See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/260739557

How to Build an Automated Fingerprint Identification System

Conference Paper · July 2013

DOI: 10.1109/ISBAST.2013.9

CITATION READS
1 2,151

4 authors, including:

Sarah Irum Hassan Afzal

COMSATS University Islamabad University of Luxembourg
4 PUBLICATIONS 49 CITATIONS 10 PUBLICATIONS 56 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Sarah Irum on 01 February 2016.

The user has requested enhancement of the downloaded file.

 2013 International Symposium on Biometrics and Security Technologies

How to build an Automated Fingerprint

Identification System
Zeeshan-ul-hassan Usmani Sarah Irum Hassan Afzal Shahid Mahmud
Riphah International University Interactive group SnT - University of Luxembourg Interactive group
Islamabad, Pakistan Islamabad, Pakistan hassan.afzal@uni.lu Islamabad, Pakistan
zeeshan.usmani@riu.edu.pk sarah.irum@iacgrp.com sm@iacgrp.com

Abstract—Fingerprint based identification is becoming increas- treatment or prescription which could be possible in sim-
ingly useful and cost effective now a days to tackle growing trend ple computerized record evolving data entry mistakes. Other
of identity theft, repeat offenses, authentication frauds, medical applications of AFIS include authentication of employees in
malpractices, and unlawful access & permissions. This paper
provides a historical background on the evolution of fingerprint different organizations.
technology and detailed the salient feature such system should This paper is organised in the following sections. Section I
possess to meet the contemporary technical challenges. This introduces the AFIS system. Section II presents a background
work also presents the industry landscape of commercial and of the biometric information based identification and corre-
non-commercial biometrics market and a selected list of its
sponding advent of AFIS. In section III we briefly discuss
various applications. The paper conclude with a demonstration
of building an open-source AFIS from scratch and the required the main characteristics/features a modern AFIS should entail.
phases and research work needed to build such a solution. AFIS standards are listed in section IV. In section V, a review
Index Terms—Image Processing, AFIS, Feature extraction, of the commercial and non-commercial AFIS related SDKs
Feature matching is presented. Section VI discusses the major phases and their
corresponding activities which would be required to develop a
I. I NTRODUCTION holistic, large-scale and end-to-end AFIS solution. In section
Developing countries in the world face the technology VII, we present a brief overview of the work done on research
challenges to deal with the surge of terrorism, unlawful and development of eAFIS (Efficient Automated Fingerprint
activities, crimes, target killings, and absence of transparency Identification System) for Civil and Criminal applications
from daily businesses of government and private ventures. As based on open-source AFIS tools. This paper ends with a
the population growth rate increases rapidly the developing summary regarding the work done.
countries badly need the self-reliance on identity management
technologies to take care of terrorism and unlawful activities II. BACKGROUND
and keep an account of their growing population and to bring In the late nineteenth century a Parisian police clerk
the transparency in all government and commercial systems. Alphonse Bertillon invented a system called Bertillon system
Identity Management & Biometrics with its most used form that relied primarily on an elaborate set of eleven carefully
of fingerprint identification has been used in few countries recorded anatomical measurements - the length and width of
for over a decade now in many forms. National Database the head, length of the left middle finger, the distance from
Registration Authority (NADRA) is using it to account for the left elbow to the tip of the left middle finger, and so on
adult population; National Police Bureau under the flagship [13]. The system was a great success in enabling to catch
of Federal Investigation Agency (FIA) is using it to keep the repeat offenders. The beauty of the system was to be able to
tally of criminals, to prevent and punish the repeat offenders, code the features in ”one collective number”, to be able to
and several other law enforcement agencies are using it for massively search and narrow down the possible matches.
purposes including border control, passport issuance, issuance As Bertillon system was gaining popularity, Sir Francis
of driving licenses and for visa inquiries etc. Galton (The Galton point is termed after his name) discovered
In Biometric based identification systems, fingerprints are the sui generis nature of human fingerprints [13]. He wrote
most cost effective as well as reliable. For this reason, AFIS the master piece ”Finger Prints” in 1892. In 1893, the Home
deployment is getting common in developing countries. In Ministry Office, UK, endorsed that no two individual have
many countries, AFIS has been used in election for identifica- same fingerprints [13]. Galton’s system works on the method
tion for legitimate voters and for criminals AFIS used by law- of exclusion, when we find even a single dissimilarity we
enforcement agencies to verify them with the help of latent call it a ”mismatch”. And to be able to ”match” the prints
prints. it should match on multiple points Rigs, Forks, Minutiae
Hospitals are also adopting AFIS for patient and physician etc. The United Kingdom and Pakistan have been using 9
verification to ensure authorized access to medical records and points system, Italy has been using 12 points, US varies state
facilities. This also eliminates the chance of wrong diagnostics, by state from 9 to 15. A difficulty with this system is the

978-0-7695-5010-7/13 $26.00 © 2013 IEEE 40

DOI 10.1109/ISBAST.2013.9
absence of ”common compound value” to be able to optimally 1) Enrollment: AFIS would be able to efficiently and
search the database for match (as we have in Bertillon system). accurately register a person’s information in its database.
Advent of this system led to extensive usage of fingerprint This process begins with the capture of a fingerprint by a
data by law enforcement departments for record keeping of fingerprint scanner followed by a quality check. After that
criminals and for later identification, for use in forensics i.e. the critical information (also called features) are extracted
for matching the latent prints found in crime scenes with the from that scan and arranged in standard format to form a
already available record of professional criminals or suspects. feature template. Sometimes more than one fingerprint scans
A great deal of effort was put on developing scientific manual for one finger would be used to extract the best information.
matching techniques and training experts in this domain. Sometimes scans of more than one finger would used to create
The efforts spent on the improvement of manual techniques a composite template. Moreover this process can be repeated
for fingerprint indexing and matching were soon overwhelmed for all ten fingers in some cases resulting in ten separate feature
by the increased demand for fingerprint recognition. Due to templates.
this factor and also with the advent of digital computing which The enrollment of biometric information would be compli-
gained popularity in the last few decades, the focus shifted mented by a person’s personal or demographic information
towards digitizing the whole process of fingerprint acquisition as well. Each person will be assigned a unique ID. This ID
and automation of the fingerprint matching techniques incor- may be the person’s National Identity Card’s number or a
porating them in an end-to-end system known as Automated company assigned ID. Moreover a person’s facial image may
Fingerprint Identification System (AFIS). also be captured which can also be used for identification
Initially the fingerprints were recorded in the JPEG format purposes later. Similarly a person’s signature could also be
(10 MB file for each finger print). In late eighties US Defense enrolled. Therefore AFIS will not only support the enrollment
organizations overcame this hurdle with a solution based on of fingerprint related information of a person but also other
Fouriers Transformation theory or Wavelets [8]. They were types of biometric information such as facial image and
able to reduce down the size of the file from 10 MB to demographic information such as gender, date of birth, address
500 KB or less using the technique called WSQ (Wavelet etc related to him/her.
Scalar Quantization) [10] that stores the image in wavelet co- 2) Verification: It is the process of verifying the claim of
efficient (similar to one we use in neural image processing identity of a subject [13]. The information uniquely identifying
inside the brain). Most of the AFIS in the world use WSQ a subject such as the ID would be given as an input to
or its subsequent versions for storage of fingerprint scan files. the system together with the fingerprint information. The
Computerization and automation of fingerprint scanning and system would perform the feature extraction and template
matching produce very fast and accurate results unparalleled creation process. After that it would use a (classification
to manual systems in terms of both speed and accuracy. Au- and/or) matching algorithm to match the input template with
tomated Fingerprint Identification is a process of identifying the reference template stored against the unique identifier of
someone based on his/her fingerprints. the subject in it databases. The result of the matching process
The development of AFIS not only met the requirements would be either a Match or a No Match.
in terms of efficiency and accuracy in the domains for which
manual fingerprint matching was used previously such as crim- 3) Identification: It is the process of determining the iden-
inals record keeping etc, its role in these domains became more tity of a subject using only the biometric i.e. fingerprint infor-
and more dominant. A critical development was that it also mation provided to the system. The system is only provided
created a huge market for such systems to be used in various with a scan, it would perform the feature extraction and
other domains previously unthought-of. These domains span template creation phases and then would match the template
from security access or maintenance of personnel attendance with the ones (all or a subset) stored in AFIS databases. The
records for a small scale organization on one end to the record output would either be an empty list or information related
keeping of whole population at the other end. to one (or more) matches. Since this is a computationally
expensive process a pre-selection or pre-filtering commonly
known as classification of the input pattern could be performed
III. C HARACTERISTICS OF AN AFIS S YSTEM
as mentioned before to reduce the search space after which the
While selecting the best AFIS solution one must consider pattern will undergo the matching process with the patterns in
the characteristics like speed, accuracy, database size, hard- the reduced search space.
ware and software dependencies, licensing, transaction time, AFIS will have to perform the above processes using
related data storage, storage formats, encryption techniques, both offline and online data acquisition techniques. Online
security, flexibility to integrate with other law enforcement techniques refer to those in which digital fingerprint data is
systems, easiness and user friendliness, quality and statistical acquired using live fingers. Whereas off-line techniques refer
reports, compliance with ISO/NIST standards and the reliabil- to those in which fingerprint information is captured from
ity of the overall solution. latent prints or from ten-print cards typically used by law
An AFIS system should be able to perform following three enforcement agencies. To capture such data a different set of
main functions: scanners are required.

