Method and apparatus for texture optimization of fine three-dimensional surface model

Abstract

The application discloses a fine three-dimensional surface model texture optimization method, comprising the following steps: clustering textures in a fine three-dimensional surface model based on a density clustering algorithm to obtain a first clustering result; judging each point in the textures, if the current point is an unclassified point, assigning the current point a classification number, and judging whether the rest of the points and the current point belong to the same triangular patch, if they belong to the same triangular patch, no matter which cluster of the first clustering result the rest of the points are located in, the classification number of the rest of the points is modified to the classification number of the current point, and a second clustering result is obtained; calculating the minimum circumscribed rectangle of each cluster in the second clustering result to determine effective textures. The method has a better redundant texture removal effect, and the optimized textures are suitable for efficient transmission and rapid visualization in a Web environment.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional data processing. More specifically, this invention relates to a method and apparatus for optimizing the texture of a fine three-dimensional surface model. Background Technology

[0002] In addition to complex geometric data, detailed 3D surface models often contain numerous high-resolution textures. The large number of textures and their high memory consumption lead to low network transmission efficiency. Furthermore, rendering each texture requires executing a GPU instruction, severely increasing the GPU load. This ultimately results in excessively long model loading times and low rendering efficiency, making efficient transmission and rapid visualization in a web environment impossible.

[0003] Existing texture optimization methods typically involve directly calculating the minimum bounding rectangle of the model's texture coordinates, using this bounding rectangle as the effective region to extract valid texture, and then performing 2D binning on the valid texture. While this method can optimize texture to some extent, directly calculating the minimum bounding rectangle of the texture coordinates to determine the effective texture region does not guarantee that the effective texture region will be minimized. Experiments show that the texture data corresponding to the geometric parts of many detailed building models are discretely distributed on the texture surface. This means that if the minimum bounding rectangle is directly calculated for the model's texture coordinates, the truly effective texture region only exists in a few parts of this rectangle. Therefore, the effective texture region determined by directly calculating the bounding rectangle still contains a large number of redundant texture regions. Summary of the Invention

[0004] One object of the present invention is to provide a method for optimizing the texture of a fine three-dimensional surface model to solve the above-mentioned problems.

[0005] To achieve the objectives and other advantages of this invention, a method for optimizing the texture of a fine three-dimensional surface model is provided, comprising:

[0006] The density clustering algorithm is used to cluster the texture in the fine 3D surface model to obtain the first clustering result;

[0007] For each point in the texture, if the current point is an unclassified point, then assign a classification number to the current point and determine whether the remaining points belong to the same triangular facet as the current point. If they belong to the same triangular facet, then regardless of which cluster the remaining points are in the first clustering result, modify the classification number of the remaining points to the classification number of the current point to obtain the second clustering result.

[0008] Calculate the minimum bounding rectangle of each cluster in the second clustering result to determine the effective texture.

[0009] Preferably, the fine three-dimensional surface model texture optimization method further includes: before clustering the texture, standardizing the texture coordinates of the texture so that the texture coordinates are in the range of [0,1].

[0010] Preferably, the fine three-dimensional surface model texture optimization method further includes: recombining the effective texture using a two-dimensional rectangular binning method to obtain a recombined texture.

[0011] Preferably, the fine three-dimensional surface model texture optimization method further includes: remapping the texture blocks in the reconstructed texture to establish a mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

[0012] The present invention also provides a fine three-dimensional surface model texture optimization device, comprising: a texture redundancy removal module, which includes a first clustering module, a second clustering module, and an effective texture extraction module, wherein,

[0013] The first clustering module is used to cluster the textures in the fine three-dimensional surface model based on the density clustering algorithm to obtain the first clustering result;

[0014] The second clustering module is used to judge each point in the texture. If the current point is an unclassified point, it assigns a classification number to the current point and judges whether the remaining points belong to the same triangular facet as the current point. If they belong to the same triangular facet, the classification number of the remaining points is changed to the classification number of the current point, regardless of which cluster the remaining points are in the first clustering result, to obtain the second clustering result.

[0015] The effective texture extraction module is used to calculate the minimum bounding rectangle of each cluster in the second clustering result to determine the effective texture.

