<p>The underwater environment is complex and ever-changing, with light being easily absorbed and scattered, resulting in blurred and color-shifted images. This poses challenges for target detection, often leading to false positives and missed detections. To enhance the accuracy and stability of underwater target detection, an improved YOLO11 algorithm is proposed. First, the C3k2 module is enhanced by introducing a Dynamic Feature Fusion (DFF) mechanism, resulting in the C3k2-DFF module. Leveraging its dynamic weighting capability, this module enhances the representation of multi-scale features, enabling more precise localization of underwater targets. Second, by introducing a Global Context (GC) mechanism to enhance the Visual Feature Enhancement (VFE) module, the VFE-GC module is designed to adaptively recalibrate features using global information, suppressing background noise while amplifying key semantic information of targets. Finally, the Shape-IoU loss function is improved by incorporating the Normalized Wasserstein Distance (NWD), resulting in the Shape-NWD loss function. The NWD loss function optimizes bounding box regression by enforcing shape consistency constraints, thereby enhancing localization accuracy for irregular or elongated underwater targets. Results show that DVS-YOLO11 achieves an mAP@0.5 of 84.78% and an mAP@0.5:0.95 of 64.67%, along with 87.44% Precision and 75.57% Recall, while maintaining low computational costs. Notably, for scallop targets characterized by blurry edges and irregular shapes, the mAP@0.5 improved from 63.9% to 68.6%. Compared with the baseline YOLO11, all metrics show improvements, indicating that the proposed method works well for underwater target detection.</p>