Hi there, I'm looking around at different fingerprinting algorithms, but I'm
not familiar with .net so I could easily miss things in your code. I'm
curious, is this a straight-forward bozorth3 implementation? And if not, how
does it differ?
Both SourceAFIS and Bozorth3 use the same concept of compatible edges
between minutiae, but the underlying implementation is totally different which
gives both algorithms different performance and accuracy characteristics.
Extractors use many common concepts, but they differ significantly in others.
Most notably, SourceAFIS constructs fingerprint skeleton and performs error
correction on the skeleton while Bozorth3 extractor (MINDTCT).goes straight
from binarized image to minutiae.
I'm working on the scoring algorithm for fingerprint matching as a part of my
degree. Could you please explain to me how sourceAFIS calculates the final
score to decide whether the probe and the candidate fingerprints are a match?
Is there an intermediate step, where the software has a set of matching and
non-matching minutia points, using which it then generates a final score?
SourceAFIS has complicated multi-factor scoring function. I don't remember all
the details. You would do best to take a look into the sources. I am sure
there's intermediate minutia pairing information.
I am doing research in finger print recognition methods. Can you suggest the titles for the alogirthms used in AFIS to know more about that algorithms. Do you have any patents for those algorithms in AFIS.
I don't hold any patents. SourceAFIS algorithms are quite unique, but they fall into some broad categories. Extractor uses smoothing, binarization, and thinning. Matcher is a global minutia matcher that builds minutia tree similar to the one Bozorth3 builds.
I am going through the sources but am finding it difficult to pinpoint to this
part of the code -- once the matching minutiae pairs have been identified (and
the non-matching ones) using any number of algorithms, how is the final score
generated in Afis.Verify?
It would be of great help even if you would point me to the source file for
this, if not the algorithm itself.
Check out classes MatchAnalysis and MatchScoring.
Do you have any patent rights for the methods used in AFIS. Want to know the titles of the algorithms
See above for my reply to your previous question.
SourceAFIS works very well.
I am interest in a further studying at it.Can you tell me how multi_stage matching works? Or give me some paper about it ?
And I wonder how the verify score comes out?
Waiting for your reply.
multi-stage matching simply means applying fast inaccurate algorithm first and slow accurate algorithm second to combine benefits of both. SourceAFIS score function is rather complicated. You will have to take a look into the source code to find out more about score calculation.
Can you tell me what do MatchAnalysis.pairCount , correctTypeCount ,supportedCount , pairFraction, and edgeCount mean in the Java sourcecode?
from my memory:
pair count = number of paired minutiae
correct type count = like above, but restricted to minutiae of matching type (ending, bifurcation)
supported count = minutiae that have more than one edge leading to them
edge count = edges connecting paired minutiae
pair fraction = pair count / minutia count
Thank you for your reply
But I still wonder which fingerprint feature is used in SourceAFIS？
Please ignored my last two questions
1 Most algorithm use ridge ending and ridge bifuraction as the minutiae，the same SourceAFIS？Or more than these two?
2 pair count= number of paired minutiae ,does paired minutiae means one from probe ，another from candidate？
3 what does edge means?Is it a EdgeShape which measured by length,referenceAngle and neighborAngle?By the way ,what do referenceAngle and neighborAngle mean?
I mean used in the extraction
what do you mean by feature? The only features of the fingerprint used are minutiae. Ridges are not preserved in the template.
I found in MinutiaType 0 stands for ending ，1 stands for bifuration，and what for 2？
The feature mean macro feature，such as arch ，tented arch，whorl ……Dose them used during the extraction ？
And does SourceAFIS classify the fingerprint before the matching to improve the matching speed？
Another question for how SourceAFIS record a minutiae‘s location？Where the （0,0）on a fingerprint image？
Thank you for your patient reply
I just wanted to ask you if you use any kind of segmentation algorith in your application.
Hi Pera Peric,
SourceAFIS cannot segment multi-finger images, but it does segment single-finger images into fingerprint area and background area. SourceAFIS does this primarily by clipping off low-contrast areas.
Log in to post a comment.
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.