A Hand-Eye Calibration Method based on Nonlinear Joint Optimization for RGBD Cameras

Abstract

Hand-eye calibration is a basic research topic in the field of robotic vision. At present, most works focus on the hand-eye calibration of RGB cameras, while methods for calibration of RGBD cameras are relatively rare. Two popular kinds of modes for RGBD cameras are based on the stereo vision and the structured light, which both contain at least two cameras for generating the color image and the depth image, respectively. Motivated by this special structure feature, the paper proposes a novel hand-eye calibration method for RGBD cameras by integrating the vision system and the hand-eye robotic system into a whole model. Specifically, the traditional hand-eye calibration method is firstly used to obtain initial values of hand-eye matrices of the two cameras in the RGBD camera separately. Then, a nonlinear joint optimization algorithm is carried out to refine hand-eye matrices, in which the objective is to minimize the joint reprojection error of feature comers in images outputting by the two cameras in the RGBD camera. Thus, the optimized hand-eye calibration matrices can be finally obtained for both cameras. The proposed calibration approach is validated on the robotic hand-eye system, and the experimental results demonstrate that the presented algorithm greatly improves the calibration accuracy compared with the traditional methods.