VISVESVARAYA TECHNOLOGICAL UNIVERSITY

“JNANA SANGAMA”, BELAGAVI - 590 018

A PROJECT REPORT
on

“FINGERPRINT MATCHING”
Submitted by

Nitesh kamat 4SF20CI042

Pavan 4SF20CI043
In partial fulfillment of the requirements for VI Sem. B. E. (CSE-AI&ML)

DIGITAL IMAGE PROCESSING LABORATORY WITH MINI

PROJECT

Under the Guidance of

Ms.Vaishnavi Rao
Assistant Professor, Department of CSE(AI&ML)
at

SAHYADRI
COLLEGE OF ENGINEERING & MANAGEMENT
An Autonomous Institution
Adyar, Mangaluru - 575 007
2022 - 23
 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING
(ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING)

CERTIFICATE
This is to certify that the Digital Image Processing Mini project work entitled “Finger-
print Matching ” has been carried out by Nitesh kamat(4SF20CI042), and Pavan
(4SF20CI043), the bonafide students of Sahyadri College of Engineering & Manage-
ment in partial fulfillment for the award of Bachelor of Engineering in Computer Science
& Engineering (Artificial Intelligence & Machine Learning) of Visvesvaraya Technological
University, Belagavi during the year 2022-23. It is certified that all corrections/suggestions
indicated for Internal Assessment have been incorporated in the Report deposited in the
departmental library. The Digital Image Processing project report has been approved as
it satisfies the academic requirements in respect of project work prescribed for the said
degree.

———————————– ————————————– ———————————–

Guide Project Coordinator HOD
Ms.Vaishnavi Rao Mr.Gurusiddaya Hiremath Dr. Pushpalatha K

External Viva:
Examiner’s Name Signature with Date

1. . . . . . . . . . . . . . . . . . . . . . .....................

2. . . . . . . . . . . . . . . . . . . . . . .....................
 SAHYADRI
COLLEGE OF ENGINEERING & MANAGEMENT
An Autonomous Institution

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

(ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING)

DECLARATION

We hereby declare that the entire work embodied in this Digital Image Processing Project
Report titled “Fingerprint Matching” has been carried out by us at Sahyadri College of
Engineering and Management, Mangaluru under the supervision of Ms.Vaishnavi Rao
for the award of Bachelor of Engineering in Computer Science & Engineering.
This report has not been submitted to this or any other University for the award of any
other degree.

Nitesh kamat(4SF20CI042)

Pavan (4SF20CI043)

Dept. of CSE (AI & ML), SCEM, Mangaluru

Abstract

Fingerprint matching has gained significant attention as a reliable biometric authentication

method due to its uniqueness and stability. In this mini-project, we propose a fingerprint

matching system that leverages a Siamese network to achieve accurate and efficient finger-

print identification. The system consists of three main stages: image preprocessing, feature

extraction, and matching. During the preprocessing stage, the captured fingerprint image

undergoes various enhancement and noise reduction operations to ensure optimal quality

for subsequent analysis. In the feature extraction stage, the Siamese network is trained to

encode and extract discriminative features from pairs of fingerprint images. These features

capture the unique patterns and characteristics of individual fingerprints. In the matching

stage, the trained Siamese network computes a similarity score between two fingerprint

images, indicating their degree of similarity. Experimental evaluations demonstrate the

effectiveness of our approach in achieving high accuracy and robustness in fingerprint

matching. This system offers a promising solution for biometric authentication, providing

enhanced accuracy and efficiency in fingerprint recognition tasks.

i
Acknowledgement

It is with great satisfaction and euphoria that we are submitting the Project Report on
“Fingerprint Matching ”. We have completed it as a part of the curriculum of Visves-
varaya Technological University, Belagavi for the award of Bachelor of Engineering in
Computer Science & Engineering.

We are profoundly indebted to our guide, Ms.Vaishnavi Rao, Assistant Professor,Department

of Computer Science & Engineering(Artificial Intelligence & Machine Learning) for innu-
merable acts of timely advice, encouragement and We sincerely express our gratitude.

We also thank Mr. Gurusiddaya Hiremath, Project Coordinator, Department of Com-

puter Science & Engineering(Artificial Intelligence & Machine Learning) for their constant
encouragement and support extended throughout.

We express our sincere gratitude to Dr. Pushpalatha K, Head & Professor,Department

of Computer Science & Engineering(Artificial Intelligence & Machine Learning)for his in-
valuable support and guidance.

We sincerely thank Dr. Rajesha S, Principal, Sahyadri College of Engineering & Man-
agement, who have always been a great source of inspiration.

