The Intel Community Wants to ID People from Hundreds of Yards Away
Face recognition alone isn't good enough, so a new IARPA program is looking to combine multiple biometric indicators to get good matches.
The intelligence community is working to build biometric identification systems that can single out individuals from hundreds of yards away or more, a feat that’s virtually impossible using the technology that exists today.
Ultimately, the tech would let spy agencies rapidly identify people using cameras deployed on far off rooftops and unmanned aircraft, according to the Intelligence Advanced Research Projects Activity, the research arm for the CIA and other intelligence agencies.
Facial recognition and other types of biometric tech have improved significantly in recent years, but even today’s most advanced systems become less reliable without a crystal clear view of their subject. Even when the person is standing nearby and looking directly into the camera, facial recognition tech can be prone to errors.
But the intelligence community is trying to overcome those limitations in two ways: gathering more extensive training data and creating systems that lean on multiple types of data to identify people.
On Friday, IARPA started looking for researchers to participate in its Biometric Recognition and Identification at Altitude and Range, or BRIAR program, which aims to develop identification tools that work from vantage points high above or far away from their subjects. While the program is still getting off the ground, the tech it seeks to develop could improve the government’s ability to surveil adversaries—and citizens—using biometric data.
“Further research in the area of biometric recognition and identification at altitude and range may support protection of critical infrastructure and transportation facilities, military force protection and border security,” officials wrote in the request for information.
Teams interested in participating in the program must respond by Oct. 21.
In the request for information, IARPA asked teams for a wide variety of datasets that could help train biometric technology to work in less than ideal conditions. Today, the range of facial recognition and other identification systems is limited by a lack of training data, they said, and more datasets would help researchers build more versatile and powerful tools.
Specifically, IARPA asked for images of individuals taken from more than 300 meters away or at pitch angles above 20 degrees, as well as biometric research datasets captured by drones and other aircraft.
The program also aims to explore technologies that take in multiple types of biometric identifiers—such as face images, movement patterns, height and gender—instead of relying on a single data point. Building this composite image a subject would allow systems to accurately identify individuals even if certain data points aren’t the best quality. In the solicitation, IARPA asked teams to provide any data that might help train systems on these other modalities.
“Imagery captured at long-range or altitude may also require that additional biometric signatures be fused with face recognition to provide the necessary accuracy or confidence to be usable for person identification,” officials said. “The fusion of multiple biometric signatures to address these limitations remains underserved by the research community.”
Under a separate research effort, IARPA is also working to develop artificial intelligence systems that can track people as they move through broad video surveillance networks.
NEXT STORY: This Formula Predicts Soldier Firepower in 2050