Channel Error Estimation Algorithm for Multichannel in Azimuth HRWS SAR System Based on a 3-D Deep Learning Scheme

Abstract

High-resolution and wide-swath (HRWS) multichannel synthetic aperture radar (SAR) provides extensive imaging coverage, playing a pivotal role in remote sensing applications. Although multichannel in azimuth SAR system has been proposed to deal with the contradiction problem between high resolution and low pulse repetition frequency, the channel errors caused by temperature, timing uncertainty and other factors may result in azimuth ambiguity and defocus. To address this issue, a deep learning-based channel calibration method is proposed in this article, in which multichannel errors can be simultaneously estimated to improve the performance of conventional separate channel estimation. Specifically, an end-to-end strategy over 3-D convolutional neural networks (CNNs) is proposed to estimate multichannel errors collaboratively by fully exploiting the correlation of both innerchannel and intrachannel signals. Furthermore, a simulation-based training data synthesis strategy is proposed to generate training samples with similar signal characteristics with the scene to be reconstructed, by which the proposed 3-D CNN can be well trained without real multichannel signals. Experiments over both simulated and real measured data demonstrate that the proposed deep learning-based channel calibration method can well estimate multichannel errors simultaneously to improve the performance of HRWS SAR imaging.

Keywords

• Channel calibration
• convolutional neural networks (CNNs)
• deep learning
• high-resolution wide-swath (HRWS)
• multichannel synthetic aperture radar (MC-SAR)