FINGERPRINT IMAGE IDENTIFICATION FOR CRIME

DETECTION
MOHD AMER1,DR. MOHD.ABDUL BARI2,DR. AKHIL KHARE3
*1. ME Student, Department Of Computer Science, ISL ENGINEERING COLLEGE.
*2. Head Of Department of Computer Science, ISL ENGINEERING COLLEGE
*3. Professor Of CSE Dept MVSR ENGINEERING COLLEGE
Abstract
With the introduction of biometrics technology which is an advanced computer techniques now
widely adopted as a front line security measure for both identity verification and crime detection,
and also others an effective crime deterrent. In an increasingly digital world, reliable personal
authentication has become an important human computer interface activity. Fingerprint
recognition could be very complex pattern recognition problem. It is difficult to design accurate
algorithms that are capable of extracting prolific features and comparing them in a robust way,
especially in poor quality fingerprint images and when low-cost acquisition devices with small
area are adopted. There is a greatest misconception that the fingerprint recognition is a fully
solved problem considering it was one of the first applications of all amongst machine pattern
recognition. A fingerprint is the feature pattern on the finger. It is proved through strong
evidences that each fingerprint is almost unique in nature. Each individual retains his own
fingerprints with the permanent and durable uniqueness. Hence fingerprints are being adapted for
identification and forensic investigation. The fingerprint recognition problem can be grouped
into two sub-categories: one is termed as fingerprint verification and while the other is termed as
fingerprint identification. Additionally, apart from the manual approach of fingerprint
recognition by the experts, fingerprint recognition here is usually referred as AFRS (Automatic
Fingerprint Recognition System), which is drastically program-based. The proposed system
presents the variation of Fast Fourier Transform on finger print recognition by fast fingerprint
minutiae extraction and recognition algorithm that improves the clarity between the ridge and
valley structures of the supposed to be provided finger print images in accordance with the
frequency as well as the orientation of the local ridges and extracting correct minutiae.
Key Words: Binarization, Segmentation, Minutiae Extraction, Minutiae marking

.1. INTRODUCTION cost efficient method for personal

In today’s computer era, one of the most identification. The components required for
commonly used biometric systems id biometric generation as well as
fingerprint which is used for personal identification is discussed as below:
identification as well as verification among Fingerprint Sensor: It is a device that is used
all other biometric system. Throughout the to extract the input data of the finger. With
world, it is one of the most acceptable respect to functionality and application
biometric system due to it fast, secure and different optical sensors are used. Feature

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 114
Extraction: Different algorithms are used for Many research works have been done to
feature extraction out of the captured extract fingerprint accurately but still there
fingerprint data. Database: All the extracted are some drawbacks that should be
features are stored in a database in particular eliminated. Many image enhancements. The
format. This database is used for fingerprint crime scene fingerprint is somehow noisy or
verification as well identification. Decision partial and very difficult to identify.
Making: Finally, decision making With the introduction of biometrics
algorithms are used for feature matching technology which is an advanced computer
from the database. techniques now widely adopted as a front
Fingerprint is the most interesting line security measure for both identity
and oldest human identity used for verification and crime detection, and also
recognition of individual. In the early others an effective crime deterrent. In an
twentieth century, fingerprint was formally increasingly digital world, reliable personal
accepted as valid signs of identity by law- authentication has become an important
enforcement agencies. The criminal human computer interface activity.
identification activities face a series of Fingerprint recognition could be very
photographic challenges to the crime scenes. complex pattern recognition problem. It is
These images can also be extremely fragile difficult to design accurate algorithms that
so hard to see. The background color or are capable of extracting prolific features
texture can easily overlap thin details. and comparing them in a robust way,
Curved, reflective or irregular surfaces may especially in poor quality fingerprint images
appear resist attempts to record all the and when low-cost acquisition devices with
details in one show. After all, pieces that small area are adopted. There is a greatest
need proof can often be difficult, if not misconception that the
impossible. In most cases under these 2. Literature Survey
circumstances, conventional photographic The goal of the segmentation algorithm is to
techniques have been successfully used to decompose the latent fingerprint image into
record such evidence discovered on the two parts Foreground and background. The
scene. foreground consists of the fingerprint and
One of the most important the background consists of the noise or
application of fingerprint identification on unwanted region. A few methods on latent
crime scene. The image obtained from crime fingerprint segmentation [1,2,3,4,5,6,7,8,10]
spot are of two types, one is patent have been purposed. In Zhang et al. [5], the
fingerprint samples and latent fingerprint author used the Adaptive Total Variation
samples. Among them, patent fingerprints Model for fingerprint image decomposition
are visible by our naked eye. In latent and feature selection of the multiscale
fingerprint image the extracted images are image. In this model, they used a total
invisible and very difficult to perceive. variant model with L1 fidelity denoted by,

