Docstoc

ear-biometrics-090915041404-phpapp01

Document Sample
ear-biometrics-090915041404-phpapp01 Powered By Docstoc
					                                        Institute of Telecommunication
                                             Image Processing Group


      Ear Biometrics for Human
Identification Based on Image Analysis


             Michal Choras
              Image Processing Group
           Institute of Telecommunication
               ATR Bydgoszcz, Poland




           Presentation for ELCVIA Journal
                                                                    Institute of Telecommunication
                                                                               Image Processing Group



       INTRODUCTION TO HUMAN IDENTIFICATION
                                                          • Disadvantages of the
• Traditional methods:                                      traditional methods:
      PIN’s                                                     hard to remember
      Logins & Passwords                                        easy to loose
      Identification Cards                                      lack of security
      Specific Keys                                                            cards and keys are
                                                                                 often stolen
                                                                                passwords can be
                                                                                 cracked
        Identification by something
       that people know or possess.                              invasiveness



                       Michał Choraś-Ear Biometrics for Human Identification
                                                                 Institute of Telecommunication
                                                                            Image Processing Group

             INTRODUCTION TO BIOMETRICS
• Definition: automatic identification of a living person based on
  physiological or behavioural characteristics.
            Identification by who people are!
• All the biometrics methods can be divided into:




                    Michał Choraś-Ear Biometrics for Human Identification
                                                            Institute of Telecommunication
                                                                       Image Processing Group

        INTRODUCTION TO BIOMETRICS
                                                PHYSIOLOGICAL
    BEHAVIOURAL                                Hand:
                                                      hand geometry
Most popular methods:
                                                      hand veins
                                                       geometry
   voice identification                              fingerprints
   signature dynamics                                palmprints
   keystroke dynamics                         Head:
                                                      eye
   motion recognition                                      iris
                                                            retina
                                                      face recognition
                                                      ear
               Michał Choraś-Ear Biometrics for Human Identification
                                                     Institute of Telecommunication
                                                                Image Processing Group

GENERAL MOTIVATION FOR EAR BIOMETRICS


     • WHERE DO WE HEAD ?
                                           passive
                                       physiological
                                         biometrics



       FACE AND EAR BIOMETRICS
         MIGHT BE THE ANSWER


        Michał Choraś-Ear Biometrics for Human Identification
                                                                   Institute of Telecommunication
                                                                              Image Processing Group

            FACE BIOMETRICS – GENERAL OVERVIEW

• Passive physiological method.
• Natural – humans recognize people by looking at their faces.
• Fast development of new algorithms.
• Still many unsolved problems including compensation of illumination
  changes and pose invariance.
• Some popular methods:
       •   2D geometry,
       •   3D models,
       •   PCA, ICA, LDA,
       •   Gabor Wavelets,
       •   Hidden Markov Models.


                      Michał Choraś-Ear Biometrics for Human Identification
                                                                 Institute of Telecommunication
                                                                            Image Processing Group

                            EAR BIOMETRICS
• Human ears have been used as major feature in the forensic science
  for many years.
• Earprints found on the crime scene have been used as a proof in over
  few hundreds cases in the Netherlands and the United States.
• Human ear contains large amount of specific and unique features that
  allows for human identification.

• Ear images can be easily taken from a distance and without
  knowledge of the examined person.
• Therefore suitable for security, surveillance, access control and
  monitoring applications.


                    Michał Choraś-Ear Biometrics for Human Identification
                                                                 Institute of Telecommunication
                                                                            Image Processing Group

             PASSIVE BIOMETRICS: EAR vs. FACE
• Ear does not change during human life, and face changes more
  significantly with age than any other part of human body.
   – cosmetics, facial hair and hair styling, emotions express different
      states of mind like sadness, happiness, fear or surprise.
• Colour distribution is more uniform in ear than in human face, iris or
  retina.
   – not much information is lost while working with the greyscale or
      binarized images.
• Ear is also smaller than face, which means that it is possible to work
  faster and more efficiently with the images with the lower resolution.
• Ear images cannot be disturbed by glasses, beard nor make-up.
  However, occlusion by hair or earrings is possible.


                    Michał Choraś-Ear Biometrics for Human Identification
                                                       Institute of Telecommunication
                                                                  Image Processing Group

SAMPLE EAR IMAGES FROM OUR DATABASE
             Ears differ „at a first glance”.
We lack in vocabulary - humans just don’t look at ears.




                    „easy ear images”


          Michał Choraś-Ear Biometrics for Human Identification
                                                     Institute of Telecommunication
                                                                Image Processing Group

SAMPLE EAR IMAGES FROM OUR DATABASE


                   „difficult ear images”



   Removing hair for access control is simple and
            takes just single seconds.




        Michał Choraś-Ear Biometrics for Human Identification
                                                   Institute of Telecommunication
                                                              Image Processing Group

EAR BIOMETRICS – OBVIOUS APPROACH

                The method based on
                geometrical distances.




                      How to find
                    specific points?


      Michał Choraś-Ear Biometrics for Human Identification
                                                                  Institute of Telecommunication
                                                                             Image Processing Group

           IANNARELLI’S MANUAL MEASUREMENTS

• The first, manual method, used by Iannarelli in the research in which
  he examined over 10000 ears and proved their uniqueness, was based
  on measuring the distances between specific points of the ear.

• Iannarelli proved that even twin’s ears are different.

• The major problem in ear identification systems is discovering
  automated method to extract those specific, key points.




