2022 B.

Tech Final Year Project (MLRIT -2022), Hyderabad, INDIA

MULTIPLE BIOMETRIC AUTHENTICATION

THROUGH IMAGE ASSESMENT USING MACHINE
LEARNING
Mrs.A.Sangeetha T Nikhitha Chowdary
Department of Computer Science and Engineering UG Student
MLR Institute of Technology Department of Computer Science and Engineering
Hyderabad,India MLR Institute of Technology
[email protected] Hyderabad,India
[email protected]

Battuwar Sai Teja G Sai Yashwanth Reddy

UG Student UG Student
Department of Computer Science and Engineering Department of Computer Science and Engineering
MLR Institute of Technology MLR Institute of Technology
Hyderabad,India Hyderabad,India
[email protected] [email protected]

B Arjun sahas
UG Student
Department of Computer Science and Engineering
MLR Institute of Technology
Hyderabad,India
[email protected]

Abstract —It is proposed to use a combination of three the retina. The colour of the iris is often referred to as "eye
biometric identities: iris, palm print, and face. Firstly, Input colour” .Iris recognition refers to the capability of verifying
images that contain noise undergo pre-processing from which the identity of a person through his eyes i.e. Iris. The human
key features are extracted. These features are compared with iris is the space located around the pupil of the eye and is
the existing images present in the database that are considered
to be Authenticated. In this stage, we obtain Matching scores.
covered by the cornea layer which comes under the unique
A module fuses these individual scores obtained. The final information of a person[3].
fusion is capable of classifying a person to be authenticated or
un-authenticated.
For each subject, we collect palm print images
Keywords — Image Quality Assessement, Pre-processing,
from both left and right hands. These images are 8 bit gray-
Feature extraction, Biometrics, DWT Segmentation, Image
Fusion.
level JPEG files by our self-developed face recognition
device. Subjects are required to put his face near to the
capturing hardware and make sure that he is in Evenly
I. INTRODUCTION coloured background. The device takes the biological
features of a person.
The process begins with image pre-processing,
which includes noise removal and then feature extraction.
We have to interpret the shapes of Iris, palm print and face
properly because we need to train the computer which
cannot distinguish the shapes as human do. Another issue is
that an image can be influenced by a variety of elements II. LITERATURE SURVEY
such as the angle at which we take the picture or the lighting
conditions in the area. [1],[7]. A.
During a literature survey, we collected some of
the information about the mechanisms of detecting fake
A. Overview biometrics .
The iris is thin layer, circular component in the eye,
is responsible for the control of diameter and size of the Survey1:- On the vulnerability of face verification systems
pupil which is proportional to the amount of light reaching to hill-climbing attacks
Using Bayesian we try to test the weaknesses of two face
recognition systems via adaptation.. The hill-climbing
assault algorithm is used here. This attack considers the
distinct properties such as eigenface based and parts based
verification systems. We may conclude that the proposed
method has successfully bi-passed over the bulk of the III. SYSTEM ARCHITECTURE
accounts that have been attacked[5 [8].

SURVEY2:- A HIGH PERFORMANCE FINGERPRINT LIVENESS

DETECTION METHOD
A new software-based liveness detection
method is proposed that takes image quality parameters into
account. This scenario is tested on many of the databases
that are comprised of both real and duplicate fingerprints.
The proposed solution is proved to be efficient to the multi-
scenario dataset, and presents accurately classified samples
that are nearly 90%. This method added the proof for
previous technologies which propose that images are
required for authentication but in a more secured format
[4].

