CN106683169A - Sparse local decomposition and restructuring algorithm of joint motion sensing - Google Patents
Sparse local decomposition and restructuring algorithm of joint motion sensing Download PDFInfo
- Publication number
- CN106683169A CN106683169A CN201611125996.4A CN201611125996A CN106683169A CN 106683169 A CN106683169 A CN 106683169A CN 201611125996 A CN201611125996 A CN 201611125996A CN 106683169 A CN106683169 A CN 106683169A
- Authority
- CN
- China
- Prior art keywords
- grid model
- tri patch
- triangle
- triangle grid
- vector
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a sparse local decomposition and restructuring algorithm of joint motion sensing; the method includes steps of 1), calculating length of sides in a triangular mesh and size of dihedral angles corresponding to sides (the two values are recorded as LA); 2), calculating residual difference of LA vector in every triangular mesh; 3), performing sparse local decomposition on the residual difference of the LA vector and acquiring an LA base; 4), linearly mixing the LA base and acquiring a new LA vector; 5), recovering a connection mapping; 6), estimating the bid price of local orthogonality; 7), restructuring the triangular mesh. The algorithm mainly solves the problem how to perform the sparse local decomposition on a given movement mesh sequence, thereby capturing deformation of a local zone of a movement sequence; moreover, by taking it as the basis, an integrated treatment framework capable of capturing the movement zone and performing gesture edition of the triangular mesh model is structured. The sparse local decomposition and restructuring algorithm is better in locality, applicable to large-scale rotating movement and deformation, overall rotation, and good in robustness.
Description
Technical field
The present invention relates to computer graphics and three-dimensional grid editor field, a kind of the dilute of joint motions perception is referred in particular to
Dredge exploded and restructing algorithm.
Background technology
Dynamic geometry or animated Mesh sequence are the triangular mesh with identical annexation, usual these triangles
Grid have recorded different attitudes of the object in a time series.This solid all right and wrong in animation and 3D game
It is often universal.A succinct representation to this triangular mesh sequence is obtained, and supports to exist the editor of solid
It is a very important problem in dynamic geometry treatment.
Traditional mode generally catches given triangle using succinct structure as such as skeleton or cage (Cage)
The operation information of shape grid sequence, and triangular mesh reconstruct is carried out by using covering technology or broad sense quality coordinate.
Although these methods can well compress triangular mesh sequence, want by operate these skeletons or cage come
Build the triangular mesh with new attitude extremely difficult.
Each frame of triangular mesh sequence is considered as sparse local shape based method the change of average shape triangular mesh
Shape, and extract one group of local shape base that can capture posture deforming.By this group of parameter value of shape base of editor control
To constitute the combination of different shape base, so as to reconstruct the triangular mesh model with different attitudes.This method is raising eyebrow
Hair, improves the facial expression aspect such as corners of the mouth and is demonstrated by powerful ability.However, what is utilized due to this group of shape base is non-locality
Changes in coordinates, this method reconstruct rotating operation on a large scale when, easy bad student's artifact.Huang et al. is attempted by triangle
The deformation gradient (DG) of shape patch grids spatially carries out sparse exploded and represents and joint motions filling local linear
The breach of nonlinear characteristic.Although this mode can to a certain extent represent rotary motion, due to DG change it is non-
Locality, its edit capability is limited by very large.The change of DG has non-locality, because firm when triangular mesh
When property rotates partially around joint, all of DG can have similar change on rigid element.Additionally, special as non-intrinsically safe
Levy, the conversion of DG depends on global rotation transformation.Therefore, this method can not process the animation sequence with global rotation.Most
Afterwards, each deformation gradient is defined by 9 parameters, when a triangular mesh has NF9N is accomplished by during individual faceFIndividual parameter carrys out table
Show an attitude.This just substantially increases the process time of local Its Sparse Decomposition.
The above method major defect has:(1) succinct mode can be compared to represent motion sequence triangular mesh
Model, but cannot but provide more succinct mode edits the attitude of triangular mesh;(2) edlin can be entered, but
Rotating range cannot be processed than larger athletic posture;(3) represent that the data volume of grid sequence is big, calculate complicated.
The content of the invention
Shortcoming and deficiency it is an object of the invention to overcome prior art, there is provided it is sparse that a kind of joint motions are perceived
Exploded and restructing algorithm, can exactly capture the position that rotates motion and reconstruct and provide rotation fortune on a large scale
Dynamic triangular mesh attitude mode, can easily carry out attitude editor and with more preferable effect and computational efficiency.
To achieve the above object, technical scheme provided by the present invention is:The sparse part point that a kind of joint motions are perceived
Solution and restructing algorithm, comprise the following steps:
1) calculate triangle gridding in while length and while corresponding dihedral angle size
Giving a series of dynamic these models of triangular mesh sequence with different attitudes all has identical topology knot
Structure, first calculate respectively each bar on each triangle grid model while length and while corresponding to dihedral angle size (the two values
It is designated as LA);
2) residual error of each triangle grid model LA vectors is calculated
To step 1) on a triangle grid model obtaining each bar while corresponding LA according to while order be arranged to make up
One LA vector, calculates the LA vectors of each triangle grid model, is made the difference with the LA vectors with reference to triangle grid model respectively,
When corresponding LA vector residual errors;
3) LA bases are extracted
To step 2) in obtain each model LA vector residual error carry out sparse exploded, obtaining one group can catch
To the LA bases of local joint motion, each LA base controls one piece of region shape on triangle grid model;
4) linear hybrid LA bases obtain new LA vectors
To step 3) in the LA bases that can capture local joint motion that obtain, edit corresponding parameter, obtain new
LA vector residual errors, then make and calculate with the LA vectors with reference to triangle grid model, obtain new LA vectors;
5) connection mapping is recovered
To step 4) in the new LA vectors that obtain carry out the local coordinate value on summit on each tri patch and calculate, obtain
The local coordinate value on each summit in local coordinate system lower panel, and then recover the company of each tri patch on triangle grid model
Connect mapping;
6) local orthogonal marked price is estimated
To step 5) in the connection mapping relations of triangle surface that obtain, recycle and minimize marked price reconstruction of function and estimate
Calculate the local orthogonal frame of triangle grid model;
7) triangle gridding is reconstructed
To step 6) in the orthogonal marked price of office that obtains, each is estimated in triangle grid model using reconstruction of function is minimized
The position on summit, obtains the triangle grid model with new attitude after the meeting.
In step 1) in, it is described calculate in triangle gridding while length and while corresponding dihedral angle size (the two values
It is designated as LA), it is specific as follows:
Assuming thatIt is the triangle grid model data set with different attitudes for giving, NPTable
Registration according to concentrate triangle grid model number,Represent k-th triangle grid model.Each triangle grid model surface
Be by closely coupled triangular into.E and F represent line set and tri patch set in triangle grid model respectively, and
And the number on side is NE, the number of tri patch is NF.These triangle grid models have identical annexation, i.e., all of
Triangle grid model has identical line set and tri patch set.Travel through tri patch on summit when, we be with
Direction counterclockwise is carried out.
