Centralised Image Processing: challenges, trends and time on task

Centralized Image Processing: challenges, trends and time on task

Remote screening has been in use for hold baggage since the beginning of this century and has now been implemented more and more for the screening of cabin baggage via centralized image processing (CIP). It has the potential to increase the efficiency of baggage screening and is therefore a valuable tool with which to organize the checkpoint more flexibly. In the light of the global COVID-19 pandemic of 2020, remote screening has also been discussed as a solution for minimizing contact during security screening and, accordingly, it is more relevant than ever. Daniela Buser and Sarah Merks give an overview of some challenges and trends of CIP and discuss initial results of laboratory and field studies on time-on-task for remote screening.

Centralised image processing (CIP) refers to the networking of baggage images generated by X-ray or 3D CT machines. It enables a more flexible screening operation by relaxing the conventional 1:1 ratio between X-ray machines and X-ray inspection officers. An additional benefit of CIP is that security screeners do not necessarily need to sit next to an X-ray machine to evaluate images, which allows for more flexibility when organising the task of image analysis at security checkpoints.

“…it enables a more flexible screening operation by relaxing the conventional 1:1 ratio between X-ray machines and X-ray inspection officers…”

In the last decade, remote cabin baggage screening (RCBS) via the use of centralised image processing (CIP) has been implemented at several European airports. It has been shown to potentially increase detection performance, throughput and employee satisfaction. Hence, implementing CIP at security checkpoints holds many potential advantages; however, there is no ‘one-size-fits-all’ approach. In the October/November 2017 issue of ASI, Milena Kuhn introduced CIP, discussed the potential benefits and pitfalls, and compared different CIP implementation options regarding effectiveness, efficiency and human factors. Within the same research project, the Center for Adaptive Security Research and Applications (CASRA) conducted a survey with experts from two European airports, one authority representative and one member of the IATA Smart Security programme. The survey investigated challenges and trends of CIP for checkpoint security. These results will be summarised in the first part of this article. In the second part, the most relevant insights from a laboratory and a field study on time-on-task using remote screening will be presented.

What are some challenges of CIP?

In the coming years, the focus of most airports (at least in Europe) will probably be on implementing explosives detection systems for cabin baggage (EDSCB) standards and on purchasing the associated new machines. If the implementation of the new machines does not include fully integrated lanes including CIP software, there will be hardly any resources left for expanding CIP. Although CIP brings the possibility of major operational benefits, airports will likely be wary of making the additional investment after the financial losses the aviation sector has had to suffer this year. The widespread introduction of CIP could therefore take longer than originally expected. Another hurdle is that many airports wish to use CIP with a remote screening room, which is often too complex or costly. The full range of possibilities and advantages of remote local or matrix-screening, compared to actual remote screening, are not well known yet. However, with the introduction of 3D CBS equipment, this type of CIP implementation, which allows the distribution of images across security lanes (see Figure 1 overleaf), will probably quickly become relevant.

Another major challenge is the risk of network or other system disruptions. Using CIP, there will always be the risk of network or other disturbances, which can cause the loss of images or other security sensitive data. This makes it necessary to create a contingency plan, which allows a quick switch to conventional screening.

Figure 1: Illustration of remote local/matrix-screening (Kuhn, 2017)
Figure 1: Illustration of remote local/matrix-screening (Kuhn, 2017)

What are some trends of CIP?

CIP offers the possibility of ‘live’ monitoring of screener performance, which is currently still used to a very limited extent and its potential is therefore not yet exhausted. CIP could also become a necessary concept when implementing new screening equipment. 3D technologies, for example, allow passengers to leave liquids and laptops in their cabin baggage, but require more time for the inspection of each image. To make full use of these machines, CIP might be necessary to send images across security lanes to the next available X-ray inspection officer and to have more than one officer per machine.

In the long run, the use of CIP will likely increase. CIP could not only move security screening away from the checkpoint, but also away from the airport, to places where labour and infrastructure costs can be reduced. This scenario would also allow for a centralised screening area that combines the X-ray image analysis of several entities at different levels. At larger airports, for example, all terminals could be linked. In countries with many small airports, like Finland for example, there are already pilot projects doing so. The possibility of linking seems to be endless and does not necessarily end at national borders. Theoretically, it would be feasible to set up a central screening farm for all airports in Scandinavia. Of course, such concepts also bear risks and should therefore be analysed carefully.

“…theoretically, it would be feasible to set up a central screening farm for all airports in Scandinavia…”

In this context, another CIP scenario discussed for a more resource-efficient security screening at airports is the combination of different screening areas like hold baggage screening (HBS) and cabin baggage screening (CBS). However, a successful combination is currently considered unlikely. The combination of CBS and HBS at an airport will only be realistic when the same or similar EDS standards are introduced in both areas. Otherwise, the differences in the images and in the image analysis are too great. If similar standards are in place in both areas, the biggest challenge will be the combination of the different prohibited article categories. However, this should be feasible in the long term, considering that an ‘alarm only’ policy, as is already in place for HBS at many airports, is a reasonable future scenario in CBS with the introduction of automated object recognition algorithms. This would mean that screeners review only images that have triggered an alarm (EDS, firearm or knife algorithm).

