Industrial Computed Tomography from Few Views

Vienne, Caroline; CEA LIST; France

Vienne, C.; CEA, LIST; France
Costin, M.; CEA, LIST; France
Stolidi, A.; CEA, LIST; France

ID: ECNDT-0335-2018
Download: PDF
Session: CT-Methods 1
Room: G3
Date: 2018-06-12
Time: 11:30 - 11:50

The introduction of robotic systems, which are capable of taking each piece from the production conveyor and to manipulate them inside an X-ray beam or to directly move the X-ray equipment around the object, open new opportunities. Especially, the inspection trajectory is no more limited to one acquisition plane and when combined with an adequate algorithm, a satisfactory CT reconstruction can be obtained from few X-ray projections, which can greatly decrease the acquisition time and make it compatible with applications in production lines.

Such 3D trajectory makes the classical analytic reconstruction algorithms inappropriate and requires the use of 3D iterative reconstruction algorithms. In this work, we focus on two iterative algorithms, the well-known Simultaneous Algebraic Reconstruction Technique (SART) and the Discrete Algebraic Reconstruction Technique (DART), which integrates prior knowledge on the object, to perform 3D reconstructions from a limited data set. The performances of both algorithms on complex trajectories are evaluated in simulation environment using the CIVA software. A specific attention is brought to the definition of a quality measure of the reconstructed 3D volume in order to truly evaluate the impact of the sampling in terms of number of projections and 3D spatial distribution of these projections. Finally real acquisitions on non standard trajectories are performed with the robotic platform installed at CEA List and reconstruction results with SART and DART algorithms are compared.