Vision-based method of kinematic calibration and image tracking of position and posture for 3-RPS parallel robot

Parallel robot needs to implement the position and orientation calibration before being put into application, due to errors from manufacture and assembly. Traditional calibration methods with the help of non-contact high-precision measuring devices, such as sensors and three coordinate measuring instrument, possess many drawbacks including difficult installation, system complexity, high cost and low automation. To overcome these issues, this paper presents a non-contact measuring method to realize the kinematic calibration and image tracking of position and posture based on computer vision for the developed 3-RPS parallel robot in laboratory. A method of pre-distinguishing feature points region is proposed to combine with PNP algorithm for realizing three-degree-of-freedom (three-DOF) measurement of position and posture under monocular vision. To implement the robotic calibration, the inverse kinematic model and parameter identification equation are established. Then, after modeling of the forward kinematics, a synthesis of the traditional Pyramid-LK algorithm and feature points region prediction method is performed to realize the image tracking, which improves the algorithm accuracy and shortens the response time. The performance of the proposed calibration and image tracking of position and posture for the parallel mechanisms is verified by several experiments.