Three decades ago, video surveillance was seen as Orwellian dystopia. Today it is everyday in subways, shopping malls or public buildings. Although re are still not as many cameras in Germany as in UK, where an estimated five million pieces follow steps of citizens. But now next stage follows: At Berlin S-Bahn station Südkreuz, face recognition has been tested via video recording since July.

If, in European cities, re is now almost a monthly rhythm of terrorist attacks, temptation is great to want to be even more densely on track of potential "Gefährdern" in ir movements in public space than cell phone surveillance allows. Electronic facial recognition is intended to make this possible. It has become ever better in last 25 years. From analysis by "self-faces" in early Nineties via 2001 featured algorithm from Paul Viola and Michael Jones to DeepFace technology of Facebook, hit ratio has increased to 97.25 percent. For comparison: People recognize faces on average with an accuracy of 97.64 percent again. Most of technical procedures try to identify characteristic facial parts such as eyes, nose and mouth. To do this, you analyze faces pixel by pixel for brightness differences. The eyes are usually darker than intermediate nasal ridge.

The US computer scientist and artist Adam Harvey has developed a protection against most widely used technique by Viola and Jones: CV Dazzle. Harvey has hairstyles and makeup out on which program fails: hair strands covering an eye area, as well as painted rectangles or triangles in black, white or red, successfully disguise facial features and complicate analysis.

In obscure music clubs, se make-ups would probably go through as last craze of pop culture. In everyday life y are unfortunately screaming conspicuously. Researchers at Carnegie Mellon University have discovered a more subtle defense: eyeglass frames in natural horn optics. At 19 of 20 test faces from a database, spectacle-wearing faces could no longer be assigned to right person. Even more: With a matching sophisticated pattern, a test face was held for actress Milla Jovovich – although it was a man.

This text comes from time Knowledge magazine 6/17. You can purchase current booklet at kiosk or here.

But it's even easier. The Chicago Optician Scott Urban has provided eyeglass frames with a reflective layer that rethrows infrared light back. Surveillance cameras in dark angles use light to illuminate scenery. If eyeglasses reflect it, y create a glittering glow on camera image that covers face. Where not illuminated, an idea of Japanese engineer Isao Echizen can help. His eyeglass frame has even light emitting diodes whose infrared light also covers face on camera image.

These inventors and privacy activists will not win arms race. The algorithms are getting better, face databases are getting fuller. There are already approaches to identify masked faces even with cloth and sunglasses. And who, for example, at a border crossing, is facing eye of law, must show up eh. Glasses, hats, wild make-ups and or things on face are all not allowed.

The source information for time-knowledge-article can be found here.

Date Of Update: 07 November 2017, 12:03