Research On Robust Reconstruction Method Of Aperture Synthesis Radiometer

Aperture synthesis radiometer (ASR), which offers the benefit of high spatial resolution and instant image circle at cost of high complexity of signal processing and hardware, is an alternative to real aperture radiometer. But the outlying point due to the transmission error and the abnormal operation of one or several of the antennas/receivers may occur in the measured visibilities which are used to reconstruct the brightness temperature image because of the high complexity. And we hope the corresponding inversion algorithm has a good robustness, it can provide a reliable solution with regardless of the outlying points. Thus, it is a quite meaningful and valuable topic to analyze and improve the robustness of the conventional inversion algorithms.In this article, firstly the state-of-art of the robust statistics, ASRs and the inversion algorithms are presented. And the inverse problem is classified into the minimum experience risk theory which has the two primary factors:function model and loss function. In the viewpoint of statistics study theory, we point out that the conventional inversion algorithms have a poor robustness since they just optimize the function model with no consideration of the loss function. And some simulations have shown that when the outlying points appear the images reconstructed by the conventional algorithms will be distorted. Specially, the effect of the outlying point will rapidly increase with the system noise and error.In the end, in order to improve the robustness of the conventional algorithms, the Huber loss function is introduced instead of the squared loss function. Then the inversion algorithm can be called robust regularized algorithm. It has been proved to be valid by simulations as well as some experimental results.