CN104766063B - A kind of living body faces recognition methods - Google Patents
A kind of living body faces recognition methods Download PDFInfo
- Publication number
- CN104766063B CN104766063B CN201510161965.3A CN201510161965A CN104766063B CN 104766063 B CN104766063 B CN 104766063B CN 201510161965 A CN201510161965 A CN 201510161965A CN 104766063 B CN104766063 B CN 104766063B
- Authority
- CN
- China
- Prior art keywords
- image
- block
- grad
- vector
- living body
- 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
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a kind of living body faces recognition methods, it is divided into training stage and cognitive phase, in the training stage, it obtains several living body faces images and photo facial image, then the characteristic vector of the gray level image of every width living body faces image is extracted as positive sample, and the characteristic vector per the gray level image of photos facial image is extracted as negative sample, all positive samples and negative sample are input in SVM classifier again and are trained, obtain SVM classifier training pattern；A frame facial image is obtained in cognitive phase, is identified first with face recognition technology, In vivo detection technology extraction characteristic vector is recycled when recognition result is validated user, then characteristic vector is input in SVM classifier training pattern and carries out In vivo detection；Advantage is to judge whether face is validated user using face recognition technology, and it is that living body faces still palm off photo face to recycle In vivo detection technology determination face source during validated user, so as to effectively eliminate the potential safety hazard that photo face is brought.
Description
Technical field
The present invention relates to a kind of face recognition technology, more particularly, to a kind of living body faces recognition methods.
Background technology
Face recognition technology is a kind of biometrics identification technology, and it is with the advantage such as convenient, fast, accurate, in recent years
Obtain the development advanced by leaps and bounds.The input input of face identification system is usually a face for containing identity to be detected
The facial image of some known identities in image, and face database, and its output is then that a series of human face similarity degrees obtain
Point, with the identity of this face for showing identification.At present, face recognition technology is widely used to criminal investigation and case detection, department of banking
The fields such as system, customs inspection, the civil affairs department, work and rest work attendance.However, with the continuous extension of face recognition technology application,
Some safety problems also occur therewith, and criminal is using the human face photo deception face identification system forged, so as to legal
User causes heavy economic losses.Therefore, the source authenticity of facial image is judged to be particularly important, here it is live body
Detection.
After Google issues Android4.0, the work(that mobile phone is unlocked by recognition of face is brought for vast machine friend
Energy, but the subsequent report for replacing true man to unlock mobile phone with regard to useful personal photo always, thus Google is always in careful and guarantor
Keep and use face recognition technology.Face identification system is set to step into maturation, this kind of photo face replaces the peace of true living body faces
Full hidden danger must be resolved, it is therefore necessary to study a kind of technology for identifying living body faces.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of living body faces recognition methods, and whether it can determine that face
For validated user, it can determine that whether face source is living body faces, and it is hidden to effectively eliminate the safety that photo face is brought again
Suffer from.
Technical scheme is used by the present invention solves above-mentioned technical problem：A kind of living body faces recognition methods, its feature
It is to comprise the following steps：
1. obtaining M width includes the living body faces image that different face object and size are 256 × 256,
The photo facial image of every width living body faces image is obtained again, and the size per photos facial image is 256 × 256；So
M width living body faces images and M photos facial images are transformed into gray level image afterwards, 2M width gray level image is formed into an instruction
Practice image collection；Then the characteristic vector of every width gray level image in training image set is calculated；Again by every width living body faces figure
The characteristic vector of the gray level image of picture identifies as a positive sample, and with+1, by the gray level image of every photos facial image
Characteristic vector as a negative sample, and with -1 mark；All positive samples and all negative samples are finally input to svm classifier
It is trained in device, obtains SVM classifier training pattern；
2. need to carry out living body faces identification, the facial image that a frame includes face object to be identified is obtained, then
The minimum rectangular area where face object is intercepted in the facial image, then professional etiquette is entered to the size of rectangular area
It is whole, the human face region image to be identified that size is 256 × 256 is obtained, then turns human face region image to be identified
It is melted into gray level image；
3. the gray level image of human face region image is identified using face recognition technology, if recognition result is legal
User, then perform step 4.；If recognition result is disabled user, refuse face verification, face verification failure；
4. utilizing In vivo detection technology, the characteristic vector of the gray level image of human face region image is first calculated, then by face area
The characteristic vector of the gray level image of area image is input in SVM classifier training pattern, if SVM classifier training pattern exports
+ 1, then it represents that the source of human face region image is living body faces, and face verification is successful；If the output of SVM classifier training pattern-
1, then it represents that the source of human face region image is photo face, refuses face verification, face verification failure.
