CN104899578B - A kind of method and device of recognition of face - Google Patents
A kind of method and device of recognition of face Download PDFInfo
- Publication number
- CN104899578B CN104899578B CN201510363785.3A CN201510363785A CN104899578B CN 104899578 B CN104899578 B CN 104899578B CN 201510363785 A CN201510363785 A CN 201510363785A CN 104899578 B CN104899578 B CN 104899578B
- Authority
- CN
- China
- Prior art keywords
- class
- sample
- matrix
- training
- local
- 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.)
- Active
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
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24143—Distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
Abstract
The invention discloses a kind of methods of recognition of face, comprising: the face image data that will acquire is as sample to be tested；The sample to be tested is mapped in low-dimensional feature space using projective transformation matrix, the test sample after being projected；In training sample set, search with the test sample apart from nearest master sample as target sample；The classification of the target sample is determined as to the classification of the test sample；Wherein, the projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to the transformation matrix that multiple samples in the training sample set are trained, so that between class distance is maximum, inter- object distance minimum.The method and device of recognition of face provided by the present invention, for orthogonal differentiation projection construct two adjacency matrix respectively: between class and class in adjacency matrix, unpack indicates between information and class in class, with the information being equalized, to realize in class maximum purpose between minimum and class.
Description
Technical field
The present invention relates to computer vision fields, more particularly to a kind of method and device of recognition of face.
Background technique
Face recognition technology has developed into for research topic popular in computer vision at present, is also simultaneously
One of the most successfully applied in art of image analysis.Human face data is typical higher-dimension Small Sample Database, to human face data into
Row Dimensionality Reduction is necessary pre-treatment step.In the development of recent decades, a series of Dimensionality Reduction skill is proposed in succession
Art.
The orthogonal differentiation projecting method proposed at present only constructs an adjacent map, includes the information in class between class.
And in the unbalanced situation of data distribution, the information in class between class also can be unbalanced in adjacent map, and will lead to cannot be real
Existing inter- object distance minimum and the maximum purpose of between class distance.
Summary of the invention
The object of the present invention is to provide a kind of method and devices of recognition of face, it is therefore intended that solution in the prior art cannot
Realize problem maximum between minimum and class in class.
In order to solve the above technical problems, the present invention provides a kind of method of recognition of face, comprising:
The face image data that will acquire is as sample to be tested；
The sample to be tested is mapped in low-dimensional feature space using projective transformation matrix, the test specimens after being projected
This；
In training sample set, search with the test sample apart from nearest master sample as target sample；
The classification of the target sample is determined as to the classification of the test sample；
Wherein, the projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to described
The transformation matrix that multiple samples in training sample set are trained, so that between class distance is maximum, inter- object distance is minimum.
Optionally, the projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to institute
Stating the transformation matrix that multiple samples in training sample set are trained includes:
Pass through adjacency matrix F in the class of constructionwAnd adjacency matrix F between classb, according to Sw=X (Dw-Fw)XTAnd Sb=X
(Db-Fb)XTLocal Scatter Matrix S is calculated in classwAnd local Scatter Matrix S between classb；
Pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between classbProjective transformation matrix P is calculated,
So that between class distance is maximum, inter- object distance is minimum；
Wherein,
T > 0,WithIt is x respectivelyiSimilar neighbour and foreign peoples neighbour set, DwAnd DbIt is diagonal matrix.
Optionally, described to pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbIt determines and throws
Shadow transformation matrix P includes:
To Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbGeneralized eigen decomposition is carried out, it will
The characteristic value of acquisition is arranged according to sequence from big to small, and the corresponding feature vector of d characteristic value is as the throwing before taking
Shadow transformation matrix P, wherein d is the dimension in the space after projective transformation.
Optionally, the training sample set is combined into the set pre-established, and the process pre-established includes:
Several facial images that will acquire are as training sample；
The training sample is mapped in low-dimensional feature space using the projective transformation matrix, the mark after being projected
Quasi- sample；
The known class of the master sample and the facial image is stored, as the training sample set
It closes.
Optionally, described in training sample set, it searches with the test sample apart from nearest master sample conduct
Target sample includes:
Using nearest neighbor classifier, in the training sample set, search with the test sample apart from nearest mark
Quasi- sample is as target sample.
