7/28/2019 Minutiae Matching

1/15

A minutia matching algorithmin fingerprint verification

2005/10/18

7/28/2019 Minutiae Matching

2/15

Outline

Introduction of minutiae

Aligned minutia matching Conclusion

7/28/2019 Minutiae Matching

3/15

Introduction of minutiae

7/28/2019 Minutiae Matching

4/15

Alignment of minutiae set

( )Tp

M

p

M

p

M

Tppp

yxyxP ),,(,,),,( 111 L= ( )

TQ

N

Q

N

Q

N

TQQQ

yxyxQ ),,(,,),,( 111 L=

PMQN

7/28/2019 Minutiae Matching

5/15

Alignment of minutiae set cont.

= =L

i ii drdRLdistDiff 0 )()(1_ = =

L

i ii araRLangDiff 0 )()(1_

IF type is not the sameor Diff_ang> Tdor Diff_dist> To rotate[i][j] = 400ELSE rotate[i][j] =dir_temp- dir_in

7/28/2019 Minutiae Matching

6/15

cont.

+

+

=

ri

ri

ri

ri

ri

i

i

i

jirotatexx

yy

yyxx

er

]][[tan

)()(

1

22

Trrr yx ),,(

7/28/2019 Minutiae Matching

7/15

Aligned minutia matching

1) For i (1 ~M) and j (1 ~N),. If rotate[i][j]=400, then repeat this stepand choose another Pi and Qj, else go to step 2). If all possibleminutia pairs have been considered, go to step 4).

2) Take Pi and Qj as reference minutia. Convert each minutia pointin the template minutia set and the input minutiae set to the polarcoordinate system.

3) Match the resulting strings Pi" and Qj" with the process which willbe introduced below to find the matching score of Pi" and Qj".Record it as m-score[i][j]. Then go to step 1).

4) Find the maximum value of m-score[i][j] and use it as the

matching score of the input and template minutiae set. If thematching score is higher than a threshold value, then the inputimage is considered to come from the same finger as the templateimage, else we would consider these two images as coming fromdifferent fingers.

( )TpMpMpMTpppsi ererP ),,(,,),,( 111 L= ( )TQMQMQMTQQQsj ererQ ),,(,,),,( 111 L=

7/28/2019 Minutiae Matching

8/15

Bounding box

angle_size= angle_high angle_lowradius_size= radius_high radius_low

7/28/2019 Minutiae Matching

9/15

Changeable sized bounding box

Fixed sized bounding box Changeable sized bounding box

7/28/2019 Minutiae Matching

10/15

Changeable sized bounding box

cont.

7/28/2019 Minutiae Matching

11/15

Process matching

1) Decide the size of the bounding box for each minutia in the templateminutia set and set m-score[i] [j] = 0.

2) While 1

7/28/2019 Minutiae Matching

12/15

Process matching cont.

7/28/2019 Minutiae Matching

13/15

Conclusion

7/28/2019 Minutiae Matching

14/15

Reference

Luo Xiping, Tian jie, A Minutia Matching algorithm in Fingerprint Verification,Proceeding of the 15th International Conference on Pattern Recognition,

Vol.4, pp.833-836, 2000 Anil J ain, Lin Hong and Ruud Bolle, On-Line Fingerprint Verification, IEEE

Trans on Pattern Analysis and Machine Intelligence, vol.19, No.4. pp302-313, 1997

http://www.fingerpass.net/

7/28/2019 Minutiae Matching

15/15

The End