基于几何特征的钢轨轨廓高精度匹配方法研究

发布时间：2018-08-28 20:47

【摘要】：钢轨轨廓的高精度检测是钢轨廓形、磨耗精确测量的基础。快速轨廓仪是一种利用多个线结构光平面共面安装的二维激光摄像式传感器组合进行钢轨轨廓非接触检测的典型方法。各激光传感器严格对称安装是保证各传感器数据正确搭接、保证快速轨廓仪测量精度的关键,这会增加测量系统制造、装配与调试的困难。为了降低对称性对测量精度的影响,对传感器线结构光平面进行标定是一种可行的解决方案。在分析与总结现有的线结构光平面共面法标定方法的基础上,本文提出一种新型的线结构光平面标定法—等高线法,该方法以00级平面度标准量块为基准,等间距提取适量的数据点,通过角度变换,并采用Levenberg-Marquardt算法对其进行非线性迭代使它们处于同一等高线上。标定试验结果表明,等高线标定法重复性可达0.00133°,标定后的轨廓仪检测精度可达7.7?m,且具有设备要求低、操作简单的特点。通常,快速轨廓仪安装在轨检车上,通过实际检测钢的轨轨廓标准钢轨轨廓匹配可实现钢轨廓形、磨耗的连续、快速精确测量。在钢轨轨廓检测过程中,由于轨道不平顺会引起轨检车振动、蛇行、侧倾、扭曲等,导致激光摄像式传感器采集的数据无法直接进行钢轨轨廓的匹配。为解决车体振动对钢轨轨廓匹配带来的影响,本文提出一种基于钢轨固有几何特性的轨廓匹配方法。用已标定的线激光传感器获取钢轨轨廓的空间位置信息;利用Hough变换检测钢轨外侧轨头曲率半径为13 mm和内、外侧轨腰处曲率半径为20 mm的圆弧曲线,采用Levenberg-Marquardt算法拟合出这3处圆弧曲线所在圆的圆心坐标作为3个特征点进行钢轨轨廓匹配;选其中2个特征点作为匹配参考点进行旋转加平移变换实现与对应标准特征点匹配,求出侧倾角消除车体侧倾和振动问题,该方法的匹配精度高达0.047 mm。不共线的三个圆心确定一个平面实现钢轨轨廓的高精度匹配,在UI界面,对检测钢轨轨廓拟合的图形与预设置的标准轨廓图形没有错位现象,同时该匹配方法已经在江西日月明轨道检测小车上得到验证。快速高精度的检测钢轨轨廓对提高列车运行平稳性,制定合理科学的钢轨轨廓修复策略有重要意义。
[Abstract]:The high precision measurement of rail profile is the basis of accurate measurement of rail profile and wear. The fast rail profiler is a typical method of non-contact detection of rail profile using a combination of two dimensional laser camera sensors mounted on multiple line structured light planes. Strictly symmetrical installation of laser sensors is the key to ensure the correct overlap of the sensor data and ensure the measuring accuracy of the fast rail profiler, which will increase the difficulties of manufacturing, assembling and debugging of the measuring system. In order to reduce the effect of symmetry on the measurement accuracy, calibration of the structured light plane of the sensor line is a feasible solution. On the basis of analyzing and summing up the existing calibration methods of line-structured light plane coplanar method, this paper presents a new line-structured light plane calibration method-contour method, which is based on the standard block of 00 grade flatness. The equal-distance data points are extracted, and the Levenberg-Marquardt algorithm is applied to the nonlinear iteration to make them at the same contour by angle transformation. The results of calibration test show that the repeatability of the contour calibration method can reach 0.00133 掳, the precision of the calibrated rail profiler can reach 7.7 m, and it has the characteristics of low equipment requirement and simple operation. Usually, the fast rail profiler is installed on the rail inspection vehicle, and the continuous, fast and accurate measurement of the rail profile and wear can be realized by the matching of the standard rail profile of the steel rail profile. In the course of rail profile detection, track irregularity will cause vibration, snake, roll, distortion and so on, so the data collected by laser camera sensor can not directly match rail profile. In order to solve the influence of car body vibration on rail profile matching, a rail profile matching method based on the inherent geometric characteristics of rail is proposed in this paper. The space position information of rail profile is obtained by using the calibrated line laser sensor, and the arc curve with the curvature radius of 13 mm for the outside rail head and 20 mm for the inner and outer rail waist is detected by Hough transform. The Levenberg-Marquardt algorithm is used to fit the circular center coordinates of the three arcs as three feature points for rail profile matching, and two of them are selected as reference points for rotation and translation transformation to match the corresponding standard feature points. The problem of car body roll and vibration is eliminated by calculating the angle of side inclination. The matching accuracy of this method is up to 0.047 mm.. The three centers of non-collinear circle determine a plane to achieve the high precision matching of rail profile. In the UI interface, there is no dislocation between the fitting figure of rail profile and the pre-set standard rail profile. At the same time, the matching method has been verified on the track detection vehicle of Jiangxi Sun, Moon and Ming. Rapid and high precision rail profile detection is of great significance to improve the train running stability and establish a reasonable and scientific repair strategy of rail profile.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U216.3;TP212

【参考文献】