CN107169410A - The structural type rarefaction representation sorting technique based on LBP features for recognition of face - Google Patents
The structural type rarefaction representation sorting technique based on LBP features for recognition of face Download PDFInfo
- Publication number
- CN107169410A CN107169410A CN201710222367.1A CN201710222367A CN107169410A CN 107169410 A CN107169410 A CN 107169410A CN 201710222367 A CN201710222367 A CN 201710222367A CN 107169410 A CN107169410 A CN 107169410A
- Authority
- CN
- China
- Prior art keywords
- lbp
- mrow
- rarefaction representation
- msub
- structural type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
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
- G06V40/172—Classification, e.g. identification
Abstract
The invention discloses a kind of structural type rarefaction representation sorting technique based on LBP features for recognition of face（LBP‑SSRC）For traditional SRC algorithms, it is equally extremely important that the locality of sample is openness compared to its, local feature is a kind of highly useful characteristic information, the algorithm concentrates the LBP histogram features for extracting sample from original sample first, in view of the partitioned organization of training sample, the algorithm model of two kinds of structural type rarefaction representations is then devised（SSRC）Finally the LBP features extracted are input in SSRC algorithm models, make full use of the locality of sample LBP features and SSRC partitioned organization, therefore the algorithm allows test sample to select to be obtained to reconstruct with the training sample generic with it as far as possible, can be very good to lift Classification and Identification effect.By the contrast experiment on public AR face databases, the validity of put forward algorithm can be fully verified.
Description
Technical field
It is more particularly to a kind of special based on LBP for recognition of face the present invention relates to a kind of sorting algorithm of rarefaction representation
The structural type rarefaction representation sorting technique levied, belongs to field of face identification.
Background technology
In the research of long duration in past, rarefaction representation is initially to be made to solve asking for many field of signal processing
Topic, such as image denoising, compression of images and image recovery etc..Pass through the exploration of numerous researchers in recent years, find to know in pattern
Other field, rarefaction representation can be very good for solving the problem of image classification is recognized due to its high efficiency.
The rarefaction representation of image how is defined, exactly this image is used based on the openness expression most approached, together
When ensure its parsimony.Rarefaction representation has identification, it is meant that it can select a kind of expression coefficient the most sparse to come very
Good expression original image, exactly because this characteristic, in field of face identification, the application of rarefaction representation is extremely successful.In base
In the sorting technique (SRC) of rarefaction representation, we concentrate from training sample and obtain an excessively complete dictionary by study, train
Sample is all training samples for including a certain classification in the atom in the dictionary, same linear subspaces.Rarefaction representation
Purpose is intended to allow certain class testing sample only by the training sample similar with its is come linear expression, is in other words exactly in rarefaction representation
Coefficient matrix in, the expression coefficient of the only training sample generic with test sample is not zero, and other are zero.Such one
Come
Traditional sorting algorithm (SRC) based on rarefaction representation, be specially：
Given sample data set Z, wherein training sample have k classes, and the quantity of the i-th class sample is ni, define zij∈
RmExpression belongs to j-th of sample of the i-th class sample set, and the dimension of characteristic vector is m, then the i-th all class samples can be represented
For sample
This collectionBy the sample set A of all k classesiCombine, obtain dictionary matrix A：
A=[A1,A2,…,Ak]∈Rm×n (1)
Test sample y can be with dictionary matrix A linear expression：
Y=Ax ∈ Rm (2)
In above formula (2)It is coefficient vector, wants to reach rarefaction representation
Effect, coefficient vector x will meet such condition, i.e., except the coefficient corresponding to the training sample generic with test sample y
Outside being not zero, its remainder is zero.Therefore we only need to solve system of linear equations y=Ax coefficient, it is possible to judge
Go out the generic of test sample y.
