Abstract: A calibration method of parallel robots was proposed based on eye in hand vision. Firstly, based on the loop increment method, the kinematics errors and calibration model of a planar 2-DOF redundant actuated parallel robot were established. Secondly, a target plate was designed for the calibration experiments, and the target image was captured by the camera and processed by the segmentation, recognition and rotation compensation, and then the pixel error values of the end positions and the actual positions of the mechanisms were obtained. Thirdly, limited by the structures of the mechanisms, using the boundary curve to identify the feature corner points, a method was proposed to determine the rotation angles of the detection point based on the feature corner points. On the basis of compensating the rotation angles of the camera, using a simplified model of the camera pinhole, the error values of the pixel errors were converted to the true position error values of the ends of the mechanisms. Finally, the calibration model and the error values obtained through the visual system were used for kinematics calibration. After 4 iterations, the errors of the mechanisms are decreased to 1/3, which verifies the feasibility of the robots. The proposed method has the advantages of short time of calibrations, small amount of data and low cost of experiments.

Key words: planar mechanism, loop increment method, kinematics calibration, vision-based measurement

摘要： 提出了一种基于手眼视觉的并联机器人标定方法。基于环路增量法，建立了平面2-DOF冗余驱动并联机器人运动学误差与标定模型；设计了一种标定实验靶板，利用相机采集靶板图像并对其进行分割、识别、旋转补偿的处理，获取机构末端目标位置和实际位置的像素误差值；针对机构自身结构的限制，利用边界曲线识别特征角点，提出了一种基于特征角点确定检测点旋转角度的方法，在补偿相机旋转角度的基础上，再利用简化后的相机针孔模型，将像素误差值通过转换得到机构末端执行器的真实位置误差值；最后利用标定模型和通过视觉系统获取的误差值进行运动学标定。经过4次迭代，机构误差减小为原来的1/3，验证了该方法的可行性。同时该方法具有标定过程用时短、数据量小、实验成本低等优点。

关键词: 平面机构, 环路增量法, 运动学标定, 视觉测量