一种结合亮温图像和子空间分解的RFI定位算法

射频干扰(radio frequency interference, RFI) 对L波段综合孔径辐射计遥感数据造成了严重污染，降低了产品质量。RFI检测定位是处理RFI的关键步骤。传统的基于亮温图像的定位算法受到仪器角分辨率的限制，无法有效分离相邻的RFI。为了实现更高的空间分辨率，基于子空间分解技术的多信号分类(multiple signal classification, MUSIC)算法被提出。然而，当亮温图像的信噪比较低时，背景和噪声对子空间分解的准确性影响较大，进而降低了MUSIC算法的定位性能。文章通过结合亮温图像和子空间分解两种方法的优点，提出了一种融合改进定位方法。该方法通过在亮温图像域中消除背景场景、增强目标射频干扰，2次提高了图像信噪比，在频域中，利用子空间分解和MUSIC算法实现超分辨率和高精度定位。通过对土壤湿度和海洋盐度（soil moisture and ocean salinity, SMOS）卫星数据进行实验和仿真验证，证明了文章提出的方法在低信噪比情况下优于传统的MUSIC算法和基于亮温的定位算法。此外，在对多个弱RFI源的定位上，该方法的定位精度也优于基于点源波纹的弱RFI检测定位算法。

A combined RFI localization algorithm of BT image and subspace decomposition for synthetic aperture interferometric radiometer

Radio frequency interference (RFI) in the L-band heavily contaminate remote sensing data and bring many challenges to the product quality of synthesis aperture interferometric radiometers. RFI detection and localization is a critical step in dealing with RFIs. The localization algorithm based on the brightness temperature (BT) image is limited by the resolution to separate closely spaced RFIs and achieve high localization accuracy in some cases. To achieve super-resolution, the Multiple Signal classification (MUSIC) method using subspace decomposition technique is suggested. Nevertheless, when the BT image SNR is low, the power of RFIs insufficient to overwhelm the background scene, and the performance of the MUSIC algorithm is unsatisfactory. In this paper, a further improvement is proposed by combining BT image and subspace decomposition. In the BT image domain, background scene cancellation and RFI target enhancement are carried out for enhancing the SNR. In the frequency domain, the MUSIC algorithm is applied utilizing subspace-decomposition to achieve super-resolution and high precision localization. Experiments and simulations based on SMOS data validate that the method presented in this paper performs better than both the classical MUSIC algorithm and BT-based localization algorithm in low SNR cases. Moreover, the localization accuracy of the method is also superior to the localization algorithm based on point source ripples in localizing multiple weak RFI sources.