Hi Robert,
I would like to appreciate your efforts in terms of building open source fingerprint matching platform.
I have taken time to look at the SourceAfis project and am please with the functionality so far. I am
developing a module for OpenMRS an open source medical records system that is meant to help identify
patients in the field here in Kenya using a mobile phone fingerprint scanner. The greatest challenge at the
moment lies with extracting the minutia. Looking at AfisEngine, i note even from the discussions that it is
not implemented in Java (Extract method) but some work seems to be going on.
I would highly appreciate if you could direct me through to the most recent effort towards developing that bit
of code(AfisEngine extract method) so that i may offer whatever possible on my end to see this through. I
have looked at Griaule\'s Grfinger java platform and i understand the implementation of minutia extraction
but the challenge is to determine orientation.
I would be more than happy to contribute in any way i can. I look forward to your feedback.
Jonah
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
First of all, I would reconsider the idea of doing extraction on the mobile device. Wouldn't it be better to send the whole image to the server for extraction? You could then use the well-supported C# version of SourceAFIS on the server.
Depending on sensor size, uncompressed grayscale fingerprint is about 100KB. JPEG image is 10-30KB depending on quality. At that level the image can be transmitted quickly even over slow 2G/EDGE networks. If you make use of a WSQ library, it can be transmitted even faster.
You need to transfer the images to the server anyway. Inevitably you will be forced to re-extract fingerprint templates from the images due to SourceAFIS upgrades. So why not transfer the image right away?
Templates are not guaranteed to be smaller than images. I would like to incorporate ridges and other data in the template in later versions of SourceAFIS, which would grow templates to several KB.
If you are already decided to do client-side extraction, you will have to wait for java port of SourceAFIS to be completed. I don't want to give any exact dates, because I am inreasingly short of time.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Just for clarity, i am for server-side extraction. The OpenMRS instance is a java server-side. I was looking forward to supporting the java extraction porting process in-case there is progress that i may stick in the java as opposed to working partly java, partly C#. I appreciate your recommendations though.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
If you are doing server-side extraction, then definitely go for the C# version of SourceAFIS. It's stable, complete, and fast. You want to run it in separate process anyway, so just interface it to the java part via some network protocol. It will also work smoothly under Mono, so Linux is not a problem.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Robert,
I would like to appreciate your efforts in terms of building open source fingerprint matching platform.
I have taken time to look at the SourceAfis project and am please with the functionality so far. I am
developing a module for OpenMRS an open source medical records system that is meant to help identify
patients in the field here in Kenya using a mobile phone fingerprint scanner. The greatest challenge at the
moment lies with extracting the minutia. Looking at AfisEngine, i note even from the discussions that it is
not implemented in Java (Extract method) but some work seems to be going on.
I would highly appreciate if you could direct me through to the most recent effort towards developing that bit
of code(AfisEngine extract method) so that i may offer whatever possible on my end to see this through. I
have looked at Griaule\'s Grfinger java platform and i understand the implementation of minutia extraction
but the challenge is to determine orientation.
I would be more than happy to contribute in any way i can. I look forward to your feedback.
Jonah
First of all, I would reconsider the idea of doing extraction on the mobile device. Wouldn't it be better to send the whole image to the server for extraction? You could then use the well-supported C# version of SourceAFIS on the server.
Depending on sensor size, uncompressed grayscale fingerprint is about 100KB. JPEG image is 10-30KB depending on quality. At that level the image can be transmitted quickly even over slow 2G/EDGE networks. If you make use of a WSQ library, it can be transmitted even faster.
You need to transfer the images to the server anyway. Inevitably you will be forced to re-extract fingerprint templates from the images due to SourceAFIS upgrades. So why not transfer the image right away?
Templates are not guaranteed to be smaller than images. I would like to incorporate ridges and other data in the template in later versions of SourceAFIS, which would grow templates to several KB.
If you are already decided to do client-side extraction, you will have to wait for java port of SourceAFIS to be completed. I don't want to give any exact dates, because I am inreasingly short of time.
Thank you for your feedback.
Just for clarity, i am for server-side extraction. The OpenMRS instance is a java server-side. I was looking forward to supporting the java extraction porting process in-case there is progress that i may stick in the java as opposed to working partly java, partly C#. I appreciate your recommendations though.
If you are doing server-side extraction, then definitely go for the C# version of SourceAFIS. It's stable, complete, and fast. You want to run it in separate process anyway, so just interface it to the java part via some network protocol. It will also work smoothly under Mono, so Linux is not a problem.
Java port has been completed meantime. You can use either C# or Java for server-side feature extraction.