Chapter Four Experimental Results and Discussion

Republic of Iraq Ministry of Higher Education And Scientific Research University of Technology Department of Computer Science

Fingerprint Identification and Verification System Based on Extraction of Unique ID A THESIS SUBMITTED TO THE DEPARTMENT OF COMPUTER SCIENCE OF UNIVERSITY OF TECHNOLOGY IN A PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN COMPUTER SCIENCE

By Esraa Qasim Naamha Supervised by Prof. Dr. Abdul Monem S. Rahma

2017

1438

‫سورة القيامة‪ /‬االية )‪)18-1‬‬

Dedication

This thesis is proudly dedicated to all member of my beloved family My father My mother My husband My sisters and brothers My children And all my friends Thanks for your endless love, sacrifices, prayers, supports and advice

Esraa Qasim

i

Acknowledgments

After thanking GOD, I would like to express my sincere gratitude and appreciation to my supervisor Prof. Abdul Monem S. Rahma for the continuous support of my MSC study and research, for his patience, motivation, enthusiasm, and immense knowledge. His guidance helped me all the time of research and writing of this thesis. I could not have imagined having a better advisor and mentor for my MSC study.

I am also grateful to all Master course staff members in the Department of Computer Science and library staff, for their various forms of support during my graduate study.

My great and deep thanks and appreciation go to my family members for their great patience and encouragement throughout the years of my study.

Special thanks and appreciation go to everyone who has helped me throughout the time of this research.

Esraa Qasim

ii

Abstract A reliable personal identification is critical in many daily transactions. For examples, access control to special locations and access to secret information in the computer are becoming increasingly important to prevent their abuse. Personal identification numbers, passwords, smart card are some of the means normally employed for this purpose. These means do not really identify a person, but only knowledge of some data or belonging of some determined object. Therefore, many of these limitations can be eliminated by incorporation of better methods that can be used to verify identity by measuring and analyzing human characteristics instead of things the individual may have (smart card) or know (password). The main goal of this thesis is using the fingerprint technology to generate the unique ID based on the core point of the fingerprint of an individual that can be used for identifying person identity. The fingerprint’s minutiae features are extracted with the core point as the reference based on the seven invariant moments, then the extracted features are converted into numerical value, this numerical value is used as the unique ID for printing in the security documents for the personal identification and verification. If the fingerprint image does not contain the core point such as the plain arch and tented arch patterns then must create the center point in this fingerprint image and this is considered as the core point. The minutiae features will be computed with this point as the reference point and this is used to generate a unique ID. Implementation of the proposed system shows that the system has satisfactory performance parameters.

iii

List of Contents Subject

Page No.

Dedication Acknowledgements Abstract Table of Contents List of Figures List of Tables List of Abbreviations

i ii iii iv vi viii x

Chapter One: General Introduction and Survey 1.1 Introduction 1.2 Literature Survey 1.3 Aim of Thesis 1.4 Layout of Thesis

2 5 9 9

Chapter Two: Biometric System 2.1 Introduction 2.2 Biometrics 2.2.1 Overview of Biometrics 2.2.2 Biometric System 2.2.3 Operation of the Biometric System 2.2.4 Advantages and Disadvantages of Biometric Techniques 2.2.5 Classification of Biometrics 2.2.6 Biometric Characteristic 2.2.7 Features of Biometrics 2.2.8 Comparison of Various Biometrics 2.2.9 Applications of Biometric Systems 2.3 Fingerprint Biometrics 2.3.1 Overview of Fingerprint Biometrics 2.3.2 Fingerprint Recognition 2.3.3 Techniques for Fingerprint Recognition 2.3.4 Stages of Fingerprint Recognition 2.3.4.1 Fingerprint Image Acquisition 2.3.4.2 Fingerprint Image Pre-processing 2.3.4.3 Fingerprint Image Feature Extraction iv

11 11 11 12 14 15 16 16 17 18 23 24 24 28 29 30 31 33 37

Subject

Page No.

2.3.4.4 Fingerprint Matching Algorithm 2.4 Fingerprint Image Database

42 43

Chapter Three: Unique ID Generation Using Fingerprint 3.1 Introduction 3.2 Fingerprint Image Database 3.3 Proposed System stages 3.3.1 Feature Extraction 3.3.1.1 Standard Fingerprint Features Generation Phase 3.3.1.2 Unique ID Template Generation Phase 3.3.2 Fingerprint Matching Algorithm

48 48 49 51 54 59 73

Chapter Four: Results of Implementation 4.1 Introduction 4.2 The Basic System Requirements 4.2.1 Hardware Requirements 4.2.2 Software Requirements 4.3 Experimental Results of the Proposed System 4.3.1 Feature Extraction 4.3.2 Fingerprint Matching Algorithm 4.4 Analysis and Discussion 4.5 Time Measurement 4.6 Comparison with Related Works

76 76 76 76 77 77 112 113 117 118

Chapter Five: Conclusions and Suggestions for Future Work 5.1 Conclusions 5.2 Suggestions for Further Work

120 121 124

References

v

List of Figures Figure No.

Caption

Pages No.

(2-1) (2-2) (2-3) (2-4) (2-5) (2-6) (2-7) (2-8) (2-9)

Enrollment, verification, and identification Examples of biometric characteristics Ridges and valleys on fingerprint image Minutiae features In fingerprint Fingerprint loop pattern Fingerprint whorl pattern Fingerprint arch pattern Fingerprint recognition system Definition of minutiae points Sample fingerprint image of database from FVC2000 database Sample fingerprint image of database from FVC2002 database Sample fingerprint image of database from FVC2004 database Sample fingerprint image of database from Neurotechnologija database Sample fingerprint image of UPEK fingerprint database Fingerprint's features samples Fingerprint ridge patterns and minutiae Overview of the proposed system Flow Chart of the feature extraction Sample images of fingerprint's features image Binarized samples images of fingerprint's features image Sample of fingerprint image binarazation Windows in first region of fingerprint image How mask (block) slides across the fingerprint image Original fingerprint image for 4 different persons The output after applying fingerprint image binarization

15 22 24 25 26 27 28 30 38

(2-10) (2-11) (2-12) (2-13) (2-14) (3-1) (3-2) (3-3) (3-4) (3-5) (3-6) (3-7) (3-8) (3-9) (4-1) (4-2)

vi

43 44 45 46 46 49 50 51 53 56 56 59 61 66 88 88

Figure No.

Caption

Pages No.

(4-3)

Features extraction results in fingerprint image The region of interest in fingerprint image for person 1 and 2 Illustration of circles in a region of interest for person 1 and 2 The region of interest in fingerprint image for person 3 and 4 Illustration of circles in a region of interest for person 3 and 4 False Acceptance Rate True Acceptance Rate False Rejection Rate True Rejection

99

(4-4) (4-5) (4-6) (4-7) (4-8) (4-9) (4-10) (4-11)

vii

103 103 111 111 115 115 116 116

List of Tables Table No.

Caption

Pages No.

(1-1) (2-1) (2-2) (2-3) (2-4) (2-5) (2-6) (2-7) (2-8)

Identification methods Advantage and disadvantage of Biometrics Physiological and Behavioral Characteristics Comparison of various biometric FVC2000 Fingerprint Database FVC2004 Fingerprint Database FVC2004 Fingerprint Database Neurotechnologija Fingerprint Database UPEK Fingerprint Database White ridge width and black ridge width in each Window of the first region from four directions Histogram of white ridge width and black ridge width for first region Binarized samples images of core feature Binarized samples images of delta feature Binarized samples images of ridge bifurcation feature Binarized samples images of ridge ending feature Seven invariant moments, max, min and mean of the core feature Seven invariant moments, max, min and mean of the delta feature Seven invariant moments, max, min and mean of the ridge bifurcation feature Seven invariant moments, max, min and mean of The ridge ending feature Variances of the core feature Variances of the delta feature Variances of the ridge bifurcation feature Variances of the ridge ending feature White ridge width, black ridge width and Total ridge width of the (ROI) image for 4 persons Fingerprint image partition pixel by pixel for 4 persons Seven invariant moment values of each block for person 1

3 15 16 23 43 44 45 46 46

(3-1) (3-2) (4-1) (4-2) (4-3) (4-4) (4-5) (4-6) (4-7) (4-8) (4-9) (4-10) (4-11) (4-12) (4-13) (4-14) (4-15)

viii

62 64 77 78 79 79 81 82 83 84 85 86 86 87 89 90 91

Table No. (4-16) (4-17) (4-18) (4-19) (4-20) (4-21) (4-22) (4-23) (4-24) (4-25) (4-26) (4-27) (4-28) (4-29) (4-30) (4-31) (4-32) (4-33) (4-34) (4-35) (4-36) (4-37) (4-38) (4-39) (4-40)

Caption Seven invariant moment values of each block for person 2 Seven invariant moment values of each block for person 3 Seven invariant moment values of each block for person 4 Variances of each block for person 1 Variances of each block for person 2 Variances of each block for person 3 Variances of each block for person 4 Largest matches of feature points for person 1 Largest matches of feature points for person 2 Largest matches of feature points for person 3 Largest matches of feature points for person 4 Distance between core point and minutia points for person 1 Distance between core point and minutia points for person 2 Unique ID for person 1 Unique ID for person 2 Distance between minutia points for person 3 Distance between minutia points for person 4 Distance between core point and minutia points for person 3 Distance between core point and minutia points for person 4 Unique ID for person 3 Unique ID for person 4 The results of applying the fingerprint matching algorithm Time for the function in the standard fingerprint features generation phase Time for the function in the unique ID template generation phase Comparison proposed method with related works

ix

Pages No. 92 93 94 95 96 97 98 100 100 101 101 102 103 104 104 105 107 110 110 111 112 112 117 117 118

List of Abbreviations

Abbreviation

Meaning

AFRS AFIS ATM PIN ID ROI RST FAR TAR FRR TR

Automated Fingerprint Recognition System Automated Fingerprint Identification System Automated Teller Machine Personal Identification Number Identification Document Region of Interest Rotation Scale Translation False Acceptance Rate True Acceptance Rate False Rejection Rate True Rejection Radio Frequency Identification Federal Bureau of Investigation Message Authentication Code

RFID FBI MAC

x

Chapter One

General Introduction and Survey

Chapter One

General Introduction and Survey

1.1 Introduction Important confidential documents such as certificates, passports, driver's licenses, and land revenue documents include a set of security qualities that protect them from forging by making that process very difficult. However, as soon as the person attempting forging and has technical practice gains document like that then they are capable of easily indulging in illegal identity theft. It’s not such a hard thing due to fact that this kind of documents contains only name, place of residence, and the individual’s picture for personal identification. Every one of these characteristics and details might be easily copied. Implementing the usage of biometrical characteristics will be helpful in reducing the danger of tampering. The usage of biometrical characteristics in this kind of documents nowadays is not quite regular. However, for the sake of preventing identity theft and avoiding forgery of this kind of documents, the correct solution is using biometrical techniques. Using the fundamental physical and behavioral people's features for the unique identity recognition is known as the biometrics. In general, biometrical techniques use a portion of the person’s body for the sake of identifying an individual and those methods are quite precise. The biometrical features a person possesses and that are extracted have to be somehow transmitted in to the documents. Computer chips and RFID tags were used to store biometrical data. For instance, electronic passports granted by Republic of Germany include an electrical chip that contains the citizen’s biometrical data. Including a document like that which has an RFID tag is quite a complicated and expensive procedure. Moreover, this kind of devices has a very limited lifetime. Therefore, developing new less expensive techniques that have longer lifespan in order to transfer the biometrical data to documents has a high priority from the technological point of view [1]. Many systems require trusted mechanism for identification purpose in order to confirm or to identify person who requests for a specific service. This mechanism is used to ensure that provided a right person accesses service. One of the mechanisms is the biometric recognition system that used human biometric features to provide 2

Chapter One

General Introduction and Survey

personal identification. Typical biometric systems use pattern recognition that takes biometric data from specific individual and then extracts the features of these biometric data, which so called template, and comparing it with other features from the database as reference [2]. Biometrics means measure of unique physiological or behavioral characteristics in order to identify a person. Biometrics is a unique trait which is a part of us, so there is no need of worrying about remembering passwords, or carries any document for identification. Biometric characteristics can be divided in two main types. Physiological character: This is related to the shape of the body and thus it varies from person to person. Examples are fingerprints, face recognition, hand geometry and iris recognition. Behavioral character: It deals with behavior of a person like signature, key stroke dynamics and voice. Behavioral characteristics can change with age [3]. Personal identification are usually divided into three types as shown in Table (11), by what one owns (e.g., a credit card or keys), by something you know (e.g., a password or a PIN code) or by physiological or behavioral characteristics. The last method is referred to as biometrics and the six most commonly used features include face, voice, iris, signature, hand geometry and of course fingerprint identification. It has been established and is commonly known, that everyone has a unique fingerprint which do not change over time. Each person’s finger has its own unique pattern; hence any finger could be used to successfully identify a person [4]. Table (1-1): Identification methods

Method

Examples

What you know

Password, PIN code, user id

What you have

Cards, keys, badges Fingerprint, face, voice, iris, signature, hand geometry

What you are (biometrics)

3

Chapter One

General Introduction and Survey

Some of the varying types in biometrical techniques, like the finger-prints, face, iris, and so on. Finger-prints are the most widely known modality and they have been uses for a long time in real-life legal applications. The researches on automated recognition of fingerprints started in the sixties of the past century, and the produced AFISs (Automated Fingerprint Identification Systems) are used all over the world in a wide variety of applications. Millions of identification systems during a hundred years of actual legal history have definitely proven that finger-prints are individual and unchangeable and therefore that the identification of fingerprints is highly sufficient. The technical improvements of the modern years have made the recognition (in other words one-to-many matching) systems rather inexpensive for daily applications [5]. Finger-prints were submitted officially as an accepted individual identification property in the beginning of the 20th century and since that time they have become a trusted identification technology in legal facilities all over the world. The FBI nowadays is in possession of over 400 million finger-print records on file. Fingerprint techniques have a number of benefits over other biometrical techniques, which are the ones listed below [6]:  High distinctiveness  Easy collectability  Wide acceptability  High permanence  High performance  High universality When a fetus is seven months old, its finger-prints are entirely generated. The fingerprint’s properties don’t change along the human’s lifetime except the cases of wounds, diseases, or decomposition after death. Nevertheless, post a minor wound on the finger-tip, the pattern grows back with the healing of the finger-tip [6].

