A Simplified Method for Preventing Texture Deformation in Oblique Photogrammetry Models
By classifying and simplifying the vertices of the oblique photogrammetry model, and combining the vertex undulation factor and texture deformation factor, the problems of texture deformation and edge damage in existing algorithms are solved, achieving efficient preservation of the geometric and texture features of the model and improving the simplified visual effect.
Patent Information
- Application Number
- CN202410779492.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-17
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2044-06-17
AI Technical Summary
Existing 3D mesh model simplification algorithms cannot effectively handle oblique photogrammetry models containing a large amount of discontinuous texture information, resulting in texture deformation and edge damage in the simplified model, and failing to retain the geometric and texture features of the model at a high simplification rate.
By classifying the vertices of the oblique photogrammetry model, simplification strategies are formulated for different types of vertices. Combining the vertex undulation factor, texture deformation factor, and quadratic error measure algorithm, the simplification cost is calculated, and the edge with the minimum cost is selected for simplification to prevent texture deformation and edge damage.
While reducing the number of triangular meshes, the simplified model avoids texture distortion and edge damage, effectively preserving the geometric and texture features of the model and improving the visual quality of the simplified model.
Smart Images

Figure CN118799625B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of oblique photogrammetry modeling technology, and specifically to a method for simplifying oblique photogrammetry models to prevent texture deformation. Background Technology
[0002] Currently, 3D mesh model simplification algorithms mainly fall into four categories: mesh redrawing, vertex clustering, edge folding, triangular face folding, and vertex deletion. While these algorithms can reduce the number of triangular meshes and preserve geometric features to some extent, they are ineffective at handling 3D mesh models containing texture information. This can easily lead to texture distortion in the simplified model, especially when processing oblique photogrammetry 3D models. Because oblique photogrammetry models contain a large amount of discrete texture information, ordinary 3D mesh model simplification algorithms cannot accurately assign reasonable texture coordinates to newly generated vertices, resulting in significant texture distortion in the simplified model. Furthermore, these simplification algorithms also fail to properly preserve the boundaries of the oblique photogrammetry model, leading to edge breakage. On the other hand, oblique photogrammetry models contain a large number of geometric and texture features, and ordinary 3D mesh model simplification algorithms cannot effectively preserve these features at high simplification rates.
[0003] With the rapid development of 3D reconstruction technology and computer graphics, oblique photogrammetry modeling technology has been widely used in many fields such as topographic mapping, engineering surveying, safety monitoring, power line inspection, digital cities, transportation planning, and BIM design due to its advantages of low cost, high efficiency, and realistic model appearance. However, oblique photogrammetry models generate a massive amount of data. This large data volume not only prolongs the computer rendering time but also poses challenges to the subsequent storage and management of the model. Therefore, simplifying the processing of oblique photogrammetry models and reducing the storage space required and rendering time has become an urgent problem to be solved.
[0004] In response, scholars at home and abroad have proposed many simplification methods for 3D mesh models. However, most of the simplification algorithms are for ordinary mesh models without texture information or mesh models with continuous texture information, and cannot effectively handle oblique photogrammetry models with a large amount of discontinuous texture information.
[0005] Chinese patent “A method and system for simplifying a three-dimensional model” (CN108460836A) discloses a method and system for simplifying a three-dimensional model. The method uses the distinctiveness of the features of the vertices and the area of the triangular mesh corresponding to the vertices as weight values, and selects the vertices to be simplified according to the magnitude of the weight values.
[0006] Chinese patent "A 3D Model Simplification Method Based on Skeletonization" (CN109934928B) discloses a 3D model simplification method based on skeletonization. This method uses one-dimensional curves to replace the original mesh, generating multiple models with different levels of detail, greatly reducing the data volume of the 3D model while maintaining visual effects. Both of these patents do not consider the model's texture information and can only simplify ordinary mesh models that lack texture information.
[0007] Chinese patent "A Method for Simplifying 3D Models Applicable to Textured Models" (CN103714577A) discloses a method for simplifying 3D models applicable to textured models. This method introduces a texture deformation coefficient to optimize the operation cost and can handle textured models with continuous texture information. However, it cannot effectively handle oblique photogrammetry models with discontinuous texture information.
[0008] Reference [1]: Li Peng, Yan Qingsong, Qu Yingjie, Deng Fei. Algorithm for simplifying real-scene 3D models by fusing texture information [J]. Science of Surveying and Mapping, 2021, 46(10): 151-158+166. DOI: 10.16251 / j.cnki.1009-2307.2021.10.020. This paper proposes a method for simplifying real-scene 3D models by fusing texture information, which can retain certain geometric details of the model. However, this method does not consider how to adjust the texture coordinates of the simplified model, so that the geometric coordinates and texture coordinates of the newly generated vertices after simplification cannot correspond correctly, resulting in obvious texture deformation in the final output model.
[0009] Reference [2]: Li Shijun, Jiang Xiaotong, Tang Hui. High-quality simplification algorithm for textured models that preserve details [J]. Computer Applications Research, 2020, 37(01):300-303+312. DOI:10.19734 / j.issn.1001-3695.2018.04.0481. The high-quality simplification method for textured models that preserve details can retain the texture features of the model to a certain extent. However, this algorithm can only process models with continuous texture information. It cannot effectively process oblique photogrammetry models with discontinuous texture information. After simplification, the model will still show obvious texture deformation and edge damage.
