FaceVision 3D Face Recognition Technology

Geometrix ActiveID systems are the first biometric solutions to fully leverage 3D facial shape. Facial shape is a new biometric, not a new buzzword or variation on known facial recognition algorithms. Geometrix thinks it is such an important advance that we'll even tell you how it works. It's a little complicated, but as they say, knowledge is power.

Why face recognition was not a strong biometric in the past

Simply put, a picture of a face is not a face; it is a collection of pixel values generated by a complex interplay between scene lighting, the face's 3D shape, the face's reflectance at each point for each color, the background in the scene, and the distortions of the camera itself. Not much of this has anything to do with a person's identity! Yet, conventional facial recognition systems are expected to compare how similar the face in one collection of these pixels is to the face in another - a mathematically daunting task.

So how good is it?

FaceVision shape results are not statistical abstractions from a pixel pattern, they are forensic quality 3D facial models like one might see on the "CSI" TV show. For automated identity matching, the models are compressed into tiny 3KB templates. If the differences between two templates are too large to be accounted for by FaceVision algorithms for interpreting facial expressions, facial hair, donning of eyeglasses, or other changes to an individual face, then they are just not the same face.

How "tight" is FaceVision's Identity verification performance? Geometrix specifies 0.03% - 3% or better equal error rate depending on application, but in both Government tests and real installations, it works a lot better than that. Any better than any facial alternative on the market.

The uncompressed models can also be displayed to a guard for manually zooming in from any angle and comparing an enrolled subject to the guard's eyeball view of the live one - a critical capability for reliably handling exceptions in any automated ID control system. Or, the live subject may be acquired by the system, and compared to the enrolled subject interactively using sub-mm precise measurement tools.

ActiveID systems with FaceVision technology provide identity verification performance rivaling fingerprint systems, with the ubiquity facial recognition has always promised. Geometrix delivers the true shape of human identification, and our performance page describes just how well.

Copyright © 2007. All rights are reserved.
The true shape of human identity.


An Image Capture from the 3D FaceVision software showing the dual 2D pictures that are converted into a 3D biometric template.

Welcome to the new shape of human identity

ActiveID systems with FaceVision technology use pairs of cameras and automated photogrammetry techniques to directly acquire the 3D shape of the human face at sub-millimeter precision, as well as the color at each point. They acquire everything visually unique about a face in an instant, completely independently of scene lighting, pose angle, camera distortion, and anything else that is irrelevant to a person's identity. They do it reliably under any lighting condition tested, from dark rooms to direct sunlight. No other sensor can do these things, and no one has ever figured out an approach for spoofing it.


A verification where the subject is wearing sunglasses. Typically verification will fail for most 2D systems.