摘要
在介绍轨道伤损检测系统组成的基础上,分析YOLOv5s算法模型原理,针对采用YOLOv5s算法模型检测钢轨轨面掉块和夹板裂纹等小目标伤损效果不佳的问题,通过数据增强策略和算法模型改进来优化YOLOv5s算法,并进行试验和现场应用验证。结果表明:改进后的YOLOv5s算法模型平均精确率提升7.7%,检测速度增加5.7帧/秒,使轨道伤损检测系统能有效检测出钢轨轨面掉块、夹板裂纹等细小伤损。YOLOv5s算法模型的优化,为轨道伤损检测系统准确检测钢轨轨面掉块和夹板裂纹等小目标伤损提供理论依据。
On the basis of introducing the composition of the track defect testing system,the principle of the YOLOv5s algorithm model is analyzed.In response to the problem of poor performance of the YOLOv5s algorithm model for testing small target defect such as rail surface drops and clamp cracks,the YOLOv5s algorithm is optimized through data augmentation strategy and algorithm model improvement,and experimental and field application verification are conducted.The results showed that the average accuracy of the improved YOLOv5s algorithm model is improved by 7.7%and its testing speed is increased by 5.7 frames per second,which enables the track defect testing system to effectively test small defects such as rail surface drops and clamp cracks.The optimized YOLOv5s algorithm model provides a theoretical basis for accurately testing small target defects such as rail surface drops and clamp cracks by applying the track defect testing system.
作者
程晗
辜刚
邴丽红
韩兵
章罕
随海亮
CHENG Han;GU Gang;BING Lihong;HAN Bing;ZHANG Han;SUI Hailiang
出处
《铁道技术监督》
2024年第9期36-38,43,共4页
Railway Quality Control
