CN108108760A - A kind of fast human face recognition - Google Patents
A kind of fast human face recognition Download PDFInfo
- Publication number
- CN108108760A CN108108760A CN201711378336.1A CN201711378336A CN108108760A CN 108108760 A CN108108760 A CN 108108760A CN 201711378336 A CN201711378336 A CN 201711378336A CN 108108760 A CN108108760 A CN 108108760A
- Authority
- CN
- China
- Prior art keywords
- mrow
- face
- mfrac
- picture
- measured
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- 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
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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 present invention relates to a kind of fast human face recognition, including:(1) face key feature points, construction standard feature vector are extracted;(2) it is several as with reference to variable in selection standard feature vector, piecemeal is carried out, and carries out binary coding;(3) PCA dimension-reduction treatment is carried out respectively to each piece, form the block after several dimensionality reductions, several corresponding piece is named with the coding to block;(4) face picture to be measured is read, extracts face key feature points to be measured, construction standard feature vector;(5) binary coding is carried out to standard feature vector, according to the block after the coding lookup to corresponding dimensionality reduction;(6) PCA dimension-reduction treatment is carried out to face picture to be measured;(7) face picture to be measured and the similarity in the block per class picture are determined, when similarity is more than the threshold value of setting, that is, judges that face picture to be measured belongs to same people with such picture, completes identification process.
Description
Technical field
The present invention relates to a kind of fast human face recognitions, belong to technical field of face recognition.
Background technology
Face recognition technology is one kind of biometrics identification technology, is widely used in security protection, finance, telecommunications, traffic etc.
Field.Correlation technique is extracted from video flowing using face common feature containing face picture, then by facial image with prestoring
Template image is compared in face database, and the identity information of face to be measured is determined using the difference between face characteristic.
Face recognition process is broadly divided into face and obtains and position, three steps of image preprocessing and recognition of face,
Middle recognition of face is a step of most critical, and the recognizer of feature based includes the method based on PCA, the method based on LDA, base
In LBP methods, the method based on rarefaction representation, utilize method of neutral net etc..
Two important indicators for weighing face identification method quality are discrimination and recognition speed respectively, influence recognition speed
A key factor be identification process generate calculation amount.Classical PCA methods by by original image project to one it is low
Dimensional space extracts the main component of image information, be greatly reduced identification comparison process in calculation amount, allow recognition of face by
It is gradually practical.But picture to be measured will one by one be compared with the face in object library, as face trains classification (number) in storehouse
Increase, recognition speed can be slack-off, simultaneously because the probability that similar face occurs becomes larger, discrimination also can rapid decrease.For one
A little special application scenarios, such as the area that station, school, store flow of the people are big, huge object library is can hardly be avoided, and is identified
Requirement of real-time it is very high, there is basic contradictions.
Therefore, the quick identification to facial image how is realized in the case of big object library, and improves recognition of face
Accuracy become it is urgently to be resolved hurrily the technical issues of.
Chinese patent literature CN105550657A discloses the improvement SIFT face feature extraction methods based on key point, this
Method employs the method for improving the extraction of SIFT face characteristics based on key point.Pass through five crucial pixels in locating human face
Point, and this five key points are described using the direction histogram in SIFT methods, so as to form the facial image feature of robust
Vector.The similarity score value between two face feature vectors is calculated with reference to bilinearity similarity function and mahalanobis distance.
Two-value classification, the higher a kind of face picture of score value, two face figures are carried out to similarity score value using KELM graders
Piece is judged to come from same person, and a kind of face picture that score value is relatively low, two face pictures are judged to come from
In different people.But there are the method that calculating similarity score one by one is used in recognition of face comparison process, every width for the patent
Picture to be measured will compare once with all pictures in object library, with the increase of object library, calculation amount will rapid growth, locating
On the premise of reason equipment is certain, it is impossible to ensure recognition of face requirement of real-time.
The content of the invention
In view of the deficiencies of the prior art, the present invention provides a kind of fast human face recognition, to improve in big training sample
Recognition speed and accuracy rate in the case of this.
