A method for judging quality of three-dimensional point cloud based on real-time grid

Abstract

The application relates to the field of three-dimensional point cloud quality judgment, and discloses a three-dimensional point cloud quality judgment method based on real-time grids, which comprises the following steps: S1, a hidden function value based on volume is calculated; S2, all voxels in a scanning space are traversed, and voxels_select capable of generating triangular faces are extracted; S3, for each voxel selected, the number of triangular faces is judged, and the judgment method is an algorithm MarchingCubes; according to the TSDF values of eight vertices of the voxle, a look-up table is searched; S4, all voxels_select are traversed, and all triangular faces are generated by using the MarchingCubes method. The three-dimensional point cloud quality judgment method based on real-time grids can be used for judging the quality of three-dimensional point clouds, the judgment method provided by the application can timely remove poor point clouds, avoid error accumulation, improve the overall scanning precision, and solve the problems in the prior art.

Description

Technical Field

[0001] This invention relates to the field of 3D point cloud quality assessment technology, and in particular to a method for assessing the quality of 3D point clouds based on real-time meshes. Background Technology

[0002] In the field of 3D scanning equipment, the workflow is as follows: 3D point clouds are reconstructed by acquiring 2D image information; the 3D point clouds are then stitched together according to their characteristics to form a complete point cloud in a unified coordinate system; finally, a 3D mesh is generated from the 3D point clouds to complete the digitization of the real-world object. In real-time 3D scanning applications, WYSIWYG real-time rendering is a crucial step.

[0003] In real-time scanning, user interaction can be achieved through point cloud rendering or mesh rendering. Point cloud rendering is faster than mesh rendering and can meet the requirements of real-time rendering. However, in many scenarios, 3D mesh is the primary method for real-time rendering and final output in the field of 3D scanning equipment. The main reasons for this are as follows:

[0004] (1) 3D mesh is the basis of 3D printing

[0005] (2) 3D meshes, through rendering, can effectively reflect the characteristics of real objects.

[0006] (3) 3D meshes have geometric topological properties and can effectively express connectivity.

[0007] (4) 3D meshes can effectively perform texture mapping and realistically reflect the surface of objects.

[0008] Real-time rendering of 3D meshes can improve scanning quality and enhance the user experience.

[0009] The real-time rendering of 3D meshes during 3D scanning, which instantly presents the scanned data, can effectively improve scanning navigation; its geometric topological characteristics can help the algorithm remove noise in a timely and effective manner, avoiding the accumulation of noisy data.

[0010] Therefore, real-time grid mapping during scanning can not only improve scanning accuracy but also greatly enhance the user experience, making it an essential feature for many 3D scanning equipment products.

[0011] During actual scanning, noise can arise due to factors such as lighting, saliva bubbles, camera calibration, and improper operation, leading to poor quality (accuracy) of some 3D reconstructed point clouds. In severe cases, this can affect the accuracy of the entire scan. However, judging the quality of a point cloud generally requires relying on the neighborhood relationship of that point, which is quite difficult.

[0012] Real-time 3D meshes are generated from 3D point clouds. The generation process involves understanding the neighborhood relationships of the point clouds. By identifying the mesh quality during generation and then mapping it to the corresponding spatial location in the point cloud, the quality of the point cloud can be assessed. This situation can generally be resolved by rescanning; more severe cases can be addressed by cropping the already scanned data and rescanning. Summary of the Invention

[0013] (a) Technical problems to be solved

[0014] To address the shortcomings of existing technologies, this invention provides a method for judging the quality of 3D point clouds based on real-time meshes. This method promptly removes poor-quality point clouds, avoids error accumulation, and improves the overall scanning accuracy. It solves the problem in existing technologies where the generation of meshes uses point clouds and their neighborhood relationships. Therefore, for point clouds with poor local quality, relatively smooth meshes are generated due to the relationship between their neighboring points, resulting in incorrect navigation effects.

[0015] (II) Technical Solution

[0016] This invention provides the following technical solution: a method for judging the quality of 3D point clouds based on real-time meshes, comprising the following steps:

[0017] S1. All voxels in the implicit function value computation space based on volume;

[0018] S2. Traverse and scan all voxels in the space, and extract voxels_select that can generate triangular faces within the voxels;

[0019] S3. For each extracted voxels_select, determine how many triangles it has. The method for this determination is the MarchingCubes algorithm, which includes searching for the number of triangles in each voxels_select based on the TSDF values ​​of the eight vertices of the voxel.

