Three-dimensional asset library construction method and system, electronic equipment and medium

Abstract

The invention provides a three-dimensional asset library construction method. The three-dimensional asset library construction method comprises the following steps: S1, acquiring an advanced multi-source data set; s2, acquiring a triangular mesh data set, a plurality of 3D Gaussian images and a plurality of rendering images corresponding to the advanced multi-source data set by adopting an SfM algorithm and a 3D Gaussian sputtering model; s3, acquiring a geometric structure data set, a rendering parameter set and a three-dimensional dense point cloud set; aligning the rendering parameter set with the multispectral information of the remote sensing image to obtain a material information set; s4, performing three-dimensional reconstruction by using the geometric structure data sets of the plurality of rendered images to obtain a plurality of reconstruction scenes; identifying and marking a plurality of monomers and a plurality of components in each reconstruction scene to obtain a label information set of each reconstruction scene; s5, constructing the triangular mesh data set, the geometric structure information set and the three-dimensional dense point cloud set into a space entity asset set; constructing the rendering parameter set into a color texture asset set; and constructing the plurality of reconstruction scenes, the material information set and the label information set into a ground object entity asset set, and forming a three-dimensional asset library.

Description

Technical Field

[0001] The present invention relates to the field of data asset processing technology, and in particular to a three-dimensional asset library construction method, system, electronic equipment and medium. Background Art

[0002] With the development of remote sensing technology, modern technologies such as drones and remote sensing satellites are being used more and more widely. 3D reconstruction technology is an interdisciplinary field between computer vision and computer graphics. It aims to restore the three-dimensional geometric structure and surface properties of a scene using sensor data (such as images and depth information). In recent years, with the development of deep learning, multi-sensor fusion, and high-performance computing, 3D reconstruction technology has moved from the laboratory to practical applications. These technologies can provide multi-source, multi-dimensional output data, such as multispectral remote sensing images, aerial images, and 3D output data. These data can be used to monitor changes in land features and risk assessment. However, due to the diversity and complexity of data sources, how to effectively integrate these 3D output data and generate meaningful decision-making recommendations has become a major challenge in remote sensing geographic informatization. Summary of the Invention

[0003] In response to the above technical problems, the present invention provides a method, system, electronic device and medium for constructing a three-dimensional asset library.

[0004] The first object of the present invention is to provide a method for constructing a three-dimensional asset library, the method comprising the following steps:

[0005] S1 acquires primary multi-source datasets by collecting multiple remote sensing images and aerial images, and performs super-resolution processing on the multi-source datasets to obtain advanced multi-source datasets.

[0006] S2 uses the SfM algorithm and 3D Gaussian sputtering model to obtain a triangular mesh dataset and multiple 3D Gaussian images corresponding to the advanced multi-source dataset; and obtains multiple rendered images based on the multiple 3D Gaussian images;

[0007] S3 obtains the geometric structure data and rendering parameters of each rendered image to obtain a geometric structure data set, a rendering parameter set, and a three-dimensional dense point cloud set; aligns the rendering parameter set corresponding to the advanced multi-source data set with the multispectral information of the remote sensing image in sequence to obtain a material information set;

[0008] S4 uses the geometric structure datasets of multiple rendered images to perform 3D reconstruction to obtain multiple reconstructed scenes; based on the semantic segmentation model, multiple monomers and multiple components in each reconstructed scene are sequentially identified and labeled to obtain a label information set for each reconstructed scene;

[0009] S5 constructs the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; constructs the rendering parameter set into a color texture asset set; constructs the multiple reconstructed scenes, label information sets and material information sets into a ground object entity asset set; constructs the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

[0010] Specifically, the advanced multi-source data set in step S1 includes a multispectral information set, an original camera position set, and an advanced multi-source image set.

[0011] Specifically, the geometric structure data set in step S3 includes multiple high-level center point cloud coordinates and multiple high-level Gaussian ellipsoid covariance matrices; the rendering parameter set includes multiple high-level color values ​​and high-level opacity.

