Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

Palm vein authentication technology was evaluated

in Round 6 of Comparative Biometric Testing (CBT)

by International Biometric Group (IBG) in 2006.

CBT of IBG evaluates the accuracy and usability of

biometric products using scenario-based testing, and

strives to understand biometric performance in the

real-world conditions. CBT Round 6 was the ﬁrst

major independent test to evaluate multiple vascular

recognition technologies. Such assessments are typi-

cally based on a comparison of recognition samples

and enrollment templates. In the case of palm vein

authentication, approximately 40,000 [8] genuine

comparisons and 50 million imposter comparisons

were executed.

Results of the IBG study revealed that palm vein

authentication performed exceptionally w ell in the

failure to enroll (FTE) testing; only one person out of

1,290 did not ﬁnish the enrollment process given

the test criteria, a failure rate was of only 0.08%.

This extremely low rate indicates that palm vein au-

thentication is highly applicable virtually for every

individual, and does not impose any physiological

restrictions when users interface with the device. This

further indicates that palm vein authentication is us-

able, is easy for the users to learn, and is ideal for use in

high-volume and large-scale applications.

Most importantly, palm vein authentication was

effective when tested for authentication accuracy.

The false acceptance rate (FAR) and false rejection

rate (FRR) were extremely low, outperforming other

products in the evaluation at standard and high secu-

rity. The perf ormance differences between same- and

different-day transactions were also minimal. There-

fore, after the users learned how to use the device, they

were able to use it successfully on an ongoing basis.

These data further conﬁrm that palm vein authentica-

tion is highly accurate and has optimal usability, both

of which are relevant to real-world conditions.

Implementation

In actual implementation, palm vein patterns can

be stored on a smartcard. The matching between the

palm vein pattern on the smartcard and the one cap-

tured for authentication can also be executed on the

smartcard. Because the palm vein pattern is protected

against external attacks by an antitampering function

of the smartcard, users can handle their own palm vein

patterns safely.

Application

Door Security Systems

Palm vein authentication sensors have been installed

on many access control units over the world (Fig. 3).

They can be used to control entry and exit into and out

of rooms and buildings. For those applications, the

combination of the following features of palm vein

authentication means that it provides the optimum

system: a hygienic contactless unit ideal for use in

public places, user-friendly operation that requires

people to simply hold their palms over a sens or, and

a method that makes impersonation difﬁcult.

In view of the P ersonal Information Protection

ActthatwentintofulleffectinJapaninApril2005,the

Department of Planning, Information and Management

of the Univ ersity of Tokyo Hospital began using palm

vein authentication for a new security system to control

room access. The security levels of the system were

divided into three access levels: ac c ess to the administra-

tive room, the development room, and the server room.

A palm vein authentication access control unit has been

installed at the entrance of each room. The system has

been able to restrict an individual’s entry in stages.

Palm Vein. Figure 3 Palm vein access control unit

implemented by Fujitsu.

1030

Palm Vein

Palm vein authentication sensors can be integrated

into PC mouse (Fig. 4). Using a mouse as a palm

vein authentication sensor offers convenience and

space-saving advantages.

Many companies and government agencies have an

internal information system which handles sensitive

personal data. Using a mouse with an integrated

palm vein authentication sensor enables advanced,

high-level security for system log-ins, beyond mere

IDs and passwords, with the high accuracy and reli-

ability of palm vein authentication.

Financial Services

In 2003, Japan saw a rapid increase in ﬁnancial damage

caused by fraudulent withdrawals from bank accounts

by spooﬁng the identity with fake bankcards that were

made using information from stolen or skimmed

cards. It was a signiﬁcant social problem. This had

caused a sharp increase in the number of lawsuits

taken out by victims against ﬁnancial institutions

for their failure to control information used for per-

sonal identiﬁcation. The ‘‘Act for the Protection of

Personal Information’’ came into effect in May 2005,

and in response, ﬁnancial institutions in Japan have

been focusing on biometric authentication methods

together with smartcards, as a way to reinforce the

security of personal identiﬁcation. Palm vein authenti-

cation is the form of biometric authentication that was

most quickly introduced for customer conﬁrmation at

banking facilities. It has been used since July 2004,

before the act came into effect.

When used for ﬁnancial ser vices, a user’s palm vein

pattern is registered at a bank counter and stored on a

smartcard. This has the advantage of allowing users to

carry their own palm vein pattern around with them,

Palm Vein. Figure 4 PC mouse containing palm vein authentication sensor.

Palm Vein. Figure 5 ATM with palm vein authentication

sensor.

Palm Vein

1031

and lets them manage the usage of their smartcard.

To verify ATM transactions, palm vein pattern of a

user’s is captured by a palm vein authentication sensor

on the ATM (Fig. 5). The captured palm vein pattern is

transferred to the user’s smartcard, and this with the

one transferred from the sensor are compared on the

smartcard. Finally, only the matching result and not

the regist ered palm vein pattern of the user is output

from the smartcard.

In additi on to Japan, Brazil has also already decided

to adopt palm vein authentication to identify users in

ATM banking transactions. Banco Brad esco S.A., the

largest private bank in Latin America, has been on

testing palm vein authentication. After researching

various biometric technologies, Bradesco chose palm

vein authentication because of its outstanding features,

such as its high level of veriﬁcation accuracy and the

fact t hat it is noninvasive and hygienic, making it easier

to be accepted by customers of the bank.

Healthcare

Palm vein authentication is being deployed through-

out t he Carolinas HealthCare System (CHS) in the

United States as par t of a solution to effectively regis-

ter patie nt information and ensure that the proper

medical care is given to the rig ht person, while pro-

tecting their medic al record and privacy from i dentity

theft and insurance fraud. To implement the system,

CHS developed a unique hand guide as a sensor.

The h and guide is adapted perfectly for a hospital

environment since it incorporates a pediatric plate

that adapts the guide so it can be used with young

children, and can accommodate all the patients of

the CHS.

The Sapporo Hospital of Keiyu Association in

Japan also adopted palm vein authentication for their

electronic medical records system for patient authenti-

cation. Patients who are to undergo an operation,

and the registered palm vein pattern and the palm vein

pattern scanned from the patient on the day of the

operation are compared. This conﬁrms that the patient

is the same as the one whose records have been input in

the electronic medical recording system by the doctor

in charge, and avoids the wrong patient being operated

on, which might occur if two patients have the same

name for example.

Some applications for healthcare could be realized

because the contactless type of palm vein authentica-

tion is excellent in terms of hygiene.

Other Uses

The Chiba Institute of Technology in Japan deployed a

student ID system that combines palm vein authenti-

cation and multifunctional smartcards to verify the

identity of stude nts, and lets them securely access

their academic transcripts and other personal records

through information kiosk terminals installed in vari-

ous locations around the campus.

An examination service can use palm vein bio-

metrics to authenticate the identity of examination

candidates. Palm vein authentication is viewed as pref-

erable to other modalities due to the reliability; it is not

easy to steal palm vein images of others, making spoof-

ing difﬁcult.

Summary

Palm vein authentication uses vein patterns on the

palm of a person as personal information. It is a highly

secure technology because palm vein pattern is infor-

mation contained within the body of someone. It is

also hig hly accurate because palm vein patterns are

complex and unique to each individual. Moreover, its

contactless feature gives it a hygiene advantage over

other authentication technologies. Many users of prac-

tical applications have highly evaluated this authenti-

cation method and experienced no psychological

resistance to using it. This is good reason for develop-

ing new products for various solutions, starting with

ﬁnancial solutions followed by access control units and

then login sensors.

Related Entries

▶ Vascular Image Data Format, Standardization

▶ Vascular Network Pattern

▶ Vei n

References

1. Rice, J.: Apparatus for the Identiﬁcation of Individuals. US

Patent 4,699,149 (1985)

1032

Palm Vein

2. Totty, M.: A better idea. Wall St. J. Oct. 24. (2005)

3. Kim, J.G., Xia, M., Liu, H.: Extinction coefﬁcients of hemoglobin

for near-infrared spectroscopy of tissue. IEEE Eng. Med. Biol.