Term is explained:
PCA dimensionality reductions, principal component analysis is a kind of statistical analysis technique for grasping things principal contradiction, it can be from polynary
Major influence factors are parsed in things, simplify the problem of complicated.Calculate principal component method be by high dimensional data project to compared with
Lower dimensional space.It is explained in detail below:If training set has the image of m width n*p sizes, a m row is first constructed, the matrix of n*p row should
Every a line of matrix be piece image full detail (pixel value from left to right, from top to bottom form containing n*p element
Row vector), each row are a stochastic variables, then all row of matrix are handled (each element subtracts the column mean,
It is 0 to realize column mean;The covariance matrix of the matrix is calculated again, and eigen vector is calculated further according to covariance matrix,
Then feature vector is sorted according to characteristic value size) so-called principal component is the feature vector with maximum eigenvalue, then
The feature vector of (k is much smaller than n*p, and k is bigger, and precision is better, and dimensionality reduction effect is poorer) forms one k before being selected according to required precision
A new matrix n*p*k, by original image matrix m*n*p and new matrix multiple to get to the image array m*k after dimensionality reduction, this is new
Matrix realizes dimensionality reduction on the premise of original image main component is retained.
The technical scheme is that:
A kind of fast human face recognition, including:
A, step (1)-(3) are performed respectively to each Target Photo in training set, establishes object library
(1) face key feature points are extracted, and size, rotation and shift invariant are had according to face key feature points construction
Property standard feature vector;
(2) several components in selection standard feature vector are used as with reference to variable, are carried out piecemeal, and are carried out two to block
Scale coding;
(3) PCA dimension-reduction treatment is carried out respectively to each block in step (2), forms the block after several dimensionality reductions, use step
(2) coding of block is named corresponding piece after several dimensionality reductions in;
B, recognition of face
(4) face picture to be measured is read, extracts face key feature points to be measured, and according to face key feature points structure to be measured
Make the standard feature vector of the face to be measured with size, rotation and shift invariant;
(5) binary coding is carried out to the standard feature vector of step (4), according to the coding lookup to corresponding step
(3) block after dimensionality reduction;
(6) PCA dimension-reduction treatment is carried out to face picture to be measured;
(7) face picture to be measured after step (6) PCA dimension-reduction treatment is found with step (5) corresponding in the block each
Class picture is compared, and determines face picture to be measured and the similarity in the block per class picture, when similarity is more than the threshold of setting
During value, that is, judge that face picture to be measured belongs to same people with such picture, complete identification process.
It is preferred according to the present invention, the step (1), including:
A, face key feature points are extracted, the face key feature points include eyes canthus point, prenasale and two mouths
Angle point, the two ear basal parts of the ear amount to 9 key feature points;
B, 10 distance feature values are formed according to 9 key feature points, including:Left eye width d1, nose and eyes line
Vertical range d2, distance d3, the width d4 of face, right eye width d5, two outside horizontal distance d6, right eyes between the two ear basal parts of the ear
Outside canthus and the distance d7 of nose, the distance d8 of the inside canthus of left eye and nose, face midpoint it is vertical with nose away from
From d9, nose and corners of the mouth distance d10;
C, the ratio between 10 distance feature values and eyes midpoint to the vertical range L between face midpoint are defined as standardizing
Characteristic value, by all standardized feature value arrangement form standard features vectorAs shown in formula (I):
Preferred according to the present invention, the step (2) sets m component in selection standard feature vector as ginseng
Examine variable, 1≤m≤10, including:
D, choose1≤i≤m, the corresponding standardized feature value of proprietary face picture that obtaining step (1) obtainsEveryone corresponding standardized feature value of face pictureIt is definite value;
E, the corresponding standardized feature value of proprietary face picture is calculatedIntermediate value;
F, the intermediate value being calculated according to step e carries out two points to the proprietary standard feature vector that step (1) obtains
Classification, is carried out at the same time binary coding, if in standard feature vectorMore than or equal to intermediate value, then 1 is encoded to;Otherwise, encode
For 0.
Successively for eachStep D-F is performed, training set is divided into 2mA block (subset).Each block is exactly a small mesh
Mark storehouse, the face picture containing multiple people.
