Expanding Defocused Ship Data Using Existing SAR Ship Datasets by Inverse Refocusing Algorithm

The synthetic aperture radar (SAR) is a crucial tool for maritime observation, with ships being the main targets at sea. Detecting these ships is foundational for other downstream tasks, making the study of ship detection algorithms highly significant. Currently, ship detection primarily relies on deep learning algorithms, and training neural networks requires a large amount of data. In various maritime observation tasks, identifying defocused ships is particularly important. However, current databases lack defocused ship data, and SAR ship image generation algorithms are not yet mature. Therefore, this letter proposes generating defocused ship data using the existing ship data. First, we introduce the signal model of defocused ships and the causes of defocusing. Next, we describe the generation of defocused ships with nonuniform rotation by introducing motion errors through resampling. We also explain the generation of defocused ships with translational motion using initial phase compensation and envelope shift. Finally, expansion experiments using spaceborne SAR data demonstrate the effectiveness of our method.