CN107729840A - One kind is based on sparse figured face recognition discriminant analysis method - Google Patents
One kind is based on sparse figured face recognition discriminant analysis method Download PDFInfo
- Publication number
- CN107729840A CN107729840A CN201710953190.2A CN201710953190A CN107729840A CN 107729840 A CN107729840 A CN 107729840A CN 201710953190 A CN201710953190 A CN 201710953190A CN 107729840 A CN107729840 A CN 107729840A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- mtd
- discriminant analysis
- sketch
- 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.)
- Withdrawn
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2136—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on sparsity criteria, e.g. with an overcomplete basis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The one kind proposed in the present invention is based on sparse figured face recognition discriminant analysis method, and its main contents includes:Adaptive sparse diagramming, discriminant analysis based on space partition zone and the discriminant analysis method for isomery face recognition, its process is, first build expression data set of the facial sketch photo to composition, patch is represented by feature descriptor, then represented by the discriminant analysis of facial match to improve adaptive sparse figure, simply connect the expression of all face-image patches, then distinguish strategy using space and discriminant analysis is carried out to each region, finally calculate the similarity score of three accurate vectors using cosine similarity metric and merged.The discriminant analysis method for isomery face recognition of the present invention, the adaptive sparse vector of facial match is improved, recognition capability and ga s safety degree is improved, effectively eliminates the influence of texture (i.e. style) difference, substantially increase face recognition performance.
Description
Technical field
The present invention relates to area of facial recognition, is differentiated more particularly, to a kind of based on sparse figured face recognition
Analysis method.
Background technology
Facial recognition techniques are as one kind in biometrics identification technology, compared with other biometrics identification technologies,
Have the characteristics that it is direct, conveniently, be easily accepted by, due to its application prospect extensively and scientific research value is great and be increasingly subject to scientific research
The attention of workers, huge progress is also achieved in the past few years.It can be widely applied to government, army, bank,
The fields such as welfare, ecommerce, safe defence, such as enterprise and the recognition of face access control and attendance system and face of house
Identify antitheft door etc., and public security, the administration of justice, criminal investigation can track down and arrest runaway convict in China using face identification system network,
It can carry out authentication in the crowded place such as airport, railway station with reference to E-Passport and identity card, safeguard public field
Safety.However, traditional facial recognition techniques are easily influenceed by big texture (i.e. style) difference, recognition performance is low
Under, it is not easy to promote the use of.
The present invention proposes one kind and is based on sparse figured face recognition discriminant analysis method, first structure face element
- expression data set of the photo to composition is retouched, patch is represented by feature descriptor, then passes through the discriminant analysis of facial match
Represented to improve adaptive sparse figure, simply connect the expression of all face-image patches, then distinguish strategy using space
Discriminant analysis is carried out to each region, finally calculates the similarity score of three accurate vectors simultaneously using cosine similarity metric
Merged.The discriminant analysis method for isomery face recognition of the present invention, improve the adaptive sparse of facial match to
Amount, recognition capability and ga s safety degree are improved, the influence of texture (i.e. style) difference is effectively eliminated, substantially increases face recognition
Performance.
The content of the invention
For recognition performance it is low the problem of, it is an object of the invention to provide one kind to be based on sparse figured face
Discriminant analysis method is identified, expression data set of the facial sketch-photo to composition is first built, spot is represented by feature descriptor
Block, then represented by the discriminant analysis of facial match to improve adaptive sparse figure, simply connect all face-images and mend
The expression of fourth, then distinguish strategy using space and discriminant analysis is carried out to each region, finally using cosine similarity metric
Calculate the similarity score of three accurate vectors and merged.
To solve the above problems, the present invention, which provides one kind, is based on sparse figured face recognition discriminant analysis method,
Its main contents includes:
(1) adaptive sparse diagramming;
(2) discriminant analysis based on space partition zone;
(3) it is used for the discriminant analysis method of isomery face recognition.
Wherein, described adaptive sparse diagramming, construct first by M facial sketch-photos to forming
Represent data set;Each face-image is divided into N number of overlapping patch, and each patch is represented by feature descriptor;
An a given sketch image t and work images g, is classified into patch, and represents each patch with feature descriptor, with
Expression data set before is identical.
Further, described feature descriptor, makes yi(i=1,2 ..., N) represents sketch image, f (yi) correspond to
yiFeature descriptor;The Euclidean distance selection of feature based descriptor surrounds yiTable in region of search R × R of position
Registration is according to sketch closest on each facial sketch of concentration;Therefore, it is possible to find that M related sketch patch, for element
Tracing is as yi;It is also possible to find photograph album photo xiRelevant picture.