Assuming thatOn a line e=(i, j) ∈ E, share this edge tri patch for τ=(j, h, i) and υ=(g,
J, i), the letter in bracket, to constitute this tri patch the Vertex Labeling, is sequentially sequence counter-clockwise.When from this common edge
Vertex viGo to vjWhen, triangle surface τ be located at left-hand side, tri patch υ be located at right-hand side, so τ and υ be referred to as side (i,
J) left tri patch and right tri patch.When calculating the length of e, the Euclidean distance between two summits is used, i.e.,
lij=| | vj-vi||2.θ represents the angle of the normal of tri patch τ and υ outward direction, i.e. θ=acos (nτ·nυ), wherein nτWith
nυThe normal of direction outward direction on respectively tri patch τ and υ, then the corresponding dihedral angle of common edge be defined as
A LA values (l achieved aboveij,φij)。
For data setUpper all of triangle grid modelThe LA values on its each side are calculated, and according to side
Number order arrangement may be constructed a 2NEThe vector of rowThat is LA vectors.
In step 2) in, the LA vector residual errors of described each triangle grid model of calculating, specific method is:In step 1)
In obtain the LA vectors of each triangle grid model after, the LA vectors respectively with first grid model are poor, draw LA to
The residual error of amount, i.e. { Dk=Xk-X1, k=1,2 ..., Np}。
In step 3) in, described extraction LA bases, concrete mode is as follows:
Using step 2) in obtain LA vector residual error, construct a 2NE×NpMatrix D=(D1,...,DNp).Will
Matrix D resolves into D=C × W.Wherein Matrix C is a 2NEThe LA basic matrixs of × K, each row in matrix all represent a LA
Each row in base, i.e. matrix all control the shape in region in triangle grid model.Here K represents the individual of LA bases
Number, can be sized, if defining K value is that triangle grid model is roughly divided into K region as needed.And matrix W is one
Individual K × NPWeight matrix, the every a line in matrix represents a hybrid weight sequence.For the ease of the attitude to triangle gridding
Enter edlin, it is necessary to as much as possible keep LA bases locality, i.e., each LA base control region between mixing portion as far as possible
Ground is few, and needs normalization hybrid weight to eliminate due to scaling the influence to attitude.Come here by function (1) is minimized
Carry out matrix decomposition.
Wherein first energy termFor the error that constraint matrix is decomposed, and second energy
Quantifier is used to adjust the locality of LA bases in Matrix C, the i.e. magnitude range in the region of each LA base control.
Second energy term is specific as follows:
Wherein ΛikRepresent triangle grid model on i-th while kth bar as in shape base while midpoint between geodetic
Linear distance, is defined as follows:
Its detailed process is:K bars side is selected first on triangle grid model as the K center of shape base, referred to as in
Heart side, is represented with Edge here, edgekRepresent kth bar center edge.Use eiThe side that the numbering on triangle grid model is i is represented,
The then d in above formulaikRepresent eiAnd edgekResult after the geodesic curve range normalization at midpoint, normalization here is actually
The midpoint of triangle gridding top and the geodesic curve distance at each center edge midpoint are exactly calculated, maximum therein is taken out and is designated as
dmax, the value work that the geodesic curve distance for then calculating the side on triangle grid model is finally obtained in this maximum
It is the geodesic curve distance after normalization.Therefore there is dik≤1。rminAnd rmaxFor support radius span for [0,1), its tool
Body value sets according to demand.
In whole decomposable process, ci,kThe data of the i-th rows of representing matrix C kth row, wkjRepresenting matrix W row ks jth row
Value;And have max { wk,j, k=1 ..., K;J=1 ..., NPAll elements in }=1, i.e. matrix W, maximum is 1.In reality
In testing, the value of η is set to 200.
In step 4) in, described linear hybrid LA bases obtain new LA vectors, and its specific method is:In step 3) in obtain
After the LA basic matrixs for taking, the numerical value of any one row, obtains vectorial w in modification W matrixesk, then Matrix C again with wkIt is multiplied, obtains
One new LA vectors residual error, finally the LA addition of vectors with first triangle grid model (i.e. with reference to triangle grid model) be
The LA vectors of the triangle grid model with new attitude can be obtained.I.e. by calculatingNew LA can be obtained
Vector.
In step 5) in, described recovery connection mapping, its process mainly includes setting up a new local coordinate system, leads to
The LA values for crossing side are calculated two coordinates on each summit of tri patch that a this edge is common edge, and each is then calculated again
The orthogonal marked price of part under new local coordinate system of dough sheet, its company is recovered finally by the local orthogonal frame of two dough sheets
Details are provided below to connect mapping relations:
From step 4) in obtain LA vectors after, we can recover its annexation to every a line.Assuming that side e=
(i, j) is the common edge of left tri patch τ and right tri patch υ, and h, g are respectively the 3rd summit of left and right tri patch,
I.e. the left and right tri patch of common edge e=(i, j) is respectively τ=(j, h, i), υ=(g, j, i).A coordinate system is created first
(r, s, t), then calculates coordinate value of two each summits of tri patch under this coordinate system.Coordinate system is with vertex viMake
It is the origin of coordinates and vertex vjCoordinate value be (lij, 0,0), the t axles of coordinate system are same with the normal of tri patch υ outward directions
To.Because tri patch υ is the right tri patch of common edge e=(i, j), so vertex vgR coordinates be necessarily smaller than 0.So
V can be calculatedgCoordinate be respectively:
In order to calculate vertex vhCoordinate value, we first assume that vertex vhFlattening followed by triangular facet are rotated around common edge
Piece υ coplanar vertex vh', then what is be similar to can calculate vertex vh'Coordinate value be:
By vertex vh'φ is rotated around r reference axisijVertex v can be obtained after-πhReference axis.
After two reference axis on each summit of tri patch are obtained, local coordinate newly-built in above process is calculated
The orthogonal marked price of part under system.The detailed process of local orthogonal marked price is as follows:
By taking tri patch τ as an example, the orthogonal marked price of part that we are defined on this tri patch is ζτ=(uτ,vτ,nτ) its
InRepresent on tri patch τ from first vertex viPoint to second vertex vjUnit vector;nτIt is triangular facet
The outside unit normal of piece;vτ=nτ×uτ.Three above unit vector just constitutes ζτ3 × 3 matrix structure.
The orthogonal marked price ζ of part on tri patch υ can similarly be obtainedυ.SoAs by under other coordinate systems other
Coordinate value under coordinate system is transformed into the transition matrix represented under υ local coordinate systems.
Can just be counted after the orthogonal marked price of part for calculating two tri patch with side e=(i, j) as common edge
Calculate the connection mapping Q of this common edgeij。QijIt is defined as follows:
So far, the connection mapping on side e=(i, j) has been calculated, the connection that can recover all sides with similar mode is reflected
Penetrate.
