Human identification using ear biometrics: A complete pipeline from detection to recognition

Author Identifier

Md Mursalin

https://orcid.org/0000-0001-6524-0759

Date of Award

2023

Document Type

Thesis - ECU Access Only

Publisher

Edith Cowan University

Degree Name

Doctor of Philosophy

School

School of Science

First Supervisor

Mohiuddin Ahmed

Second Supervisor

Paul Haskell-Dowland

Abstract

Biometrics is a critical component of cybersecurity that identifies persons by verifying their behavioral and physical traits. It is the most precise and powerful physical security solution for identity verification presently in use. In biometric-based authentication, each individual can be correctly recognized based on their intrinsic behavioral or physical features. The human ear is an important biometric with highly discriminating features such that even identical twins have different ear shapes. Owing to the easy acquisition, invariance to expression, and stable structure over a long period of time, ear image analysis has the edge over other biometric traits such as faces, fingerprints, palmprints, and iris. These advantages allow the use of ear images for numerous applications, including biometric identification, asymmetry analysis for clinical purposes, genetic relationship study, and gender recognition.

In ear-based biometrics, one of the significant steps is to localize ears in profile face images. Most ear detection approaches use 2D images for ear region localization as they require fewer computations. Due to the importance of being able to handle unconstrained images for object detection and segmentation. However, 2D image-based approaches are limited to constrained scenarios due to their sensitivity to lighting conditions and pose variations. Therefore, 3D images can be used to overcome the limitations of 2D images.

Recent developments in 3D imaging techniques open the door for 3D image-based applications, including biometrics, robotics, medical diagnosis, and autonomous driving. Generally, 3D data can be represented in various forms, such as point clouds, volumetric grids, depth images, and meshes. Point cloud representation is becoming more popular as it reserves the original geometric information in 3D domains without discretization. However, conventional convolutional neural networks cannot be applied directly to point clouds due to the irregular order of the points. Therefore, most work using 3D images generally converts point cloud data to Euclidean structured format before sending it to the CNN architectures. This representation conversion introduces unnecessarily voluminous data and wraps natural invariances of the data due to the generation of quantization artefacts.

This work presents a complete pipeline for ear biometrics that can detect ears directly from 3D point clouds of profile face images and recognize a person with the novel matching algorithm You Morph Once (YMO). This technology can be used in the field of cyber security as a form of biometric authentication, which is a way to identify and verify the identity of a person using unique physical characteristics such as ear shape. The use of ear biometrics as a form of identification can provide an additional level of security, as ears are unique to each individual and difficult to replicate or falsify. Additionally, using 3D point clouds and the 3D Morphable Ear Model (3DMEM) model allows for greater accuracy and robustness in identification, making it harder for hackers to bypass the system. The proposed pipeline can be integrated into various security systems, such as access controls, online authentication, and personal identification. These systems could be used for highly secure areas such as government buildings, military bases, and banks, where the traditional form of authentication like password and PIN is not enough to protect sensitive information.

Access Note

Access to this thesis is embargoed until 30th January 2028.

Access to this thesis is restricted. Please see the Access Note below for access details.

Share

 
COinS