The present invention also provides a kind of devices of recognition of face, comprising:
Module is obtained, the face image data for will acquire is as sample to be tested；
Mapping block is obtained for the sample to be tested to be mapped in low-dimensional feature space using projective transformation matrix
Test sample after projection；
Searching module is made with the test sample apart from nearest master sample for searching in training sample set
For target sample；
Determining module, for the classification of the target sample to be determined as to the classification of the test sample；
Wherein, adjacent square between adjacency matrix and class in the class that the projective transformation matrix passes through construction for training module
Battle array, to the transformation matrix that multiple samples in the training sample set are trained, so that between class distance is maximum, in class
Distance is minimum.
Optionally, the training module includes:
Training acquiring unit, several facial images for will acquire are as training sample；
Training map unit, for the training sample to be mapped to low-dimensional feature space using the projective transformation matrix
In, the master sample after being projected；
Training storage unit is made for storing the known class of the master sample and the facial image
For the training sample set.
Optionally, the searching module is used in training sample set, is searched nearest with test sample distance
Master sample includes: as target sample
The searching module be specifically used for utilize nearest neighbor classifier, in the training sample set, search with it is described
Test sample is apart from nearest master sample as target sample.
The method and device of recognition of face provided by the present invention, using projective transformation matrix will acquire to test sample
Originally it is mapped in low-dimensional feature space, the test sample after being projected.Then in training sample set, lookup and test specimens
The classification of target sample, as target sample, and is determined as the classification of test sample apart from nearest master sample by this, to reach
To the purpose of recognition of face.The method and device of recognition of face provided by the present invention constructs respectively for orthogonal differentiation projection
Two adjacency matrix: between class and class in adjacency matrix, in class between information and class unpack indicate, with the letter being equalized
Breath, to realize in class maximum purpose between minimum and class.
Detailed description of the invention
Fig. 1 is a kind of method flow diagram of specific embodiment of the method for recognition of face provided by the present invention；
Fig. 2 is that projective transformation matrix is true in another specific embodiment of the method for recognition of face provided by the present invention
Determine the flow chart of process；
Fig. 3 pre-establishes trained sample in another specific embodiment for the method for recognition of face provided by the present invention
The flow chart of the process of this set；
Fig. 4 is the nicety of grading of three kinds of algorithms with dimension change curve；
Fig. 5 is a kind of structural block diagram of specific embodiment of the device of recognition of face provided by the present invention.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, with reference to the accompanying drawings and detailed description
The present invention is described in further detail.Obviously, described embodiments are only a part of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
The method flow diagram of specific embodiment of the method for recognition of face provided by the present invention a kind of as shown in Figure 1,
This method comprises:
Step S101: the face image data that will acquire is as sample to be tested；
Step S102: the sample to be tested is mapped in low-dimensional feature space using projective transformation matrix, is projected
Test sample afterwards；
Step S103: it in training sample set, searches with the test sample apart from nearest master sample as mesh
Standard specimen sheet；
Step S104: the classification of the target sample is determined as to the classification of the test sample；
Wherein, the projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to described
The transformation matrix that multiple samples in training sample set are trained, so that between class distance is maximum, inter- object distance is minimum.
The method of recognition of face provided by the present invention is mapped using the sample to be tested that projective transformation matrix will acquire
Test sample into low-dimensional feature space, after being projected.Then it in training sample set, searches and test sample distance
The classification of target sample is determined as the classification of test sample as target sample by nearest master sample, to reach face
The purpose of identification.The method of recognition of face provided by the present invention constructs two adjacency matrix for orthogonal differentiation projection respectively:
Between class and class in adjacency matrix, in class between information and class unpack indicate, with the information being equalized, to realize in class
Maximum purpose between minimum and class.
It should be pointed out that the relationship between the sample of the interior finger same class of class in the present invention；Refer between class inhomogeneous
Relationship between sample.
