Facial-recognition software has taken years to mature: I investigated the technology almost 10-years ago and it was not new then. Now, it appears that about 40 law enforcement agencies across the country will be taking delivery of some 1,000 systems from B12 Technologies, to potentially screen persons they have reason otherwise to detain. The Wall Street Journal took the expected tack of questioning the purported violation of civil rights and coined the term facial profiling.
Facial recognition, fingerprinting, iris scans, body odors, thermal mapping, and similar technologies come under the general heading of biometric identifiers. The Department of Defense created its own Biometric Identity Management Agency (BIMA). Apparently, they took delivery of 7,000 facial recognition devices from L-1 Identity Solutions to use in Afghanistan and Iraq. At today’s estimated cost of $3,000 – $5,000 per unit, this would put the entire number of fielded units at under $50 million. Economists state the market as over $500 million, but I often look askance at their purported valuations. To me, that would seem to be how much is spent on R&D, development, and the like, not how much is sold as product.
The law enforcement version of the facial recognition uses a camera attached to an iPhone. The system supposedly interfaces to a database that is still being populated. A facial database is considerably more complicated than just scanning a page of mug shots from the Law and Order Criminal Intent files. Creating and maintaining a quality and broad database is an expensive asset.
Is the database information private? If Julian Assange avoids extradition to Sweden for his hormonal ambitions, he can focus on a follow up to Wikileaks and call it Facialleaks.
Facebook tried its own application for facial recognition, purportedly used to find your friends. I’m getting to the age where all my friends show pictures of themselves taken 20-years younger; that will challenge both the database and the software. My neighbor had cosmetic surgery. Now, she looks much younger than me. However, all those cosmetically reproduced faces have a certain similarity that make differentiation more difficult. Seeing a stranger fresh off the knife can produce one of those I-think-I’ve-seen-her-somewhere-before moments, when, of course, you have not.
The facial recognition algorithm requires measuring some 100+ points on the face, distance between eyes, angles, etc. This can scarcely be done with a single photograph so multiple shots are taken and put together in a stereo fashion so that 3D measurements can be extracted. I would expect that a good algorithm would permit new photographs to be added to the old ones and thus improve the learning; deciding which photographs to use for the database could require a whole new department.
My dad claims that he never forgets a face. Most people have this ability, but a small percentage of the population does not. (It has nothing whatsoever to do with the ubiquitous problem of forgetting names.) There is a special part of the brain that is wired just to recognize faces. It develops at a very young age. If for some reason it does not develop adequately, either through psychological or physiological obstacles, then that person has what is known as prosopagnosia, face-blindness, from the Greek, a “face-doubter.” The prosopagnosic cannot reliably recognize faces. I am faced with this problem, I suspect, because at a young age I had debilitatingly poor vision that was not corrected until much later in life. It produces great difficulty in reliably recognizing even people I know very well, particularly if I encounter them in an unexpected aspect or environment.
Facial recognition should be a hot market in a few years, for the technologists, the end users, and the lawyers. Expect more lumps and bruises though before the technology is ready for prime time. It is in the wings, however, ready to walk on stage. This should be an interesting technology to watch.