CN106446806A - Semi-supervised face identification method and system based on fuzzy membership degree sparse reconstruction - Google Patents
Semi-supervised face identification method and system based on fuzzy membership degree sparse reconstruction Download PDFInfo
- Publication number
- CN106446806A CN106446806A CN201610811185.3A CN201610811185A CN106446806A CN 106446806 A CN106446806 A CN 106446806A CN 201610811185 A CN201610811185 A CN 201610811185A CN 106446806 A CN106446806 A CN 106446806A
- Authority
- CN
- China
- Prior art keywords
- fuzzy membership
- matrix
- test sample
- class label
- regard
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/513—Sparse representations
Abstract
The invention discloses a semi-supervised face identification method and system based on fuzzy membership degree sparse reconstruction. The identification method comprises steps of obtaining a face image data set which comprises a training sample subset and a test sample subset; according to known category labels of training samples, obtaining a fuzzy membership degree initialization matrix of the training sample subset; according to residual errors of test samples about the category labels, obtaining a fuzzy membership degree initialization matrix of the test sample subset; obtaining a fuzzy membership degree initialization matrix of the face image data set about the category labels further; solving sparse coefficients of all test samples, and obtaining a sparse solution matrix of the test sample subset; according to the fuzzy membership degree initialization matrixes and the sparse solution matrix of the test sample subset, iteratively solving a fuzzy membership degree matrix of the updated face image data set about the category labels; and obtaining a category label corresponding to the maximal membership degree of each test sample to complete classification.
Description
Technical field
The invention belongs to area of pattern recognition, more particularly, to a kind of semi-supervised face based on the sparse reconstruct of fuzzy membership
Recognition methods and system.
Background technology
With the fast development of computer technology and image processing techniques, recognition of face is widely applied due to it, obtains
The concern of numerous researchers, becomes one of modern mode identification technology research importance.
However, mark facial image cumbersome and time consuming it is contemplated that unmarked facial image simple and easy to get, can be abundant
Using the label information in mark facial image and the characteristic information in unmarked facial image, unmarked facial image is entered
Row classification, this Learning Scheme is referred to as semi-supervised study.
It is one type Learning Scheme wherein based on the semi-supervised learning of figure, represented based on the semi-supervised learning point of use of figure
Data, while represent the weight between data and respective classes.The probability that the larger class of data weights belongs to such is more than other classes,
Image classification can be carried out on the basis of this assumption, and manifold be met based on the semi-supervised learning of figure it is assumed that and the basis in data
Levying lower dimensional space, to realize effect fine.
The effect of the semi-supervised learning based on figure depend primarily on data between similarity definition.Most common of which
Similarity Measures are the measures based on gaussian kernel function of theorem in Euclid space, have LNP (linear based on the scheme of this method
Neighbour propagates) can be gone to calculate linear space by its approximated linear reconstruction of k neighbour based on this hypothesis-each data
Data similarity；SIS (sparse induction similarity measure) goes to calculate sparse similarity, SIS and LNP phase using rarefaction representation technology
Do not need pre-defined neighbour's number than when using sparse complete coding excessively.
The problem of parameter selection of KNN (k neighbour) and ε-ball can be avoided in view of rarefaction representation technology.SIS is (sparse
Induction similarity measure) it is used for marking propagation using the method for sparse induction similarity measurement, and redefine similarity measurement side
Method, this method goes sparse linear reconstruct data itself using the remaining data of certain data itself, by the sparse coefficient obtaining
As data between similarity；LPSN (mark propagates sparse neighbour) goes to obtain using rarefaction representation technology in luv space
Obtain sparse coding, and sparse coding is used in the reconstruct of data label；MLRR (manifold low-rank representation) uses enhanced low-rank
Sparse data representation method carries out realizing data similarity measurement.
Above several method can fully obtain the local and overall feature between data, but data is sparse heavy
Structure residual error is ignored, and the residual error of data is the key factor of the semi-supervised learning based on rarefaction representation, so can drop
The classification and recognition of low semi-supervised method.
Content of the invention
In order to solve the shortcoming of prior art, the present invention provides a kind of semi-supervised people based on the sparse reconstruct of fuzzy membership
Face recognition method and system.The inventive method employs membership function and iterates to calculate test data using sparse reconstructed residual
Degree of membership, and maintain the manifold structure of data.
A kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership, including：
Obtain face image data collection；Face image data collection includes training sample subset and test sample subset；Training
The sample of sample set is facial image known to class label, and the sample of test sample subset is the unknown face of class label
Image；
Class label according to known to training sample, obtains the fuzzy membership initialization matrix of training sample subset；Meter
Calculate the reconstructed residual with regard to class label for the test sample, obtain the fuzzy membership initialization matrix of test sample subset；Enter one
Step obtains face image data collection and initializes matrix with regard to the fuzzy membership of class label；
Solve the sparse coefficient of all test samples, and then obtain the sparse dematrix of test sample subset；
Initialize the sparse dematrix of matrix and test sample subset according to fuzzy membership, iterative updates face
Image data set is with regard to the fuzzy membership matrix of class label；
Obtain the classification that the class label corresponding to the maximum membership degree of each test sample can complete test sample.
The present invention using fuzzy membership function and puts into the reconstructed residual usually ignored in membership function and carries out
Calculate the degree of membership of test sample, manifold space implicitly keep data structure, take into full account local between data and
Global characteristics, improve the image recognition degree of accuracy.
During obtaining the fuzzy membership initialization matrix of training sample subset, predefine classification mark first
Sign, then whether the training sample in training of judgement sample set belongs to predetermined class label, if so, then accordingly trains
Sample fuzzy membership is 1；Otherwise, corresponding training sample fuzzy membership is 0.
In the fuzzy membership initialization matrix of test sample subset, test sample is with regard to the fuzzy membership of label classification
Make the ratio of business for two numbers；Wherein, divisor is the inverse with regard to respective labels classification residual error for the test sample, and dividend is test
Sample with regard to all label classifications reciprocal tired of reconstructed residual and.
During iterative renewal face image data collection is with regard to the fuzzy membership matrix of class label,
According to the sparse dematrix of test sample subset, calculate the reconstructed residual with regard to class label for each test sample；
According to each test sample with regard to the reconstructed residual of class label, obtain fuzzy membership after the renewal of test sample
Degree, obtains the fuzzy membership matrix after test sample subset updates further；
Training sample subset fuzzy membership initialization matrix constant, merge training sample subset fuzzy membership and
Fuzzy membership matrix after the renewal of test sample subset, obtains face image data collection with regard to the mould after the renewal of class label
Paste subordinated-degree matrix.
The reconstructed residual usually ignored is put into the degree of membership carrying out calculating test sample in membership function by the present invention,
Implicitly keep data structure in manifold space, take into full account the local between data and global characteristics.The inventive method uses
Membership function simultaneously iterates to calculate the degree of membership of test data using sparse reconstructed residual, and maintains the manifold knot of data
Structure.
During iterative renewal face image data collection is with regard to the fuzzy membership matrix of class label, if full
The sufficient fuzzy membership matrix that this is asked for makees the institute in the matrix obtaining after difference with the front fuzzy membership matrix once asked for
The absolute value having element is respectively less than predetermined threshold value, then iteration terminates；Or iterations is more than maximum iterations, then iteration
Terminate.
A kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership, including recognition of face server, institute
State recognition of face server to include：
Face image data collection acquisition module, it is used for obtaining face image data collection；Face image data collection includes instructing
Practice sample set and test sample subset；The sample of training sample subset is facial image, test sample known to class label
The sample of subset is the unknown facial image of class label；
Fuzzy membership initializes matrix initialisation module, and it is used for class label according to known to training sample, obtains
The fuzzy membership initialization matrix of training sample subset；Calculate the reconstructed residual with regard to class label for the test sample, obtain and survey
The fuzzy membership initialization matrix of examination sample set；Obtain the fuzzy person in servitude with regard to class label for the face image data collection further
Genus degree initializes matrix；
Sparse solution Matrix Solving module, it is used for solving the sparse coefficient of all test samples, and then obtains test sample
The sparse dematrix of subset；
Fuzzy membership matrix update module, it is used for initializing matrix and test sample subset according to fuzzy membership
Sparse dematrix, iterative update face image data collection with regard to class label fuzzy membership matrix；
Class label acquisition module, it for obtaining the class label corresponding to maximum membership degree of each test sample is
The classification of test sample can be completed.
Described fuzzy membership initializes the fuzzy membership initialization that matrix initialisation module includes training sample subset
Matrix initialisation module, it is used for predefining class label first, and then the training sample in training of judgement sample set is
No belong to predetermined class label, if so, then corresponding training sample fuzzy membership is 1；Otherwise, corresponding training sample mould
Paste degree of membership is 0.
