A method and system for rapid acquisition of 3D models of virtual simulation scenes

By constructing a view frustum clipping network and using an OpenGL perspective projection matrix to filter invalid triangular faces, and combining it with multi-threaded texture mapping, the problem of low efficiency in 3D model acquisition in virtual simulation scenes is solved, achieving efficient 3D reconstruction and texture restoration.

CN116503577BActive Publication Date: 2026-11-10SOUTH SURVEYING & MAPPING INSTR
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202310372800.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-07
Publication Date
2026-11-10
Estimated Expiration
2043-04-07

AI Technical Summary

Technical Problem

Existing technologies in virtual simulation scenarios suffer from high computational resource consumption and difficulty in restoring textures in ROI regions during the acquisition of 3D models. In particular, the depth estimation and mesh reconstruction efficiency of multi-view stereo matching algorithms is low during the loading, browsing, rendering, and interaction of large virtual 3D scenes.

Method used

By capturing multi-view images of the ROI region of a virtual simulation scene, a frustum clipping network is constructed to clip the original mesh. Invalid triangular faces are filtered out using an OpenGL perspective projection matrix, the best image is selected, and block-based multi-threaded texture mapping is performed to generate a realistic 3D model.

Benefits of technology

It improves the efficiency of 3D model acquisition, reduces the computational resource consumption of redundant and overlapping triangular faces, and achieves efficient 3D reconstruction under realistic simulation conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116503577B_ABST
    Figure CN116503577B_ABST
Patent Text Reader

Abstract

This invention proposes a method and system for rapid acquisition of 3D models of virtual simulation scenes, relating to the technical field of image processing. After acquiring multi-view images of the Region of Interest (ROI) and the position and pose information of the multi-view images, an initial set of visible triangles of the multi-view images is obtained by constructing a view frustum clipping network. Then, the initial set of visible triangles is filtered using an OpenGL perspective projection matrix, and the best image of the remaining visible triangles is selected. Based on the filtered visible triangles, a triangular mesh scene model of the ROI region is generated. After dividing the triangular mesh scene model of the ROI region into blocks, multi-threaded texture mapping is performed to output a 3D model in a realistic simulation state, thereby improving the model acquisition efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically, to a method and system for rapid acquisition of 3D models of virtual simulation scenes. Background Technology

[0002] Virtual simulation is an important application scenario in 3D reconstruction. A virtual simulation scenario first needs to reconstruct the 3D model of all elements in the scene, create a simulation environment, provide real scale information of all elements, and carry out human-computer interaction, simulation and other processes in the virtual and real scene. At the same time, it can provide multi-dimensional data semantic segmentation, 3D bounding boxes and depth annotation, and can also simulate environmental rendering under different parameters, such as lighting and weather.

[0003] Due to widespread user adoption and the demand for interactive applications, virtual simulation scenes are generally composed of multiple refined 3D models pieced together to form the entire virtual simulation scene. This creates two problems: First, because the various independent refined 3D models are superimposed on each other, there is a large amount of overlapping and redundant triangular faces at the splicing points between the models, which increases the consumption of additional computing resources during the loading, browsing, rendering, and interaction processes of large virtual 3D scenes. Second, in order to perceive a more realistic 3D environment, virtual simulation scenes will add different lighting, weather, and other environmental rendering to the original models. Moreover, users often focus on the ROI (Region of Interest) scene, that is, a specific area or a small part of a large scene. At this time, it is necessary to crop, splice, and collect the triangular face models of the ROI region in the virtual simulation scene. However, each independent model uses its own different textures, and is affected by the real-time lighting, weather, and other rendering in the virtual scene, making it difficult to restore the texture of the ROI scene model in a realistic simulation state. To overcome these shortcomings, existing technologies typically use multi-view stereo matching algorithms to estimate the depth map of each view of the ROI scene, and then fuse the depth maps to obtain the final point cloud and texture model reconstruction results. However, the depth estimation and mesh reconstruction of this technology are extremely time-consuming and inefficient. Summary of the Invention

[0004] To address the problem of inefficient acquisition of realistic texture models in existing virtual simulation scenes, this invention proposes a method and system for rapid acquisition of 3D models in virtual simulation scenes. Given an existing virtual scene mesh, it eliminates the need to generate a mesh model using a multi-view stereo matching algorithm. By cropping and stitching visible triangular faces from various perspectives and utilizing optimal image texture mapping, a 3D model in a realistic simulation state can be output, thus improving model acquisition efficiency.

