TrackDef-YOLO：一种改进的YOLOv11模型用于铁轨表面缺陷检测
TrackDef-YOLO: An Improved YOLOv11 Model for Railway Track Surface Defect Detection

随着铁路运输的快速发展，铁路轨道的安全性与可靠性成为重要的研究课题。传统的人工检测方法效率低且易受人为因素影响。近年来，深度学习，尤其是YOLO系列模型在目标检测中的应用取得了显著进展。为提高铁轨表面缺陷检测的精度和实时性，本文提出了一种改进的YOLOv11模型(TrackDef-YOLO)。通过引入多尺度特征融合、自动锚框调整和加权损失函数等创新技术，TrackDef-YOLO显著提升了对微小缺陷和复杂背景的检测能力。实验结果表明，TrackDef-YOLO相比YOLOv8和YOLOv11n模型在准确率、召回率、F1-Score和误报率等方面均有显著提升。该研究为铁路轨道智能检测提供了高效、精准的解决方案，并为基于深度学习的铁路维护系统的优化奠定了基础。
With the rapid development of railway transportation, the safety and reliability of railway tracks have emerged as critical research topics. Traditional manual inspection methods are inefficient and susceptible to human factors. In recent years, deep learning, especially YOLO series models, has achieved remarkable progress in object detection. To enhance the accuracy and real-time performance of railway track surface defect detection, this paper proposes an improved YOLOv11 model, named TrackDef-YOLO. By incorporating innovative techniques such as multi-scale feature fusion, automatic anchor box adjustment, and a weighted loss function, TrackDef-YOLO significantly improves the detection capability for minute defects and complex backgrounds. Experimental results demonstrate that TrackDef-YOLO outperforms YOLOv8 and YOLOv11n models in terms of accuracy, recall, F1-Score, and false alarm rate. This study provides an efficient and precise solution for intelligent railway track inspection and lays the foundation for optimizing railway maintenance systems based on deep learning.