Gum three-dimensional model shelling method and device based on Marking Cube algorithm

Abstract

The invention discloses a gingival three-dimensional model shelling method and device based on a Marking Cube algorithm, and the method comprises the steps: (1) importing an initial gingival three-dimensional model, and carrying out the related preprocessing; (2) performing extrusion offset on the non-division points; (3) constructing a symbol distance field by adopting a grid diffusion and hierarchical construction strategy; (4) deleting redundant repeated vertexes and patches corresponding to the vertexes in the model; (5) traversing vertexes of the manifold inner shell gingival model, and removing topological errors; (6) sewing the bottom surface of the height compensation gingival model to obtain a closed shell gingival model; (7) combining the standard inner shell gingival model and the standard outer shell gingival model together to obtain an open shelling gingival model; and (8) obtaining a complete and closed standard shelled gingiva model. According to the method, effective preprocessing and post-processing processes are provided before and after MarchingCube reconstruction, the construction defect of a MarchingCube result is accurately repaired, no error is reported after the finally generated standard shelled gingival model is imported into CAD printing software, and the grid quality is uniform and healthy.

Description

Technical Field

[0001] The present invention provides a method and device for shelling an initial gingival three-dimensional model and subsequent model optimization. The classical Marching Cube mesh reconstruction algorithm is used to generate a preliminary inner shell gingival model. In view of the common mesh quality problem, a hole repairing mesh reconstruction algorithm and the characteristics of the mesh half-edge data structure are used to finally obtain a complete, normal-faceted standard shelling gingival model. Background Art

[0002] In recent years, the rapid development of stomatology, such as the application of 3D scanning, CAD / CAM technology, the use of new materials, and the introduction of artificial intelligence and digital platforms, has greatly improved the accuracy and comfort of invisible braces. In addition, the scientific basis provided by evidence-based medicine has also verified its safety and effectiveness. These advances have jointly promoted the widespread application of invisible braces in the field of dental correction, making it a popular choice for more and more people. The production of invisible braces requires 3D printing of the patient's virtual gingival model after simulated orthodontic treatment, and then pressing the mold. In order to save materials for 3D printing, it is necessary to shell the model while ensuring the strength of the virtual gingival model. Model shelling is an important operation of CAD software for 3D printing. Shelling, as the name suggests, is to extract the entity into a shell with the same wall thickness on all sides. This function can not only add a thickness to the unclosed surface model to obtain a body model, but also hollow out the inside of the closed model to obtain a shell with uniform thickness. In terms of direction, it can shrink inward or expand outward.

[0003] The most common and classic shell extraction algorithm is based on the Marching Cube mesh reconstruction algorithm. The MarchingCube algorithm is a classic algorithm for extracting isosurfaces from three-dimensional discrete data fields. It is mainly used in visualization scenes in the medical field, such as 3D reconstruction of CT scans and MRI scans. The basic idea of ​​the algorithm is to first construct a three-dimensional voxel distance field based on the geometry to be meshed and reconstructed, process the cube voxels in the distance field one by one, separate the cubes that intersect with the isosurface, and use interpolation to calculate the intersection of the isosurface and the cube edge. According to the relative position of each vertex of the cube and the isosurface, the intersection of the isosurface and the cube edge is connected in a certain way to generate an isosurface as an approximate representation of the isosurface in the cube. This is because Marching Cubes has a basic assumption: the data field along the edge of the hexahedron changes continuously. That is, if the two vertices of an edge are respectively greater than or less than the value of the isosurface, then there is one and only one point on the edge that is the intersection of the edge and the isosurface. Intuitively speaking, it is to use many small cubes to divide the space, and then use the plane inside the small cube to approximate the current isosurface. Obviously, the more small cubes there are, the better the approximation effect.

[0004] There are many problems with using the Marching Cube-based mesh reconstruction algorithm directly for shelling the initial gingival 3D model. First, because the mesh is reconstructed by constructing planes in small cubes, the model is composed of discrete patches. Adjacent patches have no common vertices, but their vertices overlap. This not only causes trouble for subsequent processing, but also wastes vertices. In addition, patches often intersect with each other, resulting in complex edges and non-manifold edges.