The present invention provides another specific embodiments of the method for recognition of face, compared with a upper embodiment, this reality
The determination process that example increases projective transformation matrix is applied, as shown in Figure 2:
Step S201: pass through adjacency matrix F in the class of constructionwAnd adjacency matrix F between classb, according to Sw=X (Dw-Fw)XT
And Sb=X (Db-Fb)XTLocal Scatter Matrix S is calculated in classwAnd local Scatter Matrix S between classb；
Step S202: pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between classbProjection is calculated
Transformation matrix P, so that between class distance is maximum, inter- object distance is minimum；
Wherein,
T > 0,WithIt is x respectivelyiSimilar neighbour and foreign peoples neighbour set, DwAnd DbIt is diagonal matrix.
Pass through interior local Scatter Matrix S as a preferred implementation manner,wAnd local Scatter Matrix S between classbIt determines and throws
Shadow transformation matrix P can be further specifically:
To Scatter Matrix S local in the classwAnd local Scatter Matrix S between classbGeneralized eigen decomposition is carried out, will be obtained
Characteristic value arranged according to sequence from big to small, take before the corresponding feature vector of d characteristic value become as the projection
Matrix P is changed, wherein d is the dimension in the space after projective transformation.
After projective transformation matrix has been determined, the present embodiment additionally provides the process for pre-establishing training sample set, such as
Shown in Fig. 3:
Step S301: several facial images that will acquire are as training sample；
Step S302: the training sample is mapped in low-dimensional feature space using the projective transformation matrix, is obtained
Master sample after projection；
Step S303: the known class of the master sample and the facial image is stored, as the instruction
Practice sample set.
The present invention also provides another specific embodiments of the method for recognition of face, in the present embodiment, ORL face
Database includes 400 facial images of 40 people；Everyone 10 images.The image of some of them face is when different
Phase shooting.The countenance and facial detail of people has a different degrees of variation, for example opens eyes or close one's eyes, wears glasses or not
Wear a pair of spectacles is laughed at or is not laughed at；Human face posture also has considerable degree of variation, and depth rotation and Plane Rotation are up to 200；Every width figure
The size of picture is 32 × 32 pixels, and each pixel is 256 tonal gradations.Random selection 50% is as training sample from database
This, remaining 50% is used as test sample, repeats stochastical sampling 10 times, reports average result.
Specifically, the present embodiment includes face training dataset being established by training and by the face training dataset
The process classified to image.
If existing face training dataset isWherein xi∈RDIt is some human face data, yi=1,2 ..., c }
Indicate xiClass label, c indicate classification number, N indicate training sample total number, D indicate training sample dimension.
N=200 in the present embodiment, c=40, D=1024.It can certainly be other numerical value, this does not affect the present invention
Realization.
In order to consider to keep the geometrical characteristic and training points information of low-dimensional coordinate simultaneously, an optimal transformation P is found, will be counted
According to collectionIt is mapped to the feature space of opposite low-dimensional, such as d dimension space, and d < < D.In the feature space of this low-dimensional,
It maximizes between class distance and minimizes inter- object distance, it may be assumed that
Wherein trace is to ask matrix trace function, SbIt is local Scatter Matrix, S between classwLocal Scatter Matrix in class.In order to count
The two local Scatter Matrixes are calculated, we construct two adjacency matrix, adjacency matrix F in classwThe adjacency matrix F between classb.Then Sw
=X (Dw-Fw)XTAnd Sb=X (Db-Fb)XT, wherein DwAnd DbIt is diagonal matrix,WithFwWith
FbIt is defined as follows:
With
Wherein t > 0 is the parameter of function,WithIt is x respectivelyiSimilar neighbour and foreign peoples neighbour set.?
In the present embodiment, t=8.
In order to obtain P, we are to SbAnd SwCarry out generalized eigen decomposition.The characteristic value of acquisition is suitable according to from big to small
Sequence is ranked up, the corresponding feature vector composition matrix P=[p of its d characteristic value before taking1,p2,…,pd], wherein piIt is feature
Feature vector after decomposition.
After having obtained projection matrix P, the sample in original sample space is projected to low-dimensional feature space, z by projectingi=
PTxi, wherein ziIt is xiIn the projection of lower dimensional space, zi∈Rd.It enablesFor the training sample set after projection.The present embodiment
In, d value changes to 50 from 1.