[0010] Oblique photogrammetry models contain a large amount of discontinuous texture information. Using existing 3D mesh model simplification techniques often results in obvious edge damage and texture deformation in the simplified model. Furthermore, the geometric and texture features of the oblique photogrammetry model cannot be effectively preserved after simplification. Summary of the Invention
[0011] To address the aforementioned technical problems, this invention proposes a method for simplifying oblique photogrammetry models to prevent texture deformation. This method can reduce the number of triangular meshes in the oblique photogrammetry model while avoiding edge damage and texture deformation in the simplified model, and effectively preserves the geometric and texture features of the oblique photogrammetry model.
[0012] The technical solution adopted in this invention is as follows:
[0013] A method for simplifying oblique photogrammetry models to prevent texture deformation includes the following steps:
[0014] Step 1: Preprocess the oblique photogrammetry model and read the topological connection relationships between the graphic elements in the oblique photogrammetry model;
[0015] Step 2: Based on the number of texture coordinates associated with each vertex in the oblique photogrammetry model, vertices are divided into three categories: smooth vertices, seam vertices, and corner vertices;
[0016] Step 3: Based on the number of triangular faces associated with all edges in the first-order neighborhood of a vertex, perform a secondary classification of the vertex into non-boundary vertices and boundary vertices;
[0017] Step 4: For edges composed of smooth vertices, seam vertices, and corner vertices, formulate simplified strategies to prevent texture deformation; at the same time, for edges composed of boundary vertices and non-boundary vertices, formulate simplified strategies to prevent edge breakage.
[0018] Step 5: Calculate the vertex undulation factor, the area factor of the first-order neighborhood triangle of the vertex, and the texture deformation factor in the oblique photogrammetry model;
[0019] Step 6: Introduce the vertex undulation factor, the vertex first-order neighborhood triangle area factor, and the texture deformation factor into the quadratic error measurement algorithm to calculate the quadratic error matrix of the edge to be simplified and the optimized simplification cost;
[0020] Step 7: Sort the simplification costs of each edge in the oblique photogrammetry model from low to high, and select the edge with the lowest simplification cost based on the simplification strategy established in Step 4 for simplification.
[0021] Step 8: Update the quadratic error of the affected edges, the coordinates of the newly generated vertices, and the simplified cost sorting. Then, based on the new simplified cost sorting, select the edge with the smallest simplified cost for the folding operation until the number of triangles in the oblique photogrammetry model is reduced to the target value.
[0022] In step two, when a vertex corresponds to only one texture coordinate t, it is defined as a smooth vertex; when a vertex has two corresponding texture coordinates t, it is defined as a seam vertex; when a vertex has three or more corresponding texture coordinates t, it is defined as a corner vertex.
[0023] In step three, edges with only one associated triangle in the oblique photogrammetry model are defined as boundary edges, and edges with two or more associated triangles are defined as non-boundary edges. When a boundary edge exists in the first-order neighborhood of a vertex, the vertex is identified as a boundary vertex; when no boundary edge exists in the first-order neighborhood of a vertex, the vertex is identified as a non-boundary vertex.
[0024] The simplification strategy for preventing texture deformation in step four specifically includes:
[0025] (a): When both vertices v1 and v2 are smooth vertices, the newly generated vertex v after simplification is also a smooth vertex. The spatial coordinates and texture coordinates of vertex v are calculated using a quadratic error measure method that considers texture coordinate information. The specific calculation process is as follows:
[0026] Place all vertices into a five-dimensional vector space R 5 That is, the vertex representation vector is [xyzuw] T Where: x, y, z represent spatial coordinate values; u, w represent normalized texture coordinate values; T represents matrix transpose.
[0027] Let R be a five-dimensional vector space. 5 The memory contains any triangle t = (v1, v2, v3), and the three vertices v1, v2, and v3 lie in space R. 5 A two-dimensional plane is constructed, and e1 and e2 are two orthogonalized unit vectors on this plane. Their calculation formulas are as follows:
[0028]
[0029] Suppose there exists any vertex v i ∈R 5 Then vertex v i The square of the distance to the triangle t is:
[0030]
[0031] In the formula, D represents vertex v i The square of the distance to the triangular face t; v i T Indicates v i The transpose of ; A represents a 5×5 matrix; B represents a 5-dimensional vector; B T Let B be the transpose of B; C represents a 5-dimensional scalar.
[0032] The formulas for calculating A, B, and C are as follows:
[0033]
[0034] In the formula, I represents a 5×5 identity matrix; e1 T e2 represents the transpose of e1; T Let e2 be the transpose vector.
[0035] Define vertex v i The quadratic error matrix is:
[0036]
[0037] In the formula: planes(v i ) represents vertex v i The set of all triangular faces within a first-order neighborhood; t denotes belonging to planes(v i The triangular facets of the set; A t Indicates the calculation of vertex v i A 5×5 matrix representing the process of moving the square of the distance t to the triangular face; B t Indicates the calculation of vertex v i A 5-dimensional vector C that travels from the squared distance of the triangle t to the triangular facet t; t Indicates the calculation of vertex v i A 5-dimensional scalar B in the process of moving from the square of the distance t to the triangular face; t T B t The transpose of .
[0038] When edge (v1, v2) simplifies to a new vertex For ease of calculation, Q(v1) + Q(v2) is taken as... The quadratic error matrix, the simplified error of edge (v1,v2) and newly generated vertices The calculation formula is as follows:
[0039]
[0040] In the formula: planes(v1,v2) represents the set of triangular faces in the first-order neighborhood of vertices v1 and v2; Q(v1) represents the quadratic error matrix of vertex v1; Q(v2) represents the quadratic error matrix of vertex v2; Indicates will The coordinate vector is increased from 5 dimensions to 6 dimensions, and all the added elements are 1.
