Accurate calibration approach for monocular multi-view vision system using rotational dual-target

Considering the complexity and potential inaccuracy caused by the improper placement of calibration devices in the existing dual-target calibration method for non-overlapping multi-camera system, a global calibration method is proposed based on geometric properties of rotational associated motion and absolute orientation between independent fields of view. Firstly, image sequences of the rotational dual-target are captured in two or more positions; and based on the geometric constraints during the target rotation process, combined with known camera intrinsic parameters, feature point images, and world coordinates, it solves for the rotation axis and reference points at different positions through global PnP optimization. Then, using the rotation axis, reference points, and length of the connecting rod, an absolute orientation matrix is constructed, and the rotation matrix and translation vector are calculated using singular value decomposition. Simulations and actual experiments show that compared with existing methods, the maximum pose and position errors of the method proposed in this paper are 0.009° and 0.138mm respectively, with the accuracy improved by 30.76% and 21.14% respectively. The calibration device in this method is simple, and there is no need for parallelism, verticality, and coplanarity between multiple rotation positions, and the calibration accuracy can meet the requirements of most application scenarios.