Free Newsletter
Register for our Free Newsletters
Access Control
Deutsche Zone (German Zone)
Education, Training and Professional Services
Government Programmes
Guarding, Equipment and Enforcement
Industrial Computing Security
IT Security
Physical Security
View All
Other Carouselweb publications
Carousel Web
Defense File
New Materials
Pro Health Zone
Pro Manufacturing Zone
Pro Security Zone
Web Lec
Editor's Blog and Industry Comments

3D face recognition technology goes beyond border protection biometrics

16 November, 2007
At the British Technion Society's seminar in London, leading biometrics professor explains the state of the art in 3D biometric facial recognition and how it can, with sufficient funding, be with us in 2-5 years in applications encompassing border control, medical technology and even an alternative to keys.
The Technion (Israel's Institute of Technology) is at the cutting edge of numerous anti-terrorist technologies. Yesterday in London at a seminar run by the British Technion Society, Technion based in Haifa, Israel, allowed the brains behind their technology prowess to show a UK audience what they are working on.

The seminar is part of a move to set up collaborative projects with various UK organizations. Four of Technion's 40 key R&D programmes were highlighted.

Although the seminar presented a brief snapshot of what Technion has to offer it also underlined some of the key technical challenges that terrorists are posing the security community today. These included:

- 3D face recognition systems
- Peroxide-based explosives.
- Interpreting monitoring and surveillance video.
- Urban Search and Rescue Applications

Perhaps the most common theme of the seminar was that security is far from a stand-alone subject for Technion's research teams.

Security and health are often two of the common concerns for most people. So it seems appropriate that these twin aspects of everyday life are responsible for driving the commercial impetus of many of the anti-terrorist technologies being developed.

Dual use was a striking motif of the presentations and it was clear that without addressing twin application markets many of the research projects could fail to develop into real life solutions due to lack of investment.

Perhaps the best example of this research duality was underscored by Prof Ron Kimmel of Technion's Computer Science Department who focused on the subject of 'Biometry and Isometry â€" 3D Face Recognition Systems'.

Prof Kimmel's research collaborators at Technion are in fact twins, Alexander and Michael Bronstein, who have been used as willing guinea pigs to test the performance of the 3D Face Recognition technology they are developing. The reasoning was that if the software algorithms can differentiate between the twins then it should also be able to spot the difference between different people.

The team have come a long way in a short time. In the early days, biometrics suffered from the 'curse of expressions'. Even an innocent smile could make fools of the biometric brigade.

Prof Kimmel's team has now shown that applying isometrics can overcome the curse. They revealed that expressions equate to extrinsic (Euclidean) geometry while identity, it seems, equates to intrinsic (Riemannian) geometry.

The researchers have shown that each face has its own geometric 'space'. Their solution was to find a space that better captures a face's intrinsic geometry yet still provides a convenient comparison property of a flat embedding space. The result, which uses spherical embedding techniques together with generalized multidimensional scaling, GMDS, offers expression-invariant face recognition using intrinsic geometry.

On the question of "Does your face represent your face over time?" Prof Kimmel said that there are several invariants of a face that do not change with time.

"My vision is that any changes in an individual's face will be added incrementally to the database over time to reflect the facial changes. It should be relatively simple to update the facial data with regular updates over time."

Storing the 3D Face Recognition data doesn't seem to be a limiting factor now either. The Technion system's use of GMDS enables it to recognise a face by capturing only 500 bytes of data.

Prof Kimmel made it clear that he believed that 3D Face Recognition technology was heading out of the research phase and was firmly heading down the development path.

As he said: "Research makes knowledge out of the data we collect while development takes that knowledge and makes money out of it."

Prof Kimmel pointed out that the independent Face Recognition Vendor Test (FRVT) that assessed the various face recognition technologies in 2006 showed just how far work had progressed towards a commercially viable solution. The FRVT 2006 report concludes that: "The best-performing face recognition algorithms were more accurate than humans.'

The goal of the FRVT 2006 was to measure state-of-the-art prototype systems/algorithms as well as commercial face recognition systems. FRVT 2006 quantified the improvement of the technologies against four key milestones. For each milestone, the false reject rate (FRR) at a false accept rate (FAR) of 0.001 (1 in 1000) was calculated for a representative state-of-the-art algorithm.

The testing showed that in the early days of 3D Face Recognition (circa 1993) the technology of the period could only achieve a rating of 0.79.

By 2006 this figure had dropped to just 0.01 - a massive improvement.

Prof Kimmel told Prosecurityzone: "It is now not a case of whether 3D Face Recognition technology can be effective. There is a belief now that it will work."

The improvement, according to Prof Kimmel, is simply down to two essential factors.

"More pixels and less noise," explained Prof Kimmel. "The speed of the new generation of processors that are being talked about such as Intel's 45nm metal gate silicon technology or the new ATI AMD technologies means that the 3D Face Recognition techniques can now become a practical reality relatively quickly."

Prof Kimmel suggested that in the next two to four years 3D sensors should be capable of computing 3D structures adequately.

According to Prof Kimmel the likelihood of 3D Face Recognition being implemented in real world applications can occur between the next five to 15 years. He clearly believes that lack of investment is now the main limiting factor to progress in this technology area.

"It depends on capital. But if someone is willing to fund the development needed then the timescale could slim down to two to five years. If people want to invest in the technology it will be the shorter of the timeframes. Capital investment can squeeze the development time."

Cash does seem to be a major stumbling block. However, venture capitalists tend to do the calculation as follows: How many airports are there in the world? So how much profit can be made from each biometrics system at these airports? The answer is not in the billions of dollars at first glance and that's where the investors are stopping for now.

"What they don't seem to see is that there is much more scope in the underlying technology concept. I want to see 3D Face Recognition replacing your car key or your house key. Every day items â€" that's where the numbers really work. It is not just about biometric systems in airports."

This is where the health dimension of the research really seems to kick into play. Technion has already helped three start-up companies get off the ground by applying the underlying 3D Face Recognition technology to medical imaging applications.

One of these companies, MediGuide, has recently started the first human clinical trials in collaboration with Siemens Medical Solutions to assess MediGuide's Medical Positioning System (MPS) at the Regensburg University Hospital in Germany.

MediGuide focuses on intra-body navigation technology that provides the capability to guide and track a tool during a medical less-invasive procedure within the human body.

Philips and GE Healthcare have also both recognised the potential of the underlying algorithms used in 3D Face Recognition. These and many other medical organizations are recognising how the technology can be transferred to their areas of expertise.

So it seems that dual use funding of many of these anti-terrorism research projects broadens the funding pool and may be the only way forward if we want to see all the theory turned quickly into commercial solutions.

However, there are now signs that face recognition technology, which claims to be the least intrusive biometric, may be about to go mainstream. Google, for one, has already started to buy up face recognition technology start-ups. The face recognition market in 2004 was worth $144 million it is expected to be worth more than $800m in 2008.

Bookmark and Share