稀疏目标的关联成像算法研究
发布时间：2018-11-13 14:27
【摘要】：基于时空两维随机辐射场的微波凝视关联成像是一种全新的雷达成像体制,它可对固定区域进行凝视成像,同时可获得突破天线孔径限制的分辨率,具有重要的应用价值。由于该成像体制中辐射场的时空两维随机特性,不同于传统的成像方式,这里需将已知的辐射场和接收的散射回波做信息的关联处理来获得目标图像。关联处理的方法有多类,如基于格拉姆-施密特(Gram-Schmidt)正交化的信息处理方法以及基于正则化的信息处理方法等。本文针对稀疏目标场景,研究了基于压缩感知(Compressive Sensing, CS)的关联成像信息处理方法。 压缩感知理论表明,若信号具有稀疏性,便可通过少数随机测量的数据,通过稀疏恢复算法重构信号。本文研究的出发点正是利用目标散射点分布的稀疏性这一先验,通过稀疏重构技术使系统获得更好的成像效果。 论文首先研究了经典稀疏恢复算法在关联成像系统中的应用。由基于时空两维随机辐射场的微波凝视关联成像系统模型推导得到了系统中接收回波、测量矩阵与目标后向散射系数的表达式以及相互关系,建立了关联成像的数学模型,并将经典的FOCUSS (Focal Underdetermined System Solver)、稀疏贝叶斯学习(Sparse Bayesian Learning, SBL)等稀疏恢复算法应用于微波凝视关联成像系统,并通过仿真验证了应用稀疏恢复算法可获得超分辨特性。 其次,研究了静止目标的辐射场矩阵的失配问题。当目标的强散射点与预先划分的网格点位置存在偏差,便会在辐射场矩阵中引入“扰动”,导致原有算法失效。针对该问题,重新推导并建立了关联成像系统中目标位置存在扰动的情况下的回波模型,提出了一种改进的基于约束总体最小平方(constrained least squares,CTLS)的稀疏自适应校正反演算法,仿真验证了所提算法在散射点位置偏差引起辐射场矩阵扰动情况下可以明显提高稀疏恢复算法的恢复精度并能实现目标位置误差的自校正。 最后,研究了运动目标的关联成像算法。推导并建立了目标运动场景下的回波模型,对现有的针对运动目标的成像算法进行了系统调研和分析,指出了现有方法存在的不足并提出了一种基于速度估计的运动目标的稀疏恢复算法。通过迭代的方法交替求解运动速度和目标反射系数。在每次迭代中,通过最小化加权Lp模来进行成像,同时通过最小化残差来进行速度估计。仿真验证了所提方法可同时获得高分辨图像和精确的速度估计结果。
[Abstract]:Microwave staring correlation imaging based on spatio-temporal two dimensional random radiation field is a new radar imaging system. It can perform staring imaging in fixed area and obtain resolution of antenna aperture limitation. It has important application value. Because of the spatio-temporal stochastic characteristics of the radiation field in the imaging system, which is different from the traditional imaging method, the known radiation field and the received scattering echo need to be associated with the information to obtain the target image. There are many kinds of association processing methods, such as information processing based on Gram-Schmidt orthogonalization and regularization based information processing. In this paper, we study the method of processing the correlation imaging information based on compressed perceptual (Compressive Sensing, CS) for sparse target scene. Compression sensing theory shows that if the signal is sparse, the signal can be reconstructed by sparse recovery algorithm through a small number of randomly measured data. The starting point of this paper is to make use of a priori the sparsity of the scattering point distribution of the target and to obtain a better imaging effect by the sparse reconstruction technique. Firstly, the application of the classical sparse restoration algorithm in the correlation imaging system is studied. Based on the model of microwave gaze correlation imaging system based on two dimensional random radiation field in time and space, the expressions of received echo, measurement matrix and the backscattering coefficient of target are derived, and the mathematical model of correlation imaging is established. The classical sparse restoration algorithm such as FOCUSS (Focal Underdetermined System Solver), sparse Bayesian learning (Sparse Bayesian Learning, SBL) is applied to the microwave gaze correlation imaging system. The simulation results show that the sparse restoration algorithm can obtain superresolution characteristics. Secondly, the mismatch of the radiation field matrix of the stationary target is studied. When there is a deviation between the strong scattering point of the target and the position of the pre-divided grid point, the "perturbation" will be introduced into the radiation field matrix, which will lead to the failure of the original algorithm. In order to solve this problem, the echo model of the target position in the correlation imaging system is rederived and established, and an improved sparse adaptive correction inversion algorithm based on constrained population least square (constrained least squares,CTLS) is proposed. The simulation results show that the proposed algorithm can significantly improve the recovery accuracy of the sparse recovery algorithm and achieve self-correction of the target position error under the condition of the radiation field matrix disturbance caused by the scattering point position deviation. Finally, the correlation imaging algorithm of moving targets is studied. The echo model of target moving scene is derived and established, and the existing imaging algorithms for moving target are investigated and analyzed systematically. The shortcomings of the existing methods are pointed out and a sparse restoration algorithm for moving targets based on velocity estimation is proposed. The iterative method is used to solve the motion velocity and target reflection coefficient alternately. In each iteration, the imaging is performed by minimizing the weighted Lp mode and the velocity estimation is performed by minimizing the residual error. Simulation results show that the proposed method can simultaneously obtain high resolution images and accurate velocity estimation results.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52
