双目视觉车轮姿态提取的位移检测方法

Displacement Detection Method Based on Wheel Position Obtaining by Binocular Vision

机车车轮三维场景重建需要进行多个特征点匹配,为了避免匹配中存在的运算量大及算法复杂等缺点,采用两个光轴正交的相机来采集车轮图像信息。由两相机所在坐标的对应关系恢复出车轮两个端面中心的空间坐标信息,由车轮轴线空间位置间接得到轮轨接触场景,利用当前时刻车轮中心轴线坐标与初始时刻比较,计算车轮相对轨道的横移量、垂移量、摇头角位移量。实验结果表明,该方法实现了列车轮轨位移的非接触检测,并避免复杂的车轮的三维"面"重建,检测效果准确,系统运行时间约500 ms,综合误差0.12,能够满足检测要求,该方法为后续进一步开展监测列车运行的状态以完善列车运行的安全性机理奠定了基础。

Locomotive wheel3D scene reconstruction requires several feature points matching,in order to avoid shortcomings of large amount of computation and the algorithm complexity,the wheel image information was collected by two cameras with orthogonal optical axis,so the wheel spatial coordinates of two end faces are obtained by two cameras coordinates relationship,meanwhile the wheel rail contact scene is indirectly obtained from the position of the wheel axis.By comparing the coordinates of the wheel center axis with the initial,it can calculate the lateral displacement,vertical displacement and yaw angle displacement.The experimental results show that this method achieves non-contact detection of train wheel rail displacement and avoids to reconstruct three-dimensional“surface”,and simplifies the calculation.Furthermore,having a012comprehensive error,the detection system runs about500ms and have a accurate detection result.This research established foundation for further monitoring the train operation state to improve the safety mechanism.