It is preferred according to the present invention, the step (7), including:
G, the picture to be measured after step (6) PCA dimension-reduction treatment and step (5) are found pair by calculating cosine similarity
All kinds of pictures in the block answered are matched, shown in the calculation formula such as formula (II) of cosine similarity ξ:
In formula (II), XjRepresent j-th of component of the vector generated after face picture PCA dimension-reduction treatment to be measured, RjRepresent step
Suddenly (5) find j-th of component for generating vector in corresponding piece after certain class face picture dimensionality reduction, and n represents the dimension after dimensionality reduction;
H, the similarity ξ of face picture to be measured and target face picture is obtained according to step g, if ξ is more than 80%, then sentence
Fixed face picture to be measured belongs to same people with such picture, otherwise, then judges that face picture to be measured is not belonging to together with such picture
One people.
Beneficial effects of the present invention are:
On the technical foundation of conventional face's identification, present invention employs presorted simultaneously to object library with key feature points
Big object library using the priori to face characteristic, is divided into Small object storehouse by the method for coding, effectively reduces face knowledge
Not Bi Dui link calculation amount, in the case that processing equipment is constant while all kinds of recognizer advantages are retained, can significantly carry
High recognition of face speed meets the requirement of real-time of different application scene.Also, the present invention can be with other face identification methods
It is used in conjunction with, presorting for object library in training link is completed, is not take up real-time comparison time.
Description of the drawings
Fig. 1 is face key feature points of the present invention and the schematic diagram of distance feature value;
Fig. 2 is blocking process flow diagram of the present invention;
Fig. 3 is the flow diagram of fast human face recognition of the present invention.
Specific embodiment
The present invention is further qualified with reference to Figure of description and embodiment, but not limited to this.
Embodiment 1
A kind of fast human face recognition, as shown in figure 3, including:
A, step (1)-(3) are performed respectively to each Target Photo in training set, establishes object library
(1) face key feature points are extracted, and size, rotation and shift invariant are had according to face key feature points construction
Property standard feature vector;
(2) several components in selection standard feature vector are used as with reference to variable, are carried out piecemeal, and are carried out two to block
Scale coding,;
(3) PCA dimension-reduction treatment is carried out respectively to each block in step (2), forms the block after several dimensionality reductions, use step
(2) coding of block is named corresponding piece after several dimensionality reductions in;
The calculation amount of recognition of face can substantially be described with a simplified example, if there are 1000 100*100 in object library
Face picture, we will judge whether the face of front end camera shooting is member in storehouse, we will by the photo with
1000 photos compare calculating one by one, since computer disposal digital image information is in units of pixel, if some is calculated
Method is to the processing of single pixel only with 1 step, then the amount of calculation 1000*100*100, it is actual to be much larger than this.The effect of PCA is
100*100 below is solved, on the premise of main information ingredient is retained, dimension is reduced, for example 100*100 is become into 100*
10, reduce operand.Meaning of the present invention be reduce front that 1000, what is utilized is that people know the priori of face characteristic
Know, for example according to nose size, determine a median, 1000 faces in object library are divided into russian and spur two
Point, it is assumed that respectively account for half, 500 people of russian, 500 people of spur, before picture recognition to be measured, first by his nose and median
Compare, if belonging to russian, the identification process belonged to afterwards only carries out in russian object library, can save the calculating of half
Amount, similarly, can also be certain as reference value etc. according to eye widths, this is merely illustrative, more multiple than this in the application
It is much miscellaneous, according to priori, cannot effectively it be classified by single face organ's size, we are to different face differences
Understanding is more based on the restriction relation between organ, carries out two points of maximum difficult points of classification and is that searching feature, being will be first
Find to expression, illumination-insensitive, can to the apparent feature of face classification, chosen in embodiment be all organ geometry it is special
Sign, further optimization algorithm, it is necessary to largely test, and finds the feature of more multi-environment consistency, the feature of selection is more, mesh
Mark storehouse just get it is thinner, calculate benefit it is better.
B, recognition of face
(4) face picture to be measured is read, extracts face key feature points to be measured, and according to face key feature points structure to be measured
Make the standard feature vector of the face to be measured with size, rotation and shift invariant;
(5) binary coding is carried out to the standard feature vector of step (4), according to the coding lookup to corresponding step
(3) block after dimensionality reduction;
(6) PCA dimension-reduction treatment is carried out to face picture to be measured;
(7) face picture to be measured after step (6) PCA dimension-reduction treatment is found with step (5) corresponding in the block each
Class picture is compared, and determines face picture to be measured and the similarity in the block per class picture, when similarity is more than the threshold of setting
During value, that is, judge that face picture to be measured belongs to same people with such picture, complete identification process.
