CN101388074A - Human face identification method based on personal ICA base image reconstruction error - Google Patents

Abstract

The invention discloses a face identification method based on an individual ICA based image reconstruction error, considering personal features of the face of an individual person and based on an independent component analytical algorithm, the invention uses a similar internal covariance matrix as a generating matrix to obtain a sub-space of the face features of the individual person, then performs mapping on extract characteristics to each characterized sub-space of a to-be-identified image, performs image reconstruction by the characteristic values, and utilizes an obtained image reconstruction error as a new characteristic vector. Based on the characteristic vector, we utilize a classification and identification algorithm which supports vector machines to realize the face identification. Compared with the existing characterized sub-space method, the method utilizes face features of different persons more sufficiently, and efficiently increases the identification accuracy.

Description

Face identification method based on personal ICA base image reconstruction error

Technical field

The invention belongs to area of pattern recognition, be specifically related to a kind of face identification method, is that extract and a kind of method of discerning about face characteristic in the living things feature recognition field.

Background technology

Living things feature recognition is a technology of utilizing human distinctive biological characteristic to carry out identification, and it provides the identity of a kind of high reliability, high stability to differentiate approach.Biological characteristic has uniqueness, is can measure or discernible physiological property or behavior, is divided into physiological characteristic and behavioural characteristic.Biological recognition system extracts its feature and changes into digital code by biological characteristic is taken a sample, and then these codes is formed its feature templates by different way.When the person of being identified carried out authentication alternately with recognition system, recognition system was obtained its feature and is compared with feature templates, and determining whether coupling, thereby this people is accepted or refuses in decision.Because living things feature recognition is based on individual physiological characteristic or behavioural characteristic,, and be convenient for people to use so it possesses stronger anti-counterfeit capability than traditional recognition method.In all living things feature recognition methods; recognition of face is a branch that paid close attention to by people at present; it is a computer vision and a very active research direction of area of pattern recognition, is widely used in identity identification systems such as national security, public security, the administration of justice, government, finance, commerce, safety check, security personnel.Simultaneously, because the non-infringement of recognition of face is compared and other living things feature recognition method, such as fingerprint recognition, the identification of palm shape, eye iris recognition and voice recognition etc., have characteristics such as direct, friendly, convenient, it also is the most acceptable identity identification methods of people.Compare and other living things feature recognition method, recognition of face has following 2 unique characteristics:

(1) recognition of face does not need people's interoperation, makes it be easier to use, and is particularly suitable for requiring the occasion of hidden implementation;

(2) people's face as a kind of high ubiquity, can contactless collection the important biomolecule feature, can be more directly perceived, verify a people's identity more easily.

Rely on technology such as image understanding, pattern-recognition, computer vision and neural network, the computer face recognition technology has obtained success within the specific limits, and is pushed to application.Such as: the additional clause authentication; The security control of building turnover; The safety detection in important place and monitoring; Authentication in the smart card etc.Simultaneously, because the utilization of Internet resources such as ecommerce, the security control of network also becomes a urgent day by day major issue, the appearance of these problems has also proposed new requirement to authentication, and utilize face recognition technology, just can realize login control, application security use, database security visit and the file encryption etc. of computing machine, to guarantee the security of ecommerce and network.In addition, face recognition technology also can be used in numerous areas such as image library retrieval technique, sense of reality virtual game.International (the International Biometric Group of biological tissue, the report in a biological identification technology market of IBG) doing shows, the gross income in whole biological identification technology market, the whole world in 2007 is 30.126 hundred million dollars, and face recognition technology shared ratio in whole biological identification technology market is 12.9%, simultaneously, this tissue is also predicted biological identification technology market gross income in 2012 will reach 74.077 hundred million dollars, and the shared market share in recognition of face field also has the development trend that increases year by year.

Summary of the invention

The objective of the invention is to: proposed a kind of with the recognition of face new method of image reconstruction error as the differentiation feature, this method will be tested facial image and be carried out projection and reconstruct to each proper subspace by the image construction of ICA base, realize recognition of face with the reconstructed image error as feature.

