Research Of The Tripod Parallel Robot's Kinematic Calibration Based On Monocular Vision

Aiming at the problem that the position accuracy of the Tripod parallel robot's end effector,in order to achieve the improvement of the position accuracy of the Tripod parallel robot economically and quickly,a kinematics calibration method of the Tripod parallel robot based on monocular vision is proposed.The research involves camera calibration technology,kinematics calibration technology,visual positioning technology and image processing technology.The kinematics calibration process of robots is mainly divided into four steps: error modeling,error measurement,parameter identification and error compensation.Each part is researched respectively.First of all,in order to ensure the accuracy of error measurement based on the visual system,the calibration process of the camera's main parameters is designed.The main parameters to be calibrated are obtained,the camera parameter calibration process for obtaining the initial solution of the camera parameter model and then optimizing the nonlinear parameters of the camera model is determined.Secondly,for the error modeling of kinematic calibration,the robot kinematics forward and inverse solution model is established based on structure analysis of the Tripod parallel robot,and the robot's initial error mapping model is established based on analyze the geometric error source that affects the position accuracy of the robot,the error distribution under each geometric error source is analyzed by simulation,and the sensitivity and redundancy of each geometric error source are analyzed respectively,finally optimized and identifiable error mapping model is obtained by deleting the geometric error sources that have little effect on the position accuracy and have linear correlation with other geometric error sources.Then,in terms of error measurement,an error measurement method based on the P4 P position measurement algorithm is determined,an overall vision measurement scheme is designed,and for image processing of feature points of the planar calibration target,a weighted least-squares line fit algorithm based on the RANSAC algorithm is proposed,which constructs a more accurate initial solution of the fitted line by the RANSAC algorithm and then obtains the optimal solution by the weighted least-squares method.The simulation results show that the optimization algorithm is more robust and accurate than RANSAC algorithm and weighted least-squares line fit algorithm.Next,the parameter identification and error compensation are studied.The parameter identification model and error compensation model are established,and each model is simulated respectively.The simulation results verify the accuracy of models.Finally,the Tripod parallel robot platform based on vision system is built.On the one hand,combining with the established model and the proposed methods,the accuracy of visual measurement is detected,and the kinematic calibration experiment is carried out with visual measurement by obtaining the position error of the end effector.On the other hand,experiments were carried out on the repeated positioning accuracy and the positioning accuracy of the robot before and after calibration by a laser tracker.The comparison of the positioning accuracy experiments before and after calibration shows that the average comprehensive positioning accuracy of the Tripod parallel robot's end position has been significantly improved,confirming the effectiveness of the Tripod parallel robot's kinematic calibration research based on the vision system.