I wonder if someone could find Bin Laden with these…
Brazilian police will use futuristic ‘Robocop-style’ glasses fitted with facial recognition equipment to identify and root out troublemakers at the 2014 World Cup. A small camera fitted to the glasses can capture 400 facial images per second and send them to a central computer database storing up to 13 million faces.
The system can compare biometric data at 46,000 points on a face and will immediately signal any matches to known criminals or people wanted by police. If there is a match a red signal will appear on a small screen connected to the glasses, alerting the police officer of the need to take further action or make an arrest.
[…]
The camera will generally be used to scan faces in crowds up to 50 metres (164ft) away but can be adjusted, if searching for a specific target, to recognise faces as far as 12 miles away.
This sounds like B.S. Frankly, it’s more likely that some company and some government officials are making some money than this will actually work in a crowd.
> “The camera will (…) scan faces in crowds up to 50 metres (164ft) away but can be adjusted (…) as far as 12 miles away”
Suuure… Where’s that BS meter graphic?
I’d think stationary camera could do this better–hard wired and all.
they must be doing this in the UK/London “all the time?” or do the bad guys get away before RoboCop shows up, hence the glasses?
I support this. Total BS LIEberTARDian crap to call this efficient use of police resources “Big Brother” ((or progressive?–ha, ha)). The “harm” if any there will be, and there will be without doubt, is the “laws” that are passed and violated to begin with.
But the best way to get a law changed, is to fully enforce it. I don’t believe that, but its often said. So many BS laws. We should have the MOST EFFECTIVE law enforcement possible===and then only the laws needed?
I know–doesn’t fit on a bumper sticker. “Born too Loose”
ORLY? Here in Brazil? Total BS… The government is having trouble on reforming 10 of 14 airports which will be used in the soccer cup in 2014. If they can’t do it right on airports they won’t have any robocop glasses. Believe me, this is the most awfull government we have ever had in gitmo amazon jungle…
Leo–Rio is not in the Amazon Jungle. I believe the rest of your critique though===and if they do get such glasses/ID system, the money really could be better used elsewhere, argument to follow?
I thought Brazil was riding the crest of one of the best governments emerging? Which ox of yours is getting gored?
Self Interest or Patriotism==whats at play?
They’ve been watching too much CSI. Facial recognition and searching a 13 million face database is not something that can be done in a few seconds.
chuck–I don’t know. I’m amazed by google everytime I use it. I’d think properly written the ID system would be wicked fast.
But I bow to your expressed expertise.
Always smile. They can only recognize your face when you are not smiling. Hence the reason they won’t let you smile for your driver’s licence picture.
Somewhere a salesman for a facial recognition system company is celebrating the coup of the century. I have a feeling that the US taxpayers are going to end up covering the cost of this system for Brazil, using national security as an excuse.
If the vendor is a US based company, then the “bailout” is sure the happen. It is simply a transfer of money from US taxpayers to the US defense industry.
This is classic MO from the “Confessions of an economic hitman”.
I suspect they have fallen for a scam. It’ll work but not very often and not very well. There’ll be lots of false positives and even more false negatives. In the end, some contractor will walk off with millions of Reals and the cops will have a pair of cool sunglasses.
smiling good, so is frowning, or twisting your jaw to the side, or putting some bits of paper under your lips to give you semi buck teeth, filling your cheeks with air,lifting one eyebrow up way higher than the other.
maybe even a fake nose and mustache glasses set.
i like to practice this when going to pubs and clubs that require a photo before entry, they will never get my straight face on record, and if they refuse you for looking like a weirdo, then you accuse them of discrimination against retards.
I’d buy that for a dollar!
That is not fair … it is discrimination against the English ….
In #5, chuck said: not something that can be done in a few seconds.
Actually, it can. (I worked on something similar years ago.)
With internet enabled cameras by Canon, servers by Sun and software by Oracle.
As far as distance goes. facial features are extracted on a reference scale of the face itself, which takes pixel mapping out of the equation.
It works by taking the measurement relative to each other for:
• eyebrows, (color, thickness [or bushiness], length and angles),
• eyes (color, size,)
• nose (length, proboscis protuberance, and angles),
• cheekbones,
• mouth (lip thickness and length),
• overall color and
• any and all scars (same measurement points as eyebrows).
All measurements are given a range ± and looked up on a database.
The speed of the matching depends on your budget.
The more of the database that can be RAM resident, the faster you can match. If you have a Sun server with a few dozen gigabytes of RAM, you can match faces detected in a crowd (you’ve seen the cameras that outline faces haven’t you?,) against quite a few suspects,
I forgot to mention that it works about as well as a mother picking out her kid in a picture of a crowd of toddlers.
It also doesn’t work if you’ve had reconstructive (or drastic) plastic surgery done since the reference picture taken.
Further to the Big Brotherism, at the time the system wasn’t cheap and it wasn’t fool-proof.
A few years of iterative application of Moore’s Law, cost is no longer an issue.
The algorithm for detecting and extracting the facial features has been perfected. Now it works slightly better than a trained human being.
The algorithm for self-referential scaling has been perfected (it wasn’t very difficult to do in the first place. {How do you think AFIS works, matching thousands of requests from all over the ‘States against tens of millions of fingerprints in its database? {If you left a print at a crime scene, don’t do it again ’cause you will be caught if you’re arrested by major metropolitan police force.}])
The database matching was perfected years ago.
The algorithm for ranking the degrees of certainty when matching features has been perfected to 80%-90%.
The system can detect a face (in a crowd it detects every face,) searches against its database, drops you or flags you as a person of interest (with links as to why you’re interesting.)
In other news,congress to be fitted with advanced embedded AI goggles to allow them to spot obvious total BS attempts to extract huge funding for security.
The goggles are believed to be able to spot waste at the level of one micro-tsa (defined as 1millionth the amount of money wasted by the TSA)
It´s ridiculous, Rio de Janeiro has a Murder Solution rate of 4%.
In #17, Michael said: Rio de Janeiro has a Murder Solution rate of 4%.
Rio has “third world problems.” So does most of this planet. (I’d like to insure that this was applied to Washington as well. [Fat chance of that happening.])
I wonder why, huh? 🙂
One major problem with surveillance in a machismo society is that the people you want to keep an eye on are the sort of people who take great offense at, and are more than willing to get in your grill about, your keeping an eye on them.
Its one of life’s little ironies.
The story is complete BS.
Facial recognition via a head mounted camera, up to 12 miles away? Right. Facial recognition with anything smaller than the Hubble Space telescope at 12 miles away? Still unlikely.
If you have a small database of maybe 500 subjects, then it might be possible, using some kind of supercomputer, to do a real-time facial search, with lots of false positives and negatives, on a small steady flow of subjects. such as at a single entry point. But it certainly won’t happen via multiple cops wearing head mounted visors.
DISCLOSURE: I am employed in the biometrics industry.
@msbpodcast, certainly facial recognition has gotten better over the years, and certainly systems can be scaled to provide very fast responses. And a case could probably be made that a mobile camera, while less accurate, can monitor places that a stationary camera cannot monitor.
However, there was one thing in the original article that gave me pause:
“The system can compare biometric data at 46,000 points on a face….”
Huh?
As you know, biometric comparisons are usually made with far fewer points of comparison (often less than 100, sometimes much less than 100). Yet that same number appears later in the article:
“Major Leandro Pavani Agostini, of Sao Paulo’s Military Police, said: ‘…To the naked eye two people may appear identical but with 46,000 points compared, the data will not be beaten.'”
Unfortunately the vendor was not mentioned, but I am not aware of any vendor that claims to use tens of thousands of comparison points on a face, or on any biometric specimen (DNA excluded).
See the link to this comment for a link to an earlier Brazilian article which appears to be the source for the English-language articles. This original article also includes the “46,000 points” claim, but only says that an Israeli company is involved.
The ctrueltd.com website talks of its technology by referring to “[f]ace acquisition at exceptional quality, greater 1,000,000 points per scan.” However, this is probably a reference to pixels (e.g. 1000 x 1000), rather than feature comparison points.
Update – I am a member of the Biometrics group on Yahoo!, and I posed my question there. Dr. James Wayman posted a thorough response, and while I won’t attempt to reproduce the whole thing here, Dr. Wayman did say that one way to perform facial recognition calculations is via vector analysis in which a single vector may consist of the pixels from the image. Thus, it is correct to state that it is possible to use 46,000 “features” in facial recognition comparisons. Dr. Wayman referred to the definition of “feature” in ISO/IEC SC37 N3971, and although I haven’t reviewed that document, I trust Dr. Wayman on this one…
I suspect this article is mainly a rewritten press release from the company or group of companies trying to sell their systems to security organizations. While many skeptical posters here seem to have forgotten the power of Moore’s Law, as mbspodcast pointed out, the numbers quoted (400 faces per second, 50 meters, 12 miles) are probably exaggerations of best-possible-case scenarios.
On the other hand, European police, including Interpol, have been using systems like this to identify and foil soccer hooligans at the games before they can cause any trouble for more than a decade now. So I believe the central idea of the story is true, but the numbers have been exaggerated to push sales.
#6, #13, I won’t claim to be an expert on facial recognition. So I’ll defer to your expertise.
But I do know a bit about databases. They are getting very fast. Scanning a few million records in a few seconds is possible. But a single face doesn’t take up a single record. It will consist of thousands of data points, all of which have to be scanned and matched.
I’m also very skeptical of the results. Fingerprint matching technology is most effective (for police) if they are scanning within a subset of the population. i.e. if they are looking for a fingerprint match on the break-and-enter, then they scan the criminal database of known B&E criminals. It’s quicker and it reduces the number of false-positives. If they had a database of fingerprints for everyone in the USA (and I’m sure Homeland Security would like that) they’d get literally 1000s of false-positive results, which would have to be checked by actual human beings.
Faces are even harder. White people all look alike. (ha!) Many of the facial datapoints can be fooled by people simply smiling or wearing sunglasses. Imagine a sports event where many fans decide to paint their faces to match the team colors.
2 more thoughts:
13 million faces x 46,000 data points, that’s 598 billion records. Is there a database system that exists which can handle that many records?
Also, wouldn’t they need to collect the 46,000 data points from the 13 million people first in order to create the database? So the next time you’re at the DMV they ask you to stick you face in a scanner which can map your face from 46,000 points?
I’m sorry but some of this has to be BS. To get a clear enough image to recognize a face at 12 miles is going to take better lenses than you can easily transport as part of a helmet.
#24–chuck==like me, do you get a warm feeling when opining on areas you have total lack of expertise in?
I have a vague suspicion that a google search is very different from a criminal face match system even though I assume there are also many similarities?
I trust msbpodcasts opinion on the issue, until an equally sounding expert opinion comes in on the other side and we are thrown into another global warming dispute.
Reminds me of what my Top Sarge used to say whenever he could: “Every swinging dick has an opinion.” Still cracks me up.
What’s even funnier is this is coming to America.
TSA was testifying to Congress about this last month.
I watched the House Committee Chair get confused by the concept of false positives.
Guys, keep pretending it’s not coming to America.
Blue pills for all, and for all a good time!
#27 – you know I do get such a good warm feeling spouting off on almost any subject.
Most of the time I just sit around waiting for idiots to close the pie-holes long enough so that I can explain just how stupid they are.
I consider it a public service.
And if, at any point, you say something I agree with (never gonna happen) I’ll appreciate your contribution too.
I trust msbpodcasts opinion on the issue, until an equally sounding expert opinion comes in on the other side and we are thrown into another global warming dispute.