41
 IV. AFIS S TANDARDS case of fingerprint matching [13]. A lot of research has
The AFIS system should be able to support the following been carried out in improving the efficiency and quality of
data encoding standards for various types of data it is supposed each of these processes as mentioned by Davide et al. in
to handle: [13]. A lot of these techniques for each of the processes
mentioned above have been implemented in the open-source
• ISO/IEC 19794-2:2005 (Fingerprint Minutiae Data)
domain and are available for free in the form of open-source
• ISO/IEC 19794-4:2005 (Finger Image Data)
SDKs (or corresponding application programming interfaces
• ANSI/INCITS 378-2004 (Finger Minutiae Format for
i.e. APIs). Similarly there are many commercial organiza-
Data Interchange)
tions which have provide such SDKs (or APIs) providing
• ANSI/INCITS 381-2004 (Finger Image-Based Data In-
required functionality to perform/implement all the processes
terchange Format)
related to fingerprint enrollment and matching. Using such
The AFIS system should also support the following match- commercial/non-commercial SDKs (or APIs) custom designed
ing techniques: AFIS solutions can be easily developed for a variety of
• Latent print to latent print domains. Then there are organizations which provide end-
• Ten-print to latent print to-end biometrics (mainly fingerprints) based solutions with
• Latent print to ten-print complete functionality and user interface. These solutions
• Ten-print to ten-print can be customized according to the client’s needs. In this
• Live print to latent print section we give a brief overview of such commercial and
• Latent print to live print non-commercial SDKs which are available together with end-
• Ten-print to live print to-end commercial products. We begin with open-source/non-
• Live print to ten-print commercial tools. First is the SourceAFIS which is an open-
• Live print to live print source SDK for .Net developers (an experimental version is
Latent prints are the prints taken from the crime scene. Ten- also available for Java) [6]. It provides an easy-to-use API with
prints are taken from the arrested criminal on Ten-print card detailed documentation for C#/.Net and experimentally for
and then further scanned to the system. Live prints are taken Java. It provides functionality for both fingerprint verification
from the arrested criminal on fingerprint scanners. Latent print and identification [6]. It has good accuracy (independently
to latent print matching is a search of a latent from a crime tested accuracy of 3.6% EER (equal error rate)) and has a
scene against latents on file from other crime scenes. Ten-print matching speed of 10,000 fingerprints/second. It can accept
to ten-print matching is a search of a set of known fingerprints and export ISO/IEC 19794-2 templates. It also includes tools
(ten-prints) against an existing ten-print database. Ten-print to for fingerprint visualization and analysis. The only shortfall
latent print matching is a search of a new ten-print addition is that it does not have an acquisition module both from
to the database against all unsolved latent prints in file. AFIS live scanners and from standard ten-print cards and takes
matching system should be compatible to all types of input fingerprint images as input [6]. Next is the BioMetric SDK
prints. that can match (identify) people through their fingerprints [7].
The other important feature is how many different queries It is currently available in C# though the author mentions that
the system can support at one time, and what is the con- his goal is to make this SDK available in C++ and Java as
nectivity and bandwidth requirements in case of a remote well. It also does not have a fingerprint acquisition module
system connects to the central site. Number of users who and takes fingerprint images as input [7]. Next is the Free
can connect at parallel, disaster recovery, management, and Fingerprint Verification SDK provided by Neurotechnology for
planning should be given due consideration as well. Figure C#. Though it has live fingerprint acquisition and verification
1 gives an overview of the workflow of an AFIS system capabilities, it otherwise provides very limited functionality for
deployed in a typical Law Enforcement agency. independent storage of acquired images and allows storage of
data related to only ten users at a time [1].
V. R EVIEW OF AFIS SDK S Tables I presents a comparison of the various features
Globally AFIS has a 2.6 Billion USD market with an supported by the non-commercial and commercial fingerprint
average growth of 20% per year. The biometrics market related SDKs reviewed above. The features include processing
(other than AFIS) for attendance and authentication services (enrollment and matching etc) support for fingerprints obtained
is another 1.34 Billion USD with an annual growth of 31%. via various means such as from live-scanning (support for live-
Given the cost and the requirements of growing population scanning means support for both the method of acquisition and
and technological savvy inventions, more than 30 countries for images obtained from this method), from ten-print cards
are developing their own AFIS systems. and latent-prints which are obtained from crime scenes etc.
Fingerprint enrollment and matching involve many pro- Similarly the ability of the SDKs to support both the rolled
cesses which include image acquisition followed by fingerprint and flat fingerprints is also shown. These tables also present a
segmentation, image enhancement, feature extraction, template comparison of the support provided by the SDKs for the two
creation, fingerprint classification, template storage (in the most commonly used programming languages/frameworks i.e.
case of fingerprint enrollment) or template matching in the .Net and Java, the two most commonly used operating systems

