CN114820985A - A three-dimensional mesh model differentiation method, device, equipment and storage medium - Google Patents

Abstract

The invention relates to the computer graphic processing technology, and discloses a method, a device, equipment and a storage medium for solving the difference of a three-dimensional grid model; a three-dimensional grid model difference solving method is characterized in that a reference model A and an extraction model B are registered; respectively calculating the point cloud density and the bounding box of the registered model to obtain the point cloud density avg of the reference model A _A And the point cloud density avg of bounding box ABox and extraction model B _B And a bounding box BBox, and dividing the voxels of the reference model to obtain A _grid (ii) a Determining a region ROI to be extracted according to the bounding box of the reference model A and the bounding box of the extraction model B; the point of the ROI of the region to be extracted is positioned at the voxel A of the reference model A _grid The difference points are searched and combined to obtain a difference result Diff. The invention can be used for the non-closed modelOr different point cloud sampling densities, or three-dimensional models with different patch numbers are differentiated without precision loss.

Description

A three-dimensional mesh model differentiation method, device, equipment and storage medium

technical field

The invention relates to computer graphics processing technology, in particular to a method, device, equipment and storage medium for differentiating a three-dimensional mesh model.

Background technique

The 3D model can represent surface information by point cloud data and triangular patches formed by connecting these 3D points. The 3D model of the object can be obtained by scanning the object with sensors. There are various types of sensors, such as vision sensors, computed tomography, etc. The 3D model of the same object acquired by the device may be a non-closed model (such as a 3D scanned tooth and jaw model), or have different point cloud densities and number of patches.

As in the prior art, the article "Accurate, Robust, Efficient Regularization of Boolean Values for General 3D Grids" in Computer and Mathematical Applications Technology 70 (2015) 1235-1254 discloses that in 3D model processing, in order to obtain two The difference between the models is realized by Boolean operation. The Boolean operation requires the grid to be closed and oriented (with normal direction). It is not applicable to the non-closed model mentioned in the above one. During the Boolean operation of the closed model, the Boundary meshes often require triangulation to determine intersecting boundaries, which can change the existing mesh structure and cause accuracy degradation.

SUMMARY OF THE INVENTION

The present invention is not applicable to the non-closed model in the prior art. In the process of executing the Boolean operation, the boundary grid often needs to be triangulated to determine the intersecting boundary, which will change the existing grid structure and cause the problem of precision degradation. Provided are a method, apparatus, device and storage medium for differentiating a three-dimensional mesh model.

In order to solve the above-mentioned technical problems, the present invention is solved by the following technical solutions:

A method for differentiating a three-dimensional mesh model, the model includes a reference model A and an extraction model B, and the method includes:

For model registration, the reference model A and the extraction model B are registered, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlaps;

Calculation of point cloud density and bounding box, the point cloud density and bounding box are calculated for the registered reference model A and extraction model B, respectively, to obtain the point cloud density avg _A of the reference model A, the bounding box ABox and the extraction model B The point cloud density avg _B and the bounding box BBox;

The voxels of the benchmark model A are obtained A _grid , and the voxels of the benchmark model are divided to obtain A _grid ;

Determine the ROI of the region to be extracted, and determine the ROI of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B;

To obtain the difference result Diff, the points of the ROI of the region to be extracted are searched in the voxel A _grid of the benchmark model A, the difference points are extracted, and the difference points are merged to obtain the difference result Diff.

Preferably, the registration of the model includes fine registration and coarse registration; the fine registration is performed by the ICP method.

Preferably, the calculation method of the bounding box includes:

Obtain all points in the benchmark model A to form a point set P _A , each of which contains three coordinate components of X, Y, and Z;

Randomly select points P _a1 and P _a2 in P _A , and calculate the initial outer bounding boxes minP _ABox and maxP _ABox of model A;

minP _ABox = min(P _a1 , P _a2 )

maxP _ABox = max(P _a1 , P _a2 )

Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value;

Traverse each point P _a in the point set P _A , and update the minP _ABox and maxP _ABox coordinates;

minP _ABox = min(minP _ABox , P _a );

maxP _ABox = max(maxP _ABox , P _a );

Thereby, the bounding box ABox of the benchmark model A is obtained;