Further, described patch, K nearest neighbor neighborhood is selected from the fragment of correlation, and detects sketch piece
Section yiIt may be considered by column vectoryiThe linear combination of K nearest neighbor neighborhood of weighting;
It may then pass through joint and model all sketch image patches and its neighborhood to build Marko's husband's network model:
Wherein, (i, j) ∈ Ξ represent i-th of sketch image, j-th of sketch image neighborhood.
Further, described linear combination,Represent linear group of the feature descriptor of K nearest neighbor neighborhood
Close, such as
Wherein, w isConnection;Matrix Q and c are quadratic parameters;Problem (2) can pass through cascade decomposition
Method solves.
Further, the problem of described (2), can be shown below with optimization problem (2):
Constraint in function (3), such asWithFollowing constraint it is identical;
This is non-negative sparse regularization.
Wherein, the discriminant analysis based on space partition zone, the adaptive of two sketch images and work images is being obtained
After answering sparse figure expression, these expressions are improved by the discriminant analysis of facial match;The table of all face-image patches
Showing simply to link together, then using classical subspace analysis, such as principal component analysis (PCA) and linear discriminant analysis
(LDA) identifying information for matching is extracted;Discriminant analysis based on space partition zone includes three kinds of spaces and distinguishes strategy.
Further, strategy is distinguished in three kinds of described spaces, by the K of image blockcRow are combined as a space partition zone;Then
Discriminant analysis can be carried out respectively on each space partition zone region;Merge extracted feature to be matched;In order to utilize base
In capable space partition zone strategy, KrCapable image block can also carry out discriminant analysis as space partition zone to each region.
Further, described discriminant analysis, experimental section will discuss different Kc,KrAnd KlInfluence, and using optimal
Kc,KrAnd Kl;During the discriminant analysis carried out to each space partition zone, retain first using PCA using 99% variance;
Then, LDA is carried out further to reduce dimension and improve distinguishability;Finally, by all projection vectors of same facial image
Connect, and the similarity scores between sketch image and work images are calculated using cosine similarity metric.
Wherein, the described discriminant analysis method for isomery face recognition, first, face-image is divided it is blocking, and
And each image block is represented using public characteristic descriptor;Secondly, for sketch image (or work images), data are being represented
Markov is built in the sketch image patch (or picture library photo patch) of concentration and the feature of sketch patch (or photo patch)
Network model;It may then pass through solution formula (3) and represented to generate the adaptive sparse figure of input picture;3rd, application
Represent based on row, based on row and space segmentation strategy based on study optimizing adaptive sparse figure and improve it to distinguish
Property;Finally, the similarity score of three accurate vectors is calculated using cosine similarity metric, is then merged.
Brief description of the drawings
Fig. 1 is a kind of system framework figure based on sparse figured face recognition discriminant analysis method of the present invention.
Fig. 2 is a kind of space partition zone strategy based on sparse figured face recognition discriminant analysis method of the present invention.
Fig. 3 is that a kind of isomery face that is used for based on sparse figured face recognition discriminant analysis method of the present invention is known
Other discriminant analysis method.
Embodiment
It should be noted that in the case where not conflicting, the feature in embodiment and embodiment in the application can phase
Mutually combine, the present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.
Fig. 1 is a kind of system framework figure based on sparse figured face recognition discriminant analysis method of the present invention.It is main
To include adaptive sparse diagramming, the discriminant analysis based on space partition zone and the discriminant analysis for isomery face recognition
Method.
Adaptive sparse diagramming, construct first by M facial sketch-photos to the expression data set that forms;
Each face-image is divided into N number of overlapping patch, and each patch is represented by feature descriptor;Give a sketch
An image t and work images g, patch is classified into, and each patch is represented with feature descriptor, with expression number before
It is identical according to collecting.
Make yi(i=1,2 ..., N) represents sketch image, f (yi) correspond to yiFeature descriptor;Feature based describes
The Euclidean distance selection of symbol surrounds yiIn expression data set in region of search R × R of position on each facial sketch away from
From nearest sketch;Therefore, it is possible to find that M related sketch patch, for sketch image yi;It is also possible to find photograph album
Photo xiRelevant picture.
K nearest neighbor neighborhood is selected from the fragment of correlation, and detects sketch fragment yiMay be considered from row to
Amount yiThe linear combination of K nearest neighbor neighborhood of weighting;It may then pass through joint modeling
All sketch image patches and its neighborhood build Marko's husband's network model:
Wherein, (i, j) ∈ Ξ represent i-th of sketch image, j-th of sketch image neighborhood.
