High-precision stereo calibration for 3D measurement via speckle-enhanced circular control point matching

High-precision stereo calibration is essential for achieving accurate 3D vision measurements. Marker-based calibration methods, such as the checkerboard and circular markers, are widely adopted for their robustness and high accuracy. However, the checkerboard is relatively sensitive to noise, while circular markers suffer from eccentricity errors. This introduces two significant issues: control point mismatch across different images and misalignment between the projections and detected results of control points. These challenges become especially pronounced in high-precision calibration tasks, where even minor discrepancies can substantially degrade calibration accuracy. Specifically, the primary difficulties are ensuring accurate matching of control points across multiple images despite eccentricity errors and performing precise calibration without exact correspondence between 3D control points and their detected image counterparts. To address these limitations in high-precision stereo calibration, we introduce a speckle-enhanced circular board by integrating circular markers with speckle patterns, aiming to improve the matching accuracy of the corresponding points through speckle pattern matching. Additionally, we introduce a bundle adjustment algorithm that compensates for eccentricity errors and manufacturing imperfections in the calibration target, which can significantly enhance the calibration accuracy without the need for high-precise manufactured calibration boards. Extensive simulations and real-world experiments demonstrate the proposed stereo calibration approach's high accuracy and effectiveness.