基于稀疏贝叶斯学习的雷达目标成像技术  

Radar Target Imaging Techniques Based on Sparse Bayesian Learning

作　　者：张西托[1] 杜小勇[1] 王壮[1] 

机构地区：[1]国防科技大学ATR实验室,长沙410073

出　　处：《计算机仿真》2008年第8期227-230,共4页Computer Simulation

摘　　要：利用傅立叶变换进行雷达目标成像,分辨率受瑞利准则的限制;超分辨成像技术能显著改善雷达图像的分辨率,但算法的复杂性急剧增加并且正则化参数不易选取。以稀疏贝叶斯学习为基础,针对雷达成像系统的结构特点,提出了一种基于快速傅立叶变换(FFT)和分块托普里兹(Toeplitz)系统的快速超分辨成像算法。算法无需存储系数矩阵,极大地降低了存储量和运算量。进一步,通过寻找拟合误差曲线和稀疏性度量函数曲线的交点实现了正则化参数的方便选择。仿真结果表明,算法对雷达目标图像具有良好的分辨率增强能力。Rayleigh criterion is a resolution limit for radar target imaging by Fourier transfornl. Super - resolution imaging can remarkably improve the image quality, but the algorithm＇s complexity increases rapidly and it is difficult to select the regularization parameter. On the basis of sparse Bayesian learning, the structure features of radar imaging system are discussed and a fast super- resolution imaging algorithm based on fast Fourier transform（FFF） and block Toeplitz system is proposed. The proposed algorithm need not store the coefficients matrix, and can dramatically reduce the memory and computation. Moreover, a convenient method for selecting the regularization parameter is presented by pursuiting the intersection point of fitting error curve and sparsity measure curve. The simulations demonstrate that the algorithm proposed in this paper does well in enhancing the resolution of radar target image.

关 键 词：超分辨成像 稀疏贝叶斯学习 快速傅立叶变换 分块托普里兹系统 正则化参数 

分 类 号：TN958[电子电信—信号与信息处理]