Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Iris Recognition - DOC by KzOLGr

VIEWS: 80 PAGES: 18

									Iris Recognition




    Anders Hur
    COEN 150
     Holliday
      2-22-05
       The attacks of September 11th, 2001 demonstrated the immediate need

for improved security in our society. People around the world realized that what

they thought was safe and secure was in fact vulnerable and in danger. At no

other time in history has there been such a desperate call for security. In the

scramble to implement better security came the realization of new solutions.

One such solution was biometrics.

       Biometrics is the recognition of a person based on physiological or

behavioral characteristics1. The most commonly known biometric is the

fingerprint. It was perceived as a method of identification in the1800’s and has

since been the main biometric used in modern society. Other biometrics include

voice patterns, facial recognition, and retinal and iris scanning. Of these, iris

scanning, also known as iris recognition, has become accepted by many as a

superior method of identification.



History

       One of the first applications of iris analysis was Iridology. Starting in the

1800’s, Iridology was the belief that parts of the body were represented by

corresponding areas in the iris. According to this belief, analysis of the color and

texture of the iris could lead to diagnosis of a person’s health and diseases.

Iridology also implies that the iris changes over time depending on the health of a

person. However, all these beliefs have been proven groundless. Multiple

studies have shown that iridologists cannot accurately diagnose a person’s well-

              1
                 findBiometrics.com. “What are Biometrics?” 18 February 2005
       <http://www.findbiometrics.com/Pages/guide1.html>.
being2. Nevertheless, iridologists still practice and people still consult with them

about health.

       Alphonse Bertillon was the first to study eye color as a means of

identification. Frank Burch expanded this claim at a meeting of the American

Academy of Ophthalmology in 1936. He claimed that the complex patterns in the

iris could be used as an optical fingerprint. This idea became accepted in the

field but was not realized for decades until two ophthalmologists named Leonard

Flom and Aran Safir patented it in 1987. They then went to John Daugman for

help with the algorithm and mathematical theory necessary for automated iris

recognition. In 1994 Daugman patented a method for iris reading, analysis, and

comparison3. These patents are held by Iriscan Inc. which is now Iridian

Technologies. Another major contributor to iris recognition is Richard Wildes.

His method of iris reading, analysis and matching differ from that of Daugman’s.



Anatomy of the Iris

       The iris is the colored part of the eye. The function of the iris is to control

the amount of light that enters the eye. It has tiny muscles that allow it to dilate

and constrict the size of the pupil for light regulation. It is flat and separates the

front and back of the eye4. It is protected in the front of the eye by the cornea.




       2
           Barrett, Stephen. “Iridology is Nonsense.” Quackwatch. 10 September 2004. 18
February 2005 <http://www.quackwatch.org/01QuackeryRelatedTopics/iridology.html>.
         3
           Daugman, John. “Iris Recognition.” August 2004. 18 February 2005
<http://www.securimetrics.com/articles/gfx/Iris_PDF_file.pdf>.
         4
           “Iris.” StLukesEye.com. 18 February 2005
<http://www.stlukeseye.com/anatomy/Iris.asp>.
       The basic structure of the iris is genetically determined. However, the

exact composition of the iris is based on the conditions in the embryo during

development and is thus very unique. In rare cases, the iris develops

abnormally. Some parts of the iris are defined at birth but other parts do not

solidify until about two years of age. In addition, the iris size and coloring

continue to develop until adolescence. The iris changes very little after

adolescence with the exception of depigmentation and a decrease in size of the

pupil in advanced age5.

       It is necessary to delve into the structure of the iris in order to realize its

uniqueness. It is made up of four different layers. The back layer is heavily

pigmented and makes the iris opaque so that light only reaches the eye through

the pupil. The next layer out contains the sphincter and dilator muscles that

allow for constriction and dilation. The third layer is the stromal layer which,

“consists of collagenous connective tissue in arch-like processes6.” In this layer

are corkscrew-like blood vessels that span out radially. The exterior layer is

called the “anterior border layer” and is denser than the previous layer with more

pigmentation.

       The color of the iris is created by different levels of light absorption in the

anterior border layer. Little pigmentation in this layer results in a blue

appearance because light reflects from the back layer of the iris. The more




       5
         Wildes, Richard P. “Iris Recognition: An Emerging Biometric Technology.” 18 February
2005 <http://www.cs.yorku.ca/~wildes/wildesPIEEE1997.pdf>, 1349.
       6
         Ibid. 1348.
pigmentation a person has in the anterior border layer, the darker their iris is.

The overall visual appearance of the iris is due to its multi-layered structure7.



Why Iris Recognition?

        There are many different biometrics to use for identification. What makes

iris recognition better than other biometrics? To begin with, the iris is an internal

organ of the eye. This means that the iris is protected from the external

environment because it is situated behind the cornea and aqueous humor.

Although the iris is an internal organ, it is still externally visible and can be read at

reasonable distances.

        The irises of a person are also very unique. The basic structure and color

of the iris is determined genetically, but the minute details of the iris are created

based on conditions in the embryo8. The chances of similar irises are 1 in 1078 9.

Even identical twins have different irises because of the randomness of creation

in the embryo10. In fact, the left and right irises of the same person are different.

        Iris patterns have also been shown to be stable for the life of an individual

(the exceptions being preadolescence and advanced age). This means that only

one initial enrollment reading needs to be taken. Thereafter, no more readings

are necessary because of the stability of the iris. This eliminates extra time spent

updating that is often required for other biometrics.

        7
           Ibid. 1348.
        8
           Ibid. 1349.
         9
           LG Electronics. “How it Compares.” Iris Technology Division. 18 February 2005
<http://www.lgiris.com/iris/compares.html>.
         10
            Daugman, John. “Genetic penetrance and iris recognition.” 18 February 2005
<http://www.cl.cam.ac.uk/users/jgd1000/genetics.html>.
       The nature of the iris also helps in detecting frauds. It is impossible to

surgically modify the iris without unacceptable risk to vision because it is located

behind the cornea and aqueous humor. The iris’s physiological response to light

also prevents the use of paper imitations or even removed eyes11. A simple

addition or reduction of light to the eye can determine if an iris is living or not.

These two characteristics of the eye and iris make iris recognition very hard to

falsify or change.



How Iris Recognition Works

       The basic procedure for iris recognition is the same for most products.

The first phase is enrollment. A person is taken to a controlled and monitored

environment in which their identity is manually confirmed. Once their identity is

confirmed one or both of their irises are recorded, depending on the system

used. During the image acquisition stage irises are recorded using black and

white video cameras (they are not “scanned” as the common term “iris scan”

suggests).

       The recorded image then goes through iris localization. In this stage a

computer takes the recorded image and filters out everything but the iris. The

localization process is done through complex algorithms which eliminates

eyebrows, eyelashes, and the sclera (the white part of the eyes). Once the

pictures are localized they are stored in a binary format.



       11
         Daugman, John. “High Confidence Visual Recognition of Persons by Test of Statistical
Independence.” 18 February 2005 <http://www.cl.cam.ac.uk/users/jgd1000/PAMI93.pdf>.
       The last stage of iris recognition occurs after the enrollment process.

Once a person has been enrolled they can proceed to an iris recognition unit. At

the recognition unit, the iris is recorded and the two steps from enrollment are

processed. An additional pattern matching step is required in the recognition

phase. In the pattern matching stage the localized iris picture is compared with

other irises in the database. If a match is found a desired outcome such as

opening a door or granting access to a computer is achieved.



Challenges and Solutions in Iris Recognition

       Although the steps of iris recognition sound straight forward, the actual

details of how each step is achieved is very complex. This section details the

challenges in iris recognition and then provides the solutions that two main iris

recognition technologies use. The first of these iris recognition technologies was

created by John Daugman in 1994. The other was by Richard Wildes in 199812.

The Daugman technology is the more widely used of the two.

       The first challenge of iris recognition is to get a high-resolution image of an

iris while at the same time being noninvasive. This is a challenge because the

iris is small (about 1cm in diameter) and dark, combined with the fact that human

users are sensitive to intense light. There needs to be enough light in order to

get a high quality reading with enough detail and contrast. However, iris

recognition methods cannot shine too much light on the iris because humans are

sensitive to light.

       12
           LG Electronics. “How it Works.” 18 February 2005
<http://publib.upol.cz/~obd/fulltext/Physica%2039/Physica%2039_07.pdf>.
       The Daugman system solves the resolution problem by getting video

images of the iris that have 100 to 200 pixels in diameter from a distance of 15-

46 cm. It also solves the light issue by using a LED light source and a normal

video camera. This method results in a compact design and also allows for the

use of glasses because the LED does not reflect on the glasses. However,

because the LED is located below the user, a refection is captured on the lower

portion of the iris. This means that a portion of the lower iris must be omitted

during matching because it is affected by glare13.

