Rotation axis calibration is crucial for high-precision automatic point
cloud stitching in turntable-based 3D scanning systems. To achieve a
360° sampling with a 2D calibrator in rotation axis calibration, this
paper proposes a dual-turntable angle cancellation (DTAC) method. DTAC
introduces an auxiliary turntable to keep a proper relative angle
between the 3D sensor and the calibrator during the calibration
process. The auxiliary turntable rotates at the same and opposite
angle as the main turntable and cancels the increment of the relative
angle. By projecting the feature points on the planar calibrator from
real-world space to virtual calibration space, the projected points
all share the same rotation axis of the main turntable. Further, a
layered circle center extraction (LCCE) algorithm is applied to deal
with outlier data points. The algorithm uses a two-step robust
estimation strategy combining RANSAC circle fitting with a median
noise filter for circle center selection. The standard ball
reconstruction experiment shows that the 3D system calibrated by the
method achieves a mean absolute error of 0.022 mm and root mean
square error of 0.025 mm within the measurement distance of
60–70 cm. Point cloud stitching experiments of different types
of objects show that our method outperforms other state-of-the-art
methods in stitching accuracy. The DTAC method and LCCE algorithm can
improve turntable-based 3D scanning systems.