[0041] (b): If one of vertices v1 and v2 is a smooth vertex and the other is a corner vertex, and vertex v2 is a corner vertex, then after simplification, edge (v1, v2) continues to use v2 as the position of the newly generated vertex.
[0042] (c): If one of vertices v1 and v2 is a smooth vertex and the other is a seam vertex, as shown in Figure 4(c), then after the edge (v1,v2) is simplified, v2 will continue to be the newly generated vertex.
[0043] (d): If vertices v1 and v2 are both seam vertices, and edge (v1, v2) is located at the seam between two different texture blocks, then Also, for the seam vertex, take the midpoint of edge (v1, v2) as... spatial coordinates, The texture coordinates are taken as the average of the v1 and v2 texture coordinates:
[0044]
[0045] Step four includes a simplified strategy to prevent edge breakage, specifically:
[0046] 1) If both vertex v1 and vertex v2 are non-boundary vertices or both are boundary vertices, the corresponding simplifications can be performed directly according to cases (a)-(d) above;
[0047] 2) There is one boundary vertex and one non-boundary vertex in v1 and v2. If the edge is simplified, the edge of the model will be damaged. Therefore, the simplification of the edge is prohibited.
[0048] Step five involves calculating the vertex undulation factor, the area factor of the first-order neighborhood triangle of the vertex, and the texture deformation factor in the oblique photogrammetry model. Specifically, this includes:
[0049] The peak undulation factor Flu is introduced, and its calculation formula is as follows:
[0050]
[0051] In the formula: Represents vertex v i Vertex variability factor; vert(v i ) is vertex v i The set of adjacent vertices within a first-order neighborhood; Represents vertex v i normal vector and vertex v j The angle between the normal vectors, Indicates v i The vertex normal vector; Indicates v j Vertex normal vector; ||v i -v j || is vertex v i and vertex v j The Euclidean distance; k is the vertex vi The number of adjacent vertices in a first-order neighborhood; v j This indicates that it belongs to vertex v i The vertices of the set of adjacent vertices within a first-order domain.
[0052] As shown in the above formula, the accuracy of the vertex normal vector directly affects the effectiveness of introducing the vertex undulation factor, and thus affects the preservation of the geometric features of the simplified model. Therefore, in order to calculate the vertex normal vector in the model more accurately, the area S of the triangular mesh in the first-order neighborhood of the vertex and the interior angle θ associated with the vertex are introduced into the vertex normal vector calculation formula to obtain the optimized vertex normal vector calculation formula:
[0053]
[0054] In the formula: Indicates v i Vertex normal vectors; planes(v i ) represents vertex v i The set of all triangular faces within a first-order neighborhood; θ p For the p-th triangle and vertex v i Associated interior angle; S p n represents the geometric area of the p-th triangle; p Let represent the normal vector of the p-th triangle face.
[0055] To measure vertex v i The density of the surrounding triangular mesh better preserves the geometric features of the model, and an area factor for the first-order neighborhood triangles at the vertices is introduced. The calculation formula is as follows:
[0056]
[0057] In the formula: S p Let k represent the geometric area of the p-th triangle face; k is the vertex v. i The number of adjacent vertices in a first-order neighborhood; planes(v i ) represents vertex v i The set of all triangular faces within a first-order domain.
[0058] To preserve the texture features of the oblique photogrammetry model and reduce the risk of texture distortion after model simplification, a texture distortion factor Tex is introduced, which is calculated as follows:
[0059]
[0060] In the formula: Tex(v i ,v j ) represents the edge (v i ,v jTexture deformation factor; S tex,p α represents the texture area of the p-th triangle face; den(v) represents the average texture density of the triangular mesh within the first-order neighborhood of vertex v. Its function is to measure the proportion of the texture area within the first-order neighborhood of a vertex to the geometric area. The larger den(v) is, the denser the texture is in the region surrounding vertex v; α i and α j They represent v respectively i and v j The vertex type adjustment factor, which takes a real number greater than zero, is used to distinguish the importance of vertices of different categories.
[0061] den(v i ) represents vertex v i Average texture density of triangular mesh within a first-order neighborhood; den(v) j ) represents vertex v j Average texture density of a triangular mesh within a first-order neighborhood; n represents the number of triangular faces within the first-order neighborhood of vertex v.
[0062] In step six, the specific expression of the quadratic error matrix is shown in formula (4);
[0063] Edge simplification cost C(v) i ,v j The calculation formula is as follows:
[0064]
[0065] In the formula: Represents vertex v i First-order neighborhood triangle area factor; Represents vertex v j First-order neighborhood triangle area factor; Q(v) i ) represents vertex v i The quadratic error matrix; Q(v j ) represents vertex v j The quadratic error matrix; Represents vertex v i Vertex variability factor; Represents vertex v j The vertex undulation factor.
[0066] In step seven, let (v1, v2) be the edge in the model that currently conforms to the simplification strategy and has the minimum simplification cost. Let v1, v2 be the newly generated vertex position after simplification, and v3 to v8 be the adjacent vertices of edge (v1, v2); delete the two original vertices v1 and v2 of edge (v1, v2), and also delete the edges connected to v1 and v2, then use... Replace edge (v1, v2) and rebuild. The topological connections with adjacent vertices are used to complete a simplified operation.
[0067] In step eight, the quadratic error of the affected edges, the new vertices, and the simplification cost ranking are updated. Then, based on the new simplification cost ranking, the edge with the smallest simplification cost is selected for simplification, until the number of triangles in the oblique photogrammetry model reaches N. F The simplification ends after satisfying equation (12).