Technical scheme of the present invention is: based on the independent component analytical algorithm, we have proposed with covariance matrix in the class of individual facial image serves as to produce its face characteristic subspace of ICA base image construction that matrix obtains single people, then image to be identified is shone upon the extraction feature to each proper subspace, and carry out image reconstruction with this eigenwert, at last the image minimal reconstruction error that obtains is used for support vector machine with the image reconstruction error sequence as the characteristics of image parameter of extracting and carries out Classification and Identification.Its specific implementation step is as follows:

Step 1, image pre-service;

(size is w * h) carry out certain pre-service, mainly comprises the normalized of image smoothing and gradation of image and variance to facial image matrix I.

Step 2, obtain single people's training image matrix V _j

Each width of cloth image array I of single people is launched into the vector x of n=w*h dimension by row or column, and go average to handle the vector x and albefaction is handled, make that the variable covariance matrix after the albefaction is a unit matrix, utilize covariance to carry out characteristic value decomposition, i.e. E (xx ^T)=PEP ^T, wherein E is orthogonal matrix E (xx ^T) eigenwert, P is the characteristic of correspondence vector, the albefaction matrix that obtains is:

M＝PE ^-1/2P ^T (1)

Obtain the data after the albefaction:

x＝Mx (2)

At last with all training images of individual with n * s (s is everyone face training image quantity of a people) matrix V _jExpression.Step 3, with training matrix V _jAs producing matrix, extract its ICA base image, and constitute its proper subspace W _jHypothesis according to Bartlett, each width of cloth facial image is mixed by the independently basic image linearity of some statistics, promptly the facial image X that hypothesis is actual can regard as by some implicit separate basic image S and form by hybrid matrix A linear hybrid, just S can be recovered out by separation matrix M.For the sake of simplicity, consider that hybrid matrix A is the situation of square formation, can suppose that then this N width of cloth facial image is by N independently basic image S=[s ₁, s ₂..., s _N] ^T, S ∈ R ^{N * n}Linear hybrid forms, that is:

X＝AS (3)

Wherein each row of X is represented a width of cloth facial image, and each row of S is represented a width of cloth base image, A ∈ R ^{N * N}Be hybrid matrix.Obtain separation matrix M with the FastICA method, make and export:

Y＝MX＝MAS (4)

Y=[y wherein ₁, y ₂..., y _N] ^T, Y ∈ R ^{N * n}Row vector separate, then Y is exactly the estimation of independent basic image, each row of Y represent the basic image of width of cloth estimation.After obtaining Y, just can be proper subspace W of proper vector structure with the capable vector of Y _j, the facial image of training and test is projected on this sub spaces.

Step 4, repeating step 2-3 extract everyone face characteristic subspace W _j, j=1,2 ..., m, in the formula m for be used for the training and know others face classification number, in the following steps roughly the same;

Step 5, with people's face training image X _iWith formula (5) to W _jShine upon, extract its characteristics of image H corresponding to everyone face proper subspace _Ij

H _j＝(X _i-μ _j)×W _j，i＝1，2，...，N，j＝1，2，...，m (5)

Step 6, in the mode of formula (6), with proper vector H _IjTo W _jOppositely mapping, reconstruct obtains new facial image X _Ij

X _ij＝W _j×H _ij+μ _j，i＝1，2，...，N，j＝1，2，...，m (6)

Step 7, will be corresponding to the reconstructed image X of each proper subspace _IjDeduct original image X _i, ask for its two norm and divided by original image X _iTwo norms, to calculate the reconstructed error ε of this image corresponding to each proper subspace _Ij, as formula (7):

${\mathrm{\ϵ}}_{\mathrm{ij}}=\frac{\left|\right|X_{\mathrm{ij}}-X_i\left|\right|}{\left|\right|X_i\left|\right|},j=\mathrm{1,2},...,m---\left(7\right)$

Step 8, repeating step 4-7 extract the reconstructed error ε of all training images _Ij, i=1,2 ..., N, j=1,2 ..., m;

Step 9, all training image reconstructed errors are carried out normalization with formula (8);

${\mathrm{\ϵ}}_{\mathrm{ij}}={\mathrm{\ϵ}}_{\mathrm{ij}}/\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{\mathrm{ij}},j=\mathrm{1,2},...,m---\left(8\right)$

Step 10, with the training vector of the training image reconstructed error vector after the normalization as support vector machine, the training supporting vector machine model;

Step 11, with facial image to be identified with step 5-9 to proper subspace W _j, j=1,2 ..., m shines upon and reconstruct, obtains the image reconstruction error vector;

Step 12, the image reconstruction error to be identified after the normalization is vectorial as the identification vector is carried out the Classification and Identification of facial image with supporting vector machine model.

The present invention compared with prior art has the following advantages:

1, because people's face is the entity of a general character and characteristic coexistence, and everyone has certain similarity and rule by face, but different people's faces exists tangible difference again.The face characteristic subspace that traditional face identification method produces all is a general subspace that is produced by all training samples in the face database usually, what this subspace comprised more is the common feature of everyone face sample, and has ignored for some even more important personal characteristics of recognition of face.Consider some personal features of individual human face, based on the independent component analytical algorithm, we serve as to produce matrix with covariance matrix in its class, for everyone has set up independent proper subspace model, ratio proper subspace method has in the past been utilized the face characteristic of different people more fully, effectively raises recognition accuracy.

2, this method is with good expansibility; Because the eigenface of different people is relatively independent, when in face database, adding new people's face, only need carry out the training of eigenface to people's face of new interpolation, and need not be as traditional proper subspace method training characteristics subspace again, so this method has better extensibility.

Description of drawings

Fig. 1 system flowchart of the present invention

The recognition result figure of Fig. 2 the present invention on the ORL face database

Embodiment

With reference to figure 1, it is the process flow diagram of performing step of the present invention, in conjunction with this figure implementation process of the present invention is done detailed explanation.Embodiments of the invention are being to implement under the prerequisite with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention are not limited to following embodiment.

Embodiment has adopted a public face database, the ORL face database of univ cambridge uk.The ORL storehouse comprises the facial image of 40 people's 400 112 * 92 sizes, everyone 10 width of cloth.These images are taken at different time, and variations such as attitude, angle, yardstick, expression and glasses are arranged.Concrete face recognition process is as follows:

1, image pre-service

Facial image to 112 * 92 sizes carries out pre-service, mainly comprises the normalized of figure image intensifyings such as image smoothing and contrast correction and gradation of image and variance.Through after the pre-service, the gray scale of all images is unified to standard level, and gray-level is clearly more demarcated, and simultaneously, for the time and the memory space of saving computing, we adopt bilinear interpolation to compress image to 24 * 24 sizes.In addition, eigenface extracts because the people's face sample that is based on single people based on the face identification method of image reconstruction carries out, the experiment sample number is less, for more efficiently extraction face characteristic subspace, mirror image about we have carried out all images in the face database has enlarged the quantity of training and test sample book.

2, feature extraction

(1) we adopt direct access to be used for training according to half image in the storehouse, the way that second half is used to discern, be everyone 5 samples and mirror image (totally 10 training samples) way of training thereof, corresponding samples remaining and mirror image thereof (10 test sample books) are tested.At first people's face training image is handled, to obtain the original training sample matrix V in the former space _jPeople's face training image is stacked as the vector of 576 dimensions by windrow, and adopts and go average and albefaction processing that its value is normalized between 0 to 1.Form 576 * 5 * 40 training sample matrix V={ V so altogether ₁, V ₂..., V _j, j=1,2 ..., 40, V in the formula _jIt is 576 * 5 training sample matrix;

(2) corresponding to each V _j, j=1,2 ..., 40, use the independent component analytic approach and ask for its characteristic of correspondence personal ICA base image, and constitute its personal characteristics subspace W _j, j=1,2 ..., 40;

(3) with training image X _i, i=1,2 ..., 200 to W _jShine upon, extract its characteristics of image H corresponding to each proper subspace _Ij, i=1,2 ..., 200, afterwards, with H _Ij, i=1,2 ..., 200 to W _jCounter asking to obtain the reconstructed image X of himself _Ij, and with formula ${\mathrm{\ϵ}}_{\mathrm{ij}}=\frac{\left|\right|X_{\mathrm{ij}}-X_i\left|\right|}{\left|\right|X_i\left|\right|},i=\mathrm{1,2},...,200;$

(4) the same, obtain all training images corresponding to all W _j, j=1,2 ..., 40 reconstructed error;

(5) with reconstructed error with formula ${\mathrm{\ϵ}}_{\mathrm{ij}}={\mathrm{\ϵ}}_{\mathrm{ij}}/\underset{j=1}{\overset{c}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{\mathrm{ij}},j=\mathrm{1,2},...,m$ Carry out normalization, obtain to be used for the reconstructed error proper vector of recognition of face.

3, training and identification

(1) the reconstructed error proper vector of training image being obtained is trained supporting vector machine model as the training vector of support vector machine.In this example, the kernel function of support vector machine adopts be radial basis function K (x, y)=exp (γ ‖ x-y ‖ ²).

(2) will test facial image equally to proper subspace W _j, j=1,2 ..., 40 shine upon and reconstruct, and normalization is to obtain test person face image reconstruction error vector.

(3) the image reconstruction error vector to be identified after the normalization is vectorial as identification, the supporting vector machine model that obtains with aforementioned training image training carries out the class test of facial image, and outputs test result.

For the validity of algorithm better is described, on the basis of aforesaid identical data sample, adopt respectively based on overall scatter matrix S _tAnd scatter matrix S between class _bThe PCA feature extracting method, adopt arest neighbors method and support vector machine method to discern respectively the feature that extracts, and it carried out feature extraction with this paper proposes with the PCA image reconstruction, the method for discerning with support vector machine compares experiment.We repeat this experiment 50 times, and the mean value of getting these 50 discriminations is as final experimental result.In this experiment, the number of training that is used to extract the personal characteristics face has 10 width of cloth images, and the order of the covariance matrix of resulting same class sample mostly is 9 most, that is to say, the nonzero eigenvalue of this covariance matrix has only 9 at most, and promptly the intrinsic dimensionality maximum of PCA can only be got 9 dimensions.Therefore, the PCA intrinsic dimensionality d that the present invention extracts is 4,5,6,7,8,9, and for ease of comparing, when adopting traditional PCA face identification method to extract intrinsic dimensionality, we get its intrinsic dimensionality d is 4 ², 5 ², 6 ², 7 ², 8 ², 9 ², experimental result as shown in Figure 2.

Fig. 2 is the experimental result of the present invention on the ORL face database, and transverse axis is represented the intrinsic dimensionality of ICA proper subspace, and the longitudinal axis is represented corresponding recognition of face accuracy.As can be seen from Figure 2, in recognition of face experiment based on ORL, the correct recognition rata that obtains based on the face identification method of ICA base image reconstruction error is 96.48%, is enhanced based on overall scatter matrix and based on the PCA face identification method of scatter matrix between class than traditional.

Claims (1)

1, based on the face identification method of personal ICA base image reconstruction error, comprise with facial image pre-service, facial image feature extraction and image reconstruction and based on people's face Classification and Identification of SVM, it is characterized in that its identification step is:

(1) normalized of image smoothing and gradation of image and variance;

(2) read training storehouse facial image and become the gray matrix form, and training sample is carried out classification and storage with the individual become V _j

(3) with individual's facial image matrix V _jAs producing matrix, adopt the ICA method to extract its proper subspace W _j

(4) repeating step 2-3 extracts the proper subspace W of everyone face _j, j=1,2 ..., m;

(5) with training image X _iTo W _jShine upon, extract its feature H _Ij

(6) with proper vector H _IjTo W _jCounter asking, reconstruct obtain new facial image X _Ij

(7) with reconstructed image X _IjDeduct original image X _iTwo norms divided by original image X _iTwo norm calculation go out reconstructed error ε _Ij

(8) repeating step 5-7 extracts the reconstructed error ε of training image _Ij, i=1,2 ..., N, j=1,2 ..., m;

(9) reconstructed error with all training images carries out normalization;

(10) with the training vector of the training image reconstructed error vector after the normalization as support vector machine, the training supporting vector machine model;

(11) to extract the reconstructed error of facial image to be identified with the same step 5-9 of training image;

(12) the image reconstruction error vector to be identified after the normalization is carried out Classification and Identification with supporting vector machine model.

Images