The fuzzy membership that described fuzzy membership initialization matrix initialisation module also includes test sample subset is initial
Change matrix initialisation module, it is used for for two numbers asking for the fuzzy membership with regard to label classification for the test sample as business；Wherein,
Divisor is the inverse with regard to respective labels classification residual error for the test sample, and dividend is the weight with regard to all label classifications for the test sample
Reciprocal tired of structure residual error and.
Described fuzzy membership matrix update module is additionally operable to the sparse dematrix according to test sample subset, calculates each
Test sample is with regard to the reconstructed residual of class label；
According to each test sample with regard to the reconstructed residual of class label, obtain fuzzy membership after the renewal of test sample
Degree, obtains the fuzzy membership matrix after test sample subset updates further；
Training sample subset fuzzy membership initialization matrix constant, merge training sample subset fuzzy membership and
Fuzzy membership matrix after the renewal of test sample subset, obtains face image data collection with regard to the mould after the renewal of class label
Paste subordinated-degree matrix.
Described display device is also included based on the semi-supervised face identification system of the sparse reconstruct of fuzzy membership, described display
Device is connected with recognition of face server；Described display device is used for showing corresponding to the maximum membership degree of each test sample
Class label.
The invention has the beneficial effects as follows：
The present invention takes full advantage of the characteristics of image of global and local and is used for obscuring sparse degree of membership by sparse residual error
In calculating.Because face digital picture has manifold structure in itself, reconstructed residual is combined with data fuzzy membership solution
Data membership degree can by reconstructed residual consider mark propagate in it is considered to reconstructed residual mark than only consider sparse coding or
Person only considers that the mark propagation of data similarity has more judgement information, therefore can improve the image recognition degree of accuracy.
Brief description
Fig. 1 is the semi-supervised face identification method flow chart based on the sparse reconstruct of fuzzy membership of the present invention；
Fig. 2 is initialization sample data fuzzy membership matrix construction schematic diagram；
Fig. 3 is being shown based on semi-supervised face identification system embodiment one structure of the sparse reconstruct of fuzzy membership of the present invention
It is intended to；
Fig. 4 is being shown based on semi-supervised face identification system embodiment two structure of the sparse reconstruct of fuzzy membership of the present invention
It is intended to.
Specific embodiment
The present invention will be further described with embodiment below in conjunction with the accompanying drawings：
The invention provides a kind of semi-supervised recognition methods of the face digital picture based on figure.Traditional based on figure
In semi-supervised method, or only using simple similarity measurement method, or at one point when execution flag is propagated
Two kinds of architectural features of global and local of data are simply considered in generic task.Scheme proposed by the present invention uses fuzzy membership
Function and put into the reconstructed residual usually ignored carries out in membership function calculating the degree of membership of test sample, empty in manifold
Between in implicitly keep data structure, take into full account the local between data and global characteristics.
The thinking of the semi-supervised recognition methods of the face digital picture based on figure of the present invention is：
First, initialize the subordinated-degree matrix with regard to all corresponding class labels of all samples；
Then calculate the sparse reconstructed residual with regard to all categories label for the test sample；
Re-define the method for iterative modifications degree of membership and obtain the person in servitude with regard to all categories for the test sample through iterative
Genus degree matrix, obtains the classification that the class label corresponding to the maximum membership degree of each test sample can complete test sample.
Fig. 1 is the semi-supervised face identification method flow chart based on the sparse reconstruct of fuzzy membership of the present invention, as schemed institute
The semi-supervised face identification method based on the sparse reconstruct of fuzzy membership showing, including：
Step 1：Obtain face image data collection；Face image data collection includes training sample subset and test specimens book
Collection；The sample of training sample subset is facial image known to class label, and the sample of test sample subset is for class label not
The facial image known.
In specific implementation process：
Obtain face image data collection X=[XlXu], each sample is xi∈Rq{ i=1,2 ..., n }, n is sample
Number, sample dimension is q, whereinFor training sample set,For test sample
Collection, nlFor training sample number, nuFor test sample number；
Step 2：Class label according to known to training sample, obtains the fuzzy membership initialization square of training sample subset
Battle array；Calculate the reconstructed residual with regard to class label for the test sample, obtain the fuzzy membership initialization matrix of test sample subset；
The fuzzy membership obtaining face image data collection further with regard to class label initializes matrix.