We extend our sincere regards and respect to Dr. Manjunath Bhandary, Chairman,
SCEM, having provided all the facilitites that helped us in the timely completion of this
project report.

Finally, yet importantly, We express our heartfelt thanks to our family & friends for their
wishes and encouragement throughout the work.

Nitesh kamat(4SF20CI042)
Pavan (4SF20CI043)

ii
Table of Contents

Abstract i

Acknowledgement ii

Table of Contents iv

List of Figures 1

1 Introduction 2
1.1 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Literature Survey 4

3 Problem Definition 8

4 Methodology 9
4.1 Fingerprint Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 Standardization of images: . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Architecture for Fingerprint Matching . . . . . . . . . . . . . . . . . . . . 12
4.4 Training Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

5 Datasets 15
5.1 Sokoto Coventry Fingerprint Dataset (SOCOFing) . . . . . . . . . . . . . 15

6 Results and Analysis 16

6.1 Training Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.1.1 Compatibility Test . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.1.2 Accuracy Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
6.2 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.2.1 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 17

iii
 6.3 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.5 Potential Areas for Further Improvement . . . . . . . . . . . . . . . . . . . 18
6.6 Future Research Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

7 Conclusion and Future work 20

References 21

iv
List of Figures

4.1 Flowchart of the Fingerprint Matching System . . . . . . . . . . . . . . . . 9

4.2 Workflow of Fingerprint Matching . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Left: original image. Right: equalized image . . . . . . . . . . . . . . . . . 11
4.4 Left: filtered image Right: enhanced Fingerprint image . . . . . . . . . . . 12
4.5 Siamese Network Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 13

5.1 Some Sample images from the dataset . . . . . . . . . . . . . . . . . . . . 15

6.1 Fingerprint Matching Experiment . . . . . . . . . . . . . . . . . . . . . . . 17

1
Chapter 1

Introduction

Fingerprint recognition is a widely used biometric technology that has gained significant
attention and adoption in various fields. With the unique and distinct patterns present
on each individual’s fingertips, fingerprint recognition has proven to be a reliable and
efficient method for personal identification and authentication. This technology utilizes
advanced algorithms and pattern matching techniques to capture, process, and analyze
the distinctive features of fingerprints, enabling accurate identification and verification.
The application of fingerprint recognition has seen tremendous growth in recent years,
encompassing a diverse range of sectors such as law enforcement, access control systems,
banking and finance, border security, and mobile device authentication. Its popularity can
be attributed to its non-invasive nature, high accuracy, and the fact that fingerprints are
difficult to forge, making it an ideal choice for ensuring security and identity verification.
In this report, we will explore the fundamental principles behind fingerprint recognition,
the various components involved in the process, and the challenges faced in implementing
this technology. Additionally, we will delve into the potential applications and benefits of
fingerprint recognition systems, as well as discussing some emerging trends and advance-
ments in the field. By understanding the intricacies of fingerprint recognition, we can gain
insights into its effectiveness, limitations, and its impact on enhancing security measures
in our increasingly digital and interconnected world.

1.1 Purpose
The purpose of fingerprint recognition is to accurately and reliably identify individuals
based on the unique patterns and characteristics present in their fingerprints. This bio-
metric authentication method serves several key objectives. Firstly, it aims to provide

2
Fingerprint Matching Chapter 1

identification by comparing a person’s fingerprint with a database of known fingerprints,

enabling the determination of their identity based on the distinctiveness of their fingerprint
patterns. Secondly, fingerprint recognition is used for authentication purposes, verifying
the claimed identity of an individual by comparing a presented fingerprint with a stored
reference template.

1.2 Scope
The scope of fingerprint recognition is broad, encompassing a wide range of domains and
applications. It is a versatile technology that finds application in various fields, includ-
ing security systems, forensic investigations, identity verification, financial transactions,
employee management, mobile devices, healthcare, and more.
In the realm of security systems, fingerprint recognition is extensively used for access
control, ensuring that only authorized individuals can gain entry to secure areas or elec-
tronic devices. It provides a high level of security and convenience compared to traditional
methods like passwords or access cards.In forensic investigations, fingerprint recognition
plays a crucial role in linking suspects to crime scenes. It aids in the identification and
matching of fingerprints, helping law enforcement agencies solve crimes and provide evi-
dence in court.

1.3 Overview
Fingerprint recognition is a widely used biometric technology that aims to identify and
authenticate individuals based on their unique fingerprint patterns. It has gained signifi-
cant attention and adoption across various industries and applications due to its reliability,
security, and convenience.Fingerprint recognition has a broad scope and finds application
in numerous domains. It is extensively used in access control systems to secure physical
spaces, electronic devices, and confidential information. By relying on the uniqueness of
fingerprints, it provides a highly secure method for granting authorized access.In forensic
investigations, fingerprint recognition plays a crucial role in criminal identification. It en-
ables the comparison of fingerprints found at crime scenes with those in a database, aiding
law enforcement agencies in linking suspects to criminal activities and providing valuable
evidence in court.

