CN105006020B - A kind of conjecture face generation method based on 3D models - Google Patents
A kind of conjecture face generation method based on 3D models Download PDFInfo
- Publication number
- CN105006020B CN105006020B CN201510411239.2A CN201510411239A CN105006020B CN 105006020 B CN105006020 B CN 105006020B CN 201510411239 A CN201510411239 A CN 201510411239A CN 105006020 B CN105006020 B CN 105006020B
- Authority
- CN
- China
- Prior art keywords
- msub
- face
- characteristic point
- abscissa
- posture
- 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 invention discloses a kind of conjecture face generation method based on 3D models, comprise the following steps:The posture P1 and posture P2 of two image for deflecting into different angles formation are selected from image data base as training sample；N characteristic point is demarcated under posture P1 and posture P2 respectively, n characteristic point pair is constituted altogether, n is positive integer；Using characteristic point to calculating the corresponding parameter a of each characteristic pointiAnd ri, riFor, using neck as axle center, the point on face is the radius of turn on border in three-dimensional space；aiFor the abscissa positions put on front face image；Test sample is calculated from front deflection angle θjWhen abscissa array；Test sample deflection θ is respectively obtained using abscissa arrayjConjecture face image.The conjecture face image and real human face image that the present invention is generated are much like, and especially in the case where angle is relatively low, generation figure and artwork almost can not be identified.
Description
Technical field
The present invention relates to a kind of mode identification technology, more particularly to a kind of conjecture face generation based on 3D models
Method.
Background technology
In recent decades, research of the people to biometric technology obtains some achievements, wherein face recognition technology more
It is to achieve breakthrough progress.Face recognition technology is a kind of important biometrics identification technology, relative to fingerprint, iris
Etc. identification technology, it is with more handling characteristics such as convenience, close friends, therefore it has been obtained extensively in many fields such as safety, business
Using.Due to human face posture change, ambient lighting change, and human face expression change, appearance transition, wearing spectacles and cap etc.
Shelter, existing face recognition technology accuracy rate can decline to a great extent.Wherein human face posture change is caused under recognition accuracy
One of major reason of drop, therefore fast, accurately core identification method of the research based on multi-pose just seems extremely important.
At present, the method for solving pose problem in recognition of face is substantially broadly divided into two classes：Method based on two-dimensional signal
With the method based on 3D information.Wherein, the face identification method based on 3D information, mainly using shape information, texture letter
The essential attribute of the faces such as breath is identified.Curvature analysis is for example carried out come real based on high-quality laser scanning depth data
Existing recognition of face, but because accurate 3D data need the hardware devices such as laser scanning to obtain, in acquisition process to
The interference at family is than larger, and data obtaining time is long, and three-dimensional data computation complexity is far above 2-D data, amount of calculation compared with
Greatly, thus which greatly limits based on 3D information carry out recognition of face application.For another example 3D transform model method, this method is by 3D
The projection of distorted pattern and computer graphics and lighting simulation are combined, and finally realize human face rebuilding, but whole process is needed
Want iteration optimization to complete, calculate complicated, than relatively time-consuming.And the face multi-pose problem-solving approach based on two-dimensional signal, substantially
Two classes can be divided them into：(1) method based on view, i.e., it is necessary under everyone multi-pose in training/enrolled set
Several views are identified；(2) method standardized based on posture, i.e., determine test (or enrolled set) image rectification to advance
It is identified under the standard posture of justice, posture standardization is included the appearance under attitude updating to front or other deflection angles
State.The method wherein standardized based on posture achieves certain experiment effect, but there is also some problems simultaneously.These problems
Can not be too big including such as side image deflection angle, or some face parts of direct picture are not deposited in side
, it is impossible to the information of loss is found from the face of side, causes the direct picture of generation to differ too big with true picture.
The content of the invention
It is contemplated that at least solving technical problem present in prior art, especially innovatively propose a kind of based on 3D
The conjecture face generation method of model.
