Analysis of scanned images, of both baggage and passengers, is one of the most important tasks in aviation security, but inevitably, also one of the most repetitive.
Whilst this repetition might make it problematic for human undertaking, it is what makes it perfectly suited to AI, with tools available that can rapidly analyse scanned images, identify abnormalities and classify threat type to facilitate human intervention.
Our solutions can be integrated with industry standard systems, including wide area based remote operations for both hold and hand baggage processing and people screening systems.
Threat Detection in Baggage and People Scanning
With airports now deploying remote and multiplexed monitoring of screened images, threat detection operators are faced with a variety of images from a range of scanners.
Our AI tools provide vital assistance, carrying out high-speed visual checks that can expose anomalies, highlight risks and prompt closer inspection.
In threat detection applications, dedicated AIs are trained on industry-standard Threat Image Projection (TIP) images whilst enhancing the overall detection capability by recognising non-threat specific differences between threat/no threat image data. It is a proven approach that has parallels in the use of AI to analyse and identify anomalies in medical images.
The emergence of auto-encoders to learn what normal looks like and then raise alerts when abnormal patterns are detected is transforming our capability, as ultimately this approach overcomes the issues with traditional methods as a way to keep pace with emerging threats as well as those we have seen before.
We are collaborating with government bodies and leading industry partners, building very large image data sets with which to train our neural networks.
For this application, the AI has been tested using imagery from the threat image projection (TIP) library approved for use in UK aviation baggage-screening equipment as well as very large numbers of benign, no threat images so an initial normal/abnormal check can be carried out. Images containing abnormalities are flagged for re-check, with specific known threats from the TIP imagery classified.
An abnormal image with no classification based on the TIP may still contain a threat, just one which has not been included in the TIP image set as yet. This methodology is using the constantly evolving nature of AI to potentially detect threats which have yet to be encountered.
Today’s Passenger Screening equipment using millimetre wave scanners must be image free to the operator for reasons of privacy. In this application, our AI has been trained to identify concealed threats and direct operators to their location.
As part of the Future Aviation Security Solutions program (FASS), we have also collaborated with the developers of a revolutionary new type of scanner, which uses terahertz imaging to screen passengers who do not have to remove outer clothing or stand still due to its sensitivity.
News and Insights
Visit our news page for more information.
Whitepaper: The Common Departure Lounge
Download our whitepaper to find out more about using biometrics to manage increasing passenger numbers.