JP2001351119A - Method for generating triangular mesh, and recording medium for implementing the method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a triangular mesh without loosing the edge of an acute angle in an angular shape especially the edge of the acute angle of a recess in the generation of a triangular mesh from distance image obtained by performing measurement by a three-dimensional measuring device, etc. SOLUTION: A volume model having the three attribute values of a contour, inside and outside is generated from the distance image obtained by performing measurement by the three-dimensional measuring device, etc., (step S1). A basic triangular mesh is generated by the combination of the attribute value of 8 adjacent voxels in the volume model (step S2). The triangular mesh is obtained by detecting a thin plate surface and removing it from the basic triangular mesh (step S3).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating a triangular mesh and a recording medium for executing the method, and more particularly, to a CAD / CAM system applicable to a shape measuring device and a virtual reality device. The present invention relates to a method for generating a triangular mesh applied to an automatic three-dimensional shape model generating apparatus.

[0002]

2. Description of the Related Art In recent years, with the spread of multimedia on the Internet, the use of three-dimensional expressions has expanded, and the importance of three-dimensional modeling has increased. Recently, a signed distance function has been generated based on three-dimensional point cloud data and a distance image taken from a three-dimensional measuring device or the like, and an in-phase surface such as a marching cubes method (refer to the following reference [1]). A method of automatically generating a shape model represented by a triangular mesh using a generation method has been proposed (see References [2] and [3] below).

[0003] These techniques have the following advantages. -Shapes with topological structures such as through holes can be generated. -Even if there is a missing point cloud, there is an effect of generating a surface that fills the hole.・ Even if there is noise due to the characteristics of the measuring instrument or measurement conditions,
Less susceptible to noise. -Even in the case where the distribution of measured sample points is non-uniform, the sample points having a uniform distribution can be obtained by functionalization.

[0004] However, the above method has the following disadvantages.・ The reproducibility of features at acute angles is weak. Even artificially created mechanical parts such as machine parts with sharp edges are always reproduced as rounded shapes. This is because in the marching cubes method, the positions of the vertices of the triangular mesh are limited on the lattice axis, so that features such as acute ridge lines cannot be sufficiently captured. In addition, there is a grid point connection method (refer to the following reference [4]) in which vertex positions of a triangular mesh can be freely arranged in voxels. FIG. 2 is a diagram showing an example of generation of the grid point connection method as described above, where 1 indicates a three-dimensional coordinate expression (point group) of a distance image, and 2 indicates a triangle mesh reconstructed by the grid point connection method. . However, the grid point connection method has a problem that an acute ridge line of a convex portion can be reproduced, but an acute ridge line of a concave portion cannot be reproduced.

(References) [1] William E. Lorensen and Harvey E. Cline: "Ma
rching Cubes: A High Resolution 3D Surface Constru
ction Algorithm ", SIGGRAPH'87 Proceedings, pp.163-
169, Jul 1987. [2] Hugues Hoppe, Tony DeRose, Tom Duchamp, John
McDonald and Werner Stuetzle: "Surface reconstruc
tion from unorganized points ", SIGGRAPH '92 Proceed
ings, pp.71-78, Jul 1992. [3] Brian Curless and Marc Levoy: "A Volumetric
Method for Building Complex Models from Range Imag
es ", SIGGRAPH'96 Proceedings, pp.303-312, Aug 199
6. [4] Matsuoka, Ueda, Hayano: "Automatic reconstruction of solid models from point cloud data using grid point connection and local operations",
IPSJ Transactions, vol.40, No.5, 1999. [5] Jean-Daniel Boissonnat: "Geometric structure
s for three-dimentional shape representation ", ACM
Transactions on Graphics, Vol. 3, No. 4, Oct1984, p
p.266-286.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described circumstances, and has a problem in that when a triangular mesh is generated from a distance image obtained by measuring with a three-dimensional measuring device or the like, an angular shape is obtained. It is an object of the present invention to provide a method for generating a triangular mesh that eliminates the problem of losing acute ridges of a shape, particularly acute ridges of concave portions.

