Experiments with INCITS-378 template matching
Fingerprint recognition library (.NET port)
Brought to you by:
robert-vazan
Menu
▾
▴
Hi,
I've read the previous discussions of INCITS-378 vs ISO 19794-2:2005 templates, but based on a a paper that claims 19794 is based on INCITS-378, I decided to see if they were close enough for AFIS. We have lots of existing I378 templates that need to be compared against scanned fingerprints, so don't have the option of using an image-only approach. What I found was that the I378 templates are very close indeed to the format produced by AsIsoTemplate; with very minor changes to the I378 template data I can pass it to AsIsoTemplate, and the containing Finger object to Afis.Verify, which does return a score. The resultant match scores are in the 10 to 13 range, but also do not appear to correlate with the data; that is, I can scan the wrong finger, and get just as good a match as when scanning the correct finger. No complaints; this is an experiment, and I thought I'd share the results. if anyone elese has tried something similar I'd be happy to hear about it.
Robert: I also saw your previous comments on not supporting INCITS-378, and why. I would appreciate a chance to discuss this with you further. And, thank you for making this software available; I found your extensive comments in the source very helpful in deciphering the template data encoding.
Well, if scores are the same for matching and non-matching pairs, then the templates are garbled. It could be as simple as big-endian vs low-endian encoding or a small offset in file position. Unfortunately I don't have time to address this at the moment. Maybe someday in the future.
ANSI/INCITS 378 is now supported by the latest SourceAFIS.