• Sonuç bulunamadı

TABLE OF CONTWNTS

N/A
N/A
Protected

Academic year: 2021

Share "TABLE OF CONTWNTS"

Copied!
1
0
0

Yükleniyor.... (view fulltext now)

Tam metin

(1)

TABLE OF CONTWNTS

ACKNOWLEDGEMENTS...i

ABSTRACT...ii

TABLE OF CONTENTS...iii

LIST OF TABLES...v

LIST OF FIGURES...vi

LIST OF ABBREVIATIONS... viii

INTRODUCTION...1

CHAPTER ONE: LITERATURE REVIEW 1.1 Overview...3

1.2 Literature Background...3

1.3 Forensic History Timeline...3

1.4 Forensic Dentistry...6

1.5 Dental Radiograph...8

1.6 Summary...9

CHAPTER TWO: NEURAL NETWORK 2.1 Overview...10

2.2 Artificial Neural Networks...10

2.3 Human Brain...12

2.4 Neural Network Components ...14

2.5 Models of Neurons...14

2.5.1 Types if Activation Function ...15

2.6 Neural Network Architectures ...18

2.6.1. Single Layer Perceptron...18

2.6.2 Multi Layer Perceptron ...19

2.6.3. Recurrent Networks ...20

2.7 Artificial Neural Network Learning...21

2.7.1 Learning Paradigms...21

2.7.1.1. Supervised Learning...22

2.7.1.2. Unsupervised Learning (Self Organization Learning)...22

2.7.2. Learning Rule...22

2.8 Advantages of Artificial Neural Networks...24

2.9 Back Propagation Algorithm ...25

2.9.1 Feed Forward Network Stage...27

2.9.2 Error Signal...28

2.9.3 Mean Squared Error Function...29

2.9.4 Learning Rate and Momentum Factor...29

2.9.5 Weight Adjusting...29

2.9.6 Back Propagation Training...30

2.10 Summary...32

CHAPTER THREE: DENTAL IDENTIFICATION SYSTEM 3.1 Overview...33

3.2 Dental Identification System Operations...33

3.2.1 Database Collecting...34

3.2.2 Preprocessing...35

3.2.2.1 Image Normalization...35

3.2.2.2 Image Compression ...36

III

(2)

3.2.2.3 Image Reading...39

3.2.2.4 Image Vectorization...40

3.3 Using Matlab...40

3.4 Implementation of Neural Network...40

3.4.1 Define Back Propagation Architecture and Parameters...41

3.4.2 Formulation of Weight Adjustment...41

3.4.3 Define the Learning Rate and Momentum Factor...41

3.4.4 Define the Maximum Error...41

3.5 Training Using Back Propagation...42

3.6 Testing the Dental Identification System...48

3.7 Summary...49

CHAPTER FOUR EXPERIMENTAL RESULTS AND COMPARISON 4.1 Overview...50

4.2 General Experimental Setup...50

4.3 First Experiment...50

4.3.1 First Quarter First Experiment ...50

4.3.2 Second Quarter First Experiment ...52

4.3.3 Third Quarter First Experiment ...54

4.3.4 Fourth Quarter First Experiment ...56

4.4 Second Experiment...59

4.4.1 First Quarter Second Experiment ...59

4.4.2 Second Quarter Second Experiment ...61

4.4.3 Third Quarter Second Experiment ...63

4.4.4 Fourth Quarter Second Experiment ...65

4.5 Extra Testing Experiment...67

4.6 Discussion...68

4.7 Summary...70

CONCLUSION AND FUTURE RESEARCH...71

REFERENCES...72

APPENDIX I DATABASE ...75

APPENDIX II EXPERIMENT RESULTS...80

APPENDIX III MATLAB SOURSE CODE...100

IV

Referanslar

Benzer Belgeler

Ülkemiz kolemanit cevherlerinin özellikle arsenikli olan örnekleri toplanarak yapılan çalışmalar arsenik miktarının en fazla 500 ppm civarında olduğunu .ortaya koymuştur. Bu

En son kısımda ise yaygın olarak çevremizde yaygın olarak kullanılan bazı cep telefonları üzerinde yapılan elektrik alan ölçüm sonuçları verilecek ve elde

trutta populasyonunun erkek bireylerinde kolesterol seviyesi ile GSİ değerleri arasındaki ilişkisi incelendi ve bu ilişki de kolesterol seviyesinin üremeden önceki nisan ayında

Araştırma süresince akım, su sıcaklığı, pH, elektriksel iletkenlik, çözünmüş oksijen arazide yapılan ölçümlerle belirlenirken, toplam sertlik, toplam azot,

environment, humans started to change their environment for their future goals (Broeckmann, 2016: 18). Operational tools called machines were created -more or less there are

Bugün, müze için gerek­ li ödenek sağlanırsa, kapsamlı, kalıcı bir restorasyon, balam çalışmasının yapılacağını söylüyor Tunç Tüfekçi ve müzenin

However, when we adjust the impact of bank loans to the size of the local economy (provin- cial GDP), increases in banking activities are found to lower the output per capita as

Mean number of trials taken to complete context and single colour training across the three days of training and on the testing day (refresher training).. Neural pathways from the