Iris Presentation Attack Detection Based On Photometric Stereo Features
Tech ID: 19-013
Inventors: Dr. Adam Czajka, Dr. Kevin Bowyer, and Zhaoyuan Fang
Date Added: August 22, 2019
A software-based method for detection of textured contact lenses by iris recognition systems.
Iris recognition systems can be susceptible to presentation attacks, where an individual uses cosmetic contact lenses to conceal their identity or impersonate the identity of another. The performance of current commercial recognition systems is significantly lowered when presented with a textured contact lens compared to a normal iris.
University of Notre Dame researchers have developed a new algorithm for detecting iris presentation attacks using a photometric stereo method. Importantly, this method uses only the existing hardware found in commercial iris recognition systems and thus can be easily implemented. The algorithm analyzes the iris pictures captured by iris recognition systems, with a focus on the eye’s three-dimensional features. The surface of a normal iris is flat while the surface of an iris with a cosmetic contact lens is concave, so their shadows, when illuminated, are different. The photometric stereo method examines the shadow of the iris to detect whether a cosmetic contact lens is present.
- Software-based solution for improving iris recognition systems
- No additional hardware required for implementation
- Photometric stereo method has better generalization capabilities than other methods for use with all contact brands and patterns
Technology Readiness Status
TRL 3 - Proof of Concept
US Provisional 62/749,864 (Method of Ocular Contact Lens Detection)
Iris Presentation Attack Detection Based on Photometric Stereo Features