[0016] Preferably, the fine three-dimensional surface model texture optimization device further includes: a standardization module, which is used to standardize the texture coordinates of the texture before clustering the texture, so that the texture coordinates are in the range of [0,1].

[0017] Preferably, the fine three-dimensional surface model texture optimization device further includes: a texture recombination module, which is used to reconstruct the effective texture using a two-dimensional rectangular binning method to obtain a reconstructed texture.

[0018] Preferably, the fine three-dimensional surface model texture optimization device further includes: a texture remapping module, which is used to remap the texture blocks in the recombined texture to establish a mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

[0019] The present invention also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the method of any one of claims 1-4.

[0020] The present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1-4.

[0021] The present invention has at least the following beneficial effects:

[0022] The texture optimization method of the present invention, after performing the first clustering of the texture based on the density algorithm, determines whether the unclassified points and the remaining points after clustering belong to the same triangular facet, and then reclassifies the unclassified points and the remaining points belonging to the same triangular facet. That is, it adds triangle constraints on the basis of the existing density-based clustering algorithm, and finally determines the effective texture region, which effectively improves the removal rate of redundant textures, and the optimized texture is suitable for efficient transmission and fast visualization in the Web environment.

[0023] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0024] Figure 1 This is a flowchart of a fine three-dimensional surface model texture optimization method according to an embodiment of the present invention;

[0025] Figure 2 This is the result of remapping according to another embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram of a repeating texture according to another embodiment of the present invention;

[0027] Figure 4 This is a schematic diagram of the texture coordinates translated along the positive UV axis according to another embodiment of the present invention;

[0028] Figure 5 This is a schematic diagram of the Delaunay triangulation of the constraint edge according to another embodiment of the present invention;

[0029] Figure 6 This is a schematic diagram illustrating the CP algorithm for determining the optimal position according to another embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of texture reconstruction according to another embodiment of the present invention;

[0031] Figure 8 This is a schematic diagram of texture remapping according to another embodiment of the present invention. Detailed Implementation

[0032] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings, so that those skilled in the art can implement it based on the description.

[0033] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0034] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0035] like Figure 1 As shown, the present invention provides a method for optimizing the texture of a fine three-dimensional surface model, comprising:

[0036] S400, texture redundancy removal.

[0037] S401. Cluster the textures in the fine three-dimensional surface model based on the density clustering algorithm to obtain the first clustering result;

[0038] S402. For each point in the texture, if the current point is an unclassified point, assign a classification number to the current point and determine whether the remaining points belong to the same triangular facet as the current point. If they belong to the same triangular facet, regardless of which cluster the remaining points are in the first clustering result, modify the classification number of the remaining points to the classification number of the current point to obtain the second clustering result.

[0039] S403. Calculate the minimum bounding rectangle of each cluster in the second clustering result to determine the effective texture.

[0040] When modeling detailed 3D surface models, an octree is typically used to construct the spatial index of the model. Each node of the octree corresponds to one or more textures. Only a portion of each texture is actually referenced by the model; this portion is called the effective texture, and its corresponding minimum bounding rectangle is called the effective texture region. The image portion outside the effective texture region is called redundant texture. Redundant textures not only consume network transmission resources and increase model loading time during model loading and visualization, but also increase the burden on the GPU for texture processing, thus reducing the system's rendering performance.

[0041] Existing texture redundancy removal methods often employ density-based clustering algorithms (DBSCAN clustering algorithm). This algorithm directly calculates the bounding box of each cluster based on the clustering results to determine the effective texture region. Then, effective texture is extracted based on these regions. After re-texturing using this effective texture, a comparison with the original model reveals that some areas of the re-textured model have missing textures, such as… Figure 2 (a) Analysis reveals that the reason is that for larger or narrower triangular faces (such as walls or pillars) in a 3D model, the texture coordinates of their three vertices usually also form a larger or narrower triangle in 2D space. During clustering, if one vertex of this triangle has already been assigned to a cluster, but the distance of the other two vertices to this cluster is greater than the given clustering threshold eps, then these two vertices cannot be assigned to this cluster, resulting in the loss of the effective texture region of the triangle.

