군사용 SAR 이미지 초해상화를 위한 딥러닝 기반의 네트워크 구조에 관한 연구

• Published in: 한국CDE학회 논문집 Vol. 26; no. 2; pp. 154 - 162
• Main Authors: 류제민(Jemin Ryu), 마정목(Jungmok Ma)
• Format: Journal Article
• Language: Korean
• Published: (사)한국CDE학회 01.06.2021; 한국CDE학회
• Subjects: 기계공학; Deep learning; Synthetic aperture radar image; Super-resolution
• ISSN: 2508-4003; 2508-402X
• DOI: 10.7315/CDE.2021.154

The Republic of Korea military is using SAR(Synthetic Aperture Radar) geographic intelligence to deal with security threats. However, human experts have difficulty on analyzing acquired SAR images and identifying military targets due to low resolution. In this paper, we study the deep learning-based network architecture fit for the super-resolution of military SAR images. Previous military SAR image super-resolution studies mainly conducted on improving the results of super-resolution, but it was difficult to find studies on network architecture. The proposed neural network is a deep learning-based super-resolution networks. And it consists of input, learning, upsampling, and output layers with real military SAR images. We show and experiment with networks for super-resolution of military SAR images, while focusing on the input and upsampling layers. Experiment results show that we able to find a suitable architecture of input and upsampling layers is discussed. KCI Citation Count: 3