Grid optimization method based on subject correlation

Abstract

The invention discloses a grid optimization method based on subject correlation, and the method specifically comprises the following steps: S1, obtaining a bitmap Array: inputting a to-be-optimized grid and a point cloud adjacency list, and carrying out the subject correlation construction of the to-be-optimized grid; s2, obtaining an effective main body grid: carrying out cardinal number sorting on the bitmap to obtain a vertex subscript with the maximum cardinal number, retaining the vertex of the part, obtaining a neighborhood radius r according to an adjacency list, and further filtering the grid by selecting a proper threshold value alpha; s3, reducing the storage occupation of the grid: traversing the grid to find an optimal adjacent point pair meeting the condition that the normal included angle is less than 30 degrees, and performing edge shrinkage on the optimal adjacent point pair; according to the method, the main grid is simplified, the number of grid vertexes and triangular faces is further reduced, the conciseness and processing efficiency of the grid are improved, the processing and analysis difficulty of the grid model is reduced, the application efficiency and performance are improved, and the processing speed is increased.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and specifically provides a mesh optimization method based on subject relevance. Background Art

[0002] Museums are places for collecting, protecting, and displaying various important cultural relics and specimens. They are an important part of cultural infrastructure and an important educational base for improving people's cultural qualities. Using 3D reconstruction to digitally transform museums and build a web-based digital museum system can better solve the contradiction between rich exhibits and limited exhibition space and time, further explore new fields for the protection, research, and display of museum collections, and is of extremely important practical significance and necessity for realizing resource sharing, protecting precious museum resources, spreading ancient Chinese culture, and strengthening exchanges and cooperation among experts in domestic and foreign fields.

[0003] Deficiencies of the prior art:

[0004] In the actual application of meshes, many challenges are often faced. First, when meshes are used for display on web pages, they face the problem of excessive storage space occupation, resulting in slow transmission, which affects user experience and data transmission efficiency. Second, the reconstructed meshes often contain a large amount of irrelevant background content, increasing the complexity of the mesh model and storage space occupation. These irrelevant background contents not only waste storage resources but also increase the difficulty of processing and analyzing the mesh model, reducing the efficiency and performance of the application. In addition, when the mesh model is too large, it will cause a long waiting time for secondary applications or even a stuck phenomenon, making secondary applications impossible. For example, when the mesh model is used for problems such as mesh line drawing extraction and model retrieval, an overly large mesh model will slow down the processing speed of secondary applications and even affect the usability and stability of the application. Summary of the Invention

[0005] The purpose of the present invention is to provide a mesh optimization method based on subject relevance to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A mesh optimization method based on subject relevance, and this mesh optimization method specifically includes the following steps:

[0007] S1. Obtain the bitmap Array: Input the mesh to be optimized and the point cloud adjacency list, and construct the subject relevance of the mesh to be optimized.

[0008] S2. Obtain the effective subject mesh: Perform radix sorting on the bitmap, obtain the vertex subscript with the largest radix and retain this part of the vertices, and then obtain the neighborhood radius r according to the adjacency list, and select an appropriate threshold α to further filter the mesh.

[0009] S3. Reduce the storage occupancy of the mesh: Traverse the mesh to find the optimal adjacent point pairs with a normal angle less than 30 degrees and perform edge contraction on them.

[0010] Preferably, the step S1 specifically includes the following steps:

[0011] a1. Input the mesh to be optimized and the point cloud adjacency list. The mesh to be optimized is obtained by Poisson reconstruction from the dense point cloud, and the point cloud adjacency list records the indices of the 40 nearest adjacent vertices for each vertex in the mesh.

[0012] a2. Create a bitmap array Array and a queue Queue. The size of the array is equal to the number of vertices in the mesh, and each element of the array is initialized to its current subscript value.

[0013] a3. Obtain the bitmap Array, randomly select an unmarked vertex in the mesh as the root node, add the indices of the vertices directly connected to the root node to the queue, and change the corresponding Array element value to the index of the root node. At the same time, mark the root node and the nodes connected to the root node, and process all the vertices in turn.

[0014] Preferably, the step S2 includes the following steps:

[0015] b1. Perform radix sort on Array.

[0016] b2. Find the array subscript with the largest radix and mark it as the main mesh vertex, and delete the non-main mesh vertices and edges.