Embodiment 2
A kind of fast human face recognition according to embodiment 1, difference lies in,
The step (1), including:
A, extract face key feature points, face key feature points include eyes canthus point, prenasale and two corners of the mouth points,
The two ear basal parts of the ear amount to 9 key feature points;
B, 10 distance feature values are formed according to 9 key feature points, including:Left eye width d1, nose and eyes line
Vertical range d2, distance d3, the width d4 of face, right eye width d5, two outside horizontal distance d6, right eyes between the two ear basal parts of the ear
Outside canthus and the distance d7 of nose, the distance d8 of the inside canthus of left eye and nose, face midpoint it is vertical with nose away from
From d9, nose and corners of the mouth distance d10;Face key feature points and the schematic diagram of distance feature value are as shown in Figure 1;
C, the ratio between 10 distance feature values and eyes midpoint to the vertical range L between face midpoint are defined as standardizing
Characteristic value, by all standardized feature value arrangement form standard features vectorAs shown in formula (I):
Embodiment 3
A kind of fast human face recognition according to embodiment 2, difference lies in,
The step (2), as shown in Fig. 2, m component in setting selection standard feature vector, which is used as, refers to variable, 1
≤ m≤10, including:
D, choose1≤i≤m, the corresponding standardized feature value of proprietary face picture that obtaining step (1) obtainsEveryone corresponding standardized feature value of face pictureIt is definite value;
E, the corresponding standardized feature value of proprietary face picture is calculatedIntermediate value;
F, the intermediate value being calculated according to step e carries out two points to the proprietary standard feature vector that step (1) obtains
Classification, is carried out at the same time binary coding, if in standard feature vectorMore than or equal to intermediate value, then 1 is encoded to;Otherwise, encode
For 0.
Successively for eachStep D-F is performed, training set is divided into 2mA block (subset).Each block is exactly a small mesh
Mark storehouse, the face picture containing multiple people.
Embodiment 4
A kind of fast human face recognition according to embodiment 3, difference lies in,
The step (7), including:
G, the picture to be measured after step (6) PCA dimension-reduction treatment and step (5) are found pair by calculating cosine similarity
All kinds of pictures in the block answered are matched, shown in the calculation formula such as formula (II) of cosine similarity ξ:
In formula (II), XjRepresent j-th of component of the vector generated after face picture PCA dimension-reduction treatment to be measured, RjRepresent step
Suddenly (5) find j-th of component for generating vector in corresponding piece after certain class face picture dimensionality reduction, and n represents the dimension after dimensionality reduction;
H, the similarity ξ of face picture to be measured and target face picture is obtained according to step g, if ξ is more than 80%, then sentence
Fixed face picture to be measured belongs to same people with such picture, otherwise, then judges that face picture to be measured is not belonging to together with such picture
One people.
The application employs the method presorted and encoded to object library with key feature points, utilizes the elder generation to face characteristic
Knowledge is tested, big object library is divided into Small object storehouse, effectively reduces the calculation amount that recognition of face compares link, it is all kinds of retaining
While recognizer advantage, in the case that processing equipment is constant, recognition of face speed can be significantly improved, meets different application field
The requirement of real-time of scape.
If the key feature number chosen is 10, identify that comparing the computationally intensive of link is approximately equivalent to the 1/ of former algorithm
1024, the calculation amount of identification ring successively about out subtracts the increased calculation amount of key feature points extraction process, is entire identification
The calculation amount of algorithm saving, it is typically certain value to treat mapping piece key feature extraction calculation amount, and object library is bigger, of the invention
Benefit is better.
