Angle Up

Checkpoint footprint

Human-machine interaction in future AvSec screening

Continuous development of Artificial Intelligence (AI) is part of the road towards future baggage screening in aviation security. Explosive Detection Systems for Cabin Baggage (EDSCB) aid operation in the detection of bare explosives and continue to improve performance in terms of detection rates. AI techniques, such as machine learning and deep learning, enable the development of smart and adaptable algorithms for automatic detection of threat objects - referred to as Automatic Prohibited Item Detection System (APIDS).  These algorithms enable support in security operations by detecting threat categories by shape, such as sharp items and firearms.

In this blog we take a look at the road towards future baggage screening in airport security, specifically at the human-machine interaction between operator and algorithm.


Continuous improvement of EDSCB algorithms

The continuous improvement of EDSCB already evolved to a single digit false alarm rate and an improved detection rate. Improvements of EDSCB detection can be related to the detection of minor threat quantities and threat concealments. One key element missing in current algorithms is the detection of complete Improvised Explosive Devices (IED).   

When EDSCB algorithms are able to detect IED’s, this will highly impact the security operation since operators hypothetically won’t have to search for IEDs anymore. As we described earlier in our blog about Operator performance on image analysis, the search for IEDs is the most complex detection category for operators. Because the threat category list is reduced by IEDs in operating procedures, the operator learning curve could increase. In addition, less operator training and coaching could be necessary.

Reducing the threat list with IEDs can result in an enhanced focus on other threat categories and this will most likely impact the operator satisfaction and confidence. The operational impact could be an increased clear baggage flow and fewer high threat occurrences; this expects to minimize the so-called Cry-wolf effect: ignoring warnings due to the past experience of false alarms.

Figure 1 – Operators must search for prohibited items in airport security


Automated Prohibited Item Detection and Image On Alarm

Automated Prohibited Item Detection Systems – APIDS – are the next step towards a more secure and automated security check. These advanced algorithms are able to detect and identify threat items such as firearms, sharps and blunt weapons, based on shape.

Both EDSCB and APIDS detect threats. Yet – their detection methodologies are significantly different. EDSCB detects bare explosives primarily by material composition, whereas APIDS detects threats mostly based on shape. On Screen Resolution (OSR) of EDSCB alarms is not allowed in many countries because the human eye is incapable of material distinguishment by observation. On the contrary, operators are able to perform shape detection, which is applicable on all threat categories, except explosives (see figure 2). Therefore, a so-called Image On Alarm (IOA) is an applicable CONOP scenario for APIDS: operators should only analyse images that contain a threat detected by APIDS.

Figure 2 – Material detection on explosives versus shape detection on firearms, sharps, blunts and miscellaneous

Successful implementation of APIDS requires consideration of different CONOPs scenarios. A CONOPs scenario with IOA allows for alarm resolution by the operator and could impact the operation on several aspects. Fewer images will be presented to the primary screener which results in an additional increased clear flow of baggage. Next to this, an alternative staffing model could be considered as fewer images requires analysis by an operator: remote screening with fewer operators or combining CBS and HBS operator activities.

When images are presented with alarm annotation operators can adjust their focus to a specific region of interest instead of analysing the complete image; operator focus will increase, and analysis time will decrease. Yet, it is important to consider additional operator GUI interaction due to image clutter of alarm annotations. Changes to the Operator GUI could help to increase intuitive GUI usability.

Fewer false alarms will be sent to recheck for secondary screening. Passenger experience could increase because this results in less queue time and less intrusive procedures. This also means an increased number of correct alarms that require secondary screening. Secondary screening on correct alarms requires more processing time than secondary screening on false alarms.


The challenges and opportunities of Auto-clear

Combining EDSCB with APIDS is the next step towards a more automated security checkpoint in the future. When the detection threshold enables the detection of all threat categories this will allow for implementation of auto-clear software; images can be cleared automatically if no threat is detected by EDSCB and APIDS. The implementation of auto-clear will impact the operational environment on various aspects.

Implementation of auto-clear software will result in fewer rejects of images. Therefore, operators will shift their focus towards alarm resolution procedures. The initial and recurrent training program for image analysis should then be adjusted accordingly.

The staffing model will be impacted by auto-clear implementation as well. Performing less alarm resolution also enables the opportunity to combine operator tasks. For example: combining passenger screening with alarm resolution for cabin baggage, or alarm resolution with first-line equipment support.

Rotating shifts within the security team on site will change with an auto-clear CONOP. There will be less start-stop behaviour of the lane and a passengers will experience a more continuous flow.

Auto-clear software will help in the improvement process of the security culture. Operator confidence and operator satisfaction levels might increase, different career opportunities arise, passengers will experience a less intrusive security procedures atmosphere and less need for ethnic profiling might exist.

Auto-clear will also impact the operational environment physically. Security lanes could require extended reclaim space for passengers when auto-clear software is implemented.  Divest positions could be extended as well due to a more continuous flow. Also, the position for the primary screener is no longer needed in the security lane. Therefore, this space becomes available. This could allow for narrow security lane setups, improvement in equipment sustainability and a more efficient equipment utilization. However, when changing the checkpoint footprint, the potential displacement effect of bottlenecks in the passenger journey – such as border control processes – must be considered as well. Relocating the bottleneck is never the objective in the passenger journey and must be considered in alignment to auto-clear implementation.


In this blog post we shared our thoughts on the future of AvSec Human-Machine interaction and how this can impact several factors in airport security. Dialogues are necessary among all European security stakeholders to define AI operational requirements, perform the research and guide the innovation process to strive for state-of-the-art AvSec solutions compliant with Europe’s regulations.

We guide security equipment deployment by keeping integration as our key focus. This covers technology, process and at heart: people. Every change results in a new situation.


Explore what Point FWD has to offer in guiding security stakeholders toward successful equipment implementation in airport security checkpoints.

Visit our EDSCB page for more info or send an information request via the button below.