Insight by NEC Corporation

Facial recognition keeps advancing

As facial recognition technology improves, federal agencies are finding more ways to use it in law enforcement, national security and administrative functions such as ID verification. Thanks to advances in recognition algorithms, facial recognition systems are getting better at comparing two images or obtaining accurate “vector maps” in poor lighting, at greater distances and at wider off-center camera angles.

Federal News Network and NEC Corporation of America assembled a panel of federal facial recognition expert to discuss the newest trends.

Among the group’s findings: Facial recognition is catching up to iris recognition in accuracy, without the finicky requirements necessary to get a good eyeball scan. It’s not quite up to fingerprints for identity purposes, but is increasingly used in conjunction with fingerprints.

In fact, according to Pruitt, development is moving so rapidly at NEC (and its competitors) that refreshed algorithms are coming out quarterly. Among the developments to make facial recognition more accurate and versatile: Establishing vector maps around the skull so that people can be reliably recognized in silhouette or even from behind.

Grother pointed out, the application of “convolutional” neural networks to train algorithms over large data sets is producing large gains in accuracy.

Facial also has potential in the field of customer experience, where fast and easy verification is a must. For instance, Customs and Border Protection is piloting a facial recognition application for the exit-entry for international travelers at several airports.

Current State of Biometrics and Facial Recognition

There’s been something of a revolution in facial recognition in the last year years, with a whole set of algorithms coming on line based on neural network technology. The benefits that we’ve measured at NIST are large accuracy gains.

Enhancing the Customer Experience

Because the technology is becoming much more ubiquitous, much more capable, generally lower cost, creating greater flexibility not only in how we collect but match images, the number of use cases we expect in DHS is likely going to increase.

Iris Recognition and A Glimpse into the Future

Improving face recognition alone isn’t the only answer. It’s really bolstering that entire ecosystem. So with face we’re trying to improve performance from different angles, different lightings…But then you add in things like tattoo recognition, you add in iris, you add in other information and fuse that into one result, you can get a better estimation of identity that’s more secure and convenient.

 

Listen to the full show: