转向架轮对参数的双目视觉测量方法与实现

Wheel Set Parameters Measurement Method and Realization by Binocular Vision

为实现转向架轮径参数的精确检测,提出一种基于双目立体视觉的非接触式检测方法.采用Bouguet算法对轮对双目图像平面进行重投影,实现立体校正,并应用简化的NCC算法实现特征点的立体匹配;通过三维重建和坐标变换将提取到的轮对边缘特征点映射到二维平面内,并最终拟合出轮径参数;轮对轮缘半径检测试验,结果显示该方法的误差小于0.1mm,验证了该方法具有较好的检测精度,解决了当前测量方法特征点立体匹配误差较大和难以三维拟合的问题,满足实际现场快速测量要求.

For accurate measurement of wheel set radius parameters,a non-contact detection method based on binocular stereo vision is proposed.Bouguet algorithm is used to rectify the stereo of the wheel set binocular stereo image plane,simplified NCC algorithm is applied for purpose of stereo matching of the feature points.By means of three-dimensional reconstruction and coordinate transformation,the extracted edge feature points of the wheel set are mapped into a dimensional plane and finally fit the wheel radius parameters.The wheel rim radius detection experiments show that the error is less than 0.1 mm,which means the method solves the problem that the stereo matching has great errors and 3D fitting is difficult.Therefore,a visual system meets the requirements of rapid measurement in actual field.