[0005] The standard shell-drawing gingival model generation algorithm of the present invention performs appropriate extrusion preprocessing in view of the special structure of the initial gingival three-dimensional model and the final invisible braces production requirements; and adjusts the construction process of the signed distance field, and proposes a grid diffusion + hierarchical signed distance field construction idea. In addition, the present invention proposes a series of optimization processes after the reconstruction of MarchingCube: based on the results of the preliminary inner shell gingival model, duplicate vertices are removed, vertices and facets with topological errors are deleted, and holes are repaired, and finally the bottom surface is cut and stitched. The optimized standard shell-drawing gingival model can be better recognized by 3D printing software, which improves the quality of mesh reconstruction. If the standard shell-drawing gingival model is used for further processing, the usability is also higher and more stable. Summary of the invention

[0006] The present invention aims to overcome the above-mentioned shortcomings of the prior art and provides a method and device for extracting a gingival three-dimensional model based on MarchingCube.

[0007] The objective of the present invention is to provide a shelling method specifically for an initial gingival three-dimensional model based on the classic Marching Cube mesh reconstruction algorithm, as well as a method for optimizing and repairing the preliminary inner shell gingival model after shelling, so as to solve the problem of the shelling gingival model and obtain a healthy and robust model.

[0008] The present invention uses the classic Marching Cube mesh reconstruction algorithm to generate a preliminary inner shell gingival model, and uses the hole repair mesh reconstruction algorithm and the characteristics of the mesh half-edge data structure to address the common mesh quality problems, and finally obtains a complete, normal-faceted standard shell gingival model.

[0009] The input of mesh optimization is the initial gingival 3D model; the output is a complete and mesh-healthy standard shell gingival model. The processes involved are position correction, downward extrusion, distance field and inner shell construction, reconstruction of half-edge structure, inner shell repair, Boolean operation, bottom surface cutting, and bottom edge stitching.

[0010] The first aspect of the present invention provides a method for extracting a gingival three-dimensional model based on Marching Cube, and the specific steps are as follows:

[0011] Step 1: Import the initial gingival 3D model and set it as M o The bottom surface of the model is not closed, and all the bottom edge points are in the same plane. Find the half edge that is only used by one face to get all the bottom edge points, and set the set of all bottom edge points as PS b According to PS b The bottom plane normal can be calculated, and the transformation matrix can be further calculated to transform the initial gingival 3D model to a position where the bottom surface is parallel to the XoY plane (the XoY plane is the plane orthogonal to the X-axis and Y-axis of the world coordinate system), and then translated to a position where the geometric center coincides with the origin of the world coordinate system. Finally, the reset gingival model M is obtained. r .

[0012] Step 2: Based on M r The bottom edge of the PS is extruded 2.0mm in the negative direction of the Z axis. b Make a copy and get PS bc . Put PS bc The point is translated 2.0 mm in the negative direction of the Z axis, and then compared with the PS b Sew the patch point by point and add it to M r In the figure, the highly compensated gingival model M is obtained. h .

[0013] Step 3: Next, construct the signed distance field, or SDF. The SDF is constructed using a grid diffusion + hierarchical strategy. First, a grid diffusion with a lower resolution is constructed. h The bounding box voxelization constructs a blank field, in which each voxel can be set to a value. Through the breadth-first search (BFS) algorithm, from M h The surface expands inward and outward layer by layer to fill the value of the corresponding voxel. Since the diffusion is discrete, parallel computing is used here, which is relatively efficient.

[0014] Assume the required thickness of the shell is T mm, find all voxels with a value of -T, and put them into the voxel set VS ori Next, put VS ori The voxels in are further subdivided by a hierarchical strategy to calculate the local subdivision signed distance field LDSDF. The value of each subvoxel is obtained by interpolating the values ​​of nearby voxels. Find all the subvoxels closest to -T value and use the classic MarchingCube to generate isosurfaces to obtain the preliminary inner shell gingival model, which is set as

[0015] Step 4: Traverse All the patches, record the three vertex positions of each patch. Then delete the redundant duplicate vertices in the model, leaving only one vertex at each position, and delete all the patches at the same time. At this time, the model becomes a point cloud model. Then, according to the three vertex positions of each patch recorded previously and The remaining non-duplicate vertices in the reconstruct the topological structure. With a light number of vertices and the same number of faces as before, the model is no longer a bunch of scattered faces, but a mesh model with a half-edge data structure, which is called a manifold inner shell gum model, set to Later, we can use the characteristics of the half-edge data structure to for further processing.