Linear combination,The linear combination of the feature descriptor of K nearest neighbor neighborhood is represented, such as
Wherein, w isConnection;Matrix Q and c are quadratic parameters;Problem (2) can pass through cascade decomposition
Method solves.
It can be shown below with optimization problem (2):
Constraint in function (3), such asWithFollowing constraint it is identical;
This is non-negative sparse regularization.
Discriminant analysis based on space partition zone, obtaining the adaptive sparse figure table of two sketch images and work images
After showing, these expressions are improved by the discriminant analysis of facial match;The expression of all face-image patches can simply connect
It is connected together, then using classical subspace analysis, is extracted such as principal component analysis (PCA) and linear discriminant analysis (LDA)
Identifying information for matching;Discriminant analysis based on space partition zone includes three kinds of spaces and distinguishes strategy.
Fig. 2 is a kind of space partition zone strategy based on sparse figured face recognition discriminant analysis method of the present invention.
By the K of image blockcRow are combined as a space partition zone;Then it can carry out differentiating respectively on each space partition zone region and divide
Analysis;Merge extracted feature to be matched;In order to using based on capable space partition zone strategy, KrCapable image block is as space
Subregion, discriminant analysis can also be carried out to each region.
Experimental section will discuss different Kc,KrAnd KlInfluence, and use optimal Kc,KrAnd Kl;To each space partition zone
During the discriminant analysis of progress, retain first using PCA using 99% variance;Then, LDA is carried out to tie up with further reduce
Spend and improve distinguishability;Finally, all projection vectors of same facial image are connected, and uses cosine similarity degree
Measure to calculate the similarity scores between sketch image and work images.
Fig. 3 is that a kind of isomery face that is used for based on sparse figured face recognition discriminant analysis method of the present invention is known
Other discriminant analysis method.First, face-image is divided blocking, and each image is represented using public characteristic descriptor
Block;Secondly, for sketch image (or work images), the sketch image patch (or picture library photo patch) in data set is represented
With structure markov network model in the feature of sketch patch (or photo patch);It may then pass through solution formula (3)
The adaptive sparse figure for generating input picture represents;3rd, split plan using the space based on row, based on row and based on study
Slightly represent to optimize adaptive sparse figure and improve its ga s safety degree;Finally, three are calculated using cosine similarity metric
The similarity score of accurate vector, is then merged.
For those skilled in the art, the present invention is not restricted to the details of above-described embodiment, in the essence without departing substantially from the present invention
In the case of refreshing and scope, the present invention can be realized with other concrete forms.In addition, those skilled in the art can be to this hair
Bright to carry out various changes and modification without departing from the spirit and scope of the present invention, these improvement and modification also should be regarded as the present invention's
Protection domain.Therefore, appended claims are intended to be construed to include preferred embodiment and fall into all changes of the scope of the invention
More and change.
Claims (10)
1. one kind is based on sparse figured face recognition discriminant analysis method, it is characterised in that main including adaptive dilute
Dredge diagramming (one);Discriminant analysis (two) based on space partition zone;Discriminant analysis method for isomery face recognition
(3).
2. based on the adaptive sparse diagramming (one) described in claims 1, it is characterised in that construct first by M
Expression data set of the individual facial sketch-photo to composition;Each face-image is divided into N number of overlapping patch, and passes through spy
Descriptor is levied to represent each patch;An a given sketch image t and work images g, is classified into patch, and with spy
Sign descriptor represents each patch, identical with expression data set before.
3. based on the feature descriptor described in claims 1, it is characterised in that make yi(i=1,2 ..., N) represent sketch map
Picture, f (yi) correspond to yiFeature descriptor;The Euclidean distance selection of feature based descriptor surrounds yiPosition is searched
Sketch closest on each facial sketch in data set is represented in the R × R of rope region;Therefore, it is possible to find that M related
Sketch patch, for sketch image yi;It is also possible to find photograph album photo xiRelevant picture.
4. based on the patch described in claims 3, it is characterised in that K nearest neighbor neighborhood is selected from the fragment of correlation,
And detect sketch fragment yiIt may be considered by column vectoryiK nearest neighbor of weighting is adjacent
The linear combination in domain;It may then pass through joint and model all sketch image patches and its neighborhood to build Marko's husband's network mould
Type:
<mrow>
<mi>p</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mi>N</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>y</mi>
<mi>N</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mi>&Pi;</mi>
<mi>i</mi>
</munder>
<mi>&Phi;</mi>
<mrow>
<mo>(</mo>
<mi>f</mi>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
<mo>,</mo>
<mi>f</mi>
<mo>(</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
</msub>
<mo>)</mo>
<mo>)</mo>
</mrow>
<munder>
<mi>&Pi;</mi>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>&Element;</mo>
<mi>&Xi;</mi>
</mrow>
</munder>
<mi>&Psi;</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
</msub>
<mo>,</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
</msub>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, (i, j) ∈ Ξ represent i-th of sketch image, j-th of sketch image neighborhood.
