HPRT-YOLO: Enhanced lightweight UAV maritime search and rescue model based on YOLOv11n

Small target detection is crucial for unmanned aerial vehicle (UAV) maritime search and rescue (SAR), but problems, including low target spatial resolution, dynamic marine environments, and severe background interference, often cause frequent missed/false detections and insufficient real-time performance. To address these issues, we propose HPRT-YOLO, a lightweight model based on YOLOv11n. We present the C3k2_Faster_CGLU module to enhance small-target feature representation and reduce computational complexity, strengthen multi-scale fusion via the ASF-P2 network with Zoom_cat and ScalSeq, adopt LAMP pruning to eliminate redundancy, utilize BCKD knowledge distillation to retain feature extraction capability while cutting parameters, and employ a WIoU-v2+MPDIoU hybrid loss to optimize recall and localization. On two public maritime SAR datasets, HPRT-YOLO outperforms YOLOv11n by 7% in mAP50 and 9.5% in recall, compresses to 2.3 MB (58.19% smaller), and maintains 94.9 FPS at a batch size of 1, providing efficient technical support for UAV maritime SAR.