Assessment of imaging models for volumetric tomography of fluid flows

Abstract

Volumetric tomography is a powerful tool that employs 2D projections to reconstruct unknown physical quantities in 3D fluid flows. A mathematical model of the imaging process is required to recover the desired volumetric fields. Errors in the imaging model can corrupt the reconstructions, so an accurate model is critical. This work reports the first systematic assessment of nine imaging models in terms of accuracy, computational cost, and range of applicability. A sample matrix method is developed to assess and improve the model’s accuracy. A flame chemiluminescence tomography experiment and a synthetic tomographic particle image velocimetry test were conducted. For both luminosity field reconstruction and velocity estimation, the Voxel Spread Function (VSF) model is the most accurate, but the computational cost is tens or hundreds of times higher than other models. The worst models are the Ray-length and VC Direct models. Their error increases when the spatial frequency or seeding density of the object field increases.