Smile: Software Captures Faces in Bad Surveillance Imagery
Imagery collected in a chaotic warzone isn’t always pristine, making it difficult for analysts to identify suspects in photos and videos.
Particularly at long ranges, atmospheric “noise” can wreak havoc on the quality of images being collected by surveillance cameras.
New Hampshire-based Animetrics has developed technology to create clear 3-D facial renderings from low-quality photos and videos. ForensicaGPS uses a patented system of mathematics and mapping to generate 3-D geometry from 2-D coordinates. It takes images, improves visibility of the face and creates a model specific to structure, geometry and texture.
The application also performs metric analysis that scores subjects based on facial similarities. Up to five images of each subject can be used to create the 3-D model and to compare it to other suspect images.
“The 3-D information of the face biometric is created from normal photographs, not from special 3-D sensors,” said Paul Schuepp, president and CEO of the company.
Animetrics has not tested its product on feeds from drones, one of the most popular ways U.S. forces gather visual intelligence from long distances. However, it has begun experimenting with long-range facial imagery under an Army contract with Securics, a company focused on eliminating the effects of atmospheric noise.
“It’s logical to understand that this technology would be directly applicable to drone surveillance that takes pictures at long distances,” Schuepp said.
ForensicaGPS is already being used by military intelligence and law enforcement agencies.
Photo Credit: Animetrics