                     Michał Choraś-Ear Biometrics for Human Identification
                                                                   Institute of Telecommunication
                                                                              Image Processing Group

              EAR BIOMETRICS – KNOWN METHODS

• Neighborhood graphs based on Voronoi diagrams.
Burge M., Burger W., Ear Recognition, in Biometrics: Personal Identification in
   Networked Society (eds. Jain A.K., Bolle R., Pankanti S.), 273-286, Kluwer
   Academic Publishing, 1998.
Burge M., Burger W., Ear Biometrics for Machine Vision, Proc. Of 21st Workshop of
   the Austrian Association for Pattern Recognition, Hallstatt, Austria, 1997.
Burge M., Burger W., Ear Biometrics in Computer Vision, IEEE ICPR 2000.




                      Michał Choraś-Ear Biometrics for Human Identification
                                                                 Institute of Telecommunication
                                                                            Image Processing Group

            EAR BIOMETRICS – KNOWN METHODS

• Ear Biometrics based on Force Field Transformation
Hurley D.J., Nixon M.S., Carter J.N., Automatic Ear Recognition by
  Force Field Transformations, IEE Colloquium on Biometrics, 2000.
Hurley D.J., Nixon M.S., Carter J.N., Force Field Energy Functionals for
  Image Feature Extraction, Image and Vision Computing Journal, vol.
  20, no. 5-6, 311-318, 2002.




                    Michał Choraś-Ear Biometrics for Human Identification
                                                                  Institute of Telecommunication
                                                                             Image Processing Group

             EAR BIOMETRICS – KNOWN METHODS

• Ear Biometrics based on Force Field Transformation
Application of force field transformation in order to find energy lines,
  wells and channels as ear features.




                     Michał Choraś-Ear Biometrics for Human Identification
                                                                    Institute of Telecommunication
                                                                               Image Processing Group

               EAR BIOMETRICS – KNOWN METHODS

 Ear Biometrics based on PCA and ‘eigenears’
Chang K., Victor B., Bowyer K.W., Sarkar S., Comparison and Combination of Ear and
   Face Images for Biometric Recognition, 2003.
Victor B., Bowyer K.W., Sarkar S., An Evaluation of Face and Ear Biometrics, Proc. of
   Intl. Conf. on Pattern Recognition, I: 429-432, 2002.
Chang K., Victor B., Bowyer K.W., Sarkar S., Comparison and Combination of Ear and
   Face Images in Appereance-Based Biometrics, IEEE Trans. on PAMI, vol. 25, no.
   9, 2003.

 Ear Biometrics based on compression networks
Moreno B., Sanchez A., Velez J.F., On the Use of Outer Ear Images for Personal
  Identification in Security Applications, IEEE 1999.



                       Michał Choraś-Ear Biometrics for Human Identification
                                                                    Institute of Telecommunication
                                                                               Image Processing Group

                EAR BIOMETRICS – OUR APPROACH

• Ear Biometrics Based on Geometrical Feature
  Extraction
Choras Michal, Feature Extraction Based on Contour Processing in Ear Biometrics,
  IEEE Workshop on Multimedia Communications and Services, MCS’04, 15-19,
  Cracow, 2004.
Choras Michal, Human Ear Identification Based on Image Anlysis, in L. Rutkowski et
  al. (Eds): Artificial Intelligence and Soft Computing, ICAISC 2004, Springer-Verlag
  LNAI 3070, 688-693, 2004.
Choras Michal, Ear Biometrics Based on Geometrical Method of Feature Extraction, in
  F.J Perales and B.A. Draper (Eds.): Articulated Motion and Deformable Objects,
  AMDO 2004, Springer-Verlag LNCS 3179, 51-61, 2004.



                       Michał Choraś-Ear Biometrics for Human Identification
                                                                Institute of Telecommunication
                                                                           Image Processing Group

            GEOMETRICAL FEATURE EXTRACTION

• General Overview:
     •   Contour Detection, Normalization
     •   Centroid Calculation
     •   1st Algorithm Based on Concentric Circles
     •   2nd Algorithm Based on Contour Tracing
     •   Feature Vectors Comparison and Classification




                   Michał Choraś-Ear Biometrics for Human Identification
                                                                Institute of Telecommunication
                                                                           Image Processing Group

            CONCLUSIONS & WORK-IN-PROGRESS
• Aim: Developement of the automatic algorithm based on geometrical
  features for ear identification
• So far: Algorithm calculating properties of concentic circles
  originated in the ear contour image centriod
• So far: Algoritm based on contour tracing and extracting of the
  characteristic points
• Results: Good for easy ear images.
• Remarks: Heavily dependent on contour detection.
       Now additional segmentation is used to avoid hair, glasses and
       earrings contours.
       New algorithm of selecting only 8-10 longest contours is
       proposed.

                   Michał Choraś-Ear Biometrics for Human Identification
                                                                 Institute of Telecommunication
                                                                            Image Processing Group

            CONCLUSIONS & WORK-IN-PROGRESS
• Work in progress:
         – Algorithm calculating standard geometrical curve-features
           applied to 10 longest ear contours,
         – New algorithm calculating ‘triangle ratio’ of the longest
           contour,
         – Classification to left and right ears based on longest contour
           direction,
         – New algorithm calculating ‘modified shape ratios’ of the 10
           longest contours,
         – Further developement of ear database – 20 views for a
           person (5 orientations, 2 scales, 2 illuminations).

                    Michał Choraś-Ear Biometrics for Human Identification

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:3/2/2012
language:English
pages:20