4

Chapter One

General Introduction and Survey

It’s considered that finger-prints are individual among people and across the fingers of the same person. It’s been proven that even in the case of the identical twins that have the same DNA possess non-identical finger-prints. The usage of fingerprints can be classified into three fields [6]:  Security as recognition of persons.  Legal applications, also as a recognition technique.  Individual properties and dermatoglyphics that are usually involved with horoscopes and similar predictions that have not been proven by science. The first two are by far the major fields. Systems based on finger-prints that are used in security applications, are very widely known nowadays that they have almost become synonymous for biometrical systems [6].

1.2 Literature Survey: Following are some of the past researches related to the subject:Yang J. C. and Park D. S. [7] in 2008 A fingerprint verification system based on a set of invariant moment features and a nonlinear Back Propagation Neural Network (BPNN) verifier is proposed. An image-based method with invariant moment features for fingerprint verification is used to overcome the demerits of traditional minutiae-based methods and other image-based methods. The proposed system contains two stages: an off-line stage for template processing and an on-line stage for testing with input fingerprints. The system preprocesses fingerprints and reliably detects a unique reference point to determine a Region-of-Interest (ROI). A total of four sets of seven invariant moment features are extracted from four partitioned sub-images of an ROI. Matching between the feature vectors of a test fingerprint and those of a template fingerprint in the database is evaluated by a nonlinear BPNN and its performance is compared with other methods in terms of absolute distance as a similarity measure. The experimental results show that the proposed method with BPNN matching has a higher matching accuracy, while the 5

Chapter One

General Introduction and Survey

method with absolute distance has a faster matching speed. Comparison results with other famous methods also show that the proposed method outperforms them in verification accuracy. Seshadri R., et al [8] in 2010 in this research proposed a technique for generating secret key for MAC algorithm with the use of novel method for fingerbased cryptographic system. The key is produced with the use of finger-print patterns that is unchanged along person’s life. Passwords can get hacked by the trial-and-error method. However, breaking the biometrics-based secure system is impossible. Gaddam S. VK, et al [9] in 2011 in this research proposed a sufficient mechanism for cryptographic key-generation from finger-print biometrics with the use of temporary templates. The suggested mechanism is made up of three stages that are: 1) Extracting minutiae points from the image of the finger-print, 2) Generating temporary biometrical templates that have added security and 3) Cryptographic keygenerating from the secure temporary templates. The resulted cryptographic key produced is irrevocable and unrepeated to a specific temporary template, making the producing of new temporary templates and cryptographic keys practical. The experimental results show the efficiency of the temporary templates and the generated cryptographic key. Torres G. A. [10] in 2012 in this research presents a fingerprint recognition method using a combination of the Fast Fourier Transform (FFT) with Gabor filters for image enhancement. Next, fingerprint recognition is carried out using a novel recognition stage based on Local Features and Hu invariant moments for verification. Chaorong L. [11] in 2012 in this research presents a scheme of fingerprint verification based on directional filter banks (DFB) and Hu invariant moments. The proposed scheme uses short time Fourier transform (STFT) to enhance the input fingerprint images and then circular region of interest (ROI) is built based on enhanced image. The circular ROI is divided into non overlapping blocks. After decomposing the ROI by using DFB, the seven Hu invariant moments were

6

Chapter One

General Introduction and Survey

computed from each block of ROI as the feature of fingerprint features. The proposed scheme can improve performance of verification and is more robust with respect to the fingerprint image quality. Abdul-Haleem M. G. [12] in 2014 in this research proposes a fingerprint recognition technique which uses local robust features for fingerprint representation and matching. The technique performs well in presence of partial fingerprints. The adopted local features include features extracted from the detail of Haar wavelet subheads. Experiments are performed using a database of 160 low quality fingerprint images collected from 40 subjects, (i.e., 4 images per subject). The test results indicated good system ability to signify low-quality fingerprint images even with existence of partial loss in fingerprint images. The technique has produced a recognition accuracy of 94.37% using one decomposition level and 96.87% using two decomposition levels of wavelet Rashid M. T., et al [13] in 2014 in this research achieved an individual key generating with the use of finger-print minutiae for the sake of performing high security for networking systems. The issue that was solved in this system is the use of finger-print minutiae for the generation of 1024-bit prime numbers which were utilized in RSA-cipher technique for the generation of a 2048-bit key. The minutiae of finger-print that was deduced from the image of the finger-print with the use of image processing techniques which the resolution, brightness, contrast of finger-print image were altered, moreover, unwanted portions like the noise were eliminated. This technique is easy and can be utilized by various kinds of processors such as the microcontroller, ARM, FPGA, and so on. Sreemathi M., et al [14] in 2014 in this research proposed an innovative model for cryptographic key-generation from the fusion of Electrocardiogram (ECG) and fingerprint using Elliptic Curve Cryptography (ECC). Among all the biometrics, ECG and Finger-print are used for generating the key because the ECG supplies substantial liveliness detecting and fingerprint-based recognition is very scalable.

7

Chapter One

General Introduction and Survey

After preprocessing the biometrical features, the characteristics are deduced. Then those deduced characteristics are fused for generating the cryptographic key. ECC is used as many mathematicians proved that elliptic curve provides the optimal solutions for cryptography. The produced cryptographic key using the suggested model is rather small in comparison with the RSA. Marimuthu M., et al [15] in 2015 in this research proposed a new cryptographic key-generation algorithm from dual finger-print biometrical template; the proposed mechanism has made a simpler generating of the cryptographic keys and reduced the complexity of the older crypto-system. Finger-prints are permanent along the lifetime of the individual. Keys are created from the finger-print template to encrypt and decrypt the user’s contents. This technique may be developed using MATLAB and it produces various size cryptography keys, with fixed amount of time and space complexities. This is proper for every real-time application that establishes secure information exchange among the users. The suggested system is evaluated with the use of publicly available FVC2002 data-base and produced better results. Sujatha E., et al [16] in 2016 in this research supplies the idea of integration of the double encryption cryptographic algorithm and biometrical features with generating guard key for achieving the optimal algorithm that has stronger security. Every non-biometrical algorithm drawbacks like easily lost, shared, stolen, guessed, forged or duplicated are eliminated. It gets rid of DOS, malicious attacks, theft of data or key cyber-crimes. Moreover, it prevents other weaknesses and threats that the secured transmission of data might face. It securely transmits, stores, accesses, manipulates data-content over the net-work using secure and unsecure network paths as well. Therefore, this thesis is a better method to provoke and secure the data on the networks from different attackers. RSA algorithm also makes the system stronger.

8

Chapter One

General Introduction and Survey

1.3 Aim of the Thesis Due to the uniqueness of the finger-prints during the human’s lifetime, fingerprint recognition has been used in a wide variety of applications. Typically, it deals with forensic recognitions and police work; it has now become more widely known in civilian applications as well, like the access control, security of finances and firearm buyers’ verification and driver license applicants. This thesis presents a scheme of fingerprint identification and verification based on unique ID generation for security documents depending on the core point of the person’s fingerprint. The minutiae features of the fingerprint are extracted with the core point as the reference point by using the seven invariant moments. Then the extracted features are converted into numerical value. This numerical value is used as the unique ID. This unique ID can be printed on security documents for preventing criminal impersonations.

1.4 Layout of the Thesis In addition to chapter one a “General Introduction and Survey” which provides a general introduction to biometrics and fingerprints and a literature survey of some projects on the same issue, the other chapters are as listed below:  Chapter Two “Biometric System” presents an Overview and detailed description of the biometric system and its technologies and fingerprint as a case study.  Chapter Three “Unique ID Generation using Fingerprint” describes a new invented method to generate the unique ID using fingerprint image.  Chapter Four “Results of Implementation” introduces a complete description of the proposed method results and the performance evaluation of the proposed method.  Chapter Five “Conclusions and Suggestions for Future Work” presents the derived conclusion and given some suggested ideas for future work.

9

Chapter Two

Biometric System

Chapter Two

Biometric System

2.1 Introduction This chapter contains the overview of different approaches and mechanisms which have been implemented for utilization in fingerprint identification and verification system based on extraction of unique ID. This is performed for the reason to provide directions to specific design decisions for future use throughout the project.

2.2 Biometrics The word “biometrics” has Greek origin, and it is a combination of the two ancient words “bios” (which means life) and “metron” (that means measurement); biometric identifiers are measurements of living human body [17].

2.2.1 Overview of Biometrics Biometrics which refers to identifying an individual based on his physiological or behavioral characteristics (identifiers), relies on “something which you are or you do” to make a positive personal identification. It is inherently more reliable and more capable than knowledge based and token-based techniques in differentiating between authorized person and a fraudulent impostor, because the physiological and behavioral Characteristics are unique to every person. Also the person to be identified is required to be physically present at the point of identification .Biometrics provides a solution for the security requirements of our electronically interconnected information society has the potential to become the dominant automatic personal identification in the near future [18].

11

Chapter Two

Biometric System

2.2.2 Biometric System A biometric system is essentially a pattern recognition system that acquires biometric data from an individual, extracts a salient feature set from the data, compares this feature set against the feature set(s) stored in the database, and executes an action based on the result of the comparison. Therefore, a generic biometric system can be viewed as having four main modules: a sensor module; a quality assessment and feature extraction module; a matching module; and a database module. Each of these modules is described below [19]: 1. Sensor module: A proper biometrical scanner or reader has to collect the basic biometrical data of a person. For the goal of obtaining fingerprint images, for instance, an optical sensor of fingerprints can be utilized for capturing the friction tops of the curves construction of the fingertip. The human machine interface is defined by the sensor module and therefore, it is substantial for the implementation of the biometrical system. An interface that is poorly designed may cause a deep failure to success ratio and, therefore, it may result in low user acceptability. And because most of the biometrical patterns are taken as images (except for voice that is based on audio and fragrance that is based on chemical structure), the goodness of the initial biometrical data is affected by the features of the camera that is used in the capturing process as well [19]. 2. Quality assessment and feature extraction module: The goodness of the biometrical data that is collected by the sensors is firstly evaluated for determining its appropriateness for the next processing. Usually, the data that is being collected undergoes a signal improvement process for the goal of improving its quality. Nevertheless, in some of the situations the data’s quality might be very low to the point that the person is required to enter the biometrical data once more. The biometrical data is treated afterwards and a group of prominent distinguishing characteristics get deduced in order to demonstrate the underlying feature.

12

Chapter Two

Biometric System

For instance, the index and the direction of minutiae points (local ridge and valley anomalies) in an image of a fingerprint are deduced by the module feature extracting in a fingerprint-based biometrical system. During the time of registration, this characteristic group gets stored in the database and it is usually called a template [19]. 3. Matching and decision-making module: The deduced characteristics are contrasted with the templates that are stored for generating match results. In a fingerprint-based biometrical system, the quantity of identical minutiae between the input and the template characteristics group is identified and a match result is provided. The match result might be smoothed by the quality of the provided biometrical input. Moreover, the matcher module contains a module of decision making, where the matching results are utilized for either the validation of an assumed identity or for providing a rating of the checked in identities for the goal of identifying person [19]. 4. System database module: The database acts as the repository of biometric information. During the enrollment process, the feature set extracted from the raw biometric sample (i.e., the template) is stored in the database (possibly) along with some biographic information (such as name, Personal Identification Number (PIN), address, etc.) characterizing the user. The data capture during the enrollment process may or may not be supervised by a human depending on the application. For example, a user attempting to create a new computer account in her biometric enabled workstation may proceed to enroll her biometrics without any supervision; a person desiring to use a biometric-enabled ATM, on the other hand, will have to enroll her biometrics in the presence of a bank officer after presenting her nonbiometric credentials [19].

