Houston (TX) – A research team at University of Houston believes that a new, reliable approach to face recognition is able to tackle the growing threat of ID theft. The group of the school’s Computational Biomedicine Lab (CBL) has come up with a complex technology that promises to correctly detect more than 99% of individuals.
The threat of ID theft is very real these days and has become a part of our life. Whenever we have to represent ourselves, there is some form of identification involved, mostly in number codes, pictures, access names and passwords. Most of these representation types can be manipulated with a certain effort, which requires individuals to keep a close eye on where and how their ID data is used.
Biometric types of identification are becoming more and more available, such as fingerprint sensors on notebooks that take over the function of filling out forms that require access names and passwords. There are also more sophisticated biometric technologies, such as face recognition, which however have been plagued by low detection rates – mostly due to the fact that the appearance of a face can change quickly – for example by changing a hairstyle or temporary changes such as sunburns or band-aids.
The Houston CBL research team around professor Ioannis Kakadiaris believes to have found a critical key to dramatically elevate the accuracy of face recognition. Their technology, consisting of 3D sensors, different cameras as well as the “URxD” face recognition software takes a 3D snapshot of an individual to create a unique biometric identifier. The group claims that first test have shown that the technology is able to correctly identify 99.6% of entries with a false accept rate of 0.001.
While the identification process is as effortless as taking a photograph, according to Kakadiaris, there is a fairly complex and expensive technology behind the approach: Using a 3D as well as a infrared camera, a 3D mesh is overlaid with texture, infrared and, in future, time data. The captured data represents the biometric signature of an individual. Both in verification and identification processes, data captured in real time can be compared against available data in a database containing signatures. According to Kakadiaris, URxD already has shown its capabilities in government testing and he believes that the technology could be used for everything from gaining access to secure facilities to authorizing credit card purchases.
A key technology where CBL’s approach is different from current technologies is the way how the infrared camera is used. Kakadiaris told us that the results gained from this device help the technology to sort out foreign objects such as glasses, hats or band-aids. “To fool the system, you will need to undergo plastic surgery,” he said.
At this time, the system demonstrated is a high-end solution that is quite expensive and cannot make it into consumer applications as a result. The 3D image sensor required for URxD alone carries a price tag north of $20,000. However, Kakadiaris believes that his technology is suited for consumer applications and could be adopted in coming years when 3D sensors become more affordable and are scaled down to the size of a microphone chip. Processing power available in today’s computers is already sufficient for the system: According to the scientist, the data processing necessary to identify a face is done “in a matter of seconds.”