To some sample to be tested x ∈ RD, it is mapped in low-dimensional feature space using projective transformation P, after obtaining projection
Test sample z=PTx∈Rd。
Using nearest neighbor classifier, classify to the test sample z after projection in low-dimensional feature space.That is,
In training sample setIn, it finds with test sample apart from nearest sample, then the classification of the sample is assigned again
Projective tests sample z.Thus complete the classification to x.Sample to be tested has 200 in the present embodiment, repeats categorization module 200
It is secondary.
Fig. 4 gives the nicety of grading of three kinds of algorithms with dimension change curve.Three kinds of control methods are as follows: orthogonal differentiation
It projects (ODP), differentiates that neighbour is embedded in (DNE) and the present invention.It can be seen that discrimination of the invention is above other two kinds of sides
Method.Table 1 gives the comparison of three kinds of methods top performance when dimensionality reduction number is between 1 to 50, is corresponding best in bracket
Dimension.The present invention just achieves top performance in lower dimension.
Table 1
The present invention provided by recognition of face device a kind of specific embodiment structural block diagram as shown in figure 5,
The device includes:
Module 100 is obtained, the face image data for will acquire is as sample to be tested；
Mapping block 200 is obtained for the sample to be tested to be mapped in low-dimensional feature space using projective transformation matrix
Test sample after to projection；
Searching module 300, for searching with the test sample apart from nearest master sample in training sample set
As target sample；
Determining module 400, for the classification of the target sample to be determined as to the classification of the test sample；
Wherein, adjacent square between adjacency matrix and class in the class that the projective transformation matrix passes through construction for training module
Battle array, to the transformation matrix that multiple samples in the training sample set are trained, so that between class distance is maximum, in class
Distance is minimum.
The device of recognition of face provided by the present invention is mapped using the sample to be tested that projective transformation matrix will acquire
Test sample into low-dimensional feature space, after being projected.Then it in training sample set, searches and test sample distance
The classification of target sample is determined as the classification of test sample as target sample by nearest master sample, to reach face
The purpose of identification.The device of recognition of face provided by the present invention constructs two adjacency matrix for orthogonal differentiation projection respectively:
Between class and class in adjacency matrix, in class between information and class unpack indicate, with the information being equalized, to realize in class
Maximum purpose between minimum and class.
Training module in the device of recognition of face provided by the present invention can further include:
Training acquiring unit, several facial images for will acquire are as training sample；
Training map unit, for the training sample to be mapped to low-dimensional feature space using the projective transformation matrix
In, the master sample after being projected；
Training storage unit is made for storing the known class of the master sample and the facial image
For the training sample set.
As a kind of specific embodiment, searching module is used in training sample set, is searched and the test sample
Include: as target sample apart from nearest master sample
Searching module utilizes nearest neighbor classifier, in the training sample set, searches and the test sample distance
Nearest master sample is as target sample.
Other the specific settings of the device of recognition of face provided by the invention are similar to method, and details are not described herein.
The device of recognition of face provided by the present invention is mapped using the sample to be tested that projective transformation matrix will acquire
Test sample into low-dimensional feature space, after being projected.Then it in training sample set, searches and test sample distance
The classification of target sample is determined as the classification of test sample as target sample by nearest master sample, to reach face
The purpose of identification.The device of recognition of face provided by the present invention constructs two adjacency matrix for orthogonal differentiation projection respectively:
Between class and class in adjacency matrix, in class between information and class unpack indicate, with the information being equalized, to realize in class
Maximum purpose between minimum and class.For the present invention compared with orthogonal differentiation projection algorithm, the present invention can handle data sample distribution not
Equalization problem, and discrimination is higher.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with it is other
The difference of embodiment, same or similar part may refer to each other between each embodiment.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (6)
1. a kind of method of recognition of face characterized by comprising
The face image data that will acquire is as sample to be tested；
The sample to be tested is mapped in low-dimensional feature space using projective transformation matrix, the test sample after being projected；
In training sample set, search with the test sample apart from nearest master sample as target sample；
The classification of the target sample is determined as to the classification of the test sample；
Wherein, the projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to the training
The transformation matrix that multiple samples in sample set are trained, so that between class distance is maximum, inter- object distance is minimum；
The projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to the training sample set
The transformation matrix that multiple samples in conjunction are trained includes:
Pass through adjacency matrix F in the class of constructionwAnd adjacency matrix F between classb, according to Sw=X (Dw-Fw)XTAnd Sb=X (Db-
Fb)XTLocal Scatter Matrix S is calculated in classwAnd local Scatter Matrix S between classb；
Pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between classbProjective transformation matrix P is calculated, so that
Between class distance is maximum, inter- object distance is minimum；
Wherein,
T > 0,WithIt is x respectivelyiSimilar neighbour and foreign peoples neighbour set, DwAnd DbIt is diagonal matrix；
It is described to pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbDetermine projective transformation matrix P
Include:
To Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbGeneralized eigen decomposition is carried out, will be obtained
Characteristic value arranged according to sequence from big to small, take before the corresponding feature vector of d characteristic value become as the projection
Matrix P is changed, wherein d is the dimension in the space after projective transformation.