5. based on the linear combination described in claims 4, it is characterised in thatRepresent the feature of K nearest neighbor neighborhood
The linear combination of descriptor, such as
<mrow>
<mtable>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow>
<munder>
<mi>min</mi>
<mi>w</mi>
</munder>
<msup>
<mi>w</mi>
<mi>T</mi>
</msup>
<mi>Q</mi>
<mi>w</mi>
<mo>+</mo>
<msup>
<mi>w</mi>
<mi>T</mi>
</msup>
<mi>c</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>K</mi>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>k</mi>
</mrow>
</msub>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>0</mn>
<mo>&le;</mo>
<msub>
<mi>w</mi>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>k</mi>
</mrow>
</msub>
<mo>&le;</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>..</mn>
<mo>,</mo>
<mi>N</mi>
<mo>,</mo>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>...</mn>
<mi>K</mi>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, w isConnection;Matrix Q and c are quadratic parameters;Problem (2) can be asked by cascade decomposition method
Solution.
6. the problem of based on described in claims 5 (2), it is characterised in that can be shown below with optimization problem (2):
<mrow>
<mtable>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow>
<munder>
<mi>min</mi>
<mi>w</mi>
</munder>
<msup>
<mi>w</mi>
<mi>T</mi>
</msup>
<mi>Q</mi>
<mi>w</mi>
<mo>+</mo>
<msup>
<mi>w</mi>
<mi>T</mi>
</msup>
<mi>c</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>s</mi>
<mo>.</mo>
<mi>t</mi>
<mo>.</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>m</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>M</mi>
</munderover>
<msub>
<mi>w</mi>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>m</mi>
</mrow>
</msub>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>0</mn>
<mo>&le;</mo>
<msub>
<mi>w</mi>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>m</mi>
</mrow>
</msub>
<mo>&le;</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>..</mn>
<mo>,</mo>
<mi>N</mi>
<mo>,</mo>
<mi>m</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>...</mn>
<mi>M</mi>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Constraint in function (3), such asWithFollowing constraint it is identical;
<mrow>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>w</mi>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
</msub>
<mo>|</mo>
<msub>
<mo>|</mo>
<mn>1</mn>
</msub>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>0</mn>
<mo>&le;</mo>
<msub>
<mi>w</mi>
<mrow>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<mi>m</mi>
</mrow>
</msub>
<mo>&le;</mo>
<mn>1</mn>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
This is non-negative sparse regularization.
7. based on the discriminant analysis (two) based on space partition zone described in claims 1, it is characterised in that obtaining two elements
After the adaptive sparse figure of tracing picture and work images represents, these tables are improved by the discriminant analysis of facial match
Show;The expression of all face-image patches can simply link together, then using classical subspace analysis, such as principal component
(PCA) and linear discriminant analysis (LDA) are analyzed to extract the identifying information for matching;Discriminant analysis bag based on space partition zone
Include three kinds of spaces and distinguish strategy.
8. distinguish strategy based on three kinds of spaces described in claims 7, it is characterised in that by the K of image blockcRow are combined as one
Individual space partition zone;Then discriminant analysis can be carried out respectively on each space partition zone region;Merge extracted feature to carry out
Matching;In order to using based on capable space partition zone strategy, KrCapable image block can also enter as space partition zone to each region
Row discriminant analysis.
9. based on the discriminant analysis described in claims 8, it is characterised in that experimental section will discuss different Kc,KrAnd KlShadow
Ring, and use optimal Kc,KrAnd Kl;During the discriminant analysis carried out to each space partition zone, applied first using PCA
99% variance retains;Then, LDA is carried out further to reduce dimension and improve distinguishability;Finally, by same facial image
All projection vectors connect, and calculated using cosine similarity metric similar between sketch image and work images
Property fraction.