       The Wildes system uses a low-light camera to capture about 256 pixels for

the diameter of the iris from 20cm away. It uses a more complex method of

lighting but in the process reduces glare. This is done by using circular

polarization which blocks reflecting light but allows normal (diffused) light to

penetrate the polarization. The reduced glare in this method allows for more

detail to be recorded and compared which results in a higher level of security14.

       Another problem in image acquisition is to get the iris in the right position

so that it can be recorded without requiring invasive contact (such as a eye piece

or chin rest). The Daugman system allows for this by using a small LCD screen

to show the user what is being captured. The LCD is placed directly behind a

mirror that captures the iris image. In this way, the user can see exactly what the

camera sees and adjust accordingly. While the user is adjusting the video

camera takes continuous footage until a sufficient image is obtained15.



       13
          Wildes, 1353.
       14
          Ibid. 1353.
       15
          Ibid. 1353.
       The Wildes system uses a more low-tech method of positioning the iris. In

this system, a square outline is centered around the camera lens. Another

smaller square outline is also suspended in front of the lens. These squares are

situated so that a user is in the right position when the squares overlap. Once

the user overlaps the squares, he or she pushes a button to have the camera

take a picture. One disadvantage of the Wildes system is that it uses a still

camera for image acquisition and thus may require multiple recordings in order to

obtain a good image16.

       The iris localization stage also creates problems for iris recognition

technologies. Localization is the term used for editing an image to contain

information only pertaining to the iris. Localization is needed because an image

taken by an iris recognition unit contains not only the iris but also the areas

surrounding it. These extra parts include the pupil, eyelashes, eyelids and the

sclera (the white of the eyes).

       Difficulties arise when trying to separate these parts from the iris. The

sclera is the easiest part to differentiate because of the distinct difference

between the white sclera and the darker iris. It becomes much more difficult to

distinguish between the pupil and the iris, especially when the iris is a very dark

brown. It is can also be hard to distinguish between the eyelid and the iris

because of both skin tone and the irregularity of the eyelid border due to

eyelashes.

       The Daugman and Wildes systems use similar methods in distinguishing

the iris from its surrounding area. The first way that these systems separate the
       16
            Ibid. 1353.
regions is by using first derivatives. The derivatives are used to find the change

in color between two points. The greater the change, the more likely that there is

a line between two different parts of the eye17.

       The second way that these systems distinguish regions is through

modeling. Both systems use modeling but in slightly different ways. The

Daugman system eliminates the upper and lower part of the image where eyelids

are expected to be. It also eliminates part of the iris portion where a glare is

expected to be from the LED. In contrast, the Wildes system explicitly models

the top and bottom portions of the image with parabolic curves so as to include

as much of the iris as possible. The exact mathematical differences in the

modeling are beyond the scope of this paper. The Wildes method ends up

produces a more complete image of the eye for better pattern matching but at the

same time requires more computation to achieve its result18.

       The third and final stage of iris recognition is pattern mapping. Within

pattern mapping are four subcategories; alignment, representation, goodness of

match, and decision. Alignment is the process in which an image of an iris is put

into spatial alignment so that it can be compared to other iris images.

Representation is the way in which iris patterns are represented in order to allow

for distinction between irises. The differences between the Daugman and Wildes

systems for the above two subcategories cannot be expressed in this paper

because of the complexity of the mathematics involved. Goodness of match is

the method in which irises are compared. The Daugman system compares irises


       17
            Ibid. 1354.
       18
            Ibid. 1354-1355.
by taking the binary representation of the iris and XOR’s the bits with

corresponding bits of the comparing iris. The Wildes system once uses complex

mathematical calculations to compare irises. The final subcategory is the

decision of whether two irises are from the same person or not. This again

involves mathematics that are beyond the scope of this paper and thus cannot be

compared19.

       Both the Daugman and Wildes systems have been tested. The Daugman

system takes under 1 second to complete both the enrolment and verification

processes. The Daugman system also has an identification mode which allows

for a comparison to the complete database (as opposed to verification which only

compares to a specified database). The identification mode of Daugman’s

system takes 1 second for every 4000 entries compared20.

       The Daugman system was tested twice. The first test involved 592

recorded irises from 323 people (each person’s iris was recorded more than

once). In this test, the crossover rate for false accepts and false rejects was 1 in

131,000. A conditional false reject rate of 1 in 109.6 was found and a conditional

false accept probability was 1 in 1031. No false accepts or false rejects were

experienced in the first test21.

       In the second study up to 122 people were compared to a database of

403. There were some false rejects in this test but most people were accepted

on retries and the remaining were explained to be errors in use such as position

or in eyeglass difficulties due to glare. This test showed no false accepts. These

       19
          Ibid. 1358-1359.
       20
          Ibid 1361-1362.
       21
          Ibid 1361-1362
two tests, while valid, need to be accepted with care because of the small sample

size22. More test of the Daugman systems have probably occurred but results

could not be found.

       The Wildes system takes 10 seconds for both enrollment and verification

and does not have an identification mode. The only test that this system had

was of 60 irises from 40 different people. Twins were included in this test

population. There were no false positives or false negatives in this test. Like the

other tests, this test did not have a large enough sample population to be

considered conclusive. Even though the tests of both the Daugman and Wildes

systems did not have ideal sample populations, the results do present a positive

outlook for iris recognition technology23.



Possible Improvements

       Even though current methods of iris recognition technology are sufficient

for today’s needs, improvements to the technology could always be made. One

of the first improvements would be the use of color cameras for image

acquisition. The use of color would create an additional level of security. Not

only would the patterns of the irises have to match (as is currently tested), but the

color would also have to match. Color would also give an additional way to index

images for better searching through databases24. With color indexing, only




       22
          Ibid 1361-1362
       23
          Ibid 1361-1362
       24
          Wildes, 1353.
images that have similar coloration would be considered for comparing instead of

the whole database.

       Another improvement would in the area of iris acquisition. Even though

iris recognition technology is mostly non-invasive, it still requires a certain

amount of user cooperation in order to be successful. A person has to walk up to

an iris recognition unit and intentionally position themselves to be read. An

improvement would be a system in which an iris recognition unit could record an

image of an iris with little user input. This could be done by taking a picture from

a meter or more away rather than the 15-50cm that is currently required. It would

also require the camera to distinguish between the face and the rest of the body.

Once recognized, the camera could zoom into the face and take a picture

containing the iris. This improved acquisition would allow for much more relaxed

iris recognition and could result in a more widespread application of the

technology25.



Current Applications

       There are many current applications of iris recognition technologies. The

predominate use of iris recognition technology today is at airports. The

technology was first tested in July 2000 at Charlotte/Douglas International Airport

in North Carolina and by EyeTicket Corporation26. Since then many more

airports have tested or embraced iris recognition technology as a means of


       25
          Wildes, 1353
       26
          Mehan, Michael. “Iris Scans take off at airports.” 17 July 2000. 18 February 2005
<http://www.computerworld.com/securitytopics/security/story/0,10801,47202,00.html>.
identification. Amsterdam Airport Schiphol began using the technology in

conjunction with a smart card in October 200127. The United States Homeland

Security Chief Tom Ridge recently announced that iris recognition technology will

be used at New York’s John F. Kennedy Airport as result of the success in

Amsterdam. The technology will be used at the JFK airport to speed up the

customs process for frequent travelers28.

        Other current applications of iris recognition technology include

correctional facility identification A.T.M., verification, pharmaceutical dispensing,

and border control. One of the first uses of the technology was in 1996 when

Lancaster County Prison in Pennsylvania used it to verify the identity of prisoners

to be released. In 1998 Nationwide Building Society bank in England started

using iris recognition instead of PIN numbers for their A.T.M.s29. Argus Solutions

of Australia created a system which uses iris recognition to dispense drugs to

verified patients30. More recently, the governments of Singapore and the United

Arab Emirates have used iris scan technology for board control. The technology

is used in Singapore at its border with Malaysia31. The United Arab Emirates

uses it to check if a person entering the country is one of the 544,000 people that

were expelled from the nation32. As iris recognition technology is applied to an


        27
            Henahan, Sean. “The Eyes Have It.” 17 June 2002. 18 February 2005
<http://www.accessexcellence.org/WN/SU/irisscan.html>.
         28
            Bishop, Ian. “Eye Flight: Iris Scan for JFK Passengers.” 14 January 2005. 18
February 2005 <http://www.nypost.com/news/regionalnews/38159.htm>.
         29
            Henahan
         30
            Argus Solutions. “Dispense True Identity.” 18 February 2005 <http://www.argus-
solutions.com/true_dispense.htm#>.
         31
            Iternational Biometric Group. “Typical Iris Recognition Applications.” 18 February
2005 <http://www.biometricgroup.com/reports/public/reports/iris-scan_applications.html>.
         32
            Daugman, John. “Largest Current Deployment of Iris Recognition.” 18 febr 2005
<http://www.cl.cam.ac.uk/users/jgd1000/deployments.html>.
increasing number of situations and as the technology gets better more people

will embrace its methods.