[0042] To address this issue, this invention incorporates triangle constraints when using the DBSCAN clustering algorithm on texture coordinate points. The algorithm is described below:

[0043] The `clusterPoint` structure stores the x, y, and z coordinates of a texture point (if the original texture coordinates have no z value, z is set to 0), the cluster ID, the triangle ID to which the point belongs, and whether the point is being updated (`isUpdate`). The specific structure is as follows:

[0044]

[0045] The pseudocode for the texture coordinate clustering implementation process based on the DBSCAN clustering algorithm is shown below:

[0046] Based on the clustering results under unconstrained conditions, each point in each texture is evaluated. If the current point is identified as a noise point or an unclassified point, it is assigned a classification number. Then, it is determined whether the current point belongs to the same triangle as the remaining points in the texture. If they belong to the same triangle, regardless of the cluster of the remaining points, their classification numbers are changed to the current point's classification number. This ensures the integrity of each triangular facet texture in the original model while clustering textures according to spatial density. Using this method to extract valid textures and re-texturing, correct texture referencing results can be obtained, such as... Figure 2 (b)

[0047] Furthermore, the clustering threshold eps significantly impacts the clustering results. An excessively high value reduces the number of clusters and generates more redundant texture space, while a value that is too low increases computational complexity and reduces algorithm efficiency. Through multiple experiments, eps = 0.0001 demonstrates a good balance between redundant texture removal and algorithm efficiency. Using the texture redundancy removal method of this invention (the method of this invention) and existing density-based clustering algorithms (existing methods), redundancy removal was performed on textures referenced by the model within octree nodes. The removal status of 10 nodes was statistically analyzed, and the results are shown in Table 1.

[0048] Table 1 Comparison of Redundant Texture Removal Rate

[0049]

[0050] As shown in Table 1, the method of this invention achieves a redundant texture removal rate exceeding 98%, significantly higher than existing methods. When processing the data "101_m3", the two algorithms achieve similar redundant texture removal rates, both exceeding 97%. This is because existing methods heavily rely on the clustered distribution of texture coordinates within nodes in two-dimensional space. When texture coordinates are clustered, the effective texture region represented by its minimum bounding rectangle is very close to the actually referenced texture region, resulting in a high redundant texture removal rate. In contrast, the method of this invention determines the effective texture region based on texture coordinate clustering. Regardless of the texture coordinate distribution, it can identify the smallest possible effective texture region, thus eliminating most redundant textures. Therefore, the redundant texture removal method proposed in this paper exhibits better removal performance and higher robustness than existing methods.

[0051] In the above embodiments, after the first clustering, the unclassified points and the remaining points belonging to the same triangular facet are reclassified by determining whether the unclassified points and the remaining points belong to the same triangular facet. That is, triangle constraints are added to the existing density-based clustering algorithm, and then the effective texture is determined, which effectively improves the removal rate of redundant textures.

[0052] In another embodiment, before clustering the textures, the method further includes:

[0053] S200, Texture Coordinate Standardization Processing: Obtain the original texture in the fine 3D surface model, and standardize the texture coordinates located outside the range [0,1] so that all texture coordinates are within the range [0,1], so as to facilitate subsequent texture optimization and coordinate updates.

[0054] Normally, the standardized texture coordinates u and v are located in the range of [0,1]. However, since the modeling process often uses repeated texture mapping (such as in architectural models, doors, windows, walls, etc. often have a lot of repeated textures), there are also many texture coordinates outside the range of [0,1].

[0055] Figure 3 A set of repeating texture data is presented. To facilitate observation of the relationship between their coordinates, three vertices, Point10, Point11, and Point12, are added. Observation reveals that in the repeating texture model, the texture is repeated based on the origin Point10 (0,0) along the four directions (up, down, left, and right), with textures stitched together at integer junctions. Each texture within an integer range corresponds to a complete texture within the range [0,1], but its texture coordinates are accumulated based on the offset from the origin. Therefore, texture coordinates outside the [0,1] range appear. Figure 3 The texture coordinates of each point are shown in Table 2.

