Enhancing Dense Ship Detection in SAR Images Through Cluster-Region-Based Super-Resolution

Synthetic Aperture Radar (SAR), with its all-weather and all-day imaging capabilities, has become an essential tool in ship monitoring. In recent years, deep learning has significantly enhanced the performance of ship detection in SAR images. However, in complex coastal scenarios, ships are often densely clustered with arbitrary orientations, and the speckle noise and scattering properties make it challenging to distinguish adjacent objects. To address these issues, this paper proposes an integrated ship detection method in SAR images incorporating cluster-guided super-resolution. First, a ship cluster detection pipeline is designed, which supports automatic cluster region labeling, training, and prediction. Secondly, a super-resolution module is developed to reconstruct and enhance the object features within cluster regions. Additionally, an oriented ship detection network is employed for fine-grained ship detection within cluster regions, and an ensemble learning strategy is utilized to optimize the detection results. Experimental results on the SSDD and RSDD-SAR datasets demonstrate that the proposed method significantly improves detection accuracy in nearshore scenarios, especially for densely arranged ships.