Time-on-Task using remote screening

Another topic of discussion becoming of higher interest when implementing remote screening is how long security officers can inspect X-ray images without experiencing performance deterioration. In CBS, the EU regulation restricts the inspection of X-ray images to 20 minutes of continuous screening; thereafter, security officers must change to a different position or take a ten-minute break. With remote screening in place, keeping up the 20-minute rotation can be challenging because the screening room might not be located directly at the checkpoint. Longer screening sessions could alleviate such problems. Furthermore, analysing X-ray images in a remote room with less distraction could be beneficial for performance and might enable screening for a longer period. CASRA investigated the feasibility of longer screening sessions by conducting two studies.

Stable detection performance in the lab

To investigate how detection performance changes over time, CASRA conducted a laboratory study with 71 screeners of a European airport. Screeners performed a one-hour X-ray image inspection test. While half of the screeners took ten-minute breaks after 20 minutes of screening, the other half worked for one hour continuously. We found that the groups did not differ significantly in their detection performance (Buser et al., 2020). Hence, there was no indication that the breaks affected performance. Furthermore, both groups were able to maintain their performance over the course of one hour. However, the groups did differ in terms of reported distress (negative stress). Taking breaks reduced the amount of reported distress significantly. Having said that, the amount of distress reported by both groups, was rather low overall. Nevertheless, this could be first indications of negative long-term effects of longer screening sessions and it shows the importance of also examining subjective well-being.

Detection performance and time on task in the field

To evaluate the long-term effect of longer screening sessions, a field study at another European airport was conducted. In this study, screeners performed longer screening sessions as part of their daily work, and not in a testing situation where threat items are more frequent but missed threats have no real consequences. In this field study, we implemented a similar design as in the lab. One group of screeners performed screening sessions lasting 20 minutes (20-min group). Another group was instructed to screen up to 60 minutes (60-min group); however, they were able to end a session early, if they felt tired or they could not keep up concentration.

The distribution of screening sessions per group (see Figure 2) revealed, that it was possible for the 60-min group to screen for longer than 20 minutes. Of all the sessions conducted by the 60-min group 54% of the sessions were longer than 30 minutes, and 12% lasted 60 minutes or longer. It is important to note, that not all sessions in the 60-min group were cancelled due to fatigue. Many sessions were ended due to operational reasons, e.g., screeners were sent to another position or it was the end of their shift. The 20-min group conducted screening sessions that mostly lasted around 20 minutes (95% of all sessions shorter than 25 minutes).

Figure 2: Distributions of screening durations conducted for both groups
Figure 2: Distributions of screening durations conducted for both groups

Comparing the detection performance of the two groups revealed that they did not differ regarding their average hit rate, reject rate or processing time (time needed for the inspection of an image). This shows that, even though the 60-min group conducted screening sessions that lasted longer than 20 minutes, they were able to maintain their performance. Interestingly and contrary to the lab study, screeners in the 60-min group did not report more distress. This could be due to the fact that those screeners were given more autonomy as they were able to decide on their own when they would end a screening session.

How screeners experience longer screening sessions

According to the participating screeners, longer screening sessions are generally feasible. However, screeners reported various factors that influence whether it is feasible to screen for longer than 20 minutes. How fit or well-rested one feels that day and workload were stated to be relevant. Generally, the insights from the surveys and interviews supported the notion that longer screening sessions are feasible. Further, it also highlighted the relevance of factors other than time on task on performance (e.g., workload).

However, it is important to consider that there were large inter-individual differences between screeners. While some screeners found it easy to screen longer, others had more trouble. This is certainly, though not only, related to the screening experience. The feedback we got from the screeners from this study suggests that working on the screen between 30 to 40 minutes is feasible under most circumstances (e.g., high or low workload). A fixed screening duration of 60 minutes on the other hand was perceived as too long for most screeners.

“…working on the screen between 30 to 40 minutes is feasible under most circumstances…”

Implications

Both studies showed that longer screening durations are feasible without noticeable performance decline. This holds true if screeners can decide independently when they want to end a screening session. Therefore, relaxing the EU regulation would be a possible implication. A longer fixed screening duration, however, does not seem ideal since a variety of factors can influence performance and the ability to screen longer (such as screening experience, workload, etc.). It seems more appropriate to set a certain time range during which screeners ideally can decide for themselves when they want to stop. This again can have a positive influence on screeners’ well-being as they experience more autonomy when carrying out their work.

Conclusion

Despite some challenges, CIP offers flexibility and a wide range of possible implementations. It thereby allows for a more flexible and customised security screening process, which can be adapted to the individual needs of checkpoints and airports. In addition, with CIP in place, handling a variety of different screening durations is possible. This could be advantageous in the future if longer screening durations could be legally implemented.


Daniela Buser (left) and Dr. Sarah Merks (right) are research scientists and project managers at the Institute Humans in Complex Systems (www.fhnw.ch/miks) of the School of Applied Psychology, University of Applied Sciences and Arts Northwestern Switzerland. They also work at the Center for Adaptive Security Research and Applications in Zurich (www.casra.ch).

References

Buser, D., Sterchi, Y., & Schwaninger, A. (2020). Why stop after 20 minutes? Breaks and target prevalence in a 60-minute X-ray baggage screening task. International Journal of Industrial Ergonomics, 76, 1–10. https://doi.org/10.1016/j.ergon.2019.102897
Kuhn, M. (2017). Centralised image processing: the impact on security checkpoints. Aviation Security International, 23 (5) (2017), pp. 28-30