Survey3:- Finger print verification system’s evaluation

on direct attacks
All the possible test cases that are given to finger print
verification systems that includes and excludes the user’s Fig 1 System architecture
knowledge are studied. Considering the quality of the
images, and the observations that are obtained after different
operations , we conclude the robustness of the systems to Image Acquisition:-
various direct attacks. [5] It is the first step in any vision systems, because without an
image, no processing is possible. In this process, we retrieve
our required images from the source i.e. a human being is
A. Existing System
used in this case. These retrieved images are completely
Existing systems include techniques such as unprocessed which means they consist of so much un-
Edge Detection, Segmentation and Feature Vector. necessary data like noise etc. Hence these images are now
Presently the biometrics are done using Finger printing. sent for desmissal of all such weird data.
But it is not much flexible because we can make Pre-Processing:-
duplicates of fingers and bluff people. Hence making it For Pre-processing the images, here we use Median
less efficient. We will be using PCA i.e. Principal filtering method. The main use of inculcating this method is
Component Analysis algorithm for finding the co- that it replaces each pixel value with the median value of the
variance and variance. Only spatial domain is the pixels that are in neighborhood to that particular pixel.
calculated. Previously, that was not the case where we considered the
B. Proposed System central value of the pixel. Also, it is capable of obtaining the
filtered images with less blurring rather than traditional
We can implement the proposed method, On applying
downscaling and color level transform, we pre-process the methods. Hence, it is an effective method in Data cleaning
image. Biometrics based on the consideration of iris, palm process.
print and face features for person’s authentication using Discrete Wavelet Transform:-
Sequential modified haar wavelet and Energy feature A signal is decomposed into a set of basis by
extraction using key generation analysis[6]. Wavelet transform which are known as wavelets.

Where a-> scaling parameter

b -> shifting parameter
 Figure 3: Application Interface

(ii) Fused Image formed out of input Images:

Figure 2: 1D- Discrete Wavelet Transform

Discrete wavelet transform decomposes the input image Figure 4: Fused Image is formed
into four levels which are LL,HL,LH and HH[8].

Features Extraction Process:- (iii) Output classifying Authenticate or not:

In this process, we obtain a matrix named Co-occurrence
matrix (CCM).Each element of the matrix is comprised of
the values i and j. It calculates the occurrence of i valued
pixel intensity in relationship with j valued pixel. The size
of the CCM is determined by the number of gray levels in
the image. The levels that are obtained from DWT
Segmentation are used for considering the data energy.
Based on the orientation and distance between image pixels,
co-occurrence matrix is constructed.

Figure 5: POC showing that the person is Authenticate

IV. EVALUATION AND RESULTS
Firstly, images are acquired through Image
Acquisition process and then these are pre-processed to V. CONCLUSION
remove the un-necessary data. Now, we apply DWT
transformation for decomposition. Features are extracted
and compared to that of data base images obtaining the Our existing database consists of 4 different iris, face and
matching scores that are put into CCM matrix. finger print images of 4 persons and code matrix is formed.
When the photographs of Iris, palm, and face of the
(i) GUI of the Application: individual who needs to be classified are received, a code
matrix is produced, similar to the prior technique. Once the
matching of patterns is done, then we can classify if the
person is authenticated or not.

VI. FUTURE WORK

As part of future enhancement, we have to develop

this method even for physically handicapped people who
might be in loss of any one of the biometric traits which are [4]https://hidglobal.com/sites/default/files/resource_files/hid
Iris, palm Print and face. As a result, an effective real-time -lumidigm-spoof-detection-schemes-wp-en.pdf
solution for bogus biometric detection is provided.
[5] G.Susmitha Valli “Scalar Vs Vector Color Image
Processing :An analysis” JARDCS Issue No:7 Special
ISSN No:1943–023X 2018.
[6] K.Srinivas Rao “Fast fifinger print retrieval using
minutiae neighbor structure” Springer volume 705
VII. REFERENCES
ISBN978-981-10-8569-7 2018.
[7] Dr.Illaiah Kavati “Score based indexing and
[1]https://research.ijcaonline.org/volume78/number16/pxc3 retrievaltechnique for biometric databases” IJPRAI Vol. 31,
891412.pdf No. 06, 1756009 (2017) 2017.
[8] J. Galbally, J. Fierrez, F. Alonso-Fernandez, and M.
[2]https://link.springer.com/article/10.1007/s00371-010- Martinez-Diaz, “Evaluation of direct attacks to fingerprint
0500-8 verification systems,” J. Telecommun. Syst., vol. 47, nos. 3–
4, pp. 243–254, 2011.

[3]https://www.researchgate.net/publication/317138401_Eff
icient_Score-
Based_Indexing_Technique_for_Fast_Palmprint_Retrieval