Detecting the presence of hazardous materials in suitcases and carry-on luggage is an important problem in aviation security. As the set of threats is expanding, there is a corresponding need to increase the capabilities of explosive detection systems to address these threats. However, there is a lack of principled tools for predicting the performance of alternative designs for detection systems. In this paper, we describe an approach for computing bounds on the achievable classification performance of material discrimination systems based on empirical statistics that estimate the f-divergence of the underlying features. Our approach can be used to examine alternative physical observation modalities and measurement configurations, as well as variations in reconstruction and feature extraction algorithms.

Date Published: 1 May 2017
PDF: 9 pages
Proc. SPIE 10187, Anomaly Detection and Imaging with X-Rays (ADIX) II, 101870M (1 May 2017); doi: 10.1117/12.2276654

Published in SPIE Proceedings Vol. 10187:
Anomaly Detection and Imaging with X-Rays (ADIX) II
Amit Ashok; Edward D. Franco; Michael E. Gehm; Mark A. Neifeld, Editor(s)