高级检索
于鸿宇, 刘智慧, 张承瑞, 陈赓, 高涛. 基于旁轴视觉的激光焊接系统加工定位方法[J]. 焊接学报, 2024, 45(9): 42-49, 102. DOI: 10.12073/j.hjxb.20230926002
引用本文: 于鸿宇, 刘智慧, 张承瑞, 陈赓, 高涛. 基于旁轴视觉的激光焊接系统加工定位方法[J]. 焊接学报, 2024, 45(9): 42-49, 102. DOI: 10.12073/j.hjxb.20230926002
YU Hongyu, LIU Zhihui, ZHANG Chengrui, CHEN Geng, GAO Tao. Positioning method for laser welding systems based on side-axis vision[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION, 2024, 45(9): 42-49, 102. DOI: 10.12073/j.hjxb.20230926002
Citation: YU Hongyu, LIU Zhihui, ZHANG Chengrui, CHEN Geng, GAO Tao. Positioning method for laser welding systems based on side-axis vision[J]. TRANSACTIONS OF THE CHINA WELDING INSTITUTION, 2024, 45(9): 42-49, 102. DOI: 10.12073/j.hjxb.20230926002

基于旁轴视觉的激光焊接系统加工定位方法

Positioning method for laser welding systems based on side-axis vision

  • 摘要: 针对激光焊相机标定精度低,定位精度与像素当量倍数相差较大等问题,提出了一种基于旁轴视觉的相机标定方法和定位方法,通过内参标定校正透镜畸变;通过外参标定,建立系统中像素坐标系与机床坐标系间的转换关系,设计一种二值化阈值的局部寻优算法,以保证工件准确的边缘拟合. 通过识别工件角点和一边的角度,确定工件在系统中的摆放位置与姿态,算法精度稳定在1像素内,即0.017 mm的理论精度,通过加工试验测试定位精度,在试验中对识别的工件转角角度进行修正,修正前,x方向平均误差0.050 mm,y方向平均误差0.137 mm;修正后,x方向平均误差为0.029 mm,y方向平均误差为0.026 mm,分别为像素当量的2.12倍和1.53倍,实际误差与像素当量间倍数相差较小. 结果表明,该定位方法在充分利用相机性能的同时,具有较高的算法精度和定位精度,提高加工效率满足实际生产需求.

     

    Abstract: In response to the suboptimal calibration results in laser welding, as well as significant discrepancies between the positioning accuracy and pixel equivalence, a camera calibration method and positioning approach based on side-axis vision were proposed on the basis of a self-developed laser welding system in the laboratory. The method involved the calibration of intrinsic parameters to correct lens distortion and the calibration of extrinsic parameters to establish the transformation relationship between the pixel coordinate system and the machine tool coordinate system. Additionally, a locally optimized algorithm for binary thresholding was designed to ensure accurate edge fitting of workpieces. The placement position and orientation of the workpiece in the system were determined through the identification of workpiece corner and the angle of one side, with the algorithm achieving a stable accuracy within 1 pixel, equivalent to a theoretical precision of 0.017 mm. Experimental testing of positioning accuracy involved the correction of recognized workpiece rotation angles. Prior to correction, the average errors in the x and y directions were 0.050 mm and 0.137 mm, respectively. After correction, the average errors were reduced to 0.029 mm in the x direction and 0.026 mm in the y direction, corresponding to multiples of pixel equivalence of 2.12 and 1.53, respectively. The actual errors exhibited minimal differences in multiples with pixel equivalence. The results indicated that this positioning method, while fully leveraging the camera's performance, demonstrated high algorithmic and positioning accuracy, enhancing processing efficiency and meeting practical production requirements.

     

/

返回文章
返回