[0007]

According to a first aspect of the present invention, there is provided a triangle mesh generating method for generating a triangle mesh from a distance image obtained by measuring with a three-dimensional measuring device. Generating a volume model having a value, generating a basic triangle mesh by combining attribute values of eight adjacent voxels in the volume model, and detecting a thin plate surface and removing the thin plate surface from the basic triangle mesh. It is characterized by having.

According to a second aspect of the present invention, in the first aspect, the step of generating the volume model having the three attribute values of the outline, the inside, and the outside includes detecting an outline voxel from each pixel value of the distance image. Performing a step of generating a partial triangular mesh having the pixels of the distance image as vertices; detecting an outline voxel from each triangular surface of the partial triangle mesh; and detecting an outline voxel from the shooting direction of each of the distance images. And searching for an external voxel until it is found.

According to a third aspect of the present invention, in the first aspect of the invention, the step of generating a basic triangular mesh based on a combination of the attribute values of the adjacent eight voxels includes the steps of: Extracting the relevant triangle generation function and rotation information, rotating 8 voxels based on the rotation information, and giving the rotated voxel to the triangle generation function to generate the basic triangle mesh. It is characterized by having.

According to a fourth aspect of the present invention, in the first aspect of the invention, the step of detecting the thin plate surface and removing the thin plate surface from the basic triangular mesh corresponds to a lookup table using attribute values of adjacent 12 voxels as keys. The method comprises the steps of extracting position information of a surface, determining whether or not the surface is a thin plate from the position information of the surface, and removing the thin plate surface from the basic triangular mesh.

According to a fifth aspect of the present invention, there is provided a computer-readable recording medium recording a program for executing a triangle mesh generation method for generating a triangle mesh from a distance image obtained by measuring with a three-dimensional measuring device, Generating a volume model having three attribute values of contour, interior, and exterior; generating a basic triangular mesh by combining attribute values of eight adjacent voxels in the volume model; And a computer-readable recording medium on which a program for executing a method for generating a triangle mesh having a step of removing from a basic triangle mesh is recorded.

According to a sixth aspect of the present invention, in the fifth aspect of the invention, the step of generating the volume model having the three attribute values of the contour, the inside, and the outside includes detecting a contour voxel from each pixel value of the distance image. Performing a step of generating a partial triangular mesh having the pixels of the distance image as vertices; detecting an outline voxel from each triangular surface of the partial triangle mesh; and detecting an outline voxel from the shooting direction of each of the distance images. Searching for an external voxel until it is found, and a computer-readable recording medium recording a program for executing the triangle mesh generation method.

According to a seventh aspect of the present invention, in the fifth aspect of the invention, the step of generating a basic triangular mesh based on a combination of the attribute values of the adjacent eight voxels comprises the step of: Extracting the relevant triangle generation function and rotation information, rotating 8 voxels based on the rotation information, and giving the rotated voxel to the triangle generation function to generate the basic triangle mesh. And a computer-readable recording medium on which a program for executing the triangle mesh generation method is recorded.

According to an eighth aspect of the present invention, in the fifth aspect of the present invention, the step of detecting the thin plate surface and removing the thin plate surface from the basic triangular mesh corresponds to a lookup table using attribute values of adjacent 12 voxels as keys. A program for executing a method for generating a triangular mesh having a step of extracting position information of a surface, a step of determining whether or not the surface is a thin plate from the position information of the surface, and a step of removing the thin plate surface from the basic triangular mesh is recorded. Computer-readable recording medium.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The operation of the present invention will be described based on the following assumptions. In the triangle generation lookup table used in the present invention, a pointer to each function for generating a triangle and rotation information based on a triangle plane generation pattern as shown in FIG. Assume that addresses are stored in advance in the order of addresses obtained by combining the attribute values of two voxels.

It is assumed that the surface generation position look-up table used in the present invention previously stores information on a position where a surface as shown in FIG. 7 is generated in the order of addresses shown in FIG.

