Geri Dön

Yapay sinir ağları kullanarak parmakizi analizi

Fingerprint recognition by using neural networks

  1. Tez No: 332403
  2. Yazar: SÜHELDAL GÜRDAL
  3. Danışmanlar: PROF. DR. OKYAY KAYNAK
  4. Tez Türü: Yüksek Lisans
  5. Konular: Elektrik ve Elektronik Mühendisliği, Electrical and Electronics Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2012
  8. Dil: Türkçe
  9. Üniversite: Hava Harp Okulu Komutanlığı
  10. Enstitü: Havacılık ve Uzay Teknolojileri Enstitüsü
  11. Ana Bilim Dalı: Elektronik Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Haberleşme Bilim Dalı
  13. Sayfa Sayısı: 93

Özet

Bu çalışma, Otomatik Parmak İzi Tanıma sistemlerinin en temel işlem basamağı olan parmak izi özellik vektörünün oluşturulmasında Yapay Sinir Ağı (YSA) mimarisinin kullanılabilirliğini incelemek amacıyla hazırlanmıştır.Otomatik parmak izi algoritmalarının çalışabilmesi için seçilen parmak izi karakteristik özelliklerinin belirginleştirilmesi ve sınıflandırılması gerekmektedir. Bu sayede minimum işlem ve zaman maliyetiyle, teşhis ve tanımlama safhasında ihtiyaç duyulan nitelik ve nicelikteki veriyi içeren bir öznitelik vektörü elde edilebilmektedir. Literatürde çalışmalar öznitelik vektörlerini doğru biçimde elde edebilecek algoritmaların oluşturulması ve aynı parmak izine ait farklı koşullar altında elde edilen görüntülerin optimizasyonu ile teşhis/tanıma safhasında kullanılabilecek en esnek algoritmanın elde edilmesi üzerine yoğunlaşmaktadır.Bu tez çalışması kapsamında, parmak izi öznitelik vektörünün satır sonu ve çatal noktalar kullanarak oluşturulması hedeflenmiştir. Gelişen bilgisayar işlem kapasitesi sayesinde günümüzde öznitelik vektörlerinin oluşturulmasında, parmak izi karakteristik özelliklerinin pek çoğunun bir arada ele alındığı uygulamalar yaygın olmakla birlikte, bu tür uygulamalarda çoğunlukla görüntü işleme teknikleri tercih edilmekte, bununla birlikte YSA benzeri akıllı sistemler daha çok tanıma/teşhis aşamasında tercih edilmektedir. Bununla birlikte öznitelik vektörünün oluşturulmasında akıllı sistemlerin, özellikle YSA mimarisinin kullanımı veri işlem kapasitesinin çok daha kısıtlı olduğu 1980'lere dayanmakta ve halen geçerliliğini korumaktadır. Bu yaklaşımla veri işlem bütçesinin kısıtlı olduğu koşullar altında bile günümüz algoritmalarıyla yarışabilecek niteliklere sahip olan ve tez kapsamında sunulan metodolojinin oldukça tutarlı sonuçlar ortaya koyduğu tespiti yapılmıştır. Çalışmada İleri Beslemeli Geri Yayılımlı Çok Katmalı YSA mimarisi tercih edilmiştir.

Özet (Çeviri)

A wide variety of systems requires reliable personal recognition schemes to either confirm or determine the identity of an individual requesting their services. The purpose of such schemes is to ensure that the rendered services are accessed only by a legitimate user and no one else. Examples of such applications include secure access to buildings, computer systems, laptops, cellular phones, and ATMs. In the absence of robust personal recognition schemes, these systems are vulnerable to the wiles of an impostor. Biometric recognition or, simply, biometrics refers to the automatic recognition of individuals based on their physiological and/or behavioral characteristics. By using biometrics, it is possible to confirm or establish an individual?s identity based on ?who she is,? rather than by ?what she possesses? (e.g., an ID card) or ?what she remembers? (e.g., a password).Any human physiological and/or behavioral characteristic can be used as a biometric characteristic as long as it satisfies the following requirements:Universality: Each person should have the characteristic.Distinctiveness: Any two persons should be sufficiently different in terms of the characteristic.Permanence: The characteristic should be sufficiently invariant (with respect to the matching criterion) over a period of time.Collectability: The characteristic can be measured quantitatively.A practical biometric system should meet the specified recognition accuracy, speed, and resource requirements, be harmless to the users, be accepted by the intended population, and be sufficiently robust to various fraudulent methods and attacks to the system. An finger print is the one of the ancient biometric system that satisfies all of them.A practical biometric system should meet the specified recognition accuracy, speed, and resource requirements, be harmless to the users, be accepted by the intended population, and be sufficiently robust to various fraudulent methods and attacks to the system.A fingerprint usually appears as a series of dark lines that presnt the high, peaking portion of the friction ridge skin, while tehe valleys between these rides appears as white spaceand are the low, shallow portion of the friction ridge skin. Finger print identification is based primarily on the minutiae, or the location and drection of the ride endings and bifurcations along a ride path. The types of information that can be collected from a fingerprint?s friction ridge impression include the flow of the frictions rides, the presence or absence of features along the individual friction ridge paths and their sequence, and the inrticate detail of a singel ridge. Recognation is usually based on the first and the second one or just he latter.In this study, the availability of using Neural Networks architecture in the forming fingerprint feature vectors, which is the fundamental process step on Automatic Fingerprint Recognition Systems, has been investigated and analyzed. Most of the pattern recognition systems are composed of four main steps. The first step is image acquisition, like convertin a scene into an array of numbers that can be manipulated by the computer. The second step is preprocessing, which involves removing noise, enhancing the picture and segmenting the image into meaningul regions to be analyzed separetaly. The third phase is feature extraction, in which the image is represented by a set of numerical features to remove redundancy from data and reduce its dimensions. The very last step is classification where a class label is assigned to the unknown image by examining its exracted features and compairn them with class representations that the classifier has learned during its training stage.Neural networks enable solutions to be found to problems where algoritmic methods are too computationally intesnsive or do not exist. They also offer significant speed advantages over conventional techiniques. The problems of feature extaction and classsification therefore seem to be a suitible application for neural networks.In order to make fingerprint algorithms work, characteristic features of the chosen fingerprint, also called minutiae, needs to be enriched and classified. So that it is possible to form a feature vector of related fingerprint which can also be used in the recognition phase of a designed fingerprint recognition algorithm. The crucial step in this process is to contain enough minutiae detail in the vector to ensure a correct segregation while keeping process and time cost reasonable. In studies two main areas has been defined in fingerprint recognition. One of them is to form algorithms which can form feature vectors precisely by using different fingerprint images. The other is present a method which can recognize different images of very same fingerprint, of which taken under different circumstances. In this work the first one has been considered.In this study, among the characteristics of the fingerprint images, bifurcations and ridge ends are chosen to be used to form feature vectors of related fingerprints. Thanks to improving processing power of the computer, most of the algorithms used for feature extraction and to form feature vectors consider gathering every possible characteristics of the image. Latest studies focus on consisting of more minutiae detail rather than computer processing cost and prefer using image processing algorithms instead of clever algorithms like Neural Networks. In up-to-date algorithms Neural Networks are preferred in specifying the owner of fingerprint image. On the other hand in some applications neural networks are still popular in forming feature vectors since 1980, in which processing budget was the biggest issue. In the study it is shown that proposed method can form feature vectors precisely by using Feed Forward Multi Layer Neural Network with Back Propagation.Although the backpropagation algorithm can be used very generally to train neural networks, it is most famous for applications to layered feedforward networks, or multilayer perceptrons. Simple perceptrons are very limited in their representational capabilities. In between the input layer issue. We will consider multilayer perceptrons with L layers of synaptic connections and L + 1 layers of neurons. This is sometimes called an L-layer network, and sometimes an L + 1-layer network. A network with a single layer can approximate any function, if the hidden layer is large enough. This has been proved by a number of people, generally using the Stone-Weierstrass theorem. So multilayer perceptrons are representationally powerful.In the study it is shown that proposed method can form feature vectors precisely by using Feed Forward Multi Layer Neural Network with Back Propagation. The most improtant step in fingerprint recognation has been achieved with this algoritm.

Benzer Tezler

  1. Yapay sinir ağları ile konuşmacı doğrulama sistemleri için saldırı tespiti

    Using artificial neural network on anti-spoofing for speaker verification

    BEKİR BAKAR

    Yüksek Lisans

    Türkçe

    Türkçe

    2018

    Elektrik ve Elektronik MühendisliğiBursa Teknik Üniversitesi

    Elektrik-Elektronik Mühendisliği Ana Bilim Dalı

    DOÇ. DR. CEMAL HANİLÇİ

  2. Esnek hesaplama yöntemleri ile parmak izi yüz biyometrik özelliklerinin ilişkilendirilmesi

    Investigating relationships between fingerprint and face biometrics by soft computing based methods

    NECLA ÖZKAYA

    Doktora

    Türkçe

    Türkçe

    2009

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolErciyes Üniversitesi

    Bilgisayar Mühendisliği Bölümü

    PROF. DR. ŞEREF SAĞIROĞLU

  3. Algısal özet fonksiyonları tabanlı derin öğrenme yöntemleri kullanılarak imgelerin sınıflandırılması

    Classification of images by using deep learning methods based on perceptual hash functions

    FATİH ÖZYURT

    Doktora

    Türkçe

    Türkçe

    2019

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolFırat Üniversitesi

    Yazılım Mühendisliği Ana Bilim Dalı

    PROF. DR. ENGİN AVCI

  4. Yapay sinir ağları ve parmak izi analizi yöntemi ile kimlik tayini

    Identity recognation via fingerprint analysis by neural networks

    ÖZGE BALTACI

    Yüksek Lisans

    Türkçe

    Türkçe

    2011

    Mekatronik MühendisliğiDokuz Eylül Üniversitesi

    Mekatronik Mühendisliği Ana Bilim Dalı

    DOÇ. DR. ZEKİ KIRAL

  5. Radio map estimation with neural networks and active learning for indoor localization

    Bina içi konumlama için sinir ağları ve aktif öğrenme yöntemleri ile radyo frekans haritası tahminleme

    SILA GÜLER

    Yüksek Lisans

    İngilizce

    İngilizce

    2019

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolBoğaziçi Üniversitesi

    Bilgisayar Mühendisliği Ana Bilim Dalı

    PROF. ALİ TAYLAN CEMGİL