In order to realize the above-mentioned purpose of the present invention, the invention provides a kind of conjecture face generation side based on 3D models
Method, comprises the following steps:
S1, determines that face deflects the relation between three-dimensional space and two-dimensional image；
S2, selects the image that two different angles are deflected into from front of same target, two partially from image data base
The posture P1 and posture P2 of image formation of different angles are gone to as training sample；
S3, the mark point demarcated respectively under posture P1 and posture P2 on n characteristic point, P1 and P2 is corresponded, altogether structure
Into n characteristic point pair, the n is positive integer, and characteristic point is to for P1 abscissa arraysAnd P2
Abscissa array
S4, using characteristic point to X1And X2Calculate the corresponding parameter a of each characteristic pointiAnd ri, riFor three-dimensional space
In using neck as axle center, point on face is the radius of turn on border；aiFor the abscissa positions put on front face image,
The radius of turn r ' of non-characteristic point is equal to the radius of turn of the characteristic point nearest with it；
S5, calculates test sample from front deflection angle θjWhen abscissa arrayXi
It is test sample from front deflection angle θjWhen abscissa array；Respectively be test sample from
Front deflection angle θjWhen first characteristic point abscissa, second characteristic point abscissa, the 3rd characteristic point it is horizontal sit
Mark ... ..., n-th of characteristic point abscissa；
S6, utilizes XiRespectively obtain test sample deflection θjConjecture face image.
In the preferred embodiment of the present invention, in S3, two deflections by following steps to same target
Different angular images carry out characteristic point demarcation：
S21：Select same target from positive deflecting facet θ1The image posture P1 of angle and from positive deflecting facet θ2The image of angle
Posture P2；
S22：A characteristic point is chosen on P1
S23：A characteristic point is chosen on P2
S24：Repeat step S22, S23, while choosing next characteristic point in P1When meet following condition：
Wherein, M scope be 10 ＜ M ＜ 15, the i be 1,2,3 ..., n, when the characteristic point chosen on face P1, P2
When being n, now, posture P1 abscissa arraysOrdinate array
Posture P2 abscissa arraysOrdinate arrayForm n
Individual characteristic point pair, because P1 and P2 are only deflected, Y1=Y2；Terminate to choose.
In a kind of more preferred embodiment of the present invention, the key feature points that the characteristic point on P1 is behaved on the face, institute
The characteristic point that stating key feature points includes choosing in characteristic point and P1 on nose, eyes or face, P2 is same characteristic point.
In the preferred embodiment of the present invention, in step S4, corresponding a is obtained by followingiAnd riValue：
In formula, θ1Posture P1 deflection angle is deflected into from front for face；θ2Deflect into posture P2's from front for face
Deflection angle；riFor in 3d space using neck as axle center, i-th point on face be border radius of turn；aiFor positive face face
I-th point of abscissa positions on image；It is face from positive deflecting facet θ1During angle, the abscissa position of the i-th corresponding points on image
Put；It is face from positive deflecting facet θ2During angle, the abscissa positions of the i-th corresponding points on image；The i be no more than n just
Integer；
Now obtain the radius array R=(r of characteristic point1,r2,r3,...,rn) and corresponding abscissa array A=(a1,
a2,a3,...,an)。
In the preferred embodiment of the present invention,Computational methods be：
In formula, θjPosture Pj deflection angle is deflected into from front for face；riFor in 3d space using neck as axle center,
I-th point of radius of turn for border on face；aiFor i-th point on positive face facial image of abscissa positions；For face from
Positive deflecting facet θjDuring angle, the abscissa positions of the i-th corresponding points on image；The i is no more than n positive integer, is now obtained
In the preferred embodiment of the present invention, face deflection is between three-dimensional space and two-dimensional image
Relation it is as follows：
D=r*sin θ,
In formula, θ is angle of the face from positive deflecting facet；D be two dimensional surface facial image on put from positive deflecting facet θ away from
From；R is that, using neck as axle center, the point on face is the radius of turn on border in three-dimensional space.
In summary, by adopting the above-described technical solution, the beneficial effects of the invention are as follows：The visual human that the present invention is generated
Face image and real human face image are much like, especially in the case where angle is relatively low, and generation figure and artwork almost can not be identified.
Brief description of the drawings
In Fig. 1, (a) is 3D rendering rotation schematic diagram in the present invention；(b) it is that 3D rendering of the present invention shows with two dimensional image relation
It is intended to.