13

Chapter Two

Biometric System

2.2.3 Operation of the Biometric System The process of the biometrical system relies on the context of the application. This system might be operating in the following three modules [17, 20], as shown in Figure (2-1): 1. Enrollment: In this step, the biometrical data from the users gets to be stored in the database system. Firstly, a biometrical reader does the scanning of the biometrical feature and generates the initial data form taken from this feature. After that step, a quality checker performs a pre-processing operation on the data in order to ensure the validity of that data. Afterwards, this initial data undergoes the processing by feature extractor for getting an ideal data which is referred to as template. Thus, this template is stored in the database for upcoming usage in the identification or verification steps [17]. 2. Identification: A biometrical identification system is able to investigate the whole biometrical database for determining the existence of any database inputs that match the entry sample. Step by step the template of the database corresponds to the entry sample. The result of the phase of matching is the most identical match to the person’s identity, in the opposite case; the match fails in the case where the person isn’t registered in the database [17]. 3. Verification: A biometric verification system differs from the identification system in that the user’s is compared to a single enrolled biometric entity. Unlike identification system, each template in the system database is stored with a distinct identifier. Therefore the input sample is not only a biometric sample as in the case of identification system, but it is associated to some identifier such as ID, smart cards or usernames. The output of this system is Accept or reject [17].

14

Chapter Two

Biometric System

Figure (2-1): Enrollment, verification, and identification

2.2.4 Advantages and Disadvantages of Biometric Techniques No biometric system has absolute security, however, while comparing it to a username and a password, biometric system still offers a better degree of security [21]. In general, biometric systems possess a list of advantages and disadvantages, as depicted in table (2-1). Table (2-1): Advantage and disadvantage of Biometrics Advantages

Disadvantages

Positive identification

Public Acceptance

You cannot lose , forget , or share your

Legal Issues

biometric information A biometric template is unique to the

Possible increase in hardware costs to

individual for whom it is created

current systems

Rapid identification / authentication

May require large amounts of storage

Costs, in general, are decreasing

Privacy Concerns

15

Chapter Two

Biometric System

2.2.5 Classification of Biometrics Biometrics can be classified basically in two categories namely physiological and behavioral characteristics [22]. Table (2-2): Physiological and Behavioral Characteristics Physiological

Behavioral Commonly Used

Face recognition

Signature

Fingerprint

Voice

Hand geometry

Keystroke

Iris recognition

Gait

Retina

Lip motion

DNA

Body Oder

Thermo grams Ear recognition Rarely Used Skin reflection

Hand grip

Vein pattern

Brain wave pattern

Sweat pores

Foot dynamics

Fingernail bed Footprint

2.2.6 Biometric Characteristics Any human physiological and/or behavioral feature might be utilized as a biometrical feature under the condition that it fulfills the requirements listed below:  Universality: every individual must have the feature.  Distinctiveness: no two individuals can be identical concerning that characteristic.  Permanence: the characteristic has to be highly unchangeable (concerning the matching standard) with the change of time.

16

Chapter Two

Biometric System

 Collectability: the feature might be measured quantitatively. Nevertheless, in a real life biometrical system (in other words, a system that utilizes biometric features for the sake of personal identification), there are several other matters that have to be taken into consideration, that include [23]:  Performance, it means the reachable identification precision and speed, the resources that are needed for achieving the identification precision and speed goal, plus, the operational and environmental operator influences the precision and speed;  Acceptability, refers to the degree to which users are ready to agree to the use of a specific biometrical identifier (feature) in their every day routine;  Circumvention represents to what degree it is easy to fool the system by the usage of fraudulent techniques. A functional biometric system must satisfy the required identification precision, speed, and requirements of the resource, being harmless to the people, is accepted by the intended users, and is efficiently strong against the attempts of different fraudulent techniques and attacks that target the system [23].

2.2.7 Features of Biometrics Biometrical technologies propose a wide range of benefits which other technologies cannot offer [17].  Uniqueness: Mainly, biometric solutions propose a high degree of identity (for example, fingerprints for two twins aren’t identical).  Convenience: people don’t need to memorize a number of long and complicated, repeatedly changing passwords or carry many keys anymore.  Non-repudiation: This advantage makes sure that the person is present at the point and time of identification and can’t renounce having entered the system later.  Non-transferable: That means it can’t be granted, copied, lost, stolen, shared or forgotten in contrast to the passwords, PIN codes, and smart cards.

17

Chapter Two

Biometric System

2.2.8 Comparison of Various Biometrics There are a number of human recognizable features that meet the definition of biometric principles. For the sake of using it for distinguishing a user, the person’s characteristics must be diverse and not being subject to change, for example, fingerprints. They have been used for more than a century of time and, thus, they are usually well established as an identification technique. Different techniques like the hand structure, voice and iris recognition, face are very widely accepted techniques as well. A biometrical system that might ask for a blood sample for repeated personal identification probably wouldn’t be so accepted. Implementation points of view are important. No biometric system is able to guarantee one hundred percent accuracy [24]. Figure (2-2) depicts some of biometrical features that are used in personal identification [23]. A summarized introduction to the widely known biometric systems is given below:  DNA: Deoxyribonucleic acid (DNA) is the one-dimensional (1–D) maximally individual code for a human individuality. There’s an exception, however for the identical twins that have similar DNA code. But still, it is used nowadays especially in the context of forensic applications for person recognition [23].  Ear: Researches have shown that the form of the human ear and the construction of the cartilaginous tissue of the pinna are very individual. The ear recognition techniques depends on comparing the distance of prominent spots on the pinna from a landmark index on the ear. The characteristics of the ear are not considered to be that individual for the establishment of the individual’s identity [23].  Face: The features of the face are the most common characteristics that are used by people in recognizing each other. Face recognition depends on both the form and location of the eyes, eyebrows, nose, lips, and chin or it is based on the general analyzing of the image of the face that represent the face as a number of identified faces. [20].

18

Chapter Two

Biometric System

 Facial, hand, and hand vein infrared thermogram: The pattern of temperature produced by the body of a human is a remarkable feature of a person and can be caught and recorded by the infrared camera in a non-intrusive manner which is remarkably similar to an ordinary (visible spectrum) photographing. This technique might be utilized in covert identification. A similar technique that uses near infrared image capturing is utilized in scanning the back of a clenched fist for determining the vein construction of the hand. Infrared sensor is a rather costly equipment and this is a factor that restricts the wide spread utilization of this technique [23].  Fingerprints: People have been using fingerprints for individual's recognition for many decades and this technique has established its validity and reliability quite well. A fingerprint is the manner of the distribution of ridges and furrows on the fingertip's surface, the structure of which is being created during the period of growth of a fetal. These patterns are very distinctive to the point where even the identical twins have different fingerprints, and the same applies for the prints on every finger of the same individual. [25].  Gait: Gait is the specific manner of the person's walks and it is quite a complicated spatial-temporal biometrical technique. Gait isn’t considered to be so ordinary, but is merely distinguishing for allowing confirmation in the number low-security implementations. Posture is another example of the behavioral biometric technique and may not wait invariant, unusually over a long time interval, because of the variation in the weight of the body, chief grievances involving joints or brain, or because of some intoxication [24].  Hand geometry: There are a number of different measurements of the hand of a human, like its form, and length and width of every finger, these features can be deployed in biometrical system as distinguishing characteristics. Biometrical systems based on the hand structure are now being used at a wide range of places all over the world. This mechanism is rather simple, and quite easy to use, it is relatively not that inexpensive as well. [25].

19

Chapter Two

Biometric System

 Iris: The iris is the circular part of the eye residing between the pupil and the sclera (the white region of the eye) on each of the sides. The iris’s visual texture is formed along the period of the development of the fetal and settles during the first two years of the baby’s life. The complicated texture of the iris contains very discriminatory information beneficial in personal identification. Every iris is unique and, such as the fingerprints, irises of identical twins are not similar. [23].  Keystroke: The way and the technique of typing on computer’s keyboard differ from one person to another. This type biometrical feature isn’t assumed as a distinctive feature but it still might be efficient in some of the implementations. In the use of this technique number of things can be analyzed: the interval between keypressing and key-release, the type of the keyboard that is used, plus, the individual’s emotional and bodily state. Therefore, no special hardware is needed for the keystroke analysis, merely the ordinary keyboard of the computer [20].  Odor: It is common that every object produces a smell which is a feature of its chemical composition and this can be utilized in the recognition of different objects. A hint of air that surrounds the object is spread over a set of chemical sensors, each one of which is sensitive to a specific set of (aromatic) components. It’s still not proven that the distinction in the odor of the body could be caught in spite of the smells of deodorants and perfumes, and different chemical components of the surrounding environment [23].  Palm print: The human hands’ palms consist of a structure of ridges and valleys in a way very similar to the fingerprints. The area of the palm is clearly bigger than that of a finger and, therefore, the palm print is considered to be even more complicated and unique than the single fingerprints. Due to the fact that the palm prints scanner needs capturing quite a large area, it is larger and more costly than the sensor of a fingerprint. A person’s palms contains extra discriminative characteristics as well, for example, the basic lines and wrinkles that might be caught even using scanners that have lower resolutions, which would be less expensive [23].

20

Chapter Two

Biometric System

 Retinal Scanning: This technique of individual authorization is based on the vascular patterns of the eye’s retina. In people that have relatively good health, the vascular pattern in the retina doesn’t change during the person’s life span. The schema undergoes scanning with the use of a light source that has rather low-intensity (such as the near-infrared). It requires the person to look into a device and focus on a specific spot. The capturing of the image includes coordination of the subject, followed by a contact with the eye segments [24].  Signature: Every individual has a specific pattern of handwriting. Nevertheless, no two signatures of one individual are ideally similar; the diversities from an ordinary signature rely on the physical and emotional condition of the individual as well. Two approaches of signature-based recognition exist, they are: static and dynamic. The recognitions of static signatures use only the geometrical (shapes) characteristics of the signature, while the dynamic (or online) signature identifications use both the geometrical (shape) characteristics and the dynamic characteristics like the accelerating, speed, pressing, and the trash profiles of the signature [25].  Voice: The recognition of voice is the recognition of an individual that depends on discriminative features and details of their voice. Features of the voice are the collection of the physical and behavioral biometric. On physical side, voice is unchanged for the individual since it is determined by the size or form of the mouth, lips, vocal tracts and nasal cavities, etc. Nevertheless, on the behavioral side, voice isn’t that stable. It can vary depending on the person's emotions, sickness or age. Because of the behavioral impact, voice identification systems cannot be viewed as a distinguishing biometrical mechanism [20].

21

Chapter Two

Biometric System

Figure (2-2): Examples of biometric characteristics: (a) DNA, (b) ear, (c) face, (d) facial thermo gram, (e) hand thermo gram, (f) hand vein, (g) fingerprint, (h) gait, (i) hand geometry, (j) iris, (k) palm print, (l) retina, (m) signature, and (n) voice

A brief comparison of the above biometric techniques based on seven factors is provided in Table (2-3). The applicability of a specific biometric technique depends heavily on the requirements of the application domain. No single technique can outperform all the others in all operational environments. In this sense, each biometric technique is admissible and there is no optimal biometric characteristic [23].

22

Chapter Two

Biometric System

Universality

Distinctiveness

Permanence

Collectability

Performance

Acceptability

Circumvention

Table (2-3) Comparison of various biometric, High, Medium, and Low are denoted By H, M, L, respectively

DNA

H

H

H

L

H

L

L

Ear

M

M

H

M

M

H

M

Face

H

L

M

H

L

H

H

Facial thermo gram

H

H

L

H

M

H

L

Fingerprint

M

H

H

M

H

M

M

Gait

M

L

L

H

L

H

M

Hand geometry

M

M

M

H

M

M

M

Hand vein

M

M

M

M

M

M

L

Iris

H

H

H

M

H

L

L

Keystroke

L

L

L

M

L

M

M

Oder

H

H

H

L

L

M

L

Palm print

M

H

H

M

H

M

M

Retina

H

H

M

L

H

L

L

Signature

L

L

L

H

L

H

H

Voice

M

L

L

M

L

H

H

Biometric identifier

2.2.9 Applications of Biometric Systems The implementations of biometric systems can be classified into the following three basic sets [23]:  Commercial applications: Their examples include logging into a computer network system, security of the electronic data, electronic commerce, access to the Internet, credit card transactions, controlling physical entrance, ATM, cell-phones, Personal Data Assistants, managing and controlling medical records, and distance learning [23]. 23

Chapter Two

Biometric System

 Government applications: that includes the national ID card, driver’s license, social security, correctional facility, border control, welfare disbursement, and passport control [23].  Forensic applications: that refers to corpse recognition, terrorist identification, parenthood determination, criminal investigation and missing children [23].

