In order to improve the running accuracy of the packaging mechanical arm and solve the hand-eye calibration problem in the visual servo control system, a method for calibrating kinematics parameters of the manipulator was designed based on genetic algorithm. Based on the definition of hand-eye visual coordinates, the transformation relations between different coordinate systems were given. By comparing the theoretical position with the actual position of the manipulator end-effector, the calibration error model of kinematics parameters was determined. Based on the basic principle of genetic algorithm, the fitness function was constructed and the pose of end-effector was updated according to the error compensation. Finally, the joint debugging and experimental analysis of the manipulator motion control system were carried out. The mean absolute value of the absolute deviation between the actual position and the theoretical position is about 0.8 mm, and the maximum deviation is only 1.2 mm, so the accuracy is relatively high. The hand-eye calibration method can significantly improve the motion accuracy of the mechanical arm and meet the requirements of the relevant packaging industry.