A three-dimensional model lightweight method based on complex equipment

Abstract

The application faces industrial manufacturing processes and provides a three-dimensional model lightweight method based on complex equipment, supports lossless information data transmission of a complex product model based on a hierarchical bounding box algorithm, solves the problem of reduced loading speed and unsmooth browsing when the three-dimensional model data volume increases, reduces the vertex and triangular face quantity of the model while meeting error requirements, improves the rendering display rate of the three-dimensional model, achieves the purpose of efficient loading and smooth browsing of the three-dimensional model, and meets the demand of collaborative real-time display of the complex product model.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing, specifically a method for lightweighting three-dimensional models of complex equipment. Background Technology

[0002] With the rapid development and widespread application of computer-aided design (CAD) technology, manufacturing enterprise informatization has evolved from the two-dimensional era to the three-dimensional era. Aerospace companies are gradually promoting digital design and manufacturing models, and models with Product Manufacturing Information (PMI) have become the standard in the design, process, and production stages. The process and production stages have a pressing need for Model-Based Definition (MBD) browsing. In digital manufacturing, Pro / E is used to browse the original model with PMI. However, the original model is still in editing mode during browsing, making it easy to accidentally modify the original design model. Furthermore, the 3D modeling software installed on the field terminals has high requirements for computer hardware. At the same time, the structure of satellite products is becoming increasingly complex, and the 3D models are becoming increasingly large, making it impossible for ordinary computers to open large models. The display, browsing, and use of large models and MBD models are becoming increasingly difficult. Therefore, it is necessary to lightweight large models to achieve rapid display, browsing, and convenient use.

[0003] During the product design phase, complex assembly models are broken down into multiple sub-assembly models, designed by engineers from different departments. These individual part and component models are then assembled into a complete final assembly model. However, different designers or engineering departments may use different 3D design software, leading to model heterogeneity issues. With the widespread application of lightweight technology, lightweight file formats will become a standard data format for 3D product models. Lightweight files enable the transfer and exchange of CAD data between different companies and departments, simplifying and approximating product models without missing parts, significantly improving inter-company communication. Furthermore, the small size and ease of operation of lightweight 3D models facilitate communication, coordination, and viewing between engineering and process designers, company leaders, sales personnel, and customers. Therefore, lightweight technology has wide applications in the rapid browsing and publishing of part models and complex assembly models. Summary of the Invention

[0004] To address the problems of existing technologies, this invention proposes a three-level lightweighting method for complex equipment models in industrial manufacturing processes. This method supports the lossless transfer of assembly tree algorithm information in complex product models during processing, solving the problem of reduced loading speed and unsmooth browsing when the data volume of 3D models increases. While meeting error requirements, it reduces the number of vertices and triangles in the model, improves the rendering and display rate of 3D models, and achieves the goal of efficient loading and smooth browsing of 3D models, thus meeting the needs of collaborative real-time display of complex product models.

[0005] To solve the above problems, the technical solution of the present invention is as follows:

[0006] A method for lightweighting 3D models of complex equipment involves breaking down the complex equipment into structural parts and performing lightweighting steps on each part model to achieve lightweighting of the complex equipment model. The steps include:

[0007] Step 1: Layer and merge the texture image data of the part model, recalculate and record the texture coordinates; apply the same material to texture images with the same texture.

[0008] Step 2: Perform lightweight processing on the texture map, filtering out the modeling history and feature definition parameters;

[0009] Step 3: Transform the 3D parametric model of the texture map into a discretized triangular mesh model through triangulation.

[0010] Step 4: Compress the triangular mesh model and reconstruct the model topology data;

[0011] Step 5: Compress the number of vertices in the triangular mesh model;

[0012] Step 6: Use batch rendering to simultaneously draw multiple texture patches in a single rendering task to achieve lightweighting of complex equipment models.

[0013] The layer merging is as follows: for texture images in the same layer, an edge folding algorithm is used to merge multiple small texture images into one large continuous detail layer texture image, avoiding boundary smoothing and texture separation.

[0014] The modeling history and feature definition parameters include: sketches, modeling history, tolerances, roughness, annotation text, and annotation information.

[0015] The filtering process uses the "step" format as an intermediate data format for data transfer, optimizing the deletion of modeling history and feature definition parameters.

[0016] In the process of transforming a three-dimensional parametric model into a discretized triangular mesh model through triangulation, the matching relationship between the triangulated edges and faces and the parametric edges and faces of the original accurate geometric model is recorded simultaneously.

[0017] The compressed triangular mesh model is implemented using a general lossless compression coding algorithm, resulting in a lighter model with smaller files.

[0018] The processing steps of the lossless compression coding algorithm include:

[0019] The model topology data is reconstructed by compressing the encoding width of the triangular facet parameters; the triangular facet parameters include: vertex coordinates, normals, facet indexes, and materials;

[0020] The compressed triangle texture map is written to the Tile's glb in binary mode.

[0021] The number of vertices in the compressed triangular mesh model is achieved by gradually reducing the number of vertices in the model using the LOD method, which reduces the amount of rendering data on the GPU and improves the smoothness of the 3D scene.

[0022] The batch rendering method is as follows: texture patches with consistent rendering states are combined into one large patch; placed in one DrawCall and drawn in one rendering task, so that the GPU does not need to switch rendering states multiple times, thus improving rendering efficiency; the rendering states include: materials, textures, programmable pipelines, and geometric information.

[0023] The present invention has the following beneficial effects and advantages:

[0024] 1. This invention proposes a multi-level lightweight representation method based on complex models.

[0025] 2. This invention proposes to use a general lossless compression coding algorithm to compress a triangular mesh model and reconstruct the model topology data to obtain a lighter model with smaller files, thereby achieving a higher compression ratio.

[0026] 3. Compared with traditional 3D model lightweighting, this invention transforms the 3D parametric model into a discretized triangular mesh model through triangulation, using the triangular mesh model to approximate the accurate 3D model. Triangulation of the model's parametric edge and face data accelerates model display and reduces computer memory usage. Simultaneously, it records the matching relationship between the triangulated edges and faces and the original accurate geometric model's parametric edges and faces, enabling accurate measurement and annotation during rapid model display.

[0027] 4. This invention designs a three-layer lightweight representation model loading and display mechanism, which improves the loading speed and display smoothness of large assembly models. Attached Figure Description

[0028] Figure 1 Lightweight method flowchart;

[0029] Figure 2 Vertex quantization processing flowchart;

[0030] Figure 3 Data structure diagram of complex equipment parts;

[0031] Figure 4 Structural diagram of a lightweight model for complex equipment;

[0032] Figure 5 Rendering and display mechanism diagram. Detailed Implementation

[0033] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0035] like Figure 1 The diagram shown is a flowchart of the method of this invention. A method for lightweighting a 3D model of complex equipment involves refining the complex equipment into structural parts and performing lightweighting steps on each part model to achieve lightweighting of the complex equipment model; the steps include:

[0036] Step 1: Layer and merge the texture image data of the part model, recalculate and record the texture coordinates; apply the same material to texture images with the same texture.

[0037] Step 2: Perform lightweight processing on the texture map, filtering out the modeling history and feature definition parameters;

[0038] Step 3: Transform the 3D parametric model of the texture map into a discretized triangular mesh model through triangulation.

[0039] Step 4: Compress the triangular mesh model and reconstruct the model topology data;

[0040] Step 5: Compress the number of vertices in the triangular mesh model;

[0041] Step 6: Use batch rendering to simultaneously draw multiple texture patches in a single rendering task to achieve lightweighting of complex equipment models.

[0042] Specifically as follows:

[0043] Factors affecting 3D scene rendering performance include CPU, GPU, and bandwidth. Reduced CPU performance can be caused by excessive draw calls, complex scripts, or physics simulations. Excessive GPU usage and per-vertex calculations can waste GPU resources. 3D models consist of geometric information, materials, and textures. To reduce model file size, texture sizes must be limited. Compressing texture images and geometric information reduces the 3D model file size, facilitating network transmission.

[0044] The algorithm merges texture images used in the model, recalculates and records texture coordinates, and combines multiple small texture images into a single large texture image. Using the same texture image allows the use of the same material, and sampling the texture image using different sampling coordinates achieves correct texture mapping. The CPU does not need to call the graphics API multiple times; the merged large texture image can be loaded in a single DrawCall.

[0045] Textures using the same material differ only in their model geometry, such as vertices, indices, normals, and texture coordinates. If these data are merged within an allowed range, and the merged vertex cache object is submitted to the GPU, the 3D model can be rendered in a single rendering batch.

[0046] In traditional 3D model display, bandwidth is a crucial factor affecting browser loading speed. Compressing network-transmitted data maximizes bandwidth savings and improves the 3D visual experience. Crunch is an open-source compression algorithm with fast decompression speeds. This algorithm significantly reduces texture storage size, lowers video memory usage and data request time, compresses merged textures, and writes the compressed textures into the tile's GLB in binary format. Decompression is performed during use, improving data parsing and rendering efficiency.

[0047] In the patch data, the original data type of the coordinate components of each vertex is 32-bit float. Using the compression algorithm in this paper, the storage width is reduced to 8 bits or 16 bits (determined by the error requirement) through quantization compression. Figure 2 As shown, first, the vertex list of the model is obtained, and the minimum and maximum points of the model (i.e., the two corner points of the model's bounding box) are obtained; then, the absolute coordinates of each vertex in the original world coordinate system are mapped to relative coordinates relative to the minimum and maximum points.