2.3 Fingerprint Biometrics The Finger-print biometric feature is a form of the most widely used biometrical techniques, because of its simplicity and acceptability in the everyday life [17].

2.3.1 Overview of Fingerprint Biometrics Finger-print biometric is one of the most commonly used biometrical techniques for the legal documents of evidence worldwide. A finger-print is presentation of the ridges of friction that are found on the internal side of the fingers. A finger-print consists of ridges and valleys, the ridges are represented by the dark region of the finger-print and valleys are represented by the lighter region which resides between the ridges [26], as shown in Figure (2-3).

Figure (2-3): Ridges and valleys on fingerprint image

24

Chapter Two

Biometric System

The harms that affect the fingerprint like minor burns, scratches, or cuts don't impact the deeper structure of the ridges, and the initial pattern is repeated in any new growing skin. Ridges and valleys usually flow in a parallel manner sometimes they split and other times they get disconnected [27]. Ridges may be identified as solitary curve pieces. A group of a number of ridges generates the finger-print pattern. The minor characteristics generated by overlapping and terminating of the ridges are known as minutiae. Ridge terminating and splitting are considered to be the distinguishing characteristics of the finger-print. In this technique the index and angle of the characteristic are acquired for representing the finger-print and utilized in the process of matching. Altogether, the finger-print includes two distinct kinds of characteristics known as the core point and the delta point. The core points are usually considered as referencing points for minutiae coding and they are defines as the top-most points on the inner-most periodic ridges. The core and delta points are known as the “singularity points” as well. The delta points are the center of the triangle where three various direction-flows assemble [26, 28], as shown in Figure (2-4).

Figure (2-4): Minutiae features in fingerprint

25

Chapter Two

Biometric System

Finger-prints follow three basic rules [3]: 1. A finger-print is an individual property; two different people can never have identical fingerprints. 2. A pattern of the finger-print will stay unaltered during the person’s lifetime; on the other hand, the fingerprint can be changed because of a constant scar and some kinds of skin illnesses.

3. Finger-prints have general property patterns of ridge that permit them being systematically predefined. Patterns of fingerprints are split into three basic sets that consist of Arches, Loops and Whorls. About five percent of every fingerprint is made up of Arches, thirty percent from the Whorls and the rest sixty five percent represent the Loops [29]. a. Loop Patterns: In a Loop pattern, the ridges will flow in one side, re-curve, (loop around) touch or pass through an imaginary line drawn from the delta to the core, and exit the pattern on the same side from which it entered. The loop pattern consists of one or more re-curving ridges and one delta. There are two types of loop patterns [29]: 1. Ulnar loop: the ridges flow in from the little finger side [29]. 2. Radial loop: the ridges flow in from the thumb side [29].

Core

Delta Figure (2-5) Fingerprint loop pattern

26

Chapter Two

Biometric System

b. Whorl Patterns: Any fingerprint pattern which contains two or more deltas will be a whorl pattern. A whorl pattern consists of a series of almost concentric circles. There are four types of whorl patterns [29]: 1. Plain whorls: consist of one or more ridges which make a complete circuit with two deltas, between at least one re-curving ridges within the inner pattern area is cut or touched [29]. 2. Central pocket loop whorls: consist of at least one re-curving ridge to the line of flow, with two deltas, between which when an imaginary line is drawn, no re-curving ridge within the pattern area is cut or touched [29]. 3. Double loop whorls: consist of two separate and distinct loop formations with two separate and distinct shoulders for each core, two deltas and one or more ridges which make, a complete circuit [29]. 4. The accidental pattern: will contain two points of delta. One delta will be related to a re-curve and the other will be related to an up thrust [29].

Core

Delta

Figure (2-6) Fingerprint whorl pattern

27

Chapter Two

Biometric System

c. Arch Pattern: In an arch pattern, ridges flow in one side and flow out the opposite side. There are no deltas in an arch pattern. There are two types of arch patterns [29]: 1. Plain arches: have a flow of ridges from one side to the other of the pattern, no “significant up thrusts” and the ridges enter on one side of the impression, and flow out the other with a rise or wave in the centre [29]. 2. Tented arches: have an angle, an up thrust, or two of the three basic characteristics of the loop [29].

Figure (2-7) Fingerprint arch pattern

2.3.2 Fingerprint Recognition Recognizing finger-prints is that procedure that compares questioned and identified registered finger-print with another one for the sake of determining whether the acquired prints are from the exact same finger or palm [30]. Recognizing of the finger-print issue might be divided into three sub-groups: finger-print enrollment, verification and identification of finger-print. The verification step is usually implemented for the positive recognizing, which has the aim of preventing more than one person from using the same identity. Verifying the fingerprint is performed for the sake of the verification of the authenticity of a given individual by his finger-print. There’s a one-to-one comparing in this situation. In the mode of identification, the system identifies a person through the search of templates 28

Chapter Two

Biometric System

of every user registered in the data-base for a template that matches. Thus, the system performs a one-to-many comparing for the sake of establishing the identity of a person [31]. Among all the biometrical technologies, the system of finger-print recognizing is the one of the first systems of recognizing. In the beginning of the 20th century, recognizing of finger-prints was submitted as a form personal system of identification in legal application. After that, various finger-print recognizing technologies, such as latent gain of the fingerprint, classifying fingerprints, and their matching were implemented. Today, automatic recognizing of finger-prints is being developed and implemented day after day not only in legal implementations, but in daily civilian applications as well [20].

2.3.3 Techniques for Fingerprint Recognition Finger-print recognizing can be basically divided into four methods: 1. Techniques based on Minutiae Extraction: The most widely accepted fingerprint scanning technique is Minutiae-based. Techniques that are based on Minutiae generate the finger-print taking its local characteristics, such as termination and splitting. In the case where minutiae points match between two finger-prints that means that finger-print is matching. This method has been extensively researched, and it’s the base of the finger-print recognizing products that are available nowadays [32]. 2. Techniques that are based on Pattern Matching or Ridge Features: Extracting of features is performed on a chain of ridges in contrast to various points that form the basis of the technologies of pattern matching over the Extraction of Minutiae. The points of Minutiae may change by wearing and tearing and the most significant disadvantage is that these are acute to the correct adjustment of the finger and are in need of large storage for templates [32].

29

Chapter Two

Biometric System

3. Technologies Based on Correlation: This type of techniques is used for matching two finger-prints; those finger-prints undergo adjustment and computation of the correlation for every one of the corresponding pixels. They are capable of matching the shapes of ridges, breaks, etc. The most significant drawbacks of this technique lie in the fact of the computational complexity it has and the reduced tolerating to the non-linear distortions and the changing in contrast [32]. 4. Techniques Based on Images: This type of techniques tries to perform a matching operation that depends on the overall characteristics of every finger-print image. It is a developed and recently implemented technique for recognizing fingerprints [32].

2.3.4 Stages of Fingerprint Recognition The Fingerprint recognition systems consist of the following parts [20], as shown in Figure (2-8).  Sensing or Image acquisition  Pre-processing  Extracting of features or minutiae  Matching

Fingerprint sensor

Prepeocessing

Feature Extraction

Figure (2-8): Fingerprint recognition system

30

matching

Chapter Two

Biometric System

In the sensing of fingerprints the finger-print of a person is scanned by a fingerprint scanner for producing an initial digital representation [33]. Preprocessing is a very significant stage for the (Fingerprint Recognition System) FRS. It improves the quality of the acquired image and generates an image where the minutiae may be properly identified. The final result of fingerprint recognition system is based on this stage as well [33]. Characteristic or minutiae extracting stage is involved with improving the acquired image, determining the points of minutiae and finally extracting characteristics from the image [33]. The stage of minutiae matching is involved with the matching of the image template with the image of the input. Template image is gathering along the process of enrolment and it is stored in the data-base. Along the stage of recognizing, the initial image undergoes comparison with the template image. This stage determines if the two images belong to the same finger or not [33].

2.3.4 .1Fingerprint Image Acquisition The goal of image acquisition sub-system may be depicted as the transforming of the data into the form of array of numbers that can be altered by a computer program so the general goal of vision can be reached. For the sake of achieving this goal three main problems have to be handled and these are representation, transduction (or sensing) and digitization [28]. The finger-print images acquisition has been done by two methods: 1. Online Fingerprint Images (Live Scan): For this type of acquisition, capacitance or optical finger-print scanners like the URU 4000, and others. Live scanning scanners provide relatively higher quality of the image that typically has a resolution of 512 dpi, which leads in increased reliability along the process of match compared to the inked finger-prints [34].

31

Chapter Two

Biometric System

2. Offline Fingerprint Images (Inked): In the inked method an imprint of an inked finger is first obtained on a paper, which is then scanned. This method usually produces images of very poor quality because of the non-uniform spread of ink and is therefore not exercised in online [34]. Three kinds of electronic finger-print sensors depend on the sensing technology: 1. Solid-state or silicon sensors: They are the ones made up of an array of pixels, where every one of the pixels is kind of a sensor itself. An individual places the finger on the surface of the silicon, and four methods are usually used for converting the ridge/valley data into the electric signal form: capacitive, thermal, electric field and piezoelectric. Due to the fact that solid-state kind of sensors doesn't use any optical component, their size is relatively small and it permits being easily included. Nevertheless, this kind of sensors is rather costly; therefore, the area of sensing of solid-state sensors is usually small [35]. 2. Optical: The finger touches a glass prism and this prism gets illuminated by diffusing lighting. The lighting reflects at the valley regions and is absorbed at the ridge areas. The reflecting lighting gets focused onto a CCD or CMOS sensor. Optical sensors of finger-print provide fine image quality and wide sensing region but they can’t be miniaturized due to the fact that with the minimizing of the distance between the prism and the image sensor, more optical distorting appears in the given image [35]. 3. Ultrasound: Voice signals get sent, acquiring the signals of echo which are reflected at the surface of the finger-print. The signals of voice are capable of crossing dirt and oils that may exist in the finger, therefore providing good quality images. Nevertheless, scanners of ultrasound are big and rather costly, and take some seconds for acquiring the image [35]. The newer versions of touch-less direct scanning devices which produce a 3-D representation of finger-prints are improving. A number of the finger images are taken from various views that use multi-camera systems, and a contact-free 3-D 32

Chapter Two

Biometric System

finger-print representation is generated. This modern sensor technology avoids a set of the issues that intrinsically appear in contact-based sensors like the incorrect placing of a finger, deformation of skin, sensor noise or dirt [35].

2.3.4 .2 Fingerprint Image Pre-processing Finger-print recognition turns to be a complicated computer task when involved with noise or poor image quality. Pre-processing of fingerprint images is a necessary procedure for the sake of getting a high quality image for more processing. The preprocessing of the finger-print produces an image that has no noise and that provides the accuracy needed. The aim of the preprocessing is to improve the image data that suppresses the undesired distortions or enhances some image features, which are important for further processing. Due to the pre-processing being very beneficial for suppressing data which is irrelevant to the individual processing of images or analyzing procedure [6]. The image that is taken from the sensor of the finger-print usually produces low quality images. This low quality can be the result of the following factors [28]: 1. The distortion due to elastic deformation of the finger. 2. Scratches and rubbing on the finger. 3. Dirt, oil or moist on the finger or on the scanner. 4. Partial imaging of the fingertip prints image with various rotations. 5. Non-uniform contacts between the finger and the surface of the sensor that result in discontinuity in ridges and bridges that connects them. 6. Variation of the gray scale image contrast. For overcoming the above issues preprocessing of the finger-print image is very important for improving the distinctiveness of the ridge structures of finger-print images, maintaining their integrity, avoiding the introduction of spurious structures or artifacts, and maintains the connectivity of the ridges while preserving separation

33

Chapter Two

Biometric System

between the ridges. Finger-print preprocessing indicates that some of the operations must be done prior to the minutiae extraction [36]. a. Fingerprint Image Binarazation Binarization process is an operation that performs converting grey-level images into binary images and in the binary images every one of the pixels value is either 0 or 1(255). The majority of minutiae extracting techniques work with the binary images due to the fact that there are only two levels of interest: the pixels that have the black color value correspond to ridges, and the pixels that have the white color value correspond to valleys. This develops the contrast between the ridges and valleys in the image of a finger-print, and therefore helps in facilitating the minutiae extraction [37]. There are several techniques for the binarization operation. Even though these techniques are quite different from each other, the majority of these techniques perform the transformation of finger-print images into binary images using the technique of thresholding. In general, the threshold for binarization is chosen in two general methods. The first technique is implemented through finding a global threshold and then performing the comparison on every one of the pixels of the image with the chosen threshold. The second method that is locally called adaptive binarization, is performed by the computation of a threshold for every one of the pixels depending on the data in the area that surrounds that pixel. For both ways, in the case where the pixel is of a greater gray scale value than the threshold value, in the resulted image, the pixel’s value will be equal to”1” which represents black color. On the other hand,the”0” value which represents the white color will be assigned to the pixel that has lighter gray scale values. The global threshold is either determined from the image histogram or by calculating the average values of the image [38].