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 3
Chapter 2

Literature Survey

The tremendous success of fingerprint based recognition technology in law enforcement

applications, decreasing cost of fingerprint sensing devices, increasing availability of in-
expensive computing power, and growing identity fraud/theft have all ushered in an era
of fingerprint based person recognition applications in commercial, civilian, and financial
domains. In 1893 the Home Ministry Office, UK accepted that no two individuals have the
same fingerprints [1]. Soon after this discovery many major law enforcement departments
embraced the idea of first booking the fingerprints of criminals, so that there records are
readily available and later using leftover fingerprint smudges, they could determine the
identity of criminals. These agencies sponsored a rigorous study of fingerprints, devel-
oped scientific method for visual matching of fingerprints and applied the art of fingerprint
recognition for nailing down the perpetrators.
In the early stages of Automatic Fingerprint Identification System (AFIS) the algo-
rithms more or less tried to follow the manual approach of fingerprint identification. Ex-
pensive hardware was required for the computational requirements and to match the re-
sponse time requirements. T. R. Gowrishankar [2] presented a feature-based scheme for
fingerprint identification on a massively parallel system (architectures with 1000 or more
processors). The author presented an algorithm for orientation independent identification.
This automatic fingerprint identification system considers five features namely isolated
points, terminating points, merge/branch points, segments and loops as primary elements
of a feature vector that uniquely identify a given fingerprint pattern. These minutiae and
their relative locations are found sufficient to identify a given pattern.
M Mori [3] describes the structure, function, and operation of this Automatic Switching
System for high resolution Photo-telegraphic Equipment. C. A. Gunawardena and V. K.
Sagar [4] described a high speed and low cost system for scanning, identifying and coding

4
Fingerprint Matching Chapter 2

fingerprints. In their approach the raw fingerprint image is cleaned by median filter, which
makes edge extraction from the fingerprint easier. After edge extraction line thinning is
performed. The image is processed by a new single pass noise removal and linethinning
algorithm, which is based on contour generation. This algorithm was faster and more
efficient than conventional contour tracing and parallel algorithms and was the bases for the
high-speed feature extraction. For better extraction of features fingerprint enhancement is
required. B. G. Sherlock; D. M. Monro and K. Millard [5] described a method of enhancing
fingerprint images, based upon directional Fourier domain filtering. Fingerprints are first
smoothed using a directional filter whose orientation is everywhere matched to the local
ridge orientation. The proposed technique took the advantage of Fourier domain filtering
which helped to convolve the fingerprint images with filters of full image size. A criminal
case may require a very high degree match but an access control system may use a more
relaxed matching system, simply because its database is much smaller. Sometimes it is
not possible to capture the whole of fingerprint image easily
In such circumstances, how to tolerate minor distortions becomes a critical subject.
M. B. Akhan, I. Emiroglu and E. G. Bahari [6] proposed a method to overcome some of
these difficulties by allowing the user to set the accuracy and the degree of match. Most
of the approaches to fingerprint recognition used to break down the process of ridge detec-
tion into smoothing or early pre-processing, edge detection, thresholding, binarization and
subsequently thinning. This whole procedure was computationally expensive and hence
required more expensive hardware to meet the response- time requirements. V. K. Sagar;
D. B. L. Ngo and K. C. K. Foo [7]presented an approach based on fuzzy logic technique.
This technique had the advantage of being simple and less expensive. Major advantage of
using fuzzy logic is that it requires less expensive hardware for similar performance. M. H.
Ghassemian [8] addressed the problem of the extraction of relevant macroscopic properties
of the image, in particular the location of key features (core and delta). Author explored
the transformation of the image into a reconstructed skeletal form, followed by the conver-
sion of the image into a series of tracks. These tracks latter permits the straightforward
extraction of’ essential pattern characteristics. The track curvature can also be used to
identify locations in the image that are in some sense pivotal to the ridge flow and can be
used as substitutes for the more obvious cores. This is of significance in Arch-like patterns,
which have no discernible core.
Smudged furrows and cut ridges in the image of a fingerprint are the major problem in
the fingerprint identification or verification systems. M. H. Ghassemian,[9] presented a new

