Iris recognition – detecting cosmetic contact lenses…
23 October 2013 21:24 GMT

Lens or no lens - will your biometric system be able to tell?

Question. Can iris recognition systems detect textured cosmetic contact lenses? Answer. Yes…and no. A new research paper presented by biometric experts at the University of Notre Dame suggests that while detection rates can reach as high as 90% for the most common types of lens, other lenses have very poor detection rates.

This may be worrying as textured contact lens detection is an important problem to solve in preventing spoofing in iris recognition systems. Another more likely problem is that regular soft lenses have also been shown to negatively impact the performance of an iris biometrics system – and a first step in correcting this lower true accept rate is by determining which subjects are wearing soft contact lenses.

Kevin W. Bowyer at the University’s Department of Computer Science and Engineering told Planet Biometrics: “Detection may be very sensitive to the brand of lenses. That is, if you train on Johnson+Johnson brand cosmetic lenses, then yes, you can detect them very accurately.  But if people show up wearing some other brand of cosmetic lenses, you may be very poor at detecting them.”

This is important because cosmetic lenses are made by various companies all around the world, and ideally iris recognition systems should be good at detecting all brands.

The researchers presented their paper at the BTAS conference. The paper notes that automatic detection of textured contact lenses in images acquired for iris recognition has been studied by several researchers. However, to date, the experimental results in this area have all been based on the same manufacturer of contact lenses being represented in both the training data and the test data and only one previous work has considered images from more than one iris sensor.

Experimental results in this work show that accuracy of textured lens detection can drop dramatically

when tested on a manufacturer of lenses not seen in the training data, or when the iris sensor in use varies between the training and test data.

As early as 2003, the inventor of modern iris recognition, Professor John Daugman, proposed using a Fourier transform to detect the highly periodic fake iris pattern that was prevalent in textured lenses manufactured at that time. However, the researchers noted that today newer lenses have multiple layers of printing, making the Fourier response less pronounced, and textured lens detection by this method less reliable. Additionally, not all textured lenses necessarily use a dot-matrix style printing method.

The paper outlines a potential textured lens detection algorithm.