Since when the editor of LA bases is carried out, the length of side for being LA values, i.e., directly changing in triangle gridding for directly changing
The dihedral angle size of degree triangle corresponding with side.This is possible to cause the amended length of side to be unsatisfactory for the triangle length of side
Formula, i.e., amended side is possible to that tri patch can not be constituted.Need to carry out specially treated in face of these situations:Work as lgj≥
lji+ligWhen, we unify the length l on sidejiAnd ligAmplify lgj/(lji+lig) times, thus there is lgj=lji+lig, at this time
Because the direction of side (i, j) is consistent with the direction of side (g, j), the direction of side (g, j) is (1,0,0);Similarly, l is worked asji≥
lgj+ligWhen, the direction of side (g, j) is (1,0,0), works as lig≥lgj+ljiWhen, the direction of side (g, j) is (- 1,0,0).
In step 6) in, described estimation locally orthogonal marked price, its main method is as follows:
Assuming thatIt is the set on the side in triangular mesh, each triangle is estimated by following minimum function (2)
The local orthogonal frame of dough sheet:
Wherein the left side is:
Represent connection Mapping Energy.
And the ζ in function (2)1WithThe orthogonal marked price of first tri patch that needs to estimate and given is represented respectively
First orthogonal marked price of tri patch, in order to ensure that the triangle gridding for reconstructing the rigid transformation for coming has uniqueness, we
Need the orthogonal marked price of part of given first tri patch;Subscript ' F ' represents Frobenius norms;QijFor step 5) in it is extensive
Multiple link mapping.
Step 7) in, described reconstruct triangle gridding is exactly mainly the position on each summit of triangle grid model for calculating new
Put, detailed process is as follows:
In step 6) in get local orthogonal marked price after, calculated by majorized function (3) each in triangle gridding
The position on individual summit.
εER({vi, i ∈ I })=εEV({vi,i∈I})+εVC({vi,i∈I}) (3)
Wherein the item on the right is used to constrain the position of given fixing point, and it is specially:
Wherein I*To give the numbering on fixed summit,It is the position of given fixed vertices.In order to ensure to reconstruct
The triangle grid model of the rigid transformation for coming has uniqueness, it would be desirable to which the position for fixing some points is constant, the number of fixing point
Amount is adjusted as needed.
First in function (3) is want the vertex position for restrained split-flow with step 6) in the orthogonal marked price of part that obtains
The deviation of the inside vector, specially:
Wherein egj=-lgjU, eji=lji(cos β u-sin β v), eig=lig(cos α u+sin α v), α is on tri patch
Side (g, j) and the angle of side (i, g), and β is the angle of side (g, j) and side (j, i);U in formula, v are represented by step 6) estimate
Corresponding tri patch on orthogonal marked price ζ=(u, v, n) of part on corresponding vector.
The institute of triangle grid model position a little is obtained with after majorized function (3), that is, is reconstructed after editing
Triangle grid model with new attitude.
The present invention compared with prior art, has the following advantages that and beneficial effect:
1st, more preferable locality, this is mainly reflected in the positioning to triangle gridding rigid deformation region.Relative to Huang etc.
What people proposed spatially carries out sparse exploded to obtain the method (referred to herein as DG methods) of shape base in deformation gradient, this
Invention is rotated in treatment has more preferable locality in the positioning of moving region.When grid model is clicked through around a joint
During row rigid rotation movement, because the corresponding LA values in the side for closing on joint area can just occur big change, and other are not being closed
Less, so the present invention captures the region of this change well very much, this also may be used the LA value changes on the side near nodal region
Preferably to generate corresponding LA bases so that more accurate by editing the new triangle grid model of generation, can more tackle around
The deformation of joint motions.And this point DG methods can not be processed well.
2nd, deform suitable for rotary motion on a large scale.Because the calculating of deformation gradient is separate, and deformation ladder
The angle of rotary motion that degree is adapted to can only be [0, π) between, so the triangular facet in it move triangle grid model
Around artis rotary motion, when the angle of rotation is more than π, DG methods will be unable to carry out with the deformation of the larger anglec of rotation piece
The extraction of shape base.And it is relative rotation angle between adjacent triangle that the LA values for being in the present invention are represented, and this revolves
Gyration is always less than 2 π, therefore, the present invention can catch from the triangle gridding sequence with rotary motion on a large scale well
Grasp the region (extracting LA yls) moved.
3rd, rotated suitable for the overall situation.Although global rotary motion is not very related with posture deforming, a sparse office
Part solution should also process such case.Due to being the sparse exploded to LA value residual errors in the present invention, and LA has
Locality, therefore, when triangle grid model occurs global rotary motion, for example, the whole grid model of horse is around an axle
Do global rotary motion and non-self attitude is when deforming, the triangle grid model of horse will not become substantially in attitude
Change, and LA vectors corresponding in triangle grid model will not also occur too big change, therefore rational model can be reconstructed.
And other method, such as DG methods, because deformation gradient (DG) depends on the overall situation rotationally-varying, therefore there is global rotation in model
During transformationization, the change of DG is very big, and this will cause irrational deformation.
4th, with good robustness.This mainly includes that the annexation on a summit in triangle grid model has Lu Bangxing
There is robustness with the selection to the corresponding area size of LA bases.The annexation on present invention summit in tri patch is processed
When, a connection mapping algorithm is devised, this algorithm can well keep the annexation between summit.Therefore, in reconstruct
During triangle grid model attitude after editing on a large scale, there is good robustness in the annexation on summit.At this
In the experimental result of invention, the selection to the corresponding area size of LA bases can well navigate to the region of deformation, and have
Good locality.
Brief description of the drawings
Fig. 1 is algorithm flow chart of the invention.
Fig. 2 is the structure of tri patch in intermediate cam grid model of the present invention, and part therein is orthogonal to mark the price and two faces
Angle.
Fig. 3 is the present invention to doing after local rotary motion triangle grid model carries out LA base extractions, the area that LA bases are controlled
Domain visualizes schematic diagram.Figure (a) therein is the result to carrying out the iron model extraction LA bases of flexural deformation, the LA bases of extraction
The position of bending has been navigated to well, and other need not then carry out extra treatment for the rigid position of deformation;Figure (b) exists
The triangle grid model that the horse of deformation is rotated to afterbody extracts the result of LA bases, and the LA bases of extraction have navigated to tail base
The region for rotating;Figure (c) is the result that LA base extractions are carried out to the robot for moving, and the result of extraction has navigated to hair
The ancon on hand of raw rotary motion;Figure (d) is the result that the facial model changed to facial expression carries out LA base extractions,
The result of extraction also can well navigate to the position for deforming.
Fig. 4 is structural representation of the present invention in the recovery process for being attached relation, the local coordinate that figure (a) is set up
The distribution situation schematic diagram on each summit in system and tri patch, figure (b), (c), (d) are that LA values are passed through after editor, three sides
Length violation closes the situation schematic diagram of triangle inequality, wherein figure (b) represents lgj≥lji+ligSituation, figure (c) represent lji
≥lgj+ligSituation, figure (d) represent lig≥lgj+ljiSituation.
Fig. 5 is example of the present invention when the apex coordinate of tri patch is estimated.
Fig. 6 is the result and the edited result to attitude that the present invention is extracted to motion sequence LA bases.
Fig. 7 is extraction result and the edited result of expression of the present invention to face's triangle grid model LA bases
Result and the result of DG method of Fig. 8 present invention to global rotary motion.On the right of wherein two
In row, what each row were represented is result to about similar region shape base, first result of behavior DG methods, the
Two behaviors result of the present invention.
Specific embodiment
With reference to specific embodiment, the invention will be further described.
As shown in figure 1, the sparse exploded that perceives of joint motions described in the present embodiment and restructing algorithm, including it is following
Step:
1) calculate triangle gridding in while length and while corresponding dihedral angle size (the two values are designated as LA)
Giving a series of dynamic these models of triangular mesh sequence with different attitudes all has identical topology knot
Structure, first calculate respectively each bar on each triangle grid model while length and while corresponding to dihedral angle size;
2) residual error of each triangle grid model LA vectors is calculated
To step 1) on a triangle grid model obtaining each bar while corresponding LA according to while order be arranged to make up
One LA vector, calculates the LA vectors of each triangle grid model, is made the difference with the LA vectors with reference to triangle grid model respectively,
When corresponding LA vector residual errors;
3) LA bases are extracted
To step 2) in obtain each model LA vector residual error carry out sparse exploded, obtaining one group can catch
To the LA bases of local joint motion, each LA base controls one piece of region shape on triangle grid model;
4) linear hybrid LA bases obtain new LA vectors
To step 3) in the LA bases that can capture local joint motion that obtain, edit corresponding parameter, obtain new
LA vector residual errors, then make and calculate with the LA vectors with reference to triangle grid model, obtain new LA vectors;
5) connection mapping is recovered
To step 4) in the new LA vectors that obtain carry out the local coordinate value on summit on each tri patch and calculate, obtain
The local coordinate value on each summit in local coordinate system lower panel, and then recover the company of each tri patch on triangle grid model
Connect mapping;
6) local orthogonal marked price is estimated
To step 5) in the connection mapping relations of triangle surface that obtain, recycle and minimize marked price reconstruction of function and estimate
Calculate the local orthogonal frame of triangle grid model;
7) triangle gridding is reconstructed
To step 6) in the orthogonal marked price of office that obtains, each is estimated in triangle grid model using reconstruction of function is minimized
The position on summit, obtains the triangle grid model with new attitude after the meeting.
In step 1) in, it is described calculate in triangle gridding while length and while corresponding dihedral angle size (the two values
It is designated as LA), it is specific as follows:Assuming thatIt is the triangle grid model number with different attitudes for giving
According to collection, NPThe number of data set intermediate cam grid model is represented,Represent k-th triangle grid model.Each triangle gridding mould
Type surface be all by closely coupled triangular into.E and F represent line set and tri patch in triangle grid model respectively
Set, and the number on side is NE, the number of tri patch is NF.These triangle grid models have identical annexation, i.e.,
All of triangle grid model has identical line set and tri patch set.Travel through tri patch on summit when, I
Be to be carried out with direction counterclockwise.
Assuming thatOn a line e=(i, j) ∈ E, share this edge tri patch for τ=(j, h, i) and υ=(g,
J, i), the letter in bracket, to constitute this tri patch the Vertex Labeling, is sequentially sequence counter-clockwise, as shown in Figure 2.When from this
The vertex v of bar common edgeiGo to vjWhen, triangle surface τ is located at left-hand side, and tri patch υ is located at right-hand side, so τ and υ points
Have another name called is the left tri patch and right tri patch of side (i, j).When calculating the length of e, use the Europe between two summits several
Reed distance, i.e. lij=| | vj-vi||2.θ represents the angle of the normal of tri patch τ and υ outward direction, i.e. θ=acos (nτ·
nυ), wherein nτAnd nυThe normal of direction outward direction on respectively tri patch τ and υ, the then corresponding dihedral angle definition of common edge
For
A LA values (l achieved aboveij,φij)。
For data setUpper all of triangle grid modelThe LA values on its each side are calculated, and according to side
Number order arrangement may be constructed a vector for 2NE rowThat is LA vectors.
In step 2) in, the LA vector value residual errors of described each triangle grid model of calculating, specific method is:In step
1) obtained in each triangle grid model LA vector values constitute vector after, respectively with first LA value of grid model to
Amount makes the difference, and draws the residual vector of LA vector values, i.e.,.
In step 3) in, described extraction LA bases, concrete mode is as follows:
Using step 2) in obtain LA vector residual error, construct a 2NE×NpMatrix D=(D1,...,DNp).Will
Matrix D resolves into D=C × W.Wherein Matrix C is a 2NEThe LA basic matrixs of × K, each row in matrix all represent LA shapes
Each row bar in base, i.e. matrix all controls the shape in region in triangle grid model.Here K represents LA shapes
The number of base, can be sized as needed, if defining K value is that triangle grid model is roughly divided into K region.And matrix
It is the weight matrix of, the every a line in matrix represents a hybrid weight sequence.For the ease of the attitude to triangle gridding
Enter edlin, it is necessary to keep the locality of LA shape bases as much as possible, i.e. mixing unit between the region of each LA shapes base control
Divide few as much as possible, and need normalization hybrid weight to eliminate due to scaling the influence to attitude.Here by minimum
Function (1) carries out matrix decomposition.
Wherein first energy termFor the error that constraint matrix is decomposed, and second energy
Quantifier is used to adjust the locality of shape base in Matrix C, the i.e. magnitude range of each shape base.Second energy term is specific such as
Under:
Wherein ΛikRepresent triangle grid model on i-th while kth bar as in shape base while midpoint between geodetic
Linear distance, is defined as follows:
Its detailed process is:K bars side is selected first on triangle grid model as the K center of shape base, referred to as in
Heart side, is represented with Edge here, edgekRepresent kth bar center edge.Use eiThe side that the numbering on triangle grid model is i is represented,
The then d in above formulaikRepresent eiAnd edgekResult after the geodesic curve range normalization at midpoint, normalization here is actually
The midpoint of triangle gridding top and the geodesic curve distance at each center edge midpoint are exactly calculated, maximum therein is taken out and is designated as
dmax, the value work that the geodesic curve distance for then calculating the side on triangle grid model is finally obtained in this maximum
It is the geodesic curve distance after normalization.Therefore there is dik≤1。rminAnd rmaxFor support radius span for [0,1), its tool
Body value sets according to demand.
In whole decomposable process, ci,kThe data of the i-th rows of representing matrix C kth row, wkjRepresenting matrix W row ks jth row
Value;And have max { wk,j, k=1 ..., K;J=1 ..., NPAll elements in }=1, i.e. matrix W, maximum is 1.In reality
In testing, the value of η is set to 200.The visualization result of the shape base obtained after Its Sparse Decomposition is as shown in Figure 3.
