Robotic hand-eye calibration with depth camera: A sphere model approach

This paper presents a new depth-based hand-eye calibration method to find the transformation between depth camera and robot wrist. Owing to the true hand-eye transformation is hard to obtained, indirect measurements depending on robot movements and calibration tools is adopted. In our method, A sphere, which has simple and elegant parameterized presentation, is used as calibration tool. The points cloud acquired from the depth camera viewing the calibration tool are aligned to a virtual sphere model by RANSAC to find the translation part of the homogeneous transformation between depth camera and calibration tool. The rotational part of the transformation is ambiguous thus eliminated by reformulating the hand-eye calibration problem. Then a non-linear optimization method is proposed to solve the hand-eye rotation and translation simultaneously. Due to the ground truth hand-eye transformation is not available, a quantitative error measurement is adopted to measure the accuracy.