Synthetic aperture radar (SAR) ship detection faces challenges such as complex background interference, multi-scale variations, and small target feature attenuation. To address the trade-off between accuracy and computational efficiency and the insufficient extraction of small-target features, this paper proposes HDF-YOLO, an improved YOLOv11-based high-precision detection method. Firstly, the C3k2_HetConv module replaces the standard C3k2 structure, enhancing edge texture characterization via heterogeneous convolution for better detection of nearshore and small targets. Secondly, the DySample module is integrated into the neck network to preserve target details while suppressing background noise. Thirdly, the Four-directional Adaptive Spatial Feature Fusion Head (FASFFHead) is designed for efficient multi-scale feature fusion. Finally, the Inner-MPDIoU loss function optimizes localization accuracy in complex backgrounds. Evaluations on HRSID and SSDD datasets show that HDF-YOLO achieves 94.3%/71.5% and 98.7%/71.8% (mAP@0.5 /mAP@0.5:0.95), respectively, consistently outperforming YOLOv11 and other state-of-the-art methods.