The distance image used in the present invention has a value in the depth direction for each pixel, and a camera position, a camera shooting direction,
It is assumed that photographing information such as the viewing angle is clear. Therefore, it is assumed that each pixel can be easily converted to a coordinate value in a three-dimensional space using perspective transformation.

In the volume model used in the present invention, the sizes Nx, Ny, Nz (all positive integers) in each direction are determined in advance, and the size is designated by the user before use if necessary. I do. If the size is increased, a triangular mesh with a finer resolution is finally obtained. A work memory space having a size proportional to Nx, × Ny, × Nz is required.

FIG. 1 is a flowchart for explaining a method of generating a triangular mesh from a distance image according to the present invention. First, a volume model having three attribute values of contour, inside, and outside is generated from a distance image obtained by measuring with a three-dimensional measuring device or the like (step S1). Next, a triangular mesh (hereinafter, referred to as a basic triangular mesh) is generated based on a combination of attribute values of eight adjacent voxels (step S2). The thin plate surface is detected and removed from the basic triangular mesh (step S3).

FIG. 2 is a flowchart for explaining in more detail the method of generating a volume model having three attribute values of contour, inside and outside from the distance image in step S1 of FIG. First, in step S11, a bounding box including all three-dimensional coordinates of each pixel of the range image is obtained. The bounding box is obtained by calculating the maximum value Xmin, Ymin, Zmin, for each element X, Y, Z of the three-dimensional coordinates (X, Y, Z) of each pixel.
The minimum values Xmax, Ymax, and Zmax are obtained, and the maximum value,
A rectangular parallelepiped having the eight coordinates V1-V8 as vertices in which the minimum values are combined as follows: V1: (Xmin-δ, Ymin-
δ, Zmin-δ), V2: (Xmax + δ, Ymin
−δ, Zmin−δ), V3: (Xmin−δ, Yma)
x, + δZmin-δ), V4: (Xmax + δ, Ym
ax + δ, Zmin-δ), V5: (Xmin-δ, Y
min−δ, Zmax + δ), V6: (Xmax + δ,
Ymin−δ, Zmax + δ), V7: (Xmin−
δ, Ymax + δ, Zmax + δ), V8: (Xmax
+ Δ, Ymax + δ, Zmax + δ). δ
Is a small positive number so that each pixel is distributed on the surface of the bounding box and does not cause delicate boundary determination problems. A V1 vertex (described later) of the bounding box is used as the origin of the volume model, and a volume model is created in which the bounding box is composed of voxels divided by predetermined sizes Nx, Ny, and Nz. The initial attribute values of voxels are all internal voxels.

Next, in step S12, a contour voxel attribute is set to a voxel including the three-dimensional coordinates (X, Y, Z) of each pixel of the range image. Here, (1)
The voxel containing each pixel is detected using the equation. (Xi, Yi, Zi) = ((X-Xmin) / Nx, (Y-Ymin) / Ny, (Z-Zmin) / Nz) (1) where (Xi, Yi, Zi) is a voxel At the address of, the fractional part of each element is truncated and rounded to an integer.

Next, in step S13, a partial triangle mesh having the vertices of the pixels of the distance image is generated.
When each pixel of the range image is regularly arranged, adjacent information is obtained. Therefore, a partial triangle mesh is generated by generating a triangle having three adjacent pixels as vertices. When the pixels of the range image are arranged irregularly, a partial triangular mesh is generated by using Delaunay triangulation (see the above-mentioned reference [5]) for the image coordinate values of the range image. If there are a plurality of distance images photographed from different directions, a partial triangle mesh is prepared for each distance image.

Next, in step S14, a contour voxel is detected from each triangular surface of the partial triangular mesh. Search for voxels that intersect all the triangular faces that make up the partial triangle mesh and set the contour voxel attributes. The three voxels including the vertices of the triangle are searched by the above equation (1), and the minimum voxel set including the three voxels is obtained. The intersection of the voxels in the voxel set with the triangle is checked, and the intersecting voxels are defined as contour voxels. In order to save the calculation cost, the intersection check is not performed for those already determined as the contour voxels in step S12.