[0016] Step 5. Delete The specific method is: first traverse all its vertices, check the connection of the vertices with the help of the connection characteristics of the half-edge data structure, find the vertices with structural errors, and delete them to obtain the inner shell gingival model to be optimized, set as The wrong vertices may be vertices with unreasonable topological connections, or they may be two points of non-popular edges that are reused more than twice. The wrong vertices are removed, so there will be holes. Next, traverse to find For all holes, call the hole repair algorithm to fill the holes.

[0017] The principle of the hole repair algorithm is to first calculate the average length l of the boundary edge according to the hole boundary point information; then calculate the angle between the two adjacent edges of each boundary point; find the three boundary points where the two edges have the smallest angle, calculate the distance s between the two left and right boundary points of the three points, and determine whether s<2×l is true. If so, add a triangle, if not, add two triangles. Then find new boundary points to determine whether the hole still exists. If it still exists, continue the above process, otherwise end. Because some models are more complex, the hole filling algorithm can only add a part of the face at a time and cannot be completely closed, so it is necessary to loop to find out whether there is a hole. If so, repair it until there is no hole and obtain a standard inner shell gingival model, set as

[0018] Step 6: According to PS bc Put M h The bottom surface of the suture is closed to obtain the closed shell gingival model, which is set as The reason for closing M h Conclusion This is because the parameters of the Boolean addition operation must be closed models, otherwise an error will occur. Do a Boolean addition operation to get the Boolean rear shell gingival model Then remove The sutured bottom surface is obtained to obtain the unsealed standard shell gingival model

[0019] Step 7: and The models are merged together to obtain an open shell gingival model, which is set as Since the MarchingCube algorithm is not accurate in dealing with edge problems, There are still some random cuts on the inner shell. Cut off a part of the vertex at the bottom to get the cut-out shell gum model, set as After the vertices are trimmed, The edge points of the new bottom surface are displaced in the Z-axis direction, transformed to the same plane, and the new bottom surface is made parallel to the XoY plane to prepare for subsequent stitching.

[0020] Step 8: The two circles of bottom edge points of the inner shell and the outer shell are sorted according to the edge topology connection, and the sorting starting points of the two circles are guaranteed to correspond and the sorting directions are consistent. Then the two circles of bottom edge points are stitched point by point, and the obtained face is added to Finally, a closed perfect shell model is obtained, which is called the standard shell gingival model. This is the final output result of this method.

[0021] The second aspect of the present invention provides a Marching Cube-based gingival three-dimensional model extraction device, including a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a Marching Cube-based gingival three-dimensional model extraction method of the present invention.

[0022] The technical concept of the present invention is: according to the specific structure of the initial gingival 3D model and the final actual demand for generating invisible braces, a special preprocessing strategy and a new signed distance field construction method are tailored for it; and a series of groundbreaking post-processing optimization methods are applied to the MarchingCube reconstruction results, such as deleting erroneous topology and iterative repair process. Finally, the bottom is stitched together with the shell to obtain a standard shell-drawing gingival model.

[0023] The advantages of the present invention are: realizing fast and high-performance signed distance field construction of the initial gingival three-dimensional model, and providing effective pre-processing and post-processing processes before and after MarchingCube reconstruction; accurately repairing the construction defects of the Marching Cube results, and the final generated standard shell gingival model is imported into the CAD printing software without error, and the mesh quality is uniform and healthy, which also lays the foundation for further processing the standard shell gingival model and adding production processes such as base plates and latches, making it more usable and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is the overall logical framework for generating the standard shell-drawing gingival model of the present invention.

[0025] Figures 2(a) and 2(b) are the reset gingival model of the present invention with the bottom surface aligned with the XoY plane after position correction and the XYZ axis of the world coordinate system; Figure 2(a) is a left-side top view, and Figure 2(b) is a left-side top view.