In Fig. 2, (a) is posture P1 characteristic points mark schematic diagram in a kind of preferred embodiment of the invention, and (b) is the present invention
Posture P2 characteristic points mark schematic diagram in a kind of preferred embodiment.
In Fig. 3, the direct picture schematic diagram of face in a kind of (a) preferred embodiment of the invention, (b) present invention is a kind of excellent
Select the conjecture face image schematic diagram obtained in embodiment.
Fig. 4 is the radius selection schematic diagram of non-characteristic point of the invention.
Fig. 5 is the contrast schematic diagram that different images of the present invention deflect different angles.
Embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, is only used for explaining the present invention, and is not considered as limiting the invention.
The invention provides a kind of conjecture face generation method based on 3D models, comprise the following steps:
S1, determines that face deflects the relation between three-dimensional space and two-dimensional image；In the present embodiment,
Relation of the face deflection between three-dimensional space and two-dimensional image is as follows：
D=r*sin θ,
In formula, θ is angle of the face from positive deflecting facet；D be two dimensional surface facial image on put from positive deflecting facet θ away from
From；R is that, using neck as axle center, the point on face is the radius of turn on border in three-dimensional space.
S2, selects the image that two different angles are deflected into from front of same target, two partially from image data base
The posture P1 and posture P2 of image formation of different angles are gone to as training sample；Posture is that face is deflected under certain angle
State, i.e. face deflect the two dimensional image of certain angle.
In the present embodiment, image posture P1 is same target from positive deflecting facet θ1Angle, image posture P2 is
From positive deflecting facet θ2Angle；
S3, the mark point demarcated respectively under posture P1 and posture P2 on n characteristic point, P1 and P2 is corresponded, altogether structure
Into n characteristic point pair, n is positive integer, and characteristic point is to for P1 abscissa arraysWith P2 abscissa numbers
GroupWherein X1For P1 abscissa arrays, X2For P2 abscissa arrays,According to
Secondary is respectively from front deflection angle θ1When first characteristic point abscissa, second characteristic point abscissa, the 3rd feature
Point abscissa ... ..., n-th of characteristic point abscissa,It respectively is from front deflection angle θ2When
First characteristic point abscissa, second characteristic point abscissa, the 3rd characteristic point abscissa ... ..., n-th characteristic point be horizontal
Coordinate；In the present embodiment, specific method is：
S21：A characteristic point is chosen on P1
S22：A characteristic point is chosen on P2
S23：Repeat step S21, S22, while choosing next characteristic point in P1When meet following condition：
Wherein, M scope be 10 ＜ M ＜ 15, i be 1,
2nd, 3 ..., n, when the characteristic point chosen on human face posture P1, posture P2 is n, now, posture P1 abscissa arraysOrdinate arrayPosture P2 abscissa arraysOrdinate arrayN characteristic point pair is formed, wherein
Y1For P1Ordinate array, Y2For P2Ordinate array,It respectively is from front deflection angle θ1When
First characteristic point ordinate, second characteristic point ordinate,
3rd characteristic point ordinate ... ..., n-th of characteristic point ordinate,It respectively is
From front deflection angle θ2When first characteristic point ordinate, second characteristic point ordinate, the 3rd characteristic point is vertical to sit
Mark ... ..., n-th of characteristic point ordinate, because P1 and P2 are only deflected, Y1=Y2；Terminate to choose.