Obtain all points in the extraction model B to form a point set P _B , each of which contains three coordinate components of X, Y, and Z;

Randomly select points P _b1 and P _b2 in P _B , and calculate the initial outer bounding boxes minP _BBox and maxP _BBox of model A;

minP _BBox = min(P _b1 , P _b2 )

maxP _BBox = max(P _b1 , P _b2 )

Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value;

Traverse each point P _b in the point set P _B , update the minP _BBox and maxP _BBox coordinates;

minP _BBox = min(minP _BBox , P _b );

maxP _BBox = max(maxP _BBox , P _b );

Thereby, the bounding box BBox of the benchmark model B is obtained.

Preferably, the calculation method of the point cloud density avg A of the reference model A and the point cloud density avg _B of the extraction model _B is,

avg _A = ∑A_edge_len/A_edge_num;

avg _B =∑B_edge_len/B_edge_num;

Among them, A_edge_len is the edge length of one side in benchmark model A, A_edge_num is the total number of edges in benchmark model A; B_edge_len is the edge length of one side in extraction model B, and B_edge_num is the total number of edges in extraction model B.

Preferably, the voxel acquisition A _grid of the reference model A includes;

For the determination of the voxel resolution, the larger value of the point cloud density avg A of the reference model A and the point cloud density avg _B of the extraction model _B is multiplied by the first threshold th;

Resolution=th*max(avg _A , avg _B );

The length of the bounding box A _Box and the voxel resolution Resolution in the three directions of X, Y, and Z are divided and rounded to obtain the number of voxel grids. The voxel resolutions are (n _x , _ny , n _z ) ); n _x is the voxel resolution in the X direction, n _y is the voxel resolution in the Y direction, n _z is the voxel resolution in the Z direction; acquisition of voxel A _gridD , A _gridD =n _x *n _y *n _z .

Preferably, the ROI of the region to be extracted is determined, the difference direction is determined, the difference direction between the bounding box of the reference model A and the bounding box of the extraction model B, and the difference model Diff is extracted according to the difference direction; the difference model is expanded according to the difference direction, so as to mark the place to be extracted Regional ROI.

As a preference, obtain the difference result Diff, and calculate the coordinates of the grid xyz;

For a point P1 in the ROI area and any point P2 in the grid a,

dist(P1, P2) >dist_th&& P1N * P2N <normal_th,

Among them, dist represents the Euclidean distance between two input points, P1N and P2N are the normal three-dimensional vectors of P1 and P2 points; dist_th and normal_th are the thresholds.

In order to solve the above-mentioned technical problems, the present application also provides a three-dimensional mesh model differentiating device, which includes:

The registration module of the model, the registration module of the model is used to register the reference model A and the extraction model B, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlap;

The calculation module of point cloud density and bounding box, the calculation module of point cloud density and bounding box is used to calculate the point cloud density and bounding box for the registered reference model A and extraction model B respectively, and obtain the point of reference model A. Cloud density avg _A and bounding box ABox and point cloud density avg _B and bounding box BBox of extraction model B;

The voxel of the benchmark model A obtains the A _grid module, and the voxel of the benchmark model A obtains the A _grid module for performing voxel division on the benchmark model to obtain the A _grid ;

Determine the ROI module of the region to be extracted, and determine the ROI module of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B, to determine the ROI of the region to be extracted;

The acquisition module of the difference result Diff, the acquisition module of the difference result Diff searches the point of the ROI of the region to be extracted in the voxel A _grid of the benchmark model A, extracts the difference point, and merges the difference point to obtain the difference result diff.

In order to solve the above technical problems, the present application also provides a device for differentiating a three-dimensional mesh model, which is a device for obtaining a three-dimensional mesh model that is realized by a method for differentiating a three-dimensional mesh model.

In order to solve the above technical problems, the present application also provides a storage medium, which is a storage medium realized by a three-dimensional mesh model differentiation method.

Due to the adoption of the above technical solutions, the present invention has significant technical effects: the present invention can differentiate between non-closed models or different point cloud sampling densities, or three-dimensional models with different numbers of patches without loss of precision.