[0005] To achieve the above-mentioned technical effects, the technical solution of the present invention is as follows:

[0006] A method for rapid acquisition of 3D models of virtual simulation scenes includes:

[0007] S1. Surround and photograph the ROI region of the virtual simulation scene, and collect multi-view images of the ROI region, as well as the position and pose information of the multi-view images;

[0008] S2. Construct a view frustum clipping network, and use the view frustum clipping network to clip the original mesh of the virtual simulation scene to obtain the initial set of visible triangles of the multi-view image of the ROI region;

[0009] S3. Construct an OpenGL perspective projection matrix based on the position and pose information of the multi-view images of the ROI region. Filter invalid triangles in the initial set of visible triangles using the OpenGL perspective projection matrix. Merge the remaining visible triangles after filtering to construct the triangular mesh scene model of the ROI region. Select the best image for each triangle in the set of visible triangles.

[0010] S4. Divide the triangular mesh scene model of the ROI region into blocks to obtain each sub-block model;

[0011] S5. Perform multi-threaded texture mapping on the sub-block model based on the optimal image, and output the texture-mapped 3D model.

[0012] This technical solution does not require generating a mesh model. It obtains the visible triangular faces of the multi-view image of the virtual simulation scene through view frustum clipping and OpenGL perspective projection matrix filtering, constructs the triangular mesh scene model of the ROI region, performs texture mapping after block division, and outputs a 3D model in a realistic simulation state.

[0013] Preferably, in step S2, the view frustum clipping network consists of several view frustums corresponding to multi-view images. Each view frustum consists of a far clipping plane, a near clipping plane, and six planes: top, bottom, left, and right. The eight vertices of the view frustum are solved based on the position and pose information of each multi-view image, and the six planes of the view frustum are determined. The vertex calculation formula is as follows:

[0014] ntl=(0,0,N); nbl=(0,height*N,N); ntr=(width*N,0,N); nbr=(width*N,height*N,N)

[0015] ftl=(0,0,F); fbl=(0,height*F,F); ftr=(width*F,0,F); fbr=(width*F,height*F,F)

[0016] Where ntl, nbl, ntr, nbr, ftl, fbl, ftr, and fbr represent the eight vertices of the view frustum, N is the distance from the camera center of the multi-view image to the near clipping plane, F is the distance from the camera center of the multi-view image to the far clipping plane, height represents the image height, and width represents the image width.

[0017] Determine whether any vertex of a triangle in the original triangular mesh of the virtual simulation scene is outside any of the six planes. If so, the vertex of the triangle is outside the view frustum; otherwise, the vertex of the triangle is inside the view frustum. Add the vertex of the triangle to the initial set of visible triangles.

[0018] Preferably, in step S3, the OpenGL perspective projection matrix is ​​expressed by the following formula:

[0019]

[0020] Where M represents the OpenGL perspective projection matrix, left and right represent the left and right ranges of the view frustum space, top and bottom represent the upper and lower ranges of the view frustum space, N represents the near plane of the view frustum space, and F represents the far plane of the view frustum space.

[0021] Preferably, in step S3, let P be any three-dimensional coordinate point of the virtual simulation scene model, and perform a projection transformation on M. The projection transformation formula is as follows:

[0022]

[0023] The coordinates after projection transformation are normalized to homogeneous coordinates to obtain the projection depth and two-dimensional coordinates of each vertex of the triangle in the initial set of visible triangles. This is used to filter out invalid triangles in the initial set of visible triangles.

[0024] Preferably, in step S3, based on the projection depth of each vertex, using the triangle normal vector orientation constraint and triangle interpolation depth, the initial set of visible triangles within the view frustum is occlusion judged, invalid and redundant triangles are removed, and the visible triangles after viewpoint filtering for each image are obtained.

[0025] Preferably, in step S3, the image with the largest area of ​​the two-dimensional projection of each visible triangular face is selected as the best image to prepare for subsequent texture mapping.

[0026] Here, the best images were selected while reducing redundant and overlapping triangular faces, thus improving acquisition efficiency.

[0027] Preferably, during the process of constructing the triangular mesh scene model of the ROI region, the visible triangular faces after filtering each image are added to the new scene model, and the key-value pair relationship between the triangular faces and vertex indices of the triangular faces in the original virtual simulation scene model and the triangular mesh scene model of the ROI region is constructed based on the projection depth and two-dimensional coordinates of the vertices of each visible triangular face.

[0028] Preferably, in step S4, the process of dividing the triangular mesh scene model of the ROI region into blocks includes:

[0029] SA. Calculate the bounding box coordinate range of the triangular mesh scene model of the ROI region;

[0030] SB. Set the block size, and according to the block size, uniformly cut the triangular mesh scene model of the ROI region within the coordinate range to obtain each sub-block model.

[0031] Here, by dividing the triangular mesh scene model of the ROI region into blocks, the efficiency of the texture mapping process is improved.

[0032] Preferably, in step S5, the specific process of multi-threaded texture mapping is as follows: the coordinates of the vertices of each visible triangle in the sub-block model are projected onto the corresponding best image to obtain the two-dimensional coordinates of the triangle on the corresponding best image. The best image is cropped according to the two-dimensional coordinates to obtain the colored two-dimensional triangle. The colored two-dimensional triangle is copied to the texture image. Finally, the texture images of all the two-dimensional triangles are arranged to obtain the final texture image, and the texture-mapped three-dimensional model is output.

[0033] Here, the best image within the shooting field of view is cropped to ensure that the output 3D model perfectly matches the 3D reconstruction process of the real scene, making the 3D reconstruction process of the virtual scene more reasonable.

[0034] This application also proposes a rapid acquisition system for 3D models of virtual simulation scenes, the system comprising:

[0035] The image acquisition module is used to acquire multi-view images of the ROI region and the position and pose information of the multi-view images;

[0036] The visible triangle construction module is used to build the view frustum clipping network and generate the initial set of visible triangles for multi-view images;

[0037] The visible triangle filtering module is used to filter out invalid triangles in the initial set of visible triangles and select the best image for each visible triangle.

[0038] The triangular mesh scene model building module is used to generate triangular mesh scene models of ROI regions based on the filtered visible triangular faces;

[0039] The triangular mesh scene model segmentation module is used to segment the triangular mesh scene model of the ROI region into multiple sub-block models;

[0040] The texture mapping module is used to perform multi-threaded texture mapping on the sub-block model based on the best image, and output the texture-mapped 3D model.

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

[0042] This invention proposes a method and system for rapid acquisition of 3D models of virtual simulation scenes. Based on existing mesh models of virtual simulation scenes, after acquiring multi-view images of the Region of Interest (ROI) and the position and pose information of the multi-view images, an initial set of visible triangles of the multi-view images is obtained by constructing a view frustum clipping network. Then, the initial set of visible triangles is filtered using an OpenGL perspective projection matrix, and the best image of the remaining visible triangles is selected. Based on the filtered visible triangles, a triangular mesh scene model of the ROI region is generated. After dividing the triangular mesh scene model of the ROI region into blocks, multi-threaded texture mapping is performed to output a realistic 3D model in the virtual simulation state, thereby improving the model acquisition efficiency. Attached Figure Description

[0043] Figure 1 A flowchart illustrating the rapid acquisition method for virtual simulation scene 3D models proposed in Embodiment 1 of the present invention;

[0044] Figure 2 This illustrates the surround shooting diagram proposed in Embodiment 1 of the present invention;

[0045] Figure 3 This represents a clear sky multi-view image acquisition diagram of the ROI region proposed in Embodiment 1 of the present invention;

[0046] Figure 4 This represents a multi-view image acquisition diagram of the ROI region on a cloudy day, as proposed in Embodiment 1 of the present invention.

[0047] Figure 5 This diagram illustrates the view frustum proposed in Embodiment 1 of the present invention.

[0048] Figure 6 This is a schematic diagram of the visible triangular surface filter proposed in Embodiment 1 of the present invention;

[0049] Figure 7 This diagram illustrates the output of the three-dimensional model of a sunny day as proposed in Embodiment 1 of the present invention.

[0050] Figure 8 This diagram illustrates the output of the three-dimensional model for cloudy weather proposed in Embodiment 1 of the present invention.

[0051] Figure 9 This diagram illustrates the block division of the triangular mesh scene model proposed in Embodiment 2 of the present invention.

[0052] Figure 10 This is a schematic diagram of the structure of the rapid acquisition system for virtual simulation scene 3D models proposed in Embodiment 3 of the present invention. Detailed Implementation

[0053] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent.

[0054] To better illustrate this embodiment, some parts of the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions;

[0055] It is understandable to those skilled in the art that some well-known details may be omitted from the accompanying drawings.

[0056] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0057] The positional relationships depicted in the accompanying drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0058] Example 1

[0059] like Figure 1 As shown in the figure, this embodiment proposes a method for rapid acquisition of 3D models of virtual simulation scenes, including the following steps:

[0060] S1. As Figure 2 As shown, the virtual simulation scene's Region of Interest (ROI) is surrounded and photographed, and multi-view images of the ROI, along with the position and pose information of these multi-view images, are collected. (See [link]). Figure 3 and Figure 4 , Figure 3 A clear sky image of the ROI region. Figure 4 A cloudy image of the ROI region;

[0061] S2. Construct a view frustum clipping network, and use the view frustum clipping network to clip the original mesh of the virtual simulation scene to obtain the initial set of visible triangles of the multi-view image of the ROI region;

[0062] The view frustum cropping network consists of several view frustums corresponding to multi-view images. See [link / reference] Figure 5 Each view frustum consists of a far clipping plane, a near clipping plane, and six planes: top, bottom, left, and right. Based on the position and pose information of each multi-view image, the eight vertices of the view frustum are calculated, and the six planes of the view frustum are determined. The formula for calculating the vertices is as follows:

[0063] ntl=(0,0,N); nbl=(0,height*N,N); ntr=(width*N,0,N); nbr=(width*N,height*N,N)

[0064] ftl=(0,0,F); fbl=(0,height*F,F); ftr=(width*F,0,F); fbr=(width*F,height*F,F)

[0065] Where ntl, nbl, ntr, nbr, ftl, fbl, ftr, and fbr represent the eight vertices of the view frustum, N is the distance from the camera center of the multi-view image to the near clipping plane, F is the distance from the camera center of the multi-view image to the far clipping plane, height represents the image height, and width represents the image width.

[0066] Determine whether any vertex of a triangle in the original triangular mesh of the virtual simulation scene is outside any of the six planes. If so, the vertex of the triangle is outside the view frustum; otherwise, the vertex of the triangle is inside the view frustum. Add the vertex of the triangle to the initial set of visible triangles.

[0067] S3. Construct an OpenGL perspective projection matrix based on the position and pose information of the multi-view images of the ROI region. Filter invalid triangles in the initial set of visible triangles using the OpenGL perspective projection matrix. Merge the remaining visible triangles after filtering to construct the triangular mesh scene model of the ROI region. Select the best image for each triangle in the set of visible triangles.

[0068] The formula for expressing the OpenGL perspective projection matrix is ​​as follows:

[0069]

[0070] Where M represents the OpenGL perspective projection matrix, left and right represent the left and right ranges of the view frustum space, top and bottom represent the upper and lower ranges of the view frustum space, N represents the near plane of the view frustum space, and F represents the far plane of the view frustum space.

[0071] Let P be any three-dimensional coordinate point in the virtual simulation scene model. The projection transformation of M is given by the following formula:

[0072]

[0073] The coordinates after projection transformation are normalized to homogeneous coordinates to obtain the projection depth and two-dimensional coordinates of each vertex of the triangle in the initial set of visible triangles, which are then used to filter out invalid triangles in the initial set of visible triangles.

[0074] Based on the projected depth of each vertex, occlusion is determined for the initial set of visible triangles within the view frustum using the triangle facet normal orientation constraint and the triangle facet interpolation depth. Figure 6 As shown, a visible triangle with an angle between the triangle normal vector N and the ray S greater than 0° and less than 90° is considered unoccluded; otherwise, it is considered occluded. Invalid and redundant triangles are removed to obtain the visible triangles after viewpoint filtering for each image. The image with the largest area of ​​the two-dimensional projection of each visible triangle is selected as the best image.

[0075] In the process of constructing the triangular mesh scene model of the ROI region, the visible triangular faces after filtering each image are added to the new scene model. Based on the projection depth and two-dimensional coordinates of the vertices of each visible triangular face, the key-value pair relationship between the triangular faces and vertex indices of the triangular faces in the original virtual simulation scene model and the triangular mesh scene model of the ROI region is constructed.

[0076] S4. Divide the triangular mesh scene model of the ROI region into blocks to obtain each sub-block model.

[0077] S5. Perform multi-threaded texture mapping on the sub-block model based on the optimal image, and output the texture-mapped 3D model. (See [link]) Figure 7 , Figure 8 ,in, Figure 7 The output is a 3D model of a sunny day. Figure 8 The output is a 3D model of a cloudy day;

[0078] The specific process of multi-threaded texture mapping is as follows: Project the coordinates of the vertices of each visible triangle in the sub-block model onto the corresponding best image to obtain the two-dimensional coordinates of the triangle on the best image. Crop the best image according to the two-dimensional coordinates to obtain the colored two-dimensional triangle. Copy the colored two-dimensional triangle to the texture image. Finally, arrange the texture images of all the two-dimensional triangles to obtain the final texture image and output the texture-mapped 3D model.

[0079] Example 2

[0080] In this embodiment, as Figure 9 As shown, the process of dividing the triangular mesh scene model of the ROI region into blocks includes:

[0081] SA. Calculate the bounding box coordinate range of the triangular mesh scene model of the ROI region;

[0082] SB. Set the block size, and according to the block size, uniformly cut the triangular mesh scene model of the ROI region within the coordinate range to obtain each sub-block model.

[0083] Example 3

[0084] In this embodiment, as Figure 10 As shown, a computer system for rapid acquisition of 3D models of virtual simulation scenes is proposed. The system includes:

[0085] Image acquisition module 101 is used to acquire multi-view images of the ROI region and the position and pose information of the multi-view images;

[0086] Visible triangle construction module 102 is used to construct the view frustum clipping network and generate an initial set of visible triangles for multi-view images;

[0087] The visible triangle filtering module 103 is used to filter invalid triangles in the initial set of visible triangles and select the best image for each visible triangle.

[0088] The triangular mesh scene model building module 104 is used to generate a triangular mesh scene model of the ROI region based on the filtered visible triangular faces.

[0089] The triangular mesh scene model segmentation module is used to segment the triangular mesh scene model of the ROI region into multiple sub-block models 105.

[0090] The texture mapping module 106 is used to perform multi-threaded texture mapping on the sub-block model based on the best image and output the texture-mapped 3D model.

[0091] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for rapid acquisition of 3D models of virtual simulation scenes, characterized in that, include: S1. Surround and photograph the ROI region of the virtual simulation scene, and collect multi-view images of the ROI region, as well as the position and pose information of the multi-view images; S2. Construct a view frustum clipping network, and use the view frustum clipping network to clip the original mesh of the virtual simulation scene to obtain the initial set of visible triangles of the multi-view image of the ROI region; S3. Construct an OpenGL perspective projection matrix based on the position and pose information of the multi-view images of the ROI region. Filter invalid triangles in the initial set of visible triangles using the OpenGL perspective projection matrix. Merge the remaining visible triangles after filtering to construct the triangular mesh scene model of the ROI region. Select the best image for each triangle in the set of visible triangles. S4. Divide the triangular mesh scene model of the ROI region into blocks to obtain each sub-block model; S5. Perform multi-threaded texture mapping on the sub-block model based on the optimal image, and output the texture-mapped 3D model. In step S2, the view frustum clipping network consists of several view frustums corresponding to multi-view images. Each view frustum consists of a far clipping plane, a near clipping plane, and six planes: top, bottom, left, and right. The eight vertices of the view frustum are calculated based on the position and pose information of each multi-view image, and the six planes of the view frustum are determined. The vertex calculation formula is as follows: ntl=(0,0,N);nbl=(0,height N,N);ntr=(width N,0,N);nbr=(width N,height N,N) ftl=(0,0,F);fbl=(0,height F,F);ftr=(width F,0,F);fbr=(width F,height F,F) Where ntl, nbl, ntr, nbr, ftl, fbl, ftr, and fbr represent the eight vertices of the view frustum, N is the distance from the camera center of the multi-view image to the near clipping plane, F is the distance from the camera center of the multi-view image to the far clipping plane, height represents the image height, and width represents the image width. Determine whether any vertex of a triangle in the original triangular mesh of the virtual simulation scene is outside any of the six planes. If so, the vertex of the triangle is outside the view frustum; otherwise, the vertex of the triangle is inside the view frustum. Add the vertex of the triangle to the initial set of visible triangles. In step S3, the image with the largest area of ​​the two-dimensional projection of each visible triangular face is selected as the best image.

2. The method for rapid acquisition of 3D models of virtual simulation scenes according to claim 1, characterized in that, In step S3, the OpenGL perspective projection matrix is ​​expressed as follows: Where M represents the OpenGL perspective projection matrix. left and right These represent the left and right extents of the view frustum space, respectively. top and bottom These represent the upper and lower ranges of the view frustum space, respectively. N is the distance from the photographic center of the multi-view image to the near cropping plane, and F is the distance from the photographic center of the multi-view image to the far cropping plane.

3. The method for rapid acquisition of 3D models of virtual simulation scenes according to claim 2, characterized in that, In step S3, let For any 3D coordinate point in the virtual simulation scene model, perform a projection transformation on M. The projection transformation formula is as follows: The coordinates after projection transformation are normalized to homogeneous coordinates to obtain the projection depth and two-dimensional coordinates of each vertex of the triangle in the initial set of visible triangles. This is used to filter out invalid triangles in the initial set of visible triangles.

4. The method for rapid acquisition of 3D models of virtual simulation scenes according to claim 3, characterized in that, In step S3, based on the projection depth of each vertex, the initial set of visible triangles within the view frustum is occlusion judged using the triangle normal vector orientation constraint and triangle interpolation depth. Invalid and redundant triangles are removed to obtain the visible triangles after viewpoint filtering for each image.

5. The method for rapid acquisition of 3D models of virtual simulation scenes according to claim 4, characterized in that, In the process of constructing the triangular mesh scene model of the ROI region, the visible triangular faces after filtering each image are added to the new scene model. Based on the projection depth and two-dimensional coordinates of the vertices of each visible triangular face, the key-value pair relationship between the triangular faces and vertex indices of the triangular faces in the original virtual simulation scene model and the triangular mesh scene model of the ROI region is constructed.

6. The method for rapid acquisition of 3D models of virtual simulation scenes according to claim 1, characterized in that, In step S4, the process of dividing the triangular mesh scene model of the ROI region into blocks includes: SA. Calculate the bounding box coordinate range of the triangular mesh scene model of the ROI region; SB. Set the block size, and according to the block size, uniformly cut the triangular mesh scene model of the ROI region within the coordinate range to obtain each sub-block model.

7. The method for rapid acquisition of 3D models of virtual simulation scenes according to any one of claims 1-6, characterized in that, In step S5, the specific process of multi-threaded texture mapping is as follows: Project the coordinates of the vertices of each visible triangle in the sub-block model onto the corresponding best image to obtain the two-dimensional coordinates of the triangle on the best image. Crop the corresponding best image according to the two-dimensional coordinates to obtain the colored two-dimensional triangle. Copy the colored two-dimensional triangle to the texture image. Finally, arrange the texture images of all the two-dimensional triangles to obtain the final texture image and output the texture-mapped three-dimensional model.

8. A rapid acquisition system for 3D models of virtual simulation scenes, characterized in that, The system is based on the method of claim 1, comprising: The image acquisition module is used to acquire multi-view images of the ROI region and the position and pose information of the multi-view images; The visible triangle construction module is used to build the view frustum clipping network and generate the initial set of visible triangles for multi-view images; The visible triangle filtering module is used to filter out invalid triangles in the initial set of visible triangles and select the best image for each visible triangle. The triangular mesh scene model building module is used to generate triangular mesh scene models of ROI regions based on the filtered visible triangular faces; The triangular mesh scene model segmentation module is used to segment the triangular mesh scene model of the ROI region into multiple sub-block models; The texture mapping module is used to perform multi-threaded texture mapping on the sub-block model based on the best image, and output the texture-mapped 3D model.

Citation Information

Patent Citations

  • Scene data acquisition method and device, model training method and device and computer equipment

    CN112150575A

  • Texture block reconstruction method suitable for wide-range inclined three-dimensional model

    CN114742947A