Mag. 24, 118–121 (2005)

4. Wray, S., Cope, M., Delpy, D.T., Wyatt, J.S., and Reynolds, E.O.:

Characterization of the near infrared absorption spectra of cyto-

chrome aa3 and haemoglobin for the non-invasive monitoring

of cerebral oxygenation. Biochim. Biophys. Acta. 933(1),

184–192 (1988)

5. Cope, M.: The application of near infrared spectroscopy to non

invasive monitoring of cerebral oxygenation in the newborn

infant. Ph.D. thesis, University College London, Appendix B

316–323 (1991)

6. Editorial Board for Visualization: Techniques of Biological

Information Visualization Techniques of Biological Information

(in Japanese), Corona Publishing Co., Ltd., p. 86 (1997)

7. Watanabe, M., Endoh, T., Shiohara, M., Sasaki, S.: Palm vein

authentication technology and its applications, In: Proceedings

of Biometrics Symposium. 37–38 (2005)

8. International Biometric Group: Comparative Biometric Testing

Round 6 – Public Report. http://www.biometricgroup.com/

reports/public/reports/CBT6_report.htm. Accessed 19 Mar,

2009

9. Sasaki, S., Hiroaki Kawai, H., Wakabayashi, A.: Business expan-

sion of palm vein pattern authentication technology, FUJITSU

Sci. Tech. J. 41(3), 341–347 (2005)

Palm Vein Authentication

▶ Palm Vein

Palm Vein Authentication Sensor

▶ Palm Vein Image Sensor

Palm Vein Image Sensor

MASAKI WATANABE

Fujitsu Laboratories Ltd., Kawasaki, Japan

Synonyms

Palm vein authentication sensor; Palm vein scanner

Definition

The palm vein image sensor is used for palm vein

authentication. The device captures an image of the

vein pattern in the palm by emitting near-infrared rays

that are absorbed by the deoxygenated hemoglobin in

the veins and then reﬂected back to the device for

image capturing.

The palm vein image sensor is commercially avail-

able as a device for more secured personal identiﬁcation.

Example applications include door security systems, PC

systems, and hospital patient conﬁrmation systems.

Introduction

The palm vein image sensor is used for palm vein auth-

entication, a vascular pattern recognition technolog y.

The device uses the vein pattern of the palm as person-

al identiﬁcation data. Therefore, the palm vein image

sensor must scan the position of the veins with the

highest degree of accuracy.

The technology for noninvasive scanning of blood

vessels is primarily categorized as in vivo meas ure-

ment. In this ﬁeld, near-infrared spectroscopy (NIRS)

and imaging has been investigated for the last ten years

[1]. The palm vein image sen sor was developed based

on this NIRS technology.

Infrared scanning of the pattern of subcutaneous

blood vessels for the identiﬁcation of individuals w as

ﬁrst disclosed in Rice’s patent in 1985 [2]. In 2001,

Peterson et al. patented a device for palm vein authen-

tication [3], in which a plurality of light-emitting ele-

ments is arranged in an array with light-detecting

elements in a ﬂexible mat. The mat was intended

to be put into a device to be grasped by the hand of the

person to be identiﬁed following to Stiver’s patent [4].

The device comprises of an elongated transparent

cylindrical exterior shell and scans the hand grasping

the device. In 2003, Fujitsu realized a box-type device

for practical use and it was launched in Japan the

following year. In 2008, Snowﬂake Technologies [5]

released a prototype in which a palm vein sensor is

embedded to the upper part of equipment for a door

security system. The palm vein image patterns can also

be simultaneously acquired with palmprint images and

employed to achieve improvement in performance of

palmprint authentication as detailed in [12].

Palm Vein Image Sensor

1033

Hemoglobin in vessels is grouped into two types:

oxygenated hemoglobin that is in ar teries and contains

oxygen; and deoxygenated hemoglobin that is in veins

and does not contain oxygen, and in particular, absorbs

light with a wavelength of about 760 nm [6, 7]. When

the palm of the hand is illuminated by near-infrared

light, the rays will be scattered by structures under the

surface of the ski n. Much of this illumination will be

reﬂected bac k towards the illumination source, how-

ever little will be reﬂected from the veins since these

absorb near-infrared.

Palm vein patterns are preferably acquired using

the reﬂection method [3] and also in commercially

available products from Fujitsu [9] and Snowﬂake

[5], whereby the palm is illuminated from the front

and the image is also captured from the front. If the

transmission method is used, whereby the palm is

illuminated from the back of the hand and the image

is captured from the front, a very strong light would

be needed.

