CN105447432B - A kind of face method for anti-counterfeit based on local motion mode - Google Patents
A kind of face method for anti-counterfeit based on local motion mode Download PDFInfo
- Publication number
- CN105447432B CN105447432B CN201410428040.6A CN201410428040A CN105447432B CN 105447432 B CN105447432 B CN 105447432B CN 201410428040 A CN201410428040 A CN 201410428040A CN 105447432 B CN105447432 B CN 105447432B
- Authority
- CN
- China
- Prior art keywords
- face
- motion
- region
- key point
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of face method for anti-counterfeit based on local motion mode, comprising the following steps: 1) detect the facial image region of camera acquisition, and face key point is positioned;2) motion information of the statistics of the regional area locating for face key point face and non-face region;3) according to the local motion information at all key points obtained, the local motion mode of face is calculated;4) based on the local motion mode of face, judged using the true and false of the preconfigured pattern classifier to the face.The invention has the benefit that can effectively be combined with actual face identification system, real human face is fast and effeciently identified in the case where not needing user's interaction substantially and forge face.
Description
Technical field
The present invention relates to computer vision and area of pattern recognition, the face method for anti-counterfeit in living things feature recognition field is ground
Study carefully more particularly to a kind of face method for anti-counterfeit based on local motion mode.
Background technique
Currently, biometrics identification technology has been widely used in the every aspect in daily life.Face biology
Feature identification technique, since it is with easy-to-use, user friendly, it is contactless the advantages that, achieve prominent fly in recent years
The development pushed ahead vigorously, these development have been embodied in each research field, including Face datection, face characteristic extract, classifier design
And hardware device manufacture etc..However, some tests are still faced on application based on the living things feature recognition of face,
Wherein, it is the most outstanding be exactly identifying system safety issue;As a kind of device for identification, they are easy to
By an illegal person personation at legal user, and true face all cannot be distinguished in current most of face identification system
And photo, as long as having got the photo of legitimate user, then this kind of identifying system that can easily out-trick, and now hair
The social networks reached becomes abnormal easy by this attack pattern;In addition, may with the mask of the video or forgery recorded
Attack is generated to face identification system.
The anti-fake also known as face In vivo detection of face, gradually receives the attention from academia and industry;Face is anti-fake
Main purpose be to discriminate between the facial image of real human face and above-mentioned forgery, identification dummy's face image attacks face identification system
It hits, to improve the safety of face identification system;It is different according to the clue used, face method for anti-counterfeit can be divided into three
Class:
1, based on the face method for anti-counterfeit of skin reflex characteristic: from the reflection characteristic of face skin, some researchers
It is anti-fake that face is carried out using multispectral acquisition means;Using both true man's skin and the face skin of forgery under different spectrum
This feature of reflectivity difference, achievees the purpose that face is anti-fake;The research contents of such methods be find suitable spectrum so that
The difference of true and false face skin is maximum;However, such methods are with following clearly disadvantageous: 1) only in very small amount of number
According to upper test, therefore performance can not be fully assessed;2) spectral band chosen can not be incuded by common camera,
It needs to dispose special sensor devices, increases hardware spending;3) additional sensor devices need to develop targeted signal
Conversion circuit increases the compatibility issue with existing system.
2, the face method for anti-counterfeit based on texture difference: the face method for anti-counterfeit based on microtexture, which has, to be assumed: same
Equipment acquisition is forged face and is compared with the true man's face acquired with the equipment there are loss in detail or difference, and the difference in these details
The different difference just caused in image microtexture;The hypothesis is in most cases to set up, and the face of forgery is by making
It is formed with real human face picture making, by taking the photo of printing as an example, photo is printed upon on paper by illegal user first, then will
The human face photo of printing is attacked before being placed in face identification system;In this process, it can at least be caused there are two link
Difference, first is that printing link, printer can not reappear without distortion photo content;Second is that the secondary imaging of photograph print, is adopted
Collection equipment can not capture the content perfection on photo;In addition to this, real human face and printing face are in surface shape
Difference, the difference etc. of local bloom can all cause difference of the two in microtexture.