Companies

       There are many companies that are involved with iris recognition

technology. LG Electronics is one of the largest manufacturers of iris recognition

technology with over 1000 systems in six continents33. Their system is called

IrisAccess 3000 and involves three main components. Two separate iris readers

are used, one for enrollment into the database and one for iris image acquisition.

A control unit is used in conjunction with the acquisition unit to verify a person’s

identity and then perform the desired operation upon verification. LG’s system

provides both audio and visual cues to help a person use the system. An

additional feature of their system is that the iris images are stored on the control

unit and not the iris readers.

       IrisGuard Incorporated focuses on large scale uses of iris recognition such

as border crossings, airport security, and law enforcement. Their technology,

called Iris Farm Architecture (IFA), allows for the splitting of one Iris database

between an unlimited number of database storage units. It also supports slow

speed communications between databases and image acquisition units, allowing

recognition to occur even when communications are slow. Not only does it allow

for slow connection, but is also has become, “the fastest and most scalable

server architecture that exists today for large-scale IrisCode database

       33
          LG Electronics. “Overview.” 18 February 2005
<http://www.lgiris.com/products/index.html>.
searches34.” One aspect that distinguishes IrisGuard’s product, the H100, from

others is the variety of configurations in which it can be used. It can be placed on

a tripod, attached to a swinging metallic arm, or even operated by hand, in

addition to being used in the standard wall mounted position.

       Oki Electric Industries produces two models for iris recognition. The

IRISPASS-H is a small hand-held device that is used with a personal computer to

identify a user. Oki’s more comprehensive unit is the IRISPASS-WG. It is wall

mounted and can be used in conjunction with 126 other units to maintain security

at a site. The IRISPASS-WG includes a feature that automatically detects the

location of the iris, meaning that users only have to stand in front of the unit to

get identified and not peer into a small recognition unit35.

       EyeTicket Corporation focuses on travel processing and access control

using iris recognition. Its two lines of products are EyePass and JetStream. The

EyePass line has a standard iris recognition unit called EP-1 but also has two

more complex units, the EP-2 and EP-3. The EP-2 provides “piggyback”

prevention, meaning that no one else can get through the access point other than

the verified scanned user. This is achieved with a two door system and a

pressure sensitive floor in between. The unit identifies the user, opens the first

door, and allows the user to walk into the middle chamber. This chamber then

uses a pressure sensitive floor to verify that only one person has walked through.

Once this condition is met, the second door is opened and the user is allowed to


       34
          Daugman, Deployments.
       35
          Oki Electric Industries. “Oki Introduces the IRISPASS®-WG Iris Recognition System
with Automatic Iris Scanning Function.” 12 June 2002. 18 February 2005
<http://www.oki.com/en/press/2002/z02011e.html>.
pass. EyeTicket’s second line of products is JetStream. Jetstream is a system

in which users can be positively identified and quickly processed for applications

such as airline, hotel, rail, passport and visa36.

        Iridian Technologies is the leading producer of iris recognition technology.

They research and develop iris recognition technology and hold U.S. and

international patents on the fundamental concepts and technologies behind iris

recognition37. Iridian’s technology is realized in two ways. The first is by their

PrivateID software that is used in iris recognition cameras made by companies

such as Oki and Panasonic. The second way that Iridian’s techonology is used

is through their Proof Positive product certification. This certification assures that

iris recognition products meet specified standards.

        Additional companies that deal with iris recognition technology include

Panasonic, IBM, and SecuriMetrics. Panasonic has multiple iris recognition

products that are used in a number of applications. IBM supplies databases and

servers for iris recognition. SecuriMetrics produces Portable Iris Enrollment and

Recognition (PIER) devices. These units are primarily used for military and law

enforcement and have been used in Iraq and worldwide.




        36
            EyeTicket Corporation. “Index.” 18 February 2005
<http://www.eyeticket.com/index.html>.
         37
            iridian technologies. “About Iridian.” 18 February 2005
<http://www.iridiantech.com/about.php>.
Conclusion

       The demand for security technology increases as the world becomes

more dangerous. Many emerging technologies such as facial recognition and

vocal patterns have tried to gain a share of the industry. However, iris

recognition technology has stood at the forefront. The benefits of the technology

include speed, reliability, and ease of use. Its implementation in diverse settings

has proved it to be a successful method of identification. No other technology

can offer the features and benefits of iris recognition. Expansion and continued

development of iris recognition technologies provides a promising look at a safer

and more secure future.

								
To top