To realize this imaging method, the palm vein

image sensor must have both of an illumination func-

tion and imaging function by near-infrared rays. The

illumination function must emit light in a wavelength

of about 760 nm and the imaging function must have

sufﬁciently high resolution to distinguish the vein pat-

tern. In the reﬂection method, because the direction of

illumination is the same as that of image capturing, the

illumination device and the imaging device are com-

pactly integrated.

A palm vein image sensor should ideally scan

as broad a palm area as possible for keeping high

accuracy of palm vein authe ntication because the

human palm has an extensive a nd complex vein

pattern that contains sufﬁcient information to identi-

fy an individual from among many people. But some

kinds of sensors which scan a par tial area of a palm

would be designe d fo r rea son s of usability such as the

Stiver’s patent.

Implementation

The research and development efforts for the compact

palm vein image sensors have been conﬁned to few

commercial vendors. Therefore only very limited tech-

nical details are available for the palm vein image

sensors. In Fujitsu’s implementation [8, 9], the palm

vein image sensor is in the shape of a small box 35 mm

deep by 35 mm wide by 27 mm high. Image capturing

is executed in a contactless manner. Users do not touch

the sensor; they only have to hold their palms above it.

The user places his or her hand below an optical reader,

which scans the palm.

To obtain a clear image of the palm vein pattern,

imaging is controlled according to the movement and

position of the hand above the sensor or below the

sensor, and illumination is controlled recognizing the

light around the sensor. Video-rate scanning is ty pical-

ly employed for the convenience in the palm vein

authentication so that users do not have to stop the

hand for authentication.

Any imaging devices such as CMOS sensors or

CCD sensors will be used for the capturing but it

must have sensitivities of near-infrared rays. It should

be also assembled not to capture except for the

reﬂected near-infrared rays from inner of the hand

using such as a polarizing ﬁlter, an optical ﬁlter cutting

off visible lights, and so on.

The contactless method, which is adopted by both

of sensors, eliminates the concerns of users who are

sensitive about hygiene or who are reluctant to come

into direct contact with publicly used devices. It also

enables the identiﬁcation method to be used in envir-

onments where a high standard of hygiene is required,

such as in medica l facilities or food factories.

Regarding security, a data encryption function for

the palm image should be also provided. This ensures

that image output is protected from any unauthorized

access or tampering.

Other possible substantiation of palm vein image

sensor has been also announced. NEC Corporation

developed the world’s ﬁrst contactless multi-modal

ﬁnger recognition technology [10]. The new device

quickly scans two forms of biometric information,

ﬁngerprints and vein patterns of ﬁnger. They proved

that the device might also be adapted to recognize skin

and vein patterns from any region of a human body. As

an example, the possibility of a scanning system cus-

tomized to analyze and authenticate both ﬁngerprint

and palm characteristics was also shown.

Evaluation

Comparative Biometric Testing (CBT) by the Inter-

national Biometric Group (IBG) evaluates the accu-

racy and usabilit y of biometric products using

1034

Palm Vein Image Sensor

scenario-based testing, and strives to understand bio-

metric performance in real-world conditions. Round 6

of the testing in 2006 evaluated palm vein authentica-

tion technology using Fujitsu’s sensor.

The results of the IBG study revealed that palm vein

authentication using this sensor performed exception-

ally well in the failure to enroll (FTE) testing [11].

Authentication accuracy was also good; the false accep-

tance rate (FAR) and false rejection rate (FRR) were

extremely low. Performance differences between same-

day and different-day transacti ons were also minimal.

Thus, after the users learned how the sensor worked,

they were able to use it successfully on an ongoing

basis. This data further conﬁrms that palm vein au-

thentication using this sensor is highly accurate and

has optimal usability, both of which are directly rele-

vant to real-world conditions.

Application

Palm vein image sensors are embedded in many differ-

ent types of equipment for various applications, such

as for door security systems, PC login, ﬁnancial ser-

vices, and so on.

The access control unit for door security systems is

a typical example in which the palm vein image sensor

is installed. Units throughout the world are equipped

with this sensor using various methods. Almost all the

units have an input device for the user’s ID, such as a

ten-key, smartcard reader, or both (Fig. 1).

Moreover, for the purpose of PC login, the palm

vein image sensor can be installed in the PC mouse

(Fig. 2), and in the case of ﬁnancial services, it can be

installed in almost any type of ATM (Fig . 3).

