Creases of middle phalanx of finger for person identification system
Loading...
Date
2019-06
Authors
Su Lee Ting
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Today, there are many types of biometric traits to be used in identification system, to give permission to the user to access the system, such as iris, face, and fingerprint. Biometric traits became popular in security technology due to its high reliability and uniqueness which is not easy to be imitated by others. However, two major problems of using biometric traits arises as high cost and permanence. For example, iris scanners are relatively having higher price compared to other biometric. The research carried out needs high cost in hiring experts and time-consuming. While for face, the problem occurs when the user is having expression or using the facial organs to do some tasks, such as: eating or chewing food, closing eyes or sneezing. Hence, the implementation cost used for topquality camera and advanced software to ensure accuracy is quite high. For fingerprint, it was well-known due to its simplicity and reliability. However, one common problem of fingerprint identification system is the permanence of its pattern. There are many people fade in fingerprint pattern due to loss of collagen. To overcome these problems, a new biometric trait with low in cost and high permanence was introduced and its feasibility had been investigated in this study – creases of middle phalanx of finger. The overflow of this project can be separated into four main processes: image acquisition, preprocessing, feature extraction and classification. The software part was mostly done with MATLAB, except the prototype development of image acquisition process involved SketchUp. In image acquisition, a prototype was designed and made to assist the image acquisition by phone camera. It was made to provide brightness, fix the position of finger and camera. 140 images of middle phalanx pattern were taken from 14 volunteers, where 10 images from each of them. In next stage – pre-processing, the 140 images have undergone a series of image pre-processing techniques to improve the image quality and ready for next stage, which include resizing, sharpening, grey scale conversion, image intensity adjustment and manual ROI cropping. In third stage, feature extraction, Local Binary Pattern (LBP) algorithm was used to enhance the image’s contrast and LBP code was generated to be fed into final stage. Classification is the final stage where Support Vector Machine (SVM) was used as classifier. There are 2 types of SVM used in this project: one-class SVM and multiclass SVM, to investigate their performance in classification. At the end of this study, the results of two types SVM classification are both higher than 95%. It proved that the creases of middle phalanx of finger are feasible to be used as a biometric trait in person identification system.