10. based on the discriminant analysis method (three) for isomery face recognition described in claims 1, it is characterised in that first
First, face-image is divided blocking, and each image block is represented using public characteristic descriptor;Secondly, for sketch map
As (or work images), sketch image patch (or picture library photo patch) and sketch patch in data set is represented (or photo
Patch) feature on build markov network model;Solution formula (3) be may then pass through to generate the adaptive of input picture
Sparse figure is answered to represent;3rd, optimize adaptive sparse using based on row, based on row and space segmentation strategy based on study
Figure represents and improves its ga s safety degree;Finally, obtained using cosine similarity metric to calculate the similitude of three accurate vectors
Point, then merged.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710953190.2A CN107729840A (en) | 2017-10-13 | 2017-10-13 | One kind is based on sparse figured face recognition discriminant analysis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710953190.2A CN107729840A (en) | 2017-10-13 | 2017-10-13 | One kind is based on sparse figured face recognition discriminant analysis method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107729840A true CN107729840A (en) | 2018-02-23 |
Family
ID=61211244
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710953190.2A Withdrawn CN107729840A (en) | 2017-10-13 | 2017-10-13 | One kind is based on sparse figured face recognition discriminant analysis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107729840A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110334238A (en) * | 2019-03-27 | 2019-10-15 | 特斯联(北京)科技有限公司 | A kind of Missing Persons based on recognition of face trace method and system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103544507A (en) * | 2013-10-15 | 2014-01-29 | 中国矿业大学 | High spectral data dimension reduction method based on pairwise constraint discriminant analysis-nonnegative sparse divergence |
CN104123741A (en) * | 2014-06-24 | 2014-10-29 | 小米科技有限责任公司 | Method and device for generating human face sketch |
-
2017
- 2017-10-13 CN CN201710953190.2A patent/CN107729840A/en not_active Withdrawn
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103544507A (en) * | 2013-10-15 | 2014-01-29 | 中国矿业大学 | High spectral data dimension reduction method based on pairwise constraint discriminant analysis-nonnegative sparse divergence |
CN104123741A (en) * | 2014-06-24 | 2014-10-29 | 小米科技有限责任公司 | Method and device for generating human face sketch |
Non-Patent Citations (1)
Title |
---|
CHUNLEI PENG: "Sparse Graphical Representation based Discriminant Analysis for Heterogeneous Face Recognition", 《ARXIV:1607.00137V1》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110334238A (en) * | 2019-03-27 | 2019-10-15 | 特斯联(北京)科技有限公司 | A kind of Missing Persons based on recognition of face trace method and system |
CN110334238B (en) * | 2019-03-27 | 2020-01-31 | 特斯联(北京)科技有限公司 | missing population tracing method and system based on face recognition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Carvalho et al. | Illuminant-based transformed spaces for image forensics | |
US8184914B2 (en) | Method and system of person identification by facial image | |
Demirkus et al. | Automated person categorization for video surveillance using soft biometrics | |
CN107609497A (en) | The real-time video face identification method and system of view-based access control model tracking technique | |
US11594074B2 (en) | Continuously evolving and interactive Disguised Face Identification (DFI) with facial key points using ScatterNet Hybrid Deep Learning (SHDL) network | |
Paulino et al. | Latent fingerprint indexing: Fusion of level 1 and level 2 features | |
Li et al. | Age-related factor guided joint task modeling convolutional neural network for cross-age face recognition | |
Wang et al. | S 3 D: Scalable pedestrian detection via score scale surface discrimination | |
Ng et al. | Matching of interest point groups with pairwise spatial constraints | |
Grosz et al. | Minutiae-guided fingerprint embeddings via vision transformers | |
Winter et al. | Demystifying face-recognition with locally interpretable boosted features (libf) | |
CN107729840A (en) | One kind is based on sparse figured face recognition discriminant analysis method | |
Di Domenico et al. | Combining identity features and artifact analysis for differential morphing attack detection | |
Velardo et al. | Face recognition with DAISY descriptors | |
Ram et al. | Biohashing application using fingerprint cancelable features | |
Guo et al. | Occlusion-robust face detection using shallow and deep proposal based faster R-CNN | |
Kartheek et al. | Texture based feature extraction using symbol patterns for facial expression recognition | |
Janarthanan et al. | An Efficient Face Detection and Recognition System Using RVJA and SCNN | |
Heusch et al. | A novel statistical generative model dedicated to face recognition | |
Hast | Age-invariant face recognition using face feature vectors and embedded prototype subspace classifiers | |
Chang et al. | Combination of RFID and face recognition for access control system | |
Contardo et al. | Analyzing the impact of police mugshots in face verification for crime investigations | |
Liu et al. | Set-based classification for person re-identification utilizing mutual-information | |
Hu et al. | Online handwriting signature verification based on template clustering | |
Feng et al. | Fingerprint representation and matching in ridge coordinate system |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20180223 |