Abstract: In order to avoid the influence of robot model error on the hand-eye calibration of a three-dimensional shape flexible measurement system, the hand-eye calibration method was investigated. A calibration method based on the feature points fitting was proposed. The 3D shape measurement robotic system was established with a 3D shape scanner mounted on the robot end. In the calibration, the robot end coordinate system was measured with a laser tracker to obtain their transformation relationship. Then, 3D shape scanner and laser tracker were used for measuring feature points which fixed in the measurement field, and the transformation relationship between them could be identified by the constraint of feature points and the algorithm of Rodrigues matrix. Consequently, the hand-eye relationship was established directly. Experiment of hand-eye calibration was based on ATOS 3D optical scanner, Yaskawa-Hp20D robot and API laser tracer, and accuracy verification experiments were carried out. The verification experimental results show that the repeatability accuracy (3σ) of 3D shape robotic measurement system is better than 0.1 mm, the root-mean-square (RMS) error of distance is within 0.2 mm, and the accuracy of point cloud stitching is better than 0.7 mm after calibration. Meanwhile, this method can effectively avoid robotic model error which can be introduced in the process of traditional hand-eye calibration, the solution procedure adopts the linear method, and this method adapts to the 3D shape measurement robotic system.