Robot Hand-eye Calibration Method Incorporating Filtering Techniques

Hand-eye calibration is the foundation of visual servo for robotic tasks, aiming to determine the rigid transfor-mation matrix between the robot and the vision system. The accuracy of hand-eye calibration affects the robot's ability to accurately perform tasks. The classical form of the hand-eye calibration is to solve the equation “AX = XB”. The algorithms for solving this problem mainly divided into two categorie: step-wise and simultaneous solutions for the rotation and translation components. Due to the noise in the collected data, the calculation accuracy is inevitably affected. In this paper, we introduce a filtering approach for hand-eye calibration. First, we design the movement path of the robot arm end effector, then summarize the variation patterns of parameters, and finally use these patterns to filter the actually collected data. Experimental results on our custom dataset and the robot arm platform demonstrate that our proposed method can reduce reprojection error and improve the accuracy of hand-eye calibration.