Strategies for Efficient Scanning and Reconstruction Methods on Very Large Objects with High Energy X-ray Computed Tomography

摘要

X-ray computed tomography (CT) is an established tool for industrial non-destructive testing purposes. Yet conventional CT devices pose limitations regarding specimen dimensions and material thicknesses. Here we introduce a novel CT system capable of inspecting very large objects (VLO) like automobiles or sea freight containers in 3-D and discuss strategies for efficient scanning and reconstruction methods. The system utilizes a 9 MeV linear accelerator to achieve high penetration lengths in both dense and high-Z materials. The line detector array has an overall length of 4 meters. The presented system allows for reconstruction volumes of 3.2 meters in diameter and 5 meters in height. First we outline the general capabilities of high energy CT imaging and compare it with state of the art 450 kV X-ray systems. The imaging performance is shown based on experimental results. The second part addresses the problem of considerably higher scanning times when using line detectors compared to area detectors. Reducing the number of projections considerably causes image artifacts with standard reconstruction methods like filtered back projection (FBP). Alternative methods which can provide significantly better results are algebraic reconstruction techniques (ART). One of these is compressed sensing (CS) based ART which we discuss regarding its suitability in respect to FBP. We could prove the feasibility of inspecting VLOs like complete automobiles based on experimental data. CS allows for achieving sufficient image quality in terms of spatial and contrast resolution while reducing the number of projections significantly resulting in faster scanning times.