42
 Fig. 1. Sample AFIS setup

i.e. Windows and Linux, the most commonly used fingerprint fingerprint acquisition and matching but not the fingerprint
image compression method i.e. WSQ compression and one of extraction from the standard ten-print cards or latents. WSQ
the most commonly used databases i.e. Oracle. SDK allows one to add compression, decompression and
fingerprint image exchange capabilities into your AFIS systems
TABLE I [5]. Griaule AFIS ActiveX library is supported by all IDEs that
C OMPARISON OF VARIOUS FEATURES SUPPORTED BY THE REVIEWED
NON - COMMERCIAL AFIS RELATED SDK S . fully support ActiveX. It was fully tested with Visual Basic 6,
Delphi and Visual Studio .NET, which are the most commonly
Supporting Features SourceAFIS BioMetric FreeFingerprint used IDEs for Windows. It should also work with MATLAB,
LiveScan  LabView, PowerBuilder, Ruby and VBScript etc [5]. Cost of
Ten-print Griaule AFIS SDK is $450.
Latent ExpressID AFIS by Innovatrics (Slovakian) provides a
Flat    matching server which can be composed of several nodes.
Rolled Each node can match upto 150,00,00 fingerprints/second. It
WSQ also provides an SDK for development of client-side applica-
Dot Net    tion with customer specific GUI [3]. Innovatrics provides two
Java   more SDKs with the name IDKit PC SDK for PCs and IDKit
Windows    Mobile SDK for mobiles [3]. IDKit PC SDK provides func-
Linux  tionality for fingerprint identification with a matching speed
of 200,000 fingerprint/second. IDKit Mobile SDK provides
Next we discuss the commercially available AFIS related the facility of matching 40,000 fingerprints/second on Intel
SDKs. First are the SDKs provided by Griaule (a company (Marvel) PXA270 @ 624MHz [3]. All the three SDKs support
based in USA) i.e. Fingerprint SDK, AFIS SDK and WSQ SDK wide range of live scanners. It was mentioned that ExpressID
[5]. Fingerprint SDK enables a wide range of applications to AFIS can handle a database of fingerprints from ten-print cards
use fingerprint recognition. Software developers can integrate but we did not find a mention of any extraction/detection
biometrics into their software as programming languages, functionality for fingerprints from ten-print cards. ExpressID
stand-alone applications and web applications as needed - AFIS does not support palmprints and does not support
all through a single, intuitive interface [5]. AFIS SDK helps matching of partial, latent prints from crime scenes, against
to build an Automated Fingerprint Identification System by the database of ten-prints [3]. The cost of ExpressID SDK is
providing extended feature detection/extraction functionality, $1273.
manual image analysis functionality and compatibility with MegaMatcher SDK is a Neurotechnology (USA based) prod-
a wide range of sensors. Both of theses SDKs support live uct.The SDK is available in a Standard and an Extended ver-

