Furrow And Crypt Detection Using Modified Ant Colony Optimization For Iris Recognition.

(1)

UNIVERSITI TEKNOLOGI MARA

FURROW AND CRYPT DETECTION

USING MODIFIED

ANT COLONY OPTIMIZATION

FOR IRIS RECOGNITION

ZAHEERA ZAINAL ABIDIN

PhD

(2)

UNIVERSITI TEKNOLOGI MARA

FURROW AND CRYPT DETECTION

USING MODIFIED

ANT COLONY OPTIMIZATION

FOR IRIS RECOGNITION

ZAHEERA ZAINAL ABIDIN

Thesis submitted in fulfillment

of the requirements for the degree of

Doctor of Philosophy

Faculty of Computer and Mathematical Sciences

(3)

CONFIRMATION BY PANEL OF EXAMINERS

I certify that a panel of examiners has met on 3rd November 2015 to conduct the final

examination of Zaheera binti Zainal Abidin on her Doctor of Philosophy thesis entitled ―Furrow and Crypt Detection using Modified Ant Colony Optimization for

Iris Recognition‖ in accordance with Universiti Teknologi MARA Act 1976 (Akta

173). The Panel of Examiners recommends that the student be awarded the relevant degree. The panel of Examiners was as follows:

Daud Mohamad, PhD Professor

Faculty of Computer & Mathematical Sciences Universiti Teknologi MARA

(Chairman) Anil K. Jain, PhD Professor

Department of Computer Science & Engineering

Michigan State University, East Lansing, Michigan, USA (External Examiner - International)

Abd Rahman bin Ramli, PhD Associate Professor

Faculty of Engineering Universiti Putra Malaysia (External Examiner - National) Noor Elaiza binti Abd. Khalid, Phd Senior Lecturer

Faculty of Computer and Mathematical Sciences Universiti Teknologi MARA, Malaysia

(Internal Examiner)

SITI HALIJJAH SHARIFF, PhD Associate Professor

Dean

Institute Graduate Studies

Universiti Teknologi MARA Date: 26 January 2016

(4)

iii

AUTHOR'S DECLARATION

I declare that the work in this thesis was carried out in accordance with the regulation of Universiti Teknologi MARA. It is original and is the result of my own work, unless otherwise indicated or acknowledged as referenced work. This thesis has not been submitted to any other academic institution or non-academic institution for any other degree or qualification.

I, hereby, acknowledge that I have been supplied with the Academic Rules and Regulations for Post Graduate, Universiti Teknologi MARA, regulating the conduct of my study and research.

Name of Student : Zaheera binti Zainal Abidin

Student's ID No. : 2010271824

Programme : PhD in Science

Faculty : Faculty of Computer and Mathematical Sciences

Thesis Title : Furrow and Crypt Detection using

Modified Ant Colony Optimization for Iris Recognition

Signature of Student : ……….

(5)

ABSTRACT

Iris recognition has been widely recognized as one of the most performing biometric system. The accuracy performance of iris recognition system is measured by FAR (False Accept Rate) and FRR (False Reject Rate). FRR measures the genuine that is incorrectly denied by the system due to the changes in iris features (such as aging and health condition) and external factors that affected the iris image to be high in noise rate. The external factors such as technical fault, occlusion, and source of lighting caused the image acquisition which produce distorted iris images problem hence incorrectly rejected by the system. The current way of reducing FRR are wavelets and Gabor filters, cascaded classifiers, ordinal measure, multiple biometric modality and selection of unique iris features. Iris structure consists of unique features such as crypts, furrows, collarette, pigment blotches, freckles and pupil that are distinguishable among human. Previous research has been done in selecting the unique iris features however it shows low accuracy performance. As a solution, to improve the accuracy performance, this research proposes a new approach called as Modified Ant Colony Optimization that uses ant colony algorithm which search for crypts and radial furrow. The method consists of two tasks in obtaining the crypt and radial furrow features from the iris texture. The first task is the artificial ants that scan the pixel values according to the range of selected crypt or radial furrow. Then, the

scanned pixels value is searched based on degree of angle (0o_{, 45}o_{, 90}o_{and 135}o_{). The}

