An SVD-based turntable calibration method with multiple reference planes segmented from point clouds

Abstract

Turntables play an essential role in various industrial areas due to their ability to improve the DOF (degree of freedom) of workpieces or tools. In applications where positional accuracy of the objects is required, the accuracy of the system parameters of the turntable is critical for the applications to work properly. This paper proposes a new approach to calibrating the turntable system. First, we propose an optimized RANSAC (Random Sample Consensus) algorithm to obtain high-accuracy reference planes from multiple frames of point clouds grabbed from the reference object placed on a turntable. Second, an SVD (Singular Value Decomposition)-based two-step method is proposed to calculate the rotation axis vector and the center position of the turntable based on the reference planes extracted by the modified RANSAC algorithm. Finally, several experiments are performed to verify the proposed methods. Compared with the existing methods, our approach gets the analytical results and tries to correct the rotation angles. The experimental results show that the proposed method considerably improves calibration accuracy and efficiency without complex devices.