Cameron Fevig, Tyler Heister, Aayush Jhunjhnuwala, Sean Mince, Ayushi Pradhan, Mark Shimala, Kevin Jones, Ben Petry, Steve Elliott, and Kevin Chan EXAMINING INTRA-VISIT IRIS STABILITY (VISIT 3)
Jul 15, 2015
Cameron Fevig, Tyler Heister, Aayush Jhunjhnuwala,
Sean Mince, Ayushi Pradhan, Mark Shimala, Kevin
Jones, Ben Petry, Steve Elliott, and Kevin Chan
EXAMINING INTRA-VISIT
IRIS STABILITY (VISIT 3)
•Three main ways:
• Tokens – Drivers license, passport, ID card
• Secret Knowledge – PIN, password
• Biometrics
HOW TO IDENTIFY A PERSON
•Biometrics: “a measurable, physical
characteristic or biological characteristic used
to recognize the identity or verify these claimed
identity of an enrollee” [1]
BIOMETRICS – WHAT IS IT?
•There are many instance where we need to
identify an individual before allowing them to
gain access to something.
•For example – confirming identity before
crossing a country border
BIOMETRICS – WHY CARE?
• Physiological features:
• Face
• Fingerprint
• Retinas
• Iris
• Behavioral traits:
• Voice
• Signature
• Keystroke Dynamics
BIOMETRICS – MODALITIES
•The iris is the colored part of the eye, located
between white sclera and black pupil
•Acts as the muscle that controls the light levels
allowed inside the eye [2]
WHAT IS THE IRIS?
• The chances of 2 people having matching iris’ is about 1 in 10^78
• It is an internal organ that is well protected and externally visible
• Can be captured from a distance while subjects are moving
• Can be quickly matched to templates stored in a database [2]
IRIS RECOGNITION – WHY USE IT?
•Template aging refers to technical deterioration
of saved iris template
• Iris aging refers to actual changing of physical
iris over time
TEMPLATE VS IRIS AGING
•Since Daugman states that the iris is stable,
aging should not affect iris recognition
performance [2]
STABILITY - PERFORMANCE
•Does a study of iris performance scores exhibit
statistical stability when examing a time period
of 10 or fewer minutes?
RESEARCH QUESTION
• Type or class of biometric system
• Any biological or behavioral characteristic that can be measured.
• Face
• Fingerprint
• Hand Geometry
• Keystroke Dynamics
• Etc.
MODALITIES
• Arching ligaments
• Furrows
• Ridges
• Crypts
• Rings
• Corona
• Freckles
• Zigzag collarette [2]
IRIS CHARACTERISTICS
• Formula which creates an Euclidian distance between any two points of a zoo menagerie
𝑆𝑆𝐼𝑖 =
(𝑥𝑖2 − 𝑥𝑖1)2 + (𝑦𝑖2 − 𝑦𝑖1)
2
(𝑥𝑚𝑎𝑥 − 𝑥𝑚𝑖𝑛)2 + (𝑦𝑚𝑎𝑥 − 𝑦𝑚𝑖𝑛)
2
STABILITY SCORE INDEX (SSI) [3]
•Examining images collected from different
visits from a four year study concluded that the
hamming distance varies significantly over a
larger period of time compared to a shorter
period. [4]
OTHER WORKS
• Datarun collection
• Grouping creation
• SID assignment
• Score generation
• Menagerie creation
• SSI calculation
METHODOLOGY
DATARUNS
•Collected dataruns
•Note: Subjects in dataruns were present for all
visitsDatarunID Visit Start Date End Date Days
821 1 6/11/2012 4/25/2013 318
822 2 3/29/2013 4/29/2013 31
823 3 4/11/2013 5/9/2013 48
824 4 4/22/2013 5/29/2013 37
825 5 4/26/2013 6/5/2013 40
826 6 5/6/2013 6/12/2013 37
827 7 5/14/2013 6/18/2013 35
828 8 5/28/2013 6/18/2013 21
EXAMINE DATARUNS IN DEPTH
• Identified any errors for each subject
• 821 - Different collection methodology from the other data runs and a much longer collection period.
• 822 - Many subjects post data cleaning resulted in the loss of several images. Therefore, these subjects did not have the 12 minimum required images. Cleaned data to remove subjects that did not have at least 12 images for one visit.
GROUPING CREATION FOR EACH IRIS,
EACH VISIT
•Created groupings for each iris for each visit
•Sorted images into groupings for first three left
and first three right images for the same
subject
GROUPINGS SPLIT INTO THEIR OWN
DATARUNS
•Split groupings into their own dataruns
•Took each grouping and split into individual
excel files
•Datasets were created for each grouping for
each visit
•Grouping 1 (1-2, 1-3, 1-4)
•Grouping 2 (2-1, 2-3, 2-4)
•Grouping 3 (3-1, 3-2, 3-4)
•Grouping 4 (4-1, 4-2, 4-3)
RESULTS – VISIT 3
The following information applies to the upcoming results:
H0 = the median stability scores are equal
Hα = the median stability scores are not equal
α = 0.05
RESULTS
•There was not a statistically significant
difference between the median of the
groupings (H(2) = 2.78, p = 0.249), with a
mean rank of 94.3 for grouping 1-2, 95.8 for
grouping 1-3, and 81.4 for grouping 1-4.
GROUPING 1 RESULTS
•There was not a statistically significant
difference between the median of the
groupings (H(2) = 1.28, p = 0.526), with a
mean rank of 95.2 for grouping 2-1, 91.7 for
grouping 2-3, and 84.6 for grouping 2-4.
GROUPING 2 RESULTS
•There was not a statistically significant
difference between the median of the
groupings (H(2) = 0.51, p = 0.774), with a
mean rank of 93.9 for grouping 3-1, 87.1 for
grouping 3-2, and 90.5 for grouping 3-4.
GROUPING 3 RESULTS
•There was not a statistically significant
difference between the median of the
groupings (H(2) = 2.28, p = 0.320), with a
mean rank of 86.9 for grouping 4-1, 85.9 for
grouping 4-2, and 98.8 for grouping 4-3.
GROUPING 4 RESULTS
VISIT 3 N H DF P
Group 1 60 2.78 2 0.249
Group 2 60 1.28 2 0.526
Group 3 60 0.51 2 0.774
Group 4 60 2.28 2 0.320
RESULTS
There was not a statistically significant difference between the median of
the groupings, as indicated in the summary table. For this data, we can
conclude that the iris is stable in this visit.
• There was not a statistically significant difference between the median of all of the groupings.
• The P-value was greater than the α- value of 0.5 for all the groupings, hence we fail to reject the null hypothesis which stated that the median stability scores are equal.
CONCLUSION
CONTRIBUTION AND FUTURE WORK
• Through this study we have been able to show that the Iris is a stable modality when images were collected within one visit
• There are many ways to expand on this research
• Replicate the research with different matcher and/or collection device to determine match/collection device effects on ‘aging’
• Do a similar study but with different collection periods e.g. months or years apart
[1] Association of Biometrics, 1999, p. 2
[2] J. Daugman, “How iris recognition works,” IEEE Trans. Circuits Syst. Video Technol., vol. 14, no. 1, pp. 21–30, Jan. 2004.
[3] O’Connor, K. J. (2013). Examination of stability in fingerprint recognition across force levels. Purdue University, West Lafayette, Indiana.
[4] S. P. Fenker, K. W. Bowyer, “Experimental Evidence of a Template Aging Effect in Iris Biometrics University of Notre Dame University of Notre Dame,” pp. 232–239, 2010.
REFERENCES