[0017] b3. Calculate the neighborhood radius r value according to the adjacency list:

[0018]

[0019]

[0020] where x c , y c , z c are the coordinates of the center point c, x i , y i , z i are the vertex coordinates of the 40 neighborhood vertices of point c, and n is the number of neighborhood vertices of point c.

[0021] b4. Select an appropriate α threshold. α represents the number of neighborhood vertices around vertex c with r as the radius centered at vertex c. α is related to the point cloud adjacency list: The larger α is, the more vertices will be deleted, which may damage the normal main mesh; conversely, if α is too small, the mesh removal effect is not obvious.

[0022] b5. Select a vertex v that is not in the adjacency list;

[0023] b6. Calculate the number s of vertices contained within the sphere centered at v with radius r;

[0024] b7. Judge:

[0025]

[0026] b8. Delete the vertex v and the corresponding edges.

[0027] Preferably, the step S3 specifically includes the following steps:

[0028] c1. Find adjacent point pairs v1 and v2;

[0029] c2. Judge:

[0030]

[0031] where normal1 is the normal direction of vertex v1;

[0032] c3. Merge the adjacent point pairs to obtain the contracted point The contracted point is obtained by the following formula:

[0033]

[0034] where plane(v1) is the mesh face where vertex v1 is located, and plane(v2) is the mesh face where vertex v2 is located;

[0035] c4. Update the contracted point and the mesh edges with surrounding vertices, and calculate the RGB value of the new vertex. The new RGB value is obtained by taking the average of the RGB values of the two vertices in the adjacent point pair.

[0036] Preferably, the mesh input in the step a1 is a triangular patch mesh, accompanied by the normal direction of each vertex. The point cloud adjacency list is a linked list array structure. The array length is equal to the number of three-dimensional points. The array subscript is the index of the three-dimensional point. The pointer pointed to by the array element is the address of the linked list. The linked list stores the indices of 40 adjacent three-dimensional points corresponding to the three-dimensional point in the array.

[0037] Preferably, the bitmap array Array in the step a2 is a one-dimensional array.

[0038] Preferably, the value range of α in the step b4 is [12, 36].

[0039] Preferably, the main body relevance in the step S1 refers to the mesh vertices related to the main body object, which is represented by the bitmap Array; Poisson reconstruction of a dense point cloud with irrelevant background noise will generate both the main body mesh and the background mesh simultaneously.

[0040] Preferably, for the effective main body grid in step S2, a closed grid is generated by Poisson reconstruction. The vertices of the newly generated grid are not in the original dense point cloud, and the effective main body grid is obtained by filtering the closed grid with α.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] The grid optimization method based on main body correlation specifically includes the following steps: S1. Obtain the bitmap Array: Input the grid to be optimized and the point cloud adjacency list, and construct the main body correlation for the grid to be optimized; S2. Obtain the effective main body grid: Perform radix sorting on the bitmap, obtain the vertex subscript with the largest radix and retain this part of the vertices, then obtain the neighborhood radius r according to the adjacency list, and select an appropriate threshold α to further filter the grid; S3. Reduce the storage occupancy of the grid: Traverse the grid to find the optimal adjacent point pairs that satisfy the normal angle less than 30 degrees and perform edge contraction on them. Overall, simplify the main body grid, further reduce the number of grid vertices and triangular faces, improve the simplicity and processing efficiency of the grid, reduce the processing and analysis difficulty of the grid model, increase the application efficiency and performance, and speed up the processing speed. Description of the Drawings

[0043] Figure 1 It is a flowchart of the grid optimization method based on main body correlation provided by the embodiment of the present invention;

[0044] Figure 2 It is an effect diagram of grid optimization provided by the embodiment of the present invention; Detailed Embodiments

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0046] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation of the present invention.

[0047] In the description of this patent, it should be noted that unless otherwise clearly specified and defined, the terms "installation", "connection", and "setting" should be understood in a broad sense. For example, it can be fixedly connected and set, or it can be detachably connected and set, or integrally connected and set. For those of ordinary skill in the art, the specific meanings of the above terms in this patent can be understood according to specific circumstances.

[0048] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, the meaning of "several" is two or more unless otherwise clearly and specifically defined.

