CN102324022B - Composite gradient vector-based face recognition method - Google Patents
Composite gradient vector-based face recognition method Download PDFInfo
- Publication number
- CN102324022B CN102324022B CN 201110259963 CN201110259963A CN102324022B CN 102324022 B CN102324022 B CN 102324022B CN 201110259963 CN201110259963 CN 201110259963 CN 201110259963 A CN201110259963 A CN 201110259963A CN 102324022 B CN102324022 B CN 102324022B
- Authority
- CN
- China
- Prior art keywords
- vector
- base
- gradient
- face
- base vector
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of pattern recognition, and in particular relates to a composite gradient vector-based face recognition method. The method comprises the following steps of: marking a target area in a positioned face image, dividing feature subareas in the target area, performing orthogonal sampling by using marginal singular points of the feature subareas as starting points and end points of vectors to obtain base vectors, constructing all the base vectors in the target area into a vector cluster, performing multi-dimensional compounding on the base vectors to obtain all great gradient vectors in the vector cluster, constructing a composite gradient vector by using the great gradient vectors as elements, counting the dimension and the gradient information of the composite gradient vector, and comparing the composite gradient vector and the dimension and the gradient information of the composite gradient vector with a face library to recognize face identity. Compared with other face recognition methods, the face recognition method provided by the invention has the advantages of stronger environmental suitability and feature extraction capacity and high recognition performance under the conditions of illumination intensity variation, multiple gestures and multiple expressions and can be used for face recognition under the large-range complex environment in the field of biological feature identification.
Description
Technical field
The invention belongs to mode identification technology, be specifically related to a kind of face identification method based on composite gradient vector.
Background technology
Face recognition technology is to utilize computing machine to carry out analysis and the coupling of facial image, has contactless, moderate, the simple technological merit of process of distance.Face recognition application is extensive, in security fields such as criminal's identification, identity tracking, entry and exit identity verification and gate control systems, has higher using value, is also the study on classics problem in area of pattern recognition simultaneously.
As far back as phase late 1960s, people utilize the Face geometric eigenvector filtered out to carry out Classification and Identification, and recognition effect is not satisfactory.The latter stage eighties, Kirby[1] etc. the people by the introducing of Karhunen-Loeve transformation thought, designed a kind of face recognition technology under describing based on least mean-square error.On this basis, Turk and Pentland[2] the eigenface recognition technology based on the reconstruct weight vector has been proposed, subsequently, the research boom of the subspace human face analysis method based on face characterization has been launched in the recognition of face field.Wherein more classical recognition methods has people's face principal component analysis (PCA) PCA[2], linear discriminant analysis LDA[3], independent principal component analysis (PCA) ICA[4], Baysian method [5], the Kernel PCA method [6] based on nuclear technology, Kernel LDA method [7], Variant Faces sorting technique [8], method [9] of becoming with local feature set based on the overall situation etc.
With the recognition system of organic sphere, compare, a general weakness of existing face identification method is the ability of shortage and environment adjusted in concert.And biological visually-perceptible can well be coordinated with surrounding environment naturally, to adapt to itself maximum attention degree to sensation target, the most attractive characterization information of automatic tracing target, then these the most attractive characterization information are combined, form the most obvious feature distribution constraint information in the biological vision effect, using this foundation as target identification.If existing face identification method has had the ability that biological this " self-adaptation " differentiated, will play limitless effect to the raising of recognition of face effect.
List of references
[1] Kirby M, Sirovich L. Application of the Karhunen-Loeve procedure for the characterization of human faces. IEEE Trans . on Pattern Analysis and Machine Intelligence,1990,12:103-108。
[2]Turk M,Pentland A.Eigenfaces for recognition.Journal of Cognitive Neuroscience,1991,3:71~86。
[3]Belhumeur V, Hespanda J, Kiregeman D. Eigenfacesvs. Fisherfaces:Recognition using class specific linear projection. IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19:711~720。
[4]Bartlett M. S, Movellan J.R. Sejnowski T.J. Face recognition by independent component analysis. IEEE Transactions on Neural Networks,2002,13:1450~1464。
[5]Moghaddam Brik. Principal manifolds and probabilistic subspaces for visual recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24:780~788。
[6]Kim K.I, Jung K. Kim H. J. Face recognition using kernel principal component analysis. IEEE Signal Processing Letters,2002,9:40~42。
[7]Mika S. Rat sch G. Weston J. Scholkopf B. Muller K. Fisher discriminant analysis with kernels. In: Proceedings of IEEE Workshop on Neural Network for Signal Processing,Madison, Wisconsin,USA,1999,9:41~48。
[8]Martinez A M. Recognizing Imprecisely Localized,Partially Occluded,and Expression Variant Faces from a Single Sample Per Class[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,2008,28:755-768.。
[9] Su Yu, mountain generation light, Chen Xilin, high literary composition. the recognition of face [J] become with local feature set based on the overall situation. Journal of Software, 2010,21:1849-1862.
Summary of the invention
The object of the invention is to overcome above-mentioned technical deficiency, a kind of face identification method based on composite gradient vector is provided.
The technical scheme that technical solution problem of the present invention adopts is feature extraction, Vector Fusion and recognition of face three phases:
1, feature extraction phases
The fundamental purpose in this stage is to extract the characteristic information of people's face to be identified.
(1), normalization facial image to be identified, obtain target area
, wherein
For the target area that has comprised people's face main information,
For the target area width,
For length.
(2), calculate distribution probability and the information entropy of each gray scale rank respective pixel, and the statistical information entropy distributes and the extreme point of one-dimension information entropy.When the information entropy of usining obtains maximum value in distributing, the dispersing character subregion, as the segmentation threshold of target area, is divided in corresponding gray scale rank.The edge pixel point of dispersing character subregion is singular point.
(3), in target area
In, using the pixel at width center as starting point, search for vertically downward singular point, and first singular point that will search is set up first base vector as terminal
.
(4), with base vector
Terminal as starting point, respectively vertical left, under, right counter clockwise direction searches for next singular point, and sets up respectively three follow-up vectors of first base vector using first singular point of searching on each direction as terminal.Until singular point is not found at the edge of target area T yet, do not set up the base vector of this direction.
(5), using all base vectors of setting up terminal as starting point more respectively vertical left, under, right, on counter clockwise direction search singular point, and first singular point that will search is set up respectively base vector as terminal.Each singular point only need be searched for three directions, and its starting point direction is ignored, and until target area
Edge do not find yet singular point, do not set up the base vector of this direction.
(6), all directions when all starting points have all arrived target area
Edge all search for less than singular point, stop search.Now sampled and obtained all base vectors in target area.
For the base vector in target area
, wherein
For integer;
Base vector
Position in vector bunch is the
Row, the
Row, the content of vectorial bunch refers to the Vector Fusion stage; Base vector
Dimension is designated as
, gradient is designated as
Work as base vector
Terminal be base vector
Starting point, claim
For
The precursor vector,
For
Follow-up vector; Be called the root vector of vectorial bunch without the base vector of forerunner's vector, be called the leaf vector of vectorial bunch without the base vector of follow-up vector.Base vector
Be called the root vector of vectorial bunch.
2, the Vector Fusion stage
The fundamental purpose in this stage is merged the characteristic information of people's face to be identified, obtains its maximum constrained characteristic information.
(1) base vector, sampling obtained is set up vector bunch
, wherein
It is the set of base vector;
It is the set of mutual relationship between base vector.And demarcate base vector, record base vector forerunner and follow-up restriction relation, extract base vector dimension and gradient information.
(2), demarcation, restriction relation, latitude and the gradient information of base vector in vector bunch used respectively to structure matrix
, constraint matrix
, the dimension matrix
And gradient matrix
Mean:
In structure matrix, the element of the first row
Root vector for vectorial bunch; The element of the second row is base vector
,
With
, be all
Follow-up vector, and take the subvector bunch that the three is root vector
,
,
Order in vector bunch be respectively vertical left, under, right counter clockwise direction; Base vector
, by counter clockwise direction difference composite vector
,
With
Obtain its follow-up vector
,
,
; Of structure matrix
Row, the
The base vector of row
Three follow-up vectors are arranged
,
With
, and the three in vector bunch, be according to vertical left, under, right, on counter clockwise direction arrange, have so
,
,
.
In constraint matrix, element is two tuples, and the unit of two tuples is base vectors; Two tuples
First yuan
It is second yuan
Precursor vector;
Be
Follow-up vector.
In dimension and gradient matrix, corresponding to of structure matrix
Row, the
The base vector of row
, of dimension matrix
Row, the
The element of row is base vector
Dimension
, of gradient matrix
Row, the
The element of row is base vector
Gradient
.
(3), the search constraints matrix, find and only to serve as second yuan of two tuples and do not serve as the base vector of first yuan in matrix, such base vector is called the leaf vector.
(4), in constraint matrix, two tuples that the leaf vector is second yuan are take in search, record first yuan of base vector of this two tuple.This first yuan of base vector is called the precursor vector of leaf vector.
(5), the precursor vector of search leaf vector precursor vector, until the vector searched is root vector, process stops search.The path of record from the leaf vector to root vector
.
(6), that base vectors all on path is carried out to multidimensional is compound, obtains very big gradient vectors all in target area
, wherein
,
,
With
Mean base vector
Position in vector bunch is the
Row, the
Row; Symbol
For compound, composite vector
With
For
For compound continuously; Very big gradient vector
Dimension be
, gradient is
, obtain people's face section maximum constrained characteristic information.
(7), take the vector bunch all very big gradient vector obtain the composite gradient vector of people's face section as element merges
, wherein
Quantity for very big gradient vector;
The dimension of composite gradient vector is designated as
, gradient is designated as
.
3, the recognition of face stage
The fundamental purpose in this stage is the maximum constrained characteristic information of each one face of coupling people's face to be identified and database storage, identifies people's face identity.
The quantity of the facial image of storing in face database is
, its composite gradient vector is respectively
, dimension is respectively
, gradient is
, the composite gradient vector of people's face to be identified, dimensional information and gradient information are respectively
,
With
.Calculate
With
, wherein
, statistics
With
Minimum value, can judge people's face identity of testing image.
Beneficial effect of the present invention: can find out by the base vector quadrature sampling, the demarcation of base vector has covered the key feature zone of people's face section substantially, suppressed non-critical information, thereby realize in the biological vision system the effective extraction to the obvious characteristic distribution constraint information of target to be identified, at environmental change and people's face to be identified during in different condition, the obvious characteristic distribution constraint information of the same target calibrated by base vector has kept stability preferably, under the condition of environmental change and target Self-variation, the method has shown good robustness.The data volume of the obvious characteristic distribution constraint information of the people's face simultaneously calibrated due to composite gradient vector is lower, makes the method have higher recognition speed and stronger adaptive faculty.The present invention is based on the thought of biology " self-adaptation " feature constraint identification and proposes; The composite gradient vector recognition methods is to be based upon on the general frame of subspace human face analysis, utilizing the distribution of people's face information entropy that the global structure information of people's face section is carried out to subregion cuts apart and screening, then characteristic sub-areas is carried out to quadrature sampling, vector of samples is fused into to vector bunch, base vector is carried out to multidimensional is compound obtains the most obvious feature distribution constraint information.The method is by extraction and the identification of obvious characteristic distribution constraint information that people's face to be identified is characterized, overcome the impact on recognition of face of property field rotation, intensity of illumination variation and multi-pose, multiple expression, the recognition speed that tool is higher and good robust performance.
Embodiment
Below in conjunction with embodiment, the present invention is illustrated.
Below take the CMU-PIE face database as example, the implementation process of this method is described.The face-image of 41368 different attitudes that this face database comprises 68 bit test personnel, different light and different expressions. its attitude and illumination variation are to carry out the multi-angle conversion to gather under the strict condition of controlling, people's face sample of tester in this example, everyone has expression, illumination, light and four kinds of different sets of attitude.Implementing procedure is as follows:
1, feature extraction phases
At first, the facial image of four kinds of set of normalization, obtain respectively the target area of each width facial image.
