Abstract: To eliminate the error of hand-eye calibration caused by installation error associated with the verticality of the camera-s optic axis in the automatic stiffness bonder, we proposed a novel method to calculate the homogeneous coordinate transformation between the camera frame and the robot base frame. The external parameters of camera obtained with the calibration boards, were combined with the conventional 2D calibration model, thereby extending the dimensions of the model to 3 and effectively improving the calculation accuracy. The calculation error of the calibration with the 3D model is less than 0.1m, which decreases by an order of magnitude compared with the 2D model. Using the thin films to modify the value in Z direction in coordinate transformation, the system has high location precision and the error of the calibration is less than 0.015 mm. The mean error is reduced effectively, and the performance of system is significantly improved. The proposed method is capable to obtain the unique solution of calibration equation with the finite degree of freedom system. The 3D calibration model including the whole information associated with the camera pose, has no additional constraints such as camera installation, and can effectively improve the accuracy of hand-eye calibration.