[0049] Embodiment

[0050] Please refer to Figure 1-2 As shown, a technical solution of a grid optimization method based on object relevance provided by the present invention: This grid optimization method specifically includes the following steps:

[0051] S1. Obtain a bitmap Array: Input the grid to be optimized and the point cloud adjacency list, and construct the object relevance of the grid to be optimized. The object relevance refers to the grid vertices related to the object, which is represented by the bitmap Array; Poisson reconstruction of the dense point cloud with irrelevant background noise will generate both the object grid and the background grid simultaneously.

[0052] a1. Input the grid to be optimized and the point cloud adjacency list. The grid to be optimized is obtained by Poisson reconstruction of the dense point cloud. The point cloud adjacency list records the indices of the 40 nearest adjacent vertices of each vertex in the grid. The input grid is a triangular mesh with the normal direction of each vertex attached. The point cloud adjacency list is a linked list array structure, the length of the array is equal to the number of three-dimensional points, the array subscript is the index of the three-dimensional point, and the pointer pointed to by the array element is the address of the linked list. The linked list stores the indices of the 40 adjacent three-dimensional points corresponding to the three-dimensional point in the array.

[0053] a2. Create a bitmap array Array and a queue Queue. The bitmap array Array is a one-dimensional array, and the size of the array is equal to the number of vertices in the grid. Each element of the array is initialized to its current subscript value. The queue is used to store the indices of the vertices to be processed. The array and the queue are created by numpy.

[0054] a3. Obtain the bitmap Array. Randomly select a vertex from the unmarked grid vertices as the root node. Add the indices of the vertices directly connected to the root node to the queue, and change the corresponding Array element value to the index of the root node. At the same time, mark the root node and the nodes connected to the root node, and process all vertices in sequence;

[0055] S2. Obtain the effective main grid: Perform radix sorting on the bitmap to obtain the vertex subscript with the largest radix and retain this part of the vertices. Then, obtain the neighborhood radius r according to the adjacency list, select an appropriate threshold α to further filter the grid, and generate a closed grid through Poisson reconstruction. The newly generated grid vertices are not in the original dense point cloud. Filter the closed grid through α to obtain the effective main grid;

[0056] b1. Perform radix sorting on Array;

[0057] b2. Find the array subscript with the largest radix and mark it as the main grid vertex. Delete the non-main grid vertices and edges;

[0058] b3. Calculate the neighborhood radius r value according to the adjacency list:

[0059]

[0060] where x c , y c , z c are the coordinates of the center point c, x i , y i , z i are the vertex coordinates among the 40 neighborhood vertices of point c, and n is the number of neighborhood vertices of point c;

[0061] b4. Select an appropriate α threshold. α represents the number of neighborhood vertices around vertex c with r as the radius centered at vertex c. α is related to the point cloud adjacency list, and the value range of α is [12, 36]; the larger α is, the more vertices will be deleted, which may damage the normal main grid; conversely, if α is too small, the effect of grid removal is not obvious;

[0062] b5. Select a vertex v not in the adjacency list;

[0063] b6. Calculate the number s of vertices contained in the sphere with v as the center and r as the radius;

[0064] b7. Judge:

[0065]

[0066] b8. Delete vertex v and the corresponding edges;

[0067] S3. Reduce the storage occupancy of the mesh: Traverse the mesh to find the optimal adjacent point pair with a normal angle less than 30 degrees and perform edge contraction on it;

[0068] c1. Find adjacent point pair v1 and v2 with the shortest distance between them;

[0069] c2. Judge:

[0070]

[0071] where normal1 is the normal direction of vertex v1;

[0072] c3. Merge the adjacent point pair to obtain the contracted point The contracted point is obtained by the following formula:

[0073]

[0074] where plane(v1) is the mesh face where vertex v1 is located, and plane(v2) is the mesh face where vertex v2 is located;

[0075] c4. Update the mesh edges of the contracted point and the surrounding vertices, and calculate the RGB value of the new vertex. The new RGB value is obtained by taking the average of the RGB values of the two vertices in the adjacent point pair.

