CN108052867A - A kind of single sample face recognition method based on bag of words - Google Patents
A kind of single sample face recognition method based on bag of words Download PDFInfo
- Publication number
- CN108052867A CN108052867A CN201711155556.8A CN201711155556A CN108052867A CN 108052867 A CN108052867 A CN 108052867A CN 201711155556 A CN201711155556 A CN 201711155556A CN 108052867 A CN108052867 A CN 108052867A
- Authority
- CN
- China
- Prior art keywords
- msub
- mrow
- sub
- facial image
- neighbour
- 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
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of single sample face recognition methods based on bag of words, extract middle level semantic feature using bag of words to reduce " semantic gap " in the case of single sample.Face is divided into multiple sub-blocks by this method first, then extracts the SIFT feature of all sub-blocks, and the SIFT feature of all sub-blocks is clustered, and then constructs vision word dictionary；View-based access control model word dictionary proposes the multistagekNeighbour cooperates with presentation code method, and the local feature of each sub-block is projected to semantic space；In order to describe spatial information, while intrinsic dimensionality is reduced, pond is carried out to feature after coding using spatial pyramid model, generates the histogram of view-based access control model word to describe face.Finally the feature of Chi Huahou is merged together, is classified using the SVM classifier based on linear kernel.List sample face recognition method of the invention is to expression, illumination variation and blocks etc. and to have good robustness, and accuracy of identification is high.
Description
Technical field
The present invention relates to single sample face recognition methods, and in particular to each object to be identified only has a width training image
Single sample face recognition method based on bag of words, belongs to technical field of face recognition.
Background technology
By the development of nearly 50 years, Face recognition technology had been achieved for considerable progress, under controlled condition
Face recognition technology has been achieved with satisfactory performance.However under the conditions of non-controllable, due to by illumination, expression, posture, making an uproar
Sound, the influence for the factors such as blocking, the precision of face recognition technology drastically decline, and far can not meet application demand.Solve these
The most direct method of problem is exactly to increase training sample, but in practical applications as identity card identification, passport identification, the administration of justice are true
Recognize, in many practical applications such as admission control, there is usually one training samples to be acquired, recognition of face in this case
Problem is referred to as single sample recognition of face problem (single sample per person, SSPP), is further exacerbated by
The difficulty of recognition of face under the conditions of non-controllable.
The difficult point of single sample recognition of face is it is difficult to distinguish essential change and illumination, expression between different faces, block
Caused variation, that is to say, that there are semantic gaps between face characteristic and its identity.Bag of words are in image point in recent years
Excellent properties in generic task cause the widely studied interest of scholars, and this model is introduced into recognition of face research
Field.Such as Li et al. people (Z.S.Li, J.I.Imai, and M.Kaneko, " Robust face recognition using
block-based Bag of Words,”Proceedings of the 26th International Conference on
Pattern Recognition, pp.1285-1288) bag of words are clearly applied to recognition of face first, propose a kind of base
In the robust human face recognizer of piecemeal bag of words；Xie et al. (S.F.Xie, S.G.Shan, X.L.Chen, X.Meng and
W.Gao,“Learned local Gabor pattern for face representation and recognition,”
Signal Processing, vol.89, no.3, pp.2333-2344,2009) it is multiple dimensioned multidirectional to image progress first
The image of each passage is densely divided into several fritters and builds visual dictionary by Gabor transformation, using nearest neighbor method to figure
As fritter is encoded, then the word occurrence frequency histogram obtained in each passage is stitched together phenogram picture.Recently,
Cui et al. (Z.Cui, W.Li, D.Xu, S.G.Shan, and X.L.Chen, " Fusing robust face region
descriptors via multiple metric learning for face recognition in the wild,”
Proceedings of the 26th International Conference on Computer Vision and
Pattern Recognition, pp.3554-3561,2013) propose that one kind describes operator based on space human face region
The face recognition algorithms of (spatial face region descriptor, SFRD), the algorithm are equally retouched using gray feature
Image fritter is stated, but each fritter is encoded using the good dictionary of non-negative coding method control off-line training, then will
Image is divided into several subregions, uses the pond feature in the every sub-regions of metric learning algorithm fusion.
Bag of words may be considered a kind of middle level semantic feature of extraction, can weaken high-level semantic and bottom in a way
Semantic gap between layer feature.Therefore, it theoretically can also be used to be promoted the recognition of face performance in the case of single sample, but mesh
The sparse coding or non-negative sparse coding computational complexity being commonly used in preceding bag of words are too high.
The content of the invention
The technical problems to be solved by the invention are：A kind of single sample face recognition method based on bag of words is provided,
Single sample recognition of face is solved the problems, such as using the higher coding method of more robust, computational efficiency.
The present invention uses following technical scheme to solve above-mentioned technical problem：
A kind of single sample face recognition method based on bag of words, includes the following steps：
Step 1, each trained facial image is divided into a series of sub-block, and it is local special to each sub-block extraction SIFT
Sign obtains the SIFT local feature set X ∈ R of all trained facial imagesD×N, wherein, D be SIFT local features dimension, N
For the sum of all trained facial image SIFT local features；
Step 2, the subset X s of the set X of the SIFT local features composition of all trained facial images, antithetical phrase are randomly choosed
Collection Xs does K mean cluster, obtains a vision word dictionary V=[v1,v2,…,vK]∈RD×K；
Step 3, to any one facial image, sub-block is divided according to the dividing mode of training facial image in step 1, and
The SIFT local features of sub-block are extracted, form set Xr={ x1,x2,…,xM}∈RD×M, wherein, M owns for a facial image
The number of sub-block；Based on the vision word dictionary V that step 2 obtains, to the SIFT local feature set X of the facial imagerIn
SIFT local features xmIt carries out multistage k neighbour and cooperates with presentation code, obtain its coding vector cm, m=1,2 ..., M；
Step 4, the facial image of step 3 is divided into 2 using spatial pyramid modell×2lSon under a different scale
Block, l=0,1 ..., L, L are the positive integer more than 0；It sets in l layers of the H sub-block and contains MHA coding vector, to these
Coding vector carries out the operation of maximum pondization, obtains the feature behind pond；To all under spatial pyramid model different scale
The coding vector that sub-block contains all carries out the operation of maximum pondization, and the feature behind all sub-block ponds is merged together, is obtained
The face characteristic of the facial image represents；
Step 5, all trained facial images and all test facial images are all carried out with the operation of step 3 and step 4, is obtained
Face characteristic to all trained facial images represents and the face characteristic of all test facial images represents, utilizes all training
The face characteristic of facial image represents SVM classifier of the structure based on linear kernel function, using the SVM classifier built to institute
There is test facial image to be identified.
As a preferred embodiment of the present invention, the vision word dictionary V based on step 2 acquisition described in step 3, to the people
The SIFT local feature set X of face imagerIn SIFT local features xmIt carries out multistage k neighbour and cooperates with presentation code, obtain it
Coding vector cm, detailed process is as follows：
(1) k neighbour V is found from vision word dictionary Vk=[v1,v2,…,vk]∈RD×k, wherein, D is local for SIFT
The dimension of feature, k are neighbour's number；
(2) k neighbour V is utilizedk=[v1,v2,…,vk]∈RD×kCooperate with SIFT local features xm, according to equation below meter
The collaboration for calculating k neighbour represents coefficient vector cm *：
Wherein, λ is regularization coefficient, | | | |2For 2 norms；
(3) collaboration of k neighbour is represented into coefficient vector cm *It is converted into the expression coefficient vector of K × 1Middle correspondence
K neighbour value be its collaboration represent coefficient, other values then be 0；
(4) k neighbour is reduced to k-1 neighbour, calculating its collaboration according to the step of (1)-(3) represents coefficientDirectly
To k=1；As k=1,In corresponding neighbour value be 1, other are all 0；
(5) the expression coefficient vector by the k calculating from k to 1 is summed, and is obtained multistage k neighbour and is cooperateed with presentation code
Vectorial cm, formula is as follows：
As a preferred embodiment of the present invention, maximum pondization described in step 4 operates, and calculation formula is as follows：
Wherein, BlHFor the feature of Chi Huahou, chFor the H son h-th of coding vector in the block of l layers, MHFor l layers
Number containing coding vector in the H sub-block.
As a preferred embodiment of the present invention, the feature behind all sub-block ponds is merged together described in step 4, is obtained
Face characteristic to the facial image represents that formula is as follows：
Bi=[Bi1；Bi2；…；BiS]
Wherein, BiThe face characteristic of facial image is trained to represent for i-th,2l×2lIt represents face
Image is divided into 2l×2lSub-block under a different scale.
As a preferred embodiment of the present invention, the optimal classification function of SVM classifier described in step 5 is：
Wherein, f (Bj) for optimal classification function, BjFor the face characteristic expression of j-th of test facial image, BiFor i-th
Training facial image face characteristic represent, n be training facial image number, yiFor the classification of i-th of training facial image
Label, αiFor Lagrange coefficient, κ () is kernel function, b*For the threshold value of SVM classifier.
As a preferred embodiment of the present invention, the K in K mean cluster described in step 2 is the positive integer more than 1.
The present invention compared with prior art, has following technique effect using above technical scheme：
1st, the present invention weakens single sample to a certain extent due to introducing feature of the bag of words extraction with middle level semanteme
In the case of semantic gap problem, to expression, illumination variation and block etc. there is good robustness, thus with higher knowledge
Other precision.
2nd, the present invention devises more efficient multistage k neighbour and cooperates with presentation code method, and simple and practicable, computational efficiency is more
It is high.
Description of the drawings
Fig. 1 is the flow chart of single sample face recognition method the present invention is based on bag of words.
Fig. 2 is that multistage k neighbour cooperates with presentation code in single sample face recognition method the present invention is based on bag of words
Method fundamental diagram.
Fig. 3 is space pyramid model schematic diagram in single sample face recognition method the present invention is based on bag of words.
Fig. 4 is the experimental result picture in LFW face databases the present invention is based on single sample face recognition method of bag of words.
Specific embodiment
Embodiments of the present invention are described below in detail, the example of the embodiment is shown in the drawings.Below by
The embodiment being described with reference to the drawings is exemplary, and is only used for explaining the present invention, and is not construed as limiting the claims.
The difficult point of single sample recognition of face problem is to solve the problems, such as the semantic gap between face characteristic and identity, therefore
It is taken into full account in single sample recognition of face with semantic face characteristic, semantic gap can be weakened, and bag of words can carry
Middle level semantic feature is taken, is theoretically suitble to solve the problems, such as single sample recognition of face.Based on this idea, the present invention proposes a kind of base
In single sample face recognition method of bag of words.
As shown in Figure 1, a kind of single sample face recognition method based on bag of words of the present invention, comprises the following steps：
1st, each trained facial image is divided into a series of sub-block, and SIFT feature is extracted to each sub-block, obtained
The local feature set X ∈ R of all trained facial imagesD×N, D is the dimension of SIFT local features, and N is all trained face figures
As the sum of sub-block local feature；
2nd, the subset X s of the set X of the local feature composition of all sub-blocks of all trained face figures, antithetical phrase are randomly choosed
Collection Xs does K mean cluster, obtains a vision word dictionary V=[v1,v2,…,vK]∈RD×K；
3rd, to any one facial image, divide sub-block according to the dividing mode of training facial image in step 1 and extract
Sub-block SIFT feature forms set Xr={ x1,x2,…,xM}∈RD×M, wherein M is the number of all sub-blocks of facial image.
Based on the vision word dictionary V that step 2 obtains, to the local feature set X of facial imager={ x1,x2,…,xM}∈RD×MIn
Local feature xmIt carries out multistage k neighbour and cooperates with presentation code, obtain it and encode cm, detailed process is as shown in Fig. 2, specifically such as
Under：
(1) from vision word dictionary V=[v1,v2,…,vK]∈RD×KK neighbour V of middle searchingk=[v1,v2,…,vk]∈
RD×k；
(2) k neighbour V is utilizedk=[v1,v2,…,vk]∈RD×kCooperate with local feature xm, calculate and assist according to equation below
With expression coefficient：
(3) collaboration that k neighbour is calculated according to (2) represents coefficient vector cm *Afterwards, it is translated into the vector of k × 1In corresponding k neighbour value be its collaboration represent coefficient, other values then be 0；
(4) k neighbour is reduced to k-1 neighbour, calculating its collaboration according to the step of (1)-(3) represents coefficientDirectly
To k=1；As k=1, the coefficient value of arest neighbors is 1, other are all 0；
(5) the expression coefficient vector of the k calculating from k to 1 is summed, forms final multistage k neighbour by equation below
Cooperate with presentation code：
4th, feature pool is carried out to M coding vector after being encoded in step 3, utilizes spatial pyramid as shown in Figure 3
Facial image is divided into 2 by modell×2lSub-block under a different scale, l=0,1 ..., L.Assuming that in l layers of the H son
There is M in blockHA coding vector then carries out these coding vectors the operation of maximum pondization, obtains the feature behind pond, calculates public
Formula is as follows：
All sub-blocks under pyramid model different scale are all carried out with pondization operation, by the feature behind all sub-block ponds
It is merged together, is represented as final face characteristic.
