Naked Science Forum
Non Life Sciences => Geek Speak => Topic started by: Karen W. on 10/02/2008 15:17:37
-
My son works for an Indian Casino here in humboldt county. He is a trained of their security Guard.
The Casino employs a computer program which identifies people who come in by taking a picture of them
with security cameras, The program then files the picture and stores info on that person..If that
person is a regular member and has a photo on file, it matches that person to the file photo with
all the pertinent info gathered from previous visits, ie were they a trouble maker or just a good
customer etc. The camera pulls up those who enter the building, then checks it with previous records
of people entering, Records new piccy's and rules out weather they have in the past
basically been trouble of some kind or another and then a security guard is placed in a manner to
keep an eye on them or to remove them immediately.
The cameras do the Identifying, without human assistance.The program is turned on and if it recognizes
a trouble maker or banned person it immediaely pulls the file and head security is then buzzed and
guards are sent My question is How long will it be before this kind of security is being used
regularly by government agencies or police departments, or is it already? I would think Casinos would
be low on the pole in instigating these measures, so I would assume that many places other then
Casinos already employ such software in their businesses! Is this already common practice or will it
be?
-
That type of face recognition software is already widely used.
-
I did not know that.. that sure makes things easier doesn't it? Do you guys use it there also!
-
So do government agencies already use it too?
-
The police use it, airport security uses it.
It is not 100% reliable (including sometimes giving false positives, where is mistakes an innocent person for someone on file), but it does make automated operation (e.g. tracking people using CCTV) possible with the minimum of manpower.
-
A lot of the work on face recognition was done at Brunel University which is where I got my BSc in computer science.
-
That is very cool Doc.. It is amazing to me how that works..!
-
A lot of the latest generation of face recognition software can also tell whether you're smiling, frowning etc.
-
Thats pretty amazing. I wonder how much it is connected to these software programs that take a lost child's face(on a picture and can age it two years or ten and using the basic facial structure and muscle tone can present a mock up of what that child would look like updated? I know these just match compare etc.. but to be able to distinguish between a smile or frown.. amazing!
-
I wonder how much it is connected to these software programs that take a lost child's face(on a picture and can age it two years or ten and using the basic facial structure and muscle ton can present a mock up of what that child would look like updated? I know these just match compare etc..
That's a very different thing.
-
I was wondering if it was as it is able to identify a certain facial structure.. or am I mistaken.. How does that work Doc. How the heck does software recognize a face? Is it similar to fingerprint scanning.. where it takes a picture of the print and compares it to prints in a larger data base? How do they do it?
-
It's complicated - and bear in mind that my knowledge of the workings of this type of software may be well out of date by now.
It maps many points on the face (eyes, tip of nose, corners of mouth, etc) and calculates ratios between them. The more points it maps, the more accurate the recognition. Depth of shadow is another parameter that is measured.
-
Thats interesting.so it realy does use numbers calculations etc. for measurements. how do they measure depyh of shadows? What does that mean?
-
Size & blackness
-
Ok.. that is easy. .. lol !!ere I thought it was more complicated!
-
Karen - I thought you might be interested in this article from The Times Online
The rise of the digital camera has lead to a huge increase in photography, but has also consigned many pictures to the depths of a hard disk.
Part of the problem is that with so many photographs on file, it’s hard to find what you’re looking for. Filenames rarely give any indication of the subject matter, and image thumbnails may be of little help if you can’t remember which folder contains the image you’re trying to track down.
That’s where Facebubble, an application developed by HP Labs, comes in. Its face-recognition software analyses all the photographs stored on a PC and sorts them into groups according to who appears in each image. It will also allow you to find subsets of pictures that contain two or more people.
Demonstrated on a touchscreen PC, the program is slick and intuitive. It displays all of the faces it finds within the photographs in bubbles, with the size of each bubble determined by the number of photos in which the person appears. The bigger the bubble, the more that person has been photographed.
Touching another face brings up a second set of bubbles, showing all the pictures in which both this second person and the first appear.
On its own the application would provide a powerful tool for organising photographs and allowing people to find what they’re looking for, and combined with Facebook or other social networks it could create a huge and automatically updated network of images.
The application is not yet perfect – it may consider a baby in a hat to be a different person from the baby without its hat, for example – but researchers are working on solving these problems. They also say that they are hoping to develop the software so that it can recognise photographs of a person taken over a decade or more, despite the inevitable signs of ageing that emerge over time.
Facebubble is not yet publically available, but researchers said that a finished product was not far off.