Simulated measurements of traditional shipping container screening infrastructure based on large-area polyvinyl-toluene (PVT) or sodium-iodide detectors (NaI) are used alongside an iterative reconstruction algorithm to characterize the activity and location of a radioactive point source concealed within a shipping container loaded with cargo. A maximum likelihood expectation maximization reconstruction method is employed to reconstruct the source distribution under the assumption that there exists a single point source in the scenario. To account for shielding by the cargo, it is assumed that the encompassing cargo, which was chosen to represent iron cargo, such as scrap metal or machine parts, is homogeneously distributed throughout the 32.2-m3 container at realistic loaded container densities of 0.0, 0.2, or 0.6 gcm−3. When the material properties of the cargo are assumed known and provided to the algorithm, the method is capable of localizing the source to within 40.5 cm and estimating the activity to the correct order of magnitude for cases with no cargo and 0.2 gcm−3 iron cargo completely filling the 32.2-m3 volume. With iron cargo at a density of 0.6 gcm−3, the localization and activity estimation is significantly worse, which is attributed to the method of accounting for attenuation in the cargo, a decreased signal-to-noise ratio, and the use of gross-count data that include the effect of buildup radiation. Using 662-keV photopeak data from a NaI-based radiation portal monitor (RPM) achieves better results than gross-count data from a PVT- or NaI-based RPM with the correct order of magnitude activity estimates for all cargo densities. For scenarios where the material of the cargo is unknown, but its density and distribution are known, a brute-force search is performed to find the optimum mass attenuation coefficient that describes the cargo. From the range of mass attenuation coefficients obtained, the method is not capable of differentiating between different types of common cargo, but demonstrates the principle of the method for characterizing shipping container cargo. Ultimately, the largest limiting factor in this method is the use of a simple average to estimate the path length traveled from a point in the container through the cargo in the direction of a detector. The large area detectors result in a high variance in this path length, and the degree of attenuation is exponentially dependent on this value. Despite the simple method of accounting for attenuation in the cargo, the maximum likelihood expectation maximization point source (MLEM PS) method is able to characterize a concealed point source well in the case with a PVT-based RPM and 0.2 gcm−3, which is cargo above the average density of a shipping container, drastically reducing the search area in secondary screening processes. The MLEM PS algorithm, therefore, represents a means of enhancing shipping container screening procedures without requiring significant changes in infrastructure and hardware.
Read full abstract