Performance of Hasty and Consistent Multi Spectral Iris Segmentation using Deep Learning

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 7 Issue 5, September-October 2023 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD59853 | Volume – 7 | Issue – 5 | Sep-Oct 2023 Page 11
Performance of Hasty and Consistent Multi
Spectral Iris Segmentation using Deep Learning
Ram Niwas Sharma1, Ankit Kumar Navalakha2, Neha Sharma3
1
M Tech Scholar, Department of Computer Science Engineering, Mewar University, Chittorgarh, Rajasthan, India
2
Assistant Professor, Department of Computer Science Engineering, Mewar University, Chittorgarh, Rajasthan, India
3
Lecturer, Department of Computer Science Engineering, MLVTE, Bhilwara, Rajasthan, India
ABSTRACT
The recognition system is composed of seven phases: acquisition,
preprocessing, segmentation, normalization, feature extraction,
feature selection, and classification. In the acquisition phase, iris
images are captured, followed by preprocessing to enhance the
quality of the images. The segmentation phase involves separating
the iris region from the background, and the normalized iris region is
shaped into a rectangle in the normalization phase. Iris segmentation
is a critical step in iris recognition systems and has a direct impact on
authentication and recognition results. However, standard
segmentation techniques may not perform well in noisy iris databases
captured under challenging conditions. Moreover, the lack of large
iris databases hinders the performance improvement of convolution
neural networks. The proposed method addresses these challenges by
effectively handling irregular iris images captured under visible light.
The iris region is processed and evaluated to generate a unique
feature vector, which is then used for person identification. VGG16, a
well-known deep learning model, is employed for image
classification, and the feature vector is fed into VGG16 for
classification purposes.
KEYWORDS: Deep Learning, Multi Spectral Iris, neural networks,
VGG16
How to cite this paper: Ram Niwas
Sharma | Ankit Kumar Navalakha |Neha
Sharma "Performance of Hasty and
Consistent Multi Spectral Iris
Segmentation using Deep Learning"
Published in
International Journal
of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-7 |
Issue-5, October
2023, pp.11-15, URL:
www.ijtsrd.com/papers/ijtsrd59853.pdf
Copyright © 2023 by author (s) and
International Journal of Trend in
Scientific Research and Development
Journal. This is an
Open Access article
distributed under the
terms of the Creative Commons
Attribution License (CC BY 4.0)
(http://creativecommons.org/licenses/by/4.0)
INTRODUCTION
A biometric system is an automated system that
recognizes individuals based on unique features or
characteristics. Various biometric systems have been
developed using different features such as
fingerprints, facial features, voice, hand geometry,
handwriting, the retina, and the focus of this thesis,
the iris [1]. Biometric systems typically operate in
two modes: an enrollment mode for adding templates
to a database and an identification mode where a
template is created for an individual and matched
against the pre-enrolled templates in the database. .
The iris comprises several layers, including the
epithelium layer with densely pigmented cells and the
stromal layer with blood vessels, pigment cells, and
iris muscles [1]. The color of the iris is determined by
the density of stromal pigmentation.
The epigenetic nature of iris patterns results in each
eye of an individual having independent patterns,
even in the case of identical twins. The iris is an
externally visible and protected organ, maintaining its
stable epigenetic pattern throughout adulthood. The
visual system plays a crucial role in processing
information within the human body [2]. An integral
component of the visual system is the eye, which
consists of three concentric layers. Biometric research
and development require the analysis of human data.
However, it is impractical to perform algorithm
testing on real-time captured data due to various
constraints [3]. Therefore, standardized biometric
databases play a crucial role in the testing and
evaluation of recognition methods.
Biometrics encompasses quantifiable data related to
human characteristics and traits. It is used for
identification and access control in computer science
and surveillance applications. In the initial phase of
biometric processing, various characteristics are
captured. However, for automated capturing and
IJTSRD59853

International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470
comparison with stored data, the biometric
characteristics should satisfy certain criteria.
Multimodal biometric systems utilize more than one
physiological or behavioral characteristic for
enrollment, verification, or identification. These
systems are employed in applications such as border
entry/exit, access control, civil identification, and
network security to reduce false match and false non-
match rates, provide alternative means of enrollment
or identification, and counter fraudulent attempts [2].
The selection of a specific biometric for an
application requires weighing these factors, as no
single biometric can meet all requirements for every
situation.
LITERATURE REVIEW
The concept of iris recognition was first proposed by
Dr Frank Burch, an ophthalmologist, in 1936. In
1987, ophthalmologists Aran Safir and Leonard Flom
patented the idea, and they enlisted John Daugman in
1989 to develop practical algorithms for iris
recognition [12].
Cheng-Shun Hsiao, Chih-Peng Fan, and Yin-
Tsung Hwang (2022) proposed an algorithm for
locating the pupil center using an edge detection
method. They recorded the grey level values on
virtual concentric circles and constructed a zero-
crossing representation based on a one-dimensional
dyadic wavelet transform.
Viktor Varkarakis and Shabab Bazrafkan (2018)
developed an effective deep learning method for iris
biometric authentication. The proposed system
utilizes semantic segmentation technology based on
the UnEAT model to locate and extract the ROI
(Region of Interest) of the iris in an eye image.
Mahmut Karakaya (2018) conducted a study on
biometric authentication using an effective low-
complexity YOLOv3 tiny-based deep learning
inference network.
METHODOLOGY
The presence of noise regions in captured images
increases the need for robust and adaptable
segmentation techniques [11]. Various iris
segmentation proposals have been described,
including statistical approaches, alpha trimming
operation, and checking the gray level transition.
Image segmentation refers to the division of an image
into multiple components. It is a critical step in
automated image processing systems as it forms the
foundation for subsequent operations such as
description or recognition [4].
VGG16 is a convolutional neural network (CNN)
model proposed by K. Simonyan and A. Zisserman
from the University of Oxford in their paper titled
"Very Deep Convolution Networks for Large-Scale
Image Recognition." This model gained popularity
for its remarkable performance in the ImageNet
dataset, achieving a top-5 test accuracy of 92.7%.
Figure 01: VGG16 Architecture
The VGGNet configurations are labeled as A, B, C, D, and E. All configurations share the same overall
architecture and differ primarily in terms of depth. In terms of the width of the convolution layers [5], the
number of channels starts at 64 in the first layer. After each max-pooling layer, the number of channels doubles,
progressively increasing until it reaches 512 in the deeper layers [10].

Software Description
MATLAB 19 (b) is a high-level technical computing language and interactive environment that offers various
capabilities such as algorithm development, data visualization, data analysis, and numerical computation [6]. It
provides strong support for working with matrices and matrix operations, making it a powerful tool for data
analysis and visualization.
One of the key features of MATLAB 19 (b) is its extensive collection of toolboxes, which are sets of programs
designed to support specific tasks. The image processing toolbox is particularly relevant for handling images and
offers functions, commands, and techniques tailored to image analysis [6]. In MATLAB, functions play a crucial
role and are used to perform specific tasks by accepting input parameters and producing output results.
Additionally, users have the flexibility to create their own functions when necessary. MATLAB treats matrices
as its standard data type, considering all data to be matrices in some form. Images, while strings are essentially
matrices of characters, with the string length determining the matrix dimensions.
EXPERIMENTAL RESULTS
The performance of various CNN architectures, including VGG16 and SVM models, was evaluated in this
study. The models were initialized with zero biases and random weights with zero mean [7]. Dropout was
applied to the fully connected layers to mitigate over fitting. By considering these performance parameters, the
classifier's accuracy, precision, sensitivity, and specificity can be evaluated, providing insights into its
classification capabilities.
Table 01: Comparison Result with Existing Work
Database Approach Accuracy (%)
CASIA-INTERVAL
EXISTING SVM 74
PROPOSED VGG16 90.91
Figure 02: Comparison Result with Existing Work
This implies that the VGG16 model demonstrated the highest classification accuracy among the tested
architectures. Additionally, Table 02 presents a comparison of different factors derived from various datasets,
providing further insights into the evaluation and comparison process [7].
Table 02 Comparison Result with Different Dataset
Approach
Accuracy
(%)
Precision
(%)
Recall
(%)
F1
(%)
Specificity
(%)
CASIA-INTERVAL
VGG16
90.91 91.44 91.0 89.49 91.38
UBRIS.V2 84.25 83.99 86.95 86.15 84.39
MMU Database 87.15 83.26 84.12 85.15 86.00

Figure 03: Performance Graph
The evaluation was performed on the CASIA-INTERVAL dataset and compared with the UBRIS.V2 and MMU
databases [9]. The results indicated that CASIA-INTERVAL achieved the highest accuracy and specificity.
Furthermore, an attempt was made to utilize the conventional VGG16 architecture with pre-trained weights
obtained from a model in ImageNet [8].
CONCLUSION
The proposed system utilizes the GLCM and VGG16
for iris recognition. By combining these features with
other texture generation processes, the feature
extraction process becomes more effective. The
recognition of iris is achieved using the kernel
function of the VGG16 classifier. The system
accurately recognizes iris features in video, even in
the presence of challenges such as illumination and
contrast variations. It also detects the liveliness and
race of the person with high accuracy. Extensive
experiments are conducted using the CASIA Iris-,
UBIRIS.v2, and MMU databases to evaluate the
performance of the iris and eye recognition system
using VGG16 architecture. The proposed system
outperforms other existing technologies on each
database. Additionally, an iris segmentation
architecture based on CNN combined with VGG16
and VGG19 is proposed to further improve the
accuracy of the recognition system.
FUTURE WORK
The Biometric Authentication can be improved by
including more number of possible features and other
valid measures. Pupil dilation is found to have an
effect on the accuracy of iris recognition particularly
if the number of dilation is completely different at
enrollment than at verification. Wearing contact
lenses also modify the color and look of the attention
also can decrease the recognition rates.
REFERNCES
[1] Caiyong Wang; Jawad Muhammad; Yunlong
Wang; Zhaofeng; Zhenan Sun, "Towards
Complete and Accurate Iris Segmentation
Using Deep Multi-Task Attention Network for
Non-Cooperative Iris Recognition" IEEE
Transactions on Information Forensics and
Security, 2022.
[2] Cheng-Shun Hsiao; Chih-Peng Fan; Yin-Tsung
Hwang, "Design and Analysis of Deep-
Learning Based Iris Recognition Technologies
by Combination of U-Net and EfficientNet"9th
International Conference on Information and
Education Technology (ICIET), 2022.
[3] Chia-Wei Chuang; Chih-Peng Fan; Robert
Chen-Hao Chang, "Design of Low-Complexity
YOLOv3-Based Deep-Learning Networks with
Joint Iris and Sclera Messages for Biometric
Recognition Application", IEEE 9th Global
Conference on Consumer Electronics (GCCE),
2022.
[4] Shabab; Peter Corcoran, "Enhancing iris
authentication on handheld devices using deep
learning derived segmentation techniques",
IEEE International Conference on Consumer
Electronics (ICCE),2022.
[5] Mousumi Sardar; Subhash Banerjee; Sushmita
Mitra, "Iris Segmentation Using Interactive
Deep Learning", IEEE Access, 2021.

[6] Gürkan Şahin; Orkun Susuz, "Encoder-Decoder
Convolutional Neural Network Based Iris-
Sclera Segmentation", 27th
Signal Processing
and Communications Applications Conference
(SIU), 2021.
[7] Ehsan eddin Jalilian; Mahmut Karakaya;
Andreas, "Iris Recognition Using CNN Based
Iris Segmentation", International Conference of
the Biometrics Special Interest Group
(BIOSIG), 2021.
[8] Ananya Zabin, Thiri machos Bourlai, "A Deep
Learning Based Approach to Iris Sensor
Identification", IEEE/ACM International
Conference on Advances in Social Networks
Analysis and Mining (ASONAM), 2021.
[9] Luiz A. Zanlorensi, Eduardo Luz, Rayson
Laroca, Alceu S. Britto, Luiz S. Oliveira, David
Menotti, "The Impact of Preprocessing on Deep
Representations for Iris Recognition on
Unconstrained Environments"31st SIBGRAPI
Conference on Graphics, Patterns and Images
(SIBGRAPI), 2018.
[10] Lozej; Dejan, Vitomir, Peter Peer, "Influence of
segmentation on deep iris recognition
performance"7th International Workshop on
Biometrics and Forensics (IWBF), 2020.
[11] Caiyong Wang; Yunlong Wang, Boqiang Xu;
Yong He, Zhiwei Dong; Zhenan Sun, "A
Lightweight Multi-Label Segmentation
Network for Mobile Iris Biometrics" IEEE
International Conference on Acoustics, Speech
and Signal Processing (ICASSP), 2020.
[12] Daniel Kerrigan, Mateusz Trokie lewicz,
Adam, Kevin W. Bowyer, "Iris Recognition
with Image Segmentation Employing Retrained
Off-the-Shelf Deep Neural Networks",
International Conference on Biometrics (ICB),
2020.

Performance of Hasty and Consistent Multi Spectral Iris Segmentation using Deep Learning

Recommended

More Related Content

Similar to Performance of Hasty and Consistent Multi Spectral Iris Segmentation using Deep Learning (20)

More from ijtsrd (20)

Recently uploaded (20)

Performance of Hasty and Consistent Multi Spectral Iris Segmentation using Deep Learning