Described step 1. the acquisition process of the characteristic vector of every width gray level image in middle training image set with it is described
The step of the 4. characteristic vector of the gray level image of middle human face region image acquisition process it is identical, will be every in training image set
The gray level image of width gray level image and human face region image is used as a pending image, the characteristic vector of pending image
Acquisition process is：
A, pending image is divided intoThe size of individual non-overlapping copies is 64 × 64 image block；
B, i-th of image block currently pending in pending image is defined as current image block, wherein, 1≤i≤
16, i initial value is 1；
C, using size, point slides pixel-by-pixel in current image block for 3 × 3 sliding window, by current image block
It is divided into the sub-block that the individual equitant sizes of (64-2) × (64-2) are 3 × 3；
D, the Sobel operators of eight different directions are done into convolution operation with each sub-block in current image block respectively, obtained
Each sub-block into current image block eight different directions Grad, by j-th of sub-block in current image block in kth
The Grad in individual direction is designated asWherein, the Sobel operators of eight different directions be respectively 0 ° Sobel operators, 45 °
Sobel operators, 90 ° of Sobel operators, 135 ° of Sobel operators, 180 ° of Sobel operators, 225 ° of Sobel operators, 270 °
Sobel operators, 315 ° of Sobel operators, 1≤j≤(64-2) × (64-2), 1≤k≤8；
E, the order by all sub-blocks in current image block, by all sub-blocks in current image block in each direction
Grad is arranged to make up the Grad vector that dimension of the current image block in each direction is (64-2) × (64-2), will currently scheme
As the dimension of current image block that Grad of all sub-blocks in k-th of direction in block is arranged to make up in k-th of direction is
(64-2) × (64-2) Grad vector is designated as
Wherein, it is vector representation symbol in this symbol " [] ",Represent ladder of the 1st sub-block in current image block in k-th of direction
Angle value,Grad of the 2nd sub-block in current image block in k-th of direction is represented,Represent present image
- 1 sub-block of (64-2) × (64-2) in block k-th of direction Grad,Represent in current image block
The individual sub-blocks of (64-2) × (64-2) k-th of direction Grad；
F, i=i+1 is made, next pending image block in pending image as current image block, is then back to
Step c is continued executing with, until all image blocks in pending image are disposed, obtains each image in pending image
The Grad vector that block is (64-2) × (64-2) in the dimension of eight different directions, wherein, "=" in i=i+1 is assignment
Symbol；
G, the order by all image blocks in pending image, by all image blocks each comfortable eight in pending image
The Grad vector of individual different directions is arranged to make up the characteristic vector of pending image, is designated as T, T=[TV1 1,TV2 1,…,TV8 1,
TV1 2,TV2 2,…,TV8 2,…,TV1 16,TV2 16,…,TV8 16], wherein, it is vector representation symbol in this symbol " [] ", TV1 1Table
Show the Grad vector that dimension of the 1st image block in the 1st direction is (64-2) × (64-2), TV2 1Represent the 1st image block
The Grad vector that dimension in the 2nd direction is (64-2) × (64-2), TV8 1Represent the 1st image block in the 8th direction
The Grad vector that dimension is (64-2) × (64-2), TV1 2Represent that dimension of the 2nd image block in the 1st direction is (64-2)
× (64-2) Grad vector, TV2 2Represent the ladder that dimension of the 2nd image block in the 2nd direction is (64-2) × (64-2)
Angle value vector, TV8 2Represent the Grad vector that dimension of the 2nd image block in the 8th direction is (64-2) × (64-2), TV1 16
Represent the Grad vector that dimension of the 16th image block in the 1st direction is (64-2) × (64-2), TV2 16Represent the 16th
The Grad vector that dimension of the image block in the 2nd direction is (64-2) × (64-2), TV8 16Represent the 16th image block the 8th
The Grad vector that the dimension in individual direction is (64-2) × (64-2).