[0048] The model data needs to generate multi-level LOD models. If the same high-precision texture is used for each level, it will result in texture waste. The merged large texture image is resampled to generate low-precision mipmap levels level by level, ensuring that different precision texture images are used at each level, minimizing GPU memory usage. In a 3D scene, if the camera position is far away, it is difficult for the naked eye to distinguish the details of the model. LOD technology (Level of Detail algorithm) is used to gradually reduce the number of vertices in the model, thereby reducing the amount of GPU rendering data and improving the smoothness of the 3D scene. Multiple models with different levels of detail are prepared for the same object, and the models are assigned to different LOD levels. The parent-child relationships of models with different levels of detail are organized, and camera reference values ​​are set. When loading models, Cesium automatically determines which level of model to load based on the camera position.

[0049] Triangulation transforms a 3D parametric model into a discretized triangular mesh model, which approximates the accurate 3D model. Triangulation of the model's parametric edge and face data speeds up model display and reduces computer memory usage. An edge-folding algorithm generates continuous levels of detail. This algorithm's advantage lies in removing unnecessary hard edges and texture transitions, avoiding boundary smoothing and texture separation. By improving the edge-folding algorithm, the simplification ratio of triangular faces can be set, maintaining consistency in the simplified model within acceptable limits. The first layer of simplification represents the highest model accuracy, and the nth layer represents the lowest. The first layer of lightweight model is obtained through parametric curve and surface triangulation, resulting in the largest number of triangular faces. The second to nth layers of lightweight model are obtained through edge-folding simplification of the previous layer's mesh model, with the number of triangular mesh faces gradually decreasing, reaching a minimum in the nth layer. In the lightweight model representation, each part consists of n groups of triangular mesh faces, each group comprising multiple triangular faces. Each triangular face contains triangular edges and vertices, and each triangular face group has its own properties. This includes area, number of faces, total number of sides, color, layers, bounding box, etc. The part data structure diagram is as follows: Figure 3 As shown.

[0050] To meet the need for rapid browsing and display of large and complex assembly models, theoretically, an n-layer lightweight representation model with varying precision can be designed. However, the more layers there are, the larger the model becomes, and the lower the model conversion efficiency, resulting in significant memory consumption during model loading. Therefore, this invention designs a three-layer lightweight representation model with different precision. The first layer of lightweight representation for the part model obtains triangular facets by triangulating the original parametric surfaces and curves. The second and third layers of lightweight model are obtained using an edge-folding mesh simplification algorithm, always tracking and recording the number of triangular facets after each edge-folding operation. The triangulation error (the maximum distance between a point on the surface and the replacement triangular facet) is used, with a smaller triangulation error for small models and a larger triangulation error for large models. The first and second layers of lightweight model are used to represent visible parts in the assembly model. To avoid severe model deformation while maintaining a relatively small number of facets, the number of facets in the second layer is designed to be 55% of the number in the first layer. The third layer of lightweight representation model is only used to represent invisible parts in the assembly model, ensuring that browsing and viewing are not affected even if the model undergoes significant deformation. The number of facets in the third layer is designed to be 35% of the number in the second layer.

[0051] Three-dimensional part models contain a large amount of engineering semantic information, including datum, coordinate system, feature definitions / parameters, modeling history, and PMI information. Assembly models are composed of part models. When the same component is assembled multiple times in an assembly model, the geometric information of the same part is repeatedly recorded within the assembly file. This results in an increasingly large amount of data and a more complex structure in the assembly model. Browsing these complex assembly models on ordinary computers is very slow or even impossible. To improve the problem of large data volume and complex structure in complex assembly models, this invention describes assembly data information and part data information separately. Assembly data information records the tree-like hierarchical structure information of each component, assembly position relationships, and relative reference relationships between components. When the model is displayed, the assembly directly calls the relative reference paths of components for model representation, without repeatedly recording geometric information such as part edges and faces. The part list includes the names of all parts that make up the entire assembly model, and repeated part names in the assembly model are only described once. Parts in the part list are not displayed in the assembly model; only when a part is referenced are the parts correctly displayed based on the relative path and name of the referenced part in the assembly model and the transformation matrix. Part data information records attribute lists of edges and faces, triangulated edges, triangulated faces, points, and other data. Lightweight assembly model structure diagram as follows Figure 4 As shown.

