Full Text Available

Access Repository
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Cover Image

Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline

Identifying people from their fingerprints is based on well established technology. However, a number of challenges remain, notably overcoming the low feature density of the surface fingerprint and suboptimal feature matching. 2D contact based fingerprint scanners offer low security performance, are...

Full description

Saved in:
Main Author: Pitcher, Courtney Richard
Other Authors: Marais, Patrick
Format: Thesis
Language:English
Published: Department of Computer Science 2021
Subjects:
Computer Science
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1867613158412124160
access_status_str Open Access
author Pitcher, Courtney Richard
author2 Marais, Patrick
author_browse Marais, Patrick
Pitcher, Courtney Richard
author_facet Marais, Patrick
Pitcher, Courtney Richard
author_sort Pitcher, Courtney Richard
collection Thesis
description Identifying people from their fingerprints is based on well established technology. However, a number of challenges remain, notably overcoming the low feature density of the surface fingerprint and suboptimal feature matching. 2D contact based fingerprint scanners offer low security performance, are easy to spoof, and are unhygienic. Optical Coherence Tomography (OCT) is an emerging technology that allows a 3D volumetric scan of the finger surface and its internal microstructures. The junction between the epidermis and dermis - the internal fingerprint - mirrors the external fingerprint. The external fingerprint is prone to degradation from wear, age, or disease. The internal fingerprint does not suffer these deficiencies, which makes it a viable candidate zone for feature extraction. We develop a biometrics pipeline that extracts and matches features from and around the internal fingerprint to address the deficiencies of contemporary 2D fingerprinting. Eleven different feature types are explored. For each type an extractor and Iterative Closest Point (ICP) matcher is developed. ICP is modified to operate in a Cartesiantoroidal space. Each of these features are matched with ICP against another matcher, if one existed. The feature that has the highest Area Under the Curve (AUC) on an Receiver Operating Characteristic of 0.910 is a composite of 3D minutia and mean local cloud, followed by our geometric properties feature, with an AUC of 0.896. By contrast, 2D minutiae extracted from the internal fingerprint achieved an AUC 0.860. These results make our pipeline useful in both access control and identification applications. ICP offers a low False Positive Rate and can match ∼30 composite 3D minutiae a second on a single threaded system, which is ideal for access control. Identification systems require a high True Positive and True Negative Rate, in addition time is a less stringent requirement. New identification systems would benefit from the introduction of an OCT based pipeline, as all the 3D features we tested provide more accurate matching than 2D minutiae. We also demonstrate that ICP is a viable alternative to match traditional 2D features (minutiae). This method offers a significant improvement over the popular Bozorth3 matcher, with an AUC of 0.94 for ICP versus 0.86 for Bozorth3 when matching a highly distorted dataset generated with SFinGe. This compatibility means that ICP can easily replace other matchers in existing systems to increase security performance.
format Thesis
id oai:open.uct.ac.za:11427/33923
institution University of Cape Town (South Africa)
language eng
last_indexed 2026-06-10T12:31:41.113Z
license_str Not specified — see source repository
provenance_str_mv Harvested via OAI-PMH from UCTD — University of Cape Town Open Access Repository
publishDate 2021
publishDateRange 2021
publishDateSort 2021
publisher Department of Computer Science
publisherStr Department of Computer Science
record_format dspace
source_str UCTD — University of Cape Town Open Access Repository
spelling oai:open.uct.ac.za:11427/33923 Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline Pitcher, Courtney Richard Marais, Patrick Computer Science Identifying people from their fingerprints is based on well established technology. However, a number of challenges remain, notably overcoming the low feature density of the surface fingerprint and suboptimal feature matching. 2D contact based fingerprint scanners offer low security performance, are easy to spoof, and are unhygienic. Optical Coherence Tomography (OCT) is an emerging technology that allows a 3D volumetric scan of the finger surface and its internal microstructures. The junction between the epidermis and dermis - the internal fingerprint - mirrors the external fingerprint. The external fingerprint is prone to degradation from wear, age, or disease. The internal fingerprint does not suffer these deficiencies, which makes it a viable candidate zone for feature extraction. We develop a biometrics pipeline that extracts and matches features from and around the internal fingerprint to address the deficiencies of contemporary 2D fingerprinting. Eleven different feature types are explored. For each type an extractor and Iterative Closest Point (ICP) matcher is developed. ICP is modified to operate in a Cartesiantoroidal space. Each of these features are matched with ICP against another matcher, if one existed. The feature that has the highest Area Under the Curve (AUC) on an Receiver Operating Characteristic of 0.910 is a composite of 3D minutia and mean local cloud, followed by our geometric properties feature, with an AUC of 0.896. By contrast, 2D minutiae extracted from the internal fingerprint achieved an AUC 0.860. These results make our pipeline useful in both access control and identification applications. ICP offers a low False Positive Rate and can match ∼30 composite 3D minutiae a second on a single threaded system, which is ideal for access control. Identification systems require a high True Positive and True Negative Rate, in addition time is a less stringent requirement. New identification systems would benefit from the introduction of an OCT based pipeline, as all the 3D features we tested provide more accurate matching than 2D minutiae. We also demonstrate that ICP is a viable alternative to match traditional 2D features (minutiae). This method offers a significant improvement over the popular Bozorth3 matcher, with an AUC of 0.94 for ICP versus 0.86 for Bozorth3 when matching a highly distorted dataset generated with SFinGe. This compatibility means that ICP can easily replace other matchers in existing systems to increase security performance. 2021-09-15T15:19:58Z 2021-09-15T15:19:58Z 2021 2021-09-15T02:25:03Z Master Thesis Masters MSc http://hdl.handle.net/11427/33923 eng application/pdf Department of Computer Science Faculty of Science
spellingShingle Computer Science
Pitcher, Courtney Richard
Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
thesis_degree_str Master's
title Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
title_full Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
title_fullStr Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
title_full_unstemmed Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
title_short Matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
title_sort matching optical coherence tomography fingerprint scans using an iterative closest point pipeline
topic Computer Science
url http://hdl.handle.net/11427/33923
work_keys_str_mv AT pitchercourtneyrichard matchingopticalcoherencetomographyfingerprintscansusinganiterativeclosestpointpipeline

Similar Items