Claims (4)
1. a kind of fast human face recognition, which is characterized in that including:
A, step (1)-(3) are performed respectively to each Target Photo in training set, establishes object library
(1) face key feature points are extracted, and are constructed according to face key feature points with size, rotation and shift invariant
Standard feature vector;
(2) several components in selection standard feature vector are used as with reference to variable, are carried out piecemeal, and are carried out binary system to block
Coding;
(3) PCA dimension-reduction treatment is carried out respectively to each block in step (2), the block after several dimensionality reductions is formed, in step (2)
The coding of block is named corresponding piece after several dimensionality reductions;
B, recognition of face
(4) face picture to be measured is read, face key feature points to be measured is extracted, and is constructed and had according to face key feature points to be measured
There is the standard feature vector of the face to be measured of size, rotation and shift invariant;
(5) binary coding is carried out to the standard feature vector of step (4), is dropped according to the coding lookup to corresponding step (3)
Block after dimension;
(6) PCA dimension-reduction treatment is carried out to face picture to be measured;
(7) face picture to be measured after step (6) PCA dimension-reduction treatment is found into corresponding all kinds of figures in the block with step (5)
Piece is compared, and determines face picture to be measured and the similarity in the block per class picture, when similarity is more than the threshold value of setting,
Judge that face picture to be measured belongs to same people with such picture, complete identification process.
2. a kind of fast human face recognition according to claim 1, which is characterized in that the step (1), including:
A, extract face key feature points, the face key feature points include eyes canthus point, prenasale and two corners of the mouth points,
The two ear basal parts of the ear amount to 9 key feature points;
B, 10 distance feature values are formed according to 9 key feature points, including:Left eye width d1, nose and eyes line hang down
Directly distance d3 between distance d2, the two ear basal parts of the ear, the width d4 of face, right eye width d5, two outside horizontal distance d6, right eye it is outer
The distance d7 of branch hole angle and nose, the distance d8 at the inside canthus of left eye and nose, face midpoint and nose vertical range d9,
The distance d10 of nose and the corners of the mouth;
C, the ratio between 10 distance feature values and eyes midpoint to the vertical range L between face midpoint are defined as standardized feature
Value, by all standardized feature value arrangement form standard features vectorAs shown in formula (I):
<mrow>
<mover>
<mi>A</mi>
<mo>&RightArrow;</mo>
</mover>
<mo>=</mo>
<mo>{</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>1</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>2</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>3</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>4</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>5</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>6</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>7</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>8</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>9</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>,</mo>
<mfrac>
<mrow>
<mi>d</mi>
<mn>10</mn>
</mrow>
<mi>L</mi>
</mfrac>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mo>)</mo>
</mrow>
<mo>.</mo>
</mrow>
3. a kind of fast human face recognition according to claim 1, which is characterized in that the step (2), setting are chosen
M component in standardized feature vector is used as with reference to variable, 1≤m≤10, including:
D, choose1≤i≤m, the corresponding standardized feature value of proprietary face picture that obtaining step (1) obtainsEach
The corresponding standardized feature value of face picture of peopleIt is definite value;
E, the corresponding standardized feature value of proprietary face picture is calculatedIntermediate value;
F, the intermediate value being calculated according to step e carries out two points points to the proprietary standard feature vector that step (1) obtains
Class is carried out at the same time binary coding, if in standard feature vectorMore than or equal to intermediate value, then 1 is encoded to;Otherwise, it is encoded to
0;
Successively for eachStep D-F is performed, training set is divided into 2mA block.
4. a kind of fast human face recognition according to claim 1, which is characterized in that the step (7), including:
G, found with step (5) by calculating cosine similarity by the picture to be measured after step (6) PCA dimension-reduction treatment corresponding
All kinds of pictures in the block are matched, shown in the calculation formula such as formula (II) of cosine similarity ξ:
<mrow>
<mi>&xi;</mi>
<mo>=</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mi>&theta;</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>j</mi>
</msub>
<mo>&times;</mo>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msqrt>
<mrow>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
<mo>&times;</mo>
<msqrt>
<mrow>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>R</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mi>I</mi>
<mi>I</mi>
<mo>)</mo>
</mrow>
</mrow>
In formula (II), XjRepresent j-th of component of the vector generated after face picture PCA dimension-reduction treatment to be measured ,-jRepresent step
(5) j-th of component for generating vector in corresponding piece after certain class face picture dimensionality reduction is found, n represents the dimension after dimensionality reduction;
H, the similarity ξ of face picture to be measured and target face picture is obtained according to step g, if ξ is more than 80%, then judge to treat
It surveys face picture and belongs to same people with such picture, otherwise, then judge that face picture to be measured is not belonging to same people with such picture.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711378336.1A CN108108760A (en) | 2017-12-19 | 2017-12-19 | A kind of fast human face recognition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711378336.1A CN108108760A (en) | 2017-12-19 | 2017-12-19 | A kind of fast human face recognition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108108760A true CN108108760A (en) | 2018-06-01 |
Family
ID=62210290
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711378336.1A Pending CN108108760A (en) | 2017-12-19 | 2017-12-19 | A kind of fast human face recognition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108108760A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109145842A (en) * | 2018-08-29 | 2019-01-04 | 深圳市智莱科技股份有限公司 | The method and device of chamber door based on image recognition control Intelligent storage cabinet |
CN109284591A (en) * | 2018-08-17 | 2019-01-29 | 北京小米移动软件有限公司 | Face unlocking method and device |
CN109684917A (en) * | 2018-11-14 | 2019-04-26 | 南宁学院 | A kind of fast human face recognition |
CN111091128A (en) * | 2019-12-18 | 2020-05-01 | 北京数衍科技有限公司 | Character and picture classification method and device and electronic equipment |
CN111382626A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Method, device and equipment for detecting illegal image in video and storage medium |
CN112613421A (en) * | 2020-12-26 | 2021-04-06 | 数源科技股份有限公司 | Dimension reduction feature analysis and comparison method for face picture |
CN113689604A (en) * | 2021-08-26 | 2021-11-23 | 重庆工程学院 | Access control system based on living body identification and detection method thereof |
CN113705506A (en) * | 2021-09-02 | 2021-11-26 | 中国联合网络通信集团有限公司 | Nucleic acid detection method, nucleic acid detection device, nucleic acid detection apparatus, and computer-readable storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101145261A (en) * | 2007-10-11 | 2008-03-19 | 中国科学院长春光学精密机械与物理研究所 | ATM system automatic recognition device |
CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
CN102819731A (en) * | 2012-07-23 | 2012-12-12 | 常州蓝城信息科技有限公司 | Face identification based on Gabor characteristics and Fisherface |
CN103049736A (en) * | 2011-10-17 | 2013-04-17 | 天津市亚安科技股份有限公司 | Face identification method based on maximum stable extremum area |
CN103218609A (en) * | 2013-04-25 | 2013-07-24 | 中国科学院自动化研究所 | Multi-pose face recognition method based on hidden least square regression and device thereof |
CN103839033A (en) * | 2012-11-20 | 2014-06-04 | 广东工业大学 | Face identification method based on fuzzy rule |
CN103902654A (en) * | 2014-02-28 | 2014-07-02 | 小米科技有限责任公司 | Clustering method and device and terminal device |
CN104091159A (en) * | 2014-07-14 | 2014-10-08 | 无锡市合鑫川自动化设备有限公司 | Human face identification method |
CN105868716A (en) * | 2016-03-29 | 2016-08-17 | 中国科学院上海高等研究院 | Method for human face recognition based on face geometrical features |
-
2017
- 2017-12-19 CN CN201711378336.1A patent/CN108108760A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101145261A (en) * | 2007-10-11 | 2008-03-19 | 中国科学院长春光学精密机械与物理研究所 | ATM system automatic recognition device |
CN101510257A (en) * | 2009-03-31 | 2009-08-19 | 华为技术有限公司 | Human face similarity degree matching method and device |
CN103049736A (en) * | 2011-10-17 | 2013-04-17 | 天津市亚安科技股份有限公司 | Face identification method based on maximum stable extremum area |
CN102819731A (en) * | 2012-07-23 | 2012-12-12 | 常州蓝城信息科技有限公司 | Face identification based on Gabor characteristics and Fisherface |
CN103839033A (en) * | 2012-11-20 | 2014-06-04 | 广东工业大学 | Face identification method based on fuzzy rule |
CN103218609A (en) * | 2013-04-25 | 2013-07-24 | 中国科学院自动化研究所 | Multi-pose face recognition method based on hidden least square regression and device thereof |
