Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo


Pore remover

  • Johan

    Hi Robert

    Greetings for the new season.

    I am still using sourceafis, and learning every day.

    I could not attach the images on the discussion forum , so I hope these get
    to you.

    I have attached 4 images showing the same finger .

    Two of the images "print1-f7.bmp" and print2-f7.bmp" are bmp images of the
    same finger captured on

    A biometric device.

    When matching them using the SourceAfis matcher I get a score of approx
    21 which is relatively low,

    considering that they only matched 8 pairs ( marked as green squares on
    the other two attached images namely "print1-f7-min.bmp" and
    "print2-f7-min.bmp" ).

    I have also noticed that there are a lot of false minutia points,(
    bifurcations marked as yellow circles) that

    are plotted where in actual fact they are pores.

    One can clearly see this on the other 2 images "print1-f7-min.bmp" and
    "print2-f7-min.bmp" which I have also attached.

    Also there are points which are not marked on both images.

    Is there a way to eliminate the false minutia at all the pores by
    manipulating the source code?

    I had a look at the pore remover class which has a parameter

    "public int MaxArmLength = 41;"

    and tried various values without any success .


    . Johan Luus

    . Solutions Architect (Biometric Systems)

    . Muvoni Technology Group Ltd

    ' +27 (0)12 679 2925

    Fax +27 (0)12 679 2911

    Cell +27 (0)83 441 3260

    Web http://www.muvoni.com/ www.muvoni.com


    Lower Ground (-1) |267 West Avenue |Centurion |0157

    PO Box 7416 |Centurion 0046 |South Africa

    This e-mail and its contents are subject to the Muvoni Technology Group Ltd.
    E-mail legal notice

  • Robert Važan
    Robert Važan

    I've opened the fingerprints in Fingerprint Analysis app and I see 13 matched minutiae. Perhaps you are using some older or customized version of SourceAFIS.

    False minutiae are not a big concern. It is better to have a few extra minutiae than to miss some. I agree with you that SourceAFIS extractor could use some improvements. It's not as easy as tweaking a few parameters though.

    You could run DatabaseAnalyzer process against a database of fingerprints for your fingerprint reader. That would specialize SourceAFIS for your reader and result in slightly higher accuracy. You need a good training database though. That's as far as you can get with just parameter tweaks.

    You can reduce error rates by adding in some redundancy. Either match multiple fingers or request users to retry if there is no match the first time.

    Of course, I have lots of ideas for improving SourceAFIS. Mail me if you would like me to spend a couple of weeks doing R&D. That should result in some perceptible improvements.