Human factors in X-ray image inspection of passenger Baggage – Basic and applied perspectives

The X-ray image inspection of passenger baggage contributes substantially to aviation security and is best understood as a search and decision task: Trained security officers – so called screeners – search the images for threats among many harmless everyday objects, but the recognition of objects in X-ray images and therefore the decision between threats and harmless objects can be difficult. Because performance in this task depends on often difficult recognition, it is not clear to what extent basic research on visual search can be generalized to X-ray image inspection. Manuscript 1 of this thesis investigated whether X-ray image inspection and a traditional visual search task depend on the same visual-cognitive abilities. The results indicate that traditional visual search tasks and X-ray image inspection depend on different aspects of common visual-cognitive abilities. Another gap between basic research on visual search and applied research on X-ray image inspection is that the former is typically conducted with students and the latter with professional screeners. Therefore, these two populations were compared, revealing that professionals performed better in X-ray image inspection, but not the visual search task. However, there was no difference between students and professionals regarding the importance of the visual-cognitive abilities for either task.
Because there is some freedom in the decision whether a suspicious object should be declared as a threat or as harmless, the results of X-ray image inspection in terms of hit and false alarm rate depend on the screeners’ response tendency. Manuscript 2 evaluated whether three commonly used detection measures – ${d}'$, ${A}'$, and ${d}_{a}$ – are a valid representation of detection performance that is independent from response tendency. The results were consistently in favor of da with a slope parameter of around 0.6. In Manuscript 3 it was further shown that screeners can change their response tendency to increase the detection of novel threats. Also, screeners with a high ability to recognize everyday objects detected more novel threats when their response tendency was manipulated.
The thesis further addressed changes that screeners face due to technological developments. Manuscript 4 showed that screeners can inspect X-ray images for one hour straight without a decrease in performance under conditions of remote cabin baggage screening, which means that X-ray image inspection is performed in a quiet room remote from the checkpoint. These screeners did not show a lower performance, but reported more distress, compared to screeners who took a 10 min break after every 20 min of screening.
Manuscript 5 evaluated detection systems for cabin baggage screening (EDSCB). EDSCB only increased the detection of improvised explosive devices (IEDs) for inexperienced screeners if alarms by the EDSCB were indicated on the image and the screeners had to decide whether a threat was present or not. The detection of mere explosives, which lack the triggering device of IEDs, was only increased if the screeners could not decide against an alarm by the EDSCB. Manuscript 6 used discrete event simulation to evaluate how EDSCB impacts the throughput of passenger baggage screening. Throughput decreased with increasing false alarm rate of the EDSCB. However, fast alarm resolution processes and screeners with a low false alarm rate increased throughput.
Taken together, the present findings contribute to understanding X-ray image inspection as a task with a search and decision component. The findings provide insights into basic aspects like the required visual-cognitive abilities and valid measures of detection performance, but also into applied research questions like for how long X-ray image inspection can be performed and how automation can assist with the detection of explosives.