By adding constraints, the rarefaction representation problem is converted into solution l0Norm minimum problem, it is possible to obtain
Most openness coefficient solution is obtained, shown in Optimized model such as formula (3)：
In above formula (3) | | | |0Try to achieve be nonzero element in coefficient vector x number, a namely l0Norm.But it is logical
Cross the research of a large amount of scholars it can be seen that, above-mentioned optimization problem is a classical NP problem, can not easily be tried to achieve
Sparse solution.And according to compressive sensing theory, l can be used1Norm minimum problem replaces solving above-mentioned optimization problem, certainly before
It is coefficient x to be met sparse conditions enough to carry, and the now solution of formula (3) can be converted into formula (4)：
Above formula is the convex Optimization Solution problem of a standard, in practical operation, is all that the form for being converted into formula (5) is come
Solve, wherein ε is the error for allowing to exist：
In the classification of discriminating test sample, it is assumed that for different sample class i, there is mapping relations δi：
Wherein δi(x) it is the expression coefficient that represents the i-th class in coefficient x, expression coefficient of the test sample only with the i-th class
Approximate representation is yi：
yi=A δi(x) (7)
So we, which can be obtained by, approaches residual error ri(y), namely it is so-called be used for weigh to the dilute of test sample
The reconstructed error of approximation ratio is dredged, can be in the hope of by formula (9)：
ri(y)=| | y-A δi(x)||2 (8)
Reconstructed error ri(y) the smaller classification of value, indicates that the sample of the category is more similar to test sample, because weight
What the size of structure error was represented is exactly the size to test sample approximation ratio, it is determined that the classification with minimal reconstruction error
As the classification of test sample, is embodied as shown in formula (9)：
For above-mentioned traditional SRC algorithms, the locality of sample compared to its it is openness be equally it is extremely important,
Local feature is a kind of highly useful characteristic information, and in classical SRC algorithms, uses the global characteristics of sample,
The local feature of sample is not made full use of.Simultaneously traditional SRC algorithms exist one it is potential the problem of, due to training dictionary
In contain substantial amounts of different classes of training sample, test sample just can be by from different classes of test sample table
Show and obtain, the Classification and Identification effect for algorithm can produce considerable influence naturally.So carrying out Classification and Identification using SRC algorithms
When, it is necessary to take into full account the local feature of sample and the partitioned organization of training dictionary.
The content of the invention
It is dilute that the technical problems to be solved by the invention are to provide a kind of structural type based on LBP features for recognition of face
Presentation class method (LBP-SSRC) is dredged, this method is considered to make full use of the local feature of original sample first, from original sample
This concentration extracts the LBP histogram features of sample, then in view of the partitioned organization type of training dictionary, designs two kinds of structural types
Rarefaction representation sorting algorithm model (SSRC), the LBP features of extraction are input in SSRC algorithm models, carry out point of next step
Class identification mission.Pass through the contrast experiment on public AR face databases, it was demonstrated that carry algorithm be compared to it is traditional based on
The sorting algorithm of rarefaction representation, can effectively improve discrimination.
The present invention uses following technical scheme to solve above-mentioned technical problem：
The present invention provides a kind of structural type rarefaction representation sorting technique based on LBP features for recognition of face, including
Step in detail below：
Step 1, the LBP features of original sample concentration training sample and test sample image are extracted respectively；
Step 2, structural type rarefaction representation algorithm model is built；
Step 3, the LBP features of the training sample extracted in step 1 and test sample are input to structural type rarefaction representation
In sorting algorithm model, according to the output of structural type rarefaction representation algorithm model, the classification results of test sample are obtained.
As the further technical scheme of the present invention, the extracting method of the LBP features of test sample image includes in step 1
Following steps：
1) test sample image y is divided into 16 × 16 zonule, y={ y are designated as1,y2,…,yn, wherein, n is represented
The zonule number of division；
2) respectively to yiLBP codings are carried out, the LBP code patterns of each zonule is obtained, is designated asWherein, i=1,2 ...,
n；
3) it is right respectivelyStatistics with histogram is carried out, that is, counts the frequency of each LBP encoded radios, obtains histogram statistical features
Vectorial LBP (yi), and make normalized；
4) the histogram statistical features vector series connection of all zonules is obtained into the LBP feature squares of view picture test sample image
Battle array：LBP (y)=[LBP (y1),LBP(y2),…,LBP(yn)]。
As the further technical scheme of the present invention, the extracting method of the LBP features of training sample image includes in step 1
Following steps：
1) training sample image D is divided into 16 × 16 zonule, D={ D are designated as1,D2,…,Dn, wherein, n is represented
The zonule number of division；
2) respectively to DiLBP codings are carried out, the LBP code patterns of each zonule is obtained, is designated asWherein, i=1,
2,…,n；
3) it is right respectivelyStatistics with histogram is carried out, that is, counts the frequency of each LBP encoded radios, obtains histogram statistical features
Vectorial LBP (Di), and make normalized；
4) the histogram statistical features vector series connection of all zonules is obtained into the LBP feature squares of view picture test sample image
Battle array：LBP (D)=[LBP (D1),LBP(D2),…,LBP(Dn)]。
As the further technical scheme of the present invention, LBP encoded radio is obtained by below equation：
Wherein,piIt is center pixel pcThe adjacent pixel value of surrounding, 2pIt is pixel difference weight coefficient.
