Summary:
This
article introduces a strong and effective technique for identifying finger
veins using the gray level co-occurrence matrix based on the discrete wavelet
transform. By combining the discrete wavelet transform with the local binary
pattern and gray level co-occurrence matrix, we present a fresh approach to
finger vein recognition. Simulation results demonstrate the efficiency and
speed of this technique in extracting features and performing classification.
Introduction:
Biometrics
involves recognizing individuals based on their physiological, behavioral, and
biological characteristics. It can be divided into two categories:
physiological biometrics and behavioral biometrics. Physiological biometrics
identify individuals based on attributes like face, iris, fingerprint, finger
vein, hand geometry, etc., while behavioral biometrics rely on human behaviors
such as handwriting, signature, or voice recognition. Figure 1 illustrates the
process of enrolling and authenticating individuals in a biometric system.
Recognition
of Finger Veins:
Researchers
have explored the utilization of underlying skin features due to the
limitations of fingerprint technology. Finger veins, which rely on the blood
vessels beneath the skin, provide a distinct biometric system that offers
advantages such as uniqueness among individuals, even among twins. While other
vascular properties such as the retina, face, and hands can be used for
identification, finger vein recognition devices are popular due to their
familiarity and ease of use. The absorption of infrared light by hemoglobin
plays a critical role in capturing vein patterns. Vein patterns are captured
after the infrared light is absorbed. The distance between the skin and the
blood vessels affects the absorption of infrared light, with greater distance
resulting in more noise in the captured image. Although palms, the back of
hands, and fingers can be utilized for biometric data, fingers are the preferred
choice for most people.
Devices
for Capturing Finger Vein Images:
Infrared
(IR) light is employed in finger-vein biometric systems to capture blood
vessels. The position of the infrared light source significantly impacts the
quality of the captured images. Additionally, the image acquisition device
should be compact, cost-effective, and capable of providing high-resolution
images. In captured images, veins are represented as gray patterns. Figure 3
illustrates the arrangement of fingers between the Infrared Light Emitting
Diodes (IR-LEDs) and the imaging device.
Advantages
and Disadvantages:
1.
Internal nature: Vein patterns are situated inside the skin, making them
invisible to the naked eye. The identification process is not hindered by
damaged skin, and the accuracy of the system is unaffected by dry, wet, or
dirty hands.
2.
Protection against duplication: Vein patterns are challenging to replicate as
blood flow is necessary during image capture. Studies have demonstrated that it
is impossible to cut a finger and register it in the system due to blood
seepage.
3.
Hygienic readers: Finger-vein readers are considered hygienic as users do not
directly touch the sensor, unlike fingerprint and hand geometry systems.
4.
User-friendly: Finger-vein recognition systems are easy to operate and
user-friendly.
5. No
cultural resistance: Finger-vein recognition is not tied to specific cultural
practices.
6.
Uniqueness: Finger veins remain unique even among twins and do not change with
age.
Finger
Vein Recognition in Biometric Technology:
Finger
vein recognition has gained prominence as a method in biometric technology in
recent years. Several successful methods, including Line Tracking (LT), Maximum
Curvature (MC), and Wide Line Detector (WL), have been proposed for finger vein
recognition. However, LT's feature extraction and matching processes are slow,
and all three methods are susceptible to rotation and noise. To overcome these
limitations, we propose the application of popular feature descriptors widely
used in Computer Vision or Pattern Recognition (CVPR).
These
descriptors encompass Fourier Descriptors (FD), Zernike Moments (ZM) [8],
Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), and Global
Binary Patterns (GBP). Notably, FD, ZM, HOG, LBP, and GBP have not been used in
finger vein recognition before. In this study, we compare these descriptors
against LT, MC, and WL. The novelty of this research lies in (i) the application
of new feature extraction methods that have not been previously used in finger
vein recognition and (ii) evaluating the performance of these methods under
translation, rotation, and noise. We focus on the "feature
extraction" step, while keeping the preprocessing step as simple as
possible. For matching, we employ the mismatch ratio specific to LT, MC, and
WL, while other descriptors are compared using three different distance
metrics: Euclidean distance, X2 (Chi-Square distance), and Earth Mover's Distance
(EMD).
Database:
The
finger vein database used in this study is sourced from the publicly available
SDUMLA-HMT finger-vein database [4]. This database comprises 3,816 images from
both hands, including index finger, middle finger, and ring finger images. Six
different images are captured for each finger. Figure 4 displays a small sample
from the database.
The
original images have dimensions of 320x240, but for faster analysis, they were
reduced to 160x120 using nearest neighbor interpolation. The images are
grayscale, with intensity ranging from 0 to 255. We utilize the Prewitt edge
detector to extract strong edges, identify finger boundaries, and generate a
mask image. The masking process is crucial for eliminating irrelevant areas.
The
MMCBNU_6000 vein finger database is evaluated based on average image gray
value, image contrast, and entropy. Another vein finger database, HKPU-FV,
created by Ajay and Zhou, and the UTFV FV database from the University of
Twente have been utilized. Recently, Chonbuk National University and Tsinghua
University produced two additional finger vein databases. An open finger vein
database, SDUMLAHMT, contains images of index finger, middle finger, and ring
finger from both hands, with six images captured for each finger. The largest
reported finger vein database, PKU Finger Vein Database, was established based
on the attendance checking system at Peking University.
Conclusion:
This
paper presents a comprehensive survey on human identification using finger vein
recognition. Various methods and databases have been explored, and the
literature provides diverse feature description methods for analysis. In future
work, it is worthwhile to investigate extensions of LBP, HOG, and GBP that
offer invariances such as rotation and scale invariance. Additionally,
realistic scenarios involving finger rotation, translation, and varying
camera-to-finger distances can be studied to analyze the resilience of these
methods to such factors.
Hashtag/Keyword/Labels:
finger vein recognition,
biometrics, physiological biometrics, behavioral biometrics, feature
extraction, database
References/Resources:
1. Miura, N., Nagasaka, A.,
Miyatake, T. (2004). Feature extraction of finger-vein patterns based on
repeated line tracking and its application to personal identification. Machine
Vision and Applications, 15, 194-203.
2. Vallabh, H. (2012).
Authentication using finger-vein recognition. University of Johannesburg.
3. Haralick, R.M., Shanmugam, K.,
Dinstein, I.H. (1973). Textural features for image classification. IEEE
Transactions on Systems, Man and Cybernetics, 3, 610-621.
4. SDUMLA-HMT Finger-Vein
Database. Available at: https://mla.sdu.edu.cn/sdumla-hmt.html
For more such Seminar articles click index
– Computer Science Seminar Articles list-2023.
[All images are taken from Google
Search or respective reference sites.]
…till
next post, bye-bye and take care.
No comments:
Post a Comment