survey on effectiveness of using fingerprint images in forensic research

Survey on Effectiveness of using
Fingerprint Images in Forensic
research
Third Semester M. Tech. Seminar II (10EC 7301)
by
ANJU NARAYANAN
Roll No. M174102
M. Tech. in Signal Processing and Embedded Systems
Guided by
Dr. SAJITH K
Department of Electronics & Communication Engineering
Govt. College of Engineering Kannur
on
10th September 2018

Outline
 Objective
 Fingerprints
 Classification of fingerprint
 Different identification of fingerprint
 Fingerprint sensing
 Literature Review
 Survey
 Conclusion
 References
Survey on Effectiveness of using Fingerprint Images in
Forensic research

Objective
Forensic research
 Forensic science plays a vital role in the criminal
justice system.
 The recovery of fingerprints from a crime scene is
an important method of forensic science.

Fingerprints
 Fingerprints are impressions created by ridges on skin.
 Purpose of these ridges is to give grasp and to avoid
slippage.
 All fingerprint are unique in nature.
 Fingerprints can solve crimes.
Forensic research

Fingerprints(contd…)
 Ridges contain sweat pores, allow sweat and oil to exit
from glands.
 Fingerprints are left, by transfer of oils or amino acids to
a surface.
 Ridges form under pregnancy and maintain their pattern
throughout life.
 As you grow, pattern gets larger, but does not change.
Forensic research

Classification of Fingerprints: Patterns
 Arch: Ridges enter from one side, rise in centre forming
an arc, then exit other side of finger.
 Loop: Ridges enter from one side of finger, form a
curve, then exit on that same side.
 Whorl: Ridges form circularly around a central point on
finger.
Forensic research

Classification of Fingerprints: At crime
scene
 Latent prints: Commonly Invisible image of prints.
 Visible prints: Visible to naked eye, formed when a
visible contaminants are present.
 Plastic prints: Fingers comes in contact with soft
surface such as soap, butter, wax etc.
Forensic research
Visible prints Plastic prints

Different identification on fingerprint
 Crossover: Two ridges cross
each other.
 Core: centre
 Bifurcation: Ridge separates
 Ridge ending: End point.
 Island: Small ridge b/w two
spaces.
 Delta: Space between ridges.
 Pore: Human pore
Survey on Effectiveness of using
Fingerprint Images in Forensic
research

Different identification(contd…)
Forensic research

Fingerprint Sensing
 Based on mode of acquisition, fingerprint image is
classified as
 Off line image
 Live-scan image
 Number of live-scan sensing mechanisms that can detect
ridges and valleys present in fingertip.
 Examples are
 Optical
 Capacitive
 Pressure-based
Forensic research
Optical sensor Capacitive sensor

Literature Review
Forensic research
Author, year Title Publication
Anil K. Jain, Jianjiang Feng,
Karthik Nandakumar
Fingerprint matching Published by the IEEE
Computer Society-2010
A. S. Falohun, O. D. Fenwa,
F. A. Ajala
2.A Fingerprint-based Age
and Gender Detector System
using Fingerprint Pattern
Analysis
International Journal of
Computer Applications-2016
S. F. Abdullah, A. F. N. A.
Rahman and Z. A. Abas
4. Classification of
gender by using fingerprint
ridge density in northern
part of malaysia
ARPN Journal of Engineering
and Applied Sciences-2016

Literature Review(contd…)
Forensic research
Author, year Title Publication
Arun K.S, Sarath K.S,
2013.
A Machine Learning Approach
for Fingerprint Based
Gender Identification
IEEE Recent Advances in
Intelligent Computational
Systems.
Suchita Tarare, Akhil Anjikar,
Hemant Turkar
Fingerprint Based Gender
Classification Using DWT
Transform
IEEE International Conference
on Computing Communication
Control and Automation.

1. Fingerprint matching
 Skin on our palms consists of ridges and valleys.
 Friction ridge patterns are influenced by
 Genetic factors.
 Random physical stresses and tensions during foetal development.
 Due to concerns on security government and commercial
organizations proposed fingerprint-based recognition
systems.
 Fingerprint recognition system includes verification
(1:1 match) and identification(1:N match).
Forensic research

Fingerprint matching(contd…)
 AUTOMATED FINGERPRINT RECOGNITION
 Enrolment phase and Identification/Authentication phase.
Forensic research

Fingerprint matching(contd…)
 Fingerprint Sensing
 Feature extraction:
 Level 1 , Level 2 , Level 3 features
 Matching
Forensic research

2. Gender Detector System using
Fingerprint Pattern Analysis
 Human gender detection using fingerprint analysis
trained with Back Propagation Neural Network.
 METHODOLOGY
 Fingerprint acquisition/Data collection
 Enhancement
 Performance depends on quality of fingerprint images.
 Histogram equalization
Forensic research

Gender Detection using fingerprint
pattern(contd…)
 Image Binarization
 Ridges in black and valleys are white in colour.
 Local adaptive Binarization method is performed.
 Segmentation
 Image area without effective ridges and valleys are discarded.
 Fingerprint Ridge Thinning
 Eliminate redundant pixels of ridges till ridges are one pixel wide.
 Skeltonisation method is used.
Forensic research

Gender Detection using fingerprint
pattern(contd…)
 Minutia Marking
 Crossing Number (CN) concept.
 Training with Back Propagation Neural Network.
 Total 280 fingerprint samples with various gender was
collected. 140 samples used for training system’s
Database; 70 males and 70 females.
 Result: 80% classification accuracy for females and
72.86 % for males.
Forensic research

3.Gender classification using fingerprint
ridge density in northern part of malaysia
 Ridge of fingerprint from two topological areas, radial
and ulnar can be counted and mean can be calculated.
 METHODOLOGY
 Collection of fingerprint images
 Plain technique is adopted for collection of fingerprint images.
 Material use in this data collection is Unicorn thumb print pad,
ruler, pen, measuring tape and data personal form.
 pre-processing
 Original fingerprint image is turned into the grayscale.
 Binarization
Forensic research

Gender classification using fingerprint
ridge density(contd…)
 Use method of Acree to calculate ridge density.
 A square box measured by 5 x 5 mm is placed at upper
portion of radial and ulnar border in fingerprint image.
 Value of ridges density represented in number of ridges/
25mm² square areas is calculated by using the formula:
Ridge density=
𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑖𝑑𝑔𝑒 𝑖𝑛 𝑠𝑞𝑢𝑎𝑟𝑒
25𝑚𝑚2
 Result: Male respondents have lower number of ridge
density than female respondents.
Forensic research

4. Machine Learning Approach for
Gender Identification
 Use machine learning approach.
 Feature vector consisting of ridge thickness to valley
thickness ratio (RTVTR) and ridge density values.
 RTVTR is average ratio between ridge thickness and
valley thickness of a fingerprint.
 Females have higher RTVTR value compared to male.
 150 male and 125 female fingerprint images.
Forensic research

Machine Learning Approach(contd…)
 Fingerprint image is divided into 32x32 non overlapping
blocks.
 Local ridge orientation within each block is calculated.
 Projection profile of valleys and ridges in each block is
calculated.
 Projection profile was binarized using 1D optimal
thresholding.
 Resultant binary profile represents the ridges and valleys in
this block.
 RTVTR is calculated for each block.
Forensic research

Machine Learning Approach(contd…)
.
Forensic research
FINGERPRINT ACQUISITION IMAGE NORMALIZATION
FINDING LOCAL RIDGE
ORIENTATION
IMAGE BINARIZATION
FINDING PROJECTION PROFILE CALCULATION OF RIDGE
DENSITY
SVM CLASSIFIER
MALE & FEMALE OUTPUT
Methodology

5. Fingerprint Based Gender
Classification Using DWT Transform
 Pre-processing of all dataset images.
 Calculate feature vector of training images using discrete
wavelet transform.
 Classification of testing fingerprint using KNN classifier.
 Dataset of 100 male and 100 female fingerprints.
Forensic research

Gender Classification Using DWT
Transform(contd…)
 Over all process of gender classification
Forensic research
Male Female
Training Images
Pre-processed Images
Decomposed Image
Feature database
Testing Image
Pre-processed Images
Decomposed Image
Feature vector
Knn classifier
preprocessing
6 Level DWT
Feature calculation Feature calculation
6 Level DWT
preprocessing