In step 4) in, described linear hybrid LA bases obtain new LA vector values, and its specific method is:In step 3) in
After the LA shape bases matrix of acquisition, modification W matrixes in any one row numerical value, obtain vector, then Matrix C again with phase
Multiply, obtain a new LA vector value residual vector, finally with first triangle grid model (, i.e., with reference to triangle grid model)
LA vector values be added can obtain with new attitude triangle grid model LA vector values.I.e. by calculating and can obtain
New LA vector values.
In step 5) in, described recovery connection mapping, its process mainly includes setting up a new local coordinate system, leads to
The LA values for crossing side are calculated two coordinates on each summit of tri patch that a this edge is common edge, and each is then calculated again
The orthogonal marked price of part under new local coordinate system of dough sheet, its company is recovered finally by the local orthogonal frame of two dough sheets
Details are provided below to connect mapping relations:
From step 4) in obtain LA vectors after, we can recover its annexation to every a line.Assuming that side e=
(i, j) is the common edge of left tri patch τ and right tri patch υ, and h, g are respectively the 3rd summit of left and right tri patch,
I.e. the left and right tri patch of common edge e=(i, j) is respectively τ=(j, h, i), and υ=(g, j, i) is as shown in Figure 4.One is created first
Individual coordinate system (r, s, t), then calculates coordinate value of two each summits of tri patch under this coordinate system.Coordinate system with
Vertex viAs the origin of coordinates and vertex vjCoordinate value be (lij, 0,0), t axles and the tri patch υ outward directions of coordinate system
Normal is in the same direction.Because tri patch υ is the right tri patch of common edge e=(i, j), so vertex vgR coordinates be necessarily smaller than
0.So v can be calculatedgCoordinate be respectively:
In order to calculate vertex vhCoordinate value, we first assume that vertex vhFlattening followed by triangular facet are rotated around common edge
Piece υ coplanar vertex vh', then what is be similar to can calculate vertex vh'Coordinate value be:
By vertex vh'φ is rotated around r reference axisijVertex v can be obtained after-πhReference axis.
After two reference axis on each summit of tri patch are obtained, local coordinate newly-built in above process is calculated
The orthogonal marked price of part under system.The detailed process of local orthogonal marked price is as follows:
By taking tri patch τ as an example, the orthogonal marked price of part that we are defined on this tri patch is ζτ=(uτ,vτ,nτ) its
InRepresent on tri patch τ from first vertex viPoint to second vertex vjUnit vector;nτIt is triangle
The outside unit normal of dough sheet;vτ=nτ×uτ.Three above unit vector just constitutes ζτ3 × 3 matrix structure.
The orthogonal marked price ζ of part on tri patch υ can similarly be obtainedυ.SoAs by under other coordinate systems other
Coordinate value under coordinate system is transformed into the transition matrix represented under υ local coordinate systems.The orthogonal marked price of part on tri patch is such as
Shown in Fig. 2.
Can just be counted after the orthogonal marked price of part for calculating two tri patch with side e=(i, j) as common edge
Calculate the connection mapping Q of this common edgeij。QijIt is defined as follows:
So far, the connection mapping on side e=(i, j) has been calculated, the connection that can recover all sides with similar mode is reflected
Penetrate.
Since when the editor of LA bases is carried out, the length of side for being LA values, i.e., directly changing in triangle gridding for directly changing
The dihedral angle size of degree triangle corresponding with side.This is possible to cause the amended length of side to be unsatisfactory for the triangle length of side
Formula, i.e., amended side is possible to that tri patch can not be constituted.Need to carry out specially treated in face of these situations:Work as lgj≥
lji+ligWhen (Fig. 4 (b) shown in), we unify the length l on sidejiAnd ligAmplify lgj/(lji+lig) times, thus there is lgj=
lji+lig, at this time because the direction on side (i, j) is consistent with the direction of side (g, j), so the direction of side (g, j) is (1,0,0);
Similarly, l is worked asji≥lgj+ligWhen (Fig. 4 (c) shown in), the direction of side (g, j) is (1,0,0), works as lig≥lgj+ljiWhen (Fig. 4 (d)
It is shown), the direction of side (g, j) is (- 1,0,0).
In step 6) in, described estimation locally orthogonal marked price, its main method is as follows:
Assuming thatIt is the set on the side in triangular mesh, each triangle is estimated by following minimum function (2)
The local orthogonal frame of dough sheet:
Wherein the left side is:
Represent connection Mapping Energy.
And the ζ in function (2)1WithThe orthogonal marked price of first tri patch that needs to estimate and given is represented respectively
First orthogonal marked price of tri patch, in order to ensure that the triangle gridding for reconstructing the rigid transformation for coming has uniqueness, we
Need the orthogonal marked price of part of given first tri patch;Subscript ' F ' represents Frobenius norms;QijFor step 5) in it is extensive
Multiple link mapping.
Step 7) in, described reconstruct triangle gridding is exactly mainly the position on each summit of triangle grid model for calculating new
Put, detailed process is as follows:
In step 6) in get local orthogonal marked price after, calculated by majorized function (3) each in triangle gridding
The position on individual summit.
εER({vi, i ∈ I })=εEV({vi,i∈I})+εEC({vi,i∈I}) (3)
Wherein the item on the right is used to constrain the position of given fixing point, and it is specially:
Wherein I*To give the numbering on fixed summit,It is the position of given fixed vertices.Come to ensure to reconstruct
The triangle grid model of rigid transformation there is uniqueness, it would be desirable to the position for fixing some points is constant, the quantity of fixing point
It is adjusted as needed.
First in function (3) is want the vertex position for restrained split-flow with step 6) in the orthogonal marked price of part that obtains
The deviation of the inside vector, specially:
Wherein egj=-lgjU, eji=lji(cos β u-sin β v), eig=lig(cos α u+sin α v), α is on tri patch
Side (g, j) and the angle of side (i, g), and β is the angle (as shown in Figure 5) of side (g, j) and side (j, i);U in formula, v represent by
Step 6) estimate corresponding tri patch on orthogonal marked price ζ=(u, v, n) of part on corresponding vector.
The institute of triangle grid model position a little is obtained with after majorized function (3), that is, is reconstructed after editing
Triangle grid model with new attitude.
