Amanda Sgroi

A Multialgorithm Analysis of Three Iris Biometric Sensors

The issue of interoperability between iris sensors is an important topic in large-scale and long-term applications of iris biometric systems. This work compares three commercially available iris sensors and three iris matching systems and investigates the impact of cross-sensor matching on system performance in comparison to single-sensor performance. Several factors which may impact single-sensor and cross-sensor performance are analyzed, including changes in the acquisition environment and differences in dilation ratio between iris images. The sensors are evaluated using three different iris matching algorithms, and conclusions are drawn regarding the interaction between the sensors and the matching algorithm in both the cross-sensor and single-sensor scenarios. Finally, the relative performances of the three sensors are compared.

A cross-sensor evaluation of three commercial iris cameras for iris biometrics

As iris biometrics increasingly becomes a large-scale application, the issue of interoperability between iris sensors becomes an important topic of research. This work presents experiments which compare three commercially available iris sensors and investigates the impact of cross-sensor matching on system performance. The sensors are evaluated using three different iris matching algorithms, and conclusions are drawn regarding the interaction between the sensors and the matching algorithm in a cross-sensor scenario.

The prediction of old and young subjects from iris texture

Researchers have previously studied the prediction of “soft biometric” attributes such as gender and ethnicity from iris texture images. We present the results of an initial study to predict the relative age of a person from such images. We conclude that is possible to categorize iris images as representing a young or older person at levels of accuracy statistically significantly greater than random chance. This suggests that there may in fact be age-related information available in the iris texture, and motivates further study of this topic.

Trial Somaliland voting register de-duplication using iris recognition

Face and fingerprint were used in de-duplication of the voter registration list for the 2010 Somaliland presidential election. Iris recognition was evaluated as a possible more powerful means of de-duplication of the voting register for the planned 2015 elections. On a trial dataset of 1,062 registration records, all instances of duplicate registration were detected and zero non-duplicates were falsely classified as duplicates, indicating the power of iris recognition for voting register de-duplication. All but a tiny fraction of the cases were classified by automatic matching, and the remaining cases were classified by forensic iris matching. Images in this dataset reveal the existence of unusual eye conditions that consistently cause false-non-match results. Examples are shown and discussed.

SNoW: Understanding the causes of strong, neutral, and weak face impostor pairs

The Strong, Neutral, or Weak Face Impostor Pairs problem was generated to explore the causes and impact of impostor face pairs that span varying strengths of scores. We develop three partitions within the impostor distribution for a given algorithm. The Strong partition contains image pairs that are easy to categorize as impostors. The Neutral partition contains image pairs that are less easily categorized as impostors. The Weak partition contains image pairs that are likely to cause false positives. Three algorithms, and the fusion of their scores, were used to analyze the performance of these three partitions using the same set of authentic scores employed in the Face Recognition Vendor Test (FRVT) 2006 Challenge Dataset. The results of these experiments provide evidence that varying degrees of impostor scores impact the overall performance and thus the underlying causes of weak impostor pairs are worthy of further exploration.

The impact of diffuse illumination on iris recognition

Iris illumination typically causes specular highlighting both within the pupil and iris. This lighting variation is intended to be masked in the preprocessing stage. By removing or reducing these specular highlights, it is thought that a more accurate template could be made, improving the matching results. In an attempt to reduce these specular highlights we propose a diffuse illumination system. To determine if iris recognition performance is enhanced by this diffuse illumination system, we examine whether specular highlights were reduced within the pupil and iris, as well as analyze matching results obtained by several iris algorithms.

Iris recognition technology evaluated for voter registration in Somaliland

The Somaliland government web site includes the slogan: ‘Recognition – The number one priority for the Somaliland government’. Part of the effort to achieve international recognition includes holding elections that are respected as fair and that enable peaceful, political transitions. This article summarises the role that biometrics – face, finger, and iris – is playing in creating voter registration lists in Somaliland. Specifically, this article describes Somalilands efforts to move to iris recognition in an effort to create a voter registration list for the upcoming elections scheduled in 2015.

Location matters: A study of the effects of environment on facial recognition for biometric security

The term “in the wild” has become wildly popular in face recognition research. The term refers generally to use of datasets that are somehow less controlled or more realistic. In this work, we consider how face recognition accuracy varies according to the composition of the dataset on which the decision threshold is learned and the dataset on which performance is then measured. We identify different acquisition locations in the FRVT 2006 dataset, examine face recognition accuracy for within-environment image matching and cross-environment image matching, and suggest a way to improve biometric systems that encounter images taken in multiple locations. We find that false non-matches are more likely to occur when the gallery and probe images are acquired in different locations, and that false matches are more likely when the gallery and probe images were acquired in the same location. These results show that measurements of face recognition accuracy are dependent on environment.

Strong, Neutral, or Weak: Exploring the Impostor Score Distribution

The strong, neutral, or weak (SNoW) face impostor pairs problem is intended to explore the causes and impact of impostor face pairs that are inherently strong (easily recognized as nonmatches) or weak (possible false matches). The SNoW technique develops three partitions within the impostor score distribution of a given data set. Results provide evidence that varying degrees of impostor scores impact the overall performance of a face recognition system. This paper extends our earlier work to incorporate improvements regarding outlier detection for partitioning, explores the SNoW concept for the additional modalities of fingerprint and iris, and presents methods for how to begin to reveal the causes of weak impostor pairs. We also show a clear operational difference between strong and weak comparisons as well as identify partition stability across multiple algorithms.

Metadata-based understanding of impostor pair score variations

Metadata about a given face image can include information such as the subjects year of birth, subject gender, and date of acquisition. By determining the degree of metadata matches between the gallery and probe images (such as two subjects having the same gender) we hypothesize that more metadata values that match for an impostor image pair increases the likelihood of a false match. In this work, we explore year of birth, gender, date of acquisition, and expression in an attempt to understand variations in match scores produced by impostor image pairs. Impostor pairs that fall in the weak partition as identified by the previously developed SNoW technique have a slightly larger number of matching metadata values. However, there is little to no statistically significant difference in the scores produced by image pairs with more matching metadata between strong and weak impostor pairs.