Deep Learning-based Resolution Enhancement in SAR Image for Automotive Radar Sensors

• Published in: 2023 IEEE SENSORS pp. 1 - 4
• Main Authors: Kang, Sung-wook, Cho, Hahng-Jun, Lee, Hojung, Lee, Seongwook
• Format: Conference Proceeding
• Language: English
• Published: IEEE 29-10-2023
• Subjects: automotive radar sensor; generative adversarial network; Generative adversarial networks; Image resolution; Image sensors; Radar imaging; Radar polarimetry; Sensor systems; Sensors; synthetic aperture radar
• ISSN: 2168-9229
• DOI: 10.1109/SENSORS56945.2023.10325124

Forming high-resolution synthetic aperture radar (SAR) images requires large amounts of sampled data, which increases computation time and complexity. Therefore, in this paper, we propose a method to enhance the resolution of SAR images for automotive radar sensors using a generative adversarial network (GAN). The proposed GAN is an unsupervised image-to-image translation GAN based on a variational autoencoder and can form high-resolution SAR images from a small amount of sampled data. The SAR images formed by the proposed method are compared in terms of peak signal-to-noise ratio and structural similarity index measure for performance evaluation, and they are increased by 2.75% and 4.43%, respectively, compared to the existing low-resolution SAR images.