Fingerprint on the Move (FoM)™ Identificationafisandbiometrics.com/yahoo_site_admin/assets/docs/... · Fingerprint on the Move (FoM)™ Identification ... Fingerprint on the Move

Fingerprint on the Move (FoM)™ Identification Framework for Identification on the Move

Ben Bavarian,

Frank Lou, and

Mehrad Tavakoli

Principal Consultants

AFIS and Biometrics Consulting Inc.

The research in this presentation is partly supported by the

DHS science and technology Contract No. NBCHC090025

2011© AFIS and Biometrics Consulting Inc

96th IAI Educational Conference

Milwaukee, Wisconsin

August 11, 2011

Lecture outline

2

Background / the question

Stand –still pose capture and processing

Moving hand freeze capture and processing

The laboratory prototype operation;

Video clips of how the Fingerprint on the Move

(FoM) ™ works.

Background / the Question

3

Build a prototype system for identifying subjects on the

stand-off and on the move;

Use COTS Sensors/Cameras; No special hardware.

The Question:

Is it possible to take a picture of the palmar area of the

hand and get the friction ridge details from the finger and

palm area and be used for identification and verification of

the individual reliably?

The COTS components used in the proejct

4

Digital Camera Back - A device that attaches to the back of the

camera, allowing larger sensor size to be used on the medium

format. We have used the Digital Transition Phase 1 digital camera

back with effective array size of 8964x6732 - 60mpixels, and

The Canon DSLR family of cameras – 5D Mark II; Rebel XOS and

Rebel T1; plus standard flash and LED lighting panels.

Can we make identification with non-contact hand scan processing

5

Sample Cropped and Pre-Processed Images

6

Can we make identification with non-contact hand scan processing – DSLR Camera set up

7

Hand scan processing steps–second set up

8

Image Binarization

Post-processing

Finger-palm segmentation

Pre-processing

Sample ink and livescan fingerprints from the background database

9

Sample Images with Their Corresponding Gray-Scale, Binarized and Skeletonized Images

10

Right Thumb Finger

11

Can we make identification with non-contact hand scan processing

12

Search Id: 4029f2 – Case 1, Right hand index finger

Search Print Number Minutiae: 123

Rank ID Candidate # of

Minutiae

Match Score

1 4030f2 123 262.3000

2 4029f5 98 0.0971

3 4030f5 98 0.0968

4 4029f4 82 0.0937

5 4030f4 97 0.0853

6 4029f3 122 0.0823

7 4030f3 133 0.0739

Fingerprint match report

Top 7 respondents

Segment the fingerprints

– obtained 750 ppi image

Interpolate to 500 ppi

Map to gray scale image

Use standard processing

to detect the ridge flow

and extract the minutiae

Summary results for fingerprint match score analysis Histogram of matching scores and non-matching scores, Flat captures, without thumb

13

Summary results for fingerprint match score analysis Histogram of matching scores and non-matching scores, Rolled captures, without thumb

14

Summary results for fingerprint match score analysis Histogram of the Flat and Rolled fingerprint match scores

15

Summary results for fingerprint match score analysis Sample match report for right index

16

Search Token: ID=1, Data type=Right Index, Session F

Data Type Subject ID Score

Flat-Right-Index 1 125

Rolled-Right-Index 1 65

Flat-Right-Little 53 17

Flat-Right-Little 1 17

Flat-Right-Little 35 15

Flat-Right-Middle 1 15

Flat-Right-Index 28 14

Flat-Right-Ring 35 14

Flat-Left-Little 49 14

Flat-Right-Ring 53 14

ROC plots True match vs. false match rate analysis for fingerprint match score ROC diagram for individual digits of fingers for image-based versus standard fingerprints

17

0

0.2

0.4

0.6

0.8

1

1.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Tru

e A

cce

pta

nce

Rat

e

False Acceptance Rate

Image vs FP Little Image vs FP Index Image vs FP Thumb

Image vs FP Middle Image vs FP Ring

ROC plots True match vs. false match rate analysis for fingerprint match score Quality based categorized ROC diagrams

18

0

0.2

0.4

0.6

0.8

1

1.2

0 0.2 0.4 0.6 0.8 1

Tru

e A

cce

pta

nce

Ra

te

False Acceptance Rate

Good Quality Best Quality Worst Quality

Bad Quality Average Quality

ROC plots True match vs. false match rate analysis for fingerprint match score ROC diagram for best quality image-based fingerprints and standard livescan fingerprints

19

0

0.2

0.4

0.6

0.8

1

1.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Tru

e A

cce

pta

nce

Rat

e

False Acceptance Rate

FP vs FP Best Quality Image vs FP

Next generation of identification and authentication using the non-

contact scan of the palmar area of the hand with high resolution

COTS camera and automatically getting the information needed (

fingerprints and palmprint minutiae templates) to carry out the one to

one verification and one to many identification task.