Some of the equipment installed with a palm

vein image sensor includes a hand guide on which

the wrist is placed, so that ﬁrst-time users of palm

vein authenticati on can easily understand how to

use this sensor. The adaptation of the Carolinas

HealthCare System (CHS) in the United States, which

Palm Vein Image Sensor. Figure 1 Palm vein access

control unit implemented by Fujitsu.

Palm Vein Image Sensor. Figure 2 PC mouse equipped with palm vein authentication sensor.

Palm Vein Image Sensor

1035

is designed to protect patients from identity theft and

insurance fraud, is a good example of utilizing the

hand guide. To implement the system, CHS developed

a unique hand guide, adapted perfectly for a hospital

environment since it incorporates a pediatric plate

for young children, and thus can accommodate all

CHS patients.

Summary

The palm vein image sensor is used for palm vein

authentication employing the vein pattern of the

palm of a person as personal identiﬁcation informa-

tion. An image of the palm vein pattern is captured

using near-infrared rays in the reﬂection method. The

illumination device emits light with a wavelength of

around 760 nm, which is more strongly absorbed by

the deoxygenated hemoglobin in the veins compared

to the surrounding subcutaneous tissues of the hand.

The imaging device captures the whole or partial

palm area and has a sufﬁciently high resolution to

distinguish the vein pattern. In case of sensors using

the reﬂection method for capturing, those two

devices are compactly installed. The palm image sen-

sor is used for various applications and is embedded

in many different kinds of equipment, such as the

access control unit for door securit y systems, PC

mouse for login management, ATMs for ﬁnancial

services, conﬁrmation units for hospital patients,

and so on.

References

1. Kim, J.G., Xia, M., Liu, H.: Extinction coefﬁcients of hemoglobin

for near-infrared spectroscopy of tissue, IEEE Eng. Med. Biol.

Mag. March/April, 118–121 (2005)

2. Rice, J.: Apparatus for the Identiﬁcation of Individuals,

US Patent, 4, 699, 149 (1985)

3. Peterson, D.C., Jackson, D.W., Stiver, J.A: Method and appa-

ratus for subcutaneous identiﬁcation, US Patent, 6,330,346

(2001)

4. Stiver, J.A. Peterson, D.C.: Identiﬁcation system, US Patent,

5,793,881 (1998)

5. Snowﬂake Technologies, http://snowﬂaketechnologies.com/

Accessed 19 Mar, 2009

6. Wray, S., Cope, M., Delpy, D.T., Wyatt, J.S., Reynolds, E.O.:

Characterization of the near infrared absorption spectra of

cytochrome aa3 and haemoglobin for the non-invasive moni-

toring of cerebral oxygenation, Biochim. Biophys. Acta, 933(1),

184–192 (1988)

7. Cope, M.: The application of near infrared spectroscopy to

non invasive monitoring of cerebral oxygenation in the newborn

infant, Ph.D. thesis, University College, London Appendix D.

316–323 (1991)

8. Watanabe, M., Endoh, T., Shiohara, M., Sasaki, S.: Palm vein

authentication technology and its applications, In: Proceedings

of Biometrics Symposium, pp. 37–38 (2005)

9. Sasaki, S., Hiroaki Kawai, H., Wakabayashi, A.: Business expan-

sion of palm vein pattern authentication technology, FUJITSU

Sci. Tech. J. 41(3), 341–347 (2005)

10. NEC Corporation: NEC develops the world’s ﬁrst con-

tactless multi-modal ﬁnger recognition technology (2008),

http://www.nec.co.jp/press/en/0805/1402.html. Accessed 19

Mar, 2009

11. International Biometric Group: Comparative Biometric Testing

Round 6 – Public Report (2006), http://www.biom etricgroup.

com/reports/public/reports/CBT6_ report.htm. Accessed 19

Mar, 2009

12. Wang, J.-G. Yau, W.-Y. Suwandy A. Sung, E. Person recognition

by palmprint and palm vein images based on ‘Laplacianpalm’

representation. Pattern Recogni. 41(5), 1531–1544 (2008)

Palm Vein Recognition

▶ Palm Vein

Palm Vein Scanner

▶ Palm Vein Image Sensor

Palm Vein Scanner. Figure 3 ATM with palm vein

authentication sensor.