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 115
TV-L1. The Weight coefficient in the TV- used multi-polar blocking and Gabor filter.
L1 model Adjusted accordingly with L1 To calculate the statistical characteristic of
fidelity that depends on the background gray the author used five indicators Mean
noise level.TV-L1 model decomposes the value, contrast, variance, low second-order
fingerprint image in cartoon and texture. movement, and high second-order
The cartoon contains a piecewise smooth movement but this method is not appropriate
component and texture contains the texture for too wet and too dry fingerprint image.
component of a fingerprint image.ATV does Short. Et al.[8], purposed Ridge template
not give good results for ugly latent correlation methods for latent fingerprint
fingerprints. In Lai et al [6], the Purposed segmentation. This method increases the
Novel Directional total variation (DTV) accuracy of latent fingerprint segmentation
model for latent fingerprint detection and and reduces the fingerprint area that is
segmentation. In the DTV model, the author detected as a fingerprint from 60.7% to
includes a special parameter in TV 33.6% of the total fingerprint image. It used
computation that is special –dependent a Hough-based method for the detection of
texture orientation. This parameter is structured noise such as line and removes it
appropriate for an image with oriented from the fingerprint image. It also reduces
textures.DTV not segments multiple false minutiae detection considered as
overlapped fingerprints. Zhang et al. [4], foreground and true minutiae considered as
Purposed Adaptive Directional Total background from 1.41% to 0.29%. When
Variation (ADTV) model that takes the detection of false minutiae is high its gives
advantages of both Adaptive Total Variation wrong segmentation results. Choi et al.[3],
Model and Directional total variation.ADTV purposed automatic extraction of latent
model used for latent fingerprint fingerprint based on a combined method that
enhancement and segmentation. It used uses frequency features and ridge
orientation and scale two features for image orientation. Ridge frequency of latent
segmentation. It decomposed the fingerprint fingerprint is found by the local Fourier
image into two parts texture and cartoon. analysis method and a similar pattern of
The texture contains the fingerprint part ridge fingerprint is found by the tensor
while the cartoon contains the structured method. In this model, the foreground area
noise (Character, stain, line, arch, speckle, is
etc.).ADTV is incapable of handling the 3. System Analysis
overlapped fingerprints region. Xue et al.[7], 3.1 Existing System
purposed fingerprint image segmentation Fingerprints in the crime scene plays an
based on a combined method, that combines important role to identify the criminal
statistical characteristics of gray with involved in the crime. Crime scene images
orientation field information. To calculate (CSI) are images taken from the crime spot.
the fingerprint block direction the author When crime is occurred, the investigator

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 116
takes both latent and patent sample of scene may be partial or tough to identify.
fingerprints left behind. The patent This can lead to fingerprint images of bad or
fingerprints are visible by naked eye, so they low quality. Due to the low quality of
are simply photographed. But latent fingerprint image to another systematic
fingerprints are invisible and these samples image feature, the early fingerprint image
are more difficult to perceptible. These quality may be of exterior value of
samples can be lifted through different identification. To moderate this problem, the
techniques. The use of cyanoacrylate fingerprint allows performing image pre-
vapours which sticks to prints and make processing, to feature mark up and
them visible in the present of normal light. identification analysis. Segmentation is a
This method is much difficult, so normally first step in image enhancement which
in crime scene, the investigators apply a fine converting low-level image processing
dusting powder (aluminium dust or black transforming a gray scale image into high-
granular) to the surface in which fingerprints level image description in terms of features,
to be extracted. The dust actually sticks to objects and scenes.
the fingerprint then they use clear tape to lift SYSTEM ARCHITECTURE:
the fingerprint. After the lifting the
fingerprints, the prints are scanned and
saved in the digital image form. The
fingerprints taken from the crime scene is
unintentionally made and these images are
noisy or partial prints and difficult to
identify.
3.2 Proposed System
The CNN uses successive convolutional
layers with a non-linear ReLu function for
storing the features of an image having a
specific dimension. Max pooling layers are
used for down sampling. The fully
connected layer multiplies the input by a
matrix with sigmoid activation function and
adds to a bias vector which contains the
feature map. The images obtained from the
crime scene are called crime scene images 3.2 Modules Description
(CSI). These images play important role and Pre processing stage:-
used as evidence in criminal cases. The Again pre processing stage is divided in to
fingerprint information contained in the three sub stages such as:-i) image
images collected directly from the crime enhancement ii) image binarization iii)
image segmentation. For image

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 117
enhancement we used two methods such as:- on the organization. The amount of fund that
histogram equalization and Fourier the company can pour into the research and
transform. After enhancing the image we development of the system is limited. The
need to binaries the image for that we used expenditures must be justified. Thus the
the locally adaptive threshold method. For developed system as well within the budget
image segmentation we preferred a three- and this was achieved because most of the
step approach such as :- i) block direction technologies used are freely available. Only
estimation ii) segmentation by direction the customized products had to be purchased
intensity iii) Region of Interest (ROI)
extraction by Morphological operations. 3.5 Operational Feasibility
Are you into the production of “things”?
Minutia extraction:- Perhaps, your answer would be yes. We
Minutia extraction stage is divided in to two naturally don’t call them things; instead, we
sub stages such as:-i) fingerprint ridge call them products, services, or systems.
thinning and ii) minutia marking We used Using the term “things” sounds foreign
iterative parallel thinning algorithm for because you can’t just drop them into an
minutia extraction stage. Ridge thinning is area without touching them. They need to be
used to used to eliminate the redundant connected to an existing service or business.
pixels of the ridges 16 till the ridges are of These “things” are an extension of the
one pixel wide.The minutia marking is quite organization where they are produced.
simple task. Here crossing number (CN) 3.6 Technical Feasibility
concept is used. This study is carried out to check the
3.3 Feasibility Study technical feasibility, that is, the technical
The feasibility of the project is analyzed in requirements of the system. Any system
this phase and business proposal is put forth developed must not have a high demand on
with a very general plan for the project and the available technical resources. This will
some cost estimates. During system analysis lead to high demands on the available
the feasibility study of the proposed system technical resources. This will lead to high
is to be carried out. This is to ensure that the demands being placed on the client. The
proposed system is not a burden to the developed system must have a modest
company. For feasibility analysis, some requirement, as only minimal or null
understanding of the major requirements for changes are required for implementing this
the system is essential. system
Three key considerations involved in the
feasibility analysis are, 3.7 Social Feasibility
3.4. Economic Feasibility The aspect of study is to check the
This study is carried out to check the level of acceptance of the system by the
economic impact that the system will have user. This includes the process of training

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 118
the user to use the system efficiently. The Technologies and Languages used to
user must not feel threatened by the system, Develop
instead must accept it as a necessity. The 1. Python
level of acceptance by the users solely
depends on the methods that are employed Debugger and Emulator
to educate the user about the system and to  Any Browser (Particularly
make him familiar with it. Chrome)
His level of confidence must be raised so Hardware Requirements
that he is also able to make some For developing the application the following
constructive criticism, which is welcomed, are the Hardware Requirements:
as he is the final user of the system  Processor: Pentium IV or
4. System Requirements Specification higher
4.1 Introduction  RAM: 256 MB
The project involved analyzing the design of Space on Hard Disk: minimum 512MB
few applications so as to make the 4.4 Non Functional Requirements
application more users friendly. To do so, it  Any Browser (Particularly
was really important to keep the navigations Chrome)
from one screen to the other well ordered 4.5 Input & Output Design
and at the same time reducing the amount of 4.5.1 Input design :
typing the user needs to do. In order to make The input design is the link
the application more accessible, the browser between the information system and the
version had to be chosen so that it is user. It comprises the developing
compatible with most of the Browsers specification and procedures for data
4.2 Purpose preparation and those steps are necessary to
In order to make the application more put transaction data in to a usable form for
accessible, the browser version had to be processing can be achieved by inspecting the
chosen so that it is compatible with most of computer to read data from a written or
the Browsers. printed document or it can occur by having
4.3 Functional Requirements people keying the data directly into the
For developing the application, the system. The design of input focuses on
following are the Software Requirements: controlling the amount of input required,
1. Python controlling the errors, avoiding delay,
2. anaconda avoiding extra steps and keeping the process
Operating Systems supported simple.
1. Windows 7 The input is designed in such a way so that it
2. Windows XP provides security and ease of use with
3. Windows 8 retaining the privacy. Input Design
considered the following things:

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 119
  What data should be given as outputs. In output design it is determined
input? how the information is to be displaced for
 How the data should be arranged immediate need and also the hard copy
or coded? output. It is the most important and direct
 The dialog to guide the operating source information to the user. Efficient and
personnel in providing input. intelligent output design improves the
Methods for preparing input validations and system’s relationship to help user decision-
steps to follow when error occur making.
OBJECTIVES 1. Designing computer output
1.Input Design is the process of converting a should proceed in an organized, well
user-oriented description of the input into a thought out manner; the right output must be
computer-based system. This design is developed while ensuring that each output
important to avoid errors in the data input element is designed so that people will find
process and show the correct direction to the the system can use easily and effectively.
management for getting correct information When analysis design computer output, they
from the computerized system. should Identify the specific output that is
2.It is achieved by creating user- needed to meet the requirements.
friendly screens for the data entry to handle 2.Select methods for presenting information.
large volume of data. The goal of designing 3.Create document, report, or other formats
input is to make data entry easier and to be that contain information produced by the
free from errors. The data entry screen is system.
designed in such a way that all the data The output form of an
manipulates can be performed. It also information system should accomplish one
provides record viewing facilities. or more of the following objectives.
3.When the data is entered it will check for  Convey information about past
its validity. Data can be entered with the activities, current status or
help of screens. Appropriate messages are projections of the
provided as when needed so that the user  Future.
will not be in maize of instant. Thus the  Signal important events,
objective of input design is to create an input opportunities, problems, or warnings.
layout that is easy to follow  Trigger an action.
Confirm an action
4.5.2 OUTPUT DESIGN 4.6 Hardware Requirements
A quality output is one, which meets the For developing the application the following
requirements of the end user and presents are the Hardware Requirements:
the information clearly. In any system  Processor: Pentium IV or
results of processing are communicated to higher
the users and to other system through  RAM: 256 MB

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 120
Space on Hard Disk: minimum 512MB end of the system design all the major data
4.7 Software Requirements structures,file formats, output formats,
For developing the application the following andthe major modulesin thesystemandtheir
are the Software Requirements: specificationsaredecided.
3. Python
4. anaconda During, Detailed Design, the internal logic
Operating Systems supported of each of the modules specified in system
4. Windows 7 designis decided. During this phase, the
5. Windows XP details of the data of a module are usually
6. Windows 8 specified in ahigh-level design description
Technologies and Languages used to language, which is independent of the
Develop target language in whichthesoftware will
Python eventuallybe implemented.
5. System Design
Insystemdesignthefocusisonidentifyingthe
The purpose of the design phase is to plan
modules,whereasduringdetaileddesignthe
a solution of the problem specified by
focus is on designing the logic for each of
their requirement document. This phase is
the modules.In other works, in system
the first step in moving from the problem
domain to the solution domain. In other
designtheattentionisonwhatcomponentsare
words, starting with what is needed,
needed,whileindetaileddesignhowthecomp
design takes us towardhow to satisfy the
onentscanbe implemented insoftwareis the
needs.The design of a system is perhaps
issue.
the most critical factoraffection the quality
of the software; it has a major impact on
Design is concerned with identifying
the later phase, particularlytesting,
software components specifying
maintenance.The output of this phase is
relationships amongcomponents.
the design document.This document
Specifying software structure and
issimilar to a blueprint for the solution and
providing blue print for the document
is used later during implementation,
phase.Modularity is one of the desirable
testing andmaintenance.The design
properties of large systems. It implies that
activity is often divided into two separate
the system isdivided into several parts. In
phases System DesignandDetailedDesign.
such a manner, the interaction between
parts is minimal clearlyspecified.
System Design also called top-level design
aims to identify the modules that should
During the system design activities,
be inthe system, the specifications of these
Developers bridge the gap between the
modules, and how they interact with each
requirementsspecification, produced
other toproduce the desired results.At the