The concept is like the barcode scanners on the check-out counters

in the supermarket. The surface of your hand is your unique

"barcode" and our protoyype set up will do the capture and process

and match automatically.

Fingerprint on the Move (FoM)

20

Stand off access control without using any fingerprint readers on the

access door. The product uses very high resolution digital back

camera to capture the hand image from 4 to 6 feet away process and

authenticate to provide the access for the person.

Follow up prototype concept of operation Hand-Match IDentification ™ (HaMID)

21

Sample Experimentation Fingerprint on the Move (FoM)

22

Two video clips demonstrating the FoM prototype operation

FoM images sample image and the output of the automatic segmentation from the hand picture.

23

Discussions / Q&A

The feasibility of processing hand palmar area photos to

carry out friction ridge detail feature extraction is

demonstrated;

The feasibility of freeze-capture of the moving hand is

demonstrated;

Possible future use for “your-fingerprint-your-unique-

barcode” for verification and identification on the move. 24

Dr. Bavarian is one of the pioneers and industry leading authority in the field of Biometric

Identification with over 26 years of R&D and management experience in industry and academics.

He founded the AFIS and Biometrics Consulting Inc. in 2007, providing subject matter expert

consulting and strategic management services in Biometrics Identification Industry. The company

has successfully completed more than two dozen contracts in the last three years.

Prior to ABC Inc. Dr. Bavarian was the Vice President of Motorola Biometrics, where he led the

business turn around and four fold increase in Sales by directing the development of the industry

leading Automated Biometric Identification System products with over 100 large scale

deployments.

Before moving to the industry in 1992, Dr. Bavarian was a professor in the Department of

Electrical and Computer Engineering at the University of California, Irvine, where he conducted

original research in image processing, computer vision, intelligent systems and published over

120 technical papers and received several awards for outstanding research and distinguished

teaching.

Dr. Behnam (Ben) Bavarian received his Ph.D. in Electrical and Computer Engineering from The

Ohio State University, Columbus Ohio in 1984.

Author Biography - Dr. Ben Bavarian

25

Dr. Lou is the Director of Research and Product Development at AFIS and Biometrics

Consulting Inc. Prior to joining ABC Inc. Frank was a principal research engineer at Motorola

Biometrics Business Unit where he has helped to advance the core AFIS technology and get

the best performance in the NIST and Customer Benchmarks.

At ABC Inc. he has lead the architecture and core technology development of the Biometrics

Identification on the Move System™ (BIMS), a break thorough solution to incorporate stand-

off biometric capture and matching for positive identification funded by the DHS Science and

Technology.

Dr. Lou received his Ph.D. from University of Washington, Seattle in 2004 with expertise in

Image processing, pattern recognition and Signals and Systems. He has publishes over 30

technical papers, report, Patents and disclosures, and book chapter in the related areas.

Author Biography - Dr. Frank Lou

26

Mr. Mehrad Tavakoli is PhD graduate student at University of California, Irvine California (UCI),

where he is pursuing his research in advancing Biometrics Identification science and its

application for Law Enforcement, Security and Commercial applications.

Mr. Tavakoli is the lead research scientist for the UCI project on Biometrics Identification on the

Move System™ (BIMS) where he has developed new techniques for using high resolution

imaging and the practical applications to capture of face, iris and fingerprint and palmprint friction

ridge details.

In his prior experience Mehrad has designed and build a Geo-Spatial Imaging system similar to

Google Streets, where massive amount of image database where captured and managed for

instant access. He has also led the R&D efforts for new projects in IT and Communications areas

such as new ad-hoc routing protocols, MAC-layer protocols, and underwater communication

which included the underwater surveillance security systems and solutions.

Mr. Tavakoli received his Masters of Science Degree from UCI, specializing in Computer

Architecture, Image processing, Signal and Systems and completing his thesis on Stand-Off

Biometrics Identification.

Author Biography - Mr. Mehrad Tavakoli

27