Fig. 2 is initialization sample data fuzzy membership matrix construction schematic diagram, and concrete grammar is：
(1) respectively fuzzy membership matrix is initialized to training sample and test sample first, define category set C=
{ 1,2 ..., c }, and training sample set fuzzy membership matrixWherein for training sample set data xi
∈Xl, c is the positive integer more than 1；The fuzzy membership matrix M of training sample can be initialized using following formula0l, wherein, mij0For
Fuzzy membership matrix M0lElement：
Wherein mij0The degree of membership of the jth class of i-th training sample when representing the 0th iteration, p is certain classification.
(2) for test sample collection data xk∈Xu, the fuzzy membership initialization of test sample can be obtained according to following formula
Matrix M0u, wherein, mkj0For fuzzy membership matrix M0lElement：
Wherein mkj0K-th test sample x when representing the 0th iterationkJth class degree of membership, errkj0For k-th
Test sample xkCorresponding jth class sample set XjInitial reconstitution residual error, errkjo=| | xk-Xjα′||2；Jth class sample set XjInterior
The class label of sample is jth class；α ' is k-th test sample xkWith respect to jth class sample set XjSparse coefficient.
In specific implementation process, solve k-th test sample x using following formulakWith respect to jth class sample set XjSparse
Coefficient：
min||xk-Xjα′||2+γ′||α′||1(k∈1,2,....nu；j∈C)
Wherein γ ' represents that the constraint to α ' prevents over-fitting coefficient, and γ ' is constant coefficient.
(3) merge M0lAnd M0uThe fuzzy membership obtaining face image data collection with regard to class label initializes matrix
M0：
Wherein
Step 3：Solve the sparse coefficient of all test samples, and then obtain the sparse dematrix of test sample subset.
In specific implementation process, to each test sample xk, solve sparse coefficient using following formula：
min||xk-Xα||2+γ||α1||(k∈1,2,...,nu)
Wherein xkRemoved from X, sparse coefficient α of all acquisitions is formed a coefficient matrix Scoeff ∈ Rn×n,
Wherein factor alpha ∈ Rn×1Each sample in expression X is to xkContribution rate；γ represents that the constraint to α prevents over-fitting coefficient, γ
For constant coefficient.
Step 4：Initialize the sparse dematrix of matrix and test sample subset according to fuzzy membership, iterative is more
New facial image data set is with regard to the fuzzy membership matrix of class label.
In implementation process, the fuzzy membership of iterative modifications test sample, wherein t be iterations, t be more than or wait
In 1 positive integer.
Calculate test sample x using coefficient matrix Scoeff obtained abovekThe reconstructed residual related to each class.
K-th test sample xkCorresponding jth class sample set XjReconstructed residual errkjt：
errkjt=| | xk-XjWjt-1α2||
Wherein Wjt-1Represent x when the t-1 time iterationkThe degree of membership of corresponding jth class；XjFor jth class sample set.
According to reconstructed residual err derived abovekjtGo to optimize following formula：
The degree of membership of the test sample after being updated：
Update subordinated-degree matrix：
Using mkjtGo to update Mtu, Mtl=M0l, new subordinated-degree matrix can be obtained
Until meeting Mt-Mt-1The absolute value of all elements in the matrix obtaining afterwards is respectively less than predetermined threshold value, then iteration knot
Bundle；Or current iteration number of times is more than maximum iterations, then iteration terminates.
Step 5：Obtain the class label corresponding to the maximum membership degree of each test sample and can complete test sample
Classification.
The subordinated-degree matrix of the test sample that the t time iteration is obtained carries out processing according to following formula and obtains final classification knot
Really.
To verify this face identification method of the present invention below with specific experiment：
, in the artificial classification of each of data set being used, at this time taking the YaleB data set being used as a example
C=1,2 ..., 38 }.It is 5 times of cross validations used in the present invention, sample data set is divided into uniformly at random 5 parts, often
Secondary take a copy of it as test sample collection Xu, remaining four parts then as Xl, experiment can be repeated 5 times.
Verified with the view data in human face data collection (YaleB), in YaleB face database, comprised 38 people,
Have the sample of 2414 known types, pixel is 32*32.Experiment adopts 5 times of cross validations, will be equal at random for all data
Even be divided into 10 parts, choose every time one group as test data,, as training data, experiment is repeated 5 times and takes 5 times average for remaining
, as final recognition accuracy, accuracy rate is as shown in table 1 for value.
