Research on a SAR Image Ship Detection Algorithm Based on Feature Refinement

SAR images have long been plagued by coherent speckle noise and challenges in detecting small ships within the field of ship detection. This paper introduces a synthetic aperture radar (SAR) image ship detection algorithm based on feature refinement. The algorithm employs a denoising module to mitigate noise interference and captures more spatial position information through a hierarchical feature fusion (HFF) module, enhancing the detection precision of small ships. Lastly, the algorithm leverages a feature refinement (FR) module based on deformable convolution to minimize clutter signal interference, thus enhancing the accuracy of ship positioning. Experimental results demonstrate that the algorithm exhibits high accuracy and robustness in detecting ships from SAR images.