针对机器人位姿测量立体标靶的单目视觉标定方法

Monocular vision calibration method of the stereo target for robot pose measurement

  • 摘要: 工业机器人末端位姿测量对机器人装配和机器人标定等工作具有重要价值。针对机器人位姿测量中常用标靶受环境光干扰较大、机器人运动空间有限等缺点设计了一款亮度可调、响应迅速的六面体立体标靶,并提出一种基于单目视觉的立体标靶标定方法。通过初始图像对的选择,解决单目视觉中本质矩阵分解得到的平移向量不精确的问题,并采用光束平差法对初始图像场景进行优化;向场景中添加新图像并使用光束平差法对场景进行全局优化,提高特征点重建精度;以精度为5m的平面标定板上的特征点作为真实点,解决单目视觉重建场景缺乏尺度因子的问题。实验表明:特征点三维重建的平均误差小于0.035 mm,能够有效进行立体标靶的标定;使用该标靶计算的机器人位姿信息,将机器的位置精度提高了37%。

     

    Abstract: The full pose measurement of industrial robot was valuable to robot assembly and robot calibration. In order to decrease the influence of ambient light and expand the target movement space, an active target was designed, and a method of the target calibration based on monocular vision was proposed. The initial pair was selected to enhance the accuracy of the translation vector which was decomposed from essential matrix. Then, images were added one by one and they all were used in the bundle adjustment to compute the high accuracy of 3D structure in the whole scene. Further, the points of the calibration board was adopted as the scale factor, and all the information was transformed into metric. The experimental results demonstrated that the 3D reconstruction precision is less than 0.035 mm, and the method could satisfy the requirement of stereo target calibration. Besides, the robot pose is effectively identified, and the position accuracy is improved by 37%.

     

/

返回文章
返回