[0026] Figure 3 It is the height-compensated gum model effect obtained by extruding the bottom edge downward.

[0027] FIG4(a) and FIG4(b) are side view examples of constructing a distance field; the light shaded portion of FIG4(a) is a rough initial result voxel set in the initial distance field, and the dark shaded portion of FIG4(b) is a more detailed final result voxel set after the initial result voxel is subdivided.

[0028] FIG5(a) and FIG5(b) are preliminary inner shell gingival models after MarchingCube mesh reconstruction; FIG5(a) is a bottom top view, and FIG5(b) is a right side top view.

[0029] FIG6(a) and FIG6(b) are diagrams showing the erroneous topology; FIG6(a) is an erroneous topology of the inner shell gingival model to be optimized, and FIG6(b) is a state with holes after deleting the erroneous topology.

[0030] FIG7(a) and FIG7(b) are schematic diagrams of adding vertices and facets by the hole repair algorithm; FIG7(a) is a case of adding one triangle, and FIG7(b) is a case of adding two triangles.

[0031] FIG8(a) and FIG8(b) are comparisons before and after hole repair; FIG8(a) is the state before hole repair, and FIG8(b) is the state after hole repair.

[0032] FIG9(a) and FIG9(b) are comparisons of teeth before and after Boolean addition operation; FIG9(a) is the effect diagram before Boolean addition operation, and FIG9(b) is the effect diagram after Boolean addition operation.

[0033] Fig.10 It is a display of the random cutout problem on the bottom edge.

[0034] Fig.11 This is a cutting diagram.

[0035] Figure 12(a) and Figure 12(b) are comparisons of the effects before and after the bottom edge is sutured; Figure 12(a) is the effect before the bottom is sutured, and Figure 12(b) is the effect after the bottom is sutured.

[0036] Figures 13(a) and 13(b) are the final renderings of the standard shell gingival model; Figure 13(a) is a mesh topology and facet display from the left top view, and Figure 13(b) is a facet display from the bottom top view. DETAILED DESCRIPTION

[0037] Next, the present invention will be described in further detail with reference to the accompanying drawings.

[0038] Example 1

[0039] This embodiment relates to a method for extracting a gingival three-dimensional model based on Marching Cube.

[0040] The general idea of ​​generating a standard shell gingival model is as follows: after inputting the initial gingival 3D model, the position is first corrected to the XoY plane, and the bottom edge is extruded downward to obtain a highly compensated gingival model. Then, a signed distance field is constructed based on it, and a preliminary inner shell gingival model is reconstructed using the MarchingCube method. Next, the preliminary inner shell gingival model is processed. While keeping the vertex topology of the face unchanged, duplicate vertices in the same position are deleted and the topology is reconnected to obtain a half-edge data structure; the wrong vertices are found, and they and a neighboring vertex are deleted, and then the hole repair algorithm is used to repair its holes to obtain a standard inner shell gingival model. At the same time, after the bottom surface of the highly compensated gingival model is closed, a Boolean addition operation is performed with all teeth, and the standard outer shell gingival model is obtained after the bottom surface is split. After merging them, a part of the bottom is cut, and then the bottom edge is stitched to obtain the final standard shell gingival model.

[0041] Figure 1 It is the overall logical framework for generating standard shell gingival models. The following is a detailed introduction.

[0042] A method for extracting a gingival three-dimensional model shell based on Marching Cube comprises the following steps:

[0043] Step 1: First, import the initial gingival 3D model and set it as M o, the bottom surface of the model is not closed, and all the bottom edge points are in the same plane. Find the half edge that is only used by one face to get all the bottom edge points, and set the set of all bottom edge points as PS b According to PS b The bottom plane normal can be calculated, and the transformation matrix can be further calculated to transform the initial gingival 3D model to a position where the bottom surface is parallel to the XoY plane (the XoY plane is the plane orthogonal to the X-axis and Y-axis of the world coordinate system), and then translated to a position where the geometric center coincides with the origin of the world coordinate system. Finally, the reset gingival model M is obtained. r , as shown in the two perspectives of Figure 2(a) and Figure 2(b).

