| The iris is a popular biometric that has been demonstrated to have high matching accuracy and permanence under ideal conditions. However, there are several limiting factors that have yet to be comprehensively addressed. One major consideration in iris recognition systems is the restrictions imposed by the optical parameters on the depth of field. This thesis examines issues related to the depth of field and evaluates a solution proposed by CDM Optics RTM. The solution explored involves the utilization of a Wavefront Coded(TM) element. By employing a simulation of a Wavefront Coded(TM) element, developed by Wake Forest University this thesis shows the improvement in matching performance over an increased depth of field on a database of 150 iris images. Anomalies in the average performance of the data are further explored to define the cause of the outlying behavior. A second issue associated with iris recognition systems is the effect of various levels of iris pigmentation on match performance. This thesis explores the difference in match performance observed on a dataset of images partitioned by eye color (blue and brown) on a baseline algorithm. The statistical significance of this difference in performance is established by a permutation test. A third important issue associated with all biometric systems is the scope of the law as it pertains to the implementation of the system. This thesis includes a review of current privacy laws that may have interpretations applicable to the field of biometrics. The fourth and final issue addressed in this thesis is the concern over eye safety when interacting with the acquisition system. Since iris recognition systems invariably require some form of external illumination, it is important that concern is taken to ensure the safe illumination of the subject's eye. In this regard, a review of the current Threshold Limit ValuesRTM defined by the American Conference of Governmental Industrial Hygienists (ACGIHRTM) is presented. |