Then, calculate distribution probability and the information entropy of each gray scale rank respective pixel in the target area of each width facial image, the extreme point of the distribution of statistical information entropy and one-dimension information entropy, when the information entropy of usining obtains maximum value in distributing, corresponding gray scale rank are as the segmentation threshold of target area, divide the dispersing character subregion, obtain the face characteristic subregion of four kinds of set.
Finally, the face characteristic subregion is carried out to quadrature sampling, obtain the base vector of people's face target area of four kinds of set.
2, the Vector Fusion stage
At first, the base vector that four kinds of pooled samplings are obtained is set up vector bunch, and demarcates base vector, statistics base vector forerunner and follow-up restriction relation, extraction base vector dimension and gradient information.Demarcation, restriction relation, latitude and the gradient information of base vector in the vector of four kinds of set bunch are used respectively to structure matrix
, constraint matrix
, the dimension matrix
And gradient matrix
Mean.
Then, adopt the maximum constrained feature extracting method to set up very big gradient vector to the vector bunch of four kinds of set
.
Finally, take the vector bunch all very big gradient vector obtain the composite gradient vector of every width people face in four kinds of set as element merges
.
3, the recognition of face stage
Different sets to every personnel is all demarcated 20 times repeatedly, set up the composite gradient vector database, to reduce characteristic information memory capacity and to improve recognition speed, the information bank obtained comprises 68 bit test personnel's composite gradient vector information, and every personnel have that left side is looked, right side is looked, face, low composite gradient vector people face set of looking and looking up five kinds of attitudes.The quantity of the facial image of storing in face database is
, its composite gradient vector is respectively
, dimension is respectively
, gradient is
.
Then image collection illumination invariant, attitude changed carries out Recognition test, finally by attitude and illumination all vicissitudinous image collection carry out Recognition test, each set in test is all tested 20 times repeatedly, test result is as the criterion with average.The test mode number is 8, being characterized as of test
,
With
, calculate
With
, wherein
.When
With
During for minimum value, can judge that testing image is of face database
Open the corresponding people's face identity of figure.
In present case, adopt this method to be identified facial image, obtained 98.3% discrimination.
Claims (1)
1. the face identification method based on composite gradient vector, is characterized in that it comprises feature extraction, Vector Fusion and recognition of face three phases;
(1) this stage is feature extraction phases, and fundamental purpose is to extract the characteristic information of people's face to be identified;
The facial image that normalization is to be identified, obtain target area
, wherein
For the target area that has comprised people's face main information,
For the target area width,
For target area length; Calculate distribution probability and the information entropy of each gray scale rank respective pixel, and the statistical information entropy distributes and the extreme point of one-dimension information entropy; When the information entropy of usining obtains maximum value in distributing, the dispersing character subregion, as the segmentation threshold of target area, is divided in corresponding gray scale rank, and the edge pixel point of dispersing character subregion is singular point; The quadrature sampling target area
Interior base vector
, using the pixel at width center as starting point, search for vertically downward singular point, and first singular point that will search is set up first base vector as terminal
With base vector
Terminal as starting point, respectively vertical left, under, right counter clockwise direction searches for next singular point, and set up respectively three follow-up vectors of first base vector using first singular point of searching on each direction as terminal, until singular point is not found at the edge of target area T yet, do not set up the base vector of this direction; Using all base vectors of setting up terminal as starting point more respectively vertical left, under, right, on counter clockwise direction search singular point, and first singular point that will search is set up respectively base vector as terminal, each singular point only need be searched for three directions, its starting point direction is ignored, and until target area
Edge do not find yet singular point, do not set up the base vector of this direction; When all directions of all starting points have all arrived target area
Edge all search for less than singular point, stop search, now sampled and obtained all base vectors in target area;
For the base vector in target area
, wherein
For integer,
With
Mean base vector
Position in vector bunch is the
Row, the
Row, the content of vectorial bunch refers to the Vector Fusion stage,
,
, base vector
Dimension is designated as
, gradient is designated as
, work as base vector
Terminal be base vector
Starting point, claim
For
The precursor vector,
For
Follow-up vector, be called the root vector of vector bunch without the base vector of forerunner's vector, be called the leaf vector of vectorial bunch, base vector without the base vector of follow-up vector
Be called the root vector of vectorial bunch;
(2) this stage is the Vector Fusion stage, and fundamental purpose is merged the characteristic information of people's face to be identified, obtains its maximum constrained characteristic information;
The base vector that sampling is obtained is set up vector bunch
, wherein
The set of base vector,
Be the set of mutual relationship between base vector, and demarcate base vector, record base vector forerunner and follow-up restriction relation, extract base vector dimension and gradient information; Demarcation, restriction relation, latitude and the gradient information of base vector in vector bunch used respectively to structure matrix
, constraint matrix
, the dimension matrix
And gradient matrix
Mean;
Wherein
For integer,
With
Mean base vector
Position in vector bunch is the
Row, the
Row,
,
,
,
,
,
In structure matrix, the element of the first row
For the root vector of vectorial bunch, the element of the second row is base vector
,
With
, be all
Follow-up vector, and take the subvector bunch that the three is root vector
,
,
Order in vector bunch be respectively vertical left, under, right counter clockwise direction, base vector
, by counter clockwise direction difference composite vector
,
With
Obtain its follow-up vector
,
,
, of structure matrix
Row, the
The base vector of row
Three follow-up vectors are arranged
,
With
, and the three in vector bunch, be according to vertical left, under, right, on counter clockwise direction arrange, have so
,
,
In constraint matrix, element is two tuples, and the unit of two tuples is base vectors, two tuples
First yuan
It is second yuan
Precursor vector,
Be
Follow-up vector; In dimension matrix and gradient matrix, corresponding to of structure matrix
Row, the
The base vector of row
, of dimension matrix
Row, the
The element of row is base vector
Dimension
, of gradient matrix
Row, the
The element of row is base vector
Gradient
The search constraints matrix, searching is only served as second yuan of two tuples and is not served as the base vector of first yuan in matrix, and such base vector is called the leaf vector; In constraint matrix, two tuples that the leaf vector is second yuan are take in search, record first yuan of base vector of this two tuple, and this first yuan of base vector is called the precursor vector of leaf vector; The precursor vector of search leaf vector precursor vector, until the vector searched is root vector, the process that stops search, the path of record from the leaf vector to root vector
Base vectors all on path is carried out to multidimensional compound, obtain very big gradient vectors all in target area
, wherein
,
,
,
With
Mean base vector
Position in vector bunch is the
Row, the
Row, symbol
For compound, composite vector
With
For
,
For compound continuously, very big gradient vector
Dimension be
, gradient is
, obtain people's face section maximum constrained characteristic information; Take the vector bunch all very big gradient vector obtain the composite gradient vector of people's face section as element merges
, wherein
For the quantity of very big gradient vector,
,
, the dimension of composite gradient vector is designated as
, gradient is designated as
(3) this stage behaviour face cognitive phase, fundamental purpose is the maximum constrained characteristic information of each one face of coupling people's face to be identified and database storage, identifies people's face identity;
The quantity of the facial image of storing in face database is
, its composite gradient vector is respectively
, dimension is respectively
, gradient is
, the composite gradient vector of people's face to be identified, dimensional information and gradient information are respectively
,
With
, calculate
With
, wherein
, statistics
With
Minimum value, can judge people's face identity of testing image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110259963 CN102324022B (en) | 2011-09-05 | 2011-09-05 | Composite gradient vector-based face recognition method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201110259963 CN102324022B (en) | 2011-09-05 | 2011-09-05 | Composite gradient vector-based face recognition method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102324022A CN102324022A (en) | 2012-01-18 |
CN102324022B true CN102324022B (en) | 2013-03-20 |
Family
ID=45451761
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201110259963 Expired - Fee Related CN102324022B (en) | 2011-09-05 | 2011-09-05 | Composite gradient vector-based face recognition method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102324022B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016073105A1 (en) * | 2014-11-06 | 2016-05-12 | Intel Corporation | Face recognition using gradient based feature analysis |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10248848B2 (en) | 2012-03-13 | 2019-04-02 | Nokia Technologies Oy | Method and apparatus for improved facial recognition |
CN103544495A (en) * | 2012-07-12 | 2014-01-29 | 浙江大华技术股份有限公司 | Method and system for recognizing of image categories |
CN103246870A (en) * | 2013-04-24 | 2013-08-14 | 重庆大学 | Face identification method based on gradient sparse representation |
CN103544488B (en) * | 2013-11-07 | 2016-04-13 | 湖南创合制造有限公司 | A kind of face identification method and device |
CN105956515B (en) * | 2016-04-20 | 2019-04-23 | 西安电子科技大学 | Solid-EO-1 hyperion face identification method based on pole imaging |
CN108229314B (en) * | 2017-11-28 | 2021-05-04 | 深圳市商汤科技有限公司 | Target person searching method and device and electronic equipment |
CN109977930B (en) * | 2019-04-29 | 2021-04-02 | 中国电子信息产业集团有限公司第六研究所 | Fatigue driving detection method and device |
CN114463886A (en) * | 2022-01-29 | 2022-05-10 | 海信集团控股股份有限公司 | Intelligent bookshelf and book management method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2003596A2 (en) * | 2007-06-15 | 2008-12-17 | Mitsubishi Electric Corporation | Computer implemented method for constructing classifier from training data detecting moving objects in test data using classifier |
CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
CN101777131A (en) * | 2010-02-05 | 2010-07-14 | 西安电子科技大学 | Method and device for identifying human face through double models |
-
2011
- 2011-09-05 CN CN 201110259963 patent/CN102324022B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2003596A2 (en) * | 2007-06-15 | 2008-12-17 | Mitsubishi Electric Corporation | Computer implemented method for constructing classifier from training data detecting moving objects in test data using classifier |
CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
CN101777131A (en) * | 2010-02-05 | 2010-07-14 | 西安电子科技大学 | Method and device for identifying human face through double models |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016073105A1 (en) * | 2014-11-06 | 2016-05-12 | Intel Corporation | Face recognition using gradient based feature analysis |
US9384385B2 (en) | 2014-11-06 | 2016-07-05 | Intel Corporation | Face recognition using gradient based feature analysis |
Also Published As
Publication number | Publication date |
---|---|
CN102324022A (en) | 2012-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102324022B (en) | Composite gradient vector-based face recognition method | |
CN110263774B (en) | A kind of method for detecting human face | |
CN106650693B (en) | Multi-feature fusion recognition algorithm for face comparison | |
Wang et al. | View-based discriminative probabilistic modeling for 3D object retrieval and recognition | |
CN104008395B (en) | A kind of bad video intelligent detection method based on face retrieval | |
CN102637251A (en) | Face recognition method based on reference features | |
Song et al. | Aggregating minutia-centred deep convolutional features for fingerprint indexing | |
Hu et al. | Cross dataset person re-identification | |
CN102663447B (en) | Cross-media searching method based on discrimination correlation analysis | |
CN103514443A (en) | Single sample face identification transfer learning method based on LPP feature extraction | |
CN102609693A (en) | Human face recognition method based on fuzzy two-dimensional kernel principal component analysis | |
CN103617609B (en) | Based on k-means non-linearity manifold cluster and the representative point choosing method of graph theory | |
CN102831396A (en) | Computer face recognition method | |
Ahmed | Recent review on image clustering | |
CN103345621A (en) | Face classification method based on sparse concentration index | |
CN103077405A (en) | Bayes classification method based on Fisher discriminant analysis | |
CN103927561B (en) | Behavior recognition method based on probability fusion and dimensionality reduction technology | |
Li et al. | Action recognition with spatio-temporal augmented descriptor and fusion method | |
Wang et al. | Facial expression recognition based on tensor local linear discriminant analysis | |
Huang et al. | Age-puzzle facenet for cross-age face recognition | |
Gu et al. | Face verification technology based on FaceNet similarity recognition network | |
CN111464459A (en) | Network flow characteristic extraction method based on principal component analysis and linear discriminant analysis | |
Wang et al. | Face recognition based on improved support vector clustering | |
CN110458002A (en) | A kind of lightweight fast human face recognition | |
Taffar et al. | Local appearance modeling for objects class recognition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130320 Termination date: 20130905 |