The present invention proves its feasibility by experiment, can be widely used in various models.Fig. 6 shows the present invention to LA
Base positioning and the result to being rotated editor around node.Wherein figure (a) is editor's two LA bases therein, is passed through
The hybrid weight parameter for changing the LA bases for controlling the two regions has reconfigured a triangle gridding mould with new attitude
Type, intuitively say be exactly will control left side model upper part of the body region LA bases hybrid weight parameter assignment with control mid-module
The hybrid weight parameter combination of the LA bases in lower part of the body region is got up, and constitutes a complete hybrid weight parameter, is finally reconstructed
The right has the triangle grid model of new attitude.Figure (b) is also similar result.And (c) is then by changing hybrid weight
The triangle grid model with new attitude for coming is reconstructed after parameter, these attitudes are concentrated in original triangle grid data
It is non-existent.Fig. 7 shows the edit effect to face triangle grid model, wherein from first triangle gridding of left side number
Model display is the LA bases for controlling left eye region extracted by the present invention, and second shows control right eye
The LA bases in region, the 3rd shows the LA bases for controlling left corners of the mouth region, and rightmost shows by that will control these three
The hybrid weight parameter group of LA bases is combined together the three dimensional network for constructing and being reconstructed after a complete hybrid weight parameter
Lattice model.The comparing phase of the shape with first model of the three-dimensional grid model left eye for coming is reconstructed as we can see from the figure
Seemingly, the shape of right eye is similar with the comparing of second model, and the shape of the left side corners of the mouth is with the 3rd comparing phase of model
Seemingly.Fig. 8 be with the triangle grid model of horse as example illustrate the present invention to global rotary motion treatment legitimate result and
The result of DG methods treatment.The model of horse carries out running motion around a circle in experiment.The left side is not carry out originally in figure
The model of deformation, the row of the right two shows the result obtained after global rotation editor is carried out to identical base, and first is
The deformation of head, second is the deformation of foreleg, and the 3rd deformation of back leg, experiment shows, the present invention is in the global rotation for the treatment of
During motion, the result for the treatment of is more reasonable.
The examples of implementation of the above are only the preferred embodiments of the invention, not limit implementation model of the invention with this
Enclose, therefore the change that all shapes according to the present invention, principle are made, all should cover within the scope of the present invention.
Claims (2)
1. a kind of joint motions are perceived sparse exploded and restructing algorithm, it is characterised in that comprise the following steps:
1) calculate triangle gridding in while length and while corresponding dihedral angle size
Giving a series of dynamic these models of triangular mesh sequence with different attitudes all has identical topological structure, first
Calculate respectively each bar on each triangle grid model while length and while corresponding to dihedral angle size, the two values are designated as
LA;
2) residual error of each triangle grid model LA vector values is calculated
To step 1) on a triangle grid model obtaining each bar while corresponding LA according to while order be arranged to make up one
LA vector, calculate each triangle grid model LA vector, the LA values of each triangle grid model, respectively with refer to the triangulation network
The LA vector values of lattice model make the difference, when corresponding LA vector values residual error;
3) shape LA bases are extracted
To step 2) in the LA vector values residual error of each model that obtains carry out sparse exploded, obtaining one group can capture
The LA bases of local joint motion, each LA base controls one piece of region shape on triangle grid model;
4) linear hybrid LA bases obtain new LA vector values
To step 3) in the LA bases that can capture local joint motion that obtain, edit corresponding parameter, obtain new LA to
Value residual error, then make and calculate with the LA vector values with reference to triangle grid model, obtain new LA vector values;
5) connection mapping is recovered
To step 4) in the new LA vector values that obtain carry out the local coordinate value on summit on each tri patch and calculate, acquisition office
The local coordinate value on each summit in portion's coordinate system lower panel, and then recover the connection of each tri patch on triangle grid model
Mapping relations;
6) local orthogonal marked price is estimated
To step 5) in the connection mapping relations of triangle surface that obtain, recycle and minimize marked price reconstruction of function and estimate
The local orthogonal frame of triangle grid model;
7) triangle gridding is reconstructed
To step 6) in the orthogonal price mark display of office that obtains, each top in triangle grid model is estimated using reconstruction of function is minimized
The position of point, obtains the triangle grid model with new attitude after the meeting.
2. a kind of joint motions according to belonging to claim 1 are perceived sparse exploded and restructing algorithm, it is characterised in that
In step 1) in, it is described calculate in triangle gridding while length and while corresponding dihedral angle size, it is specific as follows:
Assuming thatIt is the triangle grid model data set with different attitudes for giving, NPRepresent number
According to the number for concentrating triangle grid model,Represent k-th triangle grid model, each triangle grid model surface be by
Closely coupled triangular is into E and F represents line set and tri patch set in triangle grid model, and side respectively
Number be NE, the number of tri patch is NF, these triangle grid models have identical annexation, i.e., all of triangle
Grid model has identical line set and tri patch set;Travel through tri patch on summit when, we are with the inverse time
The direction of pin is carried out;
Assuming thatOn a line e=(i, j) ∈ E, share the tri patch of this edge for τ=(j, h, i) and
υ=(g, j, i), the letter in bracket, to constitute this tri patch the Vertex Labeling, is sequentially sequence counter-clockwise;When from this
The vertex v of bar common edgeiGo to vjWhen, triangle surface τ is located at left-hand side, and tri patch υ is located at right-hand side, so τ and υ points
Have another name called is the left tri patch and right tri patch of side (i, j);When calculating the length of e, use the Europe between two summits several
Reed distance, i.e. lij=| | vj-vi||2;θ represents the angle of the normal of tri patch τ and υ outward direction, i.e. θ=a cos
(nτ·nυ), wherein nτAnd nυThe normal of direction outward direction on respectively tri patch τ and υ, the then corresponding dihedral angle of common edge
It is defined as
A LA values (l achieved aboveij,φij);
For data setUpper all of triangle grid modelThe LA values on its each side are calculated, and according to the numbering on side
Order arrangement may be constructed a 2NEThe vector of rowThat is LA vectors;
In step 2) in, the LA vector residual errors of described each triangle grid model of calculating, specific method is:In step 1) in obtain
After obtaining the vector of LA vector compositions of each triangle grid model, the LA vectors respectively with first grid model are poor, obtain
Go out the residual error of LA vectors, i.e. { Dk=Xk-X1, k=1,2 ..., Np};
In step 3) in, described extraction LA bases, concrete mode is as follows:
Using step 2) in obtain LA vector residual error, construct a 2NE×NpMatrix D=(D1,...,DNp), by matrix D
D=C × W is resolved into, wherein Matrix C is a 2NEThe LA basic matrixs of × K, each row in matrix all represent a LA base, i.e.,
