Image-based geometry calibration for a double-tomograph

Abstract Fast tomographic acquisitions are essential to study time-dependent and rapid phenomena in materials science, e.g. viscoelasticity or crack propagation. Acquisition time can be halved by employing two x-ray sources and detectors, provided they can be switched interchangeably. This work discusses the reconstruction of 3D volumes using data acquired simultaneously from two crossed-line laboratory tomographs. For this, geometrical parameters are proposed and their sensitivity is tested, allowing to discard less significant parameters. The selected parameters are then calibrated solely from the acquired projections of the object, by minimizing the difference between the reprojected volume and the experimental data. A gypsum sample is used as a model material, and three acquisition speed configurations—ranging from 20 s to 500 s—are tested to evaluate the effect of noise. The proposed procedure was shown to be robust even under noisier acquisition conditions. This methodology is believed to be an important step towards in-situ mechanical tests to be analyzed with projection-based techniques, with potentially two- to three-orders-of-magnitude increase in acquisition rates.