本文编号：2329430
[Abstract]:Microwave staring correlation imaging based on spatio-temporal two dimensional random radiation field is a new radar imaging system. It can perform staring imaging in fixed area and obtain resolution of antenna aperture limitation. It has important application value. Because of the spatio-temporal stochastic characteristics of the radiation field in the imaging system, which is different from the traditional imaging method, the known radiation field and the received scattering echo need to be associated with the information to obtain the target image. There are many kinds of association processing methods, such as information processing based on Gram-Schmidt orthogonalization and regularization based information processing. In this paper, we study the method of processing the correlation imaging information based on compressed perceptual (Compressive Sensing, CS) for sparse target scene. Compression sensing theory shows that if the signal is sparse, the signal can be reconstructed by sparse recovery algorithm through a small number of randomly measured data. The starting point of this paper is to make use of a priori the sparsity of the scattering point distribution of the target and to obtain a better imaging effect by the sparse reconstruction technique. Firstly, the application of the classical sparse restoration algorithm in the correlation imaging system is studied. Based on the model of microwave gaze correlation imaging system based on two dimensional random radiation field in time and space, the expressions of received echo, measurement matrix and the backscattering coefficient of target are derived, and the mathematical model of correlation imaging is established. The classical sparse restoration algorithm such as FOCUSS (Focal Underdetermined System Solver), sparse Bayesian learning (Sparse Bayesian Learning, SBL) is applied to the microwave gaze correlation imaging system. The simulation results show that the sparse restoration algorithm can obtain superresolution characteristics. Secondly, the mismatch of the radiation field matrix of the stationary target is studied. When there is a deviation between the strong scattering point of the target and the position of the pre-divided grid point, the "perturbation" will be introduced into the radiation field matrix, which will lead to the failure of the original algorithm. In order to solve this problem, the echo model of the target position in the correlation imaging system is rederived and established, and an improved sparse adaptive correction inversion algorithm based on constrained population least square (constrained least squares,CTLS) is proposed. The simulation results show that the proposed algorithm can significantly improve the recovery accuracy of the sparse recovery algorithm and achieve self-correction of the target position error under the condition of the radiation field matrix disturbance caused by the scattering point position deviation. Finally, the correlation imaging algorithm of moving targets is studied. The echo model of target moving scene is derived and established, and the existing imaging algorithms for moving target are investigated and analyzed systematically. The shortcomings of the existing methods are pointed out and a sparse restoration algorithm for moving targets based on velocity estimation is proposed. The iterative method is used to solve the motion velocity and target reflection coefficient alternately. In each iteration, the imaging is performed by minimizing the weighted Lp mode and the velocity estimation is performed by minimizing the residual error. Simulation results show that the proposed method can simultaneously obtain high resolution images and accurate velocity estimation results.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52
【参考文献】
相关期刊论文 前2条
1 焦李成;杨淑媛;刘芳;侯彪;;压缩感知回顾与展望[J];电子学报;2011年07期
2 方红;杨海蓉;;贪婪算法与压缩感知理论[J];自动化学报;2011年12期
相关博士学位论文 前5条
1 杨予昊;自旋目标运动成像与静止目标凝视成像方法及关键技术研究[D];中国科学技术大学;2011年
2 徐浩;基于空间谱理论和时空两维随机辐射场的雷达成像研究[D];中国科学技术大学;2011年
3 卜丽静;基于稀疏理论的星载雷达图像超分辨率重建[D];辽宁工程技术大学;2011年
4 谢晓春;压缩感知理论在雷达成像中的应用研究[D];中国科学院研究生院（空间科学与应用研究中心）;2010年
5 何学智;微波凝视关联成像的信息处理方法与仿真[D];中国科学技术大学;2013年
，本文编号：2329430
本文链接：https://www.wllwen.com/kejilunwen/wltx/2329430.html