Each row in matrix all control the shape in region in triangle grid model, and K here represents the number of LA bases, energy
It is enough to be sized as needed, if defining K value is that triangle grid model is roughly divided into K region;And matrix W be a K ×
NPWeight matrix, the every a line in matrix represents a hybrid weight sequence;Carried out for the ease of the attitude to triangle gridding
Editor, it is necessary to keep the locality of LA bases, i.e., mixing portion is few as much as possible between each LA base, and needs are returned as much as possible
One changes hybrid weight to eliminate due to scaling the influence to attitude;Matrix decomposition is carried out here by function (1) is minimized:
Wherein, first energy termFor the error that constraint matrix is decomposed, and second energy term
The area size scope of the locality for adjusting LA bases in Matrix C, i.e. each LA base control;
Second energy term is specific as follows:
Wherein, ΛikRepresent triangle grid model on i-th while kth bar as in shape base while midpoint between geodesic curve
Distance, is defined as follows:
Its detailed process is:K bars side is selected first on triangle grid model as the K center of shape base, referred to as center edge,
Here represented with Edge, edgekRepresent kth bar center edge;Use eiThe side that the numbering on triangle grid model is i is represented, then above formula
In dikRepresent eiAnd edgekResult after the geodesic curve range normalization at midpoint, normalization here is actually meter
The midpoint of triangle gridding top and the geodesic curve distance at each center edge midpoint are calculated, maximum therein is taken out and is designated as dmax, then
The geodesic curve that side on triangle grid model is calculated is apart from the value finally obtained in this maximum as normalization
Geodesic curve distance afterwards, therefore have dik≤ 1, rminAnd rmaxFor support radius span for [0,1), its occurrence according to
Demand sets;
In whole decomposable process, ci,kThe data of the i-th rows of representing matrix C kth row, wkjThe value of representing matrix W row ks jth row;
And have max { wk,j, k=1 ..., K;J=1 ..., NPAll elements in }=1, i.e. matrix W, maximum is 1;
In step 4) in, described linear hybrid LA bases obtain new LA vectors, and its specific method is:In step 3) middle acquisition
After LA basic matrixs, the numerical value of any one row, obtains vectorial w in modification W matrixesk, then Matrix C again with wkIt is multiplied, obtains one
New LA vector residual errors, finally can obtain the triangle with new attitude with first LA addition of vectors of triangle grid model
The LA vectors of grid model, i.e., by calculatingNew LA vectors can be obtained;
In step 5) in, described recovery connection mapping, its process mainly includes setting up a new local coordinate system, by side
LA values be calculated a this edge be common edge two coordinates on each summit of tri patch, each dough sheet is then calculated again
The orthogonal marked price of the part under new local coordinate system, recover its connection finally by the local orthogonal frame of two dough sheets and reflect
Details are provided below to penetrate relation:
From step 4) in obtain LA vectors after, its annexation is recovered to every a line, it is assumed that side e=(i, j) is left three
The common edge of edged surface piece τ and right tri patch υ, and h, g are respectively the 3rd summit of left and right tri patch, i.e. common edge e=
The left and right tri patch of (i, j) is respectively τ=(j, h, i), υ=(g, j, i);A coordinate system (r, s, t) is created first, then
Calculate coordinate value of two each summits of tri patch under this coordinate system;Coordinate system is with vertex viAs the origin of coordinates and
Vertex vjCoordinate value be (lij, 0,0), the t axles of coordinate system are in the same direction with the normal of tri patch υ outward directions;Due to triangular facet
Piece υ is the right tri patch of common edge e=(i, j), so vertex vgR coordinates be necessarily smaller than 0, so calculate vgCoordinate point
It is not:
In order to calculate vertex vhCoordinate value, assume initially that vertex vhFlattening is rotated around common edge followed by tri patch υ is coplanar
Vertex vh', then the calculating vertex v being similar toh'Coordinate value be:
By vertex vh'φ is rotated around r reference axisijVertex v can be obtained after-πhReference axis;
After two reference axis on each summit of tri patch are obtained, calculate under local coordinate system newly-built in above process
The orthogonal marked price of part;The detailed process of local orthogonal marked price is as follows:
By taking tri patch τ as an example, it is ζ to define the orthogonal marked price of the part on this tri patchτ=(uτ,vτ,nτ) whereinRepresent on tri patch τ from first vertex viPoint to second vertex vjUnit vector;nτIt is tri patch
Outside unit normal;vτ=nτ×uτ;Three above unit vector just constitutes ζτ3 × 3 matrix structure;
The orthogonal marked price ζ of part on tri patch υ can similarly be obtainedυ, thenAs by other coordinates under other coordinate systems
Coordinate value under system is transformed into the transition matrix represented under υ local coordinate systems;
This can be just calculated after the orthogonal marked price of part for calculating two tri patch with side e=(i, j) as common edge
The connection mapping Q of common edgeij, QijIt is defined as follows:
So far, the connection mapping on side e=(i, j) has been calculated, the connection mapping on all sides can have just been recovered with similar fashion;
Since when the editor of shape base is carried out, the edge lengths for being LA values, i.e., directly changing in triangle gridding for directly changing
The dihedral angle size of triangle corresponding with side, this is possible to cause the amended length of side to be unsatisfactory for the triangle length of side
Formula, i.e., amended side is possible to that tri patch can not be constituted, and needs to carry out specially treated in face of these situations:Work as lgj≥
lji+ligWhen, the unified length l by sidejiAnd ligAmplify lgj/(lji+lig) times, thus there is lgj=lji+lig, at this time because
The direction on side (i, j) is consistent with the direction of side (g, j), so the direction of side (g, j) is (1,0,0);Similarly, l is worked asji≥lgj+lig
When, the direction of side (g, j) is (1,0,0), works as lig≥lgj+ljiWhen, the direction of side (g, j) is (- 1,0,0);
In step 6) in, described estimation locally orthogonal marked price, its main method is as follows:
Assuming thatIt is the set on the side in triangular mesh, each tri patch is estimated by following minimum function (2)
Local orthogonal frame:
Wherein the left side is:
Represent connection Mapping Energy;
And the ζ in function (2)1WithThe orthogonal marked price and given first of first tri patch for needing to estimate is represented respectively
The orthogonal marked price of individual tri patch, in order to ensure that the triangle gridding for reconstructing the rigid transformation for coming has uniqueness, it is necessary to given
The orthogonal marked price of part of first tri patch;Subscript ' F ' represents Frobenius norms;QijFor step 5) the middle link for recovering
Mapping;
Step 7) in, described reconstruct triangle gridding is exactly mainly the position on each summit of triangle grid model for calculating new, tool
Body process is as follows:
In step 6) in get local orthogonal marked price after, each top in triangle gridding is calculated by majorized function (3)
The position of point;
εER({vi, i ∈ I })=εEV({vi,i∈I})+εVC({vi,i∈I}) (3)
Wherein, the item on the right is used to constrain the position of given fixing point, and it is specially:
Wherein, I*To give the numbering on fixed summit,It is the position of given fixed vertices, in order to ensure to reconstruct what is come
The triangle grid model of rigid transformation there is uniqueness, it is necessary to fix some point position it is constant, the quantity of fixing point is according to need
It is adjusted;
