Scott Air Force Base is looking for verification and identification algorithms for a facial recognition pilot.
The Air Force is considering using facial recognition technology at base entrances and plans to start with a pilot program at Scott Air Force Base in Illinois.
The base has an automated system for approving vehicle access, but officials want to increase the speed and security of that system by adding facial recognition in the mix. Under the current plan, drivers would approach the access control platform and have their photo taken, which then would be matched against a verified ID or a photo gallery of people approved for access that day.
The Air Force is looking for algorithms that can both verify that a person’s face matches the ID they present—known as one-to-one (1:1) or verification—as well as match a live picture against a database of up to 10,000 images—one-to-many (1:N) or identification.
One-to-one verification matching has a much higher accuracy rate, but the ID being presented must also be in the database. One-to-many identification is more versatile but can have higher error rates.
To be eligible, algorithms must have been tested by the National Institute of Standards and Technology and placed among the top five in the Facial Recognition Vendor Test for one-to-one and one-to-many. All vendors must also be based in the U.S. to participate.
While contracting officers are specifically looking to test algorithms during the pilot phase, potential vendors should also be prepared to supply necessary hardware for use during the proof of concept.
In a question submitted to the Air Force, a vendor suggested adding license plate reading to the mix to increase accuracy with another data point. Contracting officials said that is not a necessary part of the pilot but the Air Force is “interested in this option.”
The 375th Contracting Squadron is currently gathering a list of potential offerors through a sources sought announcement. Interested vendors should respond by 4 p.m. Tuesday.