2. the method for recognition of face as described in claim 1, which is characterized in that the training sample set, which is combined into, to be pre-established
Set, the process pre-established include:
Several facial images that will acquire are as training sample；
The training sample is mapped in low-dimensional feature space using the projective transformation matrix, the standard sample after being projected
This；
The known class of the master sample and the facial image is stored, as the training sample set.
3. the method for recognition of face as described in claim 1, which is characterized in that it is described in training sample set, search with
The test sample includes: as target sample apart from nearest master sample
Using nearest neighbor classifier, in the training sample set, search with the test sample apart from nearest standard sample
This is as target sample.
4. a kind of device of recognition of face characterized by comprising
Module is obtained, the face image data for will acquire is as sample to be tested；
Mapping block is projected for the sample to be tested to be mapped in low-dimensional feature space using projective transformation matrix
Test sample afterwards；
Searching module, for searching with the test sample apart from nearest master sample as mesh in training sample set
Standard specimen sheet；
Determining module, for the classification of the target sample to be determined as to the classification of the test sample；
Wherein, adjacency matrix between adjacency matrix and class, right in the class that the projective transformation matrix passes through construction for training module
The transformation matrix that multiple samples in the training sample set are trained, so that between class distance is maximum, inter- object distance
It is minimum；
The projective transformation matrix is by adjacency matrix between adjacency matrix and class in the class of construction, to the training sample set
The transformation matrix that multiple samples in conjunction are trained includes:
Pass through adjacency matrix F in the class of constructionwAnd adjacency matrix F between classb, according to Sw=X (Dw-Fw)XTAnd Sb=X (Db-
Fb)XTLocal Scatter Matrix S is calculated in classwAnd local Scatter Matrix S between classb；
Pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between classbProjective transformation matrix P is calculated, so that
Between class distance is maximum, inter- object distance is minimum；
Wherein,
T > 0,WithIt is x respectivelyiSimilar neighbour and foreign peoples neighbour set, DwAnd DbIt is diagonal matrix；
It is described to pass through Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbDetermine projective transformation matrix P
Include:
To Scatter Matrix S local in the classwAnd local Scatter Matrix S between the classbGeneralized eigen decomposition is carried out, will be obtained
Characteristic value arranged according to sequence from big to small, take before the corresponding feature vector of d characteristic value become as the projection
Matrix P is changed, wherein d is the dimension in the space after projective transformation.
5. the device of recognition of face as claimed in claim 4, which is characterized in that the training module includes:
Training acquiring unit, several facial images for will acquire are as training sample；
Training map unit, for the training sample to be mapped in low-dimensional feature space using the projective transformation matrix,
Master sample after being projected；
Training storage unit, for storing the known class of the master sample and the facial image, as institute
State training sample set.
