作者：胡成放1, 丁昊昊2, 陈德君3, 张岩3, 刘启跃2, 王文健2, 郭俊2, 林强2,4
作者单位：1. 西南交通大学 唐山研究院，河北 唐山 063000;
2. 西南交通大学 轨道交通运载系统全国重点实验室 摩擦学研究所，四川 成都 610031;
3. 唐山百川智能机器股份有限公司，河北 唐山 063000;
4. 浙江师范大学 浙江省城市轨道交通智能运维技术与装备重点实验室，浙江 金华 321004
关键词：双目立体视觉;迭代最近点;型面参数;车轮擦伤;模式匹配
摘要：
列车轮对表面缺陷及磨耗后的车轮轮型参数对列车安全行驶具有重要影响。搭建一种基于结构光与双目立体视觉相结合的非接触式列车轮对型面检测系统，设计针对车轮滚动圆直径、轮缘高度、轮缘厚度以及车轮擦伤的双目视觉无损检测方法。首先基于迭代最近点(iterative closest point, icp)算法将各相机采集的车轮型面数据拼接为整体车轮点云三维模型；然后，从该三维模型中提取出滚动圆与轮缘顶点圆，基于最小二乘拟合法分别计算滚动圆直径、轮缘高度、轮缘厚度参数；最后，基于模式匹配方法检测车轮型面是否出现擦伤缺陷，并计算得到擦伤深度。检测结果表明：该列车轮对型面检测系统及表征方法对于滚动圆直径检测误差为0.22 mm，对于轮缘高度与轮缘厚度检测误差分别为–0.08 mm及0.07 mm，最大擦伤深度检测误差为0.18 mm。研究成果可有效检测列车车轮型面参数及擦伤缺陷，具有较强的工程应用价值。
inspection method for surface defect and shape parameter of train wheelset based on binocular vision
hu chengfang1, ding haohao2, chen dejun3, zhang yan3, liu qiyue2, wang wenjian2, guo jun2, lin qiang2,4
1. tangshan institute, southwest jiaotong university, tangshan 063000, china;
2. tribology research institute, state key laboratory of rail transit vehicle system, southwest jiaotong university, chengdu 610031, china;
3. thangshan baichuan intelligent machine co., ltd., tangshan 063000, china;
4. key laboratory of urban rail transit intelligent operation and maintenance technology & equipment of zhejiang province, zhejiang normal university, jinhua 321004, china
abstract: the surface defects of train wheelset and the shape parameters of worn wheels have an important impact on the safe operation of trains. a non-contact train wheelset shape detection system based on the combination of structured light and binocular stereo vision is built, and a binocular vision nondestructive inspection method is designed for wheel rolling circle diameter, flange height, flange thickness and wheel flats. firstly, the wheel profile data collected by each camera were spliced into a 3d model of the whole wheel point cloud based on the iterative closest point (icp) algorithm. then the rolling circle and flange vertex circle were extracted from the 3d model, and the rolling circle diameter, flange height and flange thickness parameters were calculated respectively based on the least square fitting method. finally, the method based on pattern matching was used to detect whether the wheel profile has flat defects, and the flat depth was calculated. the detection results of the detection system and characterization method of the train wheelset shape show that the measuring errors of rolling circle diameter, flange height, flange thickness and the maximum flat depth are 0.22, –0.08, 0.07 and 0.18 mm, respectively. the research result can effectively detect the wheel profile parameters and wheel flats, and has high engineering application value.
keywords: binocular stereo vision;iterative closest point;shape parameters;wheel flat;pattern matching
2024, 50(8):101-108 收稿日期: 2023-02-25;收到修改稿日期: 2023-06-05
基金项目: 国家自然科学基金（52205578）；中国博士后科学基金面上项目（2021m702711）；中央高校基本科研业务费专项资金（2682022cx009）；浙江省城市轨道交通智能运维技术与装备重点实验室开放课题基金（zsdrtkf2021003）
作者简介: 胡成放（1997-），男，山东济宁市人，硕士研究生，专业方向为机车车轮双目视觉检测技术。
参考文献
[1] 孙琦, 张兵, 李艳萍, 等. 一种波长固定的车轮多边形在线故障检测方法[j]. 铁道科学与工程学报, 2018, 15(9): 2343-2348.
sun q, zhang b, li y p, et al. wavelength-fixing mechanisms for detecting the wheel polygon-shaped fault onsite[j]. journal of railway science and engineering, 2018, 15(9): 2343-2348.