1036

Palm Vein Recognition

Palmprint

Palmprint is an impression or image left on a surface

by the friction skin of the palm.

▶ Anatomy of Friction Ridge Skin

Palmprint, 3D

DAV I D ZHANG,VIVEK KANHANGAD

Biometrics Research Centre, Department of

Computing, The Hong Kong Polytechnic University,

Hung Hom, Kowloon, Hong Kong

Synonyms

P almprint authentication, 3D; P almprint recognition, 3D

Definition

Biometric systems that rely on unique features of the

palm for personal recognition are referred to as palm-

print based biometrics. A 3D Palmprint biometric

system employs 3D imaging device to acquire surfaces

of human palm and uses this data in performing user

identiﬁcation. Extracted features from a 3D palmprint

data include depth and curvature of palmlines and

wrinkles on the palm surface.

Introduction

Human palmprints are rich in features that are unique

and stable. Major palmprint features include principal

lines, wrinkles, ridges, singular points, and minutiae

points (see Fig. 1). In addition, human palmprints

are also abundant with texture features. Apart from

being feature-rich, palmprints have advantages over

other hand-based biometric technologies [1, 2]:

1. Compared to the ﬁngerprint, the palm provides

a larger surface area so that more features can be

extracted

2. An individual is less likely to damage a palm than a

ﬁngerprint, and the line features of a palm are

stable throughout one’s lifetime

3. Small amounts of dirt or grease appearing on an

individual’ s ﬁnger may pose challenges in accurately

extracting features for a ﬁngerprint system. This

problem does not arise in the extraction of palmprint

features since a comparative low resolution palmprint

images are used to extract these features

Most of the current works in the area of automated

palmprint recognition are based on acquiring intensity

image of the user’s palm and extracting line or texture

features from it. Although these systems have been able

to achieve promising performance with low error rates

[3], there are a few inherent limitations associated with

such systems. Firstly, they are sensitive to spurious pat-

terns such as dirt, lines or text on the palm. Performance

of the palmprint recognition systems based on intensity

images can be severely affected by the spurious patterns

on the palm. An impostor may also blemish his palm

with a purpose to circumvent the system. In addition,

like most other 2D image based personal recognition

systems, palmprint systems are also vulnerable to sensor

level spoof attacks. An impostor may easily fabricate

a spoof palm resembling a genuine user’s palm and

use to it to circumvent the system. On the other hand,

a palmprint biometric system based on 3D images of

users’ palm offers higher degree of robustness against

such attacks. These systems are extremely difﬁcult to

circum v ent as they require sophisticated methods to fab-

ricate spoof 3D palmprint models. Another factor that

often has an impact over the performance of

Palmprint, 3D. Figure 1 Palmprint features.

Palmprint, 3D

1037

con ventional palmprint rec ognition systems is the

changes in the illumination. This, in fact is a major

problem for other biometric systems such as the one

employing users’ face images. The face images are often

requir ed to be imaged in outdoor environments and, as a

result,showlargeintraclassvariations.However,palm-

print systems often acquire images in a contr olled envi-

ronment and therefore are less affected by inconsistent

illumination.

One possible solution to overcome these limitations

is to make use of depth features of the palm surface

using a 3D imaging device. Such observations will pro-

vide information on depth and curvature of palmlines

and wrinkles on the palm surface. Comparable perfor-

mance and robustness make 3D palmprint biometric

system a good candidate for high security identiﬁcation

tasks. However there has not been much research

focused on exploring the utilities of 3D palmprint fea-

ture. Therefore references on this topic are limited.

3D Palmprint Recognition System

A 3D palmprint based personal recognition system

includes the following modules:

1. 3D palmprint imag e acquisition device

2. Extraction of Region of Interest (ROI) to obtain the

central part of the palmprint

3. Feature extraction

4. Feature matching, where extracted features are

matched with their respective feature templates

stored during enrolment phase, generating a simi-

larity score. In identiﬁcation applications, the

query template is matched to all templates enrolled

in the database. Therefore, a one-to-many com-

parison is performed in this case

5. Decision module, where the similarity score pro-

duced is compared to the threshold of the system

to either accept or reject the identity claim. In iden-

tiﬁcation applications, identity of the user is deter-

mined to be the one with highest similarity score

3D Palm Image Acquisition

Infrared sensors are employed to detect the presence of

the hand on the acquisition device. When a hand is

detected, the device projects multiple light patterns

onto the palm surface and acquires depth information

using active triangulation. In order to distinguish be-

tween stripes, they are coded with different brightness.

The system uses a computer controlled liquid crystal

display (LCD) projector that can generate arbitrary

stripe patterns. A CCD camera is used to acquire the

images formed on the palm side. The sequence of

images acquired by the CCD camera is then processed

to obtain the 3D palm data. Figure 2a shows the

process of acquisition of 3D palmprint data using a

3D image acquisition device based on structured

light principle. The US patent [4] describes the pro-

cess of acquisition of 3D ﬁnger and palmprint infor-

mation using multi-camera and light projection

system. This device projects multiple structured lights

from different directions and the images formed on

the object (palm or ﬁnger) are captured by cameras at

different angles. It is claimed that the system can

simultaneously obtain 3D ﬁngerprint and palmprint

information.

Feature Extraction

Acquired 3D images are processed to extract the region

of interest (ROI). The inter ﬁnger points are used as

reference points to extract a sub image of ﬁxed size

located at the center part of the palm. These 3D sub

images are further processed to extract surface curva-

ture features. To represent the curvature of every point

on the 3D palmprint image by a scalar value, the

curvedness (C) introduced in [5] is utilized. The posi-

tive value C is a measure of how sharply or gently

curved, a point is [6]. It is deﬁned in terms of principal

curvatures k

, and k

, as:

C ¼

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

ðk

þ k

ð1Þ

The principal curvatures k

and k

can be deter-

mined as:

; k

¼ H 

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

 K

ð2Þ

where K and H are Gaussian and mean curvatures

respectively. For a surface patch represented by

Xðu; vÞ¼ðu; v; f ðu; vÞÞ, the values of K and H are

computed as follows:

KðXÞ¼

 f

ð1 þ f

þ f

1038

Palmprint, 3D

and

HðXÞ¼

ð1 þ f

Þf

þð1 þ f

Þf

 2f

ð1 þ f

þ f

3=2

ð3Þ

where, f

; f

and f

; f

are ﬁrst and second order

partial derivatives of f ðu; vÞ:

The scalar value of curvature (C) is obtained

for every point on the 3D palmpri nt image and this

can be stored in a 2D matrix or an image. Set of such

scalar values is referred to as surface curvature map.

Figure 3 shows sample 3D palmprint images and corre-

sponding curvature maps. It can easily be observed

that the surface curvature maps closely resemble the

palmlines, especially the strong principal lines.

Additional features such as surface types that are

characterized using the sign of mean and Gaussian

curvature [7], shape index [8], computed using mini-

mum and maximum curvature values; may also be

explored as feature representations for the 3D palm-

print images. Shape index, a local feature computed at

each point, is deﬁ ned as:

S ¼



arctan

þ k

 k

ð4Þ

The value of shape index S lies in the interval [0,1] and

can be used to classify each point on the surface to

different surface types ranging from spherical cup

to spherical cap.

Another approach to perform user authentica-

tion is to extract 3D palmlines from the acquired

3D palmprint images and match these lines (set of

points) using point set alignment algorithms such as

iterative closest point (ICP) [9]. ICP can be employed

to iteratively estimate the transformation between the

two 3D palm lines. Alignment error (e.g., mean

squared distance) generated by the ICP can be used

as the matching score.

Feature Matching

Feature matching establishes the similarity between

two 3D palm print imag es. An image matching tech-

nique, normalized local correlation can be employed to

compare two curvature maps. Result of this matching

is a correlation value for every point in the input

curvature maps. Average of these correlation values is

considered to be the matching score. The expression

for normalized local correlation is given by:

C ¼

i¼N

j¼N

ðP





PÞðQ





QÞ

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

i¼N

j¼N

ðP





PÞ

i¼N

j¼N

ðQ





QÞ

ð5Þ

Palmprint, 3D. Figure 2 (a) Example 3D palmprint image acquisition system. (b) Receiver Operating Characteristics

(ROC) curve for a 3D palmprint authentication system.

Palmprint, 3D

1039