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 5
Fingerprint Matching Chapter 2

on-line unsupervised ridge detection method that reduced the complexity and costs associ-
ated with the fingerprint identification procedure. The author describes the algorithm for
restoration of the ridges of fingerprint. This algorithm was based on fuzzy classification
technique. D. P. Mital and Eam Khwang Teoh,[10] presented the structural matching ap-
proach which is able to perform effective rotational invariant fingerprint identification. In
this approach, each of the extracted feature is correlated with five of its nearest neighboring
features to form a local feature group for a first-stage matching. After that, the feature with
the highest match is used as a central feature whereby all the other features are correlated
to form a global feature group for a second-stage matching. Distance and relative angle is
used for finding the correlation between the features. Lin Hong and Jain A.[11] introduced
a new fingerprint enhancement algorithm, which decomposes the input fingerprint image
into a set of filtered images. From the filtered images, the orientation field is estimated and
a quality mask, which distinguishes the recoverable and unrecoverable corrupted regions
in the input image, is generated. The input fingerprint image is adaptively enhanced in
the recoverable regions. Hong Lin; A. Jian and R. Bolle [12] described the design and
implementation of an on-line fingerprint verification system, which operates in two stages:
minutia extraction and minutia matching. An improved version of the minutia extraction
algorithm proposed by Ratha et al., which was much faster and more reliable, was imple-
mented for extracting features. For minutia matching, an alignment-based elastic matching
algorithm was developed. This algorithm has the capability of finding the correspondences
between minutiae in the input image and the stored template without resorting to exhaus-
tive search and has the ability of adaptively compensating for the nonlinear deformations.
Lin Hong; Yifei Wan and Jain A [13] presented a fast fingerprint enhancement algorithm,
which can adaptively improve the clarity of ridge and valley structures of input fingerprint
images based on the estimated local ridge orientation and frequency.
There were two major shortcomings of the traditional approaches to fingerprint rep-
resentation. Firstly for a significant fraction of population, the representations based on
explicit detection of complete ridge structure in the fingerprints were difficult to extract
automatically. . A. K. Jain; S. Prabhakar; L. Hong and S. Pankanti[14] proposed filter-
based algorithm that uses a bank of even symmetric Gabor filters to capture both the local
and the global details in a fingerprint as a compact 640- byte fixed length Finger Code.
The fingerprint matching is done on the bases of Euclidean distance between the two cor-
responding Finger codes and hence is extremely fast. The accuracy of the method was
found comparable to best results of minutiae based algorithms. Lee Chih-Jen and Wang

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 6
Fingerprint Matching Chapter 2

Sheng-De [15] explained usefulness of Gabor filter based method for fingerprint recognition
on small database. The method made use of Gabor filtering technology and eliminated
computationally complex steps like smoothing, binarization, thinning and minutiae de-
tection. Gabor filter based features played a central role in the processes of fingerprint
recognition, including local ridge orientation detection, core point detection, and feature
extraction.
In the improvement to their previous method,A. K. Jain; S. Prabhakar; L Hong and
S. Pankanti proposed filter-based algorithm that uses a bank of Gabor filters to capture
both local and global details in a fingerprint as a compact fixed length Finger Code. The
verification accuracy, which was only marginally inferior to the best results of minutiae-
based algorithms, was achieved. M. Horton; P. Meene; R. Adhami and P. Cox,[16] in
their work investigated the utility of complex, 2-D Gabor filter for a fingerprint matching
system. Previous approaches had used an even symmetric 2-D Gabor filter to perform
fingerprint feature extraction processing. Several processing methods using a complex
filter were explored and compared to the even-symmetric case. The use of the complex
filter provided only marginal improvement over the real filter though the computational
cost for using the complex filter was significant. L. Sha; F. Zhao and X. Tang,[17] proposed
a new rotation invariant reference point location method, and then combine the direction
features with the AAD features to form an oriented Finger Code. To approximate the
rotation invariance features, the image was rotated cyclically to generate ten templates.
Experimental results showed that the addition of direction features led to a substantial
improvement in the overall matching performance. A. Jain; A. Ross and S. Prabhakar,
[18] presented a hybrid matching algorithm that used both minutiae (point) information
and texture (region) information for matching the fingerprints. A bank of Gabor filters was
used to extract features from the tessellated cells of the template and input images. Sen
Wang and Yangsheng Wang,[19]presented a new method and designed a new frequency
domain filter to enhance fingerprint in the singular point area. First the singular point
was detected and then filter was designed for the enhancement in this area.
Classification of fingerprint is to group similar type of fingerprints together. This helps
reducing the matching time. F. You, Y.Q. Shi and P Engler[20] described a method to
automatically classify fingerprints into three groups: whorl, loop and arch. Their work
was motivated with the idea to help medical scientists to study the relationship between
fingerprint patterns and medical disorders. They used frequency domain approach in which
features of Fourier spectrum were used.

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 7
Chapter 3

Problem Definition

The problem of fingerprint matching refers to the task of accurately identifying and verify-
ing the identity of an individual based on their fingerprint patterns. Fingerprint matching
plays a crucial role in various applications, including law enforcement, access control sys-
tems, and personal authentication. The main challenge in fingerprint matching lies in the
uniqueness and complexity of fingerprint patterns. Each person’s fingerprint consists of
ridges, valleys, and minutiae points, such as ridge endings and bifurcations, which form a
distinct pattern. The goal is to compare a captured fingerprint image with a database of
known fingerprints to determine if a match exists.
The problem involves several subtasks, including image preprocessing, feature extrac-
tion, and matching algorithms. Image preprocessing aims to enhance the quality of the
captured fingerprint image by reducing noise, improving contrast, and removing artifacts.
Feature extraction involves identifying and extracting key characteristics, such as minutiae
points, from the preprocessed image. Matching algorithms then compare the extracted
features of the query fingerprint with those in the database to determine the degree of
similarity or dissimilarity.

8
Chapter 4

Methodology

4.1 Fingerprint Image Processing

The conventional fingerprint recognition system relied on the comparison of minutiae. In
this study, however, the step of minutia point extraction was omitted, and two binary
fingerprint images were input into the Siamese neural network to obtain the similarity
between two fingerprints and get the fingerprint recognition result. The whole system
needed to complete such steps as the standardization of fingerprint image, image enhance-
ment,binarization, input into the Siamese network, and output of the matching result. It
can be divided into two parts: “fingerprint image preprocessing” and “fingerprint image
matching. based on Siamese network,” as shown in flowchart. The method proposed

Figure 4.1: Flowchart of the Fingerprint Matching System

in this study can compare fingerprint images in a direct way, so it was more flexible in
practical applications. The application method is shown in Figure 4.2, that is, to store the

9
Fingerprint Matching Chapter 4

enhanced image into the database and compare it with the input fingerprint, in order to
control the subsequent programs.

Figure 4.2: Workflow of Fingerprint Matching

4.2 Standardization of images:

Fingerprint images were standardized to eliminate the problems of inconsistent clarity,
grayscale, and the number of channels between different fingerprint images. The steps were
as follows: Fingerprint images were standardized to eliminate the problems of inconsistent
clarity, grayscale, and the number of channels between different fingerprint images. The
steps were as follows:
1. Gray Scaling: Fingerprint images were converted from color to grayscale. This
conversion simplifies the image representation and reduces computational complexity, as it

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 10
Fingerprint Matching Chapter 4

focuses solely on the intensity values of pixels, which contain important fingerprint ridge
information.
2. Histogram Equalization: Histogram equalization was performed to enhance the
contrast and normalize the grayscale values of the fingerprint images. This step addressed
the issue of inconsistent clarity and variations in grayscale due to different imaging condi-
tions or finger pressure during image acquisition.
To perform histogram equalization, the following steps were followed:

• Compute the histogram of the grayscale image, which represents the distribution of
pixel intensities.

• Calculate the cumulative distribution function (CDF) of the histogram.

• Normalize the CDF to the range of pixel intensities (usually 0 to 255) to obtain the
mapping function.

• Apply the mapping function to each pixel in the grayscale image to obtain the equal-
ized image.

3.Low-pass filtering smoothing: The image noise generated during fingerprint col-
lection was removed and the fingerprint images were smoothed. Fast Fourier transform
(FFT) was applied to the images. After the high-frequency part was eliminated, low-pass
filtered images were acquired through inverse Fourier transform.

Figure 4.3: Left: original image. Right: equalized image

4. Image Enhancement Techniques: Additional image enhancement techniques

can be applied to further improve the quality of fingerprint images. These techniques may

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 11
Fingerprint Matching Chapter 4

include noise reduction, ridge thinning, ridge orientation estimation, and ridge frequency
estimation. Noise reduction techniques, such as Gaussian filtering or median filtering,
help remove unwanted noise while preserving important ridge information. Ridge thinning
algorithms aim to reduce the width of ridges to enhance their clarity and distinguishability.
Ridge orientation estimation techniques help determine the local ridge direction, which
is essential for subsequent feature extraction and matching processes. Ridge frequency
estimation techniques estimate the frequency of ridges, allowing for adaptive filtering and
analysis.

Figure 4.4: Left: filtered image Right: enhanced Fingerprint image

performing these preprocessing steps, fingerprint images are standardized and prepared
for subsequent feature extraction and matching algorithms. These steps help to improve
the accuracy and reliability of fingerprint matching systems by enhancing the quality and
consistency of the input fingerprint images.

4.3 Architecture for Fingerprint Matching

Siamese network typically consists of twin subnetworks that share weights. The following
is an example of the architecture of a Siamese network for fingerprint matching:

1. Input Layer: Two fingerprint images are fed into the Siamese network as inputs.

2. Convolutional Layers: Each subnetwork begins with a series of convolutional lay-

ers. These layers apply filters to the input images, capturing local patterns and
features.

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 12
Fingerprint Matching Chapter 4

3. Pooling Layers: Pooling layers follow the convolutional layers to reduce the dimen-
sionality of the feature maps. Common pooling techniques include max pooling or
average pooling.

4. Fully Connected Layers: Flattened features from the pooling layers are passed
through fully connected layers. These layers learn high-level representations by com-
bining the local features.

5. Contrastive Loss Layer: The outputs of the fully connected layers from both sub-
networks are concatenated. This concatenated representation is passed through a
contrastive loss layer, which measures the similarity between the feature representa-
tions of the two input fingerprint images.

6. Output Layer: The output layer produces a similarity score or distance metric that
quantifies the similarity or dissimilarity between the input fingerprint images.

Figure 4.5: Siamese Network Architecture

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 13
Fingerprint Matching Chapter 4

4.4 Training Process

In this study, we trained the Siamese network with 63 different fingerprints using fingerprint
images of size 98 × 98. The fingerprint images of the same finger in the training set were
stored in the same folder. For the main network, VGG16 pre-trained weights were used for
subsequent network training. The training set comprised 66 different kinds of fingerprint
images belonging to different fingers, collected using an AS60x fingerprint collector. On
average, 10 images were sampled for each fingerprint in the training set. All images in the
training set underwent the fingerprint preprocessing steps described earlier. To increase the
adaptability of the training set, each fingerprint image was rotated five times, resulting in
a total of six images. Through this data augmentation technique, the number of images for
each fingerprint increased to an average of 60. Regarding the training of the comparative
network, two images of the same fingerprint kind were selected from the training set, and
the network output was calibrated to 1. Similarly, an image of a different fingerprint kind
was selected, and the output was calibrated to 0. This process was repeated with different
images from the training set. By training the network in this manner, when two fingerprint
images of the same finger were input, the network output was biased towards 1. Conversely,
when two fingerprint images of different fingers were input, the network output was biased
towards 0. The training was performed using an RTX2060 graphic processing unit (GPU).
The following training parameters and results were obtained:

• Batch size = 32

• Learning rate = 0.001

• Epochs = 15

• Total loss = 0.1149

• Training time: 1 hour

• Test time per match: 300 ms

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 14
Chapter 5

Datasets

5.1 Sokoto Coventry Fingerprint Dataset (SOCOF-

ing)
Sokoto Coventry Fingerprint Dataset (SOCOFing) is a biometric fingerprint database de-
signed for academic research purposes. SOCOFing is made up of 6,000 fingerprint images
from 600 African subjects and contains unique attributes such as labels for gender, hand
and finger name as well as synthetically altered versions with three different levels of al-
teration for obliteration, central rotation, and z-cut. Sokoto Coventry Fingerprint Dataset
(SOCOFing) is a biometric fingerprint database designed for academic research purposes.
SOCOFing is made up of 6,000 fingerprint images from 600 African subjects and contains
unique attributes such as labels for gender, hand and finger name as well as synthetically
altered versions with three different levels of alteration for obliteration, central rotation,
and z-cut.

Figure 5.1: Some Sample images from the dataset

15
Chapter 6

Results and Analysis

6.1 Training Results

6.1.1 Compatibility Test

A compatibility test was conducted to evaluate the proposed fingerprint database’s com-
patibility with conventional fingerprint recognition systems. The conventional Galton al-
gorithm was applied to the preprocessed image database proposed in this research.
The results of the compatibility test indicated that the proposed database was com-
patible with the conventional system. The targeted system, which featured the generated
database, demonstrated successful integration and compatibility.
This compatibility test confirms that the proposed fingerprint database can be effec-
tively utilized with existing fingerprint recognition systems, ensuring seamless integration
and interoperability.

6.1.2 Accuracy Test

A total of 3,826 fingerprint images of 66 different fingerprints were used to train the net-
work. The network was trained five times consecutively to ensure robustness and improve
the overall accuracy. However, it should be noted that the accuracy could potentially
befurther improved by including a larger amount of data in the training database.
The fingerprint images used in the detection program were distinct from those in the
training database. A total of 82 fingerprint images belonging to 10 different fingerprints
were used for fingerprint detection. These images had undergone the same preprocessing
and binarization steps described earlier to enhance testing efficiency and ensure compati-
bility with the detection program.

16
Fingerprint Matching Chapter 6

To facilitate organization and retrieval, fingerprint images of the same kind in the
database were placed in the same folder. This arrangement allows for easier management
and enables efficient access to specific fingerprint classes during the detection process.

Figure 6.1: Fingerprint Matching Experiment

6.2 Analysis
in this section, we present the analysis of the fingerprint matching system based on the
proposed methodology.

6.2.1 Performance Evaluation

The performance of the fingerprint matching system was assessed using various met-
rics,including accuracy, precision, recall, and receiver operating characteristic (ROC) curves.
The accuracy metric measures the overall correctness of the system in identifying matching
and non-matching fingerprint pairs. Precision reflects the proportion of correctly identified
matches out of all positive predictions, while recall represents the proportion of true evalu-
ate accurately identify matching pairs while minimizing false positives and false negatives.

6.3 Advantages
Our fingerprint matching system offers several advantages over traditional methods. Firstly,
it eliminates the need for manual minutiae extraction, reducing the processing time and

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 17
Fingerprint Matching Chapter 6

improving efficiency. Secondly, the use of Siamese network architecture allows for improved
feature representation and robust matching, resulting in higher accuracy rates. Further-
more, the system can handle partial fingerprints and is less sensitive to variations in image
quality, making it more versatile in real-world scenarios.

6.4 Limitations
Despite its strengths, our fingerprint matching system has certain limitations. Firstly, the
performance may degrade when handling low-quality or distorted fingerprint images, as
it heavily relies on the quality of input data. Secondly, the system may struggle with
matching fingerprints from individuals with severe skin conditions or unique patterns that
deviate from the norm. Additionally, the system’s performance may vary across different
fingerprint databases, requiring database-specific fine-tuning.

6.5 Potential Areas for Further Improvement

There are several potential areas for further improvement in our fingerprint matching
system. Firstly, integrating advanced preprocessing techniques, such as image enhance-
ment and noise reduction algorithms, could enhance the system’s robustness to various
image conditions. Secondly, exploring more advanced deep learning architectures, such
as attention mechanisms or recurrent neural networks, may further improve the system’s
accuracy and adaptability. Additionally, incorporating multimodal biometric fusion, such
as combining fingerprint and palmprint information, could enhance overall identification
performance.

6.6 Future Research Directions

In the future, we aim to explore several research directions to advance fingerprint matching
technology. Firstly, investigating privacy-preserving methods, such as secure and encrypted
fingerprint matching, could address privacy concerns while maintaining high accuracy.
Secondly, exploring large-scale distributed fingerprint matching systems could facilitate
efficient and scalable identification in real-time applications. Furthermore, integrating
artificial intelligence techniques, such as explainable AI or deep generative models, could
provide insights into the decision-making process and improve system transparency.

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 18
Fingerprint Matching Chapter 6

Overall, our fingerprint matching system has shown promising results, but there are still
opportunities for further evaluation, enhancement, and future research to address existing
limitations and push the boundaries of fingerprint recognition technology.

Department of Computer Science and Engineering(AI & ML), SCEM, Mangaluru Page 19
Chapter 7

Conclusion and Future work

In this study, we developed a fingerprint matching system using a Siamese network and
evaluated its performance. The system achieved a high accuracy rate in identifying match-
ing fingerprint pairs and showed robust performance across various evaluation metrics,
including accuracy, precision, recall, and ROC curves. The utilization of a Siamese net-
work allowed for direct learning of discriminative features fromfingerprint images, resulting
in improved accuracy and robustness compared to traditionalfeature-based methods. Addi-
tionally, data augmentation techniques, such as image rotation, enhanced the adaptability
and generalization capability of the system, enablingaccurate matching across different
variations in fingerprint patterns. The proposed fingerprint database demonstrated com-
patibility with existing recognition systems, facilitating seamless integration into real-world
applications without significant modifications.
The fingerprint matching project provides several key benefits. Firstly, it enhances
security measures by enabling accurate identification and verification of individuals based
on their fingerprints. This enhances access control and authentication systems, ensuring
reliable and secure operations. Secondly, the system offers a fast and efficient solution for
biometric identification tasks, improving the speed and accuracy of matching individuals
against large databases. This efficiency is valuable in various domains, including law
enforcement, border control, and secure facility access. Lastly, the system contributes to
forensic investigations by aiding in the identification of suspects and providing critical

20
References

[1] D. Maltoni, D. Maio, A. K. Jain, S. Prabhakar, et al., Handbook of fingerprint recog-

nition, vol. 2. Springer, 2009.

[2] T. Gowrishankar, “Fingerprint identification on a massively parallel architecture,”

in Proceedings 2nd Symposium on the Frontiers of Massively Parallel Computation,
pp. 331–332, IEEE Computer Society, 1988.

