Deep-Learning Method for Channel-Calibration of Multichannel in Azimuth SAR System

Multi-channel in azimuth synthetic aperture radar (SAR) system can deal with the contradiction between high-resolution and low pulse repetition frequency in high-resolution and wide-swath (HRWS) imaging. However, the channel errors caused by temperature, timing uncertainty and other factors may result in azimuth ambiguity and defocus. To address this issue, a channel-calibration method based on deep learning is proposed in this paper. Firstly, a simulation dataset is made for network training, which solves the problem of lack of SAR data. Then, an end-to-end method based on the convolutional neural network (CNN) for multi-channel SAR data is designed to estimate the channel phase errors. The network can take into account the correlation between the channels in azimuth. Finally, the experiments validate the effectiveness of the proposed calibration method. Compared with the conventional channel phase error estimation methods, the accuracy of the proposed method is higher.