S4, using characteristic point to X1And X2Calculate the corresponding parameter a of each characteristic pointiAnd ri, riFor three-dimensional space
In using neck as axle center, point on face is the radius of turn on border；aiFor the abscissa positions put on front face image,
The radius of turn r ' of non-characteristic point is equal to the radius of turn of the characteristic point nearest with it；In the present embodiment, obtained by following
Obtain corresponding aiAnd riValue：
In formula, θ1Posture P1 deflection angle is deflected into from front for face；θ2It is face from just
Deflection angle of the deflecting facet to posture P2；riFor in 3d space using neck as axle center, i-th point on face be border rotation
Radius；aiFor i-th point on positive face facial image of abscissa positions；It is face from positive deflecting facet θ1During angle, i-th on image
The abscissa positions of corresponding points；It is face from positive deflecting facet θ2During angle, the abscissa positions of the i-th corresponding points on image；I is
No more than n positive integer；Now obtain the radius array R=(r of characteristic point1,r2,r3,...,rn) and corresponding abscissa array
A=(a1,a2,a3,...,an), wherein R is characterized radius array a little, and A is the horizontal seat corresponding with the radius array of characteristic point
Mark array, a1,a2,a3,......,anIt is followed successively by respectively from front deflection angle θ1When first characteristic point abscissa,
The abscissa of two characteristic points, the abscissa ... ... of the 3rd characteristic point, the point abscissa of n-th of feature；r1,r2,
r3,......,rnIt is followed successively by respectively and a1Corresponding radius, with a2Corresponding radius, with a3Corresponding radius ... ..., with anIt is right
The radius answered.
S5, calculates test sample from front deflection angle θjWhen abscissa arrayXi
It is test sample from front deflection angle θjWhen abscissa array；Respectively be test sample from
Front deflection angle θjWhen first characteristic point abscissa, second characteristic point abscissa, the 3rd characteristic point it is horizontal sit
Mark ... ..., n-th of characteristic point abscissa；In the present embodiment,Computational methods be：
In formula, θjPosture Pj deflection angle is deflected into from front for face；riFor in 3d space using neck as axle center,
I-th point of radius of turn for border on face；aiFor i-th point on positive face facial image of abscissa positions；For face from
Positive deflecting facet θjDuring angle, the abscissa positions of the i-th corresponding points on image；The i is no more than n positive integer, is now obtained
S6, utilizes XiRespectively obtain test sample deflection θxConjecture face image.
In a kind of more preferred embodiment of the present invention, the key feature that the characteristic point on posture P1 is behaved on the face
Point, the characteristic point that key feature points include choosing in characteristic point and P1 on nose, eyes or face, posture P2 is same feature
Point.
When face is deflected, in the 3 d space, the movement locus put on face is circular arc, and the radius r of the circular arc
It is that, using neck as axle center, nose is border.Put in two dimensional surface, on face and 90 degree of movement locus is deflected into from zero degree
For straight line, straight line length when deflection angle is 0 is also 0, and as deflection angle increases, straight line also gradually increases.Such as
Shown in Fig. 1.Therefore, when face deflects different angles, the deflection distance of point abscissa meets deflection on two dimensional surface facial image
Formula, i.e.,：
A'=a+r*sin θ (1)
θ is face from front deflection angle in formula；It in three dimensions is the point (P on face using neck as axle center that r, which is,
Point) be border radius of turn；A is the abscissa positions put when face is positive face on image；A ' is face from positive deflecting facet θ angles
When spending on image corresponding points abscissa positions, c be certain point (P points are nose) deflect θ angles when 3d space movement arc length.
Embodiment：
We select 10 objects from MIT-CBCL databases.Each 1 direct picture of object is selected, totally 10 images
It is used as test sample；1 direct picture of same object is randomly choosed, 1 deflects 20 ° of images and is used as training sample to the left.
20 ° of image P2,370 points of demarcation by hand are deflected in direct picture P1 and to the left respectively, these mark points are corresponded, constituted altogether
370 characteristic points pair, demarcation effect is as shown in Figure 2.Calibration point (characteristic point) abscissa is under each posture：Posture P1 (deflection angles
Equal to 0 °) abscissa array X1=(x1 0,x2 0,x3 0,...,x370 0), ordinate array Y1=(y1 0,y2 0,y3 0,...,y370 0),
Posture P2 (deflection angle is equal to 20 °) abscissa array X2=(x1 20,x2 20,x3 20,...,x370 20), ordinate array Y2=(y1 20,
y2 20,y3 20,...,y370 20).According to formula
Calculate the radius R=(r of X1 and X2 characteristic points1,r2,r3,...,r370) and correspondence abscissa A=(a1,a2,
a3,...,a370).Then the radius of turn of other non-marked points is equal to the radius of turn of the nearest marker characteristic point of Euclidean distance, such as
Really multiple minimum distances are equal, then randomly choose.As shown in figure 3, nose mark point p1 (x1, y1) passes through formula in direct picture
(1) calculate and obtained 20 ° of corresponding characteristic point p1 ' of deflection (x1 ', y1 ').Wherein centered on mark point p1, D × D (D in figure
Non-marked point p0 radius of turn is equal to p1 radius of turn in=neighborhood 5), finally obtains virtual image such as Fig. 3 (b) institutes
Show.Wherein D is variable, typically takes 4 or 5.If the length of side for D square block in have multiple marker characteristic points, other non-markeds
The radius of turn of point is equal to the radius of turn of the nearest marker characteristic point of Euclidean distance, if multiple minimum distances are equal, with
Machine is selected.As shown in Figure 4, non-marked point p0 radius of turn is equal to p1 radius of turn.
