MINUTIA CYLINDER CODE-BASED FINGERPRINT MATCHING OPTIMIZATION USING GPU
The advancement and use of digital data are growing very rapidly along with the <br /> <br /> <br /> <br /> advancement of technology nowadays. In this digital era, lots of physical data have <br /> <br /> <br /> <br /> been transformed into th...
Saved in:
Main Author: | |
---|---|
Format: | Final Project |
Language: | Indonesia |
Online Access: | https://digilib.itb.ac.id/gdl/view/23179 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Institut Teknologi Bandung |
Language: | Indonesia |
Summary: | The advancement and use of digital data are growing very rapidly along with the <br />
<br />
<br />
<br />
advancement of technology nowadays. In this digital era, lots of physical data have <br />
<br />
<br />
<br />
been transformed into the digital ones. One example of the use of digital data is the <br />
<br />
<br />
<br />
digital biometric fingerprint data on the Electronic Identity Card (KTP-el). In 2010, <br />
<br />
<br />
<br />
the population of Indonesia is about 200 million people. We can imagine how long <br />
<br />
<br />
<br />
it would take to process their fingerprint's data if it is done linearly, which is what <br />
<br />
<br />
<br />
is currently happening. Thus, there is a need for a parallel fingerprint matching. <br />
<br />
<br />
<br />
Based on this rationale, this final project aims to improve the fingerprint matching <br />
<br />
<br />
<br />
performance, in the current state of the art linear solution, by using the Minutia <br />
<br />
<br />
<br />
Cylinder-Code (MCC) algorithm in parallel on GPU. The performance <br />
<br />
<br />
<br />
improvements are made in the area of the selection of data structures, the calculation <br />
<br />
<br />
<br />
of valid areas in parallel, the creation of cylinders and cylinders-sets minutiae in <br />
<br />
<br />
<br />
parallel, and global calculations in parallel. <br />
<br />
<br />
<br />
Based on the experiment and testing, the proposed solution has a significantly better <br />
<br />
<br />
<br />
run time compared to the state of the art linear solution while maintaining the <br />
<br />
<br />
<br />
accuracy of the fingerprint matching. The proposed solution can run 14.17 up to <br />
<br />
<br />
<br />
37.82 times faster and the accuracy has only a small difference to the previous <br />
<br />
<br />
<br />
solution with a standard deviation of 0.33%. |
---|