43
sion. After live-scanning template creation takes 0.2-0.25 sec- TABLE II
C OMPARISON OF VARIOUS FEATURES SUPPORTED BY THE REVIEWED
onds. The Standard version provides the facility of matching COMMERCIAL AFIS RELATED SDK S .
35 million fingerprints/second. The Extended version provides
the matching capability of 100 million fingerprints/second [2]. Supporting Features Griaule ExpressID
The cost of MegaMatcher SDK is $3473. LiveScan  
Sonda Tech is a Russian company which provides two Latent  
commercial fingerprint SDKs. First is the Sonda Forensic SDK Ten-print  
1.0 which is specially designed for development of Biometric Flat  
Civil and Criminal applications in Windows environment Rolled  
[4]. Second is the Sonda SE SDK 3.2 which also provides Palm  
fingerprint identification facility but is not specific to Civil Crime Scene  
and Criminal applications. Sonda Forensic SDK 1.0 provides WSQ  
processing of fingerprints, latents and palms for extraction Dot Net  
of general and particular features (and construction of the Windows  
template) [4]. It also provides estimation of image quality, Java  
and comparison of all types of fingerprints and palm-prints Linux  
information: ten-print to ten-print, latent to ten-print, ten- Auxiliary Images 
print to latent, latent to latent, live to live etc. Moreover it Interoperability  
provides the functionality of automatic segmentation/detection
Adaptability 
of fingerprints in the ten-print cards including the rolled prints
1-to-1 Fingerprint  
and also from the slap-prints (complete four-finger and palm
1-to-N Fingerprints  
prints). All in all it provides a comprehensive set of functions
Ten-print - Latent  
which can be easily used to build an efficient and automatic
Latent - Ten-print  
Criminal or Civil AFIS. Moreover Sonda Tech provides end-
Ten-print - Ten-print  
to-end AFIS solutions as well customized according to the
Latent - Latent  
client’s requirements [4].
Configurable Template Size  
Table II and III indicate the support provided by the SDKs
Other Devices Support  
for matching of fingerprints acquired via various means with
each other e.g. matching of Latent to Ten-print etc.
VI. D ESIGN AND DEVELOPMENT PHASES OF AN AFIS tailed study of the state-of-the-art research in the domain of
Following is a sequential list of phases and the activities fingerprint recognition and selection of best techniques for
associated with each of them required for the development of image enhancement, feature extraction, template creation and
an end-to-end, holistic and largescale AFIS application: fingerprint matching. An efficient and accurate algorithm can
be proposed on the basis of the research on existing AFIS
A. Phase-1: Problem analysis, research and planning
algorithms. AFIS complete solution includes following steps:
Following activities are performed in Phase-1: Initial anal-
• Image Acquisition
ysis of the problem and a broad activity plan is carried out.
Team setup and responsibilities to each member are assigned. Fingerprint image can be acquired by live scanning or scanning
The detailed study of the available fingerprint capturing and ten-print cards. It includes the development of fingerprint
matching technologies and their associated Software Develop- capture algorithm that can capture image from fingerprint
ment Kits (SDKs) if available is carried out. Analysis includes scanners and also scans image from ten-print cards. This step
the survey of the fingerprint capturing devices i.e. fingerprints also includes the quality check of image. As these images are
scanners and survey of live scanning and offline scanning of stored and then further processed for matching.
fingerprints. Plan of AFIS software development is formulized. • Image Enhancement
Complete documentation of AFIS system such as system Image enhancement is an essential step in the development
architecture plan and specification (including specifications of of AFIS solution. To increase the speed of AFIS system,
all hardware components to be purchased for the system such fingerprint images are enhanced using different algorithms.
as, biometrics sensors. Research phase also includes research
• Features Extraction
on recent algorithms, open source SDKs, sensors and short
listing of useful open source SDKs which could provide useful In the step of feature extraction, fingerprint features are
APIs for the development of complete solution. extracted from the enhanced image. Matching is performed
on the extracted features.
B. Phase-2: Design, development of fingerprint algorithms
• Matching
and their fine-tuning
The final step of AFIS solution is matching of fingerprints.
Following activities will be performed in Phase-2: Design
Matching algorithm should be able to match between different
and development of fingerprint algorithms includes the de-
types of images i.e ten-print to latent print, ten-print to live