3, based drive face method for anti-counterfeit: such methods are intended to the physiological reaction by detecting face to determine to acquire
Whether be real human face;In view of real human face is compared with face is forged, there are more independences, such methods pass through requirement
User carries out specified movement as the foundation determined;Common exchange method includes blink, is shaken the head, mouth action etc.;It removes
It is that the movement based on entire head is judged there are also a kind of method except detection method based on local motion;This kind of side
The effective reason of method is the three-dimensional structure of photo and face, and there are apparent differences, so that the head movement mode obtained is also deposited
In certain difference;It is a kind of to be suggested based on multi-modal face method for anti-counterfeit in order to further increase face anti-counterfeiting performance;It should
Whether the content of text that method requires user's reading specified, the subsequent lip motion for passing through analysis user and corresponding voice content
It coincide to judge the true and false of face;However, it is this based on the method for anti-counterfeit of human-computer interaction due to requiring user specifically to be moved
To make, the requirement to user is excessively high, so that user experience is bad, meanwhile, it is also a big drawback of the above method that authenticated time is longer.
In three of the above method, based drive face method for anti-counterfeit has the condition that is not illuminated by the light, and picture quality influences
The advantages that, however, such methods are not accurately positioned each region of face when extracting motion feature, thus nothing
The actual motion state of the acquired face of method accurate description;For example, the image of acquisition is broadly divided into face square by certain methods
Shape region and background area determine the true and false of face by the motion state of both comparisons, however, the people determined by rectangle frame
Face region includes a large amount of background area, so that real human face is very big may to be mistaken for forging face;Meanwhile in this feelings
Under condition, by folding, torsional deformation can also out-trick face anti-counterfeiting system easily for the face of forgery;Therefore, how to be precisely located
Human face region and non-face region, and find the regional area of the most distinctive local motion mode strong with Extraction and discrimination
Information is the key that can face anti-counterfeiting system be applied in practice.
Summary of the invention
The object of the present invention is to provide a kind of face method for anti-counterfeit based on local motion mode, to overcome current existing skill
Art above shortcomings.
The purpose of the present invention is be achieved through the following technical solutions:
A kind of face method for anti-counterfeit based on local motion mode, comprising:
Video image gathered in advance is analyzed, determines human face region, and analyze the human face region, really
Each face key point in the fixed human face region;
According to video frame corresponding to the video image, the direction of motion and width of pixel in the video image are obtained
Value information;
According to the direction of motion and amplitude information of the pixel of acquisition, the face key point is analyzed, determines institute
The direction of motion and amplitude information in the regional area of face key point place are stated, and determines the movement of regional area according to the information
Relationship between direction and between amplitude, to obtain the local motion mode of face;
Classified by local motion mode of the preconfigured pattern classifier to the face of acquisition, and according to classification
As a result, verifying the true and false of face in the video image.
Further, the human face region is obtained or is utilized by human-face detector and is manually specified.
Further, the human face region is analyzed, determines that each face key point includes: in the human face region
Institute is determined by the initial position message of face key point predetermined according to the position of the human face region
State the position of each face key point in human face region;
According to the position of face key point each in the human face region, extract crucial with the face on the video image
The corresponding video image characteristic in position of point;
According to the video image characteristic, by preconfigured algorithm model, update on the video image with it is described
The position of the corresponding face key point of human face region;
After meeting preset condition, the above process is terminated.
Further, according to the direction of motion and amplitude information of the pixel of acquisition, the face key point is divided
Analysis, the direction of motion and amplitude information where determining the face key point in regional area include:
According to the position of the accurate face key point, head zone in the video image is accurately divided, really
The respective image mask of head zone in the fixed video image;
According to the direction of motion and amplitude information of described image mask and the pixel of acquisition, each accurate face is extracted
The direction of motion and amplitude information of head zone and non-head region of the key point in the regional area locating for it.
Further, according to the position of the face key point, head zone in the video image is accurately drawn
Point, determine that corresponding image mask includes:
According to the position of the accurate face key point, face corresponding with the position of the accurate face key point is determined
Envelope;And the region for using the face envelope including is as the human face region of the video image;
According to the connecting line at face envelope both ends in the video image, mirror image is carried out to the face envelope, and
The face envelope is combined with its mirror image, obtains a closed curve, the region for including using the curve as
The head zone of the video image;
According to the position of the position of the human face region of the video image and the head zone of the video image, institute is determined
State the human face region of video image and the respective image mask of head zone.
