Beware of Problems With Iris Recognition
With iris recognition technology becoming more commonplace, naturally there is much discussion as to the effectiveness and accuracy behind it. This month the US Department of Homeland Security announced it would be testing new iris scanning systems at the US/Mexico border. One company we’ve written about in recent posts has plans to install iris scanning systems throughout the Mexican city of Leon to create – what they say – will be the most secure city in the world.
One paper we found in recent weeks discusses the limitations of iris scanning as well as some of the problems that arise as a result of non-ideal iris imagery. Identity recognition is impacted significantly when scanning images aren’t perfect due to lighting, motion, blur, or even physical problems like occluded irises, etc. We’ve read and written about new iris scanning systems that can apparently scan persons on the go, with movement, low light, from a distance, and in crowds, but it still seems a pretty tall order to guarantee accuracy.
When an iris scanner works as it should, the first step is to acquire the iris imagery, locating and segmenting the iris into readable information. Next the system encodes patterns in texture to create templates, which are then matched across a collected database. Excluding recent improvements we read about, most systems require a good amount of participation from the scanee. The information is captured and recorded and then stored for later cross-reference, but the accuracy and functionality of the system requires a decent quality of image. Quality images can really only be obtained, the paper points out, when the subject is cooperative, and so making scans covertly or in non-cooperative situations can be trouble.
The authors suggest that biometric iris data could be supplemented with periocular data – that is, the information of features surrounding the eye such as eyelids, lashes, eyebrows, and the skin nearby. Capturing this information doesn’t require extra hardware or sensors as the iris camera can do it. By measuring “microfeatures” like scars, marks, moles, and freckles, they can be fused with iris data to provide a more whole scan. It’s an interesting solution to the well-documented problem of not capturing the perfect scan each time.
On the Fusion of Periocular and Iris Biometrics in Non-ideal Imagery