CN102495999A - Face recognition method - Google Patents

Face recognition method Download PDF

Info

Publication number
CN102495999A
CN102495999A CN2011103583624A CN201110358362A CN102495999A CN 102495999 A CN102495999 A CN 102495999A CN 2011103583624 A CN2011103583624 A CN 2011103583624A CN 201110358362 A CN201110358362 A CN 201110358362A CN 102495999 A CN102495999 A CN 102495999A
Authority
CN
China
Prior art keywords
face
proper vector
dimensional
facial image
gabor
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN2011103583624A
Other languages
Chinese (zh)
Inventor
刘鸣宇
王金楠
谢洵
王光明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
SHENZHEN BIOCOME SAFETY TECHNOLOGY CO LTD
Original Assignee
SHENZHEN BIOCOME SAFETY TECHNOLOGY CO LTD
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by SHENZHEN BIOCOME SAFETY TECHNOLOGY CO LTD filed Critical SHENZHEN BIOCOME SAFETY TECHNOLOGY CO LTD
Priority to CN2011103583624A priority Critical patent/CN102495999A/en
Publication of CN102495999A publication Critical patent/CN102495999A/en
Pending legal-status Critical Current

Links

Images

Abstract

The invention provides a face recognition method, which is a three-dimensional face recognition algorithm by utilizing two-dimensional Gabor wavelet features for two-dimensional Gabor wavelet feature extraction based on kernel linear discrimination. Effective two-dimensional Gabor wavelet features of a face are combined with the three-dimensional face recognition algorithm, and feature vectors are analyzed by a linear discrimination analysis method, thus influence on the face recognition caused by factors including illumination, posture, expression and the like and disadvantages of single three-dimensional face recognition on the factor of extracting effective face features are overcome.

