Airborne Radar Forward-Looking Imaging Algorithm Based on Generative Adversarial Networks

Radar forward-looking imaging is gaining significance due to its convenience in various applications like battlefield reconnaissance, target surveillance and precision guidance. Although synthetic aperture radar (SAR) techniques are commonly used to achieve high azimuth resolution, they suffer from limitations in forward-looking area due to the poor Doppler resolution and the “left-right” ambiguity problem. In recent years, generative adversarial networks (GANs), a common deep learning approach that produces excellent results in image motion blur removal, has been extensively used. This letter proposes building an end-to-end forward-looking imaging network using GAN to produce high-resolution images, which increases the efficiency and quality of imaging. Compared to conventional forward-looking imaging methods such as the deconvolution-based methods, this algorithm eliminates the design and iterative processes of the observation matrix. Simulated and real radar data verified that this approach offers robust recovery and better performance.