34

Chapter Two

Biometric System

b. Median Filtering Median filtering is a non-linear filtering procedure that is implemented for getting rid of the impulsive or "salt-and-pepper" kind of noise [39]. The noise removal techniques may be even more efficient if they not only deal with the noise, but also with the spatial properties of the image. The Median Filter is a nonlinear smoothing technique that has the effect of reducing the edge blurring, where the method is replacing the current pixel in the image with the use of the median filter of the brightness in its neighboring pixels. Individual noise pieces don’t have an effect on the median of the brightness in the neighboring pixels; therefore the median smoothing gets rid of the salt and pepper kind of noise relatively well [6]. Median filters result in areas of small features (usually smudges) being removed whilst areas of larger shapes will remain untouched by the filtering action. Several repetitions of a median filter over an image will remove all small, isolated noise spikes. [40]. The simple median filter has a benefit over the mean filter because the median of the data is taken instead of the mean of an image. The pixel with the median magnitude is used afterwards for replacing the studied pixel. The median of a group is stronger considering the existence of noise [40]. Median filters can be very useful for removing noise from images. A median filter is like an averaging filter in some ways. The averaging filter examines the pixel in question and its neighbor's pixel values and returns the mean of these pixel values. The median filter looks at this same neighborhood of pixels, but returns the median value. In this way noise can be removed, but edges are not blurred as much, since the median filter is better at ignoring large discrepancies in pixel values [40].

35

Chapter Two

Biometric System

c. Region of Interest (ROI) Extraction Finger-print images that are acquired by the scanners are made up of two basic areas, the region of interest and the background. The region of interest is the region of the surface of the scanner which has been in contact with the surface of a finger and the remaining region is named background. The region of interest contains every part of information needed for the fingerprint recognition. Moreover, the background does not have any beneficial data and is a noisy area. For the sake of analyzing the image of the finger-print, it’s important for the region of interest to be segmented. Separating the region of interest from the background of the image, where the background is eliminated from the processed image, is called finger-print segmenting. Finger-print segmentation is a necessary stage in the automatic fingerprint recognition systems. For the sake of improving the effectiveness of the identification system, it’s very important that the finger-print characteristics be extracted from the region of interest [41]. Finding the Region of Interest is a procedure of filling a ROI by the interpolation of the values of the pixels from the edges of the area. This procedure can be utilized for making objects within an image seem to disappear because they are exchanged with values that blend in with the region of the background [42]. Generally speaking, only a ROI is beneficial to be recognized for every one of the finger-print images. The image region with no effective ridges and furrows is initially ignored due to the fact that it only contains information that is concerned with the background. Then the edge of the rest of the effective region is outlined out because the minutiae in the bound area is confused with that false minutiae which is produced when the ridges are out of the sensor [40].

36

Chapter Two

Biometric System

2.3.4 .3 Fingerprint Image Feature Extraction Minutiae extraction means extracting representative characteristics that are named minutiae, from the initial finger-print images. For automatic finger-print matching, a silent and proper representing of the initial finger-print images is crucial. In general, this representation must have the following characteristics:  Retaining discriminating power of the initial digital images of finger-print.  Compactness.  Amendable to matching algorithms.  Resistant to noise and distortion.  Easy to calculate. The pattern of the minutiae of the fingerprint forms a valid representation of the fingerprint. In an automatic fingerprint matching, only the two most prominent types of minute details are used for stability and robustness: ridge ending and ridge bifurcation. So typically in a lives canning fingerprint image of good quality there are about 50-100 minutiae [18]. Minutiae is a property of a finger-print that is used for the purposes of identifying. A number of automatic identification systems only consider ridge endings and bifurcations as minutiae. That is due to the fact that every one of the other structures such as bridges and island structures is assumed to be false minutiae. The benefits of using this consideration are in the fact that it doesn’t recognize those minutiae. This is due to the fact that in the case a real bifurcation is falsely separated, or in the case where an end-point falsely connects with a ridge, it generates a bifurcation. No distinction is made between the kinds of minutiae instead of their indexes which are taken in a consideration. Figure (2-9) depicts the positions of minutiae and the direction of its flow [43].

37

Chapter Two

Biometric System

(a) Bifurcation Angle

(b) Ridge termination

Figure (2-9): Definition of minutiae points

Moment invariants are important shape descriptors in computer vision. There are two types of shape descriptors: contour-based shape descriptors and region-based shape descriptors. Regular moment invariants are one of the most popular and widely used contour-based shape descriptors is a set of derived by Hu. A computer vision system recognizing objects in captured images is established using moment invariants. Moment was derived from the theory of algebraic invariant. Moment invariants technique is chosen to extract image features since the features generated are Rotation Scale Translation (RST)-invariant. Moment invariants were successfully applied in aircraft identification, texture classification and radar images to optical images matching [36].

a. Moment Invariant based Fingerprint Image Feature Extraction Moment invariants have been frequently used as features for image processing, remote sensing, shape recognition and classification. Moments are capable of providing object properties which uniquely identify its form. Invariant recognizing of shape is done by classifying in the multi-dimensional moment invariant feature space. A number of methods for the derivation of invariant characteristics from moments for recognizing objects and for representation were developed. These methods are identified by their definition of moment, like the kind of the data that is implemented and the technique for the derivation of invariant values from the image moments. Hu 38

Chapter Two

Biometric System

was the first one to determine the algebraic base for the 2D moment invariants and implemented their applications in recognizing shapes. They have been initially implemented in aircraft shape recognition and have proved to be fast and dependable. These values of moment invariants don’t change with respect to translating, scaling and rotating of the object. Hu defines seven of these shape descriptor values computed from central moments through order three that are independent to object translation, scale and orientation. Translation invariance is achieved by computing moments that are normalized with respect to the centre of gravity so that the centre of mass of the distribution is at the origin (central moments). Size invariant moments are derived from algebraic invariants but these can be shown to be the result of simple size normalization. From the second and third order values of the normalized central moments a set of seven invariant moments can be computed which are independent of rotation [44]. Moments and the related invariants have been extensively analyzed to characterize the patterns in images in a variety of applications. Hu derived six absolute orthogonal invariants and one skew orthogonal invariant based upon algebraic invariants, which are not only independent of position, size and orientation but also independent of parallel projection. The moment invariants have been proved to be the adequate measures for tracing image patterns regarding the images translation, scaling and rotation under the assumption of images with continuous functions and noise-free. Moment invariants have been extensively applied to image pattern recognition, image registration and image reconstruction [45]. Objects and pattern recognition of deformed objects have been an objective of a wide range of today’s researches. Basically, there are three significant solutions for this issue: full search, image normalization, and invariant descriptors. The approach that uses invariant descriptors seems to be the one with the maximum potential and has been widely used. The main idea of this approach is describing the object using a group of characteristics that aren’t sensitive to some specific distortions and that

39

Chapter Two

Biometric System

supply a pretty good distinguishing power for discriminating between objects that belong to various classes [46]. Representation of region moment interprets a normalized grey level image function as a probability density of a two dimensional random variable. Characteristics of this random variable may be represented with the use of statistical characteristic-moments. Supposing that a pixel with values greater than zero represents areas, moments can be used in black and white or grey level area description. A moment that has the order (p + q) depends on scaling, translating, rotating, and even on grey level transformations and is given by [47]: ∞ ∞ 𝒑 𝒒 𝒙 𝒚 𝒇 −∞ −∞

µ𝒑𝒒 =

𝒙, 𝒚 𝒅𝒙 𝒅𝒚

…………………………………... (2-1)

In digitized images we evaluate sums ∞ 𝒊=−∞

µ𝒑𝒒 =

∞ 𝒑 𝒒 𝒊=−∞ 𝒊 𝒋

𝒇(𝒊, 𝒋)

………………………………………..(2-2)

where 𝒙, 𝒚, 𝒊, 𝒋 are the coordinates of the point (coordinates of the pixel in the digitized images). Translating invariance can be reached in the case of using the central moments

µ𝑝𝑞 =

∞ ∞ (𝒙 −∞ −∞

− 𝒙𝒄 )𝒑 (𝒚 − 𝒚𝒄 )𝒒 𝒇 𝒙, 𝒚 𝒅𝒙𝒅𝒚

…………….......(2-3)

Or in digitized images

µ𝒑𝒒 =

∞ 𝒊=−∞

∞ 𝒊=−∞(𝒊

− 𝒙𝒄 )𝒑 (𝒋 − 𝒚𝒄 )𝒒 𝒇(𝒊, 𝒋)

……………………..(2-4)

where 𝒙𝒄 , 𝒚𝒄 are the co-ordinates of the region's center of gravity (centroid), which can be obtained using the following relationships:

𝒙𝒄 =

𝒎𝟏𝟎 𝒎𝟎𝟎

,

𝒚𝒄 =

𝒎𝟎𝟏 𝒎𝟎𝟎

………………………………………….(2-5)

In the binary case, 𝒎𝟎𝟎 corresponds to the region area (see equations (2-1) and (22)). Scale-invariant features can also be found in scaled central moments 𝜼𝒑𝒒 (scale change 𝒙̷ = ⍺𝒙, 𝒚̷ = ⍺ 𝒚) .

40

Chapter Two

𝜼𝒑𝒒 =

µ̷𝒑𝒒 (µ̷𝟎𝟎 )𝜸

Biometric System

,

𝜸=

𝒑+𝒒 𝟐

µ̷𝒑𝒒 =

+𝟏,

µ𝒑𝒒 ⍺𝒑+𝒒+𝟐

…………………….(2-6)

and normalized un-scaled central moments

𝝑𝒑𝒒 =

µ𝒑𝒒

……………………………..…………………………….(2-7)

(µ𝟎𝟎 )𝜸

A set of seven invariant moments can be derived from the second and third moments proposed by Hu . As the equations shown in (2-8) , Hu derived the expressions from algebraic invariants applied to the moment generating function under a rotation transformation. They consist of groups of nonlinear centralized moment expressions. The result is a set of absolute orthogonal moment invariants that can be used for scale, position, and rotation invariant pattern identification [7].

𝝋𝟏 = 𝝑𝟐𝟎 + 𝝑𝟎𝟐 , 𝝋𝟐 = 𝝑𝟐𝟎 − 𝝑𝟎𝟐

𝟐

𝝋𝟑 = 𝝑𝟑𝟎 − 𝟑𝝑𝟏𝟐 𝝋𝟒 = 𝝑𝟑𝟎 + 𝝑𝟏𝟐

𝟐

𝝋𝟓 = 𝝑𝟑𝟎 − 𝟑𝝑𝟏𝟐

+ 𝟒𝝑𝟐𝟏𝟏 𝟐

+ 𝟑𝝑𝟐𝟏 − 𝝑𝟎𝟑 𝟐,

+ 𝝑𝟐𝟏 + 𝝑𝟎𝟑 𝟐 , 𝝑𝟑𝟎 + 𝝑𝟏𝟐

𝟐

𝝑𝟑𝟎 + 𝝑𝟏𝟐

+ 𝟑𝝑𝟐𝟏 − 𝝑𝟎𝟑 𝝑𝟐𝟏 + 𝝑𝟎𝟑 𝟑 𝝑𝟑𝟎 + 𝝑𝟏𝟐 𝝋𝟔 = 𝝑𝟐𝟎 − 𝝑𝟎𝟐

𝝑𝟑𝟎 + 𝝑𝟏𝟐

𝟐

− 𝝑𝟐𝟏 + 𝝑𝟎𝟑

− 𝟑 𝝑𝟐𝟏 + 𝝑𝟎𝟑

𝟐

𝟐

𝟐

− 𝝑𝟐𝟏 + 𝝑𝟎𝟑

𝟐

+ 𝟒𝝑𝟏𝟏 𝝑𝟑𝟎 + 𝝑𝟏𝟐 𝝑𝟐𝟏 − 𝝑𝟎𝟑 , 𝝋𝟕 = 𝟑𝝑𝟐𝟏 − 𝝑𝟏𝟐 𝝑𝟑𝟎 + 𝝑𝟏𝟐

𝝑𝟑𝟎 + 𝝑𝟏𝟐

𝟐

− 𝝑𝟑𝟎 − 𝟑𝝑𝟏𝟐 𝝑𝟐𝟏 + 𝝑𝟎𝟑 𝟑 𝝑𝟑𝟎 + 𝝑𝟏𝟐

− 𝟑 𝝑𝟐𝟏 + 𝝑𝟎𝟑

𝟐

𝟐

𝟐

− 𝝑𝟐𝟏 + 𝝑𝟎𝟑

where the 𝝑𝒑𝒒 values can be computed from equation (2-7) .

41

,

(2-8)

Chapter Two

Biometric System

2.3.4 .4 Fingerprint Matching Algorithm Algorithms that extract important and efficient minutiae, will improve the performance of the fingerprint matching techniques. The features extracted of the input image are compared to one or more template that was previously stored in the system database. Therefore the system returns either a degree of similarity in case of identification or a binary decision in case of verification. Due to many factors that affect the variability of fingerprint image of the same finger, matching techniques get to be a hard problem. Some of these factors are mentioned below [20]:  Finger-print’s degree of pressure, dryness, sweat, dirt, humidity  Placement of the finger in different areas on the scanner  Rotating the finger at various angles to the scanner  Remains of the previous finger-print capturing  Errors in extracting features Minutiae-based and correlation-based approaches of matching are the most widely known approaches in matching of fingerprints. In methods based on minutiae, initially the systems deduct the minutiae in both of the images then the decision is based on the correspondence of the two groups of minutiae positions. On the other hand, in methods based on correlation, they perform a comparison of two fingerprints with respect to their grey-level values. Initially it chooses proper templates in the initial finger-print, afterwards, it uses matching of templates for locating them in the second image and comparing the locations of each fingerprint. In the methods based on correlation, the results of errors in the step of extracting minutiae are overcome. It doesn’t demand several pre-processing stages. Nevertheless, methods based on minutiae are the most widely used technique in fingerprint matching [20].

42

Chapter Two

Biometric System

2.4 Fingerprint Image Database The databases (fingerprint images) in this thesis are chosen from different database, and they are:  FVC2000 Fingerprint Database: is the First International Competition for Fingerprint Verification Algorithms. The first evaluation session was held in August 2000

and

the

results

of

the

eleven

participants

were

presented

at 15th ICPR (International Conference on Pattern Recognition). This initiative is organized by D. Maio, D. Maltoni, R. Cappelli from Biometric Systems Lab (University of Bologna), J. L. Wayman from the U.S. National Biometric Test Center (San Jose State University) and A. K. Jain from the Pattern Recognition and Image Processing Laboratory of Michigan State University. Four different databases (DB1, DB2, DB3 and DB4) were collected [48], as shown in Table (2-4).

Table (2-4): FVC2000 Fingerprint Database

DB1 DB2 DB3 DB4

Sensor Type

Image Size

Wide x deep

Resolution

Number of image

Low-cost Optical Sensor Low-cost Capacitive Sensor Optical Sensor Synthetic Generator

300x300 256x364 448x478 240x320

10x8 10x8 10x8 10x8

500 dpi 500 dpi 500 dpi about 500 dpi

80 80 80 80

Figure (2-10) shows a sample fingerprint image of each database:

Figure (2-10): Sample fingerprint image of each database from FVC2000 database

43

Chapter Two

Biometric System

 FVC2002 Fingerprint Database: is the Second International Competition for Fingerprint Verification Algorithms. The evaluation was held in April 2002 and the results of the 31 participants were presented at 16th ICPR (International Conference on Pattern Recognition). This initiative is organized by D. Maio, D. Maltoni, R. Cappelli from Biometric Systems Lab (University of Bologna), J. L. Wayman from the U.S. National Biometric Test Center (San Jose State University) and A. K. Jain from the Pattern Recognition and Image Processing Laboratory of Michigan State University. Four different databases (DB1, DB2, DB3 and DB4) were collected [49], as shown in Table (2-5). Table (2-5): FVC2004 Fingerprint Database

DB1 DB2 DB3 DB4

Sensor Type

Image Size

Wide x deep

Resolution

Number of image

Optical Sensor Optical Sensor Capacitive Sensor SFinGe v2.51

388x374 296x560 300x300 288x384

10x8 10x8 10x8 10x8

500 dpi 569 dpi 500 dpi about 500 dpi

80 80 80 80

Figure (2-11) shows a sample fingerprint image of each database:

Figure (2-11): Sample fingerprint image of each database from FVC2002 database

44

Chapter Two

Biometric System

 FVC2004 Fingerprint Database: is the Third International Competition for Fingerprint Verification Algorithms. Students (24 years old on the average) enrolled in the Computer Science degree program at the University of Bologna kindly agreed to act as volunteers for providing fingerprints. Four different databases (DB1, DB2, DB3 and DB4) were collected [50], as shown in Table (2-6).

Table (2-6): FVC2004 Fingerprint Database

Sensor Type

Optical Sensor DB1 Optical Sensor DB2 DB3 Thermal sweeping Sensor SFinGe v3.0 DB4

Image Size

Wide x deep

Resolution

Number of image

640x480 328x364 300x480 288x384

10x8 10x8 10x8 10x8

500 dpi 500 dpi 512 dpi about 500 dpi

80 80 80 80

Figure (2-12) shows a sample fingerprint image of each database:

Figure (2-12): Sample fingerprint image of each database from FVC2004 database

 Neurotechnologija Fingerprint Database: two different databases (DB1and DB2) were collected [51] as shown in Table (2-7).

45

Chapter Two

Biometric System

Table (2-7): Neurotechnologija Fingerprint Database

DB1 DB2

Sensor Type

Image Size

Wide x deep

Resolution

Number of image

Cross Match Verifier 300 DigitalPersona U.are.U 4000

504x480 326x357

54x8 70x8

500 dpi 500 dpi

432 560

Figure (2-13) shows a sample fingerprint image of each database:

DB1

DB2

Figure (2-13): Sample fingerprint image of each database from Neurotechnologija database

 UPEK Fingerprint Database: one database (DB1) was collected [52] as shown in Table (2-8). Table (2-8): UPEK Fingerprint Database

DB1

Sensor Type

Image Size

Wide x deep

Resolution

Number of image

Capacitive Sensor

284x338

16x8

500 dpi

128

Figure (2-14) shows a sample fingerprint image of UPEK Fingerprint Database:

Figure (2-14): Sample fingerprint image of UPEK fingerprint database

46

Chapter Three

Unique ID Generation Using Fingerprint

Chapter Three

Unique ID Generation Using Fingerprint

3.1 Introduction Electronic commerce applications need secure mechanism for accurate user identification, for accessing sensitive database, storing and transmitting sensitive information etc. personal identification numbers, passwords, smart cards and digital certificates are some of the means normally employed for this purpose. However, these means do not really identify a person, but only knowledge of some data or belonging of some determined object. Biometrics is one of the solutions, in this chapter a new Unique ID generation method using fingerprint technology is proposed and it can be used for the personal identification and verification.

3.2 Fingerprint Image Database The database used in this thesis consists of images of different fingerprint patterns such as (loop, whorl and arch) as described in section (2.4). There are 40 images selected for standard fingerprint features generation phase, 10 images for core feature samples, 10 images for delta feature samples, 10 images for bifurcation feature samples and 10 images for ridge ending feature samples. The fingerprint's features are characterized by global and local features. The global features include core and delta. The minutiae points such as ridge bifurcations and endings from the local features. The user partitions the fingerprint image obtained from the fingerprint database into a number of blocks, detects blocks boundaries, crops the blocks, and saves them as distinct image files since the detection, extraction, and cropping of fingerprint's features are not to be considered as one of the automated system tasks, it should be done by the user and it does not belong to the technique responsibilities. Each image file will represent one of the fingerprint's features (core, delta, ridge bifurcation and ridge ending) for input to the implemented algorithm in standard fingerprint features generation phase. Figure (3-1) shows samples of fingerprint's features.

48

Chapter Three

(a) Loop core

Unique ID Generation Using Fingerprint

(b) whorl core

(c) delta

(d) bifurcation

(e) ending

Figure (3-1): Fingerprint's features samples

3.3 Proposed System Stages The systems of person identification that depend on the patterns of fingerprints are known as Automatic Fingerprint Identification Systems, AFIS, and they are considered to be some of the most widely used biometrical techniques because they offer a high success rate. The precision of AFIS is basically a result of a set of unique characteristics known as the minutiae, which are the spots where a curve track ends, intersects with another curve track, or splits in to two. Two main kinds of minutia, the ridge ending and the ridge bifurcation are chosen for the sake of improvement of the suggested application. Just those two are chosen because the goal of this thesis is creating a unique ID for confidential documents like university certifications and some other kinds of highly important documents. Figure (3-2) illustrates the ridge pattern as well as the major minutia forms. It can be concluded from the figure that a ridge ending indicates the abrupt ending of a ridge and ridge bifurcation means a single ridge that splits into two ridges.

49

Chapter Three

Unique ID Generation Using Fingerprint

Figure (3-2): Fingerprint ridge patterns and minutiae

Subsequent to distinguishing the features which make the fingerprint individual and unrepeated, they are extracted from fingerprint images based on the seven moment invariants, then the extracted features are transformed into numeric values. Those numeric values are used as the unique ID for printing in the documents for the personal identification. The proposed system contains three stages: enrollment stage , identification stage and verification stage as shown in Figure (3-3). In the enrollment stage, fingerprint images of the different individuals to be identity are first processed by a feature extraction module and then the extracted features are converted to numerical values. These numerical values are stored as a unique ID in a database for later use. In the identification stage , the fingerprint image of an individual to be identify first processed by a feature extraction module; the extracted features are converted to numerical value and then fed to a matching module with all numerical values stored in the database to decide whether the individual is identified or not. While in the verification stage, the numerical value of an individual in card fed to a matching module with all numerical values stored in the database to decide whether the individual is accepted or rejected.

50

Chapter Three

Unique ID Generation Using Fingerprint

Enrollment stage Feature Extraction

Fingerprint image

Numerical value

Database

Identification stage

Feature Extraction

Fingerprint image

Numerical value

Matching

Database

identifier or not identified

Verification stage Numerical value in card

Matching

Database

accept/reject

Figure (3-3): Overview of the proposed system

3.3.1 Feature Extraction Feature extraction is defined as the process of converting a captured biometric sample, i.e. fingerprint, in to a unique, distinctive and compact form so that it can be compared to a reference template. Feature extraction is a significant step to improve the efficiency of personal identification. This stage consists of two phases standard fingerprint features generation phase and unique ID template generation phase. Each phase has specific steps.

51

Chapter Three

Unique ID Generation Using Fingerprint

The standard fingerprint features generation phase has four steps; the first step initializes fingerprint feature samples in a binary form. Compute the seven moments of the samples of the fingerprint features in the second step. While the third step, computes the variance of the samples of the fingerprint features. Finally the variances of the samples of the fingerprint features are stored in to the array for later use. The unique ID template generation phase has eight steps; in the first step fingerprint image is initialized in a binary form. The second step crops the binarized fingerprint image into an ROI in order to separate foreground region and background region to speed up the overall process. In the third step the white ridge width and the black ridge width of the fingerprint image are computed and the histogram of the white ridge width and the black ridge width are found and then the total ridge width is computed. In the fourth step the ROIs are divided into blocks with equal size and blocks size is detected based on the total ridge width that is computed in the third step. The fifth step computes the seven moments of each block. The sixth step computes the variances of each block. The seventh step compares and matches the variances of each block with the variances of feature samples stored in array and the largest match is found and marked as the feature points on fingerprint image. Finally the extracted features are converted to numerical values. These numerical values are stored as a unique ID in a database for later use. Figure (3-4) shows the flow chart of the feature extraction.