Description

A kind of method of recognition of face
Technical field
The present invention relates to the recognition of face field, particularly based on the three-dimensional face recognition algorithm of two-dimensional Gabor wavelet character.
Background technology
Recognition of face is the computer technology research field of a hot topic, and it belongs to biometrics identification technology, is that the biological characteristic of human body itself is distinguished the biosome individuality.The two-dimension human face recognition technology is ripe at present; But two-dimension human face identification quality receives following factor affecting: (1) illumination; Image by photo obtains is represented people's face looks with half-tone information, and under varying environment (light, background, visual angle etc.), the human face photo difference that obtains is very big; (2) how attitude, photo obtain 3 d pose and adjusts having difficulty greatly from 2-dimentional photo usually not in strict conformity with the front or the lateral attitude condition; (3) expression shape change of people's face and looks can have a strong impact on recognition result with the variation of age generation.The variation of illumination and attitude is the subject matter that recognition of face faces, same individual the facial image difference under different light and the attitude usually greater than different people same light according to and attitude under difference.Three-dimensional face identification is directly extracted people's face in three-dimensional real features information; Be different from the plane projection of people's face on certain direction in the two-dimension human face identification; Therefore; Its characteristic information is more abundant, can on original two-dimension human face base of recognition, effectively improve system identification degree of accuracy and the enhanced system robustness to attitude, illumination and expression.But people's face is a flexible article, and the flexibility of human face expression changes to the three-dimensional model influence very greatly.
The multiple small echo of two-dimensional Gabor is proposed by Daugman the earliest, is widely used in various Flame Image Process thereupon, has obtained extraordinary effect.
The multiple small echo of two-dimensional Gabor is:
Figure DEST_PATH_IMAGE001
Wherein,
Figure 436819DEST_PATH_IMAGE002
; ;
Figure 868937DEST_PATH_IMAGE004
and
Figure DEST_PATH_IMAGE005
is the scale parameter of wave filter;
Figure 87429DEST_PATH_IMAGE006
expression be frequency,
Figure DEST_PATH_IMAGE007
expression be direction.
Formula (1) at frequency domain representation is:
Wherein ;
Figure 244052DEST_PATH_IMAGE010
;
Figure DEST_PATH_IMAGE011
,
Figure 214282DEST_PATH_IMAGE012
.
At present, also less than three-dimensional face recognition algorithm based on the two-dimensional Gabor wavelet character.
Summary of the invention
The objective of the invention is to disclose a kind of three-dimensional face recognition algorithm of two-dimensional Gabor wavelet character.
The present invention in order to realize the technical scheme that its goal of the invention adopts is: a kind of face identification method, this method are a kind of three-dimensional face recognition algorithm based on the two-dimensional Gabor wavelet character, may further comprise the steps:
Steps A, align the dough figurine face and detect, locate crucial human face characteristic point in a front face and the facial image, obtain people's face original image I (x, y);
Step B, characterize about the Gabor of facial image through Gabor wavelet analysis obtain; Make described original image I (x, the individual features in y) is converted into Gabor proper vector ;
Step C, adopt little exponential polynomials (FPP) model
Figure 791675DEST_PATH_IMAGE014
, m dimension Gabor feature space
Figure DEST_PATH_IMAGE015
is projected in the higher n dimension space
Figure 821948DEST_PATH_IMAGE016
;
Step D, based on nuclear linear judgment analysis algorithm (KFDA); Matrix
Figure 692001DEST_PATH_IMAGE018
in matrix
Figure DEST_PATH_IMAGE017
and the class between in
Figure 918080DEST_PATH_IMAGE016
space, setting up type; Wherein,
Figure DEST_PATH_IMAGE019
Figure 885085DEST_PATH_IMAGE020
, the standard orthogonal characteristic of calculating
Figure 587724DEST_PATH_IMAGE018
vector
Figure DEST_PATH_IMAGE021
;
Step e, extraction facial image are significantly differentiated proper vector;
Step F, the inapparent differentiation proper vector of extraction facial image;
Step G, utilize facial image significantly to differentiate proper vector and the inapparent differentiation proper vector of facial image and a 3D face database reconstruction of three-dimensional faceform;
Step H, adopt template matches and linear discriminant analysis (FLDA) method to handle to described three-dimensional face model, difference is carried out the match people face with a type differences in the class of extraction model.
Further, in above-mentioned a kind of face identification method: in the described steps A, crucial human face characteristic point in the described facial image comprises contour feature point, left eye and right eye, mouth and the nose of people's face.
Further, in above-mentioned a kind of face identification method: comprise in the described step e:
Step e 01, make
Figure 967889DEST_PATH_IMAGE022
; Wherein, is that q the eigenwert that matrix
Figure 607818DEST_PATH_IMAGE018
is corresponding in described type is positive proper vector, ;
Step e 02, calculate
Figure DEST_PATH_IMAGE025
corresponding to the proper vector
Figure DEST_PATH_IMAGE027
of individual eigenvalue of maximum; Wherein, ;
Figure DEST_PATH_IMAGE029
;
Step e 03, significantly differentiate proper vector ; Wherein,
Figure DEST_PATH_IMAGE031
Further, in above-mentioned a kind of face identification method: comprise in the described step F:
Step F01 calculation ? corresponds to a maximum eigenvalue vector ;
Step F 02, make
Figure 455743DEST_PATH_IMAGE034
; Then inapparent differentiation proper vector
Figure DEST_PATH_IMAGE035
,
Figure 76080DEST_PATH_IMAGE036
.
Further, in above-mentioned a kind of face identification method: the 3D face database is the single face three-dimensional face database of ORL (Olivetti Research Laboratory) among the described step G.
Further, in above-mentioned a kind of face identification method: among the described step G:
Step G01, confirm a Feature Conversion matrix P;
Step G02, the Gabor that extracts is differentiated the proper vector of proper vector corresponding to m eigenvalue of maximum, form main Feature Conversion matrix P ';
Step G03, extensive face database is carried out repetition training to obtain average face S.
The present invention proposes a kind of two-dimensional Gabor wavelet character based on the linear judgement of nuclear and extracts; Adopt the three-dimensional face recognition algorithm of two-dimensional Gabor wavelet character; Through people's face efficient 2-d Gabor wavelet character is combined with three-dimensional face recognition algorithm; Use the linear discriminant analysis method that proper vector is analyzed, overcome factors such as illumination, attitude and expression influence and the single three-dimensional face that recognition of face causes is identified in the deficiency of extracting effective face characteristic aspect.
Below in conjunction with specific embodiment the present invention is done comparatively detailed description.
Description of drawings
Fig. 1, process flow diagram of the present invention.
Fig. 2 is an embodiment of the invention design sketch.
Embodiment
As shown in Figure 1, present embodiment is a kind of three-dimensional face recognition algorithm based on the two-dimensional Gabor wavelet character, and this method is a kind of three-dimensional face recognition algorithm based on the two-dimensional Gabor wavelet character, may further comprise the steps:
Steps A, align the dough figurine face and detect, locate crucial human face characteristic point in a front face and the facial image, obtain people's face original image I (x, y); Crucial human face characteristic point in the image comprises contour feature point, left eye and right eye, mouth and the nose of people's face.
Step B, characterize about the Gabor of facial image through Gabor wavelet analysis obtain; Make described original image I (x, the individual features in y) is converted into Gabor proper vector
Figure 729916DEST_PATH_IMAGE013
;
Step C, adopt little exponential polynomials (FPP) model
Figure 689781DEST_PATH_IMAGE014
, m dimension Gabor feature space is projected in the higher n dimension space
Figure 512010DEST_PATH_IMAGE016
;
Step D, based on nuclear linear judgment analysis algorithm (KFDA); Matrix
Figure 668688DEST_PATH_IMAGE018
in matrix
Figure 783908DEST_PATH_IMAGE017
and the class between in space, setting up type; Wherein,
Figure 935721DEST_PATH_IMAGE019
Figure 196938DEST_PATH_IMAGE020
, the standard orthogonal characteristic of calculating
Figure 865817DEST_PATH_IMAGE018
vector
Figure 422962DEST_PATH_IMAGE021
;
Step e, extraction facial image are significantly differentiated proper vector; Make
Figure 177292DEST_PATH_IMAGE022
; Wherein, is
Figure 765585DEST_PATH_IMAGE018
, and q corresponding eigenwert is positive proper vector;
Figure 992167DEST_PATH_IMAGE024
. calculating
Figure 233792DEST_PATH_IMAGE025
is corresponding to the proper vector of
Figure 836812DEST_PATH_IMAGE026
individual eigenvalue of maximum; Wherein,
Figure 376301DEST_PATH_IMAGE028
; ; Significantly differentiate proper vector
Figure 246354DEST_PATH_IMAGE030
; Wherein,
Figure 478753DEST_PATH_IMAGE031
Step F, extracted face image is not significant discriminant feature vector; calculation
Figure 312716DEST_PATH_IMAGE032
? largest eigenvalue corresponds to an eigenvector
Figure 263355DEST_PATH_IMAGE033
.Make
Figure 208177DEST_PATH_IMAGE034
; Then inapparent differentiation proper vector
Figure 560661DEST_PATH_IMAGE035
,
Figure 66991DEST_PATH_IMAGE036
.
Step G, utilize facial image significantly to differentiate proper vector and the inapparent differentiation proper vector of facial image and a 3D face database reconstruction of three-dimensional faceform; In order to rebuild a three-dimensional face model, use the single face three-dimensional face database of ORL (Olivetti Research Laboratory).Confirm a Feature Conversion matrix P, in original three-dimensional face identification method, the subspace analysis projection matrix that this matrix is normally obtained by the subspace analysis method is made up of corresponding to the proper vector of a preceding m eigenvalue of maximum the covariance matrix of sample.The Gabor that extracts is differentiated the proper vector of proper vector corresponding to m eigenvalue of maximum; Form main Feature Conversion matrix P '; This Feature Conversion matrix has stronger robustness than original eigenmatrix P to factors such as illumination, attitude and expressions, and promptly the characteristic of representative is more accurate and stable.Extensive face database is carried out repetition training to obtain average face S.
Step H, adopt template matches and linear discriminant analysis (FLDA) method to handle to described three-dimensional face model, difference is carried out the match people face with a type differences in the class of extraction model.
Comparative study this paper method with based on the recognition technology of Gabor characteristic quantity and kernel function principal component analysis (PCA) (KPCA), compare based on Gabor characteristic quantity and principal component analytical method (PCA) recognition technology, the result shows that the method for this paper has improved people's ability.The result is as shown in Figure 2.

