In today’s blog, we get back into the world of Biometrics, focusing upon what is deemed to be the most controversial of them all-Facial Recognition:

How Facial Recognition Technology Works

The science of facial recognition technology are generally the same as the other biometric technologies:  (1)  Acquisition of facial images; Image processing of the facial images; (3)  Unique feature extraction; and (4)  Template creation.

However, facial recognition technology can be limited by a different set of constraints when compared to other biometric systems. This is the case in the acquisition phase.  For example, it is very crucial that the enrollment template is of very high quality for future identification and verification.

Whereas other biometric technologies will allow for some margin of error (for example, as in the case of hand geometry recognition, a somewhat dirty hand can still be used to create an enrollment template), this is not true for facial recognition.  The quality of the enrollment template is impacted by a number of factors.

First, the user must look at the facial recognition camera at a close range and at certain angles so that a high quality enrollment template can be captured. The second factor is lighting.  The lighting must be perfect in order to assure a good quality template.

If the image is under exposed or over exposed in any way, this can have a negative impact on the quality of the template.  Once these constraints in the acquisition phase can be overcome, the next phase is image processing.  Here, the images of the face are converted into black and white images.

These images are also cropped, rotated (clockwise or counterclockwise), and magnified.  In the third phase, unique feature extraction, the facial recognition system looks for unique features around the sockets of the eye, cheeks, mouth, and nose.  A constraint in this phase is that the structure of the face can change.  For example, weight gain or loss can change the unique features of the face.

Also, non biological changes in the face can have a negative impact upon unique feature extraction. This can include such things as wearing eyeglasses in the enrollment phase, but not wearing them in the verification phase, the removal of any facial hair, the wearing of cosmetics, etc.

In the fourth phase, template creation, the enrollment template is composed from the unique features taken from the face.  The enrollment template varies in size—from 100 bytes to 3 kilobytes.

In tomorrow’s blog, we look at some of the various applications of Facial Recognition.