Compared with prior art, the advantage of the invention is that：
1) the inventive method adds In vivo detection technology, the inventive method on the basis of existing face recognition technology
Several living body faces images and multiple photos facial image are trained to obtain SVM classifier training pattern in the training stage,
In authentication procedures, judge whether face is validated user using face recognition technology, and it is first using In vivo detection technology
The characteristic vector of human face region image is calculated, then the characteristic vector of human face region image is input to SVM classifier training pattern
In, judgement face source is still personation photo face for living body faces, so as to effectively eliminate the peace that photo face is brought
Full hidden danger, realize the double shield of private information safety.
2) the inventive method is applicable to the platform of reduction process ability, such as android system.
3) the inventive method only needs a camera, can determine whether live body people by extracting a frame facial image
Face, greatly reduces the occupancy proportion of system resource, and need not add extra visual aids equipment, and without the active of user
Coordinate, extremely naturally.
4) the inventive method in the experimental verification stage to the image in Nanjing Aero-Space University's (NUAA) living body faces storehouse
It is identified, identifies the rate of accuracy reached of two class facial images to 98.718%.
Brief description of the drawings
Fig. 1 is that the totality of the inventive method realizes block diagram；
Fig. 2 a are the ladder of each comfortable eight different directions of all image blocks in the gray level image of a width living body faces image
Angle value vector distribution figure；
Fig. 2 b are that all image blocks in the gray level image of the photo facial image of living body faces image corresponding to Fig. 2 a are each
The Grad vector distribution figure of comfortable eight different directions；
Fig. 3 is the working characteristics ROC curve figure of the In vivo detection technology in the inventive method.
Embodiment
The present invention is described in further detail below in conjunction with accompanying drawing embodiment.
A kind of living body faces recognition methods proposed by the present invention, it is applied to android system, and its totality realizes block diagram such as
Shown in Fig. 1, it comprises the following steps：
1. it is 256 × 256 to obtain M width to include different face object and size by mobile phone camera
Living body faces image, then the photo facial image of every width living body faces image is obtained by mobile phone camera, per photos face
The size of image is 256 × 256；Then M width living body faces images and M photos facial images are transformed into gray-scale map
Picture, 2M width gray level image is formed into a training image set；Then every width gray level image in training image set is calculated
Characteristic vector；, will again using the characteristic vector of the gray level image of every width living body faces image as a positive sample, and with+1 mark
The characteristic vector of gray level image per photos facial image identifies as a negative sample, and with -1；Finally by all positive samples
This and all negative samples are input in SVM classifier and are trained, and obtain SVM classifier training pattern；Wherein, M >=50, instructing
The classification results for the SVM classifier training pattern that white silk stage usual positive sample and negative sample more at most obtain are more accurate, therefore have
M can suitably take larger value when gymnastics is made, and M=261 is can use such as in concrete operations.
Here, the face object in M width living body faces images is different, i.e., M different faces shoot
To different living body faces image；And photo facial image is the figure for once being shot to obtain again to the face on photo
Picture.
2. need to carry out living body faces identification, a frame is obtained by mobile phone camera and includes face object to be identified
Facial image, the minimum rectangular area where face object, then the chi to rectangular area are then intercepted in the facial image
Very little size progress is regular, the human face region image to be identified that size is 256 × 256 is obtained, then by people to be identified
Face area image changes into gray level image.
3. the gray level image of human face region image is identified using existing face recognition technology, if recognition result
For validated user, then step is performed 4.；If recognition result is disabled user, refuse face verification, face verification failure.
This, when recognition result is disabled user, the source of the human face region image may be living body faces, it is also possible to be photo people
People beyond face, i.e. user can not pass through face verification.
4. utilizing In vivo detection technology, the characteristic vector of the gray level image of human face region image is first calculated, then by face area
The characteristic vector of the gray level image of area image is input in SVM classifier training pattern, if SVM classifier training pattern exports
+ 1, then it represents that the source of human face region image is living body faces, and face verification is successful；If the output of SVM classifier training pattern-
1, then it represents that the source of human face region image is photo face, refuses face verification, face verification failure.
In this particular embodiment, the acquisition of the step 1. characteristic vector of every width gray level image in middle training image set
The acquisition process of process and the step 4. characteristic vector of the gray level image of middle human face region image is identical, by training image set
The gray level image of every width gray level image and human face region image be used as a pending image, due to living body faces and personation
The larger difference of photo face is that the mirror-reflection amount of the latter is far longer than the former, diffusing reflection caused by real human face surface
The light intensity of light is always directly proportional to the light intensity of face surface all directions incident light and the cosine value of incidence angle, and palms off photo face
The concavo-convex situation of face is not met, and smooth degree is high, and mirror-reflection amount is high, and photo face object would is that diffusing reflection and minute surface
The linear combination of reflecting component, mirror-reflection amount shares weight with diffusing reflection amount, therefore the inventive method draws pending image
It is divided into the image block of 16 non-overlapping copies, it is big that each image block then is divided into the individual equitant sizes of (64-2) × (64-2)
It is small be 3 × 3 sub-block, recycle the Sobel operators of each sub-block and eight different directions to do the result of convolution operation to obtain
The characteristic vector of pending image, i.e., the acquisition process of the characteristic vector of pending image are：
A, pending image is divided intoThe size of individual non-overlapping copies is 64 × 64 image block.
B, i-th of image block currently pending in pending image is defined as current image block, wherein, 1≤i≤
16, i initial value is 1.
C, using size, point slides pixel-by-pixel in current image block for 3 × 3 sliding window, by current image block
It is divided into the sub-block that the individual equitant sizes of (64-2) × (64-2) are 3 × 3.
D, the Sobel operators of eight different directions are done into convolution operation with each sub-block in current image block respectively, obtained
Each sub-block into current image block eight different directions Grad, by j-th of sub-block in current image block in kth
The Grad in individual direction is designated asFor the Sobel operators in k-th of direction and j-th of sub-block in current image block are done
Convolution operation obtains；Wherein, as listed in table 1, the Sobel operators of eight different directions be respectively 0 ° Sobel operators, 45 °
Sobel operators, 90 ° of Sobel operators, 135 ° of Sobel operators, 180 ° of Sobel operators, 225 ° of Sobel operators, 270 °
Sobel operators, 315 ° of Sobel operators, 1≤j≤(64-2) × (64-2), 1≤k≤8.
The Sobel operators of 1 eight different directions of table
E, the order by all sub-blocks in current image block, by all sub-blocks in current image block in each direction
Grad is arranged to make up the Grad vector that dimension of the current image block in each direction is (64-2) × (64-2), will currently scheme
As the dimension of current image block that Grad of all sub-blocks in k-th of direction in block is arranged to make up in k-th of direction is
(64-2) × (64-2) Grad vector is designated as
Wherein, it is vector representation symbol in this symbol " [] ",Represent the 1st sub-block in current image block in k-th direction
Grad,Grad of the 2nd sub-block in current image block in k-th of direction is represented,Represent current
- 1 sub-block of (64-2) × (64-2) in image block k-th of direction Grad,Represent present image
The Grad of individual sub-blocks of (64-2) × (64-2) in block in k-th of direction.
F, i=i+1 is made, next pending image block in pending image as current image block, is then back to
Step c is continued executing with, until all image blocks in pending image are disposed, obtains each image in pending image
The Grad vector that block is (64-2) × (64-2) in the dimension of eight different directions, wherein, "=" in i=i+1 is assignment
Symbol.
G, the order by all image blocks in pending image, by all image blocks each comfortable eight in pending image
The Grad vector of individual different directions is arranged to make up the characteristic vector of pending image, is designated as T, T=[TV1 1,TV2 1,…,TV8 1,
TV1 2,TV2 2,…,TV8 2,…,TV1 16,TV2 16,…,TV8 16], wherein, it is vector representation symbol in this symbol " [] ", TV1 1Table
Show the Grad vector that dimension of the 1st image block in the 1st direction is (64-2) × (64-2), TV2 1Represent the 1st image block
The Grad vector that dimension in the 2nd direction is (64-2) × (64-2), TV8 1Represent the 1st image block in the 8th direction
The Grad vector that dimension is (64-2) × (64-2), TV1 2Represent that dimension of the 2nd image block in the 1st direction is (64-2)
× (64-2) Grad vector, TV2 2Represent the ladder that dimension of the 2nd image block in the 2nd direction is (64-2) × (64-2)
Angle value vector, TV8 2Represent the Grad vector that dimension of the 2nd image block in the 8th direction is (64-2) × (64-2), TV1 16
Represent the Grad vector that dimension of the 16th image block in the 1st direction is (64-2) × (64-2), TV2 16Represent the 16th
The Grad vector that dimension of the image block in the 2nd direction is (64-2) × (64-2), TV8 16Represent the 16th image block the 8th
The Grad vector that the dimension in individual direction is (64-2) × (64-2).