Claims (6)

1. face identification method, this method is a kind of three-dimensional face recognition algorithm based on the two-dimensional Gabor wavelet character, it is characterized in that: may further comprise the steps:
Steps A, align the dough figurine face and detect, locate crucial human face characteristic point in a front face and the facial image, obtain people's face original image I (x, y);
Step B, characterize about the Gabor of facial image through Gabor wavelet analysis obtain; Make described original image I (x, the individual features in y) is converted into Gabor proper vector
Figure 392485DEST_PATH_IMAGE001
;
Step C, adopt little exponential polynomials (FPP) model
Figure 488617DEST_PATH_IMAGE002
, m dimension Gabor feature space
Figure 872325DEST_PATH_IMAGE003
is projected in the higher n dimension space
Figure 3092DEST_PATH_IMAGE004
;
Step D, based on nuclear linear judgment analysis algorithm (KFDA); Matrix
Figure 343135DEST_PATH_IMAGE006
in matrix
Figure 663061DEST_PATH_IMAGE005
and the class between in
Figure 79633DEST_PATH_IMAGE004
space, setting up type; Wherein,
Figure 328409DEST_PATH_IMAGE007
Figure 575851DEST_PATH_IMAGE008
, the standard orthogonal characteristic of calculating
Figure 646575DEST_PATH_IMAGE006
vector
Figure 637665DEST_PATH_IMAGE009
;
Step e, extraction facial image are significantly differentiated proper vector;
Step F, the inapparent differentiation proper vector of extraction facial image;
Step G, utilize facial image significantly to differentiate proper vector and the inapparent differentiation proper vector of facial image and a 3D face database reconstruction of three-dimensional faceform;
Step H, adopt template matches and linear discriminant analysis (FLDA) method to handle to described three-dimensional face model, difference is carried out the match people face with a type differences in the class of extraction model.
2. a kind of face identification method according to claim 1 is characterized in that: in the described steps A, crucial human face characteristic point in the described facial image comprises contour feature point, left eye and right eye, mouth and the nose of people's face.
3. a kind of face identification method according to claim 1 is characterized in that: comprise in the described step e:
Step e 01, make
Figure 477445DEST_PATH_IMAGE010
; Wherein,
Figure 895788DEST_PATH_IMAGE011
is that q the eigenwert that matrix is corresponding in described type is positive proper vector,
Figure 481544DEST_PATH_IMAGE012
;
Step e 02, calculate
Figure 175831DEST_PATH_IMAGE013
corresponding to the proper vector
Figure 810392DEST_PATH_IMAGE015
of
Figure 765075DEST_PATH_IMAGE014
individual eigenvalue of maximum; Wherein,
Figure 143284DEST_PATH_IMAGE016
;
Figure 692077DEST_PATH_IMAGE017
;
Step e 03, significantly differentiate proper vector ; Wherein,
Figure 421053DEST_PATH_IMAGE019
.
4. a kind of face identification method according to claim 3 is characterized in that: comprise in the described step F:
Step F01 calculation
Figure 354374DEST_PATH_IMAGE020
? largest eigenvalue corresponds to an eigenvector
Figure 960936DEST_PATH_IMAGE021
;
Step F 02, make
Figure 219879DEST_PATH_IMAGE022
; Then inapparent differentiation proper vector ,
Figure 586587DEST_PATH_IMAGE024
.
5. a kind of face identification method according to claim 1 is characterized in that: the 3D face database is the single face three-dimensional face database of ORL (Olivetti Research Laboratory) among the described step G.
6. a kind of face identification method according to claim 5 is characterized in that: among the described step G:
Step G01, confirm a Feature Conversion matrix P;
Step G02, the Gabor that extracts is differentiated the proper vector of proper vector corresponding to m eigenvalue of maximum, form main Feature Conversion matrix P ';
Step G03, extensive face database is carried out repetition training to obtain average face S.
CN2011103583624A 2011-11-14 2011-11-14 Face recognition method Pending CN102495999A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2011103583624A CN102495999A (en) 2011-11-14 2011-11-14 Face recognition method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2011103583624A CN102495999A (en) 2011-11-14 2011-11-14 Face recognition method

Publications (1)

Publication Number Publication Date
CN102495999A true CN102495999A (en) 2012-06-13

Family

ID=46187824

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2011103583624A Pending CN102495999A (en) 2011-11-14 2011-11-14 Face recognition method

Country Status (1)

Country Link
CN (1) CN102495999A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103745208A (en) * 2014-01-27 2014-04-23 中国科学院深圳先进技术研究院 Face recognition method and device
CN103745206A (en) * 2014-01-27 2014-04-23 中国科学院深圳先进技术研究院 Human face identification method and system
CN103902992A (en) * 2014-04-28 2014-07-02 珠海易胜电子技术有限公司 Human face recognition method
CN103971122A (en) * 2014-04-30 2014-08-06 深圳市唯特视科技有限公司 Three-dimensional human face description method and device based on depth image
CN105678341A (en) * 2016-02-19 2016-06-15 天纺标检测科技有限公司 Wool cashmere recognition algorithm based on Gabor wavelet analysis
WO2017107957A1 (en) * 2015-12-22 2017-06-29 中兴通讯股份有限公司 Human face image retrieval method and apparatus
WO2020199693A1 (en) * 2019-03-29 2020-10-08 中国科学院深圳先进技术研究院 Large-pose face recognition method and apparatus, and device
CN112784660A (en) * 2019-11-01 2021-05-11 财团法人工业技术研究院 Face image reconstruction method and system

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
孔华锋等: "基于二维Gabor 小波特征的三维人脸识别算法", 《计算机工程》 *
曹林等: "基于二维Gabor小波的人脸识别算法", 《电子与信息学报》 *
王琳等: "基于二维Gabor 小波矩阵表征人脸的识别算法", 《计算机工程》 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103745206A (en) * 2014-01-27 2014-04-23 中国科学院深圳先进技术研究院 Human face identification method and system
CN103745208A (en) * 2014-01-27 2014-04-23 中国科学院深圳先进技术研究院 Face recognition method and device
CN103745206B (en) * 2014-01-27 2018-07-10 中国科学院深圳先进技术研究院 A kind of face identification method and system
CN103902992B (en) * 2014-04-28 2017-04-19 珠海易胜电子技术有限公司 Human face recognition method
CN103902992A (en) * 2014-04-28 2014-07-02 珠海易胜电子技术有限公司 Human face recognition method
WO2015165227A1 (en) * 2014-04-28 2015-11-05 珠海易胜电子技术有限公司 Human face recognition method
CN103971122A (en) * 2014-04-30 2014-08-06 深圳市唯特视科技有限公司 Three-dimensional human face description method and device based on depth image
WO2017107957A1 (en) * 2015-12-22 2017-06-29 中兴通讯股份有限公司 Human face image retrieval method and apparatus
CN105678341A (en) * 2016-02-19 2016-06-15 天纺标检测科技有限公司 Wool cashmere recognition algorithm based on Gabor wavelet analysis
CN105678341B (en) * 2016-02-19 2018-11-13 天纺标检测认证股份有限公司 A kind of woollen and cashmere recognizer based on Gabor wavelet analysis
WO2020199693A1 (en) * 2019-03-29 2020-10-08 中国科学院深圳先进技术研究院 Large-pose face recognition method and apparatus, and device
CN112784660A (en) * 2019-11-01 2021-05-11 财团法人工业技术研究院 Face image reconstruction method and system
CN112784660B (en) * 2019-11-01 2023-10-24 财团法人工业技术研究院 Face image reconstruction method and system

Similar Documents

Publication Publication Date Title
CN106778468B (en) 3D face identification method and equipment
CN102495999A (en) Face recognition method
Al-Osaimi et al. An expression deformation approach to non-rigid 3D face recognition
Alyuz et al. 3-D face recognition under occlusion using masked projection
CN101315661B (en) Fast three-dimensional face recognition method for reducing expression influence
Heseltine et al. Three-dimensional face recognition using combinations of surface feature map subspace components
Li et al. Efficient 3D face recognition handling facial expression and hair occlusion
CN101819628A (en) Method for performing face recognition by combining rarefaction of shape characteristic
CN106599785B (en) Method and equipment for establishing human body 3D characteristic identity information base
Wang et al. Feature representation for facial expression recognition based on FACS and LBP
CN105868716A (en) Method for human face recognition based on face geometrical features
Pan et al. 3D face recognition from range data
CN104318202A (en) Method and system for recognizing facial feature points through face photograph
CN103984922B (en) Face identification method based on sparse representation and shape restriction
CN103218609A (en) Multi-pose face recognition method based on hidden least square regression and device thereof
CN105740779A (en) Method and device for human face in-vivo detection
Duan et al. Expression of Concern: Ethnic Features extraction and recognition of human faces
KR20080035711A (en) Global feature extraction method for 3d face recognition
CN106611158A (en) Method and equipment for obtaining human body 3D characteristic information
CN106778491B (en) The acquisition methods and equipment of face 3D characteristic information
Tin et al. Gender and age estimation based on facial images
Sharma et al. 3D face recognition techniques-a review
Nayak et al. Efficient face recognition with compensation for aging variations
Li et al. Face recognition by 3d registration for the visually impaired using a rgb-d sensor
Thakur et al. Identification of face age range group using neural network

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C12 Rejection of a patent application after its publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20120613