Dissertation Defense - Feng Shen
|Start:||7/31/2014 at 9:00AM|
|End:||7/31/2014 at 1:00PM|
|Location:||315 Stinson Remick|
July 31, 2014, 9:00am, 315 Stinson Remick
Dr. Patrick Flynn, Advisor
Dr. Jun Li Dr. Gregory Madey Dr. Laurel Riek
A Visually Interpretable Iris Recognition System with Crypt Features
One important barrier to the use of current iris recognition techniques in law enforcement areas such as forensics is that the iris features used by these techniques are not interpretable to human eyes. This dissertation proposes a novel iris recognition technique that determines the human identity based on the crypt, a visible feature on the iris. The major motivation is to aid the law enforcement applications that require human judgment in the process.
The crypt is a visible feature whose formation is related to both color pigmentation and iris surface structures. Thus they are stable and easier for human eyes to rely on for iris recognition tasks. To verify the human-in-the-loop system design, we carried out two human-subject experiments to evaluate the human perception of crypt features in both consistency and robustness. The results support the applicability of using crypt features for semi-automatic iris recognition.
Automatic feature detector and matcher are implemented to reduce the human workload to a manageable level for large datasets and assist human in making more informative decisions. Our feature detector deploys a multi-scale pyramid framework in order to detect crypts of various sizes. An optimization step is then carried out to improve boundary accuracies and reveal more shape details. Two crypt matchers are introduced and evaluated in this dissertation. The first matcher uses a two-stage design that approximates the overall speed of of a simple global matcher while still being able to locate common crypts between potentially matching iris image pairs. The second matcher uses local binary compatibility to evaluate crypt similarities. It is designed to be more robust to shape variations and partial detection errors.
The performance of the crypt detector and matcher are evaluated in both identification and verification scenarios. The testing dataset is composed of 3505 images from 701 irises. It is by far the largest dataset that has been used in performance evaluation for visible feature based iris recognition techniques.
A final contribution of this work is the preliminary study of a semi-automatic identity verification system. This system combines both the automatic crypt detector / matcher and the human judgment. Image pairs with automatically annotated crypt features and similarity evaluations are presented to a group of anonymous online human participants to classify each iris pair as match or non-match. The results suggest that the semi-automatic identity verification framework is feasible when the conclusion is retrieved through a voting strategy among several humans. We also expect the accuracy to be considerably higher for experienced forensic examiners