Abstract:  In the processes of automatic welding of container girders， the uneven strong reflection on the workpiece surfaces， arc light， splash， and smoke easily led to inaccurate or difficult identification of weld deviations， so a welding seam tracking system for gas metal arc welding of container girders was proposed. Based on a low-cost structured light vision sensor， a highly robust welding seam feature extraction model was established. The linear slope of the laser stripes was used by the model for optimizing the random sampling processes of the random sampling consistency algorithm which achieved stable extraction of weld feature in complex environments. The welding test results on the surface of uneven and strong reflective workpieces show that the average detection errors of the method in the Y and Z directions are as 0.22 mm respectively and 0.19 mm with the welding parameters of 300 A， 29 V， and 1m/min， which achieves stable high precision welding seam tracking. 

Key words: container girder, automated welding, welding seam tracking, structured light vision sensor, image processing

摘要： 集装箱大梁自动焊接过程中，工件表面的反射不均匀、弧光、飞溅和烟尘等因素容易造成焊缝偏差识别不准确或难以稳定识别，因此设计了一种用于集装箱大梁熔化极气体保护电弧焊的焊缝跟踪系统。基于低成本结构光视觉传感器建立了高鲁棒的焊缝特征提取模型，该模型采用激光条纹的直线斜率对随机采样一致性算法的随机采样过程进行优化，以在复杂环境下实现模型对焊缝特征的稳定提取。不均匀强反射工件表面的焊接试验结果表明：焊接参数300 A、29 V、1 m/min下，Y向、Z向的平均检测误差分别为0.22 mm和0.19 mm，实现了稳定高精度的焊缝跟踪。

关键词: 集装箱大梁, 自动化焊接, 焊缝跟踪, 结构光视觉传感器, 图像处理