The abscissa of any deflection angle is can obtain by formula (1), because the present invention only considers deflection, therefore ordinate is not sent out
It is raw to change.With reference to the abscissa and correspondence ordinate calculated, the conjecture face image of any deflection angle is obtained, such as Fig. 5 institutes
Show.Partial objects experimental result is included in Fig. 5.Wherein angle refers to the angle that face is deflected to the left, and artwork refers to claps under corresponding posture
The true picture taken the photograph, generation figure refers to utilize this by posture P1 (deflection angle is equal to 0 °) and posture P2 (deflection angle is equal to 20 °)
The conjecture face image of inventive method generation.Found out by experimental result, the conjecture face image and real human face of this method generation
Image is much like, especially in the case where angle is relatively low, and generation figure and artwork almost can not be identified.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means to combine specific features, structure, material or the spy that the embodiment or example are described
Point is contained at least one embodiment of the present invention or example.In this manual, to the schematic representation of above-mentioned term not
Necessarily refer to identical embodiment or example.Moreover, specific features, structure, material or the feature of description can be any
One or more embodiments or example in combine in an appropriate manner.
Although an embodiment of the present invention has been shown and described, it will be understood by those skilled in the art that：Not
In the case of departing from the principle and objective of the present invention a variety of change, modification, replacement and modification can be carried out to these embodiments, this
The scope of invention is limited by claim and its equivalent.
Claims (6)
1. a kind of conjecture face generation method based on 3D models, it is characterised in that comprise the following steps:
S1, determines that face deflects the relation between three-dimensional space and two-dimensional image；
S2, selects the image that two different angles are deflected into from front of same target, two deflect into from image data base
The posture P1 and posture P2 of the image formation of different angles are used as training sample；
S3, the mark point demarcated respectively under posture P1 and posture P2 on n characteristic point, P1 and P2 is corresponded, and n is constituted altogether
Individual characteristic point pair, the n is positive integer, and characteristic point is to for P1 abscissa arraysWith the horizontal seats of P2
Mark array
S4, using characteristic point to X1And X2Calculate the corresponding parameter a of each characteristic pointiAnd ri, riFor in three-dimensional space with
Neck is axle center, and the point on face is the radius of turn on border；aiFor the abscissa positions put on front face image, non-spy
The radius of turn r ' levied a little is equal to the radius of turn of the characteristic point closest with it；
S5, calculates test sample from front deflection angle θjWhen abscissa arrayXi
It is test sample from front deflection angle θjWhen abscissa array；Respectively be test sample from
Front deflection angle θjWhen first characteristic point abscissa, second characteristic point abscissa, the 3rd characteristic point it is horizontal sit
Mark ... ..., n-th of characteristic point abscissa；
S6, utilizes XiRespectively obtain test sample deflection θjConjecture face image.
2. the conjecture face generation method according to claim 1 based on 3D models, it is characterised in that in S3, pass through
Following steps carry out characteristic point demarcation to the different angular images of two deflections of same target：
S21：Select same target from positive deflecting facet θ1The image posture P1 of angle and from positive deflecting facet θ2The image posture of angle
P2；
S22：A characteristic point is chosen on P1
S23：A characteristic point is chosen on P2
S24：Repeat step S22, S23, while choosing next characteristic point in P1When meet following condition：
<mrow>
<mi>m</mi>
<mo><</mo>
<mo>|</mo>
<msup>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<msub>
<mi>&theta;</mi>
<mn>1</mn>
</msub>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&theta;</mi>
<mn>1</mn>
</msub>
</msup>
<mo>|</mo>
<mo><</mo>
<mi>M</mi>
<mo>,</mo>
<mi>m</mi>
<mo><</mo>
<mo>|</mo>
<msup>
<msub>
<mi>y</mi>
<mrow>
<mi>i</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</msub>
<msub>
<mi>&theta;</mi>
<mn>1</mn>
</msub>
</msup>
<mo>-</mo>
<msup>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&theta;</mi>
<mn>1</mn>
</msub>
</msup>
<mo>|</mo>
<mo><</mo>
<mi>M</mi>
<mo>,</mo>
<mn>0</mn>
<mo><</mo>
<mi>m</mi>
<mo><</mo>
<mi>M</mi>
<mo>,</mo>
</mrow>
Wherein, M scope be 10 ＜ M ＜ 15, the i be 1,2,3 ..., n, when the characteristic point chosen on face P1, P2 is n
When individual, now, posture P1 abscissa arraysOrdinate array
Posture P2 abscissa arraysOrdinate arrayForm n
Individual characteristic point pair, because P1 and P2 are only deflected, Y1=Y2；Terminate to choose.
3. the conjecture face generation method according to claim 2 based on 3D models, it is characterised in that the characteristic point on P1
For the key feature points of people on the face, the key feature points include choosing in characteristic point and P1 on nose, eyes or face, P2
Characteristic point be same characteristic point.
4. the conjecture face generation method according to claim 1 based on 3D models, it is characterised in that in step S4, leads to
Cross the following corresponding a of acquisitioniAnd riValue：
<mrow>
<msup>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&theta;</mi>
<mi>i</mi>
</msub>
</msup>
<mo>=</mo>
<mi>z</mi>
<mo>+</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>*</mo>
<msub>
<mi>sin&theta;</mi>
<mn>1</mn>
</msub>
</mrow>
<mrow>
<msup>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&theta;</mi>
<mn>2</mn>
</msub>
</msup>
<mo>=</mo>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>*</mo>
<msub>
<mi>sin&theta;</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
</mrow>
In formula, θ1Posture P1 deflection angle is deflected into from front for face；θ2Posture P2 deflection is deflected into from front for face
Angle；riFor in 3d space using neck as axle center, i-th point on face be border radius of turn；aiFor positive face facial image
Upper i-th point of abscissa positions；It is face from positive deflecting facet θ1During angle, the abscissa positions of the i-th corresponding points on image；It is face from positive deflecting facet θ2During angle, the abscissa positions of the i-th corresponding points on image；The i is just whole for no more than n's
Number；
Now obtain the radius array R=(r of characteristic point1,r2,r3,...,rn) and corresponding abscissa array A=(a1,a2,
a3,...,an)。
5. the conjecture face generation method according to claim 1 based on 3D models, it is characterised in that：Computational methods
For：
<mrow>
<msup>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<msub>
<mi>&theta;</mi>
<mi>j</mi>
</msub>
</msup>
<mo>=</mo>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<msub>
<mi>r</mi>
<mi>i</mi>
</msub>
<mo>*</mo>
<msub>
<mi>sin&theta;</mi>
<mi>j</mi>
</msub>
<mo>,</mo>
</mrow>
In formula, θjPosture Pj deflection angle is deflected into from front for face；riFor in 3d space using neck as axle center, face
On i-th point be border radius of turn；aiFor i-th point on positive face facial image of abscissa positions；It is face from front
Deflect θjDuring angle, the abscissa positions of the i-th corresponding points on image；The i is no more than n positive integer, is now obtained
6. the conjecture face generation method according to claim 1 based on 3D models, it is characterised in that face deflection
Relation between three-dimensional space and two-dimensional image is as follows：
D=r*sin θ,
In formula, θ is angle of the face from positive deflecting facet；D is that the distance from positive deflecting facet θ is put on two dimensional surface facial image；R is
Using neck as axle center in three-dimensional space, the point on face is the radius of turn on border.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510411239.2A CN105006020B (en) | 2015-07-14 | 2015-07-14 | A kind of conjecture face generation method based on 3D models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510411239.2A CN105006020B (en) | 2015-07-14 | 2015-07-14 | A kind of conjecture face generation method based on 3D models |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105006020A CN105006020A (en) | 2015-10-28 |
| CN105006020B true CN105006020B (en) | 2017-11-07 |
Family
ID=54378676
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510411239.2A Expired - Fee Related CN105006020B (en) | 2015-07-14 | 2015-07-14 | A kind of conjecture face generation method based on 3D models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105006020B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3993410A1 (en) * | 2020-10-28 | 2022-05-04 | Ningbo Geely Automobile Research & Development Co., Ltd. | A camera system and method for generating an eye contact image view of a person |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101101672A (en) * | 2007-07-13 | 2008-01-09 | 中国科学技术大学 | Stereo vision three-dimensional human face modelling approach based on dummy image |
| CN101320484A (en) * | 2008-07-17 | 2008-12-10 | 清华大学 | Three-dimensional human face recognition method based on human face full-automatic positioning |
| US8265399B2 (en) * | 2003-06-26 | 2012-09-11 | DigitalOptics Corporation Europe Limited | Detecting orientation of digital images using face detection information |
-
2015
- 2015-07-14 CN CN201510411239.2A patent/CN105006020B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8265399B2 (en) * | 2003-06-26 | 2012-09-11 | DigitalOptics Corporation Europe Limited | Detecting orientation of digital images using face detection information |
| CN101101672A (en) * | 2007-07-13 | 2008-01-09 | 中国科学技术大学 | Stereo vision three-dimensional human face modelling approach based on dummy image |
| CN101320484A (en) * | 2008-07-17 | 2008-12-10 | 清华大学 | Three-dimensional human face recognition method based on human face full-automatic positioning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105006020A (en) | 2015-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108334816B (en) | Multi-pose face recognition method based on contour symmetric constraint generation type countermeasure network | |
| CN106650630B (en) | A kind of method for tracking target and electronic equipment | |
| US20200334853A1 (en) | Facial features tracker with advanced training for natural rendering of human faces in real-time | |
| JP6918198B2 (en) | System and method for face position adjustment that is robust against posture fluctuations | |
| CN103678754B (en) | Information processor and information processing method | |
| CN101853523B (en) | Method for adopting rough drawings to establish three-dimensional human face molds | |
| CN103605965A (en) | Multi-pose face recognition method and device | |
| CN105069746A (en) | Video real-time human face substitution method and system based on partial affine and color transfer technology | |
| CN106485207B (en) | A kind of Fingertip Detection and system based on binocular vision image | |
| CN105719326A (en) | Realistic face generating method based on single photo | |
| CN101751689A (en) | Three-dimensional facial reconstruction method | |
| CN109584290A (en) | A kind of three-dimensional image matching method based on convolutional neural networks | |
| WO2015001791A1 (en) | Object recognition device objection recognition method | |
| CN104899563A (en) | Two-dimensional face key feature point positioning method and system | |
| CN107145224B (en) | Human eye sight tracking and device based on three-dimensional sphere Taylor expansion | |
| CN107657645B (en) | Method for calibrating parabolic catadioptric camera by using properties of conjugate diameters of straight line and circle | |
| CN109523595A (en) | A kind of architectural engineering straight line corner angle spacing vision measuring method | |
| CN101916384B (en) | Facial image reconstruction method and device and face recognition system | |
| CN111145228A (en) | Heterogeneous image registration method based on local contour point and shape feature fusion | |
| CN106652037B (en) | Face mapping processing method and device | |
| CN108615256A (en) | A kind of face three-dimensional rebuilding method and device | |
| CN102385754B (en) | A kind of object tracking methods and equipment | |
| CN105893984A (en) | Face projection method for facial makeup based on face features | |
| CN108335325A (en) | A kind of cube method for fast measuring based on depth camera data | |
| CN109325994A (en) | A method of enhanced based on three-dimensional face data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| 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: 20171107 Termination date: 20210714 |