Building structural type rarefaction representation algorithm model as the further technical scheme of the present invention, in step 2 is：
Wherein, x represents rarefaction representation coefficient vector, x={ x1,x2,…,xn}。
Building structural type rarefaction representation algorithm model as the further technical scheme of the present invention, in step 2 is：
Wherein, wherein, x represent rarefaction representation coefficient vector, x={ x1,x2,…,xn}。
As the further technical scheme of the present invention, the LBP features of training sample and test sample are input in step 3
State in structural type rarefaction representation sorting algorithm model, obtain rarefaction representation coefficient vector x；Then according to per class coefficient xiAnd LBP
(Di) LBP (y) residual error is calculated, decide that test sample belongs to that minimum class of residual error, classification formula is：
The present invention uses above technical scheme compared with prior art, with following technique effect：
Brief description of the drawings
Fig. 1 is the structural type rarefaction representation sorting algorithm based on LBP features proposed by the present invention for recognition of face
Implementation process figure.
Fig. 2 is the basic principle schematic of LBP codings.
Embodiment
Technical scheme is described in further detail below in conjunction with the accompanying drawings：
As shown in figure 1, the structural type rarefaction representation based on the LBP features classification of the present invention for recognition of face is calculated
Method, comprises the following steps：
1) the middle training sample of original sample collection and the LBP histogram features of test sample image are extracted.
Give a center pixel pc, LBP encoded radio can be obtained by formula (1)：
Wherein,piIt is center pixel pcThe adjacent pixel value of surrounding, 2pIt is pixel difference weight coefficient.
The neighborhood window definition of LBP description is 3 × 3, central point pixel value is set into threshold value, by by 8 phases of surrounding
The adjacent pixel point value binary coding in contrast for obtaining one 8, is then converted to decimal number and has just obtained central point
LBP encoded radios, the LBP values can be good at reacting the local message of sample image.Fig. 2 is the general principle of LBP codings.
LBP basic thought is the LBP encoded radios by calculating all pixels point, and the picture of original correspondence position is replaced with it
Element value, it is possible to obtain the brand-new LBP coded images of a width.In field of face identification, be not directly with LBP coded images come
Classification and Identification is carried out, but by making statistics with histogram to LBP coded images, statistics LBP encoded radios numerically occur each
Frequency, acquired results are input in algorithm model as characteristic vector and carry out Classification and Identification.Due to being carried out to whole image
The histogram statistical features classifying quality of LBP codings is not fine, so expecting that original sample image can be divided, is obtained
These sub-blocks are encoded to different subregions, then with LBP, then makees statistics with histogram respectively, obtains all subregion
Histogram statistical features, the local message of sample image can be preferably reacted by the piecemeal operation to sample image.
By taking test sample image as an example, the extraction step of LBP features is as follows：
(1) sample image y is divided into 16 × 16 zonule y={ y1,y2,…,yn, wherein, n represents the small of division
Number of regions；
(2) respectively to yiLBP codings are carried out, the LBP code patterns in each region are obtained
(3) it is right respectivelyStatistics with histogram is carried out, that is, counts the frequency of each LBP encoded radios, statistics with histogram is obtained special
Levy vectorial LBP (yi), and make normalized；
(4) the histogram statistical features vector series connection in all regions is obtained into the LBP eigenmatrixes of view picture sample image
LBP (y)=[LBP (y1),LBP(y2),…,LBP(yn)]。
2) design structure type rarefaction representation algorithm model (SSRC).
Compared to traditional SRC algorithms, traditional SRC algorithms do not consider the partitioned organization in dictionary, therefore, I
Classification more more preferable than SRC algorithm can be realized with solution formula (2)：
Wherein, xiRarefaction representation coefficient vector is represented, D represents dictionary matrix, and y is test sample, and I () is an instruction
Function, i.e., be that vacation takes 0, q >=1 when variable is really to take 1.