[0052] When browsing large and complex assembly models, the system first loads a lightweight representation model of all component models with the first level of precision. Since this first-level lightweight representation model is obtained directly through triangulation, it doesn't require simplification of the mesh model. Therefore, the model loading and display speed is fast, and the computer's memory usage is low. As the number of loaded part models increases, the number of triangular faces in the assembly model also gradually increases, making model display increasingly difficult. Eventually, this will exceed the computer's face rendering capabilities, and the assembly model will become unreadable. At this point, a lower-precision lightweight representation model (the second-level lightweight representation model) is used to replace the already loaded first-level lightweight representation model. Subsequent loaded component models are also represented using the second-level lightweight model. When model display becomes difficult, a third-level lightweight representation model is used to replace invisible part models in the assembly model. That is, visible parts in the assembly model are displayed using the second-level lightweight representation model, while internal invisible components are displayed using the third-level lightweight representation model. When this still fails to meet display requirements, the third-level lightweight representation model is hidden, i.e., invisible parts in the assembly model are hidden, ultimately achieving fast display and browsing of large and complex assembly models. The loading and display mechanism diagram is shown below. Figure 5 As shown.

[0053] Examples of lightweighting of 3D models described in this article are as follows:

[0054] This invention uses a gas turbine (containing approximately 4.5 million triangular faces), a main structural component model of a regional jet (containing approximately 14 million triangular faces), and a fuselage section model of a Boeing 777 (containing approximately 24 million triangular faces) as verification objects to verify the generation algorithm of the hierarchical bounding box tree, the model segmentation algorithm, and the model compression algorithm. The hierarchical bounding box tree generation experiment tested the time required to complete the generation of the hierarchical bounding box tree for the three models and the generation results for the gas turbine and regional jet models. It can be seen that the running time of the hierarchical bounding box tree generation algorithm for large-scale complex 3D models is roughly proportional to the number of faces. Since the hierarchical bounding box tree generation algorithm only needs to be executed once before real-time rendering, the system does not have high requirements for its generation efficiency. Experimental results show that the three tested models can all be completed within 100ms using the hierarchical bounding box tree generation algorithm, thus meeting the requirements in terms of generation efficiency.

[0055] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions made by those skilled in the art within the technical scope disclosed in the present invention are all within the scope of protection of the present invention. Therefore, the scope of protection of the present invention is determined by the scope of the claims.

Claims

1. A method for lightweighting a 3D model of complex equipment, comprising refining the complex equipment into structural parts and performing lightweighting steps on each part model to achieve lightweighting of the complex equipment model; characterized in that, The steps include: Step 1: Layer and merge the texture image data of the part model, recalculate and record the texture coordinates; apply the same material to texture images with the same texture. Step 2: Perform lightweight processing on the texture map, filtering out the modeling history and feature definition parameters; Step 3: Transform the 3D parametric model of the texture map into a discretized triangular mesh model through triangulation. Step 4: Compress the triangular mesh model and reconstruct the model topology data; Step 5: Compress the number of vertices in the triangular mesh model; Step 6: Use batch rendering to simultaneously draw multiple texture patches in a single rendering task to achieve lightweighting of complex equipment models.

2. The method for lightweighting complex equipment models according to claim 1, characterized in that, The layer merging is as follows: for texture images in the same layer, an edge folding algorithm is used to merge multiple small texture images into one large continuous detail layer texture image, avoiding boundary smoothing and texture separation.

3. The method for lightweighting complex equipment models according to claim 2, characterized in that, The modeling history and feature definition parameters include: sketches, modeling history, tolerances, roughness, annotation text, and annotation information.

4. The method for lightweighting complex equipment models according to claim 1, characterized in that, The filtering process uses the "step" format as an intermediate data format for data transfer, optimizing the deletion of modeling history and feature definition parameters.

5. The method for lightweighting complex equipment models according to claim 1, characterized in that, In the process of transforming a three-dimensional parametric model into a discretized triangular mesh model through triangulation, the matching relationship between the triangulated edges and faces and the parametric edges and faces of the original accurate geometric model is recorded simultaneously.

6. The method for lightweighting complex equipment models according to claim 1, characterized in that, The compressed triangular mesh model is implemented using a general lossless compression coding algorithm, resulting in a lighter model with smaller files.

7. The method for lightweighting complex equipment models according to claim 6, characterized in that, The processing steps of the lossless compression coding algorithm include: The model topology data is reconstructed by compressing the encoding width of the triangular facet parameters; the triangular facet parameters include: vertex coordinates, normals, facet indexes, and materials; The compressed triangle texture map is written to the Tile's glb in binary mode.

8. The method for lightweighting complex equipment models according to claim 1, characterized in that, The number of vertices in the compressed triangular mesh model is achieved by gradually reducing the number of vertices in the model using the LOD method, which reduces the amount of rendering data on the GPU and improves the smoothness of the 3D scene.

9. The method for lightweighting complex equipment models according to claim 8, characterized in that, The batch rendering method is as follows: combine texture patches with the same rendering state into one large patch; By placing it in a single DrawCall and drawing it in a single rendering task, the GPU does not need to switch rendering states multiple times, thus improving rendering efficiency. The rendering states include: materials, textures, programmable pipeline, and geometry information.

Images