CGLS 算法在综合孔径微波辐射计中的应用

高效的反演成像算法是综合孔径微波辐射计的关键技术之一。由于综合孔径微波辐射计反演成像在数学上是病态的反问题,所以需要进行正则化处理以克服其病态特性而获得稳定的解。与直接正则化方法相比,共轭梯度最小二乘(Conjugate Gradients Least Squares, CGLS)迭代法具有无须明确正则化参数、无须对传递矩阵求逆等优点。提出将共轭梯度最小二乘法应用于综合孔径辐射计成像中,并基于全极化干涉式微波辐射计(Full Polarization Interferometric Radiometer, FPIR)系统,比较了其与经典的最小范数正则化的性能。仿真结果表明:与最小范数正则化相比,CGLS 正则化算法能有效降低FPIR 系统图像反演误差,以获取高精确度的观测场景的亮温分布满足FPIR 系统探测海面风场、海表面盐度和土壤湿度等应用需求。

Application of Conjugate Gradient Least Squares Algorithm in Synthetic Aperture Microwave Radiometer

Efficient imaging algorithm is one of the key technologies of synthetic aperture microwave radiometer. Since the inverse problem of synthetic aperture microwave radiometer is mathematically ill-posed, it needs to be regularized to overcome its ill conditioned properties and provide a unique and stable solution. Compared with the direct regularization method, the Conjugate Gradient Least Squares (CGLS) iterative algorithm has several advantages on solving inverse problem: both the explicit regularization parameter and inversion of the transfer matrix are not required. It is proposed that the CGLS algorithm is applied to the imaging of synthetic aperture radiometer, and its performance is compared with the classical minimum norm regularization based on the Full Polarization Interferometric Radiometer (FPIR) system. The simulation results show that the CGLS regularization algorithm can effectively reduce the reconstruction error in the FPIR system, compared with minimum norm regularization. The CGLS regularization algorithm can obtain more accurately the brightness temperature distribution of the scene under observation in order to meet application requirement of detecting wind field of the ocean surface, sea surface salinity, soil moisture and so on.