To further illustrate the feasibility and validity of the inventive method, experimental verification is carried out to the inventive method.
NUAA In vivo detection face databases disclosed in the Sample Storehouse selected on Matlab experiment porch uses, in NUAA live bodies
522 sub-pictures, including the photo people of 261 width living body faces images and every width living body faces image are randomly selected in detection face database
Face image (i.e. photo facial image totally 261 width), the size of each image is 256 × 256.The type of each image is
RGB Three Channel Color images, convert it into gray level image, facilitate image procossing.Each image is divided intoIt is individual
The size of non-overlapping copies is 64 × 64 image block, and it is 3 × 3 sliding window in each image then to use size
Point is slided pixel-by-pixel in block, and each image block is divided into the son that the individual equitant sizes of (64-2) × (64-2) are 3 × 3
Block, then the Sobel operators of eight different directions are done into convolution operation with each sub-block in each image block respectively, obtain each
Each sub-block in image block and then obtains Grad of each image block in each direction in the Grad of eight different directions
Vector.Fig. 2 a give the ladder of each comfortable eight different directions of all image blocks in the gray level image of a width living body faces image
Angle value vector distribution, as can be seen that the percentage shared by Grad vector of all image blocks in each direction is each from Fig. 2 a
Differ.Fig. 2 b give all image blocks in the gray level image of the photo facial image of living body faces image corresponding to Fig. 2 a
The Grad vector distribution of each comfortable eight different directions, ladder of all image blocks in each direction is can be seen that from Fig. 2 b
Percentage shared by angle value vector is different.Comparison diagram 2a and Fig. 2 b, it can be seen that the level difference of the distribution shown in Fig. 2 b
Property is larger.
Each comfortable eight differences of all image blocks in the gray level image of all living body faces images obtained above
The Grad vector in direction, obtains the characteristic vector of the gray level image of all living body faces images, and randomly selectM feature
Vector is used as positive sample；And according to the institute in the gray level image of the photo facial image of all living body faces images obtained above
There is the Grad vector of each comfortable eight different directions of image block, obtain the ash of the photo facial image of all living body faces images
The characteristic vector of image is spent, and using characteristic vector corresponding to positive sample as negative sample；By all positive samples and all negative samples
It is input in SVM classifier and is trained, obtains SVM classifier training pattern.Will be remainingThe ash of M living body faces image
Spend the characteristic vector of image and remainingThe characteristic vector of the gray level image of the photo facial image of M living body faces image point
Characteristic vector not as test is input in SVM classifier training pattern, and Fig. 3 gives the In vivo detection in the inventive method
The working characteristics ROC curve of technology, abscissa represents false positive probability (False Positive Rate) in Fig. 3, and ordinate represents
Real probability (True Positive Rate), it was found from the area below ROC curve provided from Fig. 3, in the inventive method
The verification and measurement ratio of In vivo detection technology reach 98.718%, detection used time 167 seconds altogether, average every face processing it is time-consuming less than
One second, this, which fully indicates the inventive method, had relatively low complexity, and preferable reason is provided for the application of Android platform
By basis, this causes the In vivo detection technology in the inventive method to greatly improve the feasibility for improving face recognition technology, more
The sound assurance safety in utilization of face identification system.
Claims (1)
1. a kind of living body faces recognition methods, it is characterised in that comprise the following steps：
1. obtaining, M width includes different face object and size is 256 × 256 living body faces image, then obtains
The photo facial image of every width living body faces image is taken, the size per photos facial image is 256 × 256；Then by M
Width living body faces image and M photos facial images are transformed into gray level image, and 2M width gray level image is formed into a training figure
Image set closes；Then the characteristic vector of every width gray level image in training image set is calculated；Again by every width living body faces image
The characteristic vector of gray level image identifies as a positive sample, and with+1, by the spy of the gray level image of every photos facial image
Sign vector is used as a negative sample, and with -1 mark；Finally all positive samples and all negative samples are input in SVM classifier
It is trained, obtains SVM classifier training pattern；
2. need to carry out living body faces identification, the facial image that a frame includes face object to be identified is obtained, then at this
Minimum rectangular area in facial image where interception face object, then the size progress to rectangular area are regular, obtain
To the human face region image to be identified that size is 256 × 256, human face region image to be identified is then changed into ash
Spend image；
3. the gray level image of human face region image is identified using face recognition technology, if recognition result is legal use
Family, then perform step 4.；If recognition result is disabled user, refuse face verification, face verification failure；
4. utilizing In vivo detection technology, the characteristic vector of the gray level image of human face region image is first calculated, then by human face region figure
The characteristic vector of the gray level image of picture is input in SVM classifier training pattern, if SVM classifier training pattern output+1,
The source for then representing human face region image is living body faces, and face verification is successful；If SVM classifier training pattern output -1,
The source for then representing human face region image is photo face, refuses face verification, face verification failure；
Described the step 1. acquisition process of the characteristic vector of every width gray level image in middle training image set and described step
The acquisition process of the characteristic vector of the gray level image of rapid 4. middle human face region image is identical, by every width ash in training image set
The gray level image of degree image and human face region image is used as a pending image, the acquisition of the characteristic vector of pending image
Process is：
A, pending image is divided intoThe size of individual non-overlapping copies is 64 × 64 image block；
B, i-th of image block currently pending in pending image is defined as current image block, wherein, 1≤i≤16, i's
Initial value is 1；
C, using size, point slides pixel-by-pixel in current image block for 3 × 3 sliding window, and current image block is split
Into the sub-block that the individual equitant sizes of (64-2) × (64-2) are 3 × 3；
D, the Sobel operators of eight different directions are done into convolution operation with each sub-block in current image block respectively, worked as
Each sub-block in preceding image block eight different directions Grad, by j-th of sub-block in current image block in k-th of side
To Grad be designated asWherein, the Sobel operators of eight different directions be respectively 0 ° Sobel operators, 45 ° of Sobel
Operator, 90 ° of Sobel operators, 135 ° of Sobel operators, 180 ° of Sobel operators, 225 ° of Sobel operators, 270 °
Sobel operators, 315 ° of Sobel operators, 1≤j≤(64-2) × (64-2), 1≤k≤8；
E, the order by all sub-blocks in current image block, by gradient of all sub-blocks in current image block in each direction
Value is arranged to make up the Grad vector that dimension of the current image block in each direction is (64-2) × (64-2), by current image block
In dimension of the current image block that is arranged to make up of Grad of all sub-blocks in k-th of direction in k-th of direction be (64-2)
× (64-2) Grad vector is designated as Wherein, exist
This symbol " [] " is vector representation symbol,Grad of the 1st sub-block in current image block in k-th of direction is represented,Grad of the 2nd sub-block in current image block in k-th of direction is represented,Represent in current image block
- 1 sub-block of (64-2) × (64-2) k-th of direction Grad,Represent the in current image block
Grad of the individual sub-blocks of (64-2) × (64-2) in k-th of direction；
