An Optical-to-SAR Transformation Method for SAR Ship Image Augmentation

摘要

For ship detection in remote sensing, the labeled synthetic aperture radar (SAR) ship images are far less than optical ship images, which limited the development of SAR ship detection and classification. In this paper, an optical-to-SAR transformation method is proposed to transforming the well labeled optical ship images to SAR ship images. Based on electromagnetic theory, the backscattering and reflection mechanism is analyzed and modeled into the transformation. A background iteration method based on probability theory is also proposed to learn the distribution of SAR ship images. The result shows that the proposed method can transform the optical ship images to SAR ship images with the confidence coefficient of 84.643%.