52

Chapter Three

Unique ID Generation Using Fingerprint

Unique ID template generation phase

Standard fingerprint features generation phase

start

Input fingerprint image

Input fingerprint feature samples

Fingerprint image binarization

Fingerprint feature samples binarization

Region of interest (ROI) extraction

Compute the seven moments of each fingerprint feature sample

Compute the total ridge width

Compute the variance of each fingerprint feature sample Divide the ROIs into blocks with equal size and detect blocks size based on the total ridge width Variance database

Compute the seven moments of each block

Compute the variance of each block

Variance matching

Find the largest match and mark it as the feature points on fingerprint image Cconvert these feature points to numerical value

Numerical values database

Figure (3-4): Flow Chart of the Feature Extraction

53

Chapter Three

Unique ID Generation Using Fingerprint

3.3.1.1 Standard Fingerprint Features Generation Phase This phase contains three steps which can be seen in Figure (3-4). a. Sample Image Binarazation In order to transform the image in to a form which facilitates the feature extraction process, will examine the gray-level value of each pixel in the image, and, if the value is smaller than a global threshold, then the pixel value is set to a binary value zero, otherwise, it is set to one. The outcome is a binary image containing two levels of information, the foreground ridge and the background valleys. After the fingerprint's features (core, delta, ridge bifurcation and ridge ending) are provided manually, these features are processed in order to satisfy the system requirements. Ten samples for each fingerprint's feature are considered as an input for algorithm (3-1) to convert these samples images in to binary form as pre-processing step in order to be used in the seven invariant moments in subsequent steps. Binarazation algorithm first must compute the histogram of an image. The image histogram is a plot of the gray level values versus the number of pixels at that value. The horizontal axis is the gray- level value. It begins at zero and goes to the number of gray levels. Each vertical bar represents the number of times the corresponding gray level occurs in the image. Next compute the probability of the histogram by dividing the counts in each bin by the total number of pixels in the image associated with that histogram. To find the best threshold the following iterative procedure must be followed: 1. Pick an initial threshold value, (T). 2. Calculate the two mean image intensity values in the histogram (m1 and m2) below and above the threshold T. 3. For each pixel in (m1 and m2) apply this equation: 𝐱 𝟏 = 𝐬𝐮𝐦(𝐡𝐢𝐬𝐭(𝐦𝟏 ) ∗ 𝐩(𝐦𝟏 ))/𝐬𝐮𝐦(𝐩(𝐦𝟏 )) 𝐱 𝟐 = 𝐬𝐮𝐦(𝐡𝐢𝐬𝐭(𝐦𝟐 ) ∗ 𝐩(𝐦𝟐 ))/𝐬𝐮𝐦(𝐩(𝐦𝟐 ))

…………………………….(3-1)

4. Calculate new threshold, T new = (x1+ x2) / 2………………………………

54

(3-2)

Chapter Three

Unique ID Generation Using Fingerprint

5. If the threshold has stabilized (T = T new), this is the appropriate threshold level. Otherwise, T becomes T new and reiterates from step 2. Then examine the gray-level value of each pixel in the image; if the value is smaller than new threshold value (T new), then the pixel value is set to a binary value zero, otherwise, it is set to one. Algorithm (3-1) shows the steps for binarazation of each sample image. Figure (3-5) shows sample image of each fingerprint's features. Figure (3-6) shows binarized sample image of each fingerprint's features. Algorithm (3-1): Binarazation Input: Samples images of each fingerprint's feature. Output: Binarized samples images of each fingerprint's feature. Begin Step1: Find the histogram of the sample image. Step2: Find Probability for each Pixel in sample image P(x). Step3: Choose an initial threshold, T. Step4: Calculate the two mean image intensity values in the histogram (m1 and m2) below and above the threshold T. Step5: For each pixel in (m1 and m2) apply equation (3-1). Step6: Compute new threshold, T new by using equation (3-2). Step7: If the threshold has stabilized (T = T

new),

this is the appropriate

threshold level. Otherwise, T becomes T new and reiterates from step4. Step8: For each pixel in the sample image convert to white or black If pixel > 𝑇𝑛𝑒𝑤 then white Else black End

55

Chapter Three

(a)

Unique ID Generation Using Fingerprint

(b)

(c)

(d)

Figure (3-5): Sample images of fingerprint's features image a) Core b) Delta c) Bifurcation d) Ridge ending

(a)

(b)

(c)

(d)

Figure (3-6): Binarized samples images of fingerprint's features image a) Core b) Delta c) Bifurcation d) Ridge ending

b. Computing seven invariant moments of fingerprint's features samples Moments and functions of moments have been extensively employed as invariant global features of images in pattern recognition. For object recognition, regardless of orientation, size and position, feature vectors are computed with the help of nonlinear moment invariant functions. Representations of objects using twodimensional images that are taken from different angles of view are the main features leading us to our objective. After efficient feature extraction, the recognition performance in conjunction with moment–based feature sets is introduced. The binarized sample images of each fingerprint's feature which were produced by the binarazation algorithm are considered as input for algorithm (3-2) to compute the 56

Chapter Three

Unique ID Generation Using Fingerprint

seven invariant moments of each sample. The seven invariant moments are calculated by applying equations (2-2), (2-4), (2-5), (2-6), (2-7) and (2-8) respectively. Algorithm (3-2) shows the steps for computing the seven invariant moments of each sample. Algorithm (3-2): Seven Invariant Moments Input: Binarized samples images of each fingerprint's feature. Output: Seven invariant moment values of each sample. Begin Step1: Apply equation (2-2) to find the moment of order (p + q) which is dependent on scaling, translation and rotation. Step2: Apply equations (2-4) and (2-5) to find the central moments. Step3: Apply equations (2-6) and (2-7) to find the normalized central moments. Step4: Apply equation (2-8) to compute the seven invariant moments. End

c. Computing variance of fingerprint's features samples The variance of each sample can be achieved by using algorithm (3-3). The input to the variance algorithm is the result of the algorithm (3-2). The variance algorithm first must find the maximum and minimum values of each moment of the samples. Next compute the mid of each moment of the samples by applying this equation: Mid(i) = (max(i) + min(i))/2, i=( 1,……….,7)……………………………….

(3-3)

In this process each fingerprint's feature (core, delta, ridge bifurcation and ridge ending) has seven mid. For instance, the first mid of the core feature can be computed by finding the summation of the maximum value of the first moment and the minimum value of the first moment and the result of the summation is divided by two, the second mid of the core feature can be computed by finding the summation of

57

Chapter Three

Unique ID Generation Using Fingerprint

the maximum value of the second moment and the minimum value of the second moment and the result of the summation is divided by two and so on, up to seventh mid of the core feature. And so for the rest of the fingerprint's features: delta, ridge bifurcation and ridge ending. To compute the variance of each sample this equation must be applied: 𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆𝑺 (𝒋) =

𝟕 𝒊=𝟏

𝒎𝒐𝒎𝒆𝒏𝒕 𝒊𝒏𝒗𝒂𝒓𝒊𝒂𝒏𝒕𝒔 𝒊 − 𝒎𝒊𝒅(𝒊) .𝟐

j=(1,………….,10)……………………………………………………………… (3-4) In this process, each fingerprint's feature (core, delta, ridge bifurcation and ridge ending) has ten variances. For instance, the first variance of the core feature can be calculated by finding the summation of the squared difference between the seven moments of the first core sample and the seven mid of the core feature, the second variance of the core feature can be calculated by finding the summation of the squared difference between the seven moments of the second core sample and the seven mid of the core feature and so on, up to tenth variance of the core feature. And so is for the rest of the fingerprint's features: delta, ridge bifurcation and ridge ending. The seven mid and the ten variances of each fingerprint's feature are stored in the array for later use. Algorithm (3-3) shows the steps of computing the variance of each sample. Algorithm (3-3): Variances of Samples Input: Seven invariant moment values of each sample. Output: Variance of each sample. Begin Step1: Find max and min values of each moment. Step2: Compute mean of each moment by applying the equation (3-3). Step3: Compute variance of each sample by applying the equation (3-4). Step4: Store mid and variances of each fingerprint's feature in the array for later use. End 58

Chapter Three

Unique ID Generation Using Fingerprint

3.3.1.2 Unique ID Template Generation Phase This phase contains eight steps which can be seen in Figure (3-4). a. Fingerprint Image Binarazation The fingerprint image which is obtained from fingerprint database is considered as an input for algorithm (3-1) to compute the binarazation of the original fingerprint image as pre-processing step. Figure (3-7) shows a sample of original fingerprint image binarazation.

(a) original image

(b) binarized image

Figure (3-7): Sample of fingerprint image binarazation

b. Region of Interest (ROI) Extraction Region of Interest (ROI) is the process of separating the foreground regions in the image from the background regions. In fingerprint image the foreground regions correspond to the fingerprint area containing the ridges and valleys, which is our area of interest and background corresponds to the regions outside the borders of the fingerprint area, which do not contain any valid fingerprint information. When minutiae extraction algorithms are applied to fingerprint image due to the background regions of an image, extraction algorithm extract noisy and false minutiae. Therefore in this step, ROI is employed to discard these background regions, which facilitates the reliable extraction of minutiae. After computing the median filtering which is 59

Chapter Three

Unique ID Generation Using Fingerprint

considered as a pre-processing step to extract the ROI from the fingerprint image, ROI extraction will be applied by using algorithm (3-4). When performing the ROI extraction algorithm, will apply the ROI to the binarized fingerprint image but not to the noise filtered fingerprint image in order to maintain the precise details of the fingerprint image. Algorithm (3-4) shows the steps for ROI Extraction. Algorithm (3-4): ROI Extraction Input: Binarized fingerprint image. Output: ROI of the binarized fingerprint image. Begin Step1: Apply median filtering to the binarized fingerprint image. Step2: Check the left region of the binarized fingerprint image if any pixel is black then get the coordinates of the pixel (x left, y left). Step3: Check the right region of the binarized fingerprint image if any pixel is black then get the coordinates of the pixel (x right, y right). Step4: Check the top region of the binarized fingerprint image if any pixel is black then get the coordinates of the pixel (x top, y top). Step5: Check the bottom region of the binarized fingerprint image if any pixel is black then get the coordinates of the pixel (x bottom, y bottom). Step6: The new dimensions of the binarized fingerprint image ROI (x bottom - x top , y right - y left). End

c. Computing Total Ridge Width After applying the Region of Interest (ROI) extraction algorithm in the previous step, will compute the total ridge width of the fingerprint image (ROI). The total ridge width of the fingerprint image is done by using algorithm (3-5). The input for this algorithm is the result of algorithm (3-4). In the total ridge width algorithm first must determine three different regions in the fingerprint image (ROI). Next compute the white ridge width and the black ridge width for each region. This process starts 60

Chapter Three

Unique ID Generation Using Fingerprint

with 2D window at size 7×7 in each region, and then loops add an increment to window size each time round the loop, it computes the white pixels numbers (1) that are located between two black pixels (white ridge width) and the black pixels number (0) that are located between two white pixels (black ridge width) in each window boundaries from the four directions (top, left, bottom and right) as shown in Figure (3-8). Table (3-2) shows white ridge width and black ridge width in each window of the first region from four directions.

Figure (3-8): Windows in first region of fingerprint image

61

Chapter Three

Unique ID Generation Using Fingerprint

Table (3-1): white ridge width and black ridge width in each window of the first region from four directions

First region

Window

Top direction

Left direction

Bottom direction

Right direction

white

black

white

black

white

black

white

black

width

width

width

width

width

width

width

width

Window 1

0

0

0

0

0

0

0

0

Window 2

0

0

6

0

1

5

0

0

Window 3

0

0

6

0

0

0

7

0

Window 4

0

0

5

0

0

0

6

0

Window 5

0

5

5

1

9

6

5

Window 6

0

0

5

5

1

5

6

Window 7

0

0

6

0

0

5

Window 8

0

0

0

0

0

5

Window 9

0

0

6

0

0

4

Window 10

20

0

0

1

number

62

4 4 4 4 3 5 4 4 4

5

4

6

4

6 4 6 4 6 4

6

3

6

5

Chapter Three

First region

Unique ID Generation Using Fingerprint Window number

Top direction

Left direction

Bottom direction

white

black

white

black

white

black

white

black

width

width

width

width

width

width

width

width

0

0

0

6

4

5

4

0

0

5

4

5

4

5 Window 11

2

0

6 5 5

Window 12

0

0

6 7

Window 13

0

0

0

0

3 4

6

3

7

1

6 7

63

4

5

4 Window 14

Right direction

4 3

0

1

24

0

5 6

4 4 4

5

6

8

4

6

2

Chapter Three

Unique ID Generation Using Fingerprint

Then compute the histogram of the white ridge width and the black white ridge width for each region and find the maximum values of the frequency of occurrence of the white ridge width and the black ridge width for all regions. Table (3-3) shows the histogram of the white ridge width and the black ridge width for first region.

Table (3-2): Histogram of white ridge width and black ridge width for first region

First region

max

white ridge width

histogram

black ridge width

histogram

1

3

1

3

2

1

2

1

3

0

3

5

4

2

4

23

5

17

5

5

6

16

6

5

7

4

7

0

8

1

8

0

9

0

9

1

20

1

20

0

24

0

24

1

5

17

4

23

Determining an integer value (K); this value identifies how many lines that the fingerprint's feature has. Depending on trial-and-error, best integer value for K is 6. If K is more or less than 6 then the result is incorrect. The total ridge width of the fingerprint image can be computed by applying this equation: (3-5)

Total RW(i) = (max W + max B)*K, i=(1,..,3)………………………………… where total

RW

is the total ridge width of the fingerprint image, max

W

is the

maximum value of the frequency of occurrence of the white ridge width and max B is the maximum value of the frequency of occurrence of the black ridge width. 64

Chapter Three

Unique ID Generation Using Fingerprint

Algorithm (3-5) shows the steps for computing total ridge width of the fingerprint image. Algorithm (3-5): Total Ridge Width Input: Fingerprint image (ROI). Output: Total ridge width of the fingerprint image (ROI). Begin Step1: Determine three different regions in the ROI. Step2: Compute the white ridge width and the black ridge width for each region. Step3: Compute the histogram of the white ridge width and the black ridge width for each region. Step4: Find the maximum values of the frequency of occurrence of the white ridge width (max W) and the black ridge width (max B) for all regions. Step5: Choose an integer K (K=6). Step6: Compute total ridge width by using the equation (3-5). End

d. Fingerprint Image Partition After applying the (ROI) extraction algorithm and the total ridge width algorithm, will divide the fingerprint image (ROI) into blocks with equal size and detect blocks size based on the total ridge width that is computed in the previous step. The fingerprint image partition can be performed pixel by pixel, as if consider the block as the mask, which slides across the fingerprint image and performs an arithmetic operation at each pixel location as shown in Figure (3-9). For instance, start from the block that is located in the upper-left corner of the fingerprint image, slide the block over by one pixel and repeat this process until get to the end of the row. When the end of the row is reached, the block is moved down one row, and the

65

Chapter Three

Unique ID Generation Using Fingerprint

process is repeated row by row until this procedure has been performed on the entire image.