[3] M. Mori, “Automatic switching system for high-resolution phototelegraphic equip-

ment,” in IEEE International Carnahan Conference on Security Technology, Crime
Countermeasures, pp. 6–11, IEEE, 1990.

[4] C. Gunawardena and V. Sagar, “Fingerprint verification using coincident sequencing

and thinning,” in Proceedings IECON’91: 1991 International Conference on Industrial
Electronics, Control and Instrumentation, pp. 1917–1922, IEEE, 1991.

[5] B. G. Sherlock, D. Monro, and K. Millard, “Fingerprint enhancement by directional

fourier filtering,” IEE Proceedings-Vision, Image and Signal Processing, vol. 141, no. 2,
pp. 87–94, 1994.

[6] M. Akhan, I. Emiroglu, and E. Bahari, “A flexible fingerprint identification system,”

1995.

[7] V. Sagar, D. Ngo, and K. Foo, “Fuzzy feature selection for fingerprint identification,”
in Proceedings The Institute of Electrical and Electronics Engineers. 29th Annual 1995
International Carnahan Conference on Security Technology, pp. 85–90, IEEE, 1995.

[8] N. D. Tucker, “Pre-processing of fingerprint images,” in Proceedings The Institute

of Electrical and Electronics Engineers. 29th Annual 1995 International Carnahan
Conference on Security Technology, pp. 91–96, IEEE, 1995.

21
 [9] M.-H. Ghassemian, “A robust on-line restoration algorithm for fingerprint segmen-
tation,” in Proceedings of 3rd IEEE International Conference on Image Processing,
vol. 2, pp. 181–184, IEEE, 1996.

[10] D. P. Mital and E. K. Teoh, “An automated matching technique for fingerprint identi-
fication,” in Proceedings of 1st International Conference on Conventional and Knowl-
edge Based Intelligent Electronic Systems. KES’97, vol. 1, pp. 142–147, IEEE, 1997.

[11] L. Hong, A. Jian, S. Pankanti, and R. Bolle, “Fingerprint enhancement,” in Proceed-

ings Third IEEE Workshop on Applications of Computer Vision. WACV’96, pp. 202–
207, IEEE, 1996.

[12] A. Jain, L. Hong, and R. Bolle, “On-line fingerprint verification,” IEEE transactions
on pattern analysis and machine intelligence, vol. 19, no. 4, pp. 302–314, 1997.

[13] A. K. Jain, S. Prabhakar, L. Hong, and S. Pankanti, “Fingercode: a filterbank for fin-
gerprint representation and matching,” in Proceedings. 1999 IEEE Computer Society
Conference on Computer Vision and Pattern Recognition (Cat. No PR00149), vol. 2,
pp. 187–193, IEEE, 1999.

[14] A. K. Jain, S. Prabhakar, L. Hong, and S. Pankanti, “Filterbank-based fingerprint

matching,” IEEE transactions on Image Processing, vol. 9, no. 5, pp. 846–859, 2000.

[15] C.-J. Lee and S.-D. Wang, “A gabor filter-based approach to fingerprint recognition,”
in 1999 IEEE Workshop on Signal Processing Systems. SiPS 99. Design and Imple-
mentation (Cat. No. 99TH8461), pp. 371–378, IEEE, 1999.

[16] M. Horton, P. Meenen, R. Adhami, and P. Cox, “The costs and benefits of using
complex 2-d gabor filters in a filter-based fingerprint-matching system,” in Proceedings
of the Thirty-Fourth Southeastern Symposium on System Theory (Cat. No. 02EX540),
pp. 171–175, IEEE, 2002.

[17] L. Sha, F. Zhao, and X. Tang, “Improved fingercode for filterbank-based fingerprint
matching,” in Proceedings 2003 International Conference on Image Processing (Cat.
No. 03CH37429), vol. 2, pp. II–895, IEEE, 2003.

[18] A. Jain, A. Ross, and S. Prabhakar, “Fingerprint matching using minutiae and texture
features,” in Proceedings 2001 International Conference on Image Processing (Cat.
No. 01CH37205), vol. 3, pp. 282–285, IEEE, 2001.

22
[19] S. Wang and Y. Wang, “Fingerprint enhancement in the singular point area,” IEEE
signal processing letters, vol. 11, no. 1, pp. 16–19, 2004.

[20] F. You, Y. Shi, and P. Engler, “Fingerprint pattern recognition for medical uses-a
frequency domain approach,” in 1993 IEEE Annual Northeast Bioengineering Con-
ference, pp. 176–177, IEEE, 1993.

23