[0020] S4. Traverse all voxels_select and use the MarchingCubes method to generate the triangles of all voxels_select;

[0021] S5. Traverse all triangular faces;

[0022] S6. Use the UI to render and mark the triangular facets, and display the rendered and marked triangular facets in a prominent manner;

[0023] S7. Based on the real-time mesh rendering prompts, perform a rescan until the triangle face prompts disappear;

[0024] S8. If the triangular face prompt does not disappear after repeated rescanning, prompt the user to crop the area and rescan it, or crop the area and improve the scanning environment.

[0025] The present invention provides a method for judging the quality of 3D point clouds based on real-time meshes, which can be used to judge the quality of 3D point clouds. The judgment method provided by the present invention can promptly remove bad point clouds, avoid error accumulation, and improve the overall accuracy of scanning. It is used to solve the problem in the prior art that, due to the relationship between the point cloud and its neighborhood used in the generation of the mesh, a relatively smooth mesh is generated for point clouds with poor local quality due to the relationship between its neighboring points, resulting in incorrect navigation effects.

[0026] In one possible implementation, the specific steps in S1 are as follows:

[0027] S1.1 The volume method divides the scan space into small cubes called voxels; a scan space is composed of many voxels.

[0028] S1.2 Each voxel has eight vertices. Based on the coordinates and normals of the point cloud, calculate the TSDF (scalar value) of the eight vertices of the voxel in the neighborhood of the point cloud.

[0029] S1.3 The triangles of the real-time mesh are generated by all voxels, and the vertices of the triangles are all on the 12 edges of the voxel.

[0030] In one possible implementation, the voxel calculation step for the point cloud neighborhood in S1.2 is as follows:

[0031] S1.21. For each vertex of voxel, take its coordinates as q(x1,y1,z1); the coordinates of point cloud p are p(x2,y2,z2), and the normal of p is pn(xn,yn,zn).

[0032] S1.22, Calculate sdf = (pq) * q = xn(x1-x2) + yn(y1-y2) + zn(z1-z2);

[0033] S1.23. After truncating sdf, we get TSDF. The truncation rule is fmin <= sdf and sdf <= fmax, where fmin and fmax are set according to the actual situation.

[0034] In one possible implementation, the MarchingCubes algorithm in S3 is used to determine whether there are triangular faces based on the signs of the eight vertices of the voxel.

[0035] In one possible implementation, the specific operation steps in S3 are as follows:

[0036] S3.1. Traverse each triangle and calculate the gradient of the edge e(p1, p2) containing each vertex of the triangle (the difference of the TSDF of the two vertices p1 and p2 of edge e / the length of edge e).

[0037] If the gradient of S3.2 and e is less than the set threshold, then mark the vertex.

[0038] In one possible implementation, in S5, if one vertex of a triangle is marked, then the triangle is marked.

[0039] Compared with existing technologies, this invention provides a method for judging the quality of 3D point clouds based on real-time meshes, which has the following beneficial effects:

[0040] This invention assesses the quality of local meshes based on real-time mesh data and provides prompts on the mesh rendering interface (e.g., displaying poor-quality triangles in a warning color). Users can then rescan based on these prompts until the problematic mesh disappears. This effectively clears up poor-quality point clouds, preventing error accumulation and improving overall scanning accuracy. It also significantly improves scanning smoothness, reduces user assessment time, and enhances the user experience.

[0041] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit the invention. Attached Figure Description

[0042] Figure 1 This invention provides a flowchart for point cloud quality assessment based on real-time mesh in 3D point cloud quality evaluation. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0044] Examples of embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0045] like Figure 1 As shown, this invention provides a method for judging the quality of 3D point clouds based on real-time meshes, including the following steps:

[0046] S1. All voxels in the implicit function value computation space based on volume;

[0047] S1.1 The volume method divides the scan space into small cubes called voxels; a scan space is composed of many voxels.

[0048] S1.2 Each voxel has eight vertices. Based on the coordinates and normals of the point cloud, calculate the TSDF (scalar value) of the eight vertices of the voxel in the neighborhood of the point cloud.

[0049] The steps for calculating the voxel of a point cloud neighborhood are as follows:

[0050] S1.21. For each vertex of voxel, take its coordinates as q(x1,y1,z1); the coordinates of point cloud p are p(x2,y2,z2), and the normal of p is pn(xn,yn,zn).

[0051] S1.22, Calculate sdf = (pq) * q = xn(x1-x2) + yn(y1-y2) + zn(z1-z2);

[0052] S1.23. After truncating sdf, we get TSDF. The truncation rule is fmin <= sdf and sdf <= fmax, where fmin and fmax are set according to the actual situation.

[0053] S1.3. Triangles in the real-time mesh are generated by all voxels. Some voxels do not generate triangles, while others may generate multiple triangles. The vertices of the triangles are all on the 12 edges of the voxel.

