Abstract: In response to the challenges of large target scale differences and complex backgrounds leading to low detection accuracy of overpasses in remote sensing images,this paper proposes a solution based on the YOLOv11 framework.The approach introduces hypergraph computing to explore the intrinsic relationships between cross-layer features,and presents a parallel processing channel selection module based on a gating mechanism that dynamically selects and strengthens task-relevant features,enhancing the model's ability to focus on key information.Additionally,a learnable scaling factor is embedded in the regression part to construct a regression-optimized detection head,improving the adaptive ability of bounding box predictions and enhancing network performance.Experimental results show that,on the Dior dataset for the overpass category,the proposed algorithm achieves mAP＿50 and mAP＿50:95 of 80.5%and 53.1%,respectively,outperforming the comparison algorithms,effectively improving the detection accuracy and robustness of overpass targets in complex backgrounds.

Key words: remote sensing imagery, overpass, object detection, YOLOv11, hypergraph, gating mechanism