【摘要】 地面标志线检测在自动驾驶和交通场景分析中起着重要的作用，对于实现道路安全和道路智能化至关重要。然而，传统的标志线检测算法存在着检测精度较低和交通箭头标志线相关检测研究较少的问题。为应对此类问题，提出了一种基于YOLOv8n改进的交通标志识别算法。改进包括使用Timm模型库中的Res Net-18网络替换YOLOv8n模型的backbone网络，以提升图像识别精度。采用GIoU边界损失函数替代原有的CIoU损失函数，提高边界框回归性能的同时进一步提升检测效率和准确率。基于Cey Mo数据集中的2 099张地面标志线图像进行了训练和评估。实验结果表明，原始的YOLOv8n模型在精度（Precision）上为82.2%，平均精度均值（mAP）为98%，而经过该方法优化后的模型达到了88.1%的精度和99.3%的mAP，分别使模型的精度提高了5.9%，平均精度均值提高了1.3%。综合分析，在引入ResNet-18 Backbone网络和GIoU损失函数后，不仅提高了检测效率，也提高了识别精度，而且明显优于YOLOv5s和YOLOv8n算法，具有更高的有效性和检测精度。更多还原

【Abstract】 Ground marker detection plays an important role in autonomous driving and traffic scenario analysis, which is essential for road safety and road intelligence. However, the traditional sign line detection algorithm has the problems of low detection accuracy and few studies on the detection of traffic arrow markers. In order to solve such problems, an improved traffic sign recognition algorithm based on YOLOv8n was proposed. Improvements include replacing the backbone network of the YOLOv8n model with the ResNet-18 network in the Timm model library to improve the accuracy of image recognition.The GIoU boundary loss function is used to replace the original CIoU loss function to improve the regression performance of the bounding box and further improve the detection efficiency and accuracy. In the experiment, 2 099 ground marker line images in the CeyMo dataset were trained and evaluated. Experimental results show that the original YOLOv8n model has a precision of 82.2% and mean average precision of 98%, while the optimized model achieves a precision of 88.1% and mean average precision of 99.3%, which improves the precision of the model by 5.9% and the average precision by 1.3%, respectively.Comprehensive analysis shows that after the introduction of ResNet-18 Backbone network and GIoU loss function, it not only improves the detection efficiency but also improves the recognition precision, and is significantly better than the YOLOv5s and YOLOv8n algorithms, with higher effectiveness and detection precision.更多还原