[0044] Step 2: Based on M r The bottom edge of the PS is extruded 2.0mm in the negative direction of the Z axis. b Make a copy and get PS bc . Put PS bc The point is translated 2.0 mm in the negative direction of the Z axis, and then compared with the PS b Sew the patch point by point and add it to M r In the figure, the highly compensated gingival model M is obtained. h .like Figure 3 shown.

[0045] Step 3: The construction of the signed distance field (SDF) adopts the strategy of grid diffusion + hierarchical construction. First, the grid diffusion construction with a lower resolution is carried out. h The bounding box voxelization constructs a blank field, in which each voxel can be set to a value. Through the breadth-first search (BFS) algorithm, from M h The surface expands inward and outward layer by layer to fill the value of the corresponding voxel. Since the diffusion is discrete, parallel computing is used here, which is relatively efficient.

[0046] Assume the required thickness of the shell is T mm, find all voxels with a value of -T, and put them into the voxel set VS ori As shown in the shaded voxels in the cross-section of Figure 4(a), ori The voxels in are further subdivided by a hierarchical strategy to calculate the local subdivision signed distance field LDSDF. The value of each subvoxel is obtained by interpolating the values ​​of nearby voxels, as shown in the cross-section of Figure 4(b). Find all the subvoxels closest to -T value and use the classic MarchingCube to generate isosurfaces to obtain the preliminary inner shell gingival model, which is set as The effect is shown in Figure 5(a) and Figure 5(b) from two perspectives. The surface entity models in the two figures are

[0047] Step 4: Traverse All the patches, record the three vertex positions of each patch. Then delete the redundant and repeated vertices in the model, leaving only one vertex at each position, and delete all the patches at the same time. At this time, the model becomes a point cloud model. Table 1 shows the number of vertices and patches before and after deleting the redundant vertices and rebuilding the topology. It can be seen that the number of vertices in the second column of Table 1 is much less than that in the first column, and the number of patches is 0. Then, according to the three vertex positions and The remaining non-duplicate vertices in the reconstruct the topological structure. The number of vertices and faces is shown in the third column of Table 1. The model is no longer a bunch of scattered faces, but a mesh model with a half-edge data structure, which is called a manifold inner shell gum model, and is set to Later, we can use the characteristics of the half-edge data structure to for further processing.

[0048] Table 1

[0049]

[0050] Step 5: Traverse All vertices are checked for connectivity using the connectivity characteristics of the half-edge data structure, vertices with structural errors are found, and they are deleted to obtain the inner shell gingival model to be optimized, set as The wrong vertices may be vertices with unreasonable topological connections, or they may be two points of non-popular edges that are reused more than twice. Figure 6(a) shows a case of wrong vertices. The wrong vertices are removed, so holes appear, as shown in Figure 6(b). Next, traverse to find For all holes, call the hole repair algorithm to fill the holes.

[0051] The principle of the hole repair algorithm is to first calculate the average value l of the boundary edge length based on the hole boundary point information; then calculate the angle between the two adjacent edges of each boundary point; find the three boundary points where the two edges have the smallest angle, calculate the distance s between the two left and right boundary points among the three points, and determine whether s<2×l is true. If so, add a triangle as shown in Figure 7(a); if not, add two triangles as shown in Figure 7(b). Then find new boundary points to determine whether the hole still exists. If it still exists, continue the above process, otherwise end. The effects before and after repair are shown in Figure 8(a) and Figure 8(b), respectively. Because some models are relatively complex, the hole filling algorithm can only add a part of the face at a time and cannot be completely closed. Therefore, it is necessary to cyclically search for holes. If so, repair them until there are no holes and obtain a standard inner shell gingival model, set as

[0052] Step 6: Have a standard inner shell gum model Finally, a standard shell gingival model needs to be constructed. bc The point in the middle is M h The bottom edge point of PS bc Put M h The bottom surface is sutured and closed to obtain the closed shell gingival model The reason for closing M h Conclusion This is because the parameters of the Boolean addition operation must be closed models, otherwise an error will occur. Do a Boolean addition operation to get the Boolean rear shell gingival model Then remove The sutured bottom surface is obtained to obtain the unsealed standard shell gingival model As shown in Figure 9(b).

