Find a SECReT supervisor
prism apply now

Small vehicle inspection scanner imaging: SVXi

22 February 2012

Ireneos Drakos

In its simplest sense, Bruce Schneier has described security as “ensuring things that should happen do, and things that shouldn't happen, don't”. This project focuses on the principle of prevention; using dual energy, tomosynthesis and a combination of both techniques to show the potential of each in improving explosive detection in the Small Vehicle Inspection scanner (SVXi) developed by 3DX-RAY Ltd. Dual energy involves the acquisition of x-ray images at two different energies and permits the separation of attenuation into its basic components, to identify the material, and finally to produce material-specific image.

Tomosynthesis is a limited-angle tomography technique that can produce images that may contain information that is currently not visible with conventional imaging. Experiments have been designed using both simplistic materials (NaCl, C12H22O11) and phantom materials that can simulate the attenuation properties of explosives (NH4NO3) to test an algorithm that was developed to separate the mixture by giving the concentration of each substance. Using the same phantoms, ten images were collected over a 90° range and were combined using an already developed Shift-And-Add (SAA) algorithm.

Tomograms reconstructed were subsequently tested with dual energy analysis using a second developed algorithm. Field dual-energy with the SVXi was performed on 5 cars consisting of a bomb simulant placed on each driver‟s seat. In the laboratory, the curves produced using the algorithm can be normalised and used as calibration coefficients to predict the concentration of NH4NO3, NaCl and C12H22O11 in unknown mixtures containing these substances with 10% precision. The SAA has shown the potential to bring the absorbing objects within the phantom into focus.

Using the SVXi dual energy images, it was possible to estimate bomb composition; iron cylinder outside and vinyl chloride monomer packets inside, with 3% and 6% errors respectively. Finally, the combined method is shown to have great potential.