Hand Geometry

1
Hand Geometry

• CSE 717

2
Why Hand Geometry

• Acceptance UnIntrusive
• Ease of Collection Of Data
• Fast capture/processing
• Small template size
• Uses include Access control, time attendance,
  border crossing

3
Feature Extraction

• Image Capture
• Preprocessing
• Measurements
• Optimization of the template Size
• Feature Selection Feature Vector Size

4
Image Acquisition
5
Image Capture

• Flat Bed Scanner 150dpi scanner.
• CCD color camera. Color Photograph in the form of
  a Jpeg format.
• The Lateral view of the Hand can also be captured
  by the mirror placed in the side to measure
  heights.

6
Using Pegs
Raul Sanchez-Reillo Carmen Sanchez-Avila Ana
Gonazalez-Marcos
7
Preprocessing

• Binarize the Image.
• Rotation n Resizing.
• Extract the Contour of the Image.

8
Preprocessing

• 1st step Binarize the Image.
• I (bw) (Ir Ig) Ib
• Resizing n Rotation . Deviation of the hand are
  corrected
• Edge Detection Algorithm eg.Sobel Function

9
Problems with Using Pegs
Alexandra L.N. Wong1 and Pengcheng
Shi2 Deformation of the Shape of the Hand by the
Pegs. Different Placements of the Same Hand
10
Landmark Extraction Hand Alignment

Applying the border-following Algorithm 1
Alexandra L.N. Wong1 and Pengcheng
Shi2 Application of the border-following
Algorithm
11
Feature Extraction
Lengths of four fingers Widths of four fingers
at 2 locations Shapes of the fingertips
Alexandra L.N. Wong1 and Pengcheng Shi2
12
Hierarchical Recognition

• Class I 13 finger lengths and the finger widths
  Gaussian mixture Model is used to classify these
  features
• Andrew W. Moore
• Associate Professor
• School of Computer Science
• Carnegie Mellon University
  www.cs.cmu.edu/awm
• Fingertips - class II

13

• Gaussian Mixture Modeling
• Approach bet Statistical n Neural Networks
• Modeling the patterns with determined number of
  Gaussian Models
• Weighing Coefficient of Gaussian Model
• Mean Covariance Vector are the
  characteristic parameters.

14
GMM contd

• Preset Threshold Value of the GMM Probability
  Estimation.
• Group II features - Euclidian Dist Measure
  bet the sample template
• the given template. Threshold is used to
  reject the templates.(eg. 2 pixels)

15
Results

• Hit Rate Typical Methods of Comparisons
• Euclidean Distance Measure
• Group 1 Group 1 and
  2
• Hit Rate 1 0.8889
• FAR 0.1222 0.022
• Alexandra L.N. Wong1 and Pengcheng Shi2

16
Alexandra L.N. Wong1 and Pengcheng Shi2 GMM
thresholds role in reducing the FRR.
17
Different Comparison Algorithms
18
Work Done by Other Researchers

• Raul Sanchez-Reillo Carmen Sanchez-Avila
• Ana Gonazalez-Marcos have done the development
  of the GMM based comparison Algorithms.
• Alexandra L.N. Wong1 and Pengcheng Shi2 Pegs
  Free Hand based Geometry

19
Observations

• GMM obtain the best results
• the other possible comparison algorithms are
  Euclidean Hamming Distance based ,
• Radial Basis Function RBF Neural Networks.
• GMM based template require much more memory than
  the other comparison based templates.

20
Conclusion Future Research

• Ideal for Medium and Low Security based
  Biometrics.
• Can be used together with other Biometrics Palm
  Prints
• Non Geometrical hand features such as color can
  be used .

21
(No Transcript)