[2] 阮成雄, 阮秋琦. 基于log梯度加权haar-like特征的车轮裂纹检测[j]. 铁道学报, 2013, 35(5): 62-68.
ruan c x, ruan q q. wheel crack detection based on log gradient weighted haar-like features[j]. journal of the china railway society, 2013, 35(5): 62-68.
[3] 戚潇月, 宋冬利, 张卫华. 车轮多边形对车辆动力学的影响分析及在线诊断方法研究[j]. 铁道机车车辆, 2018, 38(4): 10-17.
qi x y, song d l, zhang w h. analysis influence of wheel polygonalization on vehicle dynamics and research on online diagnosis[j]. railway locomotive & car, 2018, 38(4): 10-17.
[4] zhou l, brunskill h, pletz m, et al. real-time measurement of dynamic wheel-rail contacts using ultrasonic reflectometry[j]. journal of tribology, 2019, 141(6): 061401.
[5] roveri n, carcaterrra c, sestieri a. real-time monitoring of railway infrastructures using fibre bragg grating sensors[j]. mechanical systems and signal processing, 2015, 60: 14-28.
[6] brizuela j, fritsch c, ibanez a. railway wheel-flat detection and measurement by ultrasound[j]. transportation research part c, 2011, 19(6): 975-984.
[7] pedro u, sergio m, javier f, et al. wheel-rail contact force measurement using strain gauges and distance lasers on a scaled railway vehicle[j]. mechanical systems and signal processing, 2020, 138: 106555.
[8] song y, liang l, du y l, et al. railway polygonized wheel detection based on numerical time-frequency analysis of axle-box acceleration[j]. applied sciences, 2020, 10(5): 309-321.
[9] wang y w, ni y q, wang x. real-time defect detection of high-speed train wheels by using bayesian forecasting and dynamic model[j]. mechanical systems and signal processing, 2020, 139: 106654.
[10] liu x z, xu c, ni y q. wayside detection of wheel minor defects in high-speed trains by a bayesian blind source separation method[j]. sensors, 2019, 19(18): 3981.
[11] 周茂, 胡立锦, 欧开鸿, 等. 改进稠密双目匹配算法在输电线路基础三维重建的应用研究[j]. 电子测量术, 2022, 45(7): 1-7.
zhou m, hu l j, ou k h, et al. application of 3d rebuild based on improved dense binocular matching algorithm in transmission line foundation positioning and measurement[j]. electronic measurement technology, 2022, 45(7): 1-7.
[12] 伍川辉, 邓越, 于涛, 等. 受电弓滑板磨耗检测中多视觉传感器标定方法研究[j]. 中国测试, 2023, 49(1): 7-13.
wu c h, deng y, yu t, et al. research on the calibration method of multi-vision sensors in the wear detection of pantograph skateboard[j]. china measurement & test, 2023, 49(1): 7-13.
[13] 王永信, 卢秉恒, 梁晋, 等. 一种三维人体模型快速测量方法[j]. 中国测试, 2023, 49(3): 114-119.
wang y x, lu b h, liang j, et al. a fast measuring method of three-dimensional human body model[j]. china measurement & test, 2023, 49(3): 114-119.
[14] 贾传宝, 吴开华, 陈强元. 倾斜位置成像的轮对几何参数计算方法[j]. 传感器与微系统, 2017, 36(2): 150-153.
jia c b, wu k h, chen q y. calculation method of wheel set geometrical parameters by imaging at inclined position[j]. transducer and microsystem technologies, 2017, 36(2): 150-153.
[15] chen x, sun j h, liu z. dynamic tread wear measurement method for train wheels against vibrations[j]. applied optics, 2015, 54(17): 5270-5280.
