Over a million developers have joined DZone.

Facial Recognition That Works In The Dark

DZone 's Guide to

Facial Recognition That Works In The Dark

Know about thermal camera technology, which has recently moved from deployment on aerial and ground vehicles to body-worn cameras.

· AI Zone ·
Free Resource

Facial recognition technology has come a long way in the last few years. A recent paper from the U.S. Army Research Laboratory highlights the progress. The researchers have developed facial recognition technology that can operate effectively even in the dark.

The technology utilizes AI to produce a visible facial image from the thermal image of that person's face. The approach relies on the advances made in thermal camera technology, which has recently moved from deployment on aerial and ground vehicles to body-worn cameras. The researchers hope that their work will eventually enable human-matching, even at night.

"This technology enables matching between thermal face images and existing biometric face databases/watch lists that only contain visible face imagery," they say. "The technology provides a way for humans to visually compare visible and thermal facial imagery through thermal-to-visible face synthesis."

Operating at Night

Facial recognition technology requires light to work, so currently, the software would require a flash or spotlight to function at night. By contrast, thermal cameras capture the heat signature of living skin instead, so is ideal for night-time usage.

The challenge becomes matching the data captured by the thermal camera with mugshots of people captured during the daytime. It's a problem known as "cross-spectrum" or heterogeneous face recognition.

Their method for achieving this has two main parts. The first is a non-linear regression model that allows thermal images to be mapped onto a visible latent representation, whilst the second projects the latent projection back into the image space.

The researchers have demonstrated the effectiveness of their solution in near real-time. The demonstration featured a FLIR Boson 320 thermal camera and a laptop running the software. The demonstration showcased how the thermal image could be used to produce a synthesized visible image in situ.

They plan to further develop the technology with the help of the Defense Forensics Biometrics Agency to hopefully provide the technology to soldiers in the near future.

facial recognition ,ai ,thermal imaging ,machine learning

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}