Abstract: An improved algorithm called RSC-YOLO based on YOLOv11n is proposed to address the problems of complex backgrounds, dense objects, and mutual occlusion in object detection based on uncrewed aerial vehicle (UAV) platforms. Firstly, the representative feature alignment (RFA) sensing field attention mechanism is introduced, and the Conv and C3k2 models are re-designed, which effectively solves the parameter sharing problem caused by the overlapping of spatial features. Secondly, a shallow detail fusion module(SDFM) is designed for the neck network of YOLO, which emphasizes the attention to shallow features, compensates for the missing features of small objects as well as maintains the integrity of the remaining spatial information of occluded objects. Finally, a lightweight cross-scale feature fusion network CCFF is introduced for the neck, which enhanced the model's feature interaction capability and multi-scale feature fusion capability to achieve a balance between accuracy and detection speed. Ablation and comparison experiments were conducted on the Visdrone2019 dataset, and the mAP₅₀ was improved by 4.5% compared with the baseline algorithm. The framerate of the system reached 213 FPS, which is sufficient for real-time detection. Generalization experiments were conducted on the DOTA dataset and the HIT-UAV infrared small object dataset, and the results show that the mAP₅₀ was improved by 3.1% and 3.9%, respectively, which demonstrates that the proposed algorithm has wide applicability.