Figure (3-9): How mask (block) slides across the fingerprint image

e. Computing seven moment invariants of blocks The blocks of the fingerprint image are obtained from applying the fingerprint image partition is considered as input for algorithm (3-2) to compute the seven invariant moments of each block. f. Computing variances of blocks The variance of each block of the fingerprint image can be achieved by using algorithm (3-6). The input to this algorithm is the seven moment invariant values of each block that resulted from the algorithm (3-2) and the seven mid values of the fingerprint's features stored in array. The variance of each block can be calculated from applying this equation: 𝑽𝒂𝒓𝒊𝒂𝒏𝒄𝒆𝑩 (𝒋) =

𝟕 𝒊=𝟏

𝒎𝒐𝒎𝒆𝒏𝒕 𝒊𝒏𝒗𝒂𝒓𝒊𝒂𝒏𝒕𝒔 𝒊 − 𝒎𝒊𝒅𝑫𝑩 (𝒊) .𝟐

𝒋 = (𝟏, …………..,n) , n= number of blocks ………...………………………..(3-6) where 𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒𝑩 is the variance of each block and 𝑚𝑖𝑑𝐷𝐵 is the seven mid of the fingerprint features stored in array. In this process each block has four variances, the first variance value can be calculated by finding the summation of the squared

66

Chapter Three

Unique ID Generation Using Fingerprint

difference between the seven moment values of the block and the seven mid values of the core feature stored in array, the second variance value can be calculated by finding the summation of the squared difference between the seven moment values of the block and the seven mid values of the delta feature stored in array, the third variance value can be calculated by finding the summation of the squared difference between the seven moment values of the block and the seven mid values of the bifurcation feature stored in array and the fourth variance value can be calculated by finding the summation of the squared difference between the seven moment values of the block and the seven mid values of the ridge ending feature stored in array. Algorithm (3-6) shows the steps for computing the variance of each block of the fingerprint image.

Algorithm (3-6): Variance of Blocks Input: Seven moment invariant values of each block, seven mid values of the fingerprint features stored in array. Output: Variance of each block Begin For each fingerprint's feature, compute variance of each block by applying equation (3-6). End

g. Variance Matching After calculating the ten variances of each fingerprint's feature stored in array and the four variances of each block which were produced previously by algorithm (3-3) and algorithm (3-6) respectively, now will perform the variance matching which is done by using algorithm (3-7). The input to this algorithm is the result of algorithm (3-3) and algorithm (3-6).

67

Chapter Three

Unique ID Generation Using Fingerprint

The variance matching algorithm first compares and matches the variances of each block with the variances of each fingerprint's feature stored in array. Next find the largest matches, these largest matches are considered as the fingerprint feature points and mark them on the fingerprint image. For instance, compare and match of the four variances of the block with the ten variances of the core feature, the ten variances of the delta feature, the ten variances of the bifurcation feature and the ten variances of the ending feature. In this comparison, if the largest match of the first variance value of the block exists within the ten variance values of the core feature then mark it as the core point on fingerprint image, If the largest match of the second variance value of the block exists within the ten variance values of the delta feature then mark it as the delta point on fingerprint image, If the largest match of the third variance value of the block exists within the ten variance values of the bifurcation feature then mark it as the bifurcation point on fingerprint image and if the largest match of the fourth variance value of the block exists within the ten variance values of the ending feature then mark it as the ending point on fingerprint image. Algorithm (3-7) shows the steps for fingerprint matching.

Algorithm (3-7): Variance Matching Input: The ten variances of each fingerprint's feature (variance S) stored in array from algorithm (3-3), the four variances of each block (variance B) from algorithm (3-6). Output: The largest match and mark it as the fingerprint feature points on fingerprint image. Begin Step1: Choose a real number, K (K= 0.1). Step2: For each variance

S

which is stored in array, multiply the

variance S by K. Step3: Find the range of each varianceS by adding and subtracting the variance S from the result of step2.

68

Chapter Three

Unique ID Generation Using Fingerprint

Step4: Compare each variance B with the range of each variance S to find the number of the matches. Step5: Find the percentage by dividing (100) on the result of step4. Step6: Compute the summation of the matches by dividing the variance B on the variance S and multiplying the result of the division by the result of step5. End

h. Unique ID Generation The most significant step is identifying a unique point in the fingerprint that will represent the reference point. All of the other minutia characteristics will be computed with this point as the base. Core point of the fingerprint is one such unique point. Subsequent to finding the core point and the minutia points there will be a big number of minutiae. However, just a little number of minutiae that is completely individual for that specific fingerprint is necessary in generating the unique ID. Therefore only the minutiae that lie within a Region of Interest (ROI) distributed in the area that surrounds the core point are taken. Here, the ROI is a square with an appropriate pixel area that is centered on the core point. The unique ID generation is done by using algorithm (3-8). The input to this algorithm is the result of the algorithm (3-7). There are two cases in the fingerprint image; in the first case the fingerprint image contains the core point such as loop and whorl, while in the second case the fingerprint image does not contain the core point like arch.

69

Chapter Three

Unique ID Generation Using Fingerprint

1.The fingerprint image contains the core point: In this case the unique ID generation algorithm must apply the (ROI) centered around the core point so that only the minutia points (ridge bifurcation and ridge ending points) which are inside ROI are taken, The size of the ROI can be determined by finding the distance between the core point and the rest of the other minutia points (ridge bifurcation and ridge ending points) by using the Euclidean distance equation and selecting the required number of the minutia points that have the less distance between them and the core point so that only these minutia points are inside ROI. The required number of the minutia points can be determined by the system user. By applying parametric equation of a circle, a circle can be defined as the locus of all points that satisfy the equations: x = 𝐱 𝐜 + r cos Ɵ

For 0 < Ɵ > 2π

………………………..(3-7)

y = 𝐲𝐜 + r sine Ɵ where x, y are the coordinates of any point on the circle, 𝒙𝒄 , 𝒚𝒄 are the coordinates of the center (core point), r is the radius of the circle and Ɵ is the parameter (subtended angle). What these equations do is generate the x, y coordinates of a point on the circle given the certain radius and an angle θ (theta). This process starts with theta at zero, and then theta loops add an increment to theta each time round the loop. It draws straight line segments between these successive points on the circle. The circle is thus drawn as a series of straight lines in clockwise direction. If find minutia points that lie within the boundaries of the circle then find the angle for this point. After that radius loops add an increment to radius each time round the loop in order to draw other circles by parametric equation of a circle. If find minutia points located within the boundaries of these circles then find their own angles. This process continues until the end of all the minutia points contained within the (ROI). Each angle of the minutia points is converted to fixed length which consists of three digits. For instance if the value of the angle is (60) then the fixed length of the angle is 70

Chapter Three

Unique ID Generation Using Fingerprint

(060). Depending on the priority of the finding the minutia points on the circles, all the angles are concatenated with each other to generate the numerical values. These numerical values are considered as a unique ID and stored in the database for later use. 2.The fingerprint image does not contain the core point: In this case the unique ID generation algorithm must calculate the distance between each point and the rest of the other minutia points which are marked in the fingerprint image by using the Euclidean distance equation. In the Euclidean distance, if 𝐏𝟏 (𝐱 𝟏 ,𝐲𝟏 ) and 𝐏𝟐 (𝐱 𝟐,𝐲𝟐 ) then the distance is given by: D (𝐏𝟏 , 𝐏𝟐 ) =

(𝐱 𝟐 − 𝐱 𝟏 )𝟐 + (𝐲𝟐 − 𝐲𝟏 )𝟐 …………………………………….. (3-8)

For instance if the fingerprint image contains four minutia points (p1 ,p2 , p3 and p4 ) then we compute the distance between these points: (D (P1 , P2 ), D (P1 , P3 ), D (P1 , P4 ), D (P2 , P3 ), D (P2 , P4 ) and D (P3 , P4 )). Choose the less distance between two minutia points and find the midpoint between them by using the midpoint equation, if 𝐏𝟏 (𝐱 𝟏 ,𝐲𝟏 ) and 𝐏𝟐 (𝐱 𝟐 ,𝐲𝟐 ) then the midpoint is given by: 𝐱=(𝐱 𝟏 + 𝐱 𝟐 )/𝟐 𝐲=(𝐲𝟏 + 𝐲𝟐 )/𝟐 The Midpoint is P(x, y) = ( (𝐱 𝟏 + 𝐱 𝟐 )/𝟐,(𝐲𝟏 + 𝐲𝟐 )/𝟐) ……………………….

(3-9)

This midpoint is considered as the core point of the fingerprint image. Then repeat the same steps that have been applied in the case of the fingerprint image contain the core point in order to generate a unique ID. Algorithm (3-8) shows the steps for unique ID generation.

71

Chapter Three

Unique ID Generation Using Fingerprint

Algorithm (3-8): Unique ID generation Input: A number of minutia points that are marked in the fingerprint image, Core point = true or false. Output: A unique ID. Begin Step1: If core point = true then go to step 2 else go to step7. Step2: Apply (ROI) centered around the core point so that only the minutia points which are inside ROI are taken. Step3: Draw circles in clockwise direction by applying equation (3-7) Step4: If find minutia point lie within the boundaries of the circle then find the angle of this minutia point. Step5: All angles which are found in step4 are converted to fixed length which consists of three digits. Step6: Depending on the priority of the finding the minutia points on the circles, all the angles are concatenated with each other to generate the numerical values. These numerical values are considered as a unique ID and stored in the database for later use and then go to end. Step7: Calculate the distance between each point and the rest of the other feature points which are marked in the fingerprint image by using the Euclidean distance equation (3-8), choose the less distance between two minutia points and find the midpoint between them by using the midpoint equation (3-9) and this is considered as the core point. Step8: Repeat steps 2 to 6 in order to generate a unique ID. End

72

Chapter Three

Unique ID Generation Using Fingerprint

3.3.2 Fingerprint Matching Algorithm Fingerprint recognition has been studied for many years, and numerous algorithms have been proposed to improve the performance of AFRS. After extracting the features points from fingerprint images by algorithm (3-7) and these extracted features are converted to numerical values by algorithm (3-8). The numerical value is consisting of several parts, each part represents an angle and the length of the angle is three digits. The fingerprint image matching is the most significant step in identification process. The fingerprint matching algorithm contains two stages, identification stage and verification stage. In the identification stage, a fingerprint image of the person to be identified first processed by a feature extraction module; the extracted features are converted to numerical value and then measure the similarity between the angle vectors of this numerical value with those of the numerical values stored in the database by using the absolute distance equation to decide whether this person is identified or not. While in the verification stage, will measure the similarity between the angle vectors of the input numerical value of the person which record in the card with those of the numerical values stored in the database by using the absolute distance equation in order to decide whether this person is accepted or rejected. Let Vf1 = {a1, a2….… an} and Vf2 = {b1, b2…… b n} (n=6) denote the angle vectors of the two numerical values of the fingerprints to be matched, Tm denotes the threshold used in the matching process. The difference vector V d of the two angle vectors is computed as in equation (3-10) V d=

|𝒂𝟏−𝒃𝟏|

,

|𝒂𝟐−𝒃𝟐|

𝐦𝐚𝐱⁡ (𝒂𝟏,𝒃𝟏) 𝐦𝐚𝐱⁡ (𝒂𝟐,𝒃𝟐)

……………

|𝒂𝒏−𝒃𝒏| 𝐦𝐚𝐱⁡ (𝒂𝒏,𝒃𝒏)

…...……………………..(3-10)

We define the absolute distance of the two matching vectors as in equation (3-11) R m=

|𝒂𝒊−𝒃𝒊| 𝒏 𝒊=𝟏 𝐦𝐚𝐱⁡ (𝒂𝒊,𝒃𝒊)

……...…………………………………………… …… (3-11)

73

Chapter Three

Unique ID Generation Using Fingerprint

Then the final decision of matching result is determined by equation (3-12) The person is identified if R m > Tm

……….……………………………(3-12)

The person is not identified if R m