FaceRecognition:ANovelMulti-LevelTaxonomybasedSurvey资源-CSDN文库

需积分: 2 23 浏览量 2019-01-10 11:22:54 上传评论收藏 2.18MB PDF 举报

资源推荐

资源详情

资源评论

Face Recognition: A Novel Multi-Level Taxonomy based Survey

Alireza Sepas-Moghaddam

, Fernando Pereira

, Paulo Lobato Correia

Instituto de Telecomunicações, Instituto Superior Técnico – Universidade de Lisboa, , Lisbon, Portugal

alireza@lx.it.pt

Abstract: In a world where security issues have been gaining growing importance, face recognition systems have attracted

increasing attention in multiple application areas, ranging from forensics and surveillance to commerce and entertainment.

To help understanding the landscape and abstraction levels relevant for face recognition systems, face recognition

taxonomies allow a deeper dissection and comparison of the existing solutions. This paper proposes a new, more

encompassing and richer multi-level face recognition taxonomy, facilitating the organization and categorization of available

and emerging face recognition solutions; this taxonomy may also guide researchers in the development of more efficient face

recognition solutions. The proposed multi-level taxonomy considers levels related to the face structure, feature support and

feature extraction approach. Following the proposed taxonomy, a comprehensive survey of representative face recognition

solutions is presented. The paper concludes with a discussion on current algorithmic and application related challenges which

may define future research directions for face recognition.

1. Introduction

Face recognition systems have been successfully used

in multiple application areas with high acceptability,

collectability and universality [1] [2]. After the first automatic

face recognition algorithms emerged more than four decades

ago [3], this field has attracted much research and witnesses

incredible progress, with a very large number of face

recognition solutions being used in multiple application areas.

Face recognition technology is assuming an increasingly

important role in our everyday life, what also brings ethical

and privacy dilemmas about how the captured facial

information and the corresponding identity, as a special

category of personal data, should be used, stored and shared.

According to [4], “taxonomy is the practice and

science of classification of things or concepts, including the

principles that underlie such classification”. The availability

of a taxonomy in a certain field allows to

organize/classify/abstract the ‘things’ (in this case, the face

recognition solutions) with two main benefits: i) regarding

the present, it makes it easier to discuss and analyse the

‘things’ and abstract the deeper relations between them, thus

providing a deeper knowledge and comprehension of the full

landscape, notably in terms of strengths and weaknesses;

ii) regarding the future, it makes it easier to understand the

most promising research directions and their implications as

the ‘things’ (in this case, the face recognition solutions) will

not be isolated ‘things’ but rather ‘things’ in a taxonomical

network, inheriting features, strengths and weaknesses from

their taxonomy parents and peers.

Compiling a comprehensive survey of the available

face recognition solutions is a challenging task, notably given

the large number and diversity of solutions developed in the

last decades. To help understanding the structure and

abstraction levels that may be considered in face recognition

solutions, a number of face recognition taxonomies have been

proposed so far [5] [6] [7] [8] [9] [10] [11] [12] [13] [14].

Since the available face recognition taxonomies ignore some

relevant levels of abstraction, which may be helpful for a

more complete characterization of the face recognition

landscape, this paper proposes a new, more encompassing

and richer face recognition multi-level taxonomy.

The new taxonomy can be used to better understand

the technological landscape in the area, facilitating the

characterization and organization of available solutions, and

guiding researchers in the development of more efficient face

recognition solutions for given applications.

The proposed

multi-level taxonomy considers four levels, notably face

structure, feature support, feature extraction approach, and

feature extraction sub-approach. This paper also surveys

representative state-of-the-art face recognition solutions

according to the proposed multi-level taxonomy and

discusses the current algorithmic and application related

challenges and future research directions for face recognition

systems.

The rest of the paper is organized as follows. Section

2 reviews the available face recognition taxonomies, to

understand their benefits and limitations. Section 3 proposes

a new, more encompassing and richer multi-level taxonomy

for face recognition solutions. Section 4 surveys the state-of-

the-art on face recognition under the umbrella of the proposed

multi-level face recognition taxonomy and discusses the

evolutional trends of face recognition over time. Finally,

Section 5 discusses some face recognition challenges and

identifies some relevant future research directions.

2. Reviewing existing face recognition taxonomies

Several face recognition taxonomies have been

proposed in the literature [5] [6] [7] [8] [9] [10] [11] [12] [13]

[14], as summarized in Table 1. This table includes

information about the

abstraction level(s) considered as well

as the corresponding classes – notice that some taxonomies

may use a different terminology. Excluding the taxonomy

proposed in [10], all the other taxonomies listed in Table 1

have been developed based on a single abstraction level to

organize face recognition solutions, thus proposing a specific

taxonomical point of view.

Table 1: Summary of available face recognition taxonomies

and corresponding abstraction levels and classes.

Ref. Year

Abstraction

Level(s)

Classes

[5] 2003

Feature

extraction

Appearance based;

Feature based;

Hybrid

[6] 2006

Feature

extraction

Holistic based;

Local based;

Hybrid

[7] 2009

Sensing

modality

2D;

Video;

3D;

infra-red data

[8] 2009

Modality

matching

2D vs. 2D;

2D vs. 2D via 3D;

3D vs. 3D;

Video vs. 3D;

Multimodal 2D+3D

[9] 2010

Facial

features

Soft features;

Anthropometric features;

Unstructured and micro-level

features

[10] 2011

Pose

dependency

Pose-dependant;

Pose-invariant

Feature

extraction

Appearance based;

Feature based;

Hybrid

[11] 2012

Feature

extraction

Holistic based;

Feature based;

Model based;

Hybrid

[12] 2014

Sensing

modality

2D;

3D;

Multimodal 2D+3D

[13] 2016

Illumination

dependency

Illumination normalization;

Illumination modeling;

Illumination in variation

[14] 2016

Pose

dependency

Pose-robust;

Multi-view learning;

Face synthesis;

Hybrid

The face recognition taxonomies proposed in [5], [6],

and [11] consider as abstraction level the type of feature

extraction. These taxonomies include a ‘hybrid’ class to cover

solutions combining two or more feature extraction classes to

boost the face recognition performance. The taxonomy

proposed in [9] is also related to the features used for

recognition, but relies on the type of facial features and not

on the feature extraction approach. This taxonomy considers

three classes, notably: i) soft features, such as gender, race,

and age, encompassing the global nature of the face, which

are mostly useful for indexing and reducing the search space;

ii) anthropometric features, such as facial landmarks and their

relations, thus taking the facial structures as the most

discriminative features; and iii) unstructured and micro-level

features, such as tattoos, moles and scars, which may

facilitate the discrimination between short-listed subjects.

Two recent face recognition taxonomies organize the

face recognition solutions depending on the specific

conditions addressed, such as illumination [13] and pose [14]

dependencies, not being encompassing enough to consider

every face recognition solution. Other taxonomies organize

the face recognition solutions based on the sensing modality

[7] [12], or on the adopted modality matching, to consider the

cases where the enrolment and test data are captured using

different imaging modalities [8]. To summarize, the reviewed

single-level taxonomies classify face recognition solutions

based on a specific abstraction point of view, ignoring other

abstraction levels, along with their relations.

A multi-level taxonomy, would be useful for a more

complete characterization and organization of face

recognition solutions. Accordingly, a multi-level taxonomy

has been proposed in [10], adopting two abstraction levels to

organize the face recognition solutions, notably pose-

dependency and matching features (equivalent to feature

extraction). However, this taxonomy still ignores some

relevant abstraction levels, such as the face structure and

feature support, which should be considered for a more

complete characterization of face recognition solutions.

The major improvements of the proposed multi-level

taxonomy proposed in this paper are:

1. The proposed taxonomy considers face structure and

feature support levels for a more complete

characterization of face recognition solutions. These

levels provide critical information on the way face

recognition solutions deal with the facial structure and

what is the region of spatial support for feature extraction,

impacting the selection of the corresponding approach.

2. The feature extraction approach level in the proposed

taxonomy (equivalent to the matching features level in

[10]) considers a richer and more complete set of classes,

rather than only feature based and appearance based

classes as in [10].

3. The feature extraction approach level is further divided

into feature extraction sub-approaches to more deeply

understand the technological landscape in the context of

feature extraction.

Regarding pose-dependency, this level was

considered not relevant enough as most recent face

recognition solutions are pose-invariant.

In summary, the proposed multi-level taxonomy

provides a more encompassing and richer taxonomy for face

recognition as may, thus, bring additional value in

understanding the face recognition landscape.

3. Proposing a novel multi-level face recognition

taxonomy

Following the taxonomy limitations highlighted in

Section 2, this paper proposes a novel, more encompassing

and richer multi-level face recognition taxonomy, which can

help to better understand the technological landscape in the

area, thus facilitating the characterization, organization and

comparison of face recognition solutions. The proposed

multi-level face recognition taxonomy, illustrated in Figure 1,

considers four taxonomy levels:

剩余11页未读，继续阅读

评论收藏

内容反馈

wj16account

粉丝: 3
资源: 5

Face Recognition: A Novel Multi-Level Taxonomy based Survey

Secure-Face-Recognition-System-based-on-Fully-Homomorphic-Encryption:基于全同态加密的安全人脸识别系统Secure Face Recognition System based on Fully Homomorphic Encryption

Coupled Discriminant Multi-Manifold Analysis for Low-Resolution Face Recognition:Coupled Discriminant Multi-Manifold Analysis for Low-Resolution Face Recognition的Matlab代码-matlab开发

Face-recognition-based-on-PCA-and-SVM-master_FaceOn_人脸识别PCA_源码.z

最新人脸识别论文综述

FaceRecognition:识别人脸

face-recognition-project:人脸识别项目

人脸识别项目

人脸识别

face-recognition-based-attendance-system-master.zip

Face-Recognition-Using-PCA-master_人脸识别_pca_

dlib库（face_recognition安装）-19.7.0-cp36-cp36m-win_amd64

Face-Recognition-Based-AutoAttendanceSystem-master.zip

face recognition-开源

FaceRecognition

Face Recognition

Face-Recognition-Based-Attendance-System

face-recognition-javascript-webcam-faceapi-main.zip

face_recognition_models-0.3.0-py2.py3-none-any.whl

cnn卷积神经网络论文.zip

OpenCV-Face-Recognition-master.zip

face recognization

Simple-Face-Recognition-App-CS-master_recognition_C#_facerecogni

face_recognition

最新资源