[0068] N F ≤V F (12);
[0069] In the formula: V F This represents the expected value of the triangular facet of the oblique photography model.
[0070] This invention provides a simplified method for oblique photography models to prevent texture deformation, with the following technical effects:
[0071] 1) This invention classifies vertices by the topological connection relationship of each graphic element in the model. It can formulate corresponding simplification strategies for different types of vertices, and more accurately calculate the spatial coordinates and corresponding texture coordinates of the newly generated vertices after simplification. This avoids texture deformation after model simplification. In addition, the simplification strategy also constrains the simplification of vertices at the model boundary, which can effectively avoid edge damage after model simplification.
[0072] 2) When calculating the texture deformation factor, this invention considers the influence of different types of vertices on the texture deformation factor, which can more accurately represent the importance of texture information in the first-order neighborhood of a vertex.
[0073] 3) This invention introduces a vertex undulation factor, a vertex first-order neighborhood triangle area factor, and a texture deformation factor to improve the simplification cost of edges, which can better preserve the geometric and texture details of the model.
[0074] 4) This invention defines the texture deformation factor as the product of three factors: vertex type adjustment factor, average texture density of the triangular mesh within the first-order neighborhood of a vertex, and the length of the edge to be simplified. The vertex type adjustment factor distinguishes the importance of different types of vertices, thus retaining more vertices that significantly influence the geometric and texture features of the model, thereby improving the simplification quality. The average texture density of the triangular mesh within the first-order neighborhood of a vertex measures the proportion of texture area to geometric area within the first-order neighborhood of the vertex. A higher average texture density indicates denser texture around the vertex, increasing the risk of texture deformation during simplification. The length of the edge to be simplified measures the impact of its length on texture deformation. A longer edge indicates a larger texture area spanned by the edge, encompassing more texture features, and thus a greater risk of texture deformation during simplification.
[0075] In summary, the texture deformation factor defined in this invention can effectively measure the risk of texture deformation after simplification. Introducing the texture deformation factor can better preserve the texture features of the model, avoid the risk of texture deformation, and thus better improve the visual quality of the simplified model. Attached Figure Description
[0076] The present invention will be further described below with reference to the accompanying drawings and examples;
[0077] Figure 1 A simplified flowchart of the method for creating a 3D model of a power transmission tower using oblique photography.
[0078] Figure 2 This is a diagram illustrating vertex classification.
[0079] Figure 3(a) is a schematic diagram of non-boundary vertex classification;
[0080] Figure 3(b) is a schematic diagram of boundary vertex classification.
[0081] Figure 4(a) illustrates the simplification strategy. Figure 1 ;
[0082] Figure 4(b) illustrates the simplified strategy. Figure 2 ;
[0083] Figure 4(c) is a simplified strategy diagram (3);
[0084] Figure 4(d) is a schematic diagram of the simplified strategy.
[0085] Figure 5 The diagram is for simplification.
[0086] Figure 6(a) shows the model when the QEM algorithm is simplified by 60%;
[0087] Figure 6(b) shows the model diagram when the algorithm of the present invention is simplified by 60%;
[0088] Figure 6(c) shows the model when the QEM algorithm is simplified by 80%;
[0089] Figure 6(d) shows the model diagram when the algorithm of the present invention is simplified by 80%.
[0090] Figure 7 This is a schematic diagram for reading the topological connection relationships of the model. Detailed Implementation
[0091] A method for simplifying oblique photogrammetry models to prevent texture deformation involves several steps. First, the oblique photogrammetry model is preprocessed by reading the topological connections between graphic elements. Based on the number of texture coordinates associated with each vertex, vertices are categorized into smooth vertices, seam vertices, and corner vertices. Next, based on the number of triangles associated with all edges within a vertex's first-order neighborhood, vertices are further classified into non-boundary vertices and boundary vertices. After classifying the vertices, different simplification strategies are developed for edges composed of different vertex types. Then, the vertex undulation factor, vertex first-order neighborhood triangle area factor, and texture deformation factor are calculated. These three improvement factors are combined with a quadratic error measure and used as the simplification cost for each edge. Finally, the simplification costs are sorted from smallest to largest, and simplification operations are performed on each edge to be simplified sequentially, thereby achieving the goal of simplifying the oblique photogrammetry model.
[0092] like Figure 1 As shown, a method for simplifying an oblique photogrammetry model to prevent texture deformation includes the following steps:
[0093] Step 1: Preprocess the oblique photogrammetry model and read the topological connection relationships between the graphic elements in the oblique photogrammetry model.
[0094] like Figure 7 As shown, in the preprocessing stage, the algorithm first reads the coordinate information of the vertices in the model, where x, y, and z represent the spatial coordinate values of the vertices; u and w represent the normalized texture coordinate values of the vertices. This vertex coordinate information is stored in a vertex list. Next, it reads the edge information of the model, storing the vertices that make up the edges in an edge list. Finally, it reads the triangular face information of the model, storing the numbers of the three edges that make up the triangular face in a triangular face list. By constructing vertex, edge, and triangular face lists, the algorithm can effectively read the graphic element information and topological connections in the oblique photogrammetry model, laying the foundation for subsequent model simplification.
[0095] Step 2: Based on the number of texture coordinates associated with each vertex in the oblique photogrammetry model, vertices are divided into three categories: smooth vertices, seam vertices, and corner vertices.