The purpose of the optimization problem of formula (2) is solved, is exactly to find the minimum non-zero system needed for reconstruct test sample
Several piece.Above formula is a NP problem, and its solution is extremely difficult, and similar to the solution of SRC algorithms, we can be by using
l1Relax to replace solving the optimization problem of above formula：
As q >=1, formula (3) is a convex optimization problem, as q=1, and the convex optimization problem and SRC algorithms of above formula are
Consistent, therefore under suitable condition, the convex optimization problem of SRC algorithms can be regarded a kind of as structural type sparse recovery side
The rarefaction representation of method, i.e. test sample is obtained by minimizing nonzero coefficient block.
We by way of minimizing non-zero reconstruct vector it is also contemplated that classified, shown in such as formula (4)：
Pass through l as q >=11Relaxation is converted into following convex optimization problem：
3) LBP features are input in structural type rarefaction representation sorting algorithm model P1 or P2, by solving l1Norm is most
Smallization problem, it is possible to obtain rarefaction representation coefficient vector x, then according to per class coefficient xiWith dictionary DiCalculate test sample y
Residual error, which kind of residual error is minimum, decides that y belongs to that class, classification formula is as follows：
The two kinds of the present invention SSRC algorithm model LBP-P1 and LBP-P2 based on LBP are carried out on AR face databases
Experiment, and by two kinds of models P1 and P2 of traditional SRC algorithms, LBP-SRC algorithms and SSRC algorithms algorithm as a comparison, come
Checking puies forward the validity of algorithm.
4000 multiple facial images of 126 people are included in AR face databases, everyone facial image is 26, Cong Zhongxuan
All 26 facial images of 100 people are taken to carry out contrast experiment.For everyone all sample images, we randomly select
Wherein 13 as training sample, it is remaining to be used as test sample.We carry out 20 random experiments to every kind of method, and table 1 is given
The average recognition rate of every kind of algorithm is gone out.
The average recognition rate of all algorithms on the AR storehouses of table 1
| Method | Average recognition rate (%) |
| SRC | 89.55 |
| LBP-SRC | 90.73 |
| P1 | 90.48 |
| P2 | 91.27 |
| LBP-P1 | 92.35 |
| LBP-P2 | 93.72 |
As can be seen from Table 1, recognition performances of the LBP-SSRC two kinds of model LBP-P1 and LBP-P2 on AR storehouses be all
It is better than other four kinds of control methods, wherein most preferably the minimum non-zero based on LBP features reconstructs vectorial mould to recognition performance
Type P2 LBP-P2, the discrimination than other method will at least be higher by 1.37% (93.72%-92.35%), the reality more than
Data comparison is tested, our work in the research direction are fully demonstrated meaningful.
It is described above, it is only the embodiment in the present invention, but protection scope of the present invention is not limited thereto, and appoints
What be familiar with the people of the technology disclosed herein technical scope in, it will be appreciated that the conversion or replacement expected, should all cover
Within the scope of the present invention, therefore, protection scope of the present invention should be defined by the protection domain of claims.
Claims (7)
1. the structural type rarefaction representation sorting technique based on LBP features for recognition of face, it is characterised in that including following tool
Body step：
Step 1, the LBP features of original sample concentration training sample and test sample image are extracted respectively；
Step 2, structural type rarefaction representation algorithm model is built；
Step 3, the LBP features of the training sample extracted in step 1 and test sample are input to the classification of structural type rarefaction representation
In algorithm model, according to the output of structural type rarefaction representation algorithm model, the classification results of test sample are obtained.
