Achieving Ship Target Detection in Radar Images via Dual-Route Feature Extraction and Adjacent-Layer Feature Fusion

Focusing on the problem of balancing computational cost and detection accuracy in marine SAR image target detection, a lightweight method based on Dual-Route Feature Extraction and Adjacent-Layer Feature Fusion is proposed. In the DRFE backbone, input features are equally divided in the channel dimension, and feature extraction is conducted separately, significantly reducing computational parameters. To effectively handle the scale and rotation diversity of ships, deformable convolution is employed. Furthermore, to reduce the damage caused by non-linear operations to the hierarchical correlation of the feature pyramid, the Adjacent-Layer Feature Fusion (ALFF) strategy, which generates a feature output with a pyramid-like structure. The strategy enhances the preservation of hierarchical features while maintaining detection accuracy. Experimental results on the publicly available SAR Ship Detection Dataset (SSDD) demonstrate that our proposed method not only maintains high detection accuracy but also significantly reduces the computational cost of the network, making it suitable for real-time applications.