[0096] like Figure 2 As shown, when a vertex corresponds to only one texture coordinate t, it is defined as a smooth vertex; when a vertex has two corresponding texture coordinates t, it is defined as a seam vertex; when a vertex has three or more corresponding texture coordinates t, it is defined as a corner vertex.
[0097] Step 3: Classify vertices a second time based on the number of triangles associated with all edges in the first-order neighborhood of a vertex, dividing them into non-boundary vertices and boundary vertices.
[0098] In the oblique photogrammetry model, an edge with only one associated triangle is defined as a boundary edge, and an edge with two or more associated triangles is defined as a non-boundary edge. A vertex is considered a boundary vertex if a boundary edge exists within its first-order neighborhood, and a non-boundary vertex if no boundary edge exists within its first-order neighborhood. As shown in Figures 3(a) and 3(b), all edges within the first-order neighborhood of vertex v1 have two associated triangles, therefore all edges within this first-order neighborhood are non-boundary edges, and vertex v1 is a non-boundary vertex. Vertex v3 has an edge (v3, v4) within its first-order neighborhood with only one associated triangle, meaning (v3, v4) is a boundary edge, and therefore vertex v3 is a boundary vertex.
[0099] Step 4: Based on the vertex classification method in Step 2 and Step 3, formulate corresponding simplification strategies for edges composed of different types of vertices.
[0100] ①: When both vertices of the edge to be simplified are smooth vertices, a quadratic error measure method that considers texture coordinate information is used for simplification.
[0101] ②: When one of the two vertices of the edge to be simplified is a smooth vertex and the other is a corner vertex, the newly generated vertex after simplification inherits the spatial geometric coordinates and texture coordinates of the corner vertex;
[0102] ③: When one of the two vertices of the edge to be simplified is a smooth vertex and the other is a seam vertex, the newly generated vertex after simplification inherits the spatial geometric coordinates and texture coordinates of the seam vertex;
[0103] ④: When both vertices of the edge to be simplified are seam vertices, and the edge to be simplified is located at the seam of two different texture blocks, the midpoint of the edge to be simplified is used as the geometric coordinates of the newly generated vertex, and the average of the texture coordinates of the two vertices of the edge to be simplified is used as the texture coordinates of the newly generated vertex.
[0104] ⑤: In other cases, simplification of these edges is prohibited;
[0105] ⑥: When both vertices of the edge to be simplified are non-boundary vertices or both are boundary vertices, then simplify according to the five simplification strategies ①-⑤ above;
[0106] ⑥: If an edge to be simplified contains one boundary vertex and one non-boundary vertex, simplification of such an edge is prohibited.
[0107] The specific analysis is as follows:
[0108] Assume that edge (v1, v2) is the edge to be simplified, and vertex... The newly generated vertices are defined by the simplified edge (v1, v2). To effectively preserve the texture features of the model, this invention proposes a simplification strategy to prevent texture deformation for edges composed of smooth vertices, seam vertices, and corner vertices. This simplification strategy restricts certain types of edges from participating in the simplification process and can more accurately calculate the spatial coordinates and corresponding texture coordinates of the newly generated vertices during the simplification process, thereby avoiding texture deformation after model simplification, as shown in (a) to (e) below.
[0109] (a): As shown in Figure 4(a), when both vertex v1 and vertex v2 are smooth vertices, the newly generated vertex v after simplification is also a smooth vertex. The spatial coordinates and texture coordinates are calculated using a quadratic error measure method that considers texture coordinate information. The specific calculation process is as follows:
[0110] Place all vertices into a five-dimensional vector space R 5 That is, the vertex representation vector is [xyzuv]. T x, y, and z represent spatial coordinate values; u and v represent normalized texture coordinate values. Assume space R... 5 The memory contains any triangle t = (v1, v2, v3), and the three vertices v1, v2, and v3 lie in space R. 5 A two-dimensional plane is constructed, and e1 and e2 are assumed to be two orthogonalized unit vectors on this plane. Their calculation formula is as follows:
[0111]
[0112] Suppose there exists any vertex v i ∈R 5 Then vertex v i The square of the distance to the triangle t is:
[0113] D 2 =v i T Av i +2B T v i +C (2);
[0114] The formulas for calculating A, B, and C are as follows:
[0115]
[0116] Define vertex v i The quadratic error matrix is:
[0117]
[0118] In the formula: planes(v i ) represents vertex v i The set of all triangular faces within a first-order domain.
[0119] When edge (v1, v2) simplifies to a new vertex For ease of calculation, Q(v1) + Q(v2) is taken as... The quadratic error matrix, the simplified error of edge (v1,v2) and newly generated vertices The calculation formula is as follows:
[0120]
[0121] In the formula: planes(v1,v2) represents the set of triangular faces within the first-order neighborhood of vertices v1 and v2.
[0122] (b): If one of vertices v1 and v2 is a smooth vertex and the other is a corner vertex, as shown in Figure 4(b), and vertex v2 is a corner vertex, then after simplification, edge (v1, v2) continues to use v2 as the newly generated vertex position.
[0123] (c): If one of vertices v1 and v2 is a smooth vertex and the other is a seam vertex, as shown in Figure 4(c), and vertex v2 is a seam vertex, then after simplification, edge (v1, v2) continues to use v2 as the newly generated vertex.
[0124] (d): If vertices v1 and v2 are both seam vertices, and edge (v1, v2) is located at the seam between two different texture blocks, then Also a joint vertex, as shown in Figure 4(d), the midpoint of edge (v1, v2) is taken as... spatial coordinates, The texture coordinates are taken as the average of the v1 and v2 texture coordinates:
[0125]
[0126] (e): For other cases, simplifying edges (v1,v2) will significantly increase the risk of texture distortion in the simplified model. Therefore, the algorithm of this invention prohibits the simplification of these edges.