[0076] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and do not limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A mesh optimization method based on subject correlation, characterized in that: The grid optimization method specifically includes the following steps: S1. Obtain bitmap Array: Input the mesh to be optimized and the point cloud adjacency list, and construct the main correlation of the mesh to be optimized; S2. Get a valid main mesh: sort the bitmap by radix, get the vertex subscript with the largest radix and keep the vertices, get the neighborhood radius r according to the adjacency table, and select an appropriate threshold α to further filter the mesh; S3. Reduce the storage occupancy of the grid: traverse the grid to find the optimal adjacent point pair that satisfies the normal angle less than 30 degrees and perform edge contraction on it.

2. The mesh optimization method based on subject correlation according to claim 1, characterized in that: The step S1 specifically includes the following steps: a1. Input the mesh to be optimized and the point cloud adjacency list. The mesh to be optimized is obtained by Poisson reconstruction of the dense point cloud. The point cloud adjacency list records the indexes of the 40 nearest adjacent vertices of each vertex in the mesh. a2. Create a bitmap array Array and a queue Queue. The size of the array is equal to the number of vertices in the grid, and initialize each element of the array to its current index value; a3. Get the bitmap Array, randomly select a vertex from the unmarked mesh vertices as the root node, add the index of the vertex directly connected to the root node to the queue, and change the corresponding Array element value to the index of the root node. At the same time, mark the root node and the nodes connected to the root node, and process all the vertices in sequence.

3. The mesh optimization method based on subject correlation according to claim 1, characterized in that: The step S2 comprises the following steps: b1. Sort the Array by its base number. b2. Find the array subscript with the largest cardinality and mark it as the main mesh vertex, and delete the non-main mesh vertices and edges; b3. Calculate the neighborhood radius r value based on the adjacency table: Among them, x c ,y c , z c is the coordinate of the center point c, x i ,y i , z i are the vertex coordinates of the 40 neighboring vertices of point c, and n is the number of neighboring vertices of point c; b4. Select an appropriate α threshold. α represents the number of neighborhood vertices around vertex c with r as the radius and vertex c as the center. α is related to the point cloud adjacency table: the larger α is, the more vertices will be deleted, which may destroy the normal main mesh. On the contrary, if α is too small, the mesh culling effect is not obvious. b5. Select a vertex v that is not in the adjacency list; b6. Calculate the number of vertices s contained in a sphere with v as the center and r as the radius; b7. Judgment: b8. Delete vertex v and the corresponding edge.

4. The mesh optimization method based on subject correlation according to claim 1, characterized in that: The step S3 specifically includes the following steps: c1, find the adjacent point pairs v1 and v2; c2. Judgment: Among them, normal1 is the normal direction of vertex v1; c3. Merge adjacent points to get contraction points The pinch point is obtained by: Among them, plane(v1) is the mesh surface where vertex v1 is located, and plane(v2) is the mesh surface where vertex v2 is located; c4. Update the mesh edges of the contraction point and the surrounding vertices, and calculate the RGB value of the new vertex. The new RGB value is obtained by taking the average of the RGB values ​​of the two vertices in the adjacent point pair.

5. The mesh optimization method based on subject correlation according to claim 2, characterized in that: The mesh input in step a1 is a triangular patch mesh with the normal direction of each vertex. The point cloud adjacency list is a linked list array structure. The array length is equal to the number of three-dimensional points. The array subscript is the index of the three-dimensional point. The pointer pointed to by the array element is the address of the linked list. The linked list stores the indexes of 40 adjacent three-dimensional points corresponding to the three-dimensional point in the array.

6. The mesh optimization method based on subject correlation according to claim 2, characterized in that: The bitmap array Array in step a2 is a one-dimensional array.

7. The mesh optimization method based on subject correlation according to claim 3, characterized in that: The value range of α in step b4 is [12, 36].

8. The mesh optimization method based on subject correlation according to claim 1, characterized in that: The subject correlation in step S1 refers to the mesh vertices related to the subject object, which are represented by the bitmap Arra y Representation; Poisson reconstruction of a dense point cloud with irrelevant background noise will generate both the subject mesh and the background mesh.

9. The mesh optimization method based on subject correlation according to claim 1, characterized in that: The effective main mesh in step S2 is Poisson reconstructed to generate a closed mesh. The newly generated mesh vertices are not in the original dense point cloud. The closed mesh is filtered by α to obtain a valid main mesh.

Images