F, i=i+1 is made, next pending image block in pending image as current image block, is then back to step c
Continue executing with, until all image blocks in pending image are disposed, each image block obtained in pending image exists
The Grad vector that the dimension of eight different directions is (64-2) × (64-2), wherein, "=" in i=i+1 is assignment；
G, the order by all image blocks in pending image, by each comfortable eight of all image blocks in pending image not
Equidirectional Grad vector is arranged to make up the characteristic vector of pending image, is designated as T,Wherein, it is vector in this symbol " [] "
Represent symbol, TV1 1The Grad vector that dimension of the 1st image block in the 1st direction is (64-2) × (64-2) is represented,
Represent the Grad vector that dimension of the 1st image block in the 2nd direction is (64-2) × (64-2), TV8 1Represent the 1st image
The Grad vector that dimension of the block in the 8th direction is (64-2) × (64-2), TV1 2Represent the 2nd image block in the 1st direction
Dimension be (64-2) × (64-2) Grad vector,Represent that dimension of the 2nd image block in the 2nd direction is (64-
2) × (64-2) Grad vector, TV8 2Represent that dimension of the 2nd image block in the 8th direction is (64-2) × (64-2's)
Grad vector, TV1 16The Grad vector that dimension of the 16th image block in the 1st direction is (64-2) × (64-2) is represented,Represent the Grad vector that dimension of the 16th image block in the 2nd direction is (64-2) × (64-2), TV8 16Represent the
The Grad vector that dimension of 16 image blocks in the 8th direction is (64-2) × (64-2).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510161965.3A CN104766063B (en) | 2015-04-08 | 2015-04-08 | A kind of living body faces recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510161965.3A CN104766063B (en) | 2015-04-08 | 2015-04-08 | A kind of living body faces recognition methods |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104766063A CN104766063A (en) | 2015-07-08 |
| CN104766063B true CN104766063B (en) | 2018-01-05 |
Family
ID=53647877
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510161965.3A Active CN104766063B (en) | 2015-04-08 | 2015-04-08 | A kind of living body faces recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104766063B (en) |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3369036A4 (en) * | 2015-10-30 | 2019-05-22 | Microsoft Technology Licensing, LLC | Spoofed face detection |
| CN105389553A (en) * | 2015-11-06 | 2016-03-09 | 北京汉王智远科技有限公司 | Living body detection method and apparatus |
| CN105389554B (en) * | 2015-11-06 | 2019-05-17 | 北京汉王智远科技有限公司 | Living body determination method and equipment based on recognition of face |
| CN108875473A (en) * | 2017-06-29 | 2018-11-23 | 北京旷视科技有限公司 | Living body verification method, device and system and storage medium |
| CN108573203B (en) * | 2017-03-17 | 2021-01-26 | 北京旷视科技有限公司 | Identity authentication method and device and storage medium |
| CN107122709B (en) * | 2017-03-17 | 2020-12-04 | 上海云从企业发展有限公司 | Living body detection method and device |
| CN107122744B (en) * | 2017-04-28 | 2020-11-10 | 武汉神目信息技术有限公司 | Living body detection system and method based on face recognition |
| CN107392135A (en) * | 2017-07-14 | 2017-11-24 | 广东欧珀移动通信有限公司 | Biopsy method and Related product |
| CN107368817B (en) * | 2017-07-26 | 2020-02-21 | 湖南云迪生物识别科技有限公司 | Face recognition method and device |
| CN107609494A (en) * | 2017-08-31 | 2018-01-19 | 北京飞搜科技有限公司 | A kind of human face in-vivo detection method and system based on silent formula |
| CN107679457A (en) * | 2017-09-06 | 2018-02-09 | 阿里巴巴集团控股有限公司 | User identity method of calibration and device |
| CN107680185A (en) * | 2017-09-22 | 2018-02-09 | 芜湖星途机器人科技有限公司 | The method for using robot register in meeting-place |
| CN107818313B (en) * | 2017-11-20 | 2019-05-14 | 腾讯科技（深圳）有限公司 | Vivo identification method, device and storage medium |
| CN107958236B (en) * | 2017-12-28 | 2021-03-19 | 深圳市金立通信设备有限公司 | Face recognition sample image generation method and terminal |
| CN108389053B (en) * | 2018-03-19 | 2021-10-29 | 广州逗号智能零售有限公司 | Payment method, payment device, electronic equipment and readable storage medium |
| CN108494778A (en) * | 2018-03-27 | 2018-09-04 | 百度在线网络技术（北京）有限公司 | Identity identifying method and device |
| CN111222380B (en) * | 2018-11-27 | 2023-11-03 | 杭州海康威视数字技术股份有限公司 | Living body detection method and device and recognition model training method thereof |
| CN109409344A (en) * | 2018-12-23 | 2019-03-01 | 广东腾晟信息科技有限公司 | Human face data accurately compares and judgment method |
| CN111444744A (en) * | 2018-12-29 | 2020-07-24 | 北京市商汤科技开发有限公司 | Living body detection method, living body detection device, and storage medium |