Description of drawings

Fig. 1 is the system schematic diagram of the present invention;

Fig. 2 is the voxel acquisition A _grid flow chart of the benchmark model A of the present invention;

Fig. 3 is the calculation flow chart of the bounding box of the present invention;

Fig. 4 is the benchmark model A image one of the present invention;

Fig. 5 is the extraction model B image one of the present invention;

Fig. 6 is the diff model figure of the present invention;

Fig. 7 is the diff model figure of the present invention;

Fig. 8 is the benchmark model A image two of the present invention;

Fig. 9 is the second image of the extraction model B of the present invention.

Detailed ways

The present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Example 1

A method for differentiating a three-dimensional mesh model, the model includes a reference model A and an extraction model B, and the method includes:

For model registration, the reference model A and the extraction model B are registered, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlaps;

Calculation of point cloud density and bounding box, the point cloud density and bounding box are calculated for the registered reference model A and extraction model B, respectively, to obtain the point cloud density avg _A of the reference model A, the bounding box ABox and the extraction model B The point cloud density avg _B and the bounding box BBox;

The voxels of the benchmark model A are obtained A _grid , and the voxels of the benchmark model are divided to obtain A _grid ;

Determine the ROI of the region to be extracted, and determine the ROI of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B;

To obtain the difference result Diff, the points of the ROI of the region to be extracted are searched in the voxel A _grid of the benchmark model A, the difference points are extracted, and the difference points are merged to obtain the difference result Diff.

The registration of the model includes fine registration and coarse registration; fine registration is performed by the ICP method.

The calculation methods of the bounding box include:

Obtain all points in the benchmark model A to form a point set P _A , each of which contains three coordinate components of X, Y, and Z;

Randomly select points P _a1 and P _a2 in P _A , and calculate the initial outer bounding boxes minP _ABox and maxP _ABox of model A;

minP _ABox = min(P _a1 , P _a2 )

maxP _ABox = max(P _a1 , P _a2 )

Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value;

Traverse each point P _a in the point set P _A , and update the minP _ABox and maxP _ABox coordinates;

minP _ABox = min(minP _ABox , P _a );

maxP _ABox = max(maxP _ABox , P _a );

Thereby, the bounding box ABox of the benchmark model A is obtained;

Obtain all points in the extraction model B to form a point set P _B , each of which contains three coordinate components of X, Y, and Z;

Randomly select points P _b1 and P _b2 in P _B , and calculate the initial outer bounding boxes minP _BBox and maxP _BBox of model A;

minP _BBox = min(P _b1 , P _b2 )

maxP _BBox = max(P _b1 , P _b2 )

Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value;

Traverse each point P _b in the point set P _B , update the minP _BBox and maxP _BBox coordinates;

minP _BBox = min(minP _BBox , P _b );

maxP _BBox = max(maxP _BBox , P _b );

Thereby, the bounding box BBox of the benchmark model B is obtained.

The calculation method of the point cloud density avg A of the benchmark model A and the point cloud density avg _B of the extraction model _B is,

avg _A = ∑A_edge_len/A_edge_num;

avg _B =∑B_edge_len/B_edge_num;

Among them, A_edge_len is the edge length of one side in benchmark model A, A_edge_num is the total number of edges in benchmark model A; B_edge_len is the edge length of one side in extraction model B, and B_edge_num is the total number of edges in extraction model B.

The voxel acquisition A _grid of the benchmark model A includes;

For the determination of the voxel resolution, the larger value of the point cloud density avg A of the reference model A and the point cloud density avg _B of the extraction model _B is multiplied by the first threshold th;

Resolution=th*max(avg _A , avg _B );

The length of the bounding box A _Box and the voxel resolution Resolution in the three directions of X, Y, and Z are divided and rounded to obtain the number of voxel grids. The voxel resolutions are (n _x , _ny , n _z ) ); n _x is the voxel resolution in the X direction, n _y is the voxel resolution in the Y direction, n _z is the voxel resolution in the Z direction; acquisition of voxel A _gridD , A _gridD =n _x *n _y *n _z .

Determine the ROI of the region to be extracted, determine the difference direction, the difference direction of the bounding box of the reference model A and the bounding box of the extraction model B, and extract the difference model Diff according to the difference direction; expand the difference model according to the difference direction, so as to mark the ROI of the region to be extracted.

