首页|基于移动设备机器视觉的无砟轨道钢轨爬行状态检测方法

基于移动设备机器视觉的无砟轨道钢轨爬行状态检测方法

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
无缝线路钢轨爬行是由于列车的启制动以及外部温度变化导致的钢轨沿纵向发生位移,准确检测钢轨的位移状态变化,在铁路运营安全管理中是一项重要的任务.传统的人工测量方法受限于时间、空间和精度约束,检测效果不够准确.从实际运维环境出发,提出一种基于车载式机器视觉的无砟轨道钢轨纵向位移检测方法.研究包括:在钢轨与轨道板上粘贴位移标志物,使用实验室自研探伤小车搭载线激光扫描仪获取高精度图像信息;提出一种基于改进Yolov8-Pose的关键点检测方法,对钢轨及轨道板上的位移标志物关键点进行识别并确定感兴趣区域;基于Harris亚像素检测方法对标志物的亚像素角点进行定位,进一步提高检测精度;在无砟轨道试验段进行测试,对钢轨纵向位移进行高精度检测.实验结果表明:钢轨纵向位移检测算法在不同实际位移下的均方根误差平均值为0.202 mm,最大值不超过0.5 mm,符合无砟轨道的检测要求.该方法可以在线路上多点连续进行钢轨纵向位移检测,提高了铁路运维效率,确保铁路行车的安全平稳.
Detection Method of Rail Creeping of Ballastless Track Based on Machine Vision of Mobile Equipment
The longitudinal movement of CWR is caused by the train's starting and braking,as well as external temperature fluctuations.Accurately detecting changes in the displacement state of the rails is a crucial task in railway operation safety management.Traditional manual measurement methods are limited by constraints of time,space and accuracy,leading to insufficient detection effectiveness.Based on the practical O&M environment,this paper proposes a longitudinal displacement detection method for ballastless tracks using onboard machine vision technology.The research includes pasting displacement markers on rails and track slabs,and using the laboratory's self-developed flaw detection trolley mounted with a line laser scanner to obtain high-precision image information;this paper puts forward a key point detection method based on improved Yolov8-Pose to identify the key points of displacement markers on rails and track slabs and determine the region of interest.Based on Harris subpixel detection method,the subpixel comers of markers are located to further improve the detection accuracy;in the test section of ballastless track,the longitudinal displacement of rail is detected with high precision.The experimental results demonstrate that the longitudinal displacement detection algorithm for the rails yields an average root mean square error of 0.202 mm under various actual displacements,with a maximum value not exceeding 0.5 mm.This performance meets the detection requirements for ballastless tracks.This method allows for continuous longitudinal displacement detection of the rails at multiple points along the track,improving railway O&M efficiency while ensuring the safety and smoothness of train operations.

CWRlongitudinal displacement of railmachine visionkey point detectionsubpixel corner detectiondeep learning

刘震、郑祯国、张世杰、郭积程、王平、何庆

展开 >

中铁第四勘察设计研究院集团有限公司,湖北武汉 430063

中国铁路南昌局集团有限公司福州工务段,福建福州 350013

西南交通大学土木工程学院,四川成都 610031

西南交通大学高速铁路线路工程教育部重点实验室,四川成都 610031

展开 >

无缝线路 钢轨纵向位移 机器视觉 关键点检测 亚像素角点检测 深度学习

科技部重点研发计划项目国家自然科学基金面上项目中国铁建股份有限公司重大项目中铁第四勘察设计研究院集团有限公司科技研究开发计划项目

2023YFB2603700523724002021-A032022K053

2024

10.19549/j.issn.1001-683x.2024.06.19.001

中国铁路
中国铁道科学研究院

中国铁路

北大核心
影响因子：0.407
ISSN：1001-683X
年,卷(期)：2024.(9)
  • 6