and the density of scattering centers, on the estimation of parameters.(2) MUSIC algorithm based on Propagator Method (PM-MUSIC) is applied to estimate the scattering center. The PM method instead of the matrix EVD of the MUSIC is applied to compute the noise subspace matrix. Simulations show that the PM-MUSIC can achieve good accuracy in the estimation of scattering centers and reduced the computation complexity under high SNR.2 Multidimensional scattering centers parameters estimation of radar target based on GTD model Multidimensional scattering center extraction can provide more valuable information than 1D scattering center. The thesis makes a further research based on the previous scattering center extraction methods as follows:(1) The 2D-GTD model based on the geometrical theory of diffraction is analyzed, formulated, and approximated to the 2D spectrum estimation model. Simultaneously, the relationship between 3D-GTD model and 3D spectrum estimation model is also derived. (2) 2D modified matrix enhancement and matrix pencil method (2D-MMEMP) and 2D rotational invariance techniques (2D-ESPRIT) are employed to extract the scattering center based on 2D-GTD model. The permutation matrix is used to reduce the calculation cost of the second dimension parameters in the modified methods. Meanwhile, EVD on autocorrelation matrix of enhanced matrix is introduced to decrease the computational load of SVD on enhanced matrix. According to 1D estimated parameters, the proper procedure is selected to estimate the 2D parameters in the 2D-MMEMP method. And in the 2D-ESPRIT method, an adjustable parameter is used to avoid the problem of non-unique eigenvectors caused by the repeated eigenvalues. These two modified methods can achieve good accuracy in the estimation of 2D-scattering centers and avoid the extra pairing procedure。(3) A 3D-MMEMP is developed. The method and 3D-MESPRIT methods are applied to extract the 3D scattering centers based on 3D-GTD model. In these two methods, the permutation matrix is established by utilizing the relationship of three scanning orders on space smoothing. The computational load of the other dimension parameters estimation is decreased. The 3D-MMEMP and 3D-MESPRIT methods do not require any pairing algorithm. Furthermore, they can extract the other dimensions’scattering centers accurately even if there are repeated or similar values in one dimension.3 Full-Polarization scattering centers estimation based on GTD model The thesis studies the 1D and 2D full-polarization scattering center extraction techniques based on full-polarization GTD model by combining the high-resolution technique with the full-polarization information.(1) A polarization linear transform in full-polarization scattering centers estimation is proposed. This new method avoids the parameters extraction of each channel and decreases the computational complexity in smoothing process of PP-MUSIC method. Moreover, the polarization linear transform combined with the eigenspace spatial spectrum estimation method (PL-ES) is introduced to extract the 1D-GTD scattering centers to get better estimation result.(2) A 2D-ESPRIT based on polarization linear transform (2D-PL-ESPRIT) is proposed, which is used to extract the full-polarization 2D-GTD scattering centers. This method avoids the inaccuracy of the signal channel extraction, and also avoids the great computational load of the parallel polarization MUSIC method (PP-MUSIC). Simulations show that the 2D-PL-ESPRIT can achieve good accuracy in the estimation of full-polarization 2D-GTD scattering centers and reduce the computation complexity.4 Radar target identification based on scattering centers extraction Five kinds of fighters scattering centers are used as features in target identification. The performance analyses show the feasibility of the scattering centers as the features to identify the target. The works are detailed as follows: (1) The PM-MUSIC combined with the support vector machine (SVM) is applied to target identification. The 1D scattering centers of five planes are extracted by the PM-MUSIC method, and the spurious scattering centers are excluded by using the range window around the target region. The scale training samples and testing samples are obtained according to the central moments of the distribution of 1-D scattering centers on the target. Finally, targets are identified by SVM.(2) The echo data of five fighters(F16、J6、M2000、Su27 and J8II) are calculated by the radar target back scattering simulation software which is developed by radar target character research center of Nan Jing University of Aeronautics and Astronautics. The 1D scattering centers of five fighters are extracted by the five methods-MUSIC, PM-MUSIC, improved TLS-ESPRIT, IFFT and Matrix pencil (MP) under different conditions(SNR, bandwidth, data compression ratio). And the five fighters are identified by the two kinds of identification methods–SVM and adaptive Gaussian classifier (AGC) respectively. Simulations demonstrate that the PM-MUSIC combined with the SVM has a better performance.5 RCS data compression and fitting based on scattering centers extraction This thesis studies the application of scattering centers in RCS data compression based on GTD model. The MP, improved TLS-ESPRIT and MUSIC methods are applied to extract the scattering centers parameters. Then the extracted scattering centers are used to fit and reconstruct the RCS data by the GTD model. In this way, the echo data are compressed.(1) The IFFT method and scattering extraction method based on GTD model, such as MP, improved TLS-ESPRIT and MUSIC, are used to extract the scattering centers for echo data compression.(2) The thesis compares the RCS fitting performance of several compression methods, discuses the fitting effect of the point target and volume target. The fitting performance of frequency domain and angle domain is analyzed, and the fitting performance in different azimuth is analyzed also. Simulations show a good application effect of the scattering center technique in RCS data compression and fitting field.更多还原