Obtain the difference result Diff, and calculate the coordinates of the grid xyz;

For a point P1 in the ROI area and any point P2 in the grid a,

dist(P1, P2) >dist_th&& P1N * P2N <normal_th,

Among them, dist represents the Euclidean distance between two input points, P1N and P2N are the normal three-dimensional vectors of P1 and P2 points; dist_th and normal_th are the thresholds.

Example 2

On the basis of Embodiment 1, this embodiment provides a three-dimensional mesh model differentiating device, which includes:

The registration module of the model, the registration module of the model is used to register the reference model A and the extraction model B, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlap;

The calculation module of point cloud density and bounding box, the calculation module of point cloud density and bounding box is used to calculate the point cloud density and bounding box for the registered reference model A and extraction model B respectively, and obtain the point of reference model A. Cloud density avg _A and bounding box ABox and point cloud density avg _B of extraction model B and bounding box BBox;

The voxel of the benchmark model A obtains the A _grid module, and the voxel of the benchmark model A obtains the A _grid module for performing voxel division on the benchmark model to obtain the A _grid ;

Determine the ROI module of the region to be extracted, and determine the ROI module of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B, to determine the ROI of the region to be extracted;

The acquisition module of the difference result Diff, the acquisition module of the difference result Diff searches the point of the ROI of the region to be extracted in the voxel A _grid of the benchmark model A, extracts the difference point, and merges the difference point to obtain the difference result diff.

Example 3

On the basis of Embodiment 1, this embodiment provides a three-dimensional mesh model differentiation device, which is a three-dimensional mesh model differentiation device realized by a three-dimensional mesh model differentiation method.

Example 4

On the basis of Embodiment 1, this embodiment provides a storage medium, which is a storage medium implemented by a three-dimensional mesh model differentiation method.

Example 5

On the basis of the above embodiment, in this embodiment, the number of points in the reference model A in FIG. 4 is 50922, and the number of points in the extracted model B in FIG. 5 is 210099;

Example 6

On the basis of the above embodiment, in this embodiment, the number of points of the reference model A in FIG. 8 is 50922, and the number of points of the extracted model B in FIG. 9 is 210099; the diff model diagram is obtained by differentiating them.

Claims (10)

1. a three-dimensional mesh model differentiation method, the model comprises a reference model A and an extraction model B, and its method comprises: For model registration, the reference model A and the extraction model B are registered, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlaps; Calculation of point cloud density and bounding box, the point cloud density and bounding box are calculated for the registered reference model A and extraction model B, respectively, to obtain the point cloud density avg _A of the reference model A, the bounding box ABox and the extraction model B The point cloud density avg _B and the bounding box BBox; The voxels of the benchmark model A are obtained A _grid , and the voxels of the benchmark model are divided to obtain A _grid ; Determine the ROI of the region to be extracted, and determine the ROI of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B; To obtain the difference result Diff, the points of the ROI of the region to be extracted are searched in the voxel A _grid of the benchmark model A, the difference points are extracted, and the difference points are merged to obtain the difference result Diff. 2 . The method for differentiating a three-dimensional grid model according to claim 1 , wherein the registration of the models includes fine registration and coarse registration; and the fine registration is performed by the ICP method. 3 . 3. a kind of three-dimensional mesh model differentiating method according to claim 1, is characterized in that, the calculation method of bounding box comprises: Obtain all points in the benchmark model A to form a point set P _A , each of which contains three coordinate components of X, Y, and Z; Randomly select points P _a1 and P _a2 in P _A , and calculate the initial outer bounding boxes minP _ABox and maxP _ABox of model A; minP _ABox = min(P _a1 , P _a2 ) maxP _ABox = max(P _a1 , P _a2 ) Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value; Traverse each point P _a in the point set P _A , and update the minP _ABox and maxP _ABox coordinates; minP _ABox = min(minP _ABox , P _a ); maxP _ABox = max(maxP _ABox , P _a ); Thereby, the bounding box ABox of the benchmark model A is obtained; Obtain all points in the extraction model B to form a point set P _B , each of which contains three coordinate components of X, Y, and Z; Randomly select points P _b1 and P _b2 in P _B , and calculate the initial outer bounding boxes minP _BBox and maxP _BBox of model A; minP _BBox = min(P _b1 , P _b2 ) maxP _BBox = max(P _b1 , P _b2 ) Among them, max represents the XYZ coordinate value with the largest value among the two points as the return value; min represents the XYZ coordinate value with the smallest value among the two points as the return value; Traverse each point P _b in the point set P _B , update the minP _BBox and maxP _BBox coordinates; minP _BBox = min(minP _BBox , P _b ); maxP _BBox = max(maxP _BBox , P _b ); Thereby, the bounding box BBox of the benchmark model B is obtained. 4. a kind of three-dimensional grid model differentiating method according to claim 1 is characterized in that, the calculation method of the point cloud density avg A of reference model A and the point cloud density avg _B of extraction model _B is, avg _A = ∑A_edge_len/A_edge_num; avg _B =∑B_edge_len/B_edge_num; Among them, A_edge_len is the edge length of one side in benchmark model A, A_edge_num is the total number of edges in benchmark model A; B_edge_len is the edge length of one side in extraction model B, and B_edge_num is the total number of edges in extraction model B. 5. a kind of three-dimensional grid model differentiation method according to claim 1, is characterized in that, the voxel acquisition A _grid of reference model A comprises; For the determination of the voxel resolution, the larger value of the point cloud density avg A of the reference model A and the point cloud density avg _B of the extraction model _B is multiplied by the first threshold th; Resolution=th*max(avg _A , avg _B ); The length of the bounding box A _Box and the voxel resolution Resolution in the three directions of X, Y, and Z are divided and rounded to obtain the number of voxel grids. The voxel resolutions are (n _x , _ny , n _z ) ); n _x is the voxel resolution in the X direction, n _y is the voxel resolution in the Y direction, n _z is the voxel resolution in the Z direction; acquisition of voxel A _gridD , A _gridD =n _x *n _y *n _z . 6. a kind of three-dimensional grid model differentiating method according to claim 1, is characterized in that, determining the region ROI to be extracted, determining the difference direction, the difference direction of the bounding box of the reference model A and the bounding box of the extraction model B, according to the difference Extract the difference model Diff in the direction; expand the difference model according to the difference direction, so as to mark the ROI of the region to be extracted. If the difference direction does not exist, mark the extraction model B as the ROI of the region to be extracted. 7. a kind of three-dimensional grid model difference method according to claim 1 is characterized in that, the acquisition of difference result Diff calculates the coordinates of grid xyz; For a point P1 in the ROI area and any point P2 in grid a in voxel A _gridD , dist(P1, P2) >dist_th&& P1N * P2N <normal_th, Among them, dist represents the Euclidean distance between two input points, P1N and P2N are the normal three-dimensional vectors of P1 and P2 points; dist_th and normal_th are the thresholds. 8. A three-dimensional mesh model differentiating device is characterized in that comprising: The registration module of the model, the registration module of the model is used to register the reference model A and the extraction model B, so that the reference model A and the extraction model B are aligned, and the same area of the reference model A and the extraction model B overlap; The calculation module of point cloud density and bounding box, the calculation module of point cloud density and bounding box is used to calculate the point cloud density and bounding box for the registered reference model A and extraction model B respectively, and obtain the point of reference model A. Cloud density avg _A and bounding box ABox and point cloud density avg _B of extraction model B and bounding box BBox; The voxel of the benchmark model A obtains the A _grid module, and the voxel of the benchmark model A obtains the A _grid module for performing voxel division on the benchmark model to obtain the A _grid ; Determine the ROI module of the region to be extracted, and determine the ROI module of the region to be extracted according to the bounding box of the benchmark model A and the bounding box of the extraction model B, to determine the ROI of the region to be extracted; The acquisition module of the difference result Diff, the acquisition module of the difference result Diff searches the point of the ROI of the region to be extracted in the voxel A _grid of the benchmark model A, extracts the difference point, and merges the difference point to obtain the difference result diff. 9 . A three-dimensional grid model differentiation device, characterized in that it is a three-dimensional grid model differentiation device realized by the three-dimensional grid model differentiation method according to any one of claims 1 to 6 . 10 . 10. A storage medium, characterized in that it is a storage medium realized by the method for differentiating a three-dimensional mesh model according to any one of claims 1-6.

Images