44
 TABLE III will also be developed followed by the development of an
C OMPARISON OF VARIOUS FEATURES SUPPORTED BY THE REVIEWED
COMMERCIAL AFIS RELATED SDK S .
automated test-bed for integrated testing.

Supporting Features MegaMatcher SondaTech

LiveScan  
Latent  
Ten-print  
Flat   E. Phase-5: Performance evaluation and system tuning
Rolled  
Palm  
Crime Scene  
WSQ  
Following activities will be performed in Phase-5: Matching
Dot Net  
performance and speed evaluation using standard measures
Windows  
such as FAR, FRR, EER. Integrated system architecture design
Java  
document preparation. It includes the preparation of user
manual of system interface preparation, installation guide,
Linux  
developers guide, technical report along with performance
Auxiliary Images  
analysis and comparison with other competitive biometric
Interoperability  
based personal identification systems using standard databases.
Adaptability 
After quality assurance of product AFIS system is ready for
1-to-1 Fingerprint  
final release.
1-to-N Fingerprints  
Ten-print - Latent  
Latent - Ten-print  
Ten-print - Ten-print  
Latent - Latent  
Configurable Template Size  
Other Devices Support   VII. E AFIS

print etc. The speed and accuracy of matching algorithm is

very important in the AFIS system.
In this section we present the work done on eAFIS (Efficient
C. Phase-3: Database design and development Automated Fingerprint Identification System) for Civil and
Criminal applications. The idea was to build a prototype
Following activities are performed in Phase-3:
AFIS solution based on open-source AFIS related technolo-
Brainstorming and the resultant creation of the dataflow
gies. The system itself is developed in the .Net environment
diagrams based on the System architecture and interface spec-
using C# as the programming language. We have used the
ification document developed in the first phase. It includes the
fingerprint processing algorithms through their corresponding
design of the tables, structures and the relationships between
implementations provided by the SourceAFIS SDK (and its
the tables. Creation and deployment of the database is done on
corresponding API) [6]. We have used OpenCV functions [9]
the server side and its configuration on the client side. It also
for implementing various image processing techniques via a
involves the creation of data security module which includes:
wrapper for .Net known as Emgu CV [11]. We have used
• Selection of sensitive data and encryption and decryption functions from AForge.Net library as well for blob-detection
on server as well as client side [12].
D. Phase-4: System integration, prototype development and eAFIS is able acquire fingerprints via live-scanning and also
testing via standard ten-print cards. We have developed a technique
Following activities are performed in Phase-4: to automatically extract/segment fingerprints from a ten-print
card. For now eAFIS can only perform verification of subjects
Fingerprint application developed in the second phase and
based on the unique user ID and his/her fingerprint informa-
the database created in the third phase are integrated in Phase
tion. Also eAFIS for now can only perform live-print to live-
4. Automated test-bed is designed for integrated testing using
print and ten-print to ten-print matching. Moreover eAFIS also
biometric databases. In this phase integration of the fingerprint
enrolls demographic information of a subject using basic input
application developed in the second phase and the database
devices like a mouse and a keyboard. eAFIS also enrolls mug-
created in the third phase will be carried out resulting in first
shot of a subject using face detection based on Haar features
complete prototype of the AFIS. This will be followed by
and classifiers [11]. The main features of eAFIS include:
interface testing. An integrated test plan for the system testing

