Improved identification of illicit materials using an X-ray backscattering technique

  • Sarah Fisher

Student thesis: Phd


X-ray backscatter imaging has revolutionised screening procedures at security checkpoints across the world. It is used to acquire detailed x-ray images of the inside of vehicles and cargo in real time. Security personnel use these images to detect concealed illicit materials such as drugs, explosives and currency. However, one downside to the technique is the limited ability to identify materials. Materials can only be classified as either organic or inorganic. As a result many unnecessary manual searches of vehicles and cargo are performed due to items being falsely identified as contraband. This causes extensive delays and reduced throughput at ports. Improved identification of illicit materials is needed to solve this issue. This thesis describes the development of a technique to improve material identification in x-ray backscatter imaging. The problem with current backscatter systems is that no energy spectrum information can be measured by the x-ray detectors and energy spectrum information is critical to better material identification. The method presented involved taking several detector response measurements of a backscatter beam, each with a different amount of detector filtration. An approximate energy spectrum was calculated from the response measurements by a process known as "spectrum reconstruction" or "spectrum unfolding". The idea was tested in a proof-of-concept study using a 50 keV x-ray beam and five plastic test samples, including two explosive simulants. Monte Carlo simulations were used to derive relationships between properties of the reconstructed spectrum, effective atomic number (Z) and density (p). The Z properties were calculated for the experimentally tested materials to within 0.5Z and 0.12 g per cm^3 of the true value. In the explosives material range this was improved to 0.1Z and 0.04 g per cm^3 This was a substantial improvement on the organic/inorganic separation currently achievable on commercial systems. The simulation results were also used to classify materials as explosive or inert. The two explosive simulant materials were correctly identified with a higher than 95% probability. Further analysis of simulated data suggested an up to 100% true positive detection rate and 7% false positive detection rate was possible using the technique. A particular success of the method over other techniques proposed in the literature is that it does not rely on energy-resolving detectors, and therefore offers a cheaper and more practical solution for commercial systems. Work has started to test the technique using a 200 keV x-ray beam. Future work will focus on validation of the technique for cargo imaging.
Date of Award1 Aug 2022
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorDavid Cullen (Supervisor)


  • spectrum unfolding
  • Geant4
  • Monte Carlo simulation
  • discrete inverse problems
  • explosives detection
  • contraband detection
  • energy-integrating detector
  • backscatter imaging
  • x-ray imaging

Cite this