Existing single-stage object detection models face significant challenges in maritime search and rescue scenarios.Traditional fixed-parameter models struggle to adapt to variable sea conditions

and single-stage models exhibit insufficient capability in detecting small targets

leading to frequent missed and false detections.To address these issues

a dynamic-convolution

dynamic-upsampling and Wasserstein-loss YOLO network(DDW-YOLO) is proposed for maritime target detection by improving YOLO11 model.The original convolution and upsampling modules in the model are dynamically improved to realize adaptive adjustments of convolutional kernels and sampling offsets

which effectively improves the model’s adaptability to complex maritime environments.Additionally

DDW-YOLO incorporates a new small target detection layer based on a feature fusion strategy

which significantly enhances its ability to capture small targets.Ablation experiments on the SeaDronesSee v2 dataset demonstrate that compared to the baseline

the model achieves improvements of 2.4% in precision and 5.3% in recall

providing an efficient and reliable visual detection solution for unmanned aerial vehicle-based maritime search and rescue missions.