Crypt Edge Detection Using PSO,Label Matrix And BI-Cubic Interpolation For Better Iris Recognition(PSOLB)

Hashim, Nurul Akmal (2017) Crypt Edge Detection Using PSO,Label Matrix And BI-Cubic Interpolation For Better Iris Recognition(PSOLB). Masters thesis, UTeM.

[img] Text (24 Pages)
Crypt Edge Detection Using PSO, Label Matrix And BI-Cubic Interpolation For Better Iris Recognition(PSOLB).pdf - Submitted Version

Download (542kB)


Iris identification is an automatic system to recognise an individual in biometric applications.Human iris is an internal organ that can be accessed from external view of the body.Moreover,the structure of the iris is formed in a complete random manner and has unique features such as crypts,furrows,collarets,pupil,freckles, and blotches.In fact, no iris patterns are the same.The iris structure is stable which it means the location of the iris features is permanent at certain point.Nevertheless,the shape of iris features changes slowly due to several factors which include aging,surgery,growth,emotion and dietary habits. Recently,there has been renewed interest in iris features detection.Gabor filter,cross entrophy, upport vector,and canny edge detection are methods which produce iris codes in binary codes representation.However,problems have occurred in iris recognition since low quality iris images are created due to blurriness,indoor or outdoor settings, and camera specifications.Failure was detected in 21% of the intra-class comparisons cases which were taken between intervals of three and six months intervals.However,the mismatch or False Rejection Rate (FRR) in iris recognition is still alarmingly high.Higher FRR also causes the value of Equal Error Rate (EER) to be high.The main reason for high values of FRR and EER is that there are changes in the iris due to the amount of light entering into the iris that changes the size of the unique features in the iris.One of the solutions to this problem is by finding any technique or algorithm to automatically detect the unique features.Therefore a new model is introduced which is called Crypt Edge Detection which combines PSO,Label Matrix,and Bi-Cubic Interpolation for Iris Recognition (PSOLB) to solve the problem of detection in iris features.In this research, the unique feature known as crypts has been chosen due to its accessibility and sustainability.Feature detection is performed using particle swarm optimisation (PSO) as an algorithm to select the best iris texture among the unique iris features by finding the pixel values according to the range of selected features.Meanwhile, label matrix will detect the edge of the crypt and the bi-cubic interpolation technique creates sharp and refined crypt images.In order to evaluate the proposed approach,FAR and FRR are measured using Chinese Academy of Sciences' Institute of Automation (CASIA) database for high quality images.For CASIA version 3 image databases, the crypt feature shows that the result of FRR is 21.83% and FAR is 78.17%.The finding from the experiment indicates that by using the PSOLB,the intersection between FAR and FRR produces the Equal Error Rate (EER) with 0.28%,which indicated that equal error rate is lower than previous value, which is 0.38%.Thus,there are advantages from using PSOLB as it has the ability to adapt with unique iris features and use information in iris template features to determine the user.The outcome of this new approach is to reduce the EER rates since lower EER rates can produce accurate detection of unique features.In conclusion,the contribution of PSOLB brings an innovation to the extraction process in the biometric technology and is beneficial to the communities.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Pattern recognition systems,Biometric identification, Optical pattern recognition.
Subjects: T Technology > T Technology (General)
T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: Faculty of Information and Communication Technology
Depositing User: Mohd. Nazir Taib
Date Deposited: 27 Aug 2018 02:51
Last Modified: 27 Aug 2018 02:51
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item