First in function (3) is want the vertex position for restrained split-flow with step 6) in the orthogonal marked price the inside of the part that obtains
The deviation of vector, specially:
Wherein, egj=-lgjU, eji=lji(cos β u-sin β v), eig=lig(cos α u+sin α v), α is tri patch top
The angle of (g, j) and side (i, g), and β is the angle of side (g, j) and side (j, i);U in formula, v are represented by step 6) estimate
Corresponding vector on orthogonal marked price ζ=(u, v, n) of part on correspondence tri patch;
The institute of triangle grid model position a little is just obtained in that after majorized function (3), that is, reconstruct has after editing
The triangle grid model of new attitude.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611125996.4A CN106683169B (en) | 2016-12-09 | 2016-12-09 | A kind of sparse exploded and restructing algorithm of joint motions perception |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611125996.4A CN106683169B (en) | 2016-12-09 | 2016-12-09 | A kind of sparse exploded and restructing algorithm of joint motions perception |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106683169A true CN106683169A (en) | 2017-05-17 |
CN106683169B CN106683169B (en) | 2019-10-18 |
Family
ID=58868507
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611125996.4A Active CN106683169B (en) | 2016-12-09 | 2016-12-09 | A kind of sparse exploded and restructing algorithm of joint motions perception |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106683169B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107272673A (en) * | 2017-05-18 | 2017-10-20 | 中山大学 | SLAM rear ends track optimizing method based on pose chain model |
CN109903389A (en) * | 2019-01-21 | 2019-06-18 | 江苏艾佳家居用品有限公司 | A kind of three-dimensional model gridding processing method and system |
CN110349268A (en) * | 2019-06-11 | 2019-10-18 | 华南理工大学 | A kind of reconstructing method of 3 D human body posture, expression and gesture |
CN110363833A (en) * | 2019-06-11 | 2019-10-22 | 华南理工大学 | A kind of complete human body sport parameter representation method based on local rarefaction representation |
CN111462306A (en) * | 2020-03-17 | 2020-07-28 | 华南理工大学 | Three-dimensional hair parametric model method based on volume vector field sparse localization decomposition |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101071514A (en) * | 2006-05-12 | 2007-11-14 | 中国科学院自动化研究所 | Method for directly transferring three-dimensional model attitude |
CN103630899A (en) * | 2013-03-29 | 2014-03-12 | 中国科学院电子学研究所 | Method for high-resolution radar compressed sensing imaging of moving object on ground |
CN104408767A (en) * | 2014-11-20 | 2015-03-11 | 浙江大学 | Method for building sparse consistent three-dimensional human face mesh deformation model |
CN105118093A (en) * | 2015-09-02 | 2015-12-02 | 宁波工程学院 | Parametric surface reconstruction method for false tooth discrete data |
CN106157372A (en) * | 2016-07-25 | 2016-11-23 | 深圳市唯特视科技有限公司 | A kind of 3D face grid reconstruction method based on video image |
-
2016
- 2016-12-09 CN CN201611125996.4A patent/CN106683169B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101071514A (en) * | 2006-05-12 | 2007-11-14 | 中国科学院自动化研究所 | Method for directly transferring three-dimensional model attitude |
CN103630899A (en) * | 2013-03-29 | 2014-03-12 | 中国科学院电子学研究所 | Method for high-resolution radar compressed sensing imaging of moving object on ground |
CN104408767A (en) * | 2014-11-20 | 2015-03-11 | 浙江大学 | Method for building sparse consistent three-dimensional human face mesh deformation model |
CN105118093A (en) * | 2015-09-02 | 2015-12-02 | 宁波工程学院 | Parametric surface reconstruction method for false tooth discrete data |
CN106157372A (en) * | 2016-07-25 | 2016-11-23 | 深圳市唯特视科技有限公司 | A kind of 3D face grid reconstruction method based on video image |
Non-Patent Citations (4)
Title |
---|
FEDERICA BOGO 等: "Detailed Full-Body Reconstructions of Moving People from Monocular RGB-D Sequences", 《2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
HUI HUANG 等: "L-1-Medial Skeleton of Point Cloud", 《ACM TRANSACTIONS ON GRAPHICS》 * |
孙明明: "基于自适应稀疏表示的医学图像重构算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
苏琪: "基于上下文感知的三维交互建模", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107272673A (en) * | 2017-05-18 | 2017-10-20 | 中山大学 | SLAM rear ends track optimizing method based on pose chain model |
CN109903389A (en) * | 2019-01-21 | 2019-06-18 | 江苏艾佳家居用品有限公司 | A kind of three-dimensional model gridding processing method and system |
CN109903389B (en) * | 2019-01-21 | 2022-07-08 | 江苏艾佳家居用品有限公司 | Three-dimensional model grid processing method and system |
CN110349268A (en) * | 2019-06-11 | 2019-10-18 | 华南理工大学 | A kind of reconstructing method of 3 D human body posture, expression and gesture |
CN110363833A (en) * | 2019-06-11 | 2019-10-22 | 华南理工大学 | A kind of complete human body sport parameter representation method based on local rarefaction representation |
CN110363833B (en) * | 2019-06-11 | 2021-03-30 | 华南理工大学 | Complete human motion parameterization representation method based on local sparse representation |
CN111462306A (en) * | 2020-03-17 | 2020-07-28 | 华南理工大学 | Three-dimensional hair parametric model method based on volume vector field sparse localization decomposition |
CN111462306B (en) * | 2020-03-17 | 2022-04-22 | 华南理工大学 | Three-dimensional hair parametric model method based on volume vector field sparse localization decomposition |
Also Published As
Publication number | Publication date |
---|---|
CN106683169B (en) | 2019-10-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106683169A (en) | Sparse local decomposition and restructuring algorithm of joint motion sensing | |
CN106945041B (en) | A kind of repetitive motion planning method for redundant manipulator | |
Solomon et al. | Flexible developable surfaces | |
Yang et al. | Shape space exploration of constrained meshes | |
CN104851126B (en) | Threedimensional model dividing method and device based on generalized cylinder | |
CN100543775C (en) | The method of following the tracks of based on the 3 d human motion of many orders camera | |
CN109544677A (en) | Indoor scene main structure method for reconstructing and system based on depth image key frame | |
CN106971414A (en) | A kind of three-dimensional animation generation method based on deep-cycle neural network algorithm | |
CN111553968A (en) | Method for reconstructing animation by three-dimensional human body | |
CN109544666A (en) | A kind of full automatic model deformation transmission method and system | |
CN106991722A (en) | It is a kind of that the geometric mesh model deformation method of convex closure is influenceed based on oblique ellipsoid domain | |
CN105225272B (en) | A kind of tri-dimensional entity modelling method based on the reconstruct of more contour line triangulation networks | |
CN109410335A (en) | A kind of efficient Mesh Fusion method based on reusable Laplacian Matrix | |
CN106826813A (en) | The robot stabilized motion control method of the foot point assessment that falls with multiple constraint is modeled based on curve matching | |
CN111028335B (en) | Point cloud data block surface patch reconstruction method based on deep learning | |
CN103544733B (en) | The three-dimensional human head triangular mesh model method for building up analyzed based on Statistical Shape | |
CN101877146A (en) | Method for extending three-dimensional face database | |
CN101356549B (en) | Defrobulated angles for character joint representation | |
CN102945561A (en) | Motion synthesizing and editing method based on motion capture data in computer bone animation | |
CN110189413A (en) | A kind of method and system generating clothes distorted pattern | |
CN106204635A (en) | Based on L0the human body successive frame skeleton optimization method minimized | |
Xin et al. | Efficiently computing geodesic offsets on triangle meshes by the extended Xin–Wang algorithm | |
Wang et al. | An energy-driven motion planning method for two distant postures | |
CN105957154B (en) | A kind of the threedimensional model edit methods and system of data-driven | |
CN105957138B (en) | A kind of furniture item three-dimensional modeling method and system based on assembly |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |