针对机器人位姿测量立体标靶的单目视觉标定方法

发布时间：2018-12-08 18:46

【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5μm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 measurement of terminal position and pose of industrial robot is of great value to robot assembly and robot calibration. A hexahedron stereoscopic target with adjustable brightness and quick response is designed to overcome the disadvantages of the robot's position and attitude measurement, such as the large interference of environmental light and the limited motion space of the robot. A stereo target calibration method based on monocular vision is proposed. Through the selection of initial image pairs, the problem of imprecision of translation vector obtained from the decomposition of essential matrix in monocular vision is solved, and the initial image scene is optimized by beam adjustment method. The new image is added to the scene and the global optimization of the scene is carried out by using the beam adjustment method to improve the accuracy of feature point reconstruction. The feature points on the plane calibration board with an accuracy of 5 渭 m are taken as the real points to solve the problem of the lack of scale factor in the monocular vision reconstruction scene. The experimental results show that the average error of 3D reconstruction of feature points is less than 0.035 mm, and the robot position and pose information calculated by the target can be effectively calibrated, and the position accuracy of the machine can be improved by 37%.
【作者单位】：上海大学机电工程与自动化学院;
【基金】：国家自然科学基金(51575332) 机械系统与振动国家重点实验室课题(MSV2015010) 上海市教育委员会科研创新项目(2014Z10280034)
【分类号】：TP242;TP391.41

【相似文献】