CN103902654A (en) * | 2014-02-28 | 2014-07-02 | 小米科技有限责任公司 | Clustering method and device and terminal device |
CN104091159A (en) * | 2014-07-14 | 2014-10-08 | 无锡市合鑫川自动化设备有限公司 | Human face identification method |
CN105868716A (en) * | 2016-03-29 | 2016-08-17 | 中国科学院上海高等研究院 | Method for human face recognition based on face geometrical features |
Non-Patent Citations (2)
Title |
---|
王碧泉等: "《模式识别 理论、方法和应用》", 30 November 1989 * |
赵薇 等: "用于大库人脸识别的脸型分类研究", 《电子技术》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109284591B (en) * | 2018-08-17 | 2022-02-08 | 北京小米移动软件有限公司 | Face unlocking method and device |
CN109284591A (en) * | 2018-08-17 | 2019-01-29 | 北京小米移动软件有限公司 | Face unlocking method and device |
CN109145842A (en) * | 2018-08-29 | 2019-01-04 | 深圳市智莱科技股份有限公司 | The method and device of chamber door based on image recognition control Intelligent storage cabinet |
CN109684917A (en) * | 2018-11-14 | 2019-04-26 | 南宁学院 | A kind of fast human face recognition |
CN111382626A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Method, device and equipment for detecting illegal image in video and storage medium |
CN111382626B (en) * | 2018-12-28 | 2023-04-18 | 广州市百果园信息技术有限公司 | Method, device and equipment for detecting illegal image in video and storage medium |
CN111091128A (en) * | 2019-12-18 | 2020-05-01 | 北京数衍科技有限公司 | Character and picture classification method and device and electronic equipment |
CN111091128B (en) * | 2019-12-18 | 2023-09-22 | 北京数衍科技有限公司 | Character picture classification method and device and electronic equipment |
CN112613421B (en) * | 2020-12-26 | 2023-04-14 | 数源科技股份有限公司 | Dimension reduction feature analysis and comparison method for face picture |
CN112613421A (en) * | 2020-12-26 | 2021-04-06 | 数源科技股份有限公司 | Dimension reduction feature analysis and comparison method for face picture |
CN113689604A (en) * | 2021-08-26 | 2021-11-23 | 重庆工程学院 | Access control system based on living body identification and detection method thereof |
CN113705506A (en) * | 2021-09-02 | 2021-11-26 | 中国联合网络通信集团有限公司 | Nucleic acid detection method, nucleic acid detection device, nucleic acid detection apparatus, and computer-readable storage medium |
CN113705506B (en) * | 2021-09-02 | 2024-02-13 | 中国联合网络通信集团有限公司 | Nucleic acid detection method, apparatus, device, and computer-readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108108760A (en) | A kind of fast human face recognition | |
CN108537743B (en) | Face image enhancement method based on generation countermeasure network | |
WO2017016240A1 (en) | Banknote serial number identification method | |
CN107463920A (en) | A kind of face identification method for eliminating partial occlusion thing and influenceing | |
CN105335732B (en) | Based on piecemeal and differentiate that Non-negative Matrix Factorization blocks face identification method | |
CN108090830B (en) | Credit risk rating method and device based on facial portrait | |
US11194997B1 (en) | Method and system for thermal infrared facial recognition | |
JP2014232533A (en) | System and method for ocr output verification | |
CN109255289B (en) | Cross-aging face recognition method based on unified generation model | |
CN109740572A (en) | A kind of human face in-vivo detection method based on partial color textural characteristics | |
CN107798308B (en) | Face recognition method based on short video training method | |
CN111709313B (en) | Pedestrian re-identification method based on local and channel combination characteristics | |
CN109325507A (en) | A kind of image classification algorithms and system of combination super-pixel significant characteristics and HOG feature | |
CN110826408B (en) | Face recognition method by regional feature extraction | |
CN106096517A (en) | A kind of face identification method based on low-rank matrix Yu eigenface | |
CN113449704B (en) | Face recognition model training method and device, electronic equipment and storage medium | |
Paul et al. | Extraction of facial feature points using cumulative histogram | |
CN109614866A (en) | Method for detecting human face based on cascade deep convolutional neural networks | |
CN113239839B (en) | Expression recognition method based on DCA face feature fusion | |
CN108256462A (en) | A kind of demographic method in market monitor video | |
CN111709305B (en) | Face age identification method based on local image block | |
CN111832405A (en) | Face recognition method based on HOG and depth residual error network | |
CN110188646B (en) | Human ear identification method based on fusion of gradient direction histogram and local binary pattern | |
CN111274883A (en) | Synthetic sketch face recognition method based on multi-scale HOG (histogram of oriented gradient) features and deep features | |
CN110555386A (en) | Face recognition identity authentication method based on dynamic Bayes |
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: 20180601 |
|
RJ01 | Rejection of invention patent application after publication |