摘要
针对联合像素多基线InSAR处理的大运算量问题,提出了一种降维处理方法.在分析联合像素协方差矩阵构造过程的基础上,指出SAR图像配准时信号子空间的维数是不随卫星数目的增加而改变的.利用Lanczos迭代进行信号子空间逼近,避免了对高维协方差矩阵进行特征值分解,利用信号子空间拟合代替信号子空间向噪声子空间的投影,至少能够降低运算量50%以上.对仿真多基线数据和实测单基线数据的处理结果验证了本算法能保证处理精度要求.
According to the high computational complexity of joint-pixel multibaseline interferometric synthetic aperture radar （InSAR）, a reduced-dimension method is proposed. Based on the analysis of convariance matrix construction, the conclusion is drawn that the dimension of signal-subspace is constant, when all the SAR images are accurately coregistrated in spite of the number of satellites. Lanczos iteration is utilized to estimate the signal subspace, which avoids eigen-decomposition of the high-dimensional covariance matrix, and signal subspace fitting is employed to replace the projection of the signal subspace onto the noise subspace. The operation complexity is greatly decreased by at least fifty-percent, the performance conservation ability of the proposed algorithm is proved by the processing results of simulated multi-baseline data and single baseline real data.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2009年第3期385-389,417,共6页
Journal of Xidian University
基金
重点实验室基金资助(9140C0102020602)
国家自然科学基金资助(60802074)
关键词
合成孔径雷达
干涉
多基线
联合像素处理
Lanczos迭代
降维
synthetic aperture radar
interferometry
multibaseline
joint-pixel processing
Lanczos iteration
reduced-dimension