Under the premise of can obtain face and head precise boundary, the quantity of required key point and corresponding position can
Arbitrarily to select.
Further, it according to the direction of motion and amplitude information of described image mask and the pixel of acquisition, extracts every
The direction of motion and amplitude of head zone and non-head region of a accurate face key point in the regional area locating for it are believed
Breath includes:
According to the parameter information of preconfigured regional area size, office corresponding to each accurate face key point is determined
Portion region;
According to described image mask, the pixel that head zone is fallen in the regional area is demarcated as foreground area,
The pixel fallen in except head zone in the regional area is demarcated as background area;
According to the direction of motion and amplitude information of the pixel of acquisition;Count the foreground and background area in the regional area
The respective direction of motion in domain and amplitude information.
Further, it according to the direction of motion and amplitude information of regional area where the face key point, calculates different
Relationship between the direction of motion and amplitude in region, the local motion mode for obtaining face include:
According to the direction of motion and width of foreground and background of each face key point in the regional area locating for it
Value information calculates between local foreground region, between local background region and the movement between local foreground and background area
The relationship in direction and amplitude information;
According between the local foreground region of calculating, between local background region and local foreground and background area
Between the direction of motion and amplitude information relationship, determine the local motion mode of face.
Further, the movement of the foreground and background in the regional area according to each face key point locating for it
Direction and amplitude information, calculate between local foreground region, between local background region and local foreground and background area it
Between the direction of motion and the relationship of amplitude information include:
Based on the direction of motion and amplitude information in the foreground and background region in the regional area, the direction of motion is quantified
At several sections, the motion information histogram for adding up the motion amplitude of pixel in each regional area is obtained;
According to the motion information histogram, determine between the motion information histogram of regional area described in any two
Ratio between related coefficient and motion amplitude.
Further, according between the local foreground region of calculating, between local background region and local foreground
The relationship of the direction of motion and amplitude information between background area determines that the local motion mode of face includes:
According to the ratio between the related coefficient and the motion amplitude, by the related coefficient between all regional areas
It is combined with motion amplitude ratio, determination obtains the local motion mode of face.
The invention has the benefit that based on face method for anti-counterfeit provided by the invention by carrying out accurate face and head
Portion's zone location, the strong face local motion mode of Extraction and discrimination ability, can fast and effeciently distinguish the true and false of facial image;
Effectively the defect of face and head movement information can not accurately be extracted by compensating for existing method, while use one kind can be more
Increase the local motion mode information of effect earth's surface intelligent face motion state;This method is not set substantially by image capture environment and acquisition
The influence of standby quality, meanwhile, also do not influenced substantially by the degree true to nature and forgery face deformation extent of forging photo, it can
Real human face and forgery face before effectively distinguishing camera.
Detailed description of the invention
It, below will be to embodiment following for the embodiment of the present invention or technical solution in the prior art is illustrated more clearly that
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only some of the application
Embodiment for those of ordinary skill in the art without creative efforts, can also be attached according to these
Figure obtains other attached drawings.
Fig. 1 is a kind of flow chart of the face method for anti-counterfeit based on local motion mode provided in an embodiment of the present invention;
Fig. 2 is that a kind of use of face method for anti-counterfeit based on local motion mode provided in an embodiment of the present invention cascades increasing
Strong regression model carries out the flow chart of face key point location;
Fig. 3 is a kind of face local motion for face method for anti-counterfeit based on local motion mode that inventive embodiments provide
The flow chart of schema extraction method.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art's every other embodiment obtained belong to what the present invention protected
Range.
A kind of face method for anti-counterfeit based on local motion mode described in the embodiment of the present invention, such as the flow chart institute of Fig. 1
Show, comprising the following steps:
Step 1: video image gathered in advance being analyzed, determines human face region, and carry out to the human face region
Analysis, determines each face key point in the human face region;The human face region is obtained by human-face detector or using manually
It is specified.
