Image-based kinematic calibration of a dual-robot x-ray computed tomography system via reprojection error optimization

This paper presents an image-based calibration method for a dual-robot X-ray computed tomography system with an additional rotary table that requires no external measuring instruments. Using radiographic projections of a spherical-marker object with known geometry, correction parameters for both robot kinematic chains and the turntable are jointly estimated by minimizing a multi-view reprojection error under smooth bounded limitation. The resulting corrected forward-kinematic model enables generation of a single, globally consistent projection geometry, eliminating the need for subsequent per-projection refinement. The method is quantitatively validated using modulation transfer function (MTF) measurements on reconstructed data and is compared to projection-wise geometry correction based on the Direct Linear Transformation (DLT). Results demonstrate the viability of the proposed approach for both classical Stop-and-Go and continuous FlyBy CT acquisitions.