[0056] Table 2 Repetitive Texture Coordinates

[0057]

[0058] Since translating the texture coordinates by an integer number of units does not affect the texture mapping effect, for cases where the UV coordinates of a triangle vertex are less than 0, we can translate the three UV coordinates of the triangle by an integer number of units to ensure that the texture coordinates of each point are greater than 0. Here, the integer offset along the positive U-axis is denoted as offset_u, and the integer offset along the positive V-axis is denoted as offset_v. The translation amount in each direction is the smallest integer that makes u or v greater than or equal to 0, i.e.:

[0059] offset_u = Floor(|u|)

[0060] offset_v = Floor(|v|)

[0061] By translating, all cases are transformed into solutions based on the positive UV axis; that is, after the transformation, only cases where the UV coordinate is greater than 1 need to be processed. For example... Figure 4 As shown, if a triangle has multiple textures that are repeatedly mapped and the texture coordinates u and v are located on the negative half axis, by translating offset_u and offset_v units along the positive U and V axes respectively, and then dividing the triangle along the texture boundaries (a1a2, a3a4), calculating the geometric coordinates corresponding to the intersection points a1, a2, a3, and a4, and then reconstructing the mesh based on the Delaunay triangulation with edge constraints, all texture coordinates are processed into the range [0,1].

[0062] The main idea of ​​the edge-constrained Delaunay triangulation algorithm is to embed constrained edges into an unconstrained standard Delaunay triangulation network, locally adjusting the connection relationships of the original triangles to satisfy the given constraints. The algorithm steps are as follows:

[0063] ① Calculate the intersection point of each dividing face and the cross triangle, and store it in the container v_CrossCoordinate. Store the line segment of the intersection between the dividing face and the cross triangle in the container v_SegLine.

[0064] ② Construct an unconstrained Delaunay triangulation for the vertex set v_CrossCoordinate using the divide-and-conquer algorithm and store it in v_OriTriangle;

[0065] ③ Take out the first line segment in v_SegLine and remove it from the container. Traverse all unconstrained triangles and determine whether each triangle has an intersection point with the current line segment except for the vertex. If so, store the triangle in the container v_CDTTriangle and remove it from the container v_OriTriangle.

[0066] ④ Perform LOP optimization on the triangles in v_CDTTriangle. For example... Figure 5 As shown, for the quadrilateral formed by two triangles with a common side in v_CDTTriangle, swap their diagonals to generate two new triangles, and store them in v_OriTriangle.

[0067] ⑤ Repeat steps ③ and ④ until the container v_SegLine is empty, then end the algorithm.

[0068] In another embodiment, after obtaining the effective texture, the method further includes:

[0069] S600, Texture Reconstruction: The effective texture is reconstructed using a two-dimensional rectangular binning method to obtain a reconstructed texture.

[0070] The deredundant texture significantly reduces its physical disk and memory footprint and improves network transmission efficiency by eliminating most of the invalid texels. However, during model rendering, the texture's memory usage remains unchanged compared to before deredundancy. This is because, with the bit depth unchanged, the amount of memory a texture occupies is directly proportional to its width and height. The deredundant texture still uses the original texture distribution and does not reduce the overall width and height of the texture; therefore, the effective texture area should be reorganized.

[0071] Since each effective texture is a two-dimensional rectangle, the essence of the recombination is a two-dimensional rectangle binning problem. Therefore, this invention adopts a two-dimensional rectangle binning method to recombine the effective texture.

[0072] Two-dimensional reactive angular strip packing problem (2DR-SPP) typically refers to packing several rectangles of different sizes {R1, R2, R3, ... R...} into boxes. n The goal is to pack textures into a rectangular container C with a given initial width and height, achieving a specific optimization objective after packing all rectangles, such as minimizing the area occupied by container C. This invention employs a two-dimensional rectangular binning algorithm to reorganize textures, removing redundant texture space and improving the rendering efficiency of effective textures on the GPU.

[0073] In this invention, a rectangle R = {x, y, w, h, θ} is defined, where x and y are the coordinates of the lower left corner of the rectangle, w and h are the width and height of the rectangle, and θ is the rotation angle of the rectangle. The main process of packing ordinary two-dimensional rectangles is as follows: Given a rectangular container C with initial width and height, and a set of rectangles to be packed G = {R1, R2, R3, ... R...} n}, R i Select appropriate areas for packing according to certain rules, and after packing, place container R inside container C. i The rectangular region is denoted as o i The set of occupied rectangles is O = {o1, o2, o2, ..., o}. n The remaining area in container C will be redivided into multiple rectangular areas, R i Each time, find the best location from the remaining area to place it.

[0074] The optimal location was determined using the CP algorithm. The principle of the CP algorithm is as follows: R... i Let l be the sum of the contact side lengths with the set of rectangular boxes O that are already packed. Then the set of side lengths generated at different positions is L = {l1, l2, l3...l}. n}, let l max ∈L, then when l=l max The position at that time is the optimal position. For example... Figure 6 In the given information, l2 > l1 > l3, therefore the position shown in (b) is the optimal position.

[0075] The specific process of reorganizing the effective texture region using the two-dimensional rectangular binning algorithm is as follows:

[0076] (1) Adjust the longest side of each texture block in the texture queue to the width, and sort all the texture blocks to be reassembled in descending order of width;

[0077] (2) Access the first texture in the texture queue and remove it from the queue;

[0078] (3) Determine if there is space available to place the texture block;

[0079] (4) If it does not exist, create a new texture space with the same width and height as the texture block, and recalculate the width and height of the current reassembled texture set.

[0080] (5) If it exists, select the area with the largest contact length to place the texture block. If the contact lengths are the same, place them in the order from left to right and from bottom to top.

[0081] (6) If the texture cannot be placed directly in any existing free space, rotate the texture 90° clockwise and then execute (3).

[0082] Repeat steps (2) to (6) until the texture queue is empty, at which point the algorithm ends.

[0083] Figure 7 In the middle (a), there is a texture queue to be reassembled and the reassembled texture. Figure 7 Steps (b) to (e) represent the reassembly process, strictly following steps (2) to (6), resulting in the reassembled texture set as shown below. Figure 7 As shown in (f). During the texture reconstruction process, no scaling of the original texture is involved. Therefore, the reconstructed texture not only guarantees the resolution of the original texture, but also improves the space utilization of the texture set, so as to achieve the purpose of fast network transmission and efficient GPU rendering of the texture when rendering the model.

[0084] In another embodiment, after obtaining the reconstructed texture, the method further includes:

[0085] S800, Texture remapping: Remapping the texture blocks in the reconstructed texture to establish a mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

[0086] During the texture reconstruction process, the texture coordinates change due to the translation and rotation operations performed on the texture blocks. Therefore, it is necessary to remap the texture coordinates based on the optimized texture and the model geometry to quickly extract the optimized texture and render the model.

[0087] Let the width and height of the original texture be denoted as w and h, respectively, where the coordinates of a certain texel are (p... x ,p y When rotated 90° clockwise, the width of the texture becomes h and the height becomes w. The coordinates of this texel are denoted as (p). x ',p y '),like Figure 8 From (a), we can obtain the following relation:

[0088]

[0089] The width and height of the reconstructed texture are denoted as w' and h', respectively. The position of the original texture in the new texture is as follows: Figure 8 In the middle (c), record the coordinates of its lower left corner as (x, y), then after recombination:

[0090] If no rotation is performed, the recombined texel coordinates are:

[0091]

[0092] If rotation is performed, the recombined texel coordinates are:

[0093]

[0094] The UV coordinates of the texture are then calculated using the following formula:

[0095]

[0096] Furthermore, the present invention also provides the aforementioned fine three-dimensional surface model texture optimization device, including...

[0097] The normalization module is used to normalize the texture coordinates of textures in the model so that the texture coordinates are in the range of [0,1].

[0098] The texture redundancy removal module includes a first clustering module, a second clustering module, and an effective texture extraction module. The first clustering module clusters the standardized texture coordinates to obtain a first clustering result. The second clustering module judges each point in the texture; if the current point is unclassified, it assigns a classification number and determines whether the remaining points belong to the same triangular facet. If they belong to the same triangular facet, regardless of which cluster the remaining points are in the first clustering result, the classification number of the remaining points is changed to the classification number of the current point, resulting in a second clustering result. The effective texture extraction module calculates the minimum bounding rectangle of each cluster in the second clustering result to determine the effective texture.

[0099] The texture reconstruction module is used to reconstruct the effective texture using a two-dimensional rectangular binning method to obtain a reconstructed texture;

[0100] The texture remapping module is used to remap the texture coordinates of the texture blocks in the reconstructed texture and establish the mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

[0101] The above implementation scheme is based on the same inventive concept as the fine three-dimensional surface model texture optimization method. For an understanding of this scheme, please refer to the description in the optimization method section.

[0102] Furthermore, the optimization method of the present invention can also be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0103] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for optimizing the texture of a fine three-dimensional surface model, characterized in that, include: To obtain the original texture in the fine 3D surface model, for texture coordinates that exceed the range of [0,1] due to repeated mapping, the texture coordinates are made to be greater than or equal to 0 by an integer number of units of translation. Then, the triangle is divided along the texture boundary, the geometric coordinates of the intersection point are calculated, and the mesh is reconstructed based on the Delaunay triangulation with edge constraints so that all texture coordinates are within the range of [0,1]. The standardized texture coordinate points are clustered based on the density clustering algorithm to obtain the first clustering result, where each texture coordinate point is associated with the information of the triangular facet to which it belongs; For each point in the texture, if the current point is an unclassified point, a new classification number is assigned to the current point. Then, it is determined whether the remaining points belong to the same triangular facet as the current point. If they belong to the same triangular facet, regardless of which cluster the remaining points are in in the first clustering result, and regardless of whether the spatial distance between the remaining points and the current point meets the threshold of density clustering, the classification number of the remaining points is forcibly modified to the classification number of the current point, thereby obtaining the second clustering result. Calculate the minimum bounding rectangle of each cluster in the second clustering result, and extract the effective texture region from the original texture based on the minimum bounding rectangle to determine the effective texture.

2. The method for optimizing the texture of a fine three-dimensional surface model as described in claim 1, characterized in that, Also includes: The effective texture is recombined using a two-dimensional rectangular binning method to obtain a recombined texture.

3. The method for optimizing the texture of a fine three-dimensional surface model as described in claim 2, characterized in that, Also includes: The texture blocks in the reconstructed texture are remapped to remapping of texture coordinates to establish a mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

4. The fine three-dimensional surface model texture optimization device as described in claim 1, characterized in that, include: Standardized modules; The texture redundancy removal module includes a first clustering module, a second clustering module, and an effective texture extraction module, wherein... The standardization module is used to obtain the original texture in the fine 3D surface model. For texture coordinates that exceed the range of [0,1] due to repeated mapping, the texture coordinates are made to be greater than or equal to 0 by an integer number of translation units. Then, the triangle is divided along the texture boundary, the geometric coordinates of the intersection point are calculated, and the mesh is reconstructed based on the Delaunay triangulation with edge constraints so that all texture coordinates are within the range of [0,1]. The first clustering module is used to cluster the standardized texture coordinate points based on the density clustering algorithm to obtain the first clustering result, where each texture coordinate point is associated with the information of the triangular facet to which it belongs. The second clustering module is used to judge each point in the texture. If the current point is an unclassified point, it assigns a new classification number to the current point and judges whether the remaining points belong to the same triangular facet as the current point. If they belong to the same triangular facet, regardless of which cluster the remaining points are in the first clustering result, and regardless of whether the spatial distance between the remaining points and the current point meets the threshold of density clustering, the classification number of the remaining points is forcibly modified to the classification number of the current point, thereby obtaining the second clustering result. The effective texture extraction module is used to calculate the minimum bounding rectangle of each cluster in the second clustering result, extract the effective texture region from the original texture based on the minimum bounding rectangle, and determine the effective texture.

5. The fine three-dimensional surface model texture optimization device as described in claim 4, characterized in that, Also includes: The texture reconstruction module is used to reconstruct the effective texture using a two-dimensional rectangular binning method to obtain a reconstructed texture.

6. The fine three-dimensional surface model texture optimization device as described in claim 5, characterized in that, Also includes: The texture remapping module is used to remap the texture blocks in the reconstructed texture to establish a mapping relationship between the texture blocks and the geometric data in the fine three-dimensional surface model.

7. An electronic device, characterized in that, include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform the method of any one of claims 1-3.

8. A storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method of any one of claims 1-3.

Images