PNNL

Abstract

Active X-ray inspections have great potential for non-invasive detection of illicit materials. Dual-energy X-ray radiography, in particular, has been used to determine material composition by utilizing the energy-dependence of the X-ray attenuation coefficients. However, current implementations of this method are limited in their ability to determine material composition in layered objects. Here we present an inverse algorithm that can use multi-spectral X-ray radiography data to non-invasively reconstruct the composition of a container. Adaptive regularization is used to increase the accuracy of material estimations from multi-spectral data sets. The utility of these methods is demonstrated with experimentally acquired radiographs, obtained using a 100-450 keV tunable X-ray source. The object inspected is a scale-model of a nuclear materials storage container, composed of 3D-printed plastic and stainless steel spheres inside a thin-walled steel container. Reconstructions of the steel spheres are within a root-mean-square error of 0.37 cm.