5th, after the face characteristic expression that all training sample test samples are obtained according to step 4, training sample structure is utilized
The SVM classifier based on linear kernel function is built, completes the classification of test sample.The optimal classification function of SVM classifier is：
Wherein, BjFor the face characteristic expression of j-th of test facial image, BiFace for i-th of training facial image is special
Sign represent, n be training facial image number, yiFor the class label of i-th of training facial image, αiFor Lagrange coefficient,
κ () is kernel function, b*For the threshold value of SVM classifier.
A kind of single sample face recognition method based on bag of words of the present invention, can extract middle level semantic feature, effectively
Semantic gap problem in the case of the single sample of reduction, multistage k neighbour cooperate with presentation code method compared to rarefaction representation and
Non-negative sparse represent computational efficiency higher, while to illumination, expression, block, time change all have good robustness.Such as figure
Shown in 4, nearly 10% promotion is achieved compared to traditional single sample face recognition method on LFW databases, compared to
Non-negative sparse coding method NSC_BoF, not only effect is more preferable but also computational efficiency higher.
Above example is merely illustrative of the invention's technical idea, it is impossible to protection scope of the present invention is limited with this, it is every
According to technological thought proposed by the present invention, any change done on the basis of technical solution each falls within the scope of the present invention
Within.
Claims (6)
1. a kind of single sample face recognition method based on bag of words, which is characterized in that include the following steps：
Step 1, each trained facial image is divided into a series of sub-block, and SIFT local features is extracted to each sub-block,
Obtain the SIFT local feature set X ∈ R of all trained facial imagesD×N, wherein, D is the dimension of SIFT local features, and N is institute
There is the sum of trained facial image SIFT local features；
Step 2, the subset X s of the set X of the SIFT local features composition of all trained facial images is randomly choosed, to subset Xs
K mean cluster is done, obtains a vision word dictionary V=[v1,v2,…,vK]∈RD×K；
Step 3, to any one facial image, sub-block is divided according to the dividing mode of training facial image in step 1, and is extracted
The SIFT local features of sub-block form set Xr={ x1,x2,…,xM}∈RD×M, wherein, M is all sub-blocks of facial image
Number；Based on the vision word dictionary V that step 2 obtains, to the SIFT local feature set X of the facial imagerIn SIFT
Local feature xmIt carries out multistage k neighbour and cooperates with presentation code, obtain its coding vector cm, m=1,2 ..., M；
Step 4, the facial image of step 3 is divided into 2 using spatial pyramid modell×2lSub-block under a different scale, l
=0,1 ..., L, L are the positive integer more than 0；It sets in l layers of the H sub-block and contains MHA coding vector, to these codings
Vector carries out the operation of maximum pondization, obtains the feature behind pond；To all sub-blocks under spatial pyramid model different scale
The coding vector contained all carries out the operation of maximum pondization, and the feature behind all sub-block ponds is merged together, obtains the people
The face characteristic of face image represents；
Step 5, all trained facial images and all test facial images are all carried out with the operation of step 3 and step 4, obtains institute
The face characteristic for having trained facial image represents and the face characteristic of all test facial images represents, utilizes all trained faces
The face characteristic of image represents SVM classifier of the structure based on linear kernel function, using the SVM classifier built to all surveys
Examination facial image is identified.
