针对单线结构光测量因范围有限以及遮挡等原因而无法检测整个车轮的轮缘及踏面的问题, 提出了一种列车轮对双目双线结构光检测方法。该方法采用两组双线结构光传感器进行测量和数据采集。在两传感器产生的点云进行配准的过程中使用截面点云数据还原点云矩阵, 然后对其进行傅里叶变换, 在计算旋转角度时先进行极坐标转换, 最后求取两矩阵的互功率谱, 从而得到点云的旋转和平移矩阵。在此过程中考虑到点云噪声的存在, 将辛格函数近似代替无噪声的互功率谱傅里叶反变换, 从而确定在频域配准时噪声对配准参数的确定并无影响。接着采用相对值比较的算法和主成分分析拟合基准平面, 对轮对直径方向进行标定。利用标定值对扫描数据进行处理, 得出包括车轮直径、车轮滚动圆径向跳动等对尺寸参数。实验表明: 点云频域配准的双目双线轮对检测技术对车轮直径的测量尺寸误差≤±0.05 mm, 车轮端跳的测量误差≤±0.06 mm, 车轮径跳的测量误差≤0.04 mm。与标准轮对检测数据相比可知, 该系统满足轮对检测精度的要求。

Detection method for train wheel set was proposed based on binocular and double-line structured light for solving the problem that rim and tread of the whole wheel can not be detected due to the limited measurement scope and shield of single-line structured light and other relevant reasons. Two groups of line structured sensor were employed in the measurement and data aquisition. Each group contained two laser displacement sensors. Point cloud matrix was restored with point cloud data on section in the process of registration of point cloud produced by the sensors, then was coverted by Fourier transform. In calculation of rotation angle, polar coordinates were conversed firstly and the cross-power spectrum of two matrices was solved finally to obtain rotation and translation matrices of the point cloud. During the process, Fourier inversion transform of noise-free cross-power spectrum was replaced by Sinc function approximation to determine that the noise in the registration of frequency domain had no influence on determination of registration parameters. Then datum plane was fitted with relative value comparison algorithm and Principal Component Analysis (PCA) to calibrate the diameter directions of wheel set. The canning data was processed with calibration values to acquire the dimension parameters of wheel set, such as diameter, radial runout of rolling circle. The result shows that the measurement error of the binocular and double-line detection system based on point cloud registration in frequency domain is less than ±0.05 mm in diameter, the axial runout is less than ±0.06 mm and radial runout of wheel is less than ±0.04 mm. Compared with standard wheel set detection technology, this system meets the requirements for precision of wheel set detection.