Transform(contd…)
 Pre-processing: Image resizing, Binarization
 DWT Based Feature Extraction:
 Dwt uses wavelet as its basis function.
 Decomposition of images into different frequency bands helps to
isolate different frequency components.
 2-D wavelet decomposition results in 4 decomposed sub-band
images :low–low (LL), low–high (LH), high–low (HL), and high–
high (HH).
 These sub-bands help to study different image details.
 For k level DWT, there are (3*k) + 1 sub-bands available.
Forensic research

Transform(contd…)
 Energy of all these sub-band coefficients is used as feature vector
called sub-band energy vector (E).
 Calculating features of all training images and stored in database.
 Energy of each sub band is calculated by 𝐸 𝑘 =
1
𝑀𝑁 𝑖=1
𝑁
𝐽=1
𝑀
𝑥 𝑘 ⅈ, 𝑗
 k is specific sub-band.
 M and N is the width and height of particular sub-band.
 𝑥 𝑘 ⅈ, 𝑗 represents the specific pixel of particular sub-band.
 K nearest neighbor (knn) classifier is used as a classifier.
 Uses Euclidean distance measure for classifying testing fingerprint as
male or female fingerprint.
Forensic research

Conclusion
 As fingerprints are unique for individuals in universe, it
gives a unique identification.
 Most of traditional methods used in identification of
gender gave the satisfactory results but an efficient
attempt is needed to give effective results with higher
accuracy.
 Image clarity, Frequency domain analysis, application of
neural network will have important role to increase
efficiency, still there is scope to improve results.
Forensic research

References
Forensic research
[1] Anil K. Jain, Arun Ross, Sharath Pankanti ” Biometrics: A Tool
for Information Security.”, IEEE transactions on information
forensics and security, pp. 753-764, May 2010.
[2] Gualberto, Gabriel ,” Fingerprint Recognition “, IEEE
International Conference on Internet Monitoring and
Protection, pp.1212-1215, 2012.
[3] Anil K. Jain, Jian Feng,” A multistage fingerprint recognition
method for payment verification system”, IEEE transactions on
pattern analysis and machine intelligence., pp.641-643, 2011.

References
Forensic research
[4] Arun K.S, Sarath K.S, ” A Machine Learning Approach for
Fingerprint Based Gender Identification”, IEEE Recent Advances in
Intelligent Computational Systems, pp. 110-124, December 2014.
[5] Saptarshi Rudra, Abhisek Roy, Soham Mitra, ” Gender
Classification System from Offline Survey Data Using Neural
Networks “, IEEE 7th Annual Ubiquitous Computing, Electronics &
Mobile Communication Conference ,pp.200-241, December 2016.

Thanks!
Forensic research

3.Fingerprint image normalization.
 Let N (i, j) represent the normalized grey-level value at
pixel (i, j).
 The normalized image is defined as:
 M0 and VAR0 are desired mean and variance respectively.
 M and VAR are mean and variance of image.
Forensic research

3. Finding Local Ridge Orientation
 Divide G into blocks of size WxW (32x32). Let the umber of blocks be
N.
 Compute gradients δx(i,j) and δy(i,j) at each pixel (i,j). Operator
used is Sobel operator.
 Estimate local orientation of each block centered at pixel (i, j) using:
 Where Θ(i, j) is least square estimate of local ridge orientation at block
centered at pixel (i, j).
Forensic research

3. Binarizing Fingerprint Image
 Grayscale image is categorized into only two levels, black and
white (0 and 1).
 Thresholding is used for binarizing image.
 The following algorithm is used to obtain T automatically.
 Select an initial estimate for T0.
 Segment the image using T0. This will produce two groups of pixels: G1
consisting of all pixels with gray level values greater than T0 and G2
consisting of pixels with values ≤ T0.
 Compute average gray level values μ1 and μ2 for pixels in regions G1 and
G2.
 Compute a new threshold value:
 Repeat steps 2 through 4 until difference in T in successive iterations is
smaller than a predefined parameter T0.
Forensic research

SVM CLASSIFIER
Forensic research

survey on effectiveness of using fingerprint images in forensic research

More Related Content

What's hot (19)

Similar to survey on effectiveness of using fingerprint images in forensic research (20)

Recently uploaded (20)

survey on effectiveness of using fingerprint images in forensic research

Editor's Notes