[0127] To better preserve the model's boundaries and prevent edge breakage in the simplified model, this invention proposes a simplification strategy to prevent edge breakage for edges composed of both boundary vertices and non-boundary vertices. This strategy prohibits edges composed of both boundary vertices and non-boundary vertices from participating in the simplification process, thereby effectively preserving the model's boundaries and avoiding edge breakage. Specifically, see below 1)-2):
[0128] 1) If both vertices v1 and v2 are non-boundary vertices or both are boundary vertices, they can be directly simplified according to the five cases (a)-(e) above.
[0129] 2) There is a boundary vertex and a non-boundary vertex in v1 and v2. If the edge is simplified, the edge of the model will be damaged. Therefore, the present invention prohibits the simplification of the edge.
[0130] Step 5: Calculate the vertex undulation factor, vertex first-order neighborhood triangle area factor, and texture deformation factor in the oblique photogrammetry model.
[0131] Since oblique photogrammetry models use a large number of triangular meshes to draw in uneven areas, in order to better measure the undulation of the area near the vertex, preserve the geometric features of the model as much as possible, and improve the visual quality of the simplified model, this invention introduces a vertex undulation factor Flu, the calculation formula of which is as follows:
[0132]
[0133] In the formula: Represents vertex v i vertex variability factor; ver(v i ) is vertex v i The set of adjacent vertices within a first-order neighborhood; Represents vertex v i normal vector and vertex v j The angle between the normal vectors, Indicates v i The vertex normal vector; Indicates v j Vertex normal vector; ||v i -v j || is vertex v i and vertex v j The Euclidean distance; k is the vertex v i The number of adjacent vertices within a first-order neighborhood. j This indicates that it belongs to vertex v i The vertices of the set of adjacent vertices within a first-order domain.
[0134] As can be seen from the above formula, the accuracy of the vertex normal vector directly affects the effect of introducing the vertex undulation factor, and thus affects the preservation of the geometric features of the simplified model. Therefore, in order to calculate the vertex normal vector in the model more accurately, this invention introduces the area S of the triangular mesh in the first-order neighborhood of the vertex and the interior angle θ associated with the vertex into the vertex normal vector calculation formula, resulting in the optimized vertex normal vector calculation formula:
[0135]
[0136] In the formula: planes(v i ) represents vertex v i The set of all triangular faces within a first-order neighborhood; θ p For the p-th triangle and vertex v i Associated interior angle; S p n represents the geometric area of the p-th triangle; p Let represent the normal vector of the p-th triangle face.
[0137] In oblique photogrammetry models, the triangular mesh density is generally higher in geometrically characteristic regions to measure the vertex v. i The density of the surrounding triangular mesh better preserves the geometric features of the model, and an area factor for the first-order neighborhood triangles at the vertices is introduced. The calculation formula is as follows:
[0138]
[0139] In the formula: S p Let k represent the geometric area of the p-th triangle face; k is the vertex v. i The number of adjacent vertices in a first-order neighborhood; planes(v i ) represents vertex v i The set of all triangular faces within a first-order domain.
[0140] To preserve the texture features of the oblique photogrammetry model and reduce the risk of texture distortion after model simplification, a texture distortion factor Tex is introduced, which is calculated as follows:
[0141]
[0142] In the formula: Tex(v i ,v j ) represents the edge (v i ,v j Texture deformation factor; S tex,pα represents the texture area of the p-th triangle face; den(v) represents the average texture density of the triangular mesh within the first-order neighborhood of vertex v. Its function is to measure the proportion of the texture area within the first-order neighborhood of a vertex to the geometric area. The larger den(v) is, the denser the texture is in the region surrounding vertex v; α i and α j Indicates v i and v j The vertex type adjustment factor, a real number greater than zero, is used to distinguish the importance of different types of vertices. For smooth vertices, seam vertices, and corner vertices, this invention suggests using α = 1.0, α = 1.5, and α = 1.2, respectively.
[0143] Step 6: Introduce the vertex undulation factor, the vertex first-order neighborhood triangle area factor, and the texture deformation factor into the quadratic error measure algorithm (QEM) to calculate the quadratic error matrix and simplification cost of the optimized edge to be simplified.
[0144] Combining the characteristics of the oblique photogrammetry model and comprehensively considering the above-mentioned optimization factors, the improved edge simplification cost C(v) of this invention... i ,v j The calculation formula is as follows:
[0145]
[0146] Step 7: Sort the simplification costs of each edge in the oblique photogrammetry model from low to high, and select the edge that meets the simplification strategy and has the lowest simplification cost according to the simplification strategy in Step 4 for simplification.
[0147] like Figure 5 As shown, assume that (v1, v2) is the edge that currently meets the simplification strategy and minimizes the simplification cost in the model. Let v1, v2 be the newly generated vertex position after simplification, and v3 to v8 be the adjacent vertices of edge (v1, v2). Delete the two original vertices v1 and v2 of edge (v1, v2), and simultaneously delete the edges connecting v1 and v2. Then use... Replace edge (v1, v2) and rebuild. The topological connections with adjacent vertices are used to complete a simplified operation.
[0148] Step 8: Update the quadratic error of the affected edges, the new vertices, and the simplification cost ranking. Then, based on the new simplification cost ranking, select the edge with the smallest simplification cost for simplification, and continue this process until the number of triangles in the oblique photogrammetry model reaches N. F The simplification ends after formula (12) is satisfied.
[0149] N F ≤V F (12);
[0150] In the formula: V F This represents the expected value of the triangular facet of the oblique photography model.
[0151] Example:
[0152] like Figures 6(a) to 6(d) As shown, when the model simplification rate reaches 60%, the QEM algorithm simplifies the model and exhibits texture distortion. The algorithm of this invention, however, can still effectively preserve the model's detailed features and avoid texture distortion. When the model simplification rate reaches 80%, the QEM algorithm exhibits even more severe texture distortion. The algorithm of this invention, by introducing a texture distortion coefficient to constrain the edge folding cost, better preserves the model's texture details, thus avoiding texture distortion even at high simplification rates.
[0153] As shown in Table 1, the algorithm of this invention outperforms the QEM algorithm in all aspects, including average error, standard deviation, and mesh quality. Specifically, at a 60% simplification rate, the algorithm of this invention reduces the average error by 84.4%, the standard deviation by 63.1%, and improves the mesh quality by 6.5% compared to the QEM algorithm; at an 80% simplification rate, the algorithm of this invention reduces the average error by 61.3%, the standard deviation by 33.4%, and improves the mesh quality by 5.2% compared to the QEM algorithm.
[0154] Table 1 Algorithm Simplification Error
[0155]
Claims
1. A method for simplifying an oblique photogrammetry model to prevent texture deformation, characterized in that... Includes the following steps: Step 1: Preprocess the oblique photogrammetry model and read the topological connection relationships between the graphic elements in the oblique photogrammetry model; Step 2: Based on the number of texture coordinates associated with each vertex in the oblique photogrammetry model, vertices are divided into three categories: smooth vertices, seam vertices, and corner vertices; Step 3: Based on the number of triangular faces associated with all edges in the first-order neighborhood of a vertex, perform a secondary classification of the vertex into non-boundary vertices and boundary vertices; Step 4: For edges composed of smooth vertices, seam vertices, and corner vertices, formulate simplified strategies to prevent texture deformation; at the same time, for edges composed of boundary vertices and non-boundary vertices, formulate simplified strategies to prevent edge breakage. Step 5: Calculate the vertex undulation factor, the area factor of the first-order neighborhood triangle of the vertex, and the texture deformation factor in the oblique photogrammetry model; Step 6: Introduce the vertex undulation factor, the vertex first-order neighborhood triangle area factor, and the texture deformation factor into the quadratic error measurement algorithm to calculate the quadratic error matrix of the edge to be simplified and the optimized simplification cost; Step 7: Sort the simplification costs of each edge in the oblique photogrammetry model from low to high, and select the edge with the lowest simplification cost based on the simplification strategy established in Step 4 for simplification. Step 8: Update the quadratic error of the affected edges, the coordinates of the newly generated vertices, and the simplified cost sorting. Then, based on the new simplified cost sorting, select the edge with the smallest simplified cost and perform the folding operation until the number of triangles in the oblique photogrammetry model is reduced to the target value. The simplification strategy for preventing texture deformation in step four specifically includes: (a): When the vertex and vertex When all vertices are smooth vertices, the newly generated vertices after simplification... Also a smooth vertex; vertex The spatial coordinates and texture coordinates are calculated using a quadratic error measure that takes into account texture coordinate information. The specific calculation process is as follows: Place all vertices into a five-dimensional vector space That is, the representation vector of the vertex is ,in: , , Represents spatial coordinate values; , Represents normalized texture coordinate values; Indicates matrix transpose; Let there be a five-dimensional vector space. Memory is in any triangle , , , Three vertices in space A two-dimensional plane is constructed in the middle, let and Let be two orthogonal unit vectors on this plane, and their calculation formulas are as follows: (1); Suppose there exists any vertex Then the vertex To the triangle The square of the distance is: (2); In the formula, Represents vertices To the triangle The square of the distance; express The transpose of the matrix; Represent a Matrix; Represents a 5-dimensional vector; express The transpose of ; Represents a 5-dimensional scalar; in, , , The calculation formula is as follows: (3); In the formula, express The identity matrix; express The transpose of ; express The transpose of ; Define vertices The quadratic error matrix is: (4); In the formula: Represents vertices The set of all triangular faces within a first-order neighborhood; Indicates belonging to The triangular facets of a set; Represents the calculation of vertices To the triangle One of the processes of squaring distance Matrix; Represents the calculation of vertices To the triangle A 5-dimensional vector in the process of squared distance; Represents the calculation of vertices To the triangle A 5-dimensional scalar in the process of squared distance; express The transpose of ; When the edge Simplify to new vertex At that time, for ease of calculation, take As The quadratic error matrix, edge Simplified error and newly generated vertices The calculation formula is as follows: (5); In the formula: Represents vertices and The set of triangular faces within a first-order neighborhood; Represents vertices The quadratic error matrix; Represents vertices The quadratic error matrix; Indicates will The coordinate vector increases from 5 dimensions to 6 dimensions, with all added elements being 1; (b): If the vertex and vertex One of them is a smooth vertex, and the other is a corner vertex. If it is an angle vertex, then the edge Continue after simplification As the position of the newly generated vertex; (c): If the vertex and vertex One of them is a smooth vertex, and the other is a seam vertex. If it is the vertex of the seam, then the edge Continue after simplification As newly generated vertices; (d): If the vertex and vertex All are seam vertices, and the edges If it is located at the seam between two different texture blocks, then Also the vertices of the seam, take the edge. The midpoint as spatial coordinates, Texture coordinates and Average texture coordinates: (6); Step four, the simplified strategy for preventing edge breakage, specifically includes: 1) Vertex and vertex All of them are non-boundary vertices or all of them are boundary vertices, and can be directly simplified according to the above cases (a)-(d); 2) , There is one boundary vertex and one non-boundary vertex in the model. If the model is simplified, the boundary of the model will be damaged. Therefore, the simplification of the model is prohibited. Step five involves calculating the vertex undulation factor, the area factor of the first-order neighborhood triangle of the vertex, and the texture deformation factor in the oblique photogrammetry model. Specifically, this includes: Introducing a peak fluctuation factor The calculation formula is as follows: (7); In the formula: Represents vertices Vertex variability factor; It is the vertex The set of adjacent vertices within a first-order neighborhood; Represents vertices normal vector and vertex The angle between the normal vectors, express The vertex normal vector; express The vertex normal vector; It is the vertex and vertex Euclidean distance; It is the vertex The number of adjacent vertices in a first-order neighborhood; Indicates belonging to a vertex The vertices of the set of adjacent vertices within a first-order neighborhood; In the formula for calculating the vertex normal vector, the area of the triangular mesh within the first-order neighborhood of the vertex is introduced. and the interior angles associated with the vertex. The optimized formula for calculating vertex normal vectors is obtained as follows: (8); In the formula: express The vertex normal vector; Represents vertices The set of all triangular faces within a first-order neighborhood; For the first Triangle facets and vertices Associated interior angles; Indicates the first The geometric area of each triangular facet; Indicates the first The normal vectors of the triangular faces; Introducing the area factor of the first-order neighborhood triangle at the vertex The calculation formula is as follows: (9); In the formula: Indicates the first The geometric area of each triangular facet; It is the vertex The number of adjacent vertices in a first-order neighborhood; Represents vertices The set of all triangular faces within a first-order neighborhood; Introducing texture deformation factor The calculation formula is as follows: (10); In the formula: Representing an edge Texture deformation factor; Indicates the first The texture area of each triangular facet; Represents vertices The average texture density of a triangular mesh within a first-order neighborhood measures the proportion of texture area to geometric area within the first-order neighborhood of a vertex. The larger the value, the higher the value of the vertex. The denser the texture in the surrounding area; and They represent and The vertex type adjustment factor, which takes a real number greater than zero, is used to distinguish the importance of vertices of different categories. Represents vertices Average texture density of a first-order triangular mesh within the neighborhood; Represents vertices Average texture density of a first-order triangular mesh within the neighborhood; Represents vertices The number of triangular faces within a first-order neighborhood.
2. The method for simplifying an oblique photogrammetry model to prevent texture deformation according to claim 1, characterized in that: In step one, during the preprocessing stage, the coordinate information of the vertices in the oblique photogrammetry model is first read, where... , , Represents the spatial coordinates of the vertex; , This process involves retrieving the normalized texture coordinates of the vertices and storing these coordinates in a vertex list. Then, it continues reading the edge information from the model, storing the vertices that form the edges in an edge list. Finally, it reads the triangle information from the model and stores the numbers of the three edges that form the triangles in a triangle list.
3. The method for simplifying an oblique photography model to prevent texture deformation according to claim 1, characterized in that: In step two, when a vertex corresponds to only one texture coordinate... When a vertex has two corresponding texture coordinates, it is defined as a smooth vertex; when a vertex has two corresponding texture coordinates... When a vertex has three or more corresponding texture coordinates, it is defined as a seam vertex; when a vertex has three or more corresponding texture coordinates... Define such vertices as corner vertices.
4. The method for simplifying an oblique photography model to prevent texture deformation according to claim 1, characterized in that: In step three, edges with only one associated triangle in the oblique photogrammetry model are defined as boundary edges, and edges with two or more associated triangles are defined as non-boundary edges. When a boundary edge exists in the first-order neighborhood of a vertex, the vertex is identified as a boundary vertex; when no boundary edge exists in the first-order neighborhood of a vertex, the vertex is identified as a non-boundary vertex.
5. The method for simplifying an oblique photogrammetry model to prevent texture deformation according to claim 1, characterized in that: In step six, the cost is simplified. The calculation formula is as follows: (11); In the formula: Represents vertices First-order neighborhood triangle area factor; Represents vertices First-order neighborhood triangle area factor; Represents vertices The quadratic error matrix; Represents vertices The quadratic error matrix; Represents vertices Vertex variability factor; Represents vertices The vertex undulation factor.
6. The method for simplifying an oblique photogrammetry model to prevent texture deformation according to claim 1, characterized in that: In step seven, let The edge that currently fits the simplification strategy and minimizes the simplification cost in the model. For the edge Simplified newly generated vertex positions For the edge Adjacent vertices; delete edges The two original vertices , At the same time, delete with , Connected edges, then use Replace the edge and re-establish The topological connections with adjacent vertices are used to complete a simplified operation.
7. The method for simplifying an oblique photogrammetry model to prevent texture deformation according to claim 1, characterized in that: In step eight, the quadratic error of the affected edges, the new vertices, and the simplification cost ranking are updated. Then, based on the new simplification cost ranking, the edge with the smallest simplification cost is selected for simplification, until the number of triangles in the oblique photogrammetry model is reached. Simplification terminates after satisfying equation (12); (12); In the formula: This represents the expected value of the triangular facet of the oblique photography model.
Citation Information
Patent Citations
Three-dimensional model simplification method suitable for model with textures
CN103714577A
Three-dimensional model simplification method and system
CN108460836A
3D model simplification method based on skeletonization
CN109934928B
