Application of signal detection theory to the recognition of objects in colour-encoded X-ray images

Aviation security personnel encounter problems when interpreting x-ray images of hand luggage. This research seeks to determine whether the performance of the human operator can be improved, in terms of both reliability and accuracy, through the employment of a novel multiple-view x-ray imaging tech...

Full description

Bibliographic Details
Main Author: Vassiliades, V.
Published: Nottingham Trent University 2010
Subjects:
Online Access:http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.629189
Description
Summary:Aviation security personnel encounter problems when interpreting x-ray images of hand luggage. This research seeks to determine whether the performance of the human operator can be improved, in terms of both reliability and accuracy, through the employment of a novel multiple-view x-ray imaging technique. Thus, a series of experiments were undertaken with the aim of providing evidence for the feasibility of using KDEX displays to aid in the recognition of threatening objects in airport carry-on luggage; and furthermore demonstrate the real-world value of this technique. This thesis describes experiments comparing how introducing depth information affects the performance of aviation security personnel attempting to detect various weapons in x-ray images of hand luggage. Specifically, multiple 2-dimensional (2D) x-ray luggage scans were acquired and processed to create the perception of 3-dimensionality (3D) in kinetic displays. These results were compared with weapon detection in standard static 2D scans of the same luggage. Threatening objects hidden in this luggage were more readily detected in kinetic 3-dimensional images than in the standard images. Initial results were obtained using greyscale images and limited to various types of knives. Subsequent experiments evolved to primarily use pseudo-colour x-ray images. These were encoded using the industry standard colours; orange for organic materials, green for mixture, and blue for metals. Additionally, as well as knives, the threat categories were expanded to include handguns, scissors and grenades. In view of the remarkable improvement in performance afforded by the kinetic images future studies are warranted. For example, various features of the displays (e.g. differences among hues and contrasts) should be manipulated as parameters to determine if further improvements may be achieved. Additionally, different pseudo-colour schemes could be experimented with in order to potentially maximize the gains afforded by the kinetic display. Overall, when deployed iin the field this technique may well revolutionize x-ray visual security screening.