Face key point (Face Landmark) mainly includes cheek, eyes, and eyebrow is multiple including nose and mouth
Position with certain semanteme;After user's face detection algorithm obtains the position of human face region in the picture, it can make
Crucial point location is carried out with different types of method;Currently, face key independent positioning method can be divided into multiclass, more commonly
Including active shape model (Active Shape Model, ASM), active phenomenological model (Active Appearance Mode,
AAM), the partial model (Constrained Local Model, CLM) of constraint, cascade enhancing shape regression model
(Cascaded Boosted Shape Regression Model) etc.;As shown in Fig. 2, the application chooses face cheek key
Point is used as face key point, and illustrates the basic procedure of face key point location for cascading and enhance regression model:
Step 1-1: the position based on human face region initializes the position of face key point;To the view of the human face region
Frequency image is analyzed, and the position of human face region in video image is determined, and according to the position of the human face region, by preparatory
The location information of the face key point of definition determines the position of each face key point on the human face region;Generally, using just
The shape of dough figurine face is initialized;
Step 1-2: the position of face key point each on the human face region is analyzed, is extracted on the video image
Video image characteristic corresponding with the position of the face key point;
Step 1-3: the video is determined by preconfigured image regression model according to the video image characteristic
The position of accurate face key point corresponding with the human face region on image;
Step 1-4: skipping to step 1-2, carries out the recurrence of next round, until meeting certain termination condition.
In step 1, it is necessary first to Face datection be carried out to the image currently acquired, if not detecting face figure
Picture then acquires next frame image;If detecting multiple facial images, chooses the maximum face of detection block area and carry out face
Anti-fake analysis.
Based on above face key independent positioning method, the location information QUOTE of K face key point can be obtained , wherein the position of k-th of key point is expressed as QUOTE ;Specifically, the present embodiment successively sequentially chooses 17 key points at cheek.
Step 2: according to the video frame corresponding to video image that acquired of step 1, extracting pixel in present image
The direction of motion and amplitude information.
The motion information of image refers to the image for the former frame or several frames that pixel is acquired relative to camera in image
The change in location of generation, is indicated with the direction of motion and motion amplitude;It is obtained currently, being based primarily upon light stream (Optical Flow)
Take the motion information of pixel in image;The concept of light stream is proposed by Gibson et al. in nineteen fifty earliest;It can describe by
The movement of foreground target itself, the movement of camera in scene, or both associated movement caused by a variety of different movement moulds
Formula;Currently, there are many ways to optical flow computation, such as Lucas-Kanade algorithm, Horn-Schunck algorithm and Gunnar
The method etc. based on Polynomial Expansion that Farneback is proposed;Wherein the first algorithm is for extracting sparse optical flow, latter two
Algorithm is used for computation-intensive light stream.
The application uses Gunnar Farneback algorithm;After given current frame image and previous frame image, the algorithm
The motor pattern of each pixel in current frame image can be calculated;For ith pixel point,
Its motor pattern is expressed as QUOTE , wherein QUOTE It indicates in image coordinate
The movement in X direction in system
Amplitude, QUOTE Indicate the motion amplitude on y direction.
Step 3: the direction of motion and amplitude information of the pixel calculated by step 2 --- based on light stream
It calculates and obtains, the face key point is analyzed, determines the direction of motion and width of the face key point in regional area
Value information, and according to the direction of motion and amplitude information of the face key point, it determines between the direction of motion and amplitude information
Relationship, obtain the local motion mode of face;Extract the fortune of regional area locating for 17 face key points in step 1
Dynamic information;As shown in figure 3, specific step is as follows for motor pattern extraction in order to realize more accurate face antiforge function:
Step 3-1: according to the position of the accurate face key point, to human face region and header area in the video image
Domain is accurately divided, and determines that human face region and the respective image of head zone are covered in the video image, obtain image mask
Specific steps are as follows:
Step 3-1-1: according to the position of the accurate face key point, the determining position with the accurate face key point
Corresponding face envelope;And the region for using the face envelope including is as the human face region of the video image;
Step 3-1-2: the face envelope that step 3-1-1 is obtained, it is right according to the connecting line of face envelope two-end-point
The face envelope carries out mirror image, and the face envelope is combined with its mirror image, obtains a closed curve,
The region for including using the curve is as the head zone of the video image;
Step 3-1-3: according to the head zone of the position of the human face region of the video image and the video image
Position determines the human face region of the video image and the respective image mask of head zone.