second task produces the confusion matrix and the blob of iris feature image is marked and indexed before stored into the database. In order to evaluate the performance of the proposed approach, FAR and FRR are measured with Chinese Academy of Sciences' Institute of Automation (CASIA) database for high quality images and Noisy Visible Wavelength Iris Image Databases (UBIRIS) database for noisy iris. By using CASIA version 3 image databases, the crypt feature shows that the result of FRR is 18.05% and radial furrow gives 81.5% when FAR at 0.1%. For UBIRIS version 1 database, the crypt feature indicates that the value FRR is 46.93% meanwhile the radial furrow shows the values of FRR 33.87% when FAR at 0.1%. To evaluate Modified Ant Colony Optimization, the genuine acceptance value (GAR) is measured to recognize iris features detection in low quality image environment. The experiment finding indicates that by using the Modified Ant Colony Optimization, radial furrow is able to be detected in distorted iris images with 84.62% since its own characteristics is obviously revealed. Moreover, the intersaction between FAR and FRR produces the Equal Error Rate (EER) with 0.21%, which indicated that equal error rate is lower than the previous standard value, which is 0.3%. Therefore, the advantages of using Modified Ant Colony Optimization are it has the capability to adapt with unique iris features in robust manner and use small amount of information in unique micro-characteristics of iris features to determine the user. The outcome of this new approach is to reduce the EER rates since lower EER rates indicates better accuracy performance. As a conclusion, the contribution of Modified Ant Colony Optimization extraction approach brings an innovation at the extraction process in the biometric technology and provides benefits to the communities.

(6)

ACKNOWLEDGMENT

First and foremost, I would like to thank Allah the Almighty, for his guidance, ideas and comfortable environment. I would like to extend my appreciation to the Universiti Teknikal Malaysia Melaka and Ministry of Education Malaysia for their generosity for awarding me the scholarship during this study. Thank you to my

supervisor, Professor Hj. Dr. Mazani Manaf, Faculty of Computer and Mathematical

Sciences, Universiti Teknologi MARA, and co-supervisors, Assoc Prof. Dr. Abdul Samad Shibghatullah, Department of Computer Systems and Communications, Faculty of Information and Communication Technology (FTMK), Universiti Teknikal Malaysia Melaka (UTeM). Without their sincere guidance, this work would not have been possible. A special thank and higly appreciation to Professor Dr. Anil K. Jain as an external examiner (international) with good comments, Assoc. Prof. Dr. Abdul Rahman Ramli as an external examiner (national), Dr. Noor Elaiza Abd Khalid (internal examiner), Professor Zhenan Sun and Associate Professor Dr. Sarat C. Dass for giving fruitful comments about my work. Thank you to Libor Masek who shares the matlab codes. An appreciation to Prof. Dr. Rabiah Ahmad and Associate Professor Dr. Choo Yun Huoy who have helped and supports in completing my study. A special thank to Syarulnaziah Anawar, Zakiah Ayop, Nor Azman Mat Ariff, Nurul Akmal Hashim, Dr. Zuraida Abal Abas, Ahmad Fadzli Nizam Abdul Rahman, Hidayah Rahmalan, and Mohd Zaki Mas‘ud for their constructive discussions and helps with the analysis and in thesis writing during the course of this study. Last but not least, from the bottom of my heart a gratitute to my family for their love and caring. Special thanks to my husband, Khairul Anwar Ibrahim, for his encouragements, my eternal love to all my children, Akmal al Husainy, Umairah al Husna, and Zakeeyah al Husna, for their patience and understanding. Finally, I would like to thank my beloved parents who have been the pillar of strength in all my endeavors. I am always deeply indebted to them for all their endless love and prayers that they have given me. Thank you to the individual(s) who providing me the inspiration to embark on my study.

(7)

Page

CONFIRMATION BY PANEL OF EXAMINERS ii

AUTHOR'S DECLARATION iii

ABSTRACT iv

ACKNOWLEDGMENT v

TABLE OF CONTENTS vi

LIST OF TABLES xi

LIST OF FIGURES xiii

LIST OF ABBREVIATIONS xix

CHAPTER ONE: INTRODUCTION 1

1.1 Overview of Iris Recognition 1

1.2 Background of the Research 2

1.3 Problem Statement 4

1.4 Research Aim 7

1.5 Research Scope 7

1.6 Research Significances 8

1.7 Contributions of the Research 9

1.8 Thesis Organization 11

CHAPTER TWO: LITERATURE REVIEW 14

2.1 Introduction 14

2.2 Human Iris Anatomy 15

(8)

