I have a smaller sensor with 160x160 pixels, 508dpi, that I'm trialing with the SourceAFIS algorithm.
I get really poor results (scoring <=30) for good fingerprint samples, when matching.
Is there any parameters or tuning that I can do to increase the matching score, or this is due to the inherent nature of Minutia matchers?
Br,
Tim
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The sensor area is probably too small. You need larger sensor. Storing multiple scans of the same finger and picking the best match could help somewhat.
SourceAFIS algorithm can be improved to either merge multiple fingerprints during user registration or to mine and use features from the image more aggressively. Neither of these enhancements is planned in near future. If this is a large enough commercial application, consider sponsoring one of these improvements to SourceAFIS. Contact me via email if you are interested.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Being a mintutia matcher (and not a level 3 matcher), itself poses a problem with small sensors, so I'm not suprised that the performance would be bad.
When you say "storing multiple scans" do you mean that I capture more images, and store as probes:
FingerprintTemplate probe = new FingerprintTemplate()
.dpi(500)
.create(probeImage);
And then add them to the UserDetails (candidates) list?
Or can I construct one template (probe) from multiple images with some other AFIS function?
Thanks!
Br,
Tim
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Constructing one template from multiple images is what I called fingerprint merging, which is usually implemented on template level as template merging. SourceAFIS doesn't have this feature yet.
Many phones with small sensors will ask you to put your finger on the sensor repeatedly when you enable fingerprint authentication. But they only need one scan to actually unlock the phone. This is logical since the initial setup is performed only once and it is reasonable to ask the user for more patience there. Furthermore, if the first scan used for authentication doesn't work, people retry, which provides a natural source of more scans during authentication too.
You know your application, so you know where potential sources of multiple scans are. You will likely have multiple candidate and multiple probe scans. Match all probes against all candidates and pick the highest scoring match.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I have a smaller sensor with 160x160 pixels, 508dpi, that I'm trialing with the SourceAFIS algorithm.
I get really poor results (scoring <=30) for good fingerprint samples, when matching.
Is there any parameters or tuning that I can do to increase the matching score, or this is due to the inherent nature of Minutia matchers?
Br,
Tim
The sensor area is probably too small. You need larger sensor. Storing multiple scans of the same finger and picking the best match could help somewhat.
SourceAFIS algorithm can be improved to either merge multiple fingerprints during user registration or to mine and use features from the image more aggressively. Neither of these enhancements is planned in near future. If this is a large enough commercial application, consider sponsoring one of these improvements to SourceAFIS. Contact me via email if you are interested.
Hi Robert,
Much thanks for your support.
Being a mintutia matcher (and not a level 3 matcher), itself poses a problem with small sensors, so I'm not suprised that the performance would be bad.
When you say "storing multiple scans" do you mean that I capture more images, and store as probes:
FingerprintTemplate probe = new FingerprintTemplate() .dpi(500) .create(probeImage);And then add them to the UserDetails (candidates) list?
Or can I construct one template (probe) from multiple images with some other AFIS function?
Thanks!
Br,
Tim
Constructing one template from multiple images is what I called fingerprint merging, which is usually implemented on template level as template merging. SourceAFIS doesn't have this feature yet.
Many phones with small sensors will ask you to put your finger on the sensor repeatedly when you enable fingerprint authentication. But they only need one scan to actually unlock the phone. This is logical since the initial setup is performed only once and it is reasonable to ask the user for more patience there. Furthermore, if the first scan used for authentication doesn't work, people retry, which provides a natural source of more scans during authentication too.
You know your application, so you know where potential sources of multiple scans are. You will likely have multiple candidate and multiple probe scans. Match all probes against all candidates and pick the highest scoring match.
Hi Robert,
Thanks for you explaination.
I agree fully, and it's as I expected.
Best,
Tim