2. single sample face recognition method based on bag of words according to claim 1, which is characterized in that described in step 3
Based on the vision word dictionary V that step 2 obtains, to the SIFT local feature set X of the facial imagerIn SIFT local features
xmIt carries out multistage k neighbour and cooperates with presentation code, obtain its coding vector cm, detailed process is as follows：
(1) k neighbour V is found from vision word dictionary Vk=[v1,v2,…,vk]∈RD×k, wherein, D is SIFT local features
Dimension, k be neighbour's number；
(2) k neighbour V is utilizedk=[v1,v2,…,vk]∈RD×kCooperate with SIFT local features xm, k are calculated according to equation below
The collaboration of neighbour represents coefficient vector cm*：
<mrow>
<msup>
<msub>
<mi>c</mi>
<mi>m</mi>
</msub>
<mo>*</mo>
</msup>
<mo>=</mo>
<mi>arg</mi>
<mi> </mi>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>m</mi>
</msub>
<mo>-</mo>
<msub>
<mi>V</mi>
<mi>k</mi>
</msub>
<msup>
<msub>
<mi>c</mi>
<mi>m</mi>
</msub>
<mo>*</mo>
</msup>
<mo>|</mo>
<msubsup>
<mo>|</mo>
<mn>2</mn>
<mn>2</mn>
</msubsup>
<mo>+</mo>
<mi>&lambda;</mi>
<mo>|</mo>
<mo>|</mo>
<msup>
<msub>
<mi>c</mi>
<mi>m</mi>
</msub>
<mo>*</mo>
</msup>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>2</mn>
</msub>
</mrow>
Wherein, λ is regularization coefficient, | | | |2For 2 norms；
(3) collaboration of k neighbour is represented into coefficient vector cm* it is converted into the expression coefficient vector of K × 1In corresponding k
The value of a neighbour is that its collaboration represents coefficient, and other values are then 0；
(4) k neighbour is reduced to k-1 neighbour, calculating its collaboration according to the step of (1)-(3) represents coefficientUntil k=
1；As k=1,In corresponding neighbour value be 1, other are all 0；
(5) the expression coefficient vector by the k calculating from k to 1 is summed, and is obtained multistage k neighbour and is cooperateed with presentation code vector
cm, formula is as follows：
<mrow>
<msub>
<mi>c</mi>
<mi>m</mi>
</msub>
<mo>=</mo>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</msubsup>
<msubsup>
<mi>c</mi>
<mi>m</mi>
<msup>
<mi>k</mi>
<mo>&prime;</mo>
</msup>
</msubsup>
<mo>.</mo>
</mrow>
3. single sample face recognition method based on bag of words according to claim 1, which is characterized in that described in step 4
Maximum pondization operates, and calculation formula is as follows：
<mrow>
<msub>
<mi>B</mi>
<mrow>
<mi>l</mi>
<mi>H</mi>
</mrow>
</msub>
<mo>=</mo>
<munder>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
<mrow>
<mi>h</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>M</mi>
<mi>H</mi>
</msub>
</mrow>
</munder>
<msub>
<mi>c</mi>
<mi>h</mi>
</msub>
</mrow>
Wherein, BlHFor the feature of Chi Huahou, chFor the H son h-th of coding vector in the block of l layers, MHFor l layers of H
Number containing coding vector in sub-block.
4. single sample face recognition method based on bag of words according to claim 1, which is characterized in that described in step 4
Feature behind all sub-block ponds is merged together, the face characteristic for obtaining the facial image represents that formula is as follows：
Bi=[Bi1；Bi2；…；BiS]
Wherein, BiIt is the face characteristic expression of i-th of training facial image,2l×2lIt represents facial image point
It is segmented into 2l×2lSub-block under a different scale.
5. single sample face recognition method based on bag of words according to claim 1, which is characterized in that described in step 5
The optimal classification function of SVM classifier is：
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>B</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>s</mi>
<mi>g</mi>
<mi>n</mi>
<mrow>
<mo>(</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&alpha;</mi>
<mi>i</mi>
</msub>
<mi>&kappa;</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>B</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>B</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msup>
<mi>b</mi>
<mo>*</mo>
</msup>
<mo>)</mo>
</mrow>
</mrow>
Wherein, f (Bj) for optimal classification function, BjFor the face characteristic expression of j-th of test facial image, BiFor i-th of training
The face characteristic of facial image represents, n is the number of training facial image, yiThe class label for training facial image for i-th,
αiFor Lagrange coefficient, κ () is kernel function, b*For the threshold value of SVM classifier.