vii

2.4 The Iris Recognition System 19

2.4.1 Enrollment Process 20

2.4.2 Comparison Process 33

2.5 The Existing Methods of Iris Features Detection 37

2.5.1 Wavelets and Log Gabor Filters 37

2.5.2 Cascaded Classifiers 38

2.5.3 Ordinal Measures 38

2.5.4 Multiscale 39

2.5.5 Feature Selection 41

2.6 The Theoretical Framework of Iris Recognition 43

2.7 The Basic Conceptual Framework of Iris Recognition 44

2.8 The Conceptual Framework of Iris Recognition 45

2.8.1 Swarm Intelligence Feature Extraction 46

2.8.2 The Propose Approach (modified ant colony optimization)

of Feature Extraction 50

2.9 Summary 60

CHAPTER THREE: RESEARCH METHODOLOGY 62

3.1 Introduction 62

3.2 Research Framework 63

3.3 Proposed Research Framework 65

3.3.1 Phase 1: Preliminary Study 66

3.3.2 Phase 2: Data Collection 66

3.3.3 Phase 3: Proposed Design and Prototype Construction 67

3.3.4 Phase 4: Results and Analysis 87

3.3.5 Phase 5: Performance Analysis 89

3.4 Software and Hardware Requirements 90

(9)

viii

CHAPTER FOUR: PRE-PROCESSING ANALYSIS 92

4.1 Introduction 92

4.2 The Experiment Environment 93

4.2.1 Thresholding process 94

4.2.2 Iris Segmentation Process 95

4.2.3 Iris Normalization Process 100

4.3 Experiment Results 102

4.3.1 Results of Iris Segmentation 102

4.3.2 Results of Iris Normalization 104

4.4 Experiment Findings 108

4.5 Summary 109

CHAPTER FIVE: MODIFIED ANT COLONY OPTIMIZATION FEATURE

EXTRACTION 110

5.1 Introduction 110

5.2 Experiment and Implementation 113

5.2.1 Swarm Intelligence Extraction 114

5.2.2 Experiment Setup with WEKA and Rapid Miner 115

5.2.3 Modified Ant Colony Optimization Approach 117

5.3 Experiment Results 121

5.3.1 Swarm Intelligence Extraction 126

5.3.2 Results on WEKA and Rapid Miner 130

5.3.3 Modified Ant Colony Optimization 130

5.4 Experiment Findings 146

5.4.1 The Evaluation of Current Extraction, ACO, PSO and

Modified Ant Colony Optimization 147

(10)

CHAPTER SIX: IMAGE BASED MATCHING USING CBIR 154

6.1 Introduction 154

6.2 Experiment Environment 155

6.3 Experiment Results 158

6.3.1 Ant Feature Index 160

6.3.2 Confusion Matrix 163

6.3.3 Accuracy Performance Evaluation and Validation for Iris

Image Matching 165

6.3.4 Accuracy Performance Evaluation and Validation for Iris

Image Matching 167

6.3.4.1 Iris Database Used for Validation 168

6.3.4.2 Existing Approach Used for Validation 168

6.4 Experiment Findings 169

6.5 Summary 173

CHAPTER SEVEN: THE NEW APPROACH OF IRIS RECOGNITION (Modified Ant Colony Optimization + MATCHING) AND EXPERIMENT

FINDINGS 175

7.1 Introduction 175

7.2 Experiment Environment 176

7.3 Experiment Results 178

7.3.1 ACO and Modified Ant Colony Optimization Pattern of Ant Movement 178

7.3.2 Threshold Value Setup 181

7.3.2 The ROC Curve 184

7.4 Experiment Findings 187

7.4.1 High Quality Iris Images 187

7.4.2 Low Quality Iris Images (Noisy Iris) 188

7.5 Discussion 188

7.5.1 Modified Ant Colony Optimization Implementation 188

(11)

7.6 Summary 189

CHAPTER EIGHT: CONCLUSION 191

8.1 Conclusion 193

8.2 Recommendation 195

REFERENCES 196

(12)

