To address the problems of insufficient detection accuracy and poor real-time performance caused by small detection objects, complex backgrounds, and large calculation volume in routine highway inspection tasks using unmanned aerial vehicles (UAVs), this paper proposed an improved UAV-based routine highway inspection algorithm relying on YOLOv11n, YOLOv11-UR. The algorithm incorporated the multi-level channel and attention (MLCA) mechanism into the backbone network to fuse channel and spatial dimension information, integrate local and global receptive fields, and effectively enhance feature expression capability. It achieved a significant improvement in detection accuracy with only a small increase in parameters. In the Neck part, GSConv was introduced to replace standard convolution, making the output of the convolution calculation as close as possible to standard convolution while reducing computational costs. VoVGSCSP was introduced to replace C3k2, reducing the model parameter volume and computational complexity while enhancing feature extraction capability. Experimental results show that the YOLOv11-UR algorithm has significant advantages in UAV-based routine highway inspection. It effectively reduces the parameter volume and computational overhead of the model without losing much inference speed. The model detection precision (R_P) and mean average precision (m_AP) reach 78.26% and 73.34%, respectively. The improved algorithm balances detection accuracy and inference efficiency and can better meet the needs of UAV-based routine highway inspection.