A test conducted by the Department of Homeland Security Science and Technology Directorate revealed that facial-recognition technology might reduce the need for travelers to remove their face masks at airports or other ports of entry. Contrary to our collective smartphone face-detection experiences, the platforms and configurations tested delivered accuracy rates as high as 96 percent for subjects who were wearing protective face coverings.
The research, conducted this fall at the directorate's 2020 Biometric Technology Rally in Maryland, involved six face/iris-acquisition systems and 10 matching algorithms. To address the need for greater racial and ethnic diversity in facial-recognition tech, the in-person event tested 582 diverse test volunteers, representing 60 countries.
Preliminary findings released this week were based on the systems' ability to reliably match images of each volunteer with and without a variety of masks, the processing time for each instance, and the volunteers' overall satisfaction with the experience.
Not unexpectedly, the performance varied greatly among the products. More surprising was just how effective the highest accuracy rates were:
- Without face masks, the median performance-identification rate was about 93 percent, with the best-performing system correctly identifying people 100 percent of the time.
- With face masks, the median performance-identification rate was a solid 77 percent, with the best-performing system correctly identifying people 96 percent of the time.
With numbers like that, the Covid-19 risks involved in traveler mask removal could be greatly reduced.
"This isn't a perfect, 100 percent solution," said Arun Vemury, director of S&T's Biometric and Identity Technology Center, "but it may reduce risks for many travelers, as well as the frontline staff working in airports, who no longer have to ask all travelers to remove masks."
More information about the test process can be found here. The organization expects to release final test results in the coming weeks, which will published here. The participating facial-recognition vendors all remain confidential per the terms of the test agreement, and will be referred to in the final report using aliases; anyone interested in learning more about specific top performers at that time can contact DHS S&T to be connected directly.