LIST OF TABLES

Table Title Page

Table 2.1 Statistics of CASIA 22

Table 2.2 The Critical Review Evaluation with RQ1, RQ2 and RQ3 Map 55

Table 2.3 Comparison of Data Mining Tools 56

Table 3.1 Research Methodology Mapping with RO1, RO2 and RO3 91

Table 4.1 PSNR Values of Canny and Sobel using CASIA database 98

Table 4.2 HT and IDO Techniques for Iris Segmentation Process 102

Table 4.3 Open Iris Database Accuracy Performance 103

Table 4.4 Iris Normalization 104

Table 4.5 Iris Normalization Process 104

Table 5.1 The Parameter Settings for PSO, ACO and Modified Ant Colony

Optimization 117

Table 5.2 The Sample of Scoresheet for Precision and Recall Calculation 121

Table 5.3 Texture Extraction Results 122

Table 5.4 Bio-inspired Feature Selection in Texture Analysis Extraction 127

Table 5.5 Modified Ant Colony Optimization Pheromone Table based on

numberof iteration and index for CASIA 133

Table 5.6 Modified Ant Colony Optimization Pheromone Table based on

(13)

xii

Table 5.7 Modified Ant Colony Optimization Scoresheet for Precision and

Recall Calculation 135

Table 5.8 Summary of Modified Ant Colony Optimization Precision 139

Table 5.9 Experiments Results of PSO, ACO and Modified Ant Colony

Optimization in Extraction 141

Table 5.10 The Characteristics in PSO, ACO and Modified Ant Colony

Optimization 147

Table 6.1 Iris Database Classification 156

Table 6.2 Iris Matching in Testing Set 164

Table 6.3 Results of Sample Data for Testing in Matching Process 167

Table 6.4 Comparison of ACO and Modified Ant Colony Optimization 170

Table 7.1 Score sheet results of precision using Modified Ant Colony

(14)

xiii

LIST OF FIGURES

Figures Title Page

Figure 1.1 Taxonomy of Contribution Map 11

Figure 1.2 Research Schematic Diagram 13

Figure 2.1 Iris Structure 16

Figure 2.2 Crypt 16

Figure 2.3 Furrow 17

Figure 2.4 Sample of pigment melanin 17

Figure 2.5 Sample of rare blotches in blob of iris features 17

Figure 2.6 Daugman‘s Approach of Iris Recognition 19

Figure 2.7 Iris Biometric System Phase 20

Figure 2.8 The customized or self-developed iris camera CASIA 22

Figure 2.9 Examples of iris images in CASIA-Iris-Interval 22

Figure 2.10 UBIRIS version 1 24

Figure 2.11 Circular Segmentation 25

Figure 2.12 Non Circular Segmentation 25

Figure 2.13 Partial Segmentation 26

Figure 2.14 Real and Imaginary Parts 31

Figure 2.15 Energy Histogram in DCT sub-band 31

Figure 2.16 Compression in Iris 32

Figure 2.17 Compression and Decomposition 32

Figure 2.18 Iris texture pattern 34

(15)

xiv

Figure 2.20 Problems in Iris Recognition 36

Figure 2.21 Pupilary zone contraction due to light source 37

Figure 2.22 Ordinal measures 39

Figure 2.23 Segmentation and Normalization using Multiscale 40

Figure 2.24 Feature Extraction using Multiscale 41

Figure 2.25 Feature Selection Method in Iris Recognition 42

Figure 2.26 System framework with manual inspection 42

Figure 2.27 The Theoretical Framework of Iris Recognition 43

Figure 2.28 The Basic Conceptual Framework of Iris Recognition 44

Figure 2.29 The Conceptual Framework of Iris Recognition 45

Figure 2.30 The Principle of ACO 50

Figure 2.31 The Modified Ant Colony Optimization Model 51

Figure 2.32 Formula of Precision, Recall and Accuracy 59

Figure 3.1 The Structure of Chapter 3 62

Figure 3.2 Research Framework 63

Figure 3.3 The Proposed Research Framework 65

Figure 3.4 Preliminary Study Phase 66

Figure 3.5 Data Collection Phase 66

Figure 3.6 Pre-processing Phase 67

Figure 3.7 Iris Normalization 69

Figure 3.8 Extraction Phase 71

Figure 3.9 The Overview of Modified Ant Colony Optimization Approach 72

Figure 3.10 The Procedure of Modified Ant Colony Optimization Approach 73 Figure 3.11 Region of interests for radial furrow, crypt, eyelids and eyelashes 75

(16)