6. the device of recognition of face as claimed in claim 4, which is characterized in that the searching module is used in training sample set
In conjunction, lookup includes: as target sample apart from nearest master sample with the test sample
The searching module is specifically used for utilizing nearest neighbor classifier, in the training sample set, searches and the test
Sample is apart from nearest master sample as target sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510363785.3A CN104899578B (en) | 2015-06-26 | 2015-06-26 | A kind of method and device of recognition of face |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510363785.3A CN104899578B (en) | 2015-06-26 | 2015-06-26 | A kind of method and device of recognition of face |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104899578A CN104899578A (en) | 2015-09-09 |
| CN104899578B true CN104899578B (en) | 2019-02-12 |
Family
ID=54032232
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510363785.3A Active CN104899578B (en) | 2015-06-26 | 2015-06-26 | A kind of method and device of recognition of face |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104899578B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105825236A (en) * | 2016-03-18 | 2016-08-03 | 苏州大学 | Method and system for building sample detection model |
| CN106257488B (en) * | 2016-07-07 | 2019-11-19 | 电子科技大学 | A kind of radar target identification method based on neighborhood characteristics space discriminatory analysis |
| CN107203786A (en) * | 2017-06-06 | 2017-09-26 | 苏州大学 | Image-recognizing method and device based on sparse border Fisher algorithms |
| CN107480623B (en) * | 2017-08-07 | 2020-01-07 | 西安电子科技大学 | Neighbor preserving face recognition method based on collaborative representation |
| CN110738248B (en) * | 2019-09-30 | 2022-09-27 | 朔黄铁路发展有限责任公司 | State perception data feature extraction method and device and system performance evaluation method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103679161A (en) * | 2014-01-03 | 2014-03-26 | 苏州大学 | Human-face identifying method and device |
| CN103679162A (en) * | 2014-01-03 | 2014-03-26 | 苏州大学 | Human-face identifying method and system |
| US8687880B2 (en) * | 2012-03-20 | 2014-04-01 | Microsoft Corporation | Real time head pose estimation |
| CN103870848A (en) * | 2014-04-01 | 2014-06-18 | 苏州大学 | Obtaining and sample classification method for projection transformation matrix |
-
2015
- 2015-06-26 CN CN201510363785.3A patent/CN104899578B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8687880B2 (en) * | 2012-03-20 | 2014-04-01 | Microsoft Corporation | Real time head pose estimation |
| CN103679161A (en) * | 2014-01-03 | 2014-03-26 | 苏州大学 | Human-face identifying method and device |
| CN103679162A (en) * | 2014-01-03 | 2014-03-26 | 苏州大学 | Human-face identifying method and system |
| CN103870848A (en) * | 2014-04-01 | 2014-06-18 | 苏州大学 | Obtaining and sample classification method for projection transformation matrix |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104899578A (en) | 2015-09-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104899578B (en) | A kind of method and device of recognition of face | |
| CN104573625B (en) | Data processing system and its generating device and type of face detection method | |
| CN105574527B (en) | A kind of quick object detecting method based on local feature learning | |
| CN107209853A (en) | Positioning and map constructing method | |
| CN111524132B (en) | Method, device and storage medium for identifying abnormal cells in sample to be detected | |
| Lin et al. | A visual-attention model using earth mover's distance-based saliency measurement and nonlinear feature combination | |
| CN103679161B (en) | A kind of face identification method and device | |
| Bojanić et al. | On the comparison of classic and deep keypoint detector and descriptor methods | |
| CN104657709B (en) | Facial image recognition method, device and server | |
| CN112001219A (en) | Multi-angle multi-face recognition attendance checking method and system | |
| CN109389105A (en) | A kind of iris detection and viewpoint classification method based on multitask | |
| CN110033035A (en) | A kind of AOI defect classification method and device based on intensified learning | |
| CN110532880A (en) | Screening sample and expression recognition method, neural network, equipment and storage medium | |
| CN105096293B (en) | Method and apparatus for handling the block to be processed of sediment urinalysis image | |
| CN110135380A (en) | A kind of classroom focus knowledge method for distinguishing based on Face datection | |
| CN111126347A (en) | Human eye state recognition method and device, terminal and readable storage medium | |
| CN109997199A (en) | Tuberculosis inspection method based on deep learning | |
| CN103927529B (en) | The preparation method and application process, system of a kind of final classification device | |
| CN103679162B (en) | A kind of face identification method and system | |
| CN107563324A (en) | A kind of hyperspectral image classification method and device of the learning machine that transfinited based on core basis | |
| CN108038467B (en) | A kind of sparse face identification method of mirror image in conjunction with thickness level | |
| CN104572930B (en) | Data classification method and device | |
| JP2014235033A (en) | Image processing device and program | |
| CN110502961A (en) | A kind of facial image detection method and device | |
| CN108875445A (en) | A kind of pedestrian recognition methods and device again |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |