Project Summary
HALO’s unique x-ray diffraction (XRD) technology has been shown to provide enhanced material _x000D_
discrimination capability when coupled with conventional x-ray Computed Tomography (CT)_x000D_
screening devices and used in the aviation security and customs and border protection industries. _x000D_
Regulatory standards of detection, whatever they may be, cannot encompass all possible threat _x000D_
materials, particularly homemade explosives (HMEs) where the manufacturing method is not well _x000D_
defined or controlled. The Resilient Detection project (or EDEN, for brevity), sought to build resilience _x000D_
against unseen threats into the threat detection process, by directly characterising stream-of_x0002_commerce and all outliers from this being defined as threats. The results presented in this report _x000D_
demonstrate the EDEN approach is a viable and effective method of threat detection, with enhanced _x000D_
material specificity powered by x-ray diffraction.
Project Achievements
HALO has demonstrated effective threat retention (>85%) for a range of home-made explosives _x000D_
(HMEs) for which HALO has no reference data, for an effective airport false alarm retention rate of 30-_x000D_
40%. These multipliers, coupled with initial CT false alarm rates (e.g. 30-40%), suggest an _x000D_
operationally viable, resilient detection algorithm capable of providing required threat detection in the _x000D_
face of new and emerging explosive materials and threat vectors. A CT identifies Potential Threat _x000D_
Objects (PTOs) and the fused CT/XRD EDEN algorithm running on the HXT264 resolves those _x000D_
alarms through knowledge of stream-of-commerce alone. Airports might expect to see real false _x000D_
alarm rates of 9-16%, or lower, with added resilience to threat detection.
Conclusions
The EDEN approach to threat detection (one-class outlier detection based on a comprehensive _x000D_
material description of expected stream-of-commerce) has been demonstrated to Technology _x000D_
Readiness Level (TRL) 6 under this project. _x000D_
Support Vector Data Description (SVDD) provides an effective method for outlier detection, taking _x000D_
advantage of the limited threat data available, while fully enclosing region(s) of responsibility for _x000D_
airport stream-of-commerce._x000D_
Deployment of the EDEN approach with any CT system will be increasingly effective when used in _x000D_
conjunction with a third-party object segmentation and feature generation algorithm and HALO has _x000D_
demonstrated this approach to realise a truly resilient, CT independent threat detection algorithm.
Next Steps
The EDEN algorithm, coupled with a universal CT segmentation algorithm, provides the tools _x000D_
necessary to deploy CT-agnostic, resilient detection with built-in protection against emergent threats. _x000D_
The next steps for this approach are advancement of the TRL and proof of viability across the _x000D_
landscape of CT screening systems, in a wider range of screening environments, e.g. narcotics _x000D_
detection, and progression of the prototype into a commercial product