Figure 3.13 Ant Movements Based on Degrees of Angle 77

Figure 3.14 The Ant Stigmergy 80

Figure 3.15 The Pseudocode of Co-occurrence Matrix 82

Figure 3.16 The Comparison Process in Iris Recognition 83

Figure 3.17 The Image based Matching in Iris Recognition 85

Figure 3.18 The Identification and Verification Mode in Matching Process 86

Figure 3.19 Results and Analysis 87

Figure 3.20 Performance Analysis 89

Figure 4.1 The experimental framework of pre-processing techniques 93

Figure 4.2 Iris Pre-conditioning Phase 94

Figure 4.3 Pseudocode of IDO 97

Figure 4.4 Edge Detection Techniques 98

Figure 4.5 The application of Canny edge detection to iris image 99

Figure 4.6 Pseudocode of Hough Transform 100

Figure 4.7 The iris normalization for comparison process 105

Figure 4.8 Comparison of GAR based FRR 105

Figure 4.9 Comparison of GAR based RE-rate 106

Figure 4.10 Average of circle pupil and circle iris comparison in CASIA 106

Figure 4.11 FAR versus FRR 107

Figure 4.12 FAR versus FRR Threshold 108

Figure 5.1 Experiment Planning for Iris Extraction using PSO and ACO 112

Figure 5.2 Experiment Planning for Iris Extraction using MACO 112

Figure 5.3 The General Experimental Configuration 116

Figure 5.4 The Flowchart of Extraction Phase using MACO 118

(17)

xvi

Figure 5.6 CASIA Precision based on Texture Analysis Extraction 123

Figure 5.7 UBIRIS Precision of Texture Analysis Extraction 123

Figure 5.8 Accuracy of Texture Analysis Approach in CASIA 124

Figure 5.9 Accuracy of Texture Analysis Approach in UBIRIS 124

Figure 5.10 Accuracy of iris texture extraction for CASIA.V3 125

Figure 5.11 Accuracy of iris texture extraction for UBIRIS.V1 126

Figure 5.12 Precision of ACO in CASIA 128

Figure 5.13 Precision of ACO in UBIRIS 128

Figure 5.14 Precision of PSO in CASIA 128

Figure 5.15 Precision of PSO in UBIRIS 128

Figure 5.16 Accuracy of ACO in CASIA 129

Figure 5.17 Accuracy of ACO in UBIRIS 129

Figure 5.18 Accuracy of PSO in CASIA 129

Figure 5.19 Accuracy of PSO in UBIRIS 129

Figure 5.20 Load Image Interface 131

Figure 5.21 Segment and Normalize Interface 131

Figure 5.22 Modified Ant Colony Optimization Extraction Interface 131

Figure 5.23 Modified Ant Colony Optimization Matching Interface – Match 131 Figure 5.24 Modified Ant Colony Optimization Matching Interface–Not Match 131 Figure 5.25 Precision based on Number of Ant Iteration in CASIA (Furrow) 136 Figure 5.26 Precision based on Number of Ant Iteration in UBIRIS (Furrow) 136 Figure 5.27 Precision based on Number of Ant Iteration in CASIA.v3 (Crypt) 137 Figure 5.28 Precision based on Number of Ant Iteration in UBIRIS.v1 (Crypt) 137

Figure 5.29 Ant Feature Marking of Crypts (CASIA) 138

(18)

xvii

Figure 5.31 Ant Feature Index 139

Figure 5.32 XAMPP 142

Figure 5.33 The Connection Linker 142

Figure 5.34 The MySQL 143

Figure 5.35 Ant Feature Index of Crypts (CASIA) – Marking 143

Figure 5.36 Ant Feature Index of Furrow (CASIA) – Marking 143

Figure 5.37 Indexed Iris Feature Retrieval Database 144

Figure 5.38 Confusion Matrix of Crypt using the MACO in CASIA 144

Figure 5.39 Confusion Matrix of Radial Furrow using the MACO in CASIA 145

Figure 5.40 WEKA sceen captured for MACO for Crypt in UBIRIS 145

Figure 5.41 WEKA sceen captured Results for MACO for Crypt in UBIRIS 146

Figure 5.42 Accuracy Performance of Extraction (UBIRIS.V1) 148

Figure 5.43 Accuracy Performance of Extraction (CASIA.V3) 149

Figure 5.44 Accuracy Performance of Bio-inspired Extraction 150

Figure 5.45 Accuracy Performance of MACO Extraction 151

Figure 5.46 A Comparison of Crypt based on PSO, ACO and MACO 152

Figure 5.47 A Comparison of Furrow based on PSO, ACO and MACO 153

Figure 6.1 The Process of Image Matching using CBIR 155

Figure 6.2 Flowchart of Matching Phase 158

Figure 6.3 IrisCodes using Libor Masek for User A 159

Figure 6.4 Iris codes using Libor Masek for User B 159

Figure 6.5 Modified Ant Colony Optimization for User A 160

Figure 6.6 Modified Ant Colony Optimization for User B 160

Figure 6.7 Ant Feature Index of Crypts (CASIA) – Ant Crypt Index 161

(19)

xviii

Figure 6.9 Ant Feature Marking based on in MACO 161

Figure 6.10 Ant Feature Indexed based on Hamming Distance 162

Figure 6.11 The Matching Process using Datasets 163

Figure 6.12 The Confusion Matrix 165

Figure 6.13 Accuracy of Ant-CBIR-RF-CASIA 169

Figure 6.14 Accuracy of Ant-CBIR-RF-UBIRIS 169

Figure 6.15 Accuracy of Ant-CBIR-Crypt-CASIA 169

Figure 6.16 Accuracy of Ant-CBIR-Crypt-UBIRIS 169

Figure 6.17 Summary of Crypt and Furrow using MACO 171

Figure 6.18 Crypt and Radial Furrow Evaluation of GAR before and after the Matching Process using the Modified Ant Colony Optimization

and ACO 172

Figure 7.1 The Schematic Diagram of Chapter 7 175

Figure 7.2 The New Approach of Iris Recognition 176

Figure 7.3 Genuine Score Matrix 177

Figure 7.4 FAR versus FAR for High Quality Iris (CASIA.v3) 181

Figure 7.5 FAR versus FAR for Noisy Iris (UBIRIS.v1) 182

Figure 7.6 ROC Curve of Iris Features based on Traditional, ACO and

Modified Ant Colony Optimization Approach 186

(20)

xix

LIST OF ABBREVIATIONS

Abbreviations

RP1 Research Problem 1

RP2 Research Problem 2

RP3 Research Problem 3

RQ1 Research Question 1

RQ2 Research Question 2

RQ3 Research Question 3

RO1 Research Objective 1

RO2 Research Objective 2

RO3 Research Objective 3

RC1 Research Contribution 1

RC2 Research Contribution 2

RC3 Research Contribution 3

FAR False Acceptance Rate

FRR False Rejected Rate

EER Equal Error Rate

HD Hamming Distance

ACO Ant Colony Optimization

MACO Modified Ant Colony Optimimzation

PSO Particle Swarm Optimization

CBIR Content Based Image Retrieval

Threshold Filter or Cut off Value

FMR False Match rate

(21)

Enrollment Create and Store a Biometric Enrollment Data Record with

Biometric Policy of Enrollment

Blob of Iris Features The unique micro-characteristics inside the iris features such as crypt, furrow, collarette and pigment melanin.

(22)

CHAPTER ONE

INTRODUCTION

1.1 OVERVIEW OF IRIS RECOGNITION

Biometrics has been used worldwide as a reliable source of identification system for many applications such as restricted area access control, database access, computer login, building entry, airport security, forensic application systems and automatic teller machine (ATM). Compared to existing identification system (i.e.: smartcard and RFID), biometrics offers higher accuracy, security, efficiency, availability, uniqueness and superior performance. In most application, biometric recognition system scan a person‘s body parts, extract unique features and stored them in a secured database as biometric template. Then, later, when the system is invoked again by a user (e.g. a user scans his/her body parts to gain access), the system compares the database with the existing biometric template and provides indication whether the scanned images matches any of the existing iris template. If it matches, then the system allows the user to gain access, else, the system will deny access.

In biometrics, various modalities such as facial shape, fingerprint, handwriting, and iris have been used for human identification and access control. Iris recognition stands out as a promising method for obtaining automated, secure, reliable, fast and high in accuracy for user identification which typically achieve 99% accuracy rate with equal error rates of less than 1% [1].

Iris recognition is an autonomous system that uses complex mathematical

pattern recognition, image processing and machine learning techniques for measuring the iris [2]. Inside the human iris, there are many unique features such as crypts, radial furrows, concentric furrows, collarette, freckles, pupil and pigment blotches which distinguish the genuine characteristics of a person, thus making it suitable for recognition purposes. However, the demand for higher accuracy and high speed recognition in biometric system leads to continuous proposals of new iris recognition

(23)

method since there are still concers on the arising number of impostor situation being

reported from the existing biometric system [3].

In biometric recognition process, impostor or incorrectly recognized users can be categorized into two types. The first type (Type I) of impostor is when the biometric system rejects the genuine user who wants to access the system. In most applications, a genuine user is defined as the person who is officially allowed by the system owner to gain access to a certain secured system and have his/her own biometric template captured and stored. Meanwhile, the second type (Type II) of impostor is someone who tries to penetrate the biometric systems and pretends to be the original person.

the system detects what is commonly known as ―iris distortion‖ and thus, failure to

match the scanned and stored biometrics identity. In the context of iris recognition system, iris distortion means that the iris characteristics captured by the system are detected to be significantly different and cannot be matched for similarity from any of the original iris template registered in the database. Section 1.2 shall deliberate in detail the cause of the iris distortion problem in iris recognition.

