中文摘要：

确定性的综合孔径辐射计反演方法没有考虑亮温分布先验信息的统计特性;另一方面,该类反演方法也难以分析亮温分布离散化引起的离散化误差,因此往往将其忽略。为此,提出一种多分辨率统计反演方法,该反演方法能引入先验信息和离散化误差的统计特性。多分辨率统计反演方法的思想是将未知的亮温分布和综合孔径辐射计测量的可见度视为随机变量,并采用统计推断的方法进行综合孔径辐射计图像反演。根据这一思想,提出了一种先验协方差估计方法,建立了含有超参数的高斯先验模型;采用多分辨率分析建立了亮温分布的离散模型,根据该离散模型和高斯先验模型计算了离散化误差的概率分布;将综合孔径辐射计图像反演转换为超参数估计的问题,采用Newton-Raphson迭代算法对超参数进行了估计。理论分析、仿真和实验均表明：与传统的确定性的反演方法相比,多分辨率统计反演方法通过利用先验信息和离散化误差的统计特性,可有效提高综合孔径辐射计的成像性能。

英文摘要：

As the deterministic inversion approaches（DIAs） of aperture synthesis radiometers（ASRs） does not refer to the statistical property of the prior information of brightness temperature distribution（BTD） and the discretization error due to the discretization of BTD is difficult to analyze and always ignored,this paper presented multi-resolution statistical inversion approach（MSIA）,which can introduce the statistical properties about the prior information and the discretization error.The idea of MSIA is to treat the unknown BTD and the visibilities as random variables,and to recast the image inversion of ASRs as statistical inference.According to the idea,a method for prior covariance estimation was proposed,and a Gaussian prior model was established with a hyperparameter.Then,a discrete model of BTD was constructed by the multi-resolution analysis.The probability density of the discretization error was computed based on the prior and the discrete model of BTD.Finally,the ASRs imaging was converted to a problem of hyperparameter estimation.The Newton-Raphson iterative algorithm was introduced to estimate the hyperparameter.The theoretical analysis,simulations and experiments show that as compared with the traditional DIAs,the MSIA can greatly improve the imaging performance of ASRs via utilizing the statistical properties of the prior information and the discretization error.