Butzhammer L (2026)
Publication Language: English
Publication Type: Journal article, Original article
Publication year: 2026
DOI: 10.1016/j.precisioneng.2026.03.015
Open Access Link: https://doi.org/10.1016/j.precisioneng.2026.03.015
Achieving high scan quality and dimensional accuracy in industrial Computed Tomography (CT) requires precise knowledge of the system geometry and its misalignments. The parameterization of static misalignment is done differently, depending on the geometry calibration method or reconstruction software employed. This work provides an implementation guide for the correct conversion of geometry parameters between two generic representations in cone-beam CT: (1) describing a misaligned detector in the reference frame defined by the rotary table, and (2) specifying a misaligned rotary table in the reference frame defined by the detector. The conversion rules enable a unified model for system geometry calibration and misalignment determination. Building on this unified representation, a calibration method based on the radiographic analysis of a calibrated multi-sphere reference object is proposed. The method uses the projection matrix formalism for a straightforward optimisation of the geometry parameters. To improve accuracy, sub-pixel edge detection of projected sphere contours, and a two-step optimisation process are employed. For the investigated setup, this approach yields mean reprojection errors of 0.008 pixels in simulations, and between 0.02 and 0.05 pixels in experiments. For simulations, resulting errors were in the single-digit micrometre range for the source-detector distance, sub-micron range for the rotary table position and sub-arcsecond range for the rotary axis tilt. The relative error in geometric magnification was 10-7 or lower. By comparing misalignment correction based on true versus measured parameters, it was verified that these errors are sufficiently small so as not to affect CT measurements of length and form. The method’s potential accuracy enables the detection of even small misalignments which have no significant influence on dimensional measurement results. For experiments, where a statement about (presumably higher) absolute errors is difficult due to the lack of a traceable reference and potential drifts during measurements, repeated measurements showed consistent and plausible outcomes. The proposed framework can be used to enable high-accuracy dimensional CT also for comparatively misaligned systems. It offers the possibility to be extended for future studies, such as to increase the accuracy via the usage of multiple magnifications or to extract dynamic error movements of the rotary table.
APA:
Butzhammer, L. (2026). Conversion between detector- and rotary-table-related misalignment parameterizations for unified projection-matrix-based geometry calibration in dimensional X-ray computed tomography. Precision Engineering-Journal of the International Societies For Precision Engineering and Nanotechnology. https://doi.org/10.1016/j.precisioneng.2026.03.015
MLA:
Butzhammer, Lorenz. "Conversion between detector- and rotary-table-related misalignment parameterizations for unified projection-matrix-based geometry calibration in dimensional X-ray computed tomography." Precision Engineering-Journal of the International Societies For Precision Engineering and Nanotechnology (2026).
BibTeX: Download