1.2 BACKGROUND OF THE RESEARCH

From previous works, the cause of distortion to the captured iris images can be categorized into two: i) dynamic nature of iris characteristics, ii) occlusion. The first category is defined as a situation where the iris itself changes due to human‘s biological factors, such as aging, growth, emotion, diet, health problems and eye surgery. In fact, the colour of the iris texture may also change due to inheritance and epigenetic diversity from different races. The constantly changing iris texture creates difficulties at the comparison phase to determine either the captured iris data are genuine or not. Previous studies shows that failure was detected in 21% of intra-class

comparisons cases, taken at both three and six months intervals [4].

The second category of distorted iris source simply means that the iris images could not be captured accurately due to some physical obstructions or occlusions,

(24)

such as eyelids and eyelashes [5], as well as the presence of contact lenses or spectacles. The occlusions affect some of the important and vital iris features which are unable to be captured, thus contribute to the increased error rate.

Therefore, many researchers have done studies to overcome the problem of distorted iris features and occlusion based on elimination of the unwanted noise [6], [7], enhance the noise level [8], white noise application and removal [9], non-circular segmentation process [10]–[13] and feature selection [14]. However, most of the research outcome indicates that existing solutions, to a certain extent, still incapable to reduce or eliminate noisy iris problems. Among all proposed solution, in order to solve this problem, the recommendation is to use only a certain unique part of the iris structure which remains unchanged for iris recognition. The unique part of the iris texture consists of crypts, furrows, collarette, pupil, freckles and blotches [2]. Subsequently, some studies found that the micro characteristics in iris features have been mostly stable for recognition [15], [16]. Nonetheless, the unique iris feature sustained only for a certain period of time, stated in [17], [18] and only up to six years, as stated in [19].

Studies on selecting the iris texture from its original structure have been gaining attention from some researchers [20]–[26]. Hence, feature selection is vital for choosing a subset features of available unique features by eliminating unnecessary features since the information of iris features obtained can be tremendously huge and consequently consumes a lot of computational resources [14]. In fact, in feature selection, it is observed that the existing method lacks of natural computational element in the iris recognition.

Therefore, the unique iris feature selected is based on the best features points from the entire iris texture which is required in learning the changes or instability in iris texture intelligently. The natural computational algorithms consist of artificial neural networks, artificial immune systems, evolutionary algorithms, and swarm intelligence.

In swarm intelligence, the particle swarm optimization (PSO), genetic algorithms (GA), and ant colony optimization (ACO) are the most used nature-inspired algorithms to solve optimization problems. Swarm algorithm is chosen based on winning their winninf criteria. Two of the most prominent criteria which makes the preferred algorithms are that these algorithms are able to search the elements

(1)

xviii

Figure 6.9 Ant Feature Marking based on in MACO 161

Figure 6.10 Ant Feature Indexed based on Hamming Distance 162

Figure 6.11 The Matching Process using Datasets 163

Figure 6.12 The Confusion Matrix 165

Figure 6.13 Accuracy of Ant-CBIR-RF-CASIA 169

Figure 6.14 Accuracy of Ant-CBIR-RF-UBIRIS 169

Figure 6.15 Accuracy of Ant-CBIR-Crypt-CASIA 169

Figure 6.16 Accuracy of Ant-CBIR-Crypt-UBIRIS 169

Figure 6.17 Summary of Crypt and Furrow using MACO 171 Figure 6.18 Crypt and Radial Furrow Evaluation of GAR before and after the

Matching Process using the Modified Ant Colony Optimization

and ACO 172

Figure 7.1 The Schematic Diagram of Chapter 7 175

Figure 7.2 The New Approach of Iris Recognition 176

Figure 7.3 Genuine Score Matrix 177

Figure 7.4 FAR versus FAR for High Quality Iris (CASIA.v3) 181 Figure 7.5 FAR versus FAR for Noisy Iris (UBIRIS.v1) 182 Figure 7.6 ROC Curve of Iris Features based on Traditional, ACO and

Modified Ant Colony Optimization Approach 186 Figure 8.1 The Objectives and Contributions Mapping 194