[0054] S2. Traverse and scan all voxels in the space, and extract voxels_select that can generate triangular faces within the voxels;

[0055] S3. For each extracted voxels_select, determine how many triangles it has. The method for this determination is the MarchingCubes algorithm, which includes searching for the number of triangles in each voxels_select based on the TSDF values ​​of the eight vertices of the voxel.

[0056] S3.1. Traverse each triangle and calculate the gradient of the edge e(p1, p2) containing each vertex of the triangle (the difference of the TSDF of the two vertices p1 and p2 of edge e / the length of edge e).

[0057] If the gradient of S3.2 and e is less than the set threshold, then mark the vertex.

[0058] S4. Traverse all voxels_select and use the MarchingCubes method to generate the triangles of all voxels_select;

[0059] S5. Traverse all triangular faces;

[0060] If one vertex of a triangle is marked, then that triangle is marked.

[0061] S6. Use the UI to render and mark the triangular facets, and display the rendered and marked triangular facets in a prominent manner;

[0062] S7. Based on the real-time mesh rendering prompts, perform a rescan until the triangle face prompts disappear;

[0063] S8. If the triangular face prompt does not disappear after repeated rescanning, prompt the user to crop the area and rescan it, or crop the area and improve the scanning environment.

[0064] The present invention provides a method for judging the quality of 3D point clouds based on real-time meshes, which can be used to judge the quality of 3D point clouds. The judgment method provided by the present invention can promptly remove bad point clouds, avoid error accumulation, and improve the overall accuracy of scanning. It is used to solve the problem in the prior art that, due to the relationship between the point cloud and its neighborhood used in the generation of the mesh, a relatively smooth mesh is generated for point clouds with poor local quality due to the relationship between its neighboring points, resulting in incorrect navigation effects.

[0065] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0066] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0067] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

Claims

1. A method for judging the quality of a three-dimensional point cloud based on real-time mesh, comprising the following steps: S1, calculating all voxels in a space based on the volume implicit function value; S2, traversing all voxels in the space, extracting voxels_select that can generate triangular faces in the voxels; S3, for each voxels_select extracted, judging how many triangular faces it has, the judging method being the MarchingCubes algorithm, the MarchingCubes algorithm comprising searching for the number of triangular faces of each voxels_select according to the TSDF values of the eight vertices of the voxel; S4, traversing all voxels_select, generating triangular faces of all voxels_select using the MarchingCubes method; S5, traversing all triangular faces; S6, rendering and marking the triangular faces using UI, and displaying the rendered and marked triangular faces in a conspicuous manner; S7, according to the real-time mesh rendering prompt, performing rescan until the triangular face prompt disappears; S8, if the triangular face prompt does not disappear after repeated rescan, prompting the user to crop the space region where the voxels_select corresponding to the triangular face prompt is located, and then rescan the space region where the voxels_select is located, or crop and improve the scanning environment.

2. The method for judging the quality of a three-dimensional point cloud based on a real-time grid according to claim 1, characterized in that, In S1, the specific steps are as follows: S1.1, the volume is to divide the scanning space into small cubes called voxels; a scanning space is composed of many voxels; S1.2, each voxel has eight vertices, and the TSDF values of the eight vertices of the voxel in the point cloud neighborhood are calculated according to the coordinates and normal of the point cloud; S1.3, the triangular faces of the real-time mesh are generated by all voxels, and the vertices of the triangular faces are on the 12 edges of the voxel.

3. The method for judging the quality of a three-dimensional point cloud based on a real-time grid according to claim 2, characterized in that, In S1.2, the steps of calculating the voxels in the point cloud neighborhood are as follows: S1.21, for each vertex of the voxel, taking its coordinates as q (x1, y1, z1); the coordinates of the point cloud p are p (x2, y2, z2), and the normal of p is pn (xn, yn, zn); S1.22, compute sdf = (p - q) q = xn(x1- x2) + yn(y1- y2) + zn(z1- z2); S1.23, obtaining TSDF after truncating sdf, the truncation rule being fmin<=sdf and sdf<=fmax, wherein fmin and fmax are set according to the actual situation.

4. The method for judging the quality of a three-dimensional point cloud based on a real-time grid according to claim 1, characterized in that, In S3, the MarchingCubes algorithm is used to judge whether there is a triangular face according to the sign of the eight vertices of the voxel.

5. The method for judging the quality of a three-dimensional point cloud based on real-time mesh according to claim 4, characterized in that, In S3, the specific operation steps are as follows: S3.1, traversing each triangular face, calculating the gradient of the edge e (p1, p2) where each vertex of the triangular face is located; S3.2, when the gradient of e is less than the set threshold threshold, the vertex is marked.

6. The method for judging the quality of a three-dimensional point cloud based on real-time mesh according to claim 1, characterized in that, In S5, if a triangular face has a vertex marked, the triangular face is marked.

Images