Post on 11-Jan-2016
description
People involved:People involved:- Li Fang (Lecturer) Li Fang (Lecturer) - Maylor Karhang Leung (Assoc Prof) Maylor Karhang Leung (Assoc Prof) - Kean Fatt Choon (Final Year Project student)Kean Fatt Choon (Final Year Project student)
Palmprint ClassificationPalmprint Classification
TaskTask
Create a hierarchical system to improve the speed of palmprint recognition
ContentsContents
Victor - Victor - Introduction, Research, conventional process
Tejas - Tejas - Algorithm, explanation of various categories
IntroductionIntroductionWhat is palmprint recognition?What is palmprint recognition?
Form of computer-aided personal recognition
Capturing images of palmprint and matching it with the database
Use for security purposes in many countries
DefinitionsDefinitionsIntroduction to Introduction to
principal linesprincipal lines
Life Line
Head Line
Heart Line
RationaleRationaleWhy palmprint?Why palmprint?
Widely used by many security agencies.
Cost effective
Non-intrusive
Possible to build highly accurate biometric system
RationaleRationaleWhy others methods such as iris and Why others methods such as iris and
fingerprint are not highly effectivefingerprint are not highly effective ?
Iris input devices are expensive.
Iris is intrusive
Fingerprint require high definition capturing devices.
Some may be finger deficient
Contains 1000 imagesPalmprint Capture
Input
Database
Result
Output
BEGIN
Match with user’s registered palm print in the database?
END
FalseTrue
LimitationsLimitations Image captured has to be matched with
every single image in database
Time consuming
Too high computational complexity to be applicable
Aims & ExpectationsAims & Expectations Our aim is to speed up this process by adding in 2
extra filters before the palm print is matched
We expect to increase the speed of the recognition which is one of the most deterring limitation
SurveySurvey Conducted a survey among people
living in Singapore• Gender • Age• Nationality
Survey can be used in our study and design of algorithm which will suit the residents here.
Survey ResultSurvey ResultFrom our survey,
The population palms can be classified into 6 categories (elaborated in the later slide)
Majority of the population lies in one category.
However, significant amount of the population still falls under the other categories
Studies have shown...Studies have shown... According to the algorithm proposed According to the algorithm proposed
on the research paperon the research paper
The algorithm proposed categorizes the palmprints into 6 categories
Palm Categories
cat 1cat 2cat 3cat 4cat 5 cat 6
Cat 1
Cat 5Cat 4
Cat 3Cat 2
Cat 6
Palm Categories
ResultResultAlgorithm proposed by the research Algorithm proposed by the research
paperpaper
cat 1
cat 2
cat 3
cat 4
cat 5
cat 6
Category 5Category 5
New AlgorithmNew AlgorithmWhy a new algorithm is required?Why a new algorithm is required?
78% of the people lie in the 5th category
Based on the current system, the input image has to be matched with every image in the database before the result is obtained
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
FlowchartFlowchart
In Cat. 1
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
FlowchartFlowchart
In Cat. 1NOT CAT. 5
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
In Cat. 1NOT CAT. 5
Not Cat5
Compare
Result
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
In Cat. 5
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
In Cat. 5True
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
In Cat. 5E.g cat. A
FlowchartFlowchart
Image matching with the images in same category in data base
Categories with the new algorithm
Result
Categories with the initial algorithm
Belong to Category 5?
N
Input Palmprint
Y
In Cat. 5Cat A.
Compare
Cat A
Result
New ProcessNew Process
Contains 1000 Images
Palmprint Capture
Input
Database
Result
Cat A Cat B Cat C
Cat D Cat E Not Cat5
New AlgorithmNew AlgorithmStep 1Step 1
The first line connected from the end of the little finger to the intersection of the life line and head line (green line)
The second line is connected from the end of life line to intersection of life and head line (red line)
The third line is connected from point of intersection green line and heart line to midpoint of red line (purple line)
New AlgorithmNew AlgorithmStep 2Step 2
Draw a triangle inside the triangle by connecting the mid points of the each line
Divide the two triangle into 4 parts as shown
New AlgorithmNew AlgorithmStep 3Step 3
Draw a line from end of heart line to end of life line
Draw a line from beginning of heart line to the intersection of life line and head line
The location of the point of intersection of these 2 lines can then be used to categorize the palm
ImplementationImplementation
Category A Category B
Category C
Category D Category E
ResultResult
We tried this algorithm on 100 subjects The pie chart above shows the percentage of each
category It can be concluded that algorithm proposed is
effective
Cat A
Cat B
Cat C
Cat D
Cat E
17.6%
22.3%
18.3%
23.1%
18.7%
SummarySummary Our research showed that process of Our research showed that process of
palmprint recognition is inefficient and can palmprint recognition is inefficient and can be improvedbe improved
Our survey analysis revealed that most Our survey analysis revealed that most people lie in one particular categorypeople lie in one particular category
Proposed a robust algorithm via study of Proposed a robust algorithm via study of characteristics of principal lines to reinforce characteristics of principal lines to reinforce the method of palmprint classificationthe method of palmprint classification
Tried out the proposed algorithm on 100 Tried out the proposed algorithm on 100 subjects to investigate its effectivenesssubjects to investigate its effectiveness
The EndThe End