(2)

xix

LIST OF ABBREVIATIONS

Abbreviations

RP1 Research Problem 1

RP2 Research Problem 2

RP3 Research Problem 3

RQ1 Research Question 1

RQ2 Research Question 2

RQ3 Research Question 3

RO1 Research Objective 1

RO2 Research Objective 2

RO3 Research Objective 3

RC1 Research Contribution 1

RC2 Research Contribution 2

RC3 Research Contribution 3

FAR False Acceptance Rate

FRR False Rejected Rate

EER Equal Error Rate

HD Hamming Distance

ACO Ant Colony Optimization

MACO Modified Ant Colony Optimimzation

PSO Particle Swarm Optimization

CBIR Content Based Image Retrieval Threshold Filter or Cut off Value

FMR False Match rate

(3)

Enrollment Create and Store a Biometric Enrollment Data Record with Biometric Policy of Enrollment

Blob of Iris Features The unique micro-characteristics inside the iris features such as crypt, furrow, collarette and pigment melanin.

(4)

CHAPTER ONE

INTRODUCTION

1.1 OVERVIEW OF IRIS RECOGNITION

Iris recognition is an autonomous system that uses complex mathematical pattern recognition, image processing and machine learning techniques for measuring the iris [2]. Inside the human iris, there are many unique features such as crypts, radial furrows, concentric furrows, collarette, freckles, pupil and pigment blotches which distinguish the genuine characteristics of a person, thus making it suitable for recognition purposes. However, the demand for higher accuracy and high speed recognition in biometric system leads to continuous proposals of new iris recognition

(5)

method since there are still concers on the arising number of impostor situation being reported from the existing biometric system [3].

In fact, this is a situation where the person does not officially allowed to gain access to a system and does not have any biometric template captured. However, in either situation, the reason why the system behaves in such condition is because when the system detects what is commonly known as ―iris distortion‖ and thus, failure to match the scanned and stored biometrics identity. In the context of iris recognition system, iris distortion means that the iris characteristics captured by the system are detected to be significantly different and cannot be matched for similarity from any of the original iris template registered in the database. Section 1.2 shall deliberate in detail the cause of the iris distortion problem in iris recognition.

1.2 BACKGROUND OF THE RESEARCH

The second category of distorted iris source simply means that the iris images could not be captured accurately due to some physical obstructions or occlusions,

(6)

Furrow And Crypt Detection Using Modified Ant Colony Optimization For Iris Recognition.

UNIVERSITI TEKNOLOGI MARA

FURROW AND CRYPT DETECTION

USING MODIFIED

ANT COLONY OPTIMIZATION

FOR IRIS RECOGNITION

ZAHEERA ZAINAL ABIDIN

PhD

UNIVERSITI TEKNOLOGI MARA

FURROW AND CRYPT DETECTION

USING MODIFIED

ANT COLONY OPTIMIZATION

FOR IRIS RECOGNITION

ZAHEERA ZAINAL ABIDIN

Thesis submitted in fulfillment

of the requirements for the degree of

Doctor of Philosophy

Faculty of Computer and Mathematical Sciences

CONFIRMATION BY PANEL OF EXAMINERS

AUTHOR'S DECLARATION

ABSTRACT

ACKNOWLEDGMENT

TABLE OF CONTENTS

LIST OF TABLES

LIST OF FIGURES

LIST OF ABBREVIATIONS

CHAPTER ONE

INTRODUCTION

LIST OF ABBREVIATIONS

CHAPTER ONE

INTRODUCTION

Parts

Dokumen yang terkait

Implementasi Algoritma Ant Colony Dalam Pencarian Lokasi Rumah Sakit Berbasis Mobile Gis Pada Platform Android

Studi Penggunaan Algoritma Ant Colony Dalam Pengalokasian Kanal Dinamik Pada Komunikasi Seluler

Algoritma Ant Colony System Dalam Penjadwalan Kegiatan Belajar Mengajar Di Sekolah Dasar

SEGMENTASI CITRA MENGGUNAKAN ANT COLONY OPTIMIZATION

ant-CBIR: a new method for radial furrow extraction in iris biometric.

ant-CBIR: a new method for radial furrow extraction in iris biometric.

Intelligent Vacation Planner System Using Ant Colony Optimization.

An ant colony optimization algorithm for (1)

An ant colony optimization algorithm for

ANT COLONY OPTIMIZATION DALAM PENYELESAIAN

Dokumen yang Anda mencari sudah siap untuk unduhkan