Human identification using finger images

Pao Yue-kong Library Electronic Theses Database

Human identification using finger images

 

Author: Zhou, Yingbo
Title: Human identification using finger images
Degree: M.Sc.
Year: 2010
Subject: Hong Kong Polytechnic University -- Dissertations
Biometric identification
Fingerprints -- Identification.
Fingerprints -- Data processing.
Department: Dept. of Computing
Pages: ix, 45 leaves : ill. ; 31 cm.
InnoPac Record: http://library.polyu.edu.hk/record=b2352644
URI: http://theses.lib.polyu.edu.hk/handle/200/5618
Abstract: Biometrics is a technique using human characteristics for personal identification, and this technique has been widely used for both civil and forensic applications in recent years. Methods of various identifications such as fingerprint, iris, palm print, face and so forth have been well studied to enhance the performance, effectiveness and reliability of the biometrics systems. In this thesis, we focused on civil applications, and proposed a novel approach for human identification using finger images. The approach exploits both the finger vein and frontal finger surface images, hence providing a multi-modal biometrics system. As being inside our fingers and not fully visible under normal lightening conditions, vein patterns also reduce the chance of spoofing (claimed to be another subject). Essentially, the device designed in this study relies on the hemoglobin in the blood, which absorbs the light of wavelength between 700nm to 1000nm, to detect finger vein. Therefore, a 'liveness test' is also conducted simultaneously while collecting the vein image. In doing so, the possibility of been spoofed by an artificial finger is largely reduced. Moreover, peg-free design is applied in this work to offer more user friendliness. Techniques for image segmentation, enhancement and feature extraction are investigated. In particular, we exploit various methods for finger vein and surface texture feature extraction. Noticeably, a new approach applying top-hat transformation to the feature image that was extracted by using a group of pre-defined Gabor filters (Morphological Gabor) offers promising result. The set of Gabor filters (Matched Gabor) is able to enhance and extract the underlying vein patterns as well as suppress the noise in the image. In addition, the morphological operation is used to further highlight the vein structures, especially small ones that are extracted but not clearly shown by the Gabor. We established a database of self-collected finger images with 95 subjects, and from each subject 48 images on two fingers were collected (12 images per finger per session). The data was collected twice in two sessions with an interval of one to five months. The various feature extraction methods are evaluated on this data base. The experimental resulted in best equal error rate of 2.99% for vein and 4.47% for texture and 1.71% for the final fusion score. The promising results suggested substantial applicability of this approach for human identification. Future work, such as the minutiae based approach for vein structure and texture representation and matching as well as to evaluate the method on large databases (e.g. more than 500 subjects), may help to further improve the performance, and test the reliability and stability of this method.

Files in this item

Files Size Format
b23526440.pdf 3.254Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information