Investigating training, transfer and viewpoint effects resulting from recurrent CBT of X-Ray image interpretation

Abstract

X-ray screening of passenger bags is an essential task at airport security checkpoints. In this study we investigated how well airport security screeners can detect guns, knives, improvised explosive devices (IEDs) and other threat objects in X-ray images of passenger bags before and after 3 and 6 months of recurrent (about 20 min per week) computer-based training (CBT). Two experiments conducted at different airports gave very similar results. Training with X-ray Tutor (XRT), an individually adaptive CBT, resulted in large performance increases, especially for detecting IEDs. While performance for detecting IEDs was initially substantially lower than for guns, IEDs could be detected as well as guns after several months of training. A large transfer effect was observed as well: Training with XRT helped screeners recognize new threat objects that were similar in shape as the trained objects. Threat recognition was dependent on the rotation of the objects. If depicted from an unusual viewpoint, prohibited items were more difficult to recognize. The results were compared to two conventional (not adaptive) CBT systems. For one system no training and transfer effects were observed whereas small training and transfer effects were found for the other conventional CBT system.

Improving the Efficacy of Security Screening Tasks: A Review of Visual Search Challenges and Ways to Mitigate Their Adverse Effects

“…Second, an item's visibility can be affected by its physical orientation-for example, someone might intentionally place a prohibited item at a specific orientation as objects can be more difficult to detect when viewed from an unusual, noncanonical viewpoint (Bolfing et al, 2008;Koller, Hardmeier, Michel, & Schwaninger, 2008;Schwaninger, 2006;Schwaninger et al, 2004). In an X-ray image, a knife viewed from the side will likely be easier to identify than a knife viewed such that the base is aimed at the screener; one image readily appears as a knife, whereas the latter image could appear as a block of metal.…”

“…For example, searchers are trained to identify objects from a wide variety of noncanonical viewpoints (e.g., Koller et al, 2008), such as a knife rotated with the base aimed at the screener. Effective training regimens for identifying prohibited items and breaking camouflage remain a continuing area of research (e.g., Boot, Neider, & Kramer, 2009;Chen & Hegdé, 2012), and these efforts should help improve future performance.…”

Improving customs’ border control by creating a reference database of cargo inspection X-ray images

“…Earlier studies with airport security officers (e.g. [11,12]) and customs officers [1] have shown large increases in detection performance as a result of training. In the ACXIS project, a web-based CBT was developed that simulates the image analysis environment (Figure 11).…”

Evaluating the two-component inspection model in a simplified luggage search task

“…The model predicts that, as the probability of a hit increases with time allowed to inspect an item, the probability of a false alarm increases concomitantly-although not necessarily proportionally (Spitz & Drury, 1978)-with the relative differences, resulting in improvement of overall performance. Using formulas derived from these early studies, Ghylin et al (2006) successfully applied the inspection model to the X-Ray Tutor training program being used at hundreds of airports (Koller et al, 2007). Because Ghylin et al used professional screeners and complex targets, the general application to naive participants for simple visual search was evaluated in the present study.…”