非盲源KPCA剩余噪声比阈值层析SAR成像方法  

Non-blind Source KPCA Residual Noise Ratio Threshold TomoSAR Imaging Method

作　　者：刘慧 程碧辉 庞蕾 郭子夜 王潜 LIU Hui;CHENG Bihui;PANG Lei;GUO Ziye;WANG Qian(School of Electrical and Information Engineering,Beijing University of Civil Engineering and Architecture,Beijing 102627,China)

机构地区：[1]北京建筑大学电气与信息工程学院,北京102627

出　　处：《现代雷达》2024年第5期13-18,共6页Modern Radar

基　　金：国家自然青年科学基金资助项目(61501019)。

摘　　要：合成孔径雷达(SAR)层析成像技术是基于干涉SAR测量技术发展而来的先进三维成像技术。层析SAR通过第三维反演技术将叠掩在同一距离-方位分辨单元的散射体进行分离而实现SAR的三维成像。因此,叠掩在同一距离-方位分辨单元的散射体分离是层析成像的关键。文中提出了一种非盲源散射体分离算法,结合核主成分分析和剩余噪声比阈值,估计同一距离-方位分辨单元内散射体的个数,并反演其位置。在满足完整度的同时,尽可能抑制噪声。该方法利用核主成分分析,人为增加核矩阵维度,从而减少系统的导向向量偏差;并且加入剩余成分中噪声强度比的计算作为算法的约束条件,从而降低了噪声信号误判的可能性。实验结果表明,所提方法在各个方面都优于传统的层析反演方法,并且高度重建精度得到一定程度的提高。Synthetic aperture radar(SAR)tomography is an advanced 3D imaging technology based on interferometric SAR.The tomographic SAR(TomoSAR)is realized by separating the scatterers in the same range-azimuth resolution unit by the third dimensional inversion technique.Therefore,the scatterers separation in the same range-azimuth resolution element is the key of TomoSAR.In this paper,a non-blind scatterer separation algorithm is proposed,which combines kernel-principal component analysis(KPCA)and residual noise ratio threshold to estimate the number of scatterers in the same range-azimuth pixel and to invert their positions.While satisfying the integrity,noise is suppressed as much as possible.In this method,the kernel matrix dimension is artificially added by using kernel principal component analysis to reduce the steering vector deviation of the system.The noise intensity ratio of the remaining components is added as the constraint condition of the algorithm to reduce the possibility of misjudgment of noise.Experiments results show that this method is superior to traditional TomoSAR inversion in all aspects,and the height reconstruction accuracy is improved to a certain extent.

关 键 词：非线性散射体分离 层析合成孔径雷达 核主成分分析 合成孔径雷达三维成像 

分 类 号：TN957.52[电子电信—信号与信息处理]