2. the structural type rarefaction representation sorting technique based on LBP features according to claim 1 for recognition of face, its
It is characterised by, the extracting method of the LBP features of test sample image comprises the following steps in step 1：
1) test sample image y is divided into 16 × 16 zonule, y={ y are designated as1,y2,…,yn, wherein, n represents to divide
Zonule number；
2) respectively to yiLBP codings are carried out, the LBP code patterns of each zonule is obtained, is designated asWherein, i=1,2 ..., n；
3) it is right respectivelyStatistics with histogram is carried out, that is, counts the frequency of each LBP encoded radios, histogram statistical features vector is obtained
LBP(yi), and make normalized；
4) the histogram statistical features vector series connection of all zonules is obtained into the LBP eigenmatrixes of view picture test sample image：
LBP (y)=[LBP (y1),LBP(y2),…,LBP(yn)]。
3. the structural type rarefaction representation sorting technique based on LBP features according to claim 1 for recognition of face, its
It is characterised by, the extracting method of the LBP features of training sample image comprises the following steps in step 1：
1) training sample image D is divided into 16 × 16 zonule, D={ D are designated as1,D2,…,Dn, wherein, n represents to divide
Zonule number；
2) respectively to DiLBP codings are carried out, the LBP code patterns of each zonule is obtained, is designated asWherein, i=1,2 ..., n；
3) it is right respectivelyStatistics with histogram is carried out, that is, counts the frequency of each LBP encoded radios, histogram statistical features vector is obtained
LBP(Di), and make normalized；
4) the histogram statistical features vector series connection of all zonules is obtained into the LBP eigenmatrixes of view picture test sample image：
LBP (D)=[LBP (D1),LBP(D2),…,LBP(Dn)]。
4. the structural type rarefaction representation classification side based on LBP features for recognition of face according to Claims 2 or 3
Method, it is characterised in that LBP encoded radio is obtained by below equation：
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>p</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mn>7</mn>
</munderover>
<mi>s</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>p</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&CenterDot;</mo>
<msup>
<mn>2</mn>
<mi>p</mi>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein,piIt is center pixel pcThe adjacent pixel value of surrounding, 2pIt is pixel difference weight coefficient.
5. the structural type rarefaction representation sorting technique based on LBP features according to claim 4 for recognition of face, its
It is characterised by, structural type rarefaction representation algorithm model is built in step 2 is：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mi>x</mi>
</munder>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mi>q</mi>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>D</mi>
<mo>)</mo>
</mrow>
<mi>x</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, x represents rarefaction representation coefficient vector, x={ x1,x2,…,xn}。
6. the structural type rarefaction representation sorting technique based on LBP features according to claim 4 for recognition of face, its
It is characterised by, structural type rarefaction representation algorithm model is built in step 2 is：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mi>x</mi>
</munder>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mo>|</mo>
<mo>|</mo>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>D</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mi>q</mi>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>L</mi>
<mi>B</mi>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>D</mi>
<mo>)</mo>
</mrow>
<mi>x</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, wherein, x represent rarefaction representation coefficient vector, x={ x1,x2,…,xn}。
7. the structural type rarefaction representation classification side based on LBP features for recognition of face according to claim 5 or 6
Method, it is characterised in that the LBP features of training sample and test sample are input to structural type rarefaction representation sorting algorithm in step 3
In model, rarefaction representation coefficient vector x is obtained；Then according to per class coefficient xiWith LBP (Di) calculate LBP (y) residual error, just sentence
Determine test sample and belong to that minimum class of residual error, classification formula is：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710222367.1A CN107169410A (en) | 2017-03-31 | 2017-03-31 | The structural type rarefaction representation sorting technique based on LBP features for recognition of face |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710222367.1A CN107169410A (en) | 2017-03-31 | 2017-03-31 | The structural type rarefaction representation sorting technique based on LBP features for recognition of face |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107169410A true CN107169410A (en) | 2017-09-15 |
Family
ID=59849149
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710222367.1A Pending CN107169410A (en) | 2017-03-31 | 2017-03-31 | The structural type rarefaction representation sorting technique based on LBP features for recognition of face |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107169410A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108805027A (en) * | 2018-05-03 | 2018-11-13 | 电子科技大学 | Face identification method under the conditions of low resolution |