Step 3-2: the direction of motion and the amplitude letter of the pixel based on step 3-1 image mask obtained and acquisition
Breath, extract each accurate foreground area and background area of the face key point in the regional area locating for it the direction of motion and
Amplitude information, by taking key point k as an example, the specific steps are as follows:
Step 3-2-1: according to the parameter information of preconfigured regional area size, each accurate face key point is determined
Corresponding regional area;If the width of human face region is W, a height of H, centered on face key point k, local rectangular portions are selected
QUOTE Width be 0.2 × W, a height of 0.2 × H;
Step 3-2-2: the light stream direction of all pixels point and amplitude in rectangular area are calculated, QUOTE is expressed as ;
Step 3-2-3: in rectangular area QUOTE In the regional area for including, determination falls in face and header area
The inside and outside pixel in domain, is defined as foreground and background region, respectively indicates collection and is combined into QUOTE And QUOTE ;
Step 3-2-4: QUOTE is counted respectively And QUOTE Motion information;Firstly, by light stream
Direction (0 ° to 360 °) uniform quantization is to 18 sections;Then, add up to fall in pixel in each section light stream amplitude it
With;
It obtains two histogram tables that dimension is 18 and is shown as QUOTE And QUOTE 。
Using the above method, the motion information at 17 key points of face can be obtained --- the direction of motion and amplitude letter
Breath;The motion information will be used to extract local motion mode.
Step 3-3: the direction of motion and the amplitude letter based on prospect and background at extracted key point in step 3-2
Breath calculates between local foreground region, between local background region and the direction of motion between local foreground and background area
The specific of the local motion mode of face is obtained to obtain the local motion mode of current face with the relationship of amplitude information
Steps are as follows:
Step 3-3-1: any two are calculated according to the histogram of local foreground or background area extraction where key point
Between related coefficient;
Step 3-3-2: the motion amplitude that any two are extracted according to local foreground where key point or background area is calculated
Between ratio;
Step 3-3-3: the step 3-3-1 all related coefficients being calculated and step 3-3-2 are calculated all
Motion amplitude ratio is combined, the local motion mode as current face.
By step 3-2, the histogram for indicating totally 34 18 dimensions of the local motion information at key point has been obtained;
Then, the present invention is transported by calculating 34 histograms related coefficient between any two and Amplitude Ration come the part of quantificational expression face
Dynamic pattern information;Wherein, in step 3-3-1, any two histogram is given, vector QUOTE is expressed as With
QUOTE , the calculation formula of related coefficient is as follows:
QUOTE (1)
Wherein QUOTE And QUOTE Respectively QUOTE And QUOTE It is equal
Value;Based on above-mentioned formula, 34*33/2=561 related coefficient can be obtained;It, must by calculating meanwhile in step 3-3-2
To the ratio of 561 histogram amplitudes;Wherein the amplitude of histogram is the average light stream amplitude of pixel, i.e. all pixels in region
The sum of light stream amplitude divided by the area pixel point number;So far, related coefficient and Amplitude Ration are formed to the spy of 1122 dimensions altogether
Sign, to indicate the local motion mode of face.
Step 4: after the local motion mode for obtaining face by step 3, passing through preconfigured pattern classifier pair
The local motion mode of the face of acquisition is classified, and according to classification as a result, verifying the true of face in the video image
It is pseudo-.
Use pattern classifier determines the true and false of current facial image collected;It is extracted from current face's image
To local motion mode, i.e. after 1122 dimensional feature vectors, preparatory trained support vector machines (Support can be used
Vector Machine, SVM) pattern classification model determines the true and false of current input image.
In step 4, used support vector cassification model needs training in advance;For this purpose, being acquired using camera
The video sequence of 20 real human faces and 20 forgery faces;The duration of video sequence is 30s;Wherein, true in acquisition
When the video sequence of face, it is desirable that the head of gathered person and face carry out slight movement, such as shake the head, nod, and smile, speak
Etc.;Forgery face video sequence collected is divided into two classes, and one is sequence of the acquisition from photograph print, secondly certainly for acquisition
The sequence of tablet personal computer display screen;In collection process, the face of forgery can be static, can also carry out any form of fortune
Dynamic or torsional deformation.
After obtaining above-mentioned video sequence, equally by step 1, step 2 and step 3, human face region is therefrom extracted
Local motion mode feature, and dual mode classifier is obtained using linear SVM training.
The present invention is not limited to above-mentioned preferred forms, anyone can show that other are various under the inspiration of the present invention
The product of form, however, make any variation in its shape or structure, it is all that there is skill identical or similar to the present application
Art scheme, is within the scope of the present invention.
Claims (7)
1. a kind of face method for anti-counterfeit based on local motion mode characterized by comprising
Video image gathered in advance is analyzed, human face region is determined, and analyze the human face region, determines institute
State each face key point in human face region, in which:
The human face region is analyzed, determines that each face key point includes: in the human face region
The people is determined by the initial position message of face key point predetermined according to the position of the human face region
The position of each face key point in face region;
According to the position of face key point each in the human face region, extract on the video image with the face key point
The corresponding video image characteristic in position;
According to the video image characteristic, by preconfigured algorithm model, update on the video image with the face
The position of the corresponding face key point in region determines accurate face corresponding with the human face region on the video image
The position of key point;
After meeting preset condition, the above process is terminated;
According to video frame corresponding to the video image, the direction of motion and the amplitude letter of pixel in the video image are obtained
Breath;
According to the direction of motion and amplitude information of the pixel of acquisition, the face key point is analyzed, determines the people
The direction of motion and amplitude information where face key point in regional area, and determine according to the information direction of motion of regional area
Between and amplitude between relationship, to obtain the local motion mode of face, in which:
The direction of motion and amplitude information of the pixel according to acquisition analyze the face key point, determine institute
The direction of motion and amplitude information where stating face key point in regional area include: the position according to the accurate face key point
It sets, head zone in the video image is accurately divided, determines the respective image of head zone in the video image
Mask;According to the direction of motion and amplitude information of described image mask and the pixel of acquisition, extracts each accurate face and close
The direction of motion and amplitude information of head zone and non-head region of the key point in the regional area locating for it;
Classified by local motion mode of the preconfigured pattern classifier to the face of acquisition, and according to the knot of classification
Fruit verifies the true and false of face in the video image.
2. the face method for anti-counterfeit according to claim 1 based on local motion mode, which is characterized in that the face area
Domain is obtained or is utilized by human-face detector and is manually specified.
3. the face method for anti-counterfeit according to claim 1 based on local motion mode, which is characterized in that according to the people
The position of face key point accurately divides head zone in the video image, determines that corresponding image mask includes:
According to the position of the accurate face key point, face envelope corresponding with the position of the accurate face key point is determined
Line;And the region for using the face envelope including is as the human face region of the video image;
According to the connecting line at face envelope both ends in the video image, mirror image carried out to the face envelope, and by institute
It states face envelope to be combined with its mirror image, obtains a closed curve, the region for including using the curve is as described in
The head zone of video image;
According to the position of the position of the human face region of the video image and the head zone of the video image, the view is determined
The human face region of frequency image and the respective image mask of head zone.
4. the face method for anti-counterfeit according to claim 1 based on local motion mode, which is characterized in that according to the figure
As the direction of motion and amplitude information of mask and the pixel of acquisition, each accurate office of the face key point locating for it is extracted
The direction of motion and amplitude information of head zone and non-head region in portion region includes:
According to the parameter information of preconfigured regional area size, partial zones corresponding to each accurate face key point are determined
Domain;
According to described image mask, the pixel that head zone is fallen in the regional area is demarcated as foreground area, by institute
It states the pixel fallen in except head zone in regional area and is demarcated as background area;
According to the direction of motion and amplitude information of the pixel of acquisition;The foreground and background region counted in the regional area is each
From the direction of motion and amplitude information.
5. the face method for anti-counterfeit according to claim 4 based on local motion mode, which is characterized in that according to the people
The direction of motion and amplitude information of regional area, calculate the pass between the direction of motion and amplitude of different zones where face key point
System, the local motion mode for obtaining face include:
Believed according to the direction of motion of foreground and background of each face key point in the regional area locating for it and amplitude
Breath calculates between local foreground region, between local background region and the direction of motion between local foreground and background area
With the relationship of amplitude information;
According between the local foreground region of calculating, between local background region and local foreground and background area
The direction of motion and amplitude information relationship, determine the local motion mode of face.
6. the face method for anti-counterfeit according to claim 5 based on local motion mode, which is characterized in that according to described every
The direction of motion and amplitude information of foreground and background of a face key point in the regional area locating for it calculate local foreground
The relationship of the direction of motion and amplitude information between region, between local background region and between local foreground and background area
Include:
Based on the direction of motion and amplitude information in the foreground and background region in the regional area, if the direction of motion is quantized into
Dry section, obtains the motion information histogram for adding up the motion amplitude of pixel in each regional area;
According to the motion information histogram, the correlation between the motion information histogram of regional area described in any two is determined
Ratio between coefficient and motion amplitude.
7. the face method for anti-counterfeit according to claim 6 based on local motion mode, which is characterized in that according to calculating
The direction of motion and width between the local foreground region, between local background region and between local foreground and background area
The relationship of value information determines that the local motion mode of face includes:
According to the ratio between the related coefficient and the motion amplitude, by the related coefficient and fortune between all regional areas
Dynamic amplitude ratio is combined, and determination obtains the local motion mode of face.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410428040.6A CN105447432B (en) | 2014-08-27 | 2014-08-27 | A kind of face method for anti-counterfeit based on local motion mode |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410428040.6A CN105447432B (en) | 2014-08-27 | 2014-08-27 | A kind of face method for anti-counterfeit based on local motion mode |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105447432A CN105447432A (en) | 2016-03-30 |
CN105447432B true CN105447432B (en) | 2019-09-13 |
Family
ID=55557594
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410428040.6A Expired - Fee Related CN105447432B (en) | 2014-08-27 | 2014-08-27 | A kind of face method for anti-counterfeit based on local motion mode |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105447432B (en) |
Families Citing this family (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106228137A (en) * | 2016-07-26 | 2016-12-14 | 广州市维安科技股份有限公司 | A kind of ATM abnormal human face detection based on key point location |
CN108229326A (en) | 2017-03-16 | 2018-06-29 | 北京市商汤科技开发有限公司 | Face false-proof detection method and system, electronic equipment, program and medium |
CN107358155A (en) * | 2017-06-02 | 2017-11-17 | 广州视源电子科技股份有限公司 | Method and device for detecting ghost face action and method and system for recognizing living body |
CN107688781A (en) * | 2017-08-22 | 2018-02-13 | 北京小米移动软件有限公司 | Face identification method and device |
CN107643826A (en) * | 2017-08-28 | 2018-01-30 | 天津大学 | A kind of unmanned plane man-machine interaction method based on computer vision and deep learning |
CN108537131B (en) * | 2018-03-15 | 2022-04-15 | 中山大学 | Face recognition living body detection method based on face characteristic points and optical flow field |
CN108846321B (en) * | 2018-05-25 | 2022-05-03 | 北京小米移动软件有限公司 | Method and device for identifying human face prosthesis and electronic equipment |
CN109583391B (en) * | 2018-12-04 | 2021-07-16 | 北京字节跳动网络技术有限公司 | Key point detection method, device, equipment and readable medium |
CN109766785B (en) * | 2018-12-21 | 2023-09-01 | 中国银联股份有限公司 | Living body detection method and device for human face |