45
 • Manual entry of demographic information as shown in
Fig. 4

Fig. 2. eAFIS Login Form

• Administrator/User Login as shown in Fig. 2 to restrict Fig. 5. Live fingerprint and face acquisition
the access to the application to only the people with
access permissions. After User Login the main screen
appears as shown in Fig. 3 from which the user can select
an operation he/she wants to perform
• Fingerprint information entry via live-scanning (with the
option of acquiring single, dual or ten fingerprints) as
shown in Fig. 5
• Fingerprint information entry via ten-print cards:

Fig. 3. Main Form

Fig. 6. Fingerprint acquisition via ten-print cards
• Enrollment

– Loading a ten-print card image through live-

scanning or directly from the memory, the ap-
plication automatically performs fingerprint de-
tection/segmentation based on red color detec-
tion, variance based thresholding (as done by
Raymond[15][14] ) and blob detection as shown in
Fig. 6
– Facial Image capture through face (and eyes) detec-
tion using Haar features [11] as shown in Fig. 5

Fig. 4. Manual entry of demographic information

• Verification

46
 Fig. 10. Searching based on user ID
Fig. 7. Verification based on live fingerprint scanning

• User search based on ID as shown in Fig. 10

VIII. S UMMARY
This paper presents a brief overview of the historical
background of the advent of Automated Fingerprint Identi-
fication Systems. After that it goes through the salient fea-
tures/characteristics a modern AFIS solution should incorpo-
rate. Then it provides an overview of the global landscape of
AFIS solutions and related products both commercial and non-
commercial. After that this paper provides details of the phases
required to build a large-scale and an end-to-end AFIS solution
incorporating most of the features mentioned before. In the end
this paper presents a prototype AFIS application called eAFIS
which is developed based on open-source fingerprint and
image processing related tools and technologies. The salient
features of eAFIS include live fingerprint scanning/acquisition
and verification based on these fingerprints, automatic de-
tection/extraction of fingerprints from a ten-print card and
Fig. 8. Verification Process Result verification based on these fingerprints, enrollment of mugshot
of a subject, manual entry of demographic information via
default input devices.
– Based on ID and fingerprint matching through live- R EFERENCES
scanning as shown in Fig. 7 and Fig. 8 [1] Free FingerPrint Verification SDK, 1998-2012.
[2] MegaMatcher SDK: NeuroTechnology, 1998-2012.
[3] ExpressID: ID Kit Mobile SDK: ID Kit PC SDK, 2004-2012.
[4] SondaTech SDK: Sonda SDK PE 1.2: Sonda SDK Forensic 1.0, 2006-
2012.
[5] Griaule AFIS SDK, 2011.
[6] SourceAFIS, 2011.
[7] Biometric SDK, 2012.
[8] FBI Biometric Specifications (BioSpecs), 2012.
[9] G. Bradski. The OpenCV Library. Dr. Dobb’s Journal of Software
Tools, 2000.
[10] Federal Bureau of Investigation Criminal Justice Information Ser-
vices Division. WSQ Gray-Scale Fingerprint Image Compression
Specification, 2010.
[11] Christopher Johnson. Emgu CV, 2008-2012.
[12] Andrew Kirillov. Aforge.Net.
[13] Davide Maltoni, Dario Maio, Anil K. Jain, and Salil Prabhakar. Hand-
book of Fingerprint Recognition. Springer Publishing Company, Incor-
Fig. 9. Verification based on ten-print cards porated, 2nd edition, 2009.
[14] B. Mehtre. Fingerprint image analysis for automatic identification. Ma-
chine Vision and Applications, 6:124–139, 1993. 10.1007/BF01211936.
[15] Raymond Thai. Fingerprint Image Enhancement and Minutiae Extrac-
tion, 2003.
– Based on ID and fingerprints matching via ten-print
cards as shown in Fig. 9

47

View publication stats