| CN111310819A (en) * | 2020-02-11 | 2020-06-19 | 深圳前海微众银行股份有限公司 | Data screening method, device, equipment and readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102737234A (en) * | 2012-06-21 | 2012-10-17 | 北京工业大学 | Gabor filtering and joint sparsity model-based face recognition method |
| CN102799870A (en) * | 2012-07-13 | 2012-11-28 | 复旦大学 | Single-training sample face recognition method based on blocking consistency LBP (Local Binary Pattern) and sparse coding |
| CN103632138A (en) * | 2013-11-20 | 2014-03-12 | 南京信息工程大学 | Low-rank partitioning sparse representation human face identifying method |
| CN106066994A (en) * | 2016-05-24 | 2016-11-02 | 北京工业大学 | A kind of face identification method of the rarefaction representation differentiated based on Fisher |
-
2017
- 2017-03-31 CN CN201710222367.1A patent/CN107169410A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102737234A (en) * | 2012-06-21 | 2012-10-17 | 北京工业大学 | Gabor filtering and joint sparsity model-based face recognition method |
| CN102799870A (en) * | 2012-07-13 | 2012-11-28 | 复旦大学 | Single-training sample face recognition method based on blocking consistency LBP (Local Binary Pattern) and sparse coding |
| CN103632138A (en) * | 2013-11-20 | 2014-03-12 | 南京信息工程大学 | Low-rank partitioning sparse representation human face identifying method |
| CN106066994A (en) * | 2016-05-24 | 2016-11-02 | 北京工业大学 | A kind of face identification method of the rarefaction representation differentiated based on Fisher |
Non-Patent Citations (3)
| Title |
|---|
| CHI HO CHAN ET AL: ""SPARSE REPRESENTATION OF (MULTISCALE) HISTOGRAMS FOR FACE RECOGNITION ROBUST TO REGISTRATION AND ILLUMINATION PROBLEMS"", 《PROCEEDINGS OF 2010 IEEE 17TH INTERNATIONAL CONFERENCE ON IMAGE PROCESSING》 * |
| EHSAN ELHAMIFAR ET AL: ""Robust Classification using Structured Sparse Representation"", 《IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| 朱秋双: ""基于改进LBP的人脸识别算法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108805027A (en) * | 2018-05-03 | 2018-11-13 | 电子科技大学 | Face identification method under the conditions of low resolution |
| CN108805027B (en) * | 2018-05-03 | 2020-03-24 | 电子科技大学 | Face recognition method under low resolution condition |
| CN111310819A (en) * | 2020-02-11 | 2020-06-19 | 深圳前海微众银行股份有限公司 | Data screening method, device, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104317902B (en) | Image search method based on local holding iterative quantization Hash | |
| CN105956560B (en) | A kind of model recognizing method based on the multiple dimensioned depth convolution feature of pondization | |
| CN105426919B (en) | The image classification method of non-supervisory feature learning is instructed based on conspicuousness | |
| CN106650830A (en) | Deep model and shallow model decision fusion-based pulmonary nodule CT image automatic classification method | |
| CN101256631B (en) | Method and apparatus for character recognition | |
| CN104346459B (en) | A kind of text classification feature selection approach based on term frequency and chi | |
| CN107947921A (en) | Based on recurrent neural network and the password of probability context-free grammar generation system | |
| CN104298977A (en) | Low-order representing human body behavior identification method based on irrelevance constraint | |
| CN106503672A (en) | A kind of recognition methods of the elderly's abnormal behaviour | |
| CN109978071A (en) | Hyperspectral image classification method based on data augmentation and Multiple Classifier Fusion | |
| CN107133640A (en) | Image classification method based on topography's block description and Fei Sheer vectors | |
| CN104778482A (en) | Hyperspectral image classifying method based on tensor semi-supervised scale cutting dimension reduction | |
| CN101833667A (en) | Pattern recognition classification method expressed based on grouping sparsity | |
| CN106529586A (en) | Image classification method based on supplemented text characteristic | |
| CN111861756A (en) | Group partner detection method based on financial transaction network and implementation device thereof | |
| CN102768732B (en) | Face recognition method integrating sparse preserving mapping and multi-class property Bagging | |
| CN103593674A (en) | Cervical lymph node ultrasonoscopy feature selection method | |
| CN106991049A (en) | A kind of Software Defects Predict Methods and forecasting system | |
| CN102495876A (en) | Nonnegative local coordinate factorization-based clustering method | |
| CN105550712A (en) | Optimized convolution automatic encoding network-based auroral image sorting method | |
| CN107578063B (en) | Image Spectral Clustering based on fast selecting landmark point | |
| CN108805280B (en) | Image retrieval method and device | |
| CN109522953A (en) | The method classified based on internet startup disk algorithm and CNN to graph structure data | |
| CN101877065B (en) | Extraction and identification method of non-linear authentication characteristic of facial image under small sample condition | |
| CN107169410A (en) | The structural type rarefaction representation sorting technique based on LBP features for recognition of face |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170915 |
|
| RJ01 | Rejection of invention patent application after publication |