[0053] Step 7: and The models are merged together to obtain an open shell gingival model, which is set as Since the MarchingCube algorithm is not accurate in dealing with edge problems, There are still some random cutout issues on the inner shell, such as Fig.10 As shown. Cut off a part of the vertex at the bottom to get the cut-out shell gum model, set as Fig.11 The cut section is shown. After the vertices are cut, The edge points of the new bottom surface are displaced in the Z-axis direction, transformed to the same plane, and the new bottom surface is made parallel to the XoY plane to prepare for subsequent stitching.

[0054] Step 8: The two circles of bottom edge points of the inner shell and the outer shell are sorted according to the edge topology connection, and the sorting starting points of the two circles are guaranteed to correspond and the sorting directions are consistent. Then the two circles of bottom edge points are stitched point by point, and the obtained face is added to Finally, a closed perfect shell model is obtained, which is called the standard shell gingival model. As shown in Figure 12(b). This is the final output result of the method in this paper, as shown in the two perspectives of Figure 13(a) and Figure 13(b).

[0055] Example 2

[0056] The present embodiment provides a Marching Cube-based gingival three-dimensional model extraction device, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement a Marching Cube-based gingival three-dimensional model extraction method of embodiment 1.

[0057] Example 3

[0058] This embodiment relates to a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the method for extracting a gingival three-dimensional model based on Marching Cube described in Embodiment 1 is implemented.

[0059] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for extracting a gingival three-dimensional model based on Marching Cube, characterized in that: The following steps are involved: 1) Position correction: after inputting the initial gingival 3D model, first correct the position to the XoY plane to obtain the gingival model after reset; 2) Extruding downwards, after the gingival model is reset, the bottom edge of the gingival model is extruded downwards to obtain a highly compensated gingival model; 3) construct the distance field and inner shell, construct the signed distance field according to the height-compensated gingival model, and reconstruct the preliminary inner shell gingival model using the MarchingCube method; 4) Reconstructing the half-edge structure, while keeping the vertex topology of the patch unchanged, deleting the repeated vertices at the same position of the preliminary inner shell gingiva model and reconnecting the topology to obtain a manifold inner shell gingiva model; 5) Inner shell repair: find the wrong vertices of the manifold inner shell gingival model and delete them, then use the hole repair algorithm to repair its holes to obtain a standard inner shell gingival model; 6) Boolean operation, closing the bottom surface of the height-compensated gingival model, performing Boolean addition operation with all teeth, and splitting the bottom surface to obtain a standard shell gingival model; 7) Bottom cutting: after merging the standard outer shell gingival model and the standard inner shell gingival model, a part of the bottom is cut to obtain a cut shell gingival model; 8) Suturing the bottom edge: Suturing the bottom edge of the cut shell-drawing gingival model to obtain the final standard shell-drawing gingival model.

2. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: Step (1) The initial gingival 3D model is set as o The bottom surface of is not closed, and all the bottom edge points are in the same plane; find the half edge used by only one face to get all the bottom edge points, and set the set of all bottom edge points as PS b According to PS b The normal direction of the bottom plane can be calculated, and the transformation matrix can be further calculated to transform the initial gingival 3D model to a position where the bottom surface is parallel to the XoY plane. The XoY plane is the plane orthogonal to the X-axis and Y-axis of the world coordinate system, and it is translated to a position where the geometric center coincides with the origin of the world coordinate system. Finally, the reset gingival model M is obtained. r .

3. A method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: The downward extrusion in step (2) includes: based on M r The bottom edge of the PS is extruded 2.0mm in the negative direction of the Z axis; specifically: first b Make a copy and get PS bc ; Put PS bc The point is translated 2.0 mm in the negative direction of the Z axis, and then compared with the PS b Sew the patch point by point and add it to M r In the figure, the highly compensated gingival model M is obtained. h .

4. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: The construction of the signed distance field SDF described in step (3) adopts the strategy of grid diffusion + hierarchical construction: first, the grid diffusion construction with a lower resolution is carried out, and then the M h The bounding box voxelization constructs a blank field, in which each voxel can be set to a value, and the BFS breadth-first search algorithm is used to search from M h The surface expands inward and outward layer by layer to fill the value of the corresponding voxel. Since the diffusion is discrete, parallel calculation is used; Assume the required thickness of the shell is T mm, find all voxels with a value of -T, and put them into the voxel set VS ori Next, VS ori The voxels in are further subdivided by a hierarchical strategy to calculate the local subdivision signed distance field LDSDF. The value of each subvoxel is obtained by interpolating the values ​​of nearby voxels. Find all the subvoxels closest to -T value and use the classic MarchingCube to generate isosurfaces to obtain the preliminary inner shell gingival model, which is set as 5. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: The rule for reconnecting the topology in step (4) is: traverse All the patches, record the three vertex positions of each patch; then delete the redundant duplicate vertices in the model, leaving only one vertex at each position, and delete all the patches at the same time. At this time, the model becomes a point cloud model; then according to the three vertex positions and The remaining non-duplicate vertices in the patch are rebuilt to rebuild the patch topology. It has a light number of vertices and the same number of faces as before; the model at this time is a mesh model with a half-edge data structure, called a manifold inner shell gum model Using the characteristics of the half-edge data structure for further processing.

6. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: The deleted manifold inner shell gingival model described in step (5) Specifically, all the wrong vertices of the inner shell are traversed first, and the connection characteristics of the half-edge data structure are used to check the connection of the vertices, find out the vertices with structural errors, and delete them to obtain the inner shell gingival model to be optimized, which is set as The wrong vertices may be vertices with unreasonable topological connections, or they may be two points of non-popular edges that are reused more than twice; The wrong vertices are removed, so holes will appear; next, traverse to find For all holes, call the hole repair algorithm to fill the holes. The hole repair algorithm is: first, according to the hole boundary point information, calculate the average value l of the boundary edge length; then calculate the angle between the two adjacent edges of each boundary point; find the three boundary points where the two edges have the smallest angle, calculate the distance s between the two left and right boundary points among the three points, and determine whether s<2×l is true. If so, add a triangle, if not, add two triangles; then find new boundary points to determine whether the hole still exists. If so, continue the above process, otherwise end; some models are more complex, and the hole filling algorithm can only add a part of the face at a time and cannot be completely closed, so it is necessary to loop to find out whether there is a hole, and if so, repair it until there is no hole, and get a standard inner shell gingival model 7. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: Step (6) According to PS bc Put the height compensation gingival model M h The bottom surface is sutured and closed to obtain the closed shell gingival model Combine all the tooth models of this maxillofacial face with Do a Boolean addition operation to get the Boolean rear shell gingival model Then remove The sutured bottom surface is obtained to obtain the unsealed standard shell gingival model 8. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: The step (7) of merging the standard outer shell gingival model and the standard inner shell gingival model means: the standard inner shell gingival model is combined with the standard inner shell gingival model. and standard shell gingival model The models are merged together to obtain an open shell gingival model, which is set as Bundle Cut off a part of the vertex at the bottom to get the cut-out shell gum model, set as After the vertices are trimmed, The edge points of the new bottom surface are displaced in the Z-axis direction, transformed to the same plane, and the new bottom surface is made parallel to the XoY plane to prepare for subsequent stitching.

9. The method for extracting a gingival three-dimensional model based on Marching Cube according to claim 1, characterized in that: Step (8) of sewing the bottom edge of the cut and extracted gingival model comprises: The two circles of bottom edge points of the inner shell part and the outer shell part are sorted according to the edge topology connection, and the sorting starting points of the two circles are guaranteed to correspond and the sorting directions are consistent; then the two circles of bottom edge points are stitched point by point, and the obtained facets are added to Finally, a closed perfect shell model is obtained, which is called the standard shell gingival model. This is the final output result of this method.

10. A gingival three-dimensional model extraction device based on Marching Cube, characterized in that: It comprises a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, they are used to implement a method for extracting gingival three-dimensional models based on Marching Cube as described in any one of claims 1-9.

Images