Fingerprint Recognition Checkpoint Slides

1
Fingerprint RecognitionCheckpoint Slides

• Andrew Ackerman
• Professor Ostrovsky

2
Current State

• Already Done
• Research Fingerprint Acquiring
• Research Fingerprint Analyzing
• Research Fingerprint Recognition Algorithms
• Including Minutiae detection and learned based
  templates
• Current State
• Reduced Fingerprint Recognition to Another
  Problem
• Solving for algorithm for problem

3
Fingerprint Recognition

• Taking two fingerprint images and seeing if they
  come from the same finger
• Assumption no two fingers in the world yield the
  same fingerprint and that fingerprints do not
  change in time
• Much research has been done to validate these
  assumptions
• Complex
• Due to a multitude of conditions the same
  fingerprint scanned twice can look very different
• How finger is oriented on scanner
• Condition of finger (wet, dry, scarred, etc)
• No completely accurate method exists
• Many current recognition techniques can tell with
  a degree of certainty if two fingerprint images
  match (i.e. come from same finger)

4
Our Method

• First try
• View the fingerprint edges as a graph structure.
• Bifurcation and ending points would become nodes
  and the edges would be edges
• Abandoned due to complexity involved in matching
  sub graphs of two different graphs
• Currently Explored Method
• Reduce fingerprint recognition into a problem of
  topographical equivalence

5
Reduced Problem

• Topological Equivalence
• Ridges of the fingerprint are just lines that
  when leave the frame of view go off to infinity
• If two fingerprints can be shown to be nearly
  topologically equivalent then they are most
  likely the same fingerprint

6
Topological Equivalence
Not Topologically Equivalent
Topologically Equivalent

• Two images are topologically equivalent if you
  can somehow deform the lines (without crossing
  them over) to go from one image to the other

7
Current Goal

• Come up with an algorithm that can determine
  topological equivalence for relatively simple
  cases, then generalize to reach most of the cases
  required for the fingerprint recognition.
• Mainly dealing with shapes that include
  bifurcations and edges and possibly loops
• Write software that implements the algorithm
• Will use NIST Special Database 4 (fingerprint
  database) to test algorithm

8
Self Assessment

• I feel that I have not done all the work I could
  possibly have accomplished in the fall quarter.
  However, as I noted at the beginning to Professor
  Sahai, I had a heavy course load with courses
  such as CS152B and graduate school applications.
  With these out of the way for next quarter, I
  will have much more time to dedicate to coding
  for CS194.
• That being said, both Professor Ostrovsky and I
  feel we are on track to receive good results by
  the end of the spring quarter. The best case
  scenario being that our algorithm performs faster
  and/or more accurately then current algorithms
  used.

9
Self Assessment

• Our main project focus has changed. The
  cryptology part of the project has been dropped
  for now and the main focus is on fingerprint
  recognition. I feel a good amount of work has
  been done on understanding some of the current
  algorithms for fingerprint recognition. And our
  current focus is to develop and code our own
  algorithm to test them.
• Overall, I feel I have done adequate amount of
  work for CS 194 during the fall quarter. And I
  plan to get a good amount of work done over
  winter break regarding our algorithm and coding
  it.