Automated Recursive Hand-Eye Calibration Employing 3D Point Cloud Registration

In industrial automated applications employing robot vision, hand-eye calibration is a fundamental and critical issue. With the improvement in equipment efficiency and the maturity of 3D vision technology, 3D point cloud processing is gaining increasing importance. Compared with traditional 2D images, 3D point clouds provide richer data. This paper focuses on automated recursive hand-eye calibration using 3D point cloud registration. A 3D point cloud registration approach is employed to determine the spatial relation between a 3D camera and a 3D calibration pattern. Recursive least square method is employed in conjunction with disqualification of poor quality datasets to refine the calibration results and filter out datasets causing increase of the hand-eye calibration errors, thereby minimizing inaccuracies. The proposed automated recursive hand-eye calibration system is applicable to both eye-in-hand and eye-to-hand configurations. To further facilitate sequential manipulator motion for automated hand-eye calibration, effective pose-planning approaches can be employed to enhance the efficiency of the automated calibration.