Table 1 recognition methods is compared
The present invention is respectively to be 10% using rate, in the case of 20%, 30%, 40%, 50% and 60%, to compare the present invention
This recognition methods and LNP, SIS, LPSN and MLRR recognition methods discrimination, this identification side of the present invention as can be seen from Table 1
Method, compared with other recognition methods, has higher discrimination.Wherein, LNP is Local Neighborhood Patterns's
Referred to as, full name is local neighborhood mode method；SIS is Sparsity Induced Similarity measure for
The abbreviation of label propagation, full name is the mark transmission method based on sparse similarity measure；LPSN is Label
The abbreviation of propagation through sparse neighborhood, full name is sparse next-door neighbour's mark transmission method；MLRR
For the abbreviation of manifold low-rank representation, full name is manifold low-rank representation method.
The present invention takes full advantage of the characteristics of image of global and local and is used for obscuring sparse degree of membership by sparse residual error
In calculating.Because face digital picture has manifold structure in itself, reconstructed residual is combined with data fuzzy membership solution
Data membership degree can by reconstructed residual consider mark propagate in it is considered to reconstructed residual mark than only consider sparse coding or
Person only considers that the mark propagation of data similarity has more judgement information, therefore can improve the image recognition degree of accuracy.
Fig. 3 is being shown based on semi-supervised face identification system embodiment one structure of the sparse reconstruct of fuzzy membership of the present invention
It is intended to.Semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as shown in Figure 3, including recognition of face service
Device, described recognition of face server includes：Face image data collection acquisition module, fuzzy membership initialization matrix initialisation mould
Block, sparse solution Matrix Solving module, fuzzy membership matrix update module and class label acquisition module.
(1) face image data collection acquisition module, it is used for obtaining face image data collection；Face image data collection includes
Training sample subset and test sample subset；The sample of training sample subset is facial image, test specimens known to class label
The sample of this subset is the unknown facial image of class label.
(2) fuzzy membership initialization matrix initialisation module, it is used for class label according to known to training sample, obtains
Obtain the fuzzy membership initialization matrix of training sample subset；Calculate the reconstructed residual with regard to class label for the test sample, obtain
The fuzzy membership initialization matrix of test sample subset；Obtain face image data collection further fuzzy with regard to class label
Degree of membership initializes matrix.
Fuzzy membership initializes the fuzzy membership initialization matrix that matrix initialisation module includes training sample subset
Initialization module, it is used for predefining class label first, and then whether the training sample in training of judgement sample set belongs to
In predetermined class label, if so, then corresponding training sample fuzzy membership is 1；Otherwise, corresponding training sample obscures and is subordinate to
Genus degree is 0.
Fuzzy membership initializes the fuzzy membership initialization square that matrix initialisation module also includes test sample subset
Battle array initialization module, it is used for for two numbers asking for the fuzzy membership with regard to label classification for the test sample as business；Wherein, divisor
For test sample with regard to respective labels classification residual error inverse, dividend be test sample residual with regard to the reconstruct of all label classifications
Reciprocal tired of difference and.
(3) sparse solution Matrix Solving module, it is used for solving the sparse coefficient of all test samples, and then obtains test specimens
The sparse dematrix of this subset.
(4) fuzzy membership matrix update module, it is used for initializing matrix and test sample according to fuzzy membership
The sparse dematrix of subset, iterative updates the fuzzy membership matrix with regard to class label for the face image data collection.
Fuzzy membership matrix update module is additionally operable to the sparse dematrix according to test sample subset, calculates each test
Sample is with regard to the reconstructed residual of class label；
According to each test sample with regard to the reconstructed residual of class label, obtain fuzzy membership after the renewal of test sample
Degree, obtains the fuzzy membership matrix after test sample subset updates further；
Training sample subset fuzzy membership initialization matrix constant, merge training sample subset fuzzy membership and
Fuzzy membership matrix after the renewal of test sample subset, obtains face image data collection with regard to the mould after the renewal of class label
Paste subordinated-degree matrix.
(5) class label acquisition module, it is used for obtaining the classification mark corresponding to maximum membership degree of each test sample
Label can complete the classification of test sample.
Fig. 4 is being shown based on semi-supervised face identification system embodiment two structure of the sparse reconstruct of fuzzy membership of the present invention
It is intended to.Semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as shown in Figure 4, including recognition of face server
And display device, described display device is connected with recognition of face server；
Recognition of face server includes：Face image data collection acquisition module, fuzzy membership initialization matrix initialisation
Module, sparse solution Matrix Solving module, fuzzy membership matrix update module and class label acquisition module.
Display device is used for showing the class label corresponding to maximum membership degree of each test sample.Wherein, display dress
Put including various types of display screens.
The present invention takes full advantage of the characteristics of image of global and local and is used for obscuring sparse degree of membership by sparse residual error
In calculating.Because face digital picture has manifold structure in itself, reconstructed residual is combined with data fuzzy membership solution
Data membership degree can by reconstructed residual consider mark propagate in it is considered to reconstructed residual mark than only consider sparse coding or
Person only considers that the mark propagation of data similarity has more judgement information, therefore can improve the image recognition degree of accuracy.
One of ordinary skill in the art will appreciate that realizing all or part of flow process in above-described embodiment method, it is permissible
Instruct related hardware to complete by computer program, described program can be stored in computer read/write memory medium
In, this program is upon execution, it may include as the flow process of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random
AccessMemory, RAM) etc..
Although the above-mentioned accompanying drawing that combines is described to the specific embodiment of the present invention, not model is protected to the present invention
The restriction enclosed, one of ordinary skill in the art should be understood that on the basis of technical scheme, and those skilled in the art are not
Need to pay the various modifications that creative work can make or deformation still within protection scope of the present invention.
Claims (10)
1. a kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership is it is characterised in that include：
Obtain face image data collection；Face image data collection includes training sample subset and test sample subset；Training sample
The sample of subset is facial image known to class label, and the sample of test sample subset is the unknown face figure of class label
Picture；
Class label according to known to training sample, obtains the fuzzy membership initialization matrix of training sample subset；Calculate and survey
The sample originally reconstructed residual with regard to class label, obtains the fuzzy membership initialization matrix of test sample subset；Further
Initialize matrix to face image data collection with regard to the fuzzy membership of class label；
Solve the sparse coefficient of all test samples, and then obtain the sparse dematrix of test sample subset；
Initialize the sparse dematrix of matrix and test sample subset according to fuzzy membership, iterative updates facial image
Data set is with regard to the fuzzy membership matrix of class label；
Obtain the classification that the class label corresponding to the maximum membership degree of each test sample can complete test sample.
2. a kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership as claimed in claim 1, its feature
It is, during obtaining the fuzzy membership initialization matrix of training sample subset, predefine class label first, so
Whether the training sample in training of judgement sample set belongs to predetermined class label afterwards, if so, then corresponding training sample
Fuzzy membership is 1；Otherwise, corresponding training sample fuzzy membership is 0.
3. a kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership as claimed in claim 1, its feature
It is, in the fuzzy membership initialization matrix of test sample subset, test sample with regard to the fuzzy membership of label classification is
Two numbers make the ratio of business；Wherein, divisor is the inverse with regard to respective labels classification residual error for the test sample, and dividend is test specimens
This with regard to reciprocal tired of the reconstructed residual of all label classifications and.
4. a kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership as claimed in claim 1, its feature
It is, during iterative renewal face image data collection is with regard to the fuzzy membership matrix of class label,
According to the sparse dematrix of test sample subset, calculate the reconstructed residual with regard to class label for each test sample；
According to each test sample with regard to the reconstructed residual of class label, obtain fuzzy membership after the renewal of test sample, enter
One step obtains the fuzzy membership matrix after test sample subset updates；
The fuzzy membership initialization matrix of training sample subset is constant, merges fuzzy membership and the test of training sample subset
Fuzzy membership matrix after sample set renewal, obtains face image data collection with regard to the fuzzy person in servitude after the renewal of class label
Genus degree matrix.
5. a kind of semi-supervised face identification method based on the sparse reconstruct of fuzzy membership as claimed in claim 1, its feature
It is, during iterative renewal face image data collection is with regard to the fuzzy membership matrix of class label, if meeting
The fuzzy membership matrix that this is asked for and the front fuzzy membership matrix once asked for are made all in the matrix obtaining after difference
The absolute value of element is respectively less than predetermined threshold value, then iteration terminates；Or iterations is more than maximum iterations, then iteration knot
Bundle.
6. a kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership is it is characterised in that include recognition of face
Server, described recognition of face server includes：
Face image data collection acquisition module, it is used for obtaining face image data collection；Face image data collection includes training sample
This subset and test sample subset；The sample of training sample subset is facial image known to class label, test sample subset
Sample be the unknown facial image of class label；
Fuzzy membership initializes matrix initialisation module, and it is used for class label according to known to training sample, obtains training
The fuzzy membership initialization matrix of sample set；Calculate the reconstructed residual with regard to class label for the test sample, obtain test specimens
The fuzzy membership initialization matrix of this subset；Obtain the fuzzy membership with regard to class label for the face image data collection further
Initialization matrix；
Sparse solution Matrix Solving module, it is used for solving the sparse coefficient of all test samples, and then obtains test sample subset
Sparse dematrix；
Fuzzy membership matrix update module, it is used for initializing the dilute of matrix and test sample subset according to fuzzy membership
Discongest matrix, iterative updates the fuzzy membership matrix with regard to class label for the face image data collection；
Class label acquisition module, it is used for obtaining the class label corresponding to maximum membership degree of each test sample can be complete
Become the classification of test sample.
7. a kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as claimed in claim 6, its feature
It is, described fuzzy membership initializes the fuzzy membership initialization matrix that matrix initialisation module includes training sample subset
Initialization module, it is used for predefining class label first, and then whether the training sample in training of judgement sample set belongs to
In predetermined class label, if so, then corresponding training sample fuzzy membership is 1；Otherwise, corresponding training sample obscures and is subordinate to
Genus degree is 0.
8. a kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as claimed in claim 6, its feature
It is, described fuzzy membership initializes the fuzzy membership initialization square that matrix initialisation module also includes test sample subset
Battle array initialization module, it is used for for two numbers asking for the fuzzy membership with regard to label classification for the test sample as business；Wherein, divisor
For test sample with regard to respective labels classification residual error inverse, dividend be test sample residual with regard to the reconstruct of all label classifications
Reciprocal tired of difference and.
9. a kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as claimed in claim 6, its feature
It is, described fuzzy membership matrix update module is additionally operable to the sparse dematrix according to test sample subset, calculate each survey
The sample originally reconstructed residual with regard to class label；
According to each test sample with regard to the reconstructed residual of class label, obtain fuzzy membership after the renewal of test sample, enter
One step obtains the fuzzy membership matrix after test sample subset updates；
The fuzzy membership initialization matrix of training sample subset is constant, merges fuzzy membership and the test of training sample subset
Fuzzy membership matrix after sample set renewal, obtains face image data collection with regard to the fuzzy person in servitude after the renewal of class label
Genus degree matrix.
10. a kind of semi-supervised face identification system based on the sparse reconstruct of fuzzy membership as claimed in claim 6, its feature
Be, described display device is also included based on the semi-supervised face identification system of the sparse reconstruct of fuzzy membership, described display dress
Put and be connected with recognition of face server；Described display device is used for showing the class corresponding to maximum membership degree of each test sample
Distinguishing label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610811185.3A CN106446806B (en) | 2016-09-08 | 2016-09-08 | Semi-supervised face identification method based on the sparse reconstruct of fuzzy membership and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610811185.3A CN106446806B (en) | 2016-09-08 | 2016-09-08 | Semi-supervised face identification method based on the sparse reconstruct of fuzzy membership and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106446806A true CN106446806A (en) | 2017-02-22 |
| CN106446806B CN106446806B (en) | 2018-10-12 |
Family
ID=58164609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610811185.3A Active CN106446806B (en) | 2016-09-08 | 2016-09-08 | Semi-supervised face identification method based on the sparse reconstruct of fuzzy membership and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106446806B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107194314A (en) * | 2017-04-20 | 2017-09-22 | 江苏大学 | The fuzzy 2DPCA and fuzzy 2DLDA of fusion face identification method |
| CN112052187A (en) * | 2020-10-13 | 2020-12-08 | 苏州浪潮智能科技有限公司 | Distribution method of test matrix |
| CN112131446A (en) * | 2020-08-19 | 2020-12-25 | 深圳云天励飞技术股份有限公司 | Graph node classification method and device, electronic equipment and storage medium |
| CN114330574A (en) * | 2021-12-31 | 2022-04-12 | 广东泰迪智能科技股份有限公司 | Fuzzy labeling method for pattern recognition |
| CN115131610A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609681A (en) * | 2012-01-12 | 2012-07-25 | 北京大学 | Face recognition method based on dictionary learning models |
| CN102831390A (en) * | 2012-07-02 | 2012-12-19 | 北京科技大学 | Human ear authenticating system and method |
| CN104573726A (en) * | 2015-01-12 | 2015-04-29 | 山东师范大学 | Facial image identification method for reconstructing optimal error combination based on quartering and components |
-
2016
- 2016-09-08 CN CN201610811185.3A patent/CN106446806B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609681A (en) * | 2012-01-12 | 2012-07-25 | 北京大学 | Face recognition method based on dictionary learning models |
| CN102831390A (en) * | 2012-07-02 | 2012-12-19 | 北京科技大学 | Human ear authenticating system and method |
| CN104573726A (en) * | 2015-01-12 | 2015-04-29 | 山东师范大学 | Facial image identification method for reconstructing optimal error combination based on quartering and components |
Non-Patent Citations (1)
| Title |
|---|
| 王永欣: ""基于稀疏表示的人脸特征提取与识别算法研究"", 《中国优秀硕士学位论文全文数据库-信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107194314A (en) * | 2017-04-20 | 2017-09-22 | 江苏大学 | The fuzzy 2DPCA and fuzzy 2DLDA of fusion face identification method |
| CN112131446A (en) * | 2020-08-19 | 2020-12-25 | 深圳云天励飞技术股份有限公司 | Graph node classification method and device, electronic equipment and storage medium |
| CN112131446B (en) * | 2020-08-19 | 2023-11-17 | 深圳云天励飞技术股份有限公司 | Graph node classification method and device, electronic equipment and storage medium |
| CN112052187A (en) * | 2020-10-13 | 2020-12-08 | 苏州浪潮智能科技有限公司 | Distribution method of test matrix |
| CN114330574A (en) * | 2021-12-31 | 2022-04-12 | 广东泰迪智能科技股份有限公司 | Fuzzy labeling method for pattern recognition |
| CN115131610A (en) * | 2022-06-13 | 2022-09-30 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
| CN115131610B (en) * | 2022-06-13 | 2024-02-27 | 西北工业大学 | Robust semi-supervised image classification method based on data mining |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106446806B (en) | 2018-10-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106446806A (en) | Semi-supervised face identification method and system based on fuzzy membership degree sparse reconstruction | |
| Gimelʹfarb | Image textures and Gibbs random fields | |
| Caelli et al. | An eigenspace projection clustering method for inexact graph matching | |
| CN104200240B (en) | A kind of Sketch Searching method based on content-adaptive Hash coding | |
| Shakhnarovich et al. | Nearest-neighbor methods in learning and vision | |
| Liu et al. | DeepEIT: deep image prior enabled electrical impedance tomography | |
| CN108460400A (en) | A kind of hyperspectral image classification method of combination various features information | |
| CN106228027A (en) | A kind of semi-supervised feature selection approach of various visual angles data | |
| Remil et al. | Surface reconstruction with data-driven exemplar priors | |
| Wang et al. | Building correlations between filters in convolutional neural networks | |
| CN107220656A (en) | A kind of multiple labeling data classification method based on self-adaptive features dimensionality reduction | |
| CN112734789A (en) | Image segmentation method and system based on semi-supervised learning and point rendering | |
| Leclaire et al. | A fast multi-layer approximation to semi-discrete optimal transport | |
| CN116229179A (en) | Dual-relaxation image classification method based on width learning system | |
| Antonietti et al. | Agglomeration of polygonal grids using graph neural networks with applications to multigrid solvers | |
| CN114266927A (en) | Unsupervised saliency target detection method, system, equipment and medium | |
| CN108121970A (en) | A kind of recognition methods again of the pedestrian based on difference matrix and matrix measures | |
| CN116524243A (en) | Classification method and device for fossil images of penstones | |
| Zhao et al. | Graph attribute embedding via Riemannian submersion learning | |
| CN107818327A (en) | Classifier training method and device | |
| CN105844299A (en) | Image classification method based on bag of words | |
| Wilson et al. | Consistent topographic surface labelling | |
| Sisouk et al. | Wasserstein Dictionaries of Persistence Diagrams | |
| CN113724017A (en) | Pricing method and device based on neural network, electronic equipment and storage medium | |
| Kovalevsky | Multidimensional cell lists for investigating 3-manifolds |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |