Lalithya

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 35

|| JAI SRI GURUDEV ||

SRI ADICHUNCHANAGIRI SHIKSHANA TRUST ®

ADICHUNCHANAGIRI INSTITUTE OF TECHNOL OGY


Slide No. 1

(An ISO 9001 : 2008 Certified)


Affiliated to Visvesvaraya Technological University, Belagavi ,
Approved by AICTE and Accredited by NAAC, New Delhi
Jyothinagar, Chikkamagaluru – 577 102.
DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

TECHNICAL SEMINAR (18CSS84) ON

“FINGER VEIN FEATURE EXTRACTION AND


RECOGNITION USING CNN”

Under the Guidance of:


Presented By:
Mrs. Kavya T.M, B.E., M.Tech.
Lalithya M.S
Assistant Professor
06/07/2022
Tuesday, June 7, 2022 1
[4AI18CS039]
DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. Tuesday, June 7, 20
CONTENTS
Slide No. 2

 ABSTRACT

 INTRODUCTION

 LITERATURE SURVEY

 MOTIVATION

 PROBLEM STATEMENT

 SCOPE OF THE PROJECT

 OBJECTIVES

 METHODOLOGY

 BOARD WORK

 CONCLUSION

 REFERENCES

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


ABSTRACT
Slide No. 3

 This has arisen a need for high security authentication systems


in order to safeguard personal information stored on mobile
devices.
 Since the high complexity and security of existing biometric
systems is becoming the prime concern in terms of space and
time, automated personal identification using finger-vein
biometric is emerging to be a prominent topic in both practical
as well as examination applications.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


INTRODUCTION
Slide No. 4

 The biometric identifiers know how to be duplicated and


can be made to be used as artifacts that can betray
numerous presently accessible biometric gadgets
 The aim of image preprocessing is to enhance some image
features relevant for further processing task.
 As a result, interesting details in the image are highlighted
and noise is removed from the image.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


INTRODUCTION (CONTD….)
Slide No. 5

 The finger vein images are preprocessed by applying


morphological operations and is enhanced based on the
concept of local histogram equalization.
 Then relevant features are extracted from the enhanced
finger vein image by feature extraction techniques.
 The extracted features are unique in every person and they
can be used in the authentication stage.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


LITERATURE SURVEY
Slide No. 6

SL. TITLE AUTHORS DESCRIPTION


NO
1. “Analysis of “Ruqaiya Khanam, By the image processing we get the vein
Finger Vein Ramsha Khan, image and compare it with DA and CNN
Feature Rajeev Ranjan” algorithm.
Extraction and
Recognition using
DA and CNN
Methods”

2. “A Novel Finger “Reshma Rajan”


By using the image of finger and get the
Vein Feature
Extraction ROI to find the vein pattern.
Technique for
Authentication”

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


MOTIVATION
Slide No. 7

 Live body identification: The infrared camera can capture


finger vein images only if deoxygenated hemoglobin is present in
the body and the imaging can only be done from live body. This
will provide extra security.
 High security: Unlike most of the biometric patterns in use like
fingerprint, face, iris etc., the vein pattern is seen inside the body
and is difficult to replicate and reuse it.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


PROBLEM STATEMENT
Slide No. 8

To recognize the finger vein using CNN Algorithm


Input: Finger image.
Processing:
Finger vein image processing is done by binarizing the image,
then detecting the edges using Sobel filter and after enhance the
image by using local histogram algorithm.
Apply the CNN algorithm falls under supervised machine
learning. 
Output : Recognition of finger vein pattern.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


OBJECTIVES
Slide No. 9

 To identify individuals and verify their identity.


 To get unique vein patterns
 To Attain maximum security.
 To attain maximum accuracy.
 To correctly identify the person using vein image.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


SCOPE OF THE TOPIC
Slide No. 10

 Is used in a wide variety of situations, including desktop


applications such as PC login or attendance management.
 Used in embedded system applications such as bank
automated teller machines (ATMs) or access control systems

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


SYSTEM DESIGN
Slide No. 11

Fig 1:Block Diagram of finger vein image processing.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


SYSTEM DESIGN (CONTD….)
Slide No. 12

 The acquired input finger vein image contains finger portion


and its background.
 For segmenting the ROI, first the finger vein image is
employed to binarization using a selected threshold value.
 Edge detector is applied to the input image, and the resulting
edge map is subtracted from the binarized image using sobel
filter.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


SYSTEM DESIGN (CONTD….)
Slide No. 13

 By applying histogram, the largest bounded object in the image


was extracted that resembles the ROI mask.
 Local histogram equalization is applied to the resulting image
to obtain the final enhanced vein image.
 Accuracy calculation of the features that are extracted from
enhanced image of patterns of finger veins.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK
Slide No. 14

Step 1: First, the finger image will be given as input.


Step 2: The input image will be converted into binary or to gray
scale.
Step 3: Sobel method is used to edge detection.
Step 4: The image will be enhanced using local histogram
equalization method.
Step 5: The segmentation is done based on Region of Interest
(ROI).
Step 6: Finally, pattern will be recognized using CNN.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK(CONTD….)
Slide No. 15

Step1: First, the finger image will be given as input.

Fig1: Finger Image.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 16

Step 2: The input image will be converted into binary or to gray scale.

Fig2: Gray Scale Image.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 17

FORMULA= 0.299 * R + 0.5870 * G + 0.1140 * B.

Example:

R: 224 G: 63 B: 69

= 0.299*224 + 0.5870*63 + 0.1140*69

=112

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARDWORK (CONTD….)
Slide No. 18

Conversion of Grayscale to binary

min max scalar


V’= V-min (newmax-newmin)
+newmin
max-min
= 254-112 (1-0)+0
254-112
=1

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARDWORK (CONTD….)
Slide No. 19

Step 3: The finger part will be cropped.

Fig 3: Cropped Finger Image.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 20

Step 4: The edge will be detected using Sobel detection


method

Fig 4: Edge detected Image

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 21

Step 5: The segmentation is done based on Region of Interest


(ROI).

Fig 5: Image segmentation based on ROI.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 22

Step 6: Pattern will be recognized using CNN.

Fig 6: Pattern Recognition using CNN.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD ….)
Slide No. 23

Convolution Neural Network


 Convolution neural network is a type of artificial neural network used in
image recognition and processing that is specifically designed to process
pixel data.
Step 1: Convolution

Step 2: Max Pooling

Step 3: Flattening

Step 4: Fully Connected

Figure 7 : Steps in Convolution neural network

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 24

Step 1- Convolution Layer


 The objective of the Convolution operation is to extract the high-level
features such as edges, from the input image.

Filter

Input image

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 25

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 26

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 27

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 28

Step 2 - Max Pooling

Max Pooling

Pooled Feature map


Feature map

Max Pooling

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 29

Max Pooling

Max Pooling

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 30

Step 3- Flattening

 Flattening of 2x2 image matrix into a 4x1 vector.


 Flattening is converting the data into 1-dimensional array for input to
next layer.

Flattening

Pooled Feature map

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


BOARD WORK (CONTD….)
Slide No. 31

Step 4- Fully Connected

Figure 8 : Feature Extraction and Classification.

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


CONCLUSION
Slide No. 32

 The ROI of finger images was successfully extracted by


applying morphological operations irrespective of any
illumination variations and these images were enhanced using
local histogram equalization.
 For feature map extraction, Frangi filter was used to extract
the finger vein structure from finger vein images and it gives
better results than the existing methods for the same

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


REFERENCES
 
Slide No. 33

 Yilong Yin, Lili Liu, and Xiwei Sun, “SDUMLA-HMT: A Multimodal Biometric
Database” Biometric Recognition Springer, Lecture Notes in Computer Science, vol.
7098, pp 260-268,2011.
 M. Kono, H. Ueki, and S. Umemura, “Near-infrared finger vein patterns for personal
identification,” Applied Optics, vol.41, no. 35, pp.7429–7436, December 2002.
 Kejun Wang, Hui Ma, Oluwatoyin P. Popoola and Jingyu Li, “Finger vein recognition”,
2010.
 N. Miura, A. Nagasaka, and T. Miyatake, “Feature extraction of finger vein patterns
based on repeated line tracking and its application to personal identification,” Machine
Vision and Applications, vol. 15, no. 4, pp. 194–203, October 2004.
 Reshma Rajan and Indu M.G. “A Novel Finger Vein Feature Extraction Technique for
Authentication", International Conference on Magnetics, Machines & Drives (AICERA-
2014 iCMMD).

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. 04 MAY 2022


Slide No. 34

S (Q & A)
SC U S SI ON
DI

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. Tuesday, June 7, 20


Slide No. 35

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING, A.I.T., CHIKKAMAGALURU. Tuesday, June 7, 20

You might also like