Imaging gets personal

CHAMPAIGN, IL--Biometric imaging requires individuals to willingly submit personal data such as photos, fingerprints, and retinal scans for database storage (see “Threat identification demands new imaging technologies” at www.laserfocusworld.com/articles/358147).

CHAMPAIGN, IL--Biometric imaging requires individuals to willingly submit personal data such as photos, fingerprints, and retinal scans for database storage (see “Threat identification demands new imaging technologies” at www.laserfocusworld.com/articles/358147). And even facial-recognition software requires a database of known faces against which to compare a series of images. But what about the case of an unknown threat--a person described only in terms of approximate age, hair color, height and weight, type of clothing, or other personal attributes? In such a situation, image-processing algorithms need to assess a number of different image cues in order to identify a threat within an image scene of interest.

Researchers at the University of Illinois at Urbana-Champaign utilized multiple linear regression data from the images of 1600 faces to develop an imaging algorithm that estimates a person’s age. While the technique is best applied to non-homeland-security applications such as determining what types of soft drinks teenagers like most (video cameras would identify persons of teenage status and record what they were buying), it could also be applied to a homeland-security situation in which a “person of interest” of a known age needs to be found in a crowd of individuals. The software is currently 80% accurate when estimating age to within ten years.

--Gail Overton

More in Home