Method for generating hexahedral mesh based on closed-form polycube

Abstract

A method for generating a hexahedral mesh based on a closed-form polycube includes steps of: converting a surface triangular mesh into a tetrahedral mesh; obtaining an initial 3-dimensional frame field; according surface normal constraint, generating a smooth frame field without singular line inside the model; extracting a transition relation on the cut surface based on the smooth frame field; providing Poisson optimization on cut model based on the frame field for obtaining a preliminary parameterized result; using L1-optimization for forming a final closed-form polycube; optimizing with an mixed integer for obtaining a final parameterized result; and extracting the hexahedral mesh. The method automatically generates the hexahedral mesh from the triangular mesh, and has nothing to do with an initial position of an object, which is able to better satisfy a characteristic constraint of the model; and for a model with complex topology, high quality hexahedral mesh is available.

Description

CROSS REFERENCE OF RELATED APPLICATION

This is a U.S. National Stage under 35 U.S.C 371 of the International Application PCT/CN2016/107179, filed Nov. 25, 2016, which claims priority under 35 U.S.C. 119(a-d) to CN 201610273300.6, filed Apr. 27, 2016.

BACKGROUND OF THE PRESENT INVENTION

Field of Invention

The present invention relates to a method for processing a 3-dimensional mesh, and more particularly to a method for generating a hexahedral mesh based on a closed-form polycube.

Description of Related Arts

Hexahedral meshes have a wide range of applications in computer graphics and in industries such as CAD, physical simulation, geometrical construction, and so on. However, there are still many problems in automating the generation of hexahedral meshes, such as the low degree of automation and the inability to avoid global degradation, and the resulting hexahedral mesh quality is not high. Conventionally, fully automatic hexahedral mesh generation is still a big challenge.

The conventional hexahedral mesh generation methods have their own shortcomings:

1, For a hexahedral mesh generation method based on polycubes, it is necessary to assume that the initial orientation of the model is good, see [Gregson et al. 2011 All-hex mesh generation via volumetric polycube deformation In Computer graphics forum, vol. 30, Wiley Online Library, 1407-1416]. Furthermore, the method lacks user control, see [Gregson et al. 2011 All-hex mesh generation via volumetric polycube deformation In Computer graphics forum, vol. 30, Wiley Online Library, 1407-1416. Tarini et al. 2004 Polycube-maps. In ACM Transactions on Graphics (TOG), vol. 23, ACM, 853-860].

2, For complex models, such as high genus and models with complex feature line constraint, it is impossible to get a high-quality hexahedral mesh based on polycube. [Huang et al 2014. L1 based construction of polycube maps from complex shapes. ACM TOG 33, 3, 25:1-11] discloses a polycube generation method based on L1 optimization, which can be used for hexahedral mesh generation. However, the method is based on the global uniform frame field, and the quality of the hexahedral mesh is not high.

3, [Li et al. 2012. All-hex meshing using singularity-restricted field. ACM TOG 31, 6, 177:1-11] discloses a hexahedral mesh generation method based on a three-dimensional frame field with singularity. However, the method requires a grid modification operation for many degradations, and has poor robustness, leading to difficulty in handling of complex global degradation.

SUMMARY OF THE PRESENT INVENTION

According to defects of conventional technologies, an object of the present invention is to provide a method for generating a hexahedral mesh based on a closed-form polycube which not only automatically generates a closed-form polycube by mapping from a triangular mesh, but also is independent of a position of a model, so as to obtain a high quality hexahedral mesh for a model with a complex topological structure.

Accordingly, in order to accomplish the above object, the present invention comprises steps of:

1) inputting a triangular mesh model, and converting into a tetrahedral mesh model;

2) obtaining all handles on surface meshes, and using the handles as a boundary of a cut surface; establishing a fitting surface and cutting the fitting surface to obtained the cut surface; cutting the tetrahedral mesh model into a model M_c with a first Betti number of 0;

3) defining a frame R_t in each of tetrahedrons of the model M_c, wherein the frame R_t is a 3×3 matrix; using a method in [Huang et al. 2011. Boundary aligned smooth 3d cross-frame field. ACM TOG] for obtaining an initial 3-dimensional frame field R, establishing a minimum spanning tree with the tetrahedrons as nodes in the model M_c which is cut, and globally aligning for the 3-dimensional frame field R, so as to obtain an aligned 3-dimensional frame field;

4) according to cut tetrahedral mesh model surface normal alignment constraint and cut model internal frame field smoothness requirements both inputted by a user, establishing and solving a frame field optimizing formula, so as to obtain a smooth frame field without singular line inside the model;

5) extracting a transition relation on the cut surface based on the smooth frame field obtained in the step 4);

6) providing Poisson reconstruction based on the smooth frame field obtained in the step 4) and the transition relation on the cut surface obtained in the step 5); constructing and solving a Poisson optimizing formula for obtaining a preliminary parameterized result;

7) according to the transition relation obtained in the step 5) and a polycube generating method optimized by a norm l₁ according to [Huang et la. 2014. l₁ based construction of polycube maps from complex shapes. ACM TOG], establishing and solving a L1-optimized formula for forming the closed-form polycube, wherein the L1-optimized formula is:

$\underset{\bar{\chi }}{\min }{E}_{\mathrm{arap}}+w_{\mathrm{align}}{E}_{\mathrm{align}}+w_{\mathrm{diff}}{E}_{\mathrm{diff}}$ $s.t.\stackrel{\_}{A}(\stackrel{\_}{\chi }\left(M\right))=A_{\partial M}$ ${\Pi }_{a,b}{\stackrel{\_}{\chi }}_a\left(e\right)={\stackrel{\_}{\chi }}_b\left(e\right),\forall e\in t_a\bigcap t_b\in C$

wherein Ā(χ(M)) represents a surface area of the tetrahedral mesh model in a parameter domain, A_∂M represents a surface area of an original tetrahedral mesh model, χ represents a mapping relation, C represents a collection of cut surface pieces; E_arap represents ARAP energy, E_align represents normal alignment energy, E_diff represents normal difference energy, w_align represents a normal alignment weight, w_diff represents a normal difference weight, Π_a,b represents a public plane transfer relation from a tetrahedron t_a to a tetrahedron t_b, χ_a(e) represents an image of an edge e in a parameter domain of t_a, χ_b(e) represents an image of the edge e in a parameter domain of t_b, and e represents the edge of the cut surface;

8) according to a polycube result generated in the step 7), extracting a polycube surface normal; according to the polycube surface normal and the transition relation on the cut surface, optimizing with an mixed integer for obtaining a final parameterized result; and

9) extracting according to the final parameterized result for generating the hexahedral mesh.

In the step 4), the frame field optimizing formula is:

$\underset{R}{\min }{w}_f{E}_s+w_c{E}_c+w_a{E}_A+w_d{E}_d+w_R{E}_R$

wherein w_f is a frame field smoothness weight, w_c is a frame field smoothness weight at the cut surface, w_a is a surface normal alignment weight, w_d is a surface normal smoothness weight, w_R is a frame field orthogonal constraint weight, E_s is frame field smoothness energy, E_c is frame field smoothness energy at the cut surface, E_A normal alignment energy, E_d is surface normal smoothness energy, and E_R is frame field orthogonal constraint energy.

In the step 5), the transition relation on the cut surface is extracted with a following formula, and an optimized rotation of 24 kinds of cube symmetrical rotations is selected as an optimized Π_k:

${\Pi }_k={\mathrm{argmin}}_{{\Pi }_k}\sum _{t_i\bigcap t_j\in c_k}{A}_{t_i\bigcap t_j}{{\Pi }_k{R}_{t_i}-R_{t_j}}^2$

wherein c_k is a piece collection of a number k cut surface, t_i∩t_j is a public plane of a tetrahedron t_i and a tetrahedron t_j, A_ti∩tj is an area of the public plane, R_ti is a frame defined in the tetrahedron t_i, and R_tj is a frame defined in the tetrahedron t_j.

In the step 6), the Poisson optimizing formula is:

$\underset{\bar{\chi }}{\min }{\int }_M{V\stackrel{\_}{\chi }-R}^2$ $s.t.{\Pi }_{a,b}{\stackrel{\_}{\chi }}_a\left(e\right)={\stackrel{\_}{\chi }}_b\left(e\right),\forall e\in t_a\bigcap t_b\in C$

wherein Π_a,b represents the public plane transfer relation from the tetrahedron t_a to the tetrahedron t_b, χ represents mapping defined on the model, e represents the edge of the cut surface, C represents the collection of the cut surface pieces, ∇χ represents a deformation gradient of mapping, R represents the smooth frame field obtained in the step 4), and M represents an inputted tetrahedral mesh model.

The frame field smoothness energy E_s, the frame field smoothness energy E_c at cut boundaries, the normal alignment energy E_A, the surface normal smoothness energy E_d, and the frame field orthogonal constraint energy E_R are calculated as follows:

the frame field smoothness energy E_s is:

$\sum _{t_i\bigcap t_j\in F\backslash C}\frac{V_{t_i}+V_{t_j}}{4V_F{d}_{t_i,t_j}^2}{R_{t_i}-R_{t_j}}^2$

wherein C is the collection of the cut surface pieces, d_ti,tj is a barycenter distant between a tetrahedron t_i and a tetrahedron t_i, h(r) is a rotation-irrelevant function, V_ti is a volume of t_i, V_ti is a volume of t_j, and V_F is a normalized weight of the frame field smoothness energy E_s;

the frame field smoothness energy E_c at the cutting boundaries is:

$\sum _{t_i\bigcap t_j\in C}\frac{V_{t_i}+V_{t_j}}{4V_C{d}_{t_i,t_j}^2}\sum _{k=0}^{3}h\left({\left(R_{t_i}^T{R}_{t_j}\right)}_k\right)$

wherein C is the collection of the cut surface pieces, d_ti,tj is the barycenter distant between the tetrahedron t_i and the tetrahedron t_j, h(r) is the rotation-irrelevant function, V_C is a normalized weight of the frame field smoothness energy E_c at the cut surface, k is a column number of a matrix R_ti^TR_ti, R_ti^T is transition of R_ti;

the normal alignment energy E_A is:

$\sum _{f\in \partial M}\frac{A_f}{A_{\partial M}}{R_t{n}_f}_1$

wherein ∥.∥₁ is the norm l₁, ∂M is a boundary of a tetrahedral mesh M, R_t is a frame defined in a tetrahedron t, f is a public plane of the tetrahedron t and ∂M, A_f is an area of the public plane f of the tetrahedron t and ∂M, n_f is a normal of the public plane f;

the above formula is optimized using an approximate replacement ∥u∥₁≈√{square root over (u²+ϵ)}, ϵ=10⁻³;

the surface normal smoothness energy E_d is:

$\sum _{f_i\bigcap f_j=e\in \partial M}\frac{A_{f_i}+A_{f_j}}{3A_{\partial M}}{R_{t_i}{n}_{f_i}-R_{t_j}{n}_{f_j}}^2,f_i=t_i\bigcap \partial M,f_j=t_j\bigcap \partial M$

wherein ∂M is the boundary of the tetrahedral mesh M, f_i is a public plane of the tetrahedron t_i and ∂M, f₁ is a public plane of the tetrahedron t_j and ∂M, n_fi and n_fi are a normal of f_i and a normal of f_j, A_fi is an area of f_i, A_fi is an area of f_j, R_ti is a frame defined i the tetrahedron t_i, and R_ti is a frame defined i the tetrahedron t_j;

the frame field orthogonal constraint energy E_R is:

$\sum _{t\in M}\frac{V_t}{V_M}{R_t^T{R}_t-I}^2$

wherein t is a tetrahedron in an inputted tetrahedral mesh, V_t is a volume of t, V_M is a total volume of the inputted tetrahedral mesh M, and I is a 3×3 unit matrix.

Weights of the above energy are:

$w_a=0.1,w_d=0.01,w_R=1,V_F={\Sigma }_{t_i\bigcap t_j\in F\backslash C}\frac{V_{t_i}+V_{t_j}}{4d_{t_i,t_j}^2},V_C={\Sigma }_{t_i\bigcap t_j\in C}\frac{V_{t_i}+V_{t_j}}{4d_{t_i,t_j}^2},w_f=\frac{V_F}{V_F+V_C},w_c=\frac{V_C}{V_F+V_C},$ wherein V_ti is the volume of the tetrahedron t_i, V_ti is the volume of the tetrahedron t_j, V_C is the normalized weight of the frame field smoothness energy E_c at the cut surface, and V_F is the normalized weight of the frame field smoothness energy E_s .

The rotation-irrelevant function h(r) is calculated by:h(r)=r_x²r_y²+r_y²r_z²+r_z²r_x²,r=(r_x,r_y,r_z)

wherein r=(r_x,r_y,r_z) is a 3-dimensional vector, r_x, r_y, r_z are respectively x, y and z coordinates of the 3-dimensional vector.

In the step 4), the aligned 3-dimensional frame field obtained in the step 3) is used as an initial value, and an LBFGS optimizer in a public database Alglib is used to optimize the frame field optimizing formula.

In the step 6), an IPOPT method is used for solving.

In the step 7), solving the L1-optimized formula specifically comprises a step of solving the preliminary parameterized result in the step 6) with an SQP method for obtaining the closed-form polycube.

In the step 7), energy in the L1-optimized formula is defined as follows.

The ARAP (as-rigid-as-possible) energy E_arap is:

$E_{\mathrm{arap}}=\sum _{t\in M}\frac{V_t}{V_M}{{\stackrel{\_}{V_{\chi }}}_t-\mathrm{polar}\left({\stackrel{\_}{V_{\chi }}}_t\right)}^2$

wherein polar(∇102_t) is a rotation portion of the deformation gradient ∇102_t, V_t is the volume of t, and V_M is the total volume of the inputted tetrahedral mesh M.

The normal alignment energy E_align is:

$E_{\mathrm{align}}=\sum _{f\in \partial M}\frac{1}{2A_{\partial M}}{\left({\stackrel{\_}{x}}_j^f-{\stackrel{\_}{x}}_i^f\right)\times \left({\stackrel{\_}{x}}_k^f-{\stackrel{\_}{x}}_i^f\right)}_1$

wherein A_∂M is a total surface area of M, ∂M is a surface triangular mesh of the inputted tetrahedral mesh, A_f is the area of f, and (x_i^f, x_j^f, x_k^f) is three vertex coordinates of the surface piece f in the parameter domain.

The normal difference energy E_diff is:

$E_{\mathrm{diff}}=\sum _{f_i\bigcap f_j=e\in F\backslash C}\frac{A_{f_i}+A_{f_j}}{3A_{\partial M}}{{\stackrel{\_}{n}}_i-{\stackrel{\_}{n}}_j}^2+\sum _{f_i\bigcap f_j=e\in C}\frac{A_{f_i}+A_{f_j}}{3A_{\partial M}}{{\Pi }_{\mathrm{ij}}{\stackrel{\_}{n}}_i-{\stackrel{\_}{n}}_j}^2$

wherein n_i and n_j are normals of the surface piece f_i and the surface piece f_i in the parameter domain, Π_ij is a transition relation between the two surface pieces, A_fdi is the area of f_i, A_fi is the area of f_i, and F is a surface piece collection of the model M.

Compared with conventional technologies, the present invention has advantages as follows:

The method of the present invention has nothing to do with an initial position of an object, and is able to better satisfy a characteristic constraint of the model; further, for a model with complex topology, or a model with a characteristic line, high quality hexahedral mesh is available; L1-optimization is similar to allowing the user to control details of the polycube, thereby controlling a structure of the hexahedral mesh.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flow chart of the present invention.

FIG. 2 illustrates an inputted model according to an embodiment of the present invention.

FIG. 3 illustrates a cut surface result according to the embodiment when converting the inputted model to a tetrahedral mesh.

FIG. 4 is a sketch view of an optimized 3-dimensional frame field according to the embodiment.

FIG. 5 illustrates a result of Poisson reconstruction according to the embodiment.

FIG. 6 illustrates a closed-from polycube according to the embodiment.

FIG. 7 illustrates a final hexahedral mesh according to the embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

Referring to the drawings, the present invention is further illustrated.

An embodiment of a method of the present invention is as follows:

1) inputting a triangular mesh model as shown in FIG. 2, and converting into a tetrahedral mesh model as shown in FIG. 3;

2) obtaining all handles on surface meshes, and using the handles as a boundary of a cut surface; establishing a fitting surface and cutting the fitting surface to obtained the cut surface; referring to FIG. 4, cutting the tetrahedral mesh model into a model M_c with a first Betti number of 0;

3) defining a frame R_t in each of tetrahedrons t of a tetrahedral mesh M of the model M_c, wherein the frame R_t is a 3×3 matrix; using a method in [Huang et al. 2011. Boundary aligned smooth 3d cross-frame field. ACM TOG] for obtaining an initial 3-dimensional frame field R, establishing a minimum spanning tree with the tetrahedrons as nodes in the model M_c which is cut, and globally aligning for the 3-dimensional frame field R, so as to obtain an aligned 3-dimensional frame field as shown in FIG. 5;

4) according to cut tetrahedral mesh model surface normal alignment constraint and cut model internal frame field smoothness requirements both inputted by a user, establishing a frame field optimizing formula, wherein the aligned 3-dimensional frame field obtained in the step 3) is used as an initial value, and an LBFGS optimizer in a public database Alglib is used to solve the frame field optimizing formula, so as to obtain a smooth frame field without singular line inside the model;

5) extracting a transition relation on the cut surface based on the smooth frame field obtained in the step 4), wherein according to the embodiment, the transition relation is a rotation matrix:

$\left[\begin{array}{ccc}0& 0& -1\\ 0& 1& 0\\ 1& 0& 0\end{array}\right];$

6) providing Poisson reconstruction based on the smooth frame field obtained in the step 4) and the transition relation on the cut surface obtained in the step 5); constructing and solving a Poisson optimizing formula with an IPOPT method for obtaining a preliminary parameterized result as shown in FIG. 5;

7) according to the transition relation obtained in the step 5) and a polycube generating method optimized by a norm l₁ according to [Huang et la. 2014. l₁ based construction of polycube maps from complex shapes. ACM TOG], establishing a L1-optimized formula, solving the preliminary parameterized result in the step 6) with an SQP method for obtaining the closed-form polycube as shown in FIG. 6;

8) according to a polycube result generated in the step 7), extracting a polycube surface normal; according to the polycube surface normal and the transition relation on the cut surface, optimizing with a mixed integer for obtaining a final parameterized result; and

9) extracting according to the final parameterized result for generating the hexahedral mesh as shown in FIG. 7.

Compared with conventional polycube based method, the final hexahedral mesh considers topological constraints, and has fewer angular points, wherein hexahedral units are of higher quality and features of the model are better kept.

The embodiments have been shown and described for the purposes of illustrating the functional and structural principles of the present invention and is subject to change without departure from such principles. Therefore, this invention includes all modifications encompassed within the spirit and scope of the following claims.

Claims

1. A method for generating a hexahedral mesh based on a closed-form polycube, comprising steps of: 1) inputting a triangular mesh model, and converting into a tetrahedral mesh model;2) obtaining all handles on surface meshes, and using the handles as a boundary of a cut surface; establishing a fitting surface and cutting the fitting surface to obtained the cut surface; cutting the tetrahedral mesh model into a model Mc with a first Betti number of 0;3) defining a frame Rt in each of tetrahedrons of the model Mc , wherein the frame Rt is a 3×3 matrix; using a method for obtaining an initial 3-dimensional frame field R, establishing a minimum spanning tree with the tetrahedrons as nodes in the model Mc which is cut, and globally aligning for the 3-dimensional frame field R, so as to obtain an aligned 3-dimensional frame field;4) according to cut tetrahedral mesh model surface normal alignment constraint and cut model internal frame field smoothness requirements both inputted by a user, establishing and solving a frame field optimizing formula, so as to obtain a smooth frame field without singular line inside the model;5) extracting a transition relation on the cut surface based on the smooth frame field obtained in the step 4);6) providing Poisson reconstruction based on the smooth frame field obtained in the step 4) and the transition relation on the cut surface obtained in the step 5); constructing and solving a Poisson optimizing formula for obtaining a preliminary parameterized result;7) according to the transition relation obtained in the step 5) and a polycube generating method optimized by a norm l1, establishing and solving a L1-optimized formula for forming the closed-form polycube, wherein the L1-optimized formula is:

2. The method, as recited in claim 1, wherein in the step 4), the frame field optimizing formula is:

3. The method, as recited in claim 2, wherein the frame field smoothness energy Es, the frame field smoothness energy Ec at cutting boundaries, the normal alignment energy EA, the surface normal smoothness energy Ed, and the frame field orthogonal constraint energy ER are calculated as follows: the frame field smoothness energy Es is:

4. The method, as recited in claim 3, wherein the rotation-irrelevant function h(r) is calculated by: h(r)=rx2ry2+ry2rz2+rz2rx2,r=(rx,ry,rz)wherein r=(rx, ry, rz) is a 3-dimensional vector, rx, ry, rz, are respectively x, y and z coordinates of the 3-dimensional vector.

5. The method, as recited in claim 1, wherein in the step 5), the transition relation on the cut surface is extracted with a following formula, and an optimized rotation of 24 kinds of cube symmetrical rotations is selected as an optimized Πk:

6. The method, as recited in claim 1, wherein in the step 6), the Poisson optimizing formula is:

7. The method, as recited in claim 1, wherein in the step 4), the aligned 3-dimensional frame field obtained in the step 3) is used as an initial value, and an LBFGS optimizer in a public database Alglib is used to solve the frame field optimizing formula.

8. The method, as recited in claim 1, wherein in the step 6), an IPOPT method is used for solving.

9. The method, as recited in claim 1, wherein in the step 7), solving the L1-optimized formula specifically comprises a step of solving the preliminary parameterized result in the step 6) with an SQP method for obtaining the closed-form polycube.