Many approaches have been proposed for developing fingerprint recognition systems. Some of them give inaccurate results due to low-quality images or have high time cost. The presentation focuses on singular points extraction from low quality image and then matching fingerprint within low time cost.
Potential of AI (Generative AI) in Business: Learnings and Insights
An Efficient Approach to Extract Singular Points for Fingerprint Recognition
1. An Efficient Approach to Extract
Singular Points for Fingerprint Recognition
Supervised By: Dr. Muhammad Sheikh Sadi
Associate Professor, Department of Computer Science and Engineering,
Khulna University of Engineering & Technology
Contact: sheikhsadi@gmail.com
Submitted By: MD. Mesbah Uddin Khan
Level-4, Term-2, Department of Computer Science and Engineering,
Khulna University of Engineering & Technology
Contact: mesbahuk@gmail.com
Dated: June 10, 2012
2. Problem Statement
• Over the years many approaches have been
proposed for developing fingerprint
recognition systems. But some of them give
inaccurate results due to low-quality images
or have high time cost. We will focus on
singular points extraction from low quality
image and then matching fingerprint within
low time cost.
3. Things we need to know
• Fingerprints
• Singular Points
• Fingerprint Recognition
4. Fingerprint (1/2)
•The fingerprint is a duplicate of a
fingertip epidermis.
•When a person touches a smooth
surface, the fingertip epidermis
characteristic transferred to the
surface.
•The pattern of the ridges and valleys
on the human fingertips forms the
fingerprint images.
5. Fingerprint (2/2)
• Fingerprints have remained a valuable
means of identification of an individual
because:
1. they are totally unique to the
individual
2. they never change (Immutability)
7. Ridge patterns
All fingerprints divided into 3 classes
▫ Loops
▫ Whorls
▫ Arches
8. Fingerprint Features
Two types of features
1. Local Features
▫ Ridge Ending
▫ Bifurcation
2. Global Features
▫ Core
▫ Delta
9. Singular Points
a special pattern of ridge and valleys
formed by global features like core
and delta are called singularities or
Singular Points (SP).
▫ A core is defined as the top most
point on the inner most ridge
▫ A delta defined as the center point
where three different directions
flows meet.
10. State of the Art
• Poincaré Index is the most commonly used
method for locating the singular points
• Merits
▫ easy to understand and implement
• Demerits
▫ it may lead to false detection in noisy images
11. State of the Art
• Intersection-Based Method
• proposed by Ramo et al., singular points are
taken as the intersections of transition lines
• Demerits
• As they are intersected, these intersect
points may not give accurate result always.
• This method needs high quality images
12. State of the Art
• Singular candidate method
• uses both the local and global features
• introducing singular candidate models that
indicate the positions where the probability
of the existence of singular points is high
• Demerits
• sometimes gives false candidate region
• noisy images may also be extracted to find
singular points
13. Proposed Methodology
• The proposed methodology is composed of
two main phases:
1. Singular point Extraction
2. Fingerprint Recognition.
14. Proposed Methodology
• The proposed methodology is composed of
two main phases:
1. Singular point Extraction
a) Image Filtering
b) Directional Image Extraction and detecting
DF’s angle
c) Extracting Singular points
2. Fingerprint Recognition.
15. Proposed Methodology
• The proposed methodology is composed of
two main phases:
1. Singular point Extraction
a) Image Filtering
b) Directional Image Extraction and detecting
DF’s angle
c) Extracting Singular points
2. Fingerprint Recognition.
Relative distance, variance and standard
deviation calculation for multiple singular points
16. Image Filtering
Step 1: Take an input image of defined WIDTH and HEIGHT.
Step 2: For all the pixels in the image Do the following.
a. Calculate each pixel’s RGB values
b. If R=O, G=O & B=O Then
Put a BLACK pixel i.e RGB(0,0,0) to the pixel.
Else
Put a WHITE pixel i.e RGB(255,255,255) to that pixel.
Loop.
24. Calculations
We know,
• Variance,
• Standard Deviation,
Where, x denotes the distances
Using these relations we can calculate relative
distance, variance and standard deviation for
multiple singular points
25. Experimental Setup
A short list of tools and libraries used for this
experimental setup are given below:
• OS : Microsoft Windows 7 Prof. Edition
• IDE : Microsoft Visual Studio 2010
• Framework: .NET Framework 4.0
• Language : C#
• Library : AForge.Imaging
• Database : FVC 2004 (DB4)
27. Experimental Results 1/2
The proposed method is applied on 75 fingerprint
images selected from FCV 2004 database.
• It detects 65 true and 4 false core points out of 70.
▫ accuracy rate 92.85%
• It detects 36 true and 3 false delta points out of 39.
▫ accuracy rate 92.31%
Singular Points Total Missed False
Core 70 5 4
Delta 39 3 3
28. Experimental Results 2/2
The proposed approach is applied in FCV2004(DB4)
fingerprint image database for recognizing
fingerprints.
• It recognized and matched fingerprints in of 24
runs out of 26 runs.
• The overall accuracy rate of fingerprint
recognition is found 92.31%
33. Concluding Remarks
• This thesis work proposes and implements a
technique for detecting fingerprints using Singular
Points for both high and low resolution image. It
also helps Recognizing them in low cost and less
time overhead.
• This thesis works with FCV 2004 database images,
which were much noisy. For better results
Fingerprint Scanners can be used.
34. Future plan
• This thesis work supposes that all the fingerprint
images are in straight orientation. So while a
fingerprint is rotated than sometimes it fails to
recognize it.
• All of these works can undergo further study for
better results