【作者】 曹建蜀；

【导师】 汪学刚；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2007， 博士

【摘要】 空时自适应处理(STAP)是新一代机载相控阵雷达杂波抑制与目标检测的关键技术。在推进STAP工程化进程中,运算量巨大是其面临的首要问题,除了传统的降维处理外,开发具有并行运算结构能在硬件系统中以高度并行流水方式实现的权矢量递推求解算法,是解决该问题的另一条有效途径。有鉴于此,本文第三～五章对此进行了详细的研究,所得递推算法可用于全空时处理和固定结构降维处理的权矢量快速求解。此外,实际杂波环境的非均匀特性是其面临的另一问题,这将导致杂波协方差矩阵难以准确估计。为此,第六章研究了方向矢量失配情况下的稳健的干扰目标非均匀检测方法,第七章探讨了前向阵近程杂波多普勒频移补偿的新方法。本论文的主要贡献和创新之处包括:1)研究了杂波协方差矩阵估计后的权矢量递推求解问题。根据协方差矩阵为正定Hermitian矩阵和顺序主子式均非零的性质,首先提出了基于Hermitian矩阵求逆引理的权矢量递推求解算法,并解决了迭代计算中的数值稳健性问题;推导了顺序主子式均非零矩阵的递推求逆算法,给出了基于该算法的权矢量递推求解过程。这两种权矢量递推算法中的主要计算为矩阵矢量积、矢量内积、矢量外积,均可并行实现,迭代计算所需次数等于协方差矩阵的维数。2)通过重新构建协方差矩阵的递推形式,提出了基于协方差矩阵逆更新的对角加载采样矩阵求逆(LSMI)递推算法,该算法无需采样协方差矩阵的估计且迭代计算总次数为训练样本数,主要计算为矩阵矢量积、矢量内积、矢量外积,算法的迭代步骤经优化后有效降低了计算复杂度。3)研究了利用QR和逆QR分解实现LSMI算法的问题。递推算法中的对角加载通过设置QR或逆QR递推分解的初始矩阵即可实现,无需增加额外的计算。4)推导了一种仅利用多级维纳滤波器(MWF)前向分解结构实现的采样矩阵求逆(SMI)递推算法,消除了传统方法中后向合成滤波过程,减少了时延。并详细分析了MWF结构中实现对角加载的各种方法。5)分析了非均匀杂波环境中干扰目标方向矢量与期望导向矢量失配时,传统自适应功率剩余(APR)非均匀检测法性能下降甚至失效的原理。提出了利用强对角加载先剔除强干扰目标,再利用传统APR反复检测法检出剩余的干扰目标的改进方案,改进后的方法对方向矢量失配有着很强的稳健性。此外研究了有限训练样本集中干扰目标非均匀检测的方法。6)对于机载前向阵的近程杂波频移补偿,根据向量(矩阵)相似度准则,提出了一种从杂波数据本身中估计频移补偿量的频移算法。该算法降低了雷达参数误差对补偿估值的影响,并且能在脉冲域和多普勒域中实现,具有计算量低能并行实现的优点。7)第八章介绍了一种以ADSP_TS101芯片为核心设计的振幅和差式单脉冲雷达信号处理机。本论文有关自适应权值递推求解算法和杂波非均匀性问题解决方法方面的研究将为STAP技术的工程应用提供理论和技术支持。更多还原

【Abstract】 Space-time adaptive processing (STAP) is a crucial technique applied to clutter suppression and target detection for new generation airborne phase-array radar. However, the tremendous computational complexity poses a primary challenge to implement STAP in practical engineering. Apart from traditional reduced-rank processing, an attractive suggestion for the computational problem is to develop the recursive algorithms of calculating weight vectors, which is computationally efficient and can be laid out in a highly parallel/pipeline structure in hardware. Thus, Chapter 3 to Chapter 5 make a detailed investigation of this, and the proposed fast algorithms can be exploited to solve adaptive weights associated with full-rank processing or fixed reduced-rank processing..Another major challenge for STAP application stems from the nonhomogeneity of practical clutter environments, which can significantly skew the clutter covariance matrix estimate. Therefore, a robust nonhomogeneous detection methodology for censoring the interference-targets with mismatched steering vectors is proposed in Chapter 6. Chapter 7 achieves a new technique of Doppler compensation in airborne forward-looking radar for ground short range clutter. The main contributions of this dissertation are as follows.1) The problem of weight vectors calculations in the case of estimated covariance matrix is investigated. Due to the fact that the covariance matrix is positive-definite Hermitian and its leading principal minors are all nonzero, a recursive algorithm of computing weights with numerical stability property is first developed on the basis of the Hermitian matrix inversion lemma, and then derives a fast algorithm of inversion for such a matrix whose leading principal minors are all nonzero thereby presenting a new approach for computing weights. The implementations for above algorithms involve matrix-vector multiplications, vector inner products and vector outer products, these operations are highly parallelizable, where the number of the iterations required is equal to the dimension of covariance matrix. 2) The derivation of a loaded sample-matrix inverse (LSMI) algorithm based on updating the inverse of the sample covariance matrix is conducted by reconstructing the recursive formulation of covariance matrix. The new algorithm removes the necessity of a covariance matrix estimation and needs the number of samples iterations, where the dominant operations come from matrix-vector multiplications, vector inner products and vector outer products. In addition, an improved iterative process is presented, resulting in significant computational savings.3) The computationally efficient implementation of LSMI algorithm employing the QR decomposition or inverse QR decomposition is introduced. In which the diagonal loading can be inserted by setting only initial Cholesky or inverse Cholesky factor without any addition of computation.4) A recursive sample-matrix inversion (SMI) algorithm realized only by means of a forward analysis stage of multistage Wiener filter (MWF) is developed, which reduces the time-delay by eliminating the backward synthesis stage. Furthermore, several approaches for adding diagonal loading to MWF are presented.5) The effect of a mismatch between the actual steering vector and the assumed desired one for the interference-targets signals in nonhomogeneous clutter environments is analyzed, which results in significant performance degradation or even complete failure for traditional adaptive power residue (APR) method. An enhanced methodology which first performs a strong interference-targets censoring via diagonal loading of covariance matrix with a large constant followed by a remaining weak interference-targets censoring using traditional APR method is presented here, it is robust to the steering vector mismatch of interference-targets. Additionally, an efficient methodology to eliminate the interference-targets from a limited training-sample set is developed.6) A method for Doppler compensation for ground short range clutter of airborne forward-looking radar is proposed by using the vector (matrix) similarity criteria, whereby the compensation values can be evaluated from the received clutter data. The method significantly reduces the sensitivity of compensation values estimations against radar parameter errors and can be performed both in pulse domain and Doppler domain. Moreover, it has the advantage of low complexity and parallel implementation.7) Chapter 8 introduces a signal processing system design associated with sum and difference patterns of amplitude monopulse radar based on ADSP_TS101 chips.The research on the recursive algorithms of computing adaptive weights and the solutions to the clutter nonhomogeneity problem will provide the theory and technique supports for STAP technique application in practical engineering.更多还原