Fast and Accurate 3D Eye-to-hand Calibration for Large-Scale Scene based on HALCON

Abstract

The premise of robot executing vision-based tasks is to establish the coordination between robot eye (structured light 3D sensor) and a hand (end effector). However, it is inefficient to perform the 3D eye-to-hand calibration for large-scale scene due to large amounts of point clouds data contained in the scene. This paper describes a fast and accurate 3D eye-to-hand calibration method for large-scale scene. A common tee pipe is used as calibration object. The method presented here comprises the following steps: Firstly, voxel grid filtering method is used for reducing the number of point clouds data. Secondly, employing euclidean cluster extraction method, the point clouds data of the calibration object can be extracted from the large-scale scene. Thirdly, the pose of the calibration object is acquired by surface-based 3D matching. Finally, the 3D eye-to-hand calibration is performed, and robot grasping experiments are carried out on ABB YuMi platform to validate the accuracy of calibration results. The experiments demonstrate that our method can achieve high calibration accuracy and significantly improve the efficiency of calibration.