IMAGE-BASED MATERIAL DISCRIMINATION ALGORITHMS FOR ARMS CONTROL

Year
2012
Author(s)
Sean Robinson - Pacific Northwest National Laboratory
Ken Jarman - Pacific Northwest National Laboratory
Tim White - Pacific Northwest National Laboratory
Ben McDonald - Pacific Northwest National Laboratory
Andrew Gilbert - Pacific Northwest National Laboratory
Alex Misner - Pacific Northwest National Laboratory
Abstract
The Pacific Northwest National Laboratory is developing and evaluating active radiographic image analysis techniques for verifying sensitive objects in an arms control, material control, or warhead counting regime in which sensitive information may be processed. Material discrimination algorithms which attempt to estimate the amount of specific materials present in each pixel of an image can be used to verify pertinent non- sensitive or declared attributes (e.g., the presence of special nuclear material (SNM) within an object of interest) with all image analysis performed behind an information barrier stage, allowing for reporting and storage of non-sensitive attributes only. Techniques proposed here employ spectroscopic detectors to determine the materials present between the source and a single detector pixel, and operate by fitting the attenuated spectrum to a set of expected attenuation spectra for an ensemble of materials. Practical limits on single-pixel material discrimination are defined and suggestions for the optimal use of this technique within the context of arms control or cargo scanning is presented. A limited number of free parameters are expected for analysis on a single pixel, limiting the context of material discrimination to a few representative materials. SNM may be estimated by assuming that two materials lie along the ray from source to detector: one material is parameterized by the attenuation coefficient of plutonium, and the other material is chosen such that a best-fit metric to measured data is optimized. This discrimination method may be used to verify declared material configurations, as well as for mass estimates of SNM contained within an inspected object. Results may be compared with other mass estimates (e.g., Pu mass estimates provided by gamma spectroscopy or neutron multiplicity) behind an information barrier. As only the agreement between mass estimates would be returned, this technique may allow for verification of material presence without disclosure of sensitive information.