| CN111488764B (en) * | 2019-01-26 | 2024-04-30 | 天津大学青岛海洋技术研究院 | Face recognition method for ToF image sensor |
| WO2020159437A1 (en) * | 2019-01-29 | 2020-08-06 | Agency For Science, Technology And Research | Method and system for face liveness detection |
| 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 |
| CN110321872B (en) | 2019-07-11 | 2021-03-16 | 京东方科技集团股份有限公司 | Facial expression recognition method and device, computer equipment and readable storage medium |
| CN113239761B (en) * | 2021-04-29 | 2023-11-14 | 广州杰赛科技股份有限公司 | Face recognition method, device and storage medium |
| CN113221767B (en) * | 2021-05-18 | 2023-08-04 | 北京百度网讯科技有限公司 | Method for training living body face recognition model and recognizing living body face and related device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020599A (en) * | 2012-12-12 | 2013-04-03 | 山东神思电子技术股份有限公司 | Identity authentication method based on face |
| CN103116763A (en) * | 2013-01-30 | 2013-05-22 | 宁波大学 | Vivo-face detection method based on HSV (hue, saturation, value) color space statistical characteristics |
| CN103440479A (en) * | 2013-08-29 | 2013-12-11 | 湖北微模式科技发展有限公司 | Method and system for detecting living body human face |
| CN103605958A (en) * | 2013-11-12 | 2014-02-26 | 北京工业大学 | Living body human face detection method based on gray scale symbiosis matrixes and wavelet analysis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8351662B2 (en) * | 2010-09-16 | 2013-01-08 | Seiko Epson Corporation | System and method for face verification using video sequence |
-
2015
- 2015-04-08 CN CN201510161965.3A patent/CN104766063B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103020599A (en) * | 2012-12-12 | 2013-04-03 | 山东神思电子技术股份有限公司 | Identity authentication method based on face |
| CN103116763A (en) * | 2013-01-30 | 2013-05-22 | 宁波大学 | Vivo-face detection method based on HSV (hue, saturation, value) color space statistical characteristics |
| CN103440479A (en) * | 2013-08-29 | 2013-12-11 | 湖北微模式科技发展有限公司 | Method and system for detecting living body human face |
| CN103605958A (en) * | 2013-11-12 | 2014-02-26 | 北京工业大学 | Living body human face detection method based on gray scale symbiosis matrixes and wavelet analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104766063A (en) | 2015-07-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104766063B (en) | A kind of living body faces recognition methods | |
| CN111401372B (en) | Method for extracting and identifying image-text information of scanned document | |
| Zhang et al. | Face morphing detection using Fourier spectrum of sensor pattern noise | |
| CN108537743A (en) | A kind of face-image Enhancement Method based on generation confrontation network | |
| CN109858375B (en) | Living body face detection method, terminal and computer readable storage medium | |
| RU2431190C2 (en) | Facial prominence recognition method and device | |
| KR20170006355A (en) | Method of motion vector and feature vector based fake face detection and apparatus for the same | |
| KR101443139B1 (en) | Single image-based fake face detection | |
| CN111339897B (en) | Living body identification method, living body identification device, computer device, and storage medium | |
| CN111222433A (en) | Automatic face auditing method, system, equipment and readable storage medium | |
| CN107742094A (en) | Improve the image processing method of testimony of a witness comparison result | |
| CN105488486A (en) | Face recognition method and device for preventing photo attack | |
| CN116229528A (en) | Living body palm vein detection method, device, equipment and storage medium | |
| Barni et al. | Iris deidentification with high visual realism for privacy protection on websites and social networks | |
| Mohan et al. | An intelligent recognition system for identification of wood species | |
| CN107369086A (en) | A kind of identity card stamp system and method | |
| CN104408736A (en) | Characteristic-similarity-based synthetic face image quality evaluation method | |
| CN104615985B (en) | A kind of recognition methods of human face similarity degree | |
| CN113901423B (en) | Intelligent security equipment control method and system based on face recognition | |
| Hassan et al. | Enhance iris segmentation method for person recognition based on image processing techniques | |
| CN108280468A (en) | A kind of image-recognizing method based on grid | |
| CN114266713A (en) | NonshadowGAN-based unmanned aerial vehicle railway fastener image shadow removing method and system | |
| Das et al. | Person identification through IRIS recognition | |
| Singla et al. | Challenges at different stages of an iris based biometric system. | |
| Rana et al. | Face detection system using FPGA |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |