Unmanned aerial vehicle maritime rescue object detection based on improved YOLO11

Target detection based on deep learning is a critical visual detection method for maritime rescue. However, the current methods face significant challenges in complex marine environments, including inadequate feature extraction for small-to-medium targets, suboptimal computational efficiency for unmanned aerial vehicle deployment, and ineffective multi-scale feature integration. To tackle these challenges, we introduce an improved model that builds upon YOLO11. Specifically, we integrate the spatial and channel reconstruction convolution module, which optimizes feature extraction capabilities within the C3k2 framework. Furthermore, to further augment the model’s sensitivity and selectivity toward crucial features, we incorporate the squeeze-and-excitation attention mechanism. At the same time, a simplified spatial pyramid pooling fast module is introduced, which can improve computational efficiency while ensuring accuracy. In addition, the adoption of an asymptotic feature pyramid network facilitates robust and effective multiscale feature extraction, improving YOLO11’s capacity to detect objects at various dimensions. Finally, we utilize the wise intersection over union vision 3 to achieve a harmonious balance among regression penalties, thereby enhancing the model’s overall performance even further. The analytical findings indicate that, in comparison to the baseline YOLO11 model, the enhanced version achieves a 0.75% improvement in accuracy, a 5.63% improvement in mAP50, and a 1.76% improvement in mAP50-95. In addition, the model has reduced its parameter count by 16.12%, thereby successfully attaining an equilibrium between precision in detection and efficiency in computation.