CN110223322B (en) * | 2019-05-31 | 2021-12-14 | 腾讯科技(深圳)有限公司 | Image recognition method and device, computer equipment and storage medium |
CN112269975A (en) * | 2020-03-31 | 2021-01-26 | 周亚琴 | Internet of things artificial intelligence face verification method and system and Internet of things cloud server |
CN111626101A (en) * | 2020-04-13 | 2020-09-04 | 惠州市德赛西威汽车电子股份有限公司 | Smoking monitoring method and system based on ADAS |
CN111611873B (en) * | 2020-04-28 | 2024-07-16 | 平安科技(深圳)有限公司 | Face replacement detection method and device, electronic equipment and computer storage medium |
CN112287909B (en) * | 2020-12-24 | 2021-09-07 | 四川新网银行股份有限公司 | Double-random in-vivo detection method for randomly generating detection points and interactive elements |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101169827A (en) * | 2007-12-03 | 2008-04-30 | 北京中星微电子有限公司 | Method and device for tracking characteristic point of image |
CN102750518A (en) * | 2012-05-30 | 2012-10-24 | 深圳光启创新技术有限公司 | Face verification system and method based on visible light communications |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5795979B2 (en) * | 2012-03-15 | 2015-10-14 | 株式会社東芝 | Person image processing apparatus and person image processing method |
-
2014
- 2014-08-27 CN CN201410428040.6A patent/CN105447432B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101169827A (en) * | 2007-12-03 | 2008-04-30 | 北京中星微电子有限公司 | Method and device for tracking characteristic point of image |
CN102750518A (en) * | 2012-05-30 | 2012-10-24 | 深圳光启创新技术有限公司 | Face verification system and method based on visible light communications |
Non-Patent Citations (1)
Title |
---|
基于光流的动态人脸表情识别;余棉水等;《微电子学与计算机》;20051231;第22卷(第7期);第113-115、119页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105447432A (en) | 2016-03-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105447432B (en) | A kind of face method for anti-counterfeit based on local motion mode | |
Agarwal et al. | Protecting World Leaders Against Deep Fakes. | |
Shao et al. | Deep convolutional dynamic texture learning with adaptive channel-discriminability for 3D mask face anti-spoofing | |
Patel et al. | Live face video vs. spoof face video: Use of moiré patterns to detect replay video attacks | |
Yang et al. | Learn convolutional neural network for face anti-spoofing | |
Patel et al. | Secure face unlock: Spoof detection on smartphones | |
Bharadwaj et al. | Computationally efficient face spoofing detection with motion magnification | |
CN104933414B (en) | A kind of living body faces detection method based on WLD-TOP | |
Shan | Smile detection by boosting pixel differences | |
CN102375970B (en) | A kind of identity identifying method based on face and authenticate device | |
CN110458063B (en) | Human face living body detection method for preventing video and photo cheating | |
CN108416291B (en) | Face detection and recognition method, device and system | |
CN108021892A (en) | A kind of human face in-vivo detection method based on extremely short video | |
CN102214309A (en) | Special human body recognition method based on head and shoulder model | |
Rehman et al. | Enhancing deep discriminative feature maps via perturbation for face presentation attack detection | |
Huang et al. | Deepfake mnist+: a deepfake facial animation dataset | |
JP5879188B2 (en) | Facial expression analysis apparatus and facial expression analysis program | |
Azzopardi et al. | Fast gender recognition in videos using a novel descriptor based on the gradient magnitudes of facial landmarks | |
CN113468954B (en) | Face counterfeiting detection method based on local area features under multiple channels | |
KR101727833B1 (en) | Apparatus and method for constructing composite feature vector based on discriminant analysis for face recognition | |
Putro et al. | Adult image classifiers based on face detection using Viola-Jones method | |
Liu | Face liveness detection using analysis of Fourier spectra based on hair | |
Kisku et al. | Face spoofing and counter-spoofing: a survey of state-of-the-art algorithms | |
CN111126283B (en) | Rapid living body detection method and system for automatically filtering fuzzy human face | |
Gürel | Development of a face recognition system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C41 | Transfer of patent application or patent right or utility model | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20160531 Address after: 425000 Yongzhou City, Hunan province Lengshuitan District Fushan Road Pearl Street No. 127 Applicant after: Yang Jianwei Address before: 100084 B1 floor, block A, Wan Lin Building, No. 88, Nongda South Road, Beijing, Haidian District Applicant before: QIANSOU INC. |
|
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190913 Termination date: 20210827 |