6. single sample face recognition method based on bag of words according to claim 1, which is characterized in that K described in step 2
K in mean cluster is the positive integer more than 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711155556.8A CN108052867B (en) | 2017-11-20 | 2017-11-20 | Single-sample face recognition method based on bag-of-words model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711155556.8A CN108052867B (en) | 2017-11-20 | 2017-11-20 | Single-sample face recognition method based on bag-of-words model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108052867A true CN108052867A (en) | 2018-05-18 |
| CN108052867B CN108052867B (en) | 2021-11-23 |
Family
ID=62119957
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711155556.8A Active CN108052867B (en) | 2017-11-20 | 2017-11-20 | Single-sample face recognition method based on bag-of-words model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108052867B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766810A (en) * | 2018-12-31 | 2019-05-17 | 陕西师范大学 | Recognition of face classification method based on collaboration expression and pond and fusion |
| CN109902657A (en) * | 2019-03-12 | 2019-06-18 | 哈尔滨理工大学 | A kind of face identification method indicated based on piecemeal collaboration |
| CN111723612A (en) * | 2019-03-20 | 2020-09-29 | 北京市商汤科技开发有限公司 | Face recognition and face recognition network training method and device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616291A (en) * | 2015-01-15 | 2015-05-13 | 东华大学 | Sparse coding-based fabric appearance flatness evaluation method |
| CN105184298A (en) * | 2015-08-27 | 2015-12-23 | 重庆大学 | Image classification method through fast and locality-constrained low-rank coding process |
| CN105808757A (en) * | 2016-03-15 | 2016-07-27 | 浙江大学 | Chinese herbal medicine plant picture retrieval method based on multi-feature fusion BOW model |
| CN106446774A (en) * | 2016-08-24 | 2017-02-22 | 施志刚 | Face recognition method based on secondary nearest neighbor sparse reconstruction |
-
2017
- 2017-11-20 CN CN201711155556.8A patent/CN108052867B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616291A (en) * | 2015-01-15 | 2015-05-13 | 东华大学 | Sparse coding-based fabric appearance flatness evaluation method |
| CN105184298A (en) * | 2015-08-27 | 2015-12-23 | 重庆大学 | Image classification method through fast and locality-constrained low-rank coding process |
| CN105808757A (en) * | 2016-03-15 | 2016-07-27 | 浙江大学 | Chinese herbal medicine plant picture retrieval method based on multi-feature fusion BOW model |
| CN106446774A (en) * | 2016-08-24 | 2017-02-22 | 施志刚 | Face recognition method based on secondary nearest neighbor sparse reconstruction |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766810A (en) * | 2018-12-31 | 2019-05-17 | 陕西师范大学 | Recognition of face classification method based on collaboration expression and pond and fusion |
| CN109766810B (en) * | 2018-12-31 | 2023-02-28 | 陕西师范大学 | Face recognition classification method based on collaborative representation, pooling and fusion |
| CN109902657A (en) * | 2019-03-12 | 2019-06-18 | 哈尔滨理工大学 | A kind of face identification method indicated based on piecemeal collaboration |
| CN109902657B (en) * | 2019-03-12 | 2022-07-08 | 哈尔滨理工大学 | Face recognition method based on block collaborative representation |
| CN111723612A (en) * | 2019-03-20 | 2020-09-29 | 北京市商汤科技开发有限公司 | Face recognition and face recognition network training method and device, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108052867B (en) | 2021-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Boukharouba et al. | Novel feature extraction technique for the recognition of handwritten digits | |
| CN106127196B (en) | Facial expression classification and identification method based on dynamic texture features | |
| CN100461204C (en) | Method for recognizing facial expression based on 2D partial least square method | |
| CN109063666A (en) | The lightweight face identification method and system of convolution are separated based on depth | |
| CN107463917B (en) | Improved LTP and two-dimensional bidirectional PCA fusion-based face feature extraction method | |
| CN105069447B (en) | A kind of recognition methods of human face expression | |
| CN103400154B (en) | A kind of based on the human motion recognition method having supervision Isometric Maps | |
| CN106548149A (en) | The recognition methods of the micro- facial expression image sequence of face in monitor video sequence | |
| CN103745200A (en) | Facial image identification method based on word bag model | |
| CN107122712A (en) | It polymerize the palmprint image recognition methods of description vectors based on convolutional neural networks and two-way local feature | |
| CN109325507A (en) | A kind of image classification algorithms and system of combination super-pixel significant characteristics and HOG feature | |
| CN108052867A (en) | A kind of single sample face recognition method based on bag of words | |
| CN106709419A (en) | Video human behavior recognition method based on significant trajectory spatial information | |
| CN112784722B (en) | Behavior identification method based on YOLOv3 and bag-of-words model | |
| CN106295478A (en) | A kind of image characteristic extracting method and device | |
| CN105718858A (en) | Pedestrian recognition method based on positive-negative generalized max-pooling | |
| CN113887509B (en) | Rapid multi-modal video face recognition method based on image set | |
| Karungaru et al. | Face recognition in colour images using neural networks and genetic algorithms | |
| Chen et al. | More about covariance descriptors for image set coding: Log-euclidean framework based kernel matrix representation | |
| Sankaran et al. | A multi-view approach on modular PCA for illumination and pose invariant face recognition | |
| CN108121965A (en) | Image-recognizing method based on robust joint sparse feature extraction | |
| Ardakany et al. | An extended local binary pattern for gender classification | |
| CN113688715A (en) | Facial expression recognition method and system | |
| Mydam et al. | Low rank representation based discriminative multi manifold analysis for low-resolution face recognition | |
| CN112580444B (en) | Single-sample face recognition method based on cyclic self-coding and block sparse structure representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |