Transcript ForSe - Nanyang Technological University

ForSe Overview
Forensics and Security Laboratory (ForSe Lab)
School of Computer Engineering
Nanyang Technological University
Mission of ForSe Lab
 To create a synergistic group dedicated to research in the
application of computational techniques to biometrics,
information security and forensic analysis.
 To perform cutting edge research and train and develop
talents to support Singapore’s efforts in the areas of
Homeland Security and Infocomm Security.
 To make use of strong research base to further enhance the
research contributions from NTU to the international arena
in the areas of forensic and security.
‹#›
Vision of ForSe lab
 To be one of the major research labs/centres for
research and development in the areas of forensics,
biometrics, and security technologies.
 To be a strong research arm between academic and
industry to support R&D activities in forensics and
security for Singapore.
‹#›
Facts and Figures
 Established in late 2005.
 6 active faculty members, 1 research assistant,
1 lab executive, 11 PhD students.
 6 funded projects with total amount over
S$500K.
 Supports approximately 10-15 Final Year
Projects every academic year
‹#›
Research Funding Award

InfoComm Research cluster, NTU

Institute for Infocomm Research (I2R) – Joint
Collaboration Project with I2R

Three Academic Research Fund Tier 1, Ministry of
Education, Singapore

MINDEF-NTU Joint R&D project
‹#›
Active Members
David Cho (Asst Prof)
Director
Maylor Leung
(Assoc Prof)
Sudha Natarajan
Li Fang (Lecturer)
Vinod Prasad
(Asst Prof)
Adams Kong
(Asst Prof)
(Asst Prof)
‹#›
Our Knowledge/Expertise







Pattern Recognition
Machine Learning
Digital Signal Processing
Image Processing
Embedded System
Information Security
Software Engineering
‹#›
Research Focused Areas
Forensics and Security Lab
Forensic Analysis
Forensic analysis of
digital evidence
images
Forensic analysis of
speaker voice
Forensic analysis of
image forgery
Forensic examination
of digital devices
Biometrics and
Security
Hand and Facial
Thermal pattern
analysis
Palmprint, Face, Iris
and Ear recognition
Human action
analysis for video
surveillance
Smart hidden weapon
detection
Information
Security
Digital Content
Protection
Digital Crime Scene
Reconstruction
Traffic Flow
Monitoring and
Modeling
Human Behavior
Analysis
‹#›
Key Contributions – Forensic
Analysis
 Hand Vein Pattern Analysis
 Speaker Identification
 Acoustic Voice Feature
 Image Forgery Detection
 Skin and/or Hair Analysis
‹#›
Key Contributions – Biometrics
Technology
 Facial Thermal pattern analysis
 Palmprint recognition
 Face recognition
 Iris recognition
 Ear recognition
‹#›
Key Contributions – Security
Engineering
 An Embedded Camera System for Vision Based
Surveillance
SmartEye system
Camera
Configurable
Preprocessing
Architecture
Vision-based
surveillance
framework
 Hidden Weapon Detection
 Human behaviour and brain analysis
 EEG Signal Analysis
 Emotion Recognition
‹#›
Future Plans in ForSe Lab
 To extend and build more research activities with the
research areas of the lab to attract external funding.
 To focus our staffs to prepare and submit major research
proposals to several funding agencies, such as, AcRF, A-Star,
DSTA, DSO,…etc.
 To collaborate with other major organizations, such as, I2R,
Singapore Police Force, MHA and also some companies in
security industry, …etc.
 To continue our excellent tradition of publishing our new
discoveries and theories in renowned journals,
conferences,…etc.
‹#›
Collaborators
 International





Prof. Graham Leedham, Dean of School, University of New England, Australia
Prof. M. Kamel, IEEE Fellow, University of Waterloo, Canada
Dr. Noah Caft, MD, PhD, Assistant Professor, UCAL, USA
Prof. D. Zhang, IEEE Fellow, The Hong Kong Polytechnic University, HK
Prof. Tommy Chow, City University of Hong Kong
 Local
 Dr. Li Haizhou, Dr Guan Cuntai and Dr. Vladimir Pervouchine, Institute of
Infocomm Research (I2R)
 Dr. TAY Ming Kiong Michael , Director, Physical Evidence Division, Applied
Sciences Group
 Ms. LIM Chin Chin, Head, Criminalistics Laboratory, Centre for Forensic
Science
 Dr. LOH Tsee Foong, MD, Head and Senior Consultant, KK Women’s and
Children’s Hospital
 Dr. James Wong (Application Architect), PCS Security Pte Ltd
‹#›
Recent Publication






L. Wang, G. Leedham and Siu-Yeung Cho, "Minutiae Feature Analysis for Infrared Hand Vein
Pattern Biometrics", Pattern Recognition (JCR impact factor: 3.279), 41 (3), pp. 920-929,
2008.
Lingyu Wang, Graham Leedham and Siu-Yeung Cho, “A Physiological Vein Pattern Biometric
System”, HKIE Transactions, vol. 15, iss. 4, Dec. 2008 (shortlisted paper for The HKIE
Outstanding Paper Award for Young Engineers/Researchers 2008).
L. Wang, G. Leedham and S.-Y. Cho, "Infrared imaging of hand vein patterns for biometric
purposes", IET Computer Vision (JCR impact factor: 0.667), vol. 1, Iss. 3-4, pp. 113-122,
Dec. 2007.
Siu-Yeung Cho, Lingyu Wang and Wen Jin Ong, “Thermal Imprint Feature Analysis for Face
Recognition”, in IEEE International Symposium on Industrial Electronics 2009, July
2009, Seoul, Korea.
Haishan Zhong, Siu-Yeung (David) Cho, Vladimir Pervouchine, Graham Leedham,
“Combining Novel Acoustic Features using SVM to Detect Speaker Changing Points”,
BIOSIGNALS (1) 2008: 224-227.
N.B. Puhan and N. Sudha, "A novel iris database indexing method using the iris color",
Proceedings of the IEEE International Conference on Industrial Electronics and Applications,
Singapore, June 2008.
‹#›
Thank you!
‹#›
Hand-vein pattern analysis


A vein pattern refers to the vast network of blood vessels underneath the skin of a
certain part of a person’s body
Images captured in an air-conditioned office environment (20-25°C and <50%
humidity)
FIR Image of Back of hand imaged in a normal office environment
– major veins are clearly visible
NIR images of the palms of two hands
NIR images of the back of the hand (left) and the wrist (right)
‹#›
Hand-vein pattern analysis
 We have proposed a system that recognizes the human hand
vein pattern images acquired by both far and near-infrared
camera, which consists of five individual stages
Image
Acquisition
Raw
Images
Image
Enhancement
&
ROI Selection
Finer
Images
Vein Pattern
Segmentation
Skeletonization
Vein
Pattern
Decision
Shape
Match
Template
Data Collection
Vein Pattern Extraction
Database
‹#›
Results
Skeleton and Minutiae Points of the Vein Pattern
Error Rate Curves for Minutiae
Recognition Using the Modified
Hausdorff Distance (EER=7.5% when
the threshold is set to 25)
Return
‹#›
Image Forgery Detection
 With the advent of low-cost and high-resolution digital
cameras, and sophisticated photo editing software, digital
images can be easily manipulated and altered.
 create forgeries, which are indistinguishable by naked eye
(a) Real image; (b) Forged version; (c) Duplicated regions
‹#›
Return
Skin Analysis
Skin marking system
Data collection system
‹#›
Return
Principles of Thermal Facial
Patterns for Biometrics
 The convective heat transfer from the flow of warm arterial blood in
superficial vessels is at a temperature gradient against the cooler
surrounding tissue
 Creating a characteristic thermal imprint on our face
 This thermal pattern provides an alternative feature sets in addition to
those visible features for face recognition
P. Buddharaju,et. al, “Physiology-Based face Recognition in the Thermal Infrared Spectrum”,
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.29 no.4, pp. 613-626, April 2007.
‹#›
Database
 Data Collection
 Thermal face images can be formed by capturing the temperature
profile by the NEC TH9100SL thermal camera
 Grayscale images with a resolution of 320x240 are used.
 Database provided by Equinox Corporation.
 Frontal thermal face dataset
 300 images from 30 different subjects (10 images for each subjects).
‹#›
Face Segmentation
(a) Image after enhancement
(d) Difference Image
(b) Edge detection with small
objects removed
(e) Mask is multiplied with the image
(c) Centre portion of the image is
flood filled
(f) Contrast adjusted
‹#›
Extracting Thermal Minutiae Points
 Use morphological top-hat operation to obtain the critical
edge map
Itop  I  I open ,
I open  (IS )  S
 Then extract the minutiae points by the cross numbering
concept.
8
Ntrans   | Pi 1  Pi | , where P9  P1
i 1
(a) Thermal face region
(b) Critical edge map
(c) Minutiae points
‹#›
Matching the Aligned TMPs
 Using the same MHD measurement
 Achieved 6.7% EER
Return
‹#›
Ear Recognition
 Rationale:
 Earmarks can be used as a
biometric, but a computerized
system for earmarks
identification does not existed.
 The structure of the ear does
not change radically over time,
especially after the first four
months of birth.
‹#›
Ear Recognition
 Current work:
 Build a ear profile database of 38 individuals (will be extended the
number later)
 Implement an automatic ear detection, localization and recognition
system
 A 11% Equal-Error-Rate is achieved.
‹#›
Return
Concealed Weapon Detection
 Objective: to find out the feasibility of software based image processing
techniques in detecting concealed weapons with infrared (IR) thermal
imager without violating the privacy of the people involved.
Visible image
IR image
Fused IR image
NEC Thermo Tracer
‹#›
Concealed Weapon Detection
 On-going works:





Fuzzy clustering of IR images
Advanced image registration methods
Intelligent and decision based image fusion
Robust shape matching
Collaborating with EEE staffs to work with IR and MMW image sensing
Return
‹#›
Human behaviour and brain
analysis – Emotion Recognition
Typical real-time facial expression system
Potential applications:
• Lie detection for forensics
• Crime investigation
• Understanding Criminal Psychology
Return
‹#›