With the rapid development of unmanned aerial vehicle (UAV) technology in urban security, traffic monitoring, and emergency response, object detection and recognition based on UAV imagery have become essential for supporting a wide range of intelligent applications. However, UAV-based object detection from low-altitude perspectives remains highly challenging due to the dense distribution of small objects, significant scale variations, and complex background interference. To address these issues, this paper proposes an enhanced UAV-oriented YOLO11 object detection algorithm. First, a CSP-MS module is designed to improve multi-scale feature representation through hierarchical fusion and heterogeneous convolution structures. Second, an feature-enhanced multi-scale aggregation pyramid is introduced, which combines dilated convolutions with cross-layer fusion to strengthen the model’s perception capability in complex scenes. Finally, a lightweight dynamic task-aligned detection head is integrated to reduce model parameters while improving detection accuracy for small objects. Experiments on the VisDrone dataset demonstrate improvements of 10.2% in mAP0.5and 6.7% in mAP0.5:0.95.On the CODrone dataset, the proposed method achieves gains of 5.4% and 3.7%, respectively. Overall, the results show that the improved model delivers notable advantages in detecting small objects, handling complex backgrounds, and managing multi-scale targets, highlighting its strong generalization capability and practical applicability.