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 121
during requirements elicitation and classes, interfaces, objects, components,
analysis, and the system that isdelivered to and nodes. The four structural diagrams
theuser. are −
Design is the place where the quality is Class diagram
fostered in development. Software design is Object diagram
aprocessthroughwhichrequirementsaretransl Component diagram
atedintoarepresentationofsoftware Deployment diagram
5.5 UML Diagrams (9 types)
Any complex system is best understood by Class Diagram
making some kind of diagrams or pictures. Class diagrams are the most common
These diagrams have a better impact on diagrams used in UML. Class diagram
our understanding. If we look around, we consists of classes, interfaces,
will realize that the diagrams are not a associations, and collaboration. Class
new concept but it is used widely in diagrams basically represent the object-
different forms in different industries. oriented view of a system, which is static
We prepare UML diagrams to understand in nature. Active class is used in a class
the system in a better and simple way. A diagram to represent the concurrency of
single diagram is not enough to cover all the system.
the aspects of the system. UML defines Class diagram represents the object
various kinds of diagrams to cover most of orientation of a system. Hence, it is
the aspects of a system. You can also generally used for development purpose.
create your own set of diagrams to meet isis the most widely used diagram at the
your requirements. Diagrams are generally time of system construction.
made in an incremental and iterative way.
There are two broad categories of
diagrams and they are again divided into Object Diagram
subcategories – Object diagrams can be described as an
Structural Diagrams instance of class diagram. Thus, these
Behavioral Diagrams diagrams are more close to real-life
scenarios where we implement a system.
Structural Diagrams Object diagrams are a set of objects and
The structural diagrams represent the their relationship is just like class
static aspect of the system. These static diagrams. They also represent the static
aspects represent those parts of a diagram, view of the system.
which forms the main structure and are
therefore stable. The usage of object diagrams is similar to
class diagrams but they are used to build
These static parts are represented by prototype of a system from a

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 122
practical perspective.
Component Diagram Behavioral diagrams basically capture the
Component diagrams represent a set of dynamic aspect of a system. Dynamic
components and their relationships. These aspect can be further described as the
components consist of classes, interfaces, changing/moving parts of a system.
or collaborations. Component diagrams
represent the implementation view of a UML has the following five types of
system. During the design phase, software behavioral diagrams −
artifacts (classes, interfaces, etc.) of a Use case diagram
system are arranged in different groups Sequence diagram
depending upon their relationship. Now, Collaboration diagram
these groups are known as components. Statechart diagram
Finally, it can be said component diagrams Activity diagram
are used to visualize the implementation. 6. Implementation
Deployment Diagram Pre processing stage:-
Deployment diagrams are a set of nodes Again pre processing stage is divided in to
and their relationships. These nodes are three sub stages such as:-i) image
physical entities where the components are enhancement ii) image binarization iii)
deployed. Deployment diagrams are used image segmentation. For image
for visualizing the deployment view of a enhancement we used two methods such as:-
system. This is generally used by the histogram equalization and Fourier
deployment team. Note − If the above transform. After enhancing the image we
descriptions and usages are observed need to binaries the image for that
carefully then it is very clear that all the we used the locally adaptive threshold
diagrams have some relationship with one method. For image segmentation we
another. Component diagrams are preferred a three-step approach such as :- i)
dependent upon the classes, interfaces, etc. block direction estimation ii) segmentation
which are part of class/object diagram. by direction intensity iii) Region of Interest
Again, the deployment diagram is (ROI) extraction by Morphological
dependent upon the components, which operations.
are used to make component diagrams. Minutia extraction:-
Minutia extraction stage is divided in to two
Behavioral Diagrams sub stages such as:-i) fingerprint ridge
thinning and ii) minutia marking We used
Any system can have two aspects, static iterative parallel thinning algorithm for
and dynamic. So, a model is considered as minutia extraction stage. Ridge thinning is
complete when both the aspects are fully used to used to eliminate the redundant
covered. pixels of the ridges 16 till the ridges are of

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 123
one pixel wide.The minutia marking is quite Consider the below image:
simple task. Here crossing number (CN)
concept is used.
7. Technology Description
PYTHON
Python is a general-purpose interpreted,
interactive, object-oriented, and high-level
programming language. An interpreted
language, Python has a design philosophy
that emphasizes code readability (notably
using whitespace indentation to delimit code
blocks rather than curly brackets or We can figure out that it is an image of the
keywords), and a syntax that allows New York Skyline. But, can a computer find
programmers to express concepts in this out all on its own? The answer is no!
fewer lines of code than might be used in The computer reads any image as a range of
languages such as C++or Java. It provides values between 0 and 255.
constructs that enable clear programming on For any color image, there are 3 primary
both small and large scales. Python channels – Red, green and blue. How it
interpreters are available for many operating works is pretty simple.
systems. CPython, the reference A matrix is formed for every primary color
implementation of Python, is open and later these matrices combine to provide
source software and has a community-based a Pixel value for the individual R, G, B
development model, as do nearly all of its colors.
variant implementations. CPython is Each element of the matrices provide data
managed by the non-profit Python Software pertaining to the intensity of brightness of
Foundation. Python features a dynamic the pixel.
type system and automatic memory Consider the following image:
management. It supports
multiple programming paradigms,
including object-
oriented, imperative, functional and procedu
ral, and has a large and
comprehensive standard library.
Interactive Mode Programming
Invoking the interpreter without passing a
script file as a parameter brings up the
following prompt −
How Does A Computer Read An Image?

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 124
As shown, the size of the image here can be observed CNN gives better performance
calculated as B x A x 3. compared to SVM because of its deep
11. Future Enhancement learning ability to learn relevant features
The future work is to develop a high- from the image. Using CNN classifier,
resolution fingerprint recognition framework improved fingerprint identification accuracy
that better utilizes the CNN features. In of 80% is achieved.
future, work will be focused on enhanced of 13. References
accuracy rate for partial fingerprint images. [1] H. Elouedi, W. Meliani, and Z. Elouedi,
12. Conclusion “A hybrid approach based on decision trees
We have proposed a new model that is a and clustering for breast cancer
th
fusion of morphological and neural network classification,” in 6 International
approaches for latent fingerprint Conference of Soft Computing and Pattern
segmentation. The morphological method Recognition (SoCPaR), 2014.
finds the fingerprint region from the large [2] F.Kharbat, H.Ghalayini, “New algorithm
image. The fingerprint region found by for Building Ontology from Existing Rules:
morphological method divides into A Case Study,” in International Conference
patches/blocks. Extract the features of each on Information Management and
block and used these features as an input of Engineering, 2009, pp. 12-16.
the NN. The NN model classifies the [3] P. Hamsagayathri and P. Sampath,
patches into fingerprint and non-fingerprint “Performance analysis of breast cancer
patches. We remove the non-fingerprint classification using decision tree classifier,”
patches that contain the structured noise and International Journal Of Current
reconstruct the fingerprint patches for Pharmaceutical Research (IRCPR), vol. 9,
finding the original fingerprint. Our future 2017.
work involves developing new algorithms
for improving the accuracy of the [4] R. Sumbaly, N. Vishnusri, and S.
classification and feature extraction for the Jeyalatha, “Diagnosis of breast cancer using
segmented latent fingerprints. decision tree data mining technique,”
Fingerprint identification system International Journal of Computer
used for identifies the criminal who involved Applications, vol. 98, no. 10, p. 0975 8887,
in the crime helps to automate fingerprint 2014.
identification process. Pre-processing was [5] C. P. Utomo, A. Kardina, and R.
performed with Otsu thresholding, Yuliwulandari, “Breast cancer diagnosis
fingerprint thinning and minutiae extraction using artificial neural networks with extreme
with Cross-Number method. Feature learning techniques,” International Journal
extraction will be done by the CNN of Advanced Research in Artificial
classifier. The performance of SVM and Intelligence (IJARAI), vol. 3, no. 7, pp. 10–
CNN based classifiers are analyzed. It is 14, 2014.

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 125
[6] S.Jhajharia, H.K.Varshney, S.Verma,
and R.Kumar, “A neural network based
breast cancer prognosis model with PCA
processed features,” in International
Conference on Advances in Computing,
Communications and Informatics (ICACCI),
2016.

Volume 12, Issue 10, Oct 2022 ISSN 2457 – 0362 Page 126