A global calibration method for extrinsic parameters of 3D vision system with a non-overlapping field of view based on binocular vision structure from motion

Abstract The global calibration of non-overlapping fields of view in 3D vision systems is crucial for collaborative measurements across multiple vision systems. A global calibration method for binocular vision motion structures based on localized clustering of feature similarities of coded targets (GC-BSFM) is proposed to address the challenge of global calibration of external parameters under complex noise conditions. Additionally, a multi-dimensional space-constrained pose optimization algorithm (MDSC-PNP) is proposed for high-precision estimation of the motion poses of stereo vision modules. Under complex noise conditions, the GC-BSFM method achieves a mean error of only 0.197 mm. It demonstrates strong robustness, successfully performing global calibration of non-overlapping fields of view even under challenging conditions such as partial occlusion of coded targets and strong background noise. This method holds significant importance for industrial measurement and related applications.