Next, in step S15, an external voxel is detected by searching from the shooting direction of each distance image until a contour voxel is found. A pyramid-shaped volume space (in the case of parallel projection, a prismatic volume space) consisting of each triangular surface of the partial triangular mesh and the camera position is set, and the intersection between the space and voxels is checked. Set external voxel attributes to the intersecting voxels. In order to save calculation cost, those that have already been determined to be external voxels from other distance images and those that have been determined to be contour voxels in steps S12 and S14 are left as they are without performing an intersection check.

FIG. 3 is a flowchart for more specifically explaining the method of generating a triangular mesh by combining attribute values of eight adjacent voxels in step S2 shown in FIG. First, in step S21,
The corresponding triangle generation function and rotation information are extracted from the triangle generation lookup table using the attribute values of eight adjacent voxels as keys. The triangle generation pattern is shown in FIG.
Are summarized as shown below. The rotation information is shown in FIG.
Are defined as (Rx, Ry, Rz), where the number of rotations of each of the X, Y, and Z axes with one rotation of 90 degrees is defined as an integer of 0 or more.

In step S22, rotation information (Rx, R
(8, Vz) is rotated by 8 voxels. By performing the rotation conversion, the direction is made to match the triangle generation function. Here, an area storing addresses of eight voxels is prepared, and rotation conversion is performed.

In step S23, a basic triangle mesh is generated by giving the rotated and converted 8 voxels to the triangle generation function. A point representing the voxel is calculated from the address of the contour voxel as follows, and is set as the vertex of the triangle. That is, the center point of the contour voxel is calculated, the point on the partial triangle mesh closest to the center point is obtained, and the point is set as the vertex coordinates of the triangle.

FIG. 4 is a flow chart of the step 3 shown in FIG.
It is a flowchart for demonstrating the method of detecting a thin plate surface and removing it from a basic triangular mesh. Step S31
Then, the position information of the corresponding surface is extracted from the lookup table using the attribute values of the adjacent 12 voxels as keys. As shown in FIG. 9, adjacent 12 voxels are replaced with 8 voxels +
It is interpreted as 8 voxels, and the corresponding plane generation position information is extracted from the plane generation position lookup table. The plane generation position information is composed of binary information of TRUE and FALSE for 6 bits. FIG. 7 shows the correspondence between each bit and the actual plane.

In step S32, it is determined whether or not the surface is a thin plate from the position information of the surface. Whether faces are generated from both sides at the locations that are the boundaries of the two 8 voxels shown in FIG. 9,
Check by calculating AND of bit. In step S33, the thin plate surface is removed to generate a triangular mesh. The four triangular faces determined to be thin in step S32 are removed from the basic triangular mesh. In the present invention, by performing the above-described processing, it is possible to generate a triangular mesh having an arbitrary resolution by reproducing an acute ridge line from a distance image.

[0030]

As is clear from the above description, according to the first, third, fourth, fifth, seventh and eighth aspects of the present invention,
It is possible to generate a triangular mesh of an arbitrary resolution that reproduces an acute ridge line from a distance image.

According to the second and sixth aspects of the present invention, it is possible to generate a volume model having three attributes of outline, inside and outside from a distance image. Claims 1 to 4
By realizing the present invention in a three-dimensional CAD system or the like, automatic generation of a shape model from a range image can be realized.

[Brief description of the drawings]

FIG. 1 is a flowchart illustrating a method for generating a triangle mesh from a distance image according to the present invention.

FIG. 2 is a flowchart for explaining in more detail a method of generating a volume model having three attribute values of a contour, an inside, and an outside from a distance image in step S1 of FIG.

FIG. 3 is a flowchart for more specifically describing a triangle mesh generation method based on a combination of attribute values of eight adjacent voxels in step S2 shown in FIG.

FIG. 4 is a flowchart for explaining a method of detecting a thin plate surface and removing it from a basic triangular mesh in step S3 shown in FIG. 1;

FIG. 5 is a diagram showing a triangular plane generation pattern based on a triangle generation lookup table used in the present invention.

FIG. 6 is a diagram for explaining the relationship between eight adjacent voxels applied to the present invention and attribute value addresses.

FIG. 7 is a diagram illustrating an example of eight adjacent voxels and plane generation information applied to the present invention.

FIG. 8 is a diagram for explaining rotation information of eight adjacent voxels applied to the present invention.

FIG. 9 is a diagram for describing an AND operation of surface position information of eight adjacent voxels applied to the present invention.

Claims (8)

[Claims] 1. A triangular mesh generating method for generating a triangular mesh from a distance image obtained by measuring with a three-dimensional measuring device, a step of generating a volume model having three attribute values of contour, inside, and outside; A method for generating a triangular mesh, comprising: generating a basic triangular mesh based on a combination of attribute values of eight adjacent voxels in the volume model; and detecting and removing a thin plate surface from the basic triangular mesh. 2. The triangle mesh generating method according to claim 1, wherein said step of generating a volume model having three attribute values of contour, inside and outside detects a contour voxel from each pixel value of said distance image. Steps to
Generating a partial triangular mesh having the pixels of the distance image as vertices, detecting contour voxels from each triangular surface of the partial triangle mesh, and searching until a contour voxel is found from the shooting direction of each of the distance images. Detecting external voxels. 3. The triangle mesh generation method according to claim 1, wherein the step of generating a basic triangle mesh based on a combination of attribute values of the adjacent eight voxels includes:
Extracting the corresponding triangle generation function and rotation information from a look-up table using the attribute values of the eight voxels as a key; rotating and converting the eight voxels based on the rotation information; Generating the basic triangular mesh given to a function. 4. The method for generating a triangular mesh according to claim 1, wherein the step of detecting the thin plate surface and removing it from the basic triangular mesh corresponds to an attribute value of adjacent 12 voxels as a key from a lookup table. A method for generating a triangular mesh, comprising: extracting position information of a surface; determining whether or not the surface is a thin plate from the position information of the surface; and removing the thin plate surface from the basic triangular mesh. 5. A computer-readable recording medium on which a program for executing a triangle mesh generation method for generating a triangle mesh from a distance image obtained by measuring with a three-dimensional measuring device is recorded. Generating a volume model having the following three attribute values: generating a basic triangular mesh by combining attribute values of eight adjacent voxels in the volume model; detecting a thin plate surface and removing the thin plate surface from the basic triangular mesh And a computer-readable storage medium storing a program for executing the triangle mesh generation method. 6. A computer-readable recording medium on which a program for executing the method for generating a triangle mesh according to claim 5 is recorded, wherein a volume model having the three attribute values of the contour, the inside, and the outside is generated. Detecting a contour voxel from each pixel value of the distance image; generating a partial triangular mesh having pixels of the distance image as vertices; detecting a contour voxel from each triangular surface of the partial triangle mesh And a computer-readable recording medium storing a program for executing a triangle mesh generation method, the method comprising: searching until a contour voxel is found from a shooting direction of each of the distance images and detecting an external voxel. 7. A computer-readable recording medium on which a program for executing the triangle mesh generation method according to claim 5 is recorded, wherein a step of generating a basic triangle mesh by a combination of attribute values of the adjacent eight voxels. Extracting a corresponding triangle generation function and rotation information from a look-up table using the attribute values of the 8 voxels as a key; rotating the 8 voxels based on the rotation information; Computer-readable recording medium recording a program for executing a triangle mesh generation method, the method comprising: giving a basic triangle mesh to a triangle generation function. 8. A computer-readable recording medium on which a program for executing the triangle mesh generation method according to claim 5 is recorded, wherein the step of detecting the thin plate surface and removing the thin plate surface from the basic triangle mesh is performed by using Extracting the position information of the corresponding surface from the look-up table using the attribute value of 12 voxels as a key;
A computer-readable recording medium storing a program for executing a triangle mesh generation method, comprising: determining whether or not a plane is a thin plate based on positional information of the plane; and removing the thin plane from the basic triangular mesh.