为了实时、准确、自动地实现钢轨焊缝错边量检测, 提出一种钢轨焊缝错边的视觉检测方法。首先, 采用中值滤波法去除焊缝图像中的噪声; 其次, 采用最大类间方差法和扫描法, 实现轨边的粗定位, 采用多次最小二乘直线拟合法实现轨边的精定位, 去除奇异点, 提高边缘定位精度; 再次, 采集标定图像, 利用图像处理算法, 实现相机的自动标定; 最后, 根据错边的边缘信息和相机的标定结果, 实现错边量的检测。通过对60kg/m钢轨焊缝区域、非焊缝区域和标准件的检测实验, 验证了该方法具有精度高且稳定性好的优点, 可用于焊轨基地的焊缝错边自动在线检测。

In order to measure the misalignment of the rail welding accurately and automatically in real time, a visual inspection method for the misalignment of the rail welding was proposed. Firstly, the median filtering method was used to denoise the high-resolution welding image. Secondly, the rough positioning of the track edge was conducted with the maximum interclass variance method and the scanning method, and the fine positioning of the track edge was completed by using the least squares fitting method, thus the singular points were removed effectively and the positioning accuracy was improved. Thirdly, the calibration board images were collected and the image processing algorithms were applied to automatically calibrate the camera. Finally, the misalignment measurement was realized according to the marginal information of the misalignment and the calibration results of the camera. Test experiments were performed on the standard components and the welding and non-welding area of the 60kg/m rail, and the proposed method was proved to realize a high accuracy and good stability, and can be applied to measure the misalignment of the rail welding automatically in real-time.