AI Bag Scanner Is Faster And More Accurate Than Airport Staff

One of the biggest frustrations for air passengers is standing in line, waiting for their bags to be scanned.

And one of the biggest dangers for the airline industry is that security staff, who spend their days staring at the screen, will miss something.

SeeTrue, an Israeli startup, addresses both problems with an AI system that’s trained to recognize guns, knives, explosives and other dangerous items quickly and accurately.

It plugs into an airport’s existing X-ray and CT scanning systems, and SeeTrue says it’s three times better than a human operator at spotting prohibited items, 10 times faster, and with half the error rate.

“SeeTrue uses artificial intelligence and computer vision algorithms to discover prohibited items in bags,” says Assaf Frenkel, CEO at the Tel Aviv-based company.

“It detects in real-time, faster and more accurately than most human eyes, always on, and never getting tired or distracted.”

It’s already deployed at several airports in Europe – Frenkel won’t say which – where it detected over 500,000 prohibited items last year. That covers anything from scissors and screwdrivers to drugs and banknotes.

“Nobody likes standing in line and airports care about your experience, but they also care about their costs,” he says.

“Most of the people working in an airport are actually working in security. But the whole operation of having a lot of people looking at bags, and opening them 24/7 is quite expensive.”

Recruiting security staff, especially post-Covid, is also a challenge. Thousands of flights were canceled across Europe last summer as airports struggled with ground staff shortages.

“There wasn’t a shortage of pilots, or flight attendants. It was security officers,” says Frenkel.

“They need to concentrate on fine details for hours on end, work long shifts across weekends and holidays, and the pay is not great.”

Not only that, but the results aren’t great either. He says even the world’s busiest airports (called Tier 1, with over 25 million passengers a year, such as Dubai, Istanbul and Amsterdam) perform poorly in what the industry calls “penetration testing” and “red team assessments”, when their security systems are checked for vulnerabilities.             

Human operators simply can’t maintain full concentration for a whole shift. And that’s where AI comes in.

SeeTrue’s Autonomous AI Detection can work round-the-clock, minimizing the number of times passengers have to remove items from their bags, and recognizing prohibited items regardless of the angle, size and other obstacles blocking the view.  It also raises fewer false alarms.

“We’re doing significantly more detections (of prohibited items) than human operators,” say Frenkel, “so we provide better security with a much lower error rate.

“So more passengers go through in an hour, and that means they have more leisure time, spend more money in the airport, and get to their flight on time.”

The AI still requires some human intervention, to open a bag once it’s flagged an item.   

Training the AI to recognize dangerous items – which is key to the system’s success – was a complex task. 

We all think we know what a gun looks like. Show the computer every known gun, from every possible angle, and it will learn what’s a gun and what’s not a gun.

But what about a knife? Or any bladed item, from a razor to a Samurai sword or a dagger concealed inside an umbrella (a real-life example picked up by SeeTrue) and everything in between? How do you explain to a computer to identify any item that humans regard as a bladed item?

“It’s looking for something that is not well-defined, in a chaotic environment,” says Frenkel.

Every bladed item is different, and every item of baggage is different. The AI has to be extremely robust, and be able to deal with complex objects and situations it has never encountered before.

“No automatic detection solution is going to be 100 percent,” he says. “But think about it this way: we’re required to be at least as good as an operator in order to replace him or her, and we are way beyond the typical operator in the field.

“We have a team of experts who know how to use explosives and how to conceal different items, so we train the system with hard constraints.

“There is some competition in the market, and we welcome competition. We’re privileged to be one of the first in this market, we’re very focused on what we’re doing, and we were able to get our products out and get real field experience and use that real field data.”

Airports are the biggest market for SeeTrue’s technology, but there are many more opportunities – sea ports, railways (the systems is already used by Israel Railways), US schools, custom shipments and Amazon deliveries.

The AI can be trained to recognize pretty much any object, so it can be used to stop workers stealing from their employers.

“Either you want to avoid things coming into a place or out of a place,” says Frenkel. “The AI can identify anything that a person can identify.”

Subscribe to NoCamels weekly newsletter and get our top stories