A novel binocular vision-robot hand-eye calibration method using dual nonlinear optimization and sample screening

Abstract

To reduce the effects of random errors in vision robotic systems, a double-layer LM (Levenberg- Marquardt) optimization (DLMO) method was designed to improve the calibration accuracy of the robot hand-eye matrix based on the principle of nonlinear optimization. Firstly, a hand-eye matrix equation was established, and the initial value of the hand-eye matrix was solved by using a linear least square method. Secondly, Euler angle transformation was performed on the rotation matrix part in the hand-eye matrix, to ensure the orthogonality of the rotation matrix. Then, an optimization model of the hand-eye matrix was constructed, and the initial value of the hand-eye matrix was optimized for the first time by using the traditional LM nonlinear optimization method. Finally, the optimization model of the hand-eye matrix was modified, and then the LM nonlinear optimization method with the function of sample point screening was used to optimize the hand-eye matrix for the second time. Using the proposed DLMO, the hand-eye calibration tests were carried out on two industrial robots equipped with binocular vision systems. The average position errors of the calibration results were 0.33 mm and 0.51 mm, respectively. The results showed that the accuracy of the proposed calibration method met the working requirements of the vision-robot in the industrial field.