[0012] Specifically, step S2 further includes:

[0013] S21 uses the SfM algorithm to generate sparse point cloud sets, point cloud boundary sets, and camera internal and external parameter sets for advanced multi-source raw data sets;

[0014] S22 obtains a triangular mesh dataset based on the point cloud boundary set and the original camera position set corresponding to the advanced multi-source raw data;

[0015] S23 uses a 3D Gaussian sputtering model based on the camera's internal and external parameter sets to model each sparse point cloud in the sparse point cloud set into a 3D Gaussian image;

[0016] S24 maps and renders each 3D Gaussian image onto a 2D image plane corresponding to the advanced multi-source original dataset to obtain a plurality of rendered images.

[0017] Specifically, step S3 further includes:

[0018] S31 sequentially acquires corresponding geometric structure data and rendering parameters from a plurality of rendered images to obtain a geometric structure data set and a rendering parameter set;

[0019] S32 extracts a three-dimensional dense point cloud set from the geometric structure data corresponding to each rendered image;

[0020] S33 sequentially transforms each advanced multi-source image in the advanced multi-source image set into the coordinate system where the rendered image is located, unifies the pixel coordinates, and obtains a transformed advanced image set;

[0021] S34 constructs a mapping relationship between the rendering parameter set and the multispectral information set based on the transformed advanced multi-source image set, and aligns the rendering parameter set with the multispectral information of the corresponding position in turn to obtain the material information set.

[0022] Specifically, the monomers in step S4 include buildings, bridges, roads, and woodlands; the components include doors, windows, pillars, roofs, steps, lane lines, and zebra crossings; and the label information set includes a monomer-level information set and a component-level information set.

[0023] A second object of the present invention is to provide a three-dimensional asset library construction system, the system comprising:

[0024] The module for acquiring multi-source datasets collects multiple remote sensing images and aerial images to obtain primary multi-source datasets;

[0025] The super-resolution processing module is used to perform super-resolution processing on multi-source data sets to obtain advanced multi-source data sets;

[0026] A Gaussian rendering module is used to obtain a triangular mesh dataset and multiple 3D Gaussian images corresponding to an advanced multi-source dataset based on a SfM algorithm and a 3D Gaussian sputtering model; and to obtain multiple rendered images based on the multiple 3D Gaussian images;

[0027] A material information acquisition module is used to construct the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; construct the rendering parameter set into a color texture asset set; construct the multiple reconstructed scenes, label information sets and material information sets into a ground object entity asset set; and construct the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

[0028] A label information set acquisition module is used to reconstruct multiple reconstructed scenes using the geometric structure datasets of multiple rendered images; multiple monomers and multiple components in each reconstructed scene are sequentially identified based on the semantic segmentation model, and labeling is performed to obtain a label information set for each reconstructed scene;

[0029] A three-dimensional asset library construction module is used to construct the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; construct the rendering parameter set and the material information set into a color texture asset set; construct each reconstructed scene and label information set into a ground object entity asset set; and construct the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

[0030] The third object of the present invention is to provide an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors execute any one of the methods described above.

[0031] A fourth object of the present invention is to provide a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute any one of the methods described above.

[0032] A fifth object of the present invention is to provide a computer program product, characterized in that it includes a computer program, and the computer program implements any one of the above methods when executed by a processor.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention uses super-resolution processing to improve the resolution of remote sensing and aerial flight images, solving the problem of fuzzy original data; uses a 3D Gaussian sputtering model to obtain rendering parameters. This technology is compatible with images of different resolutions / different times, making it easy to achieve physical-level realistic rendering; it can also provide multi-scale geometric expression, taking into account both macroscopic structures and microscopic details; and maps rendering parameters with multi-spectral information to solve the problem of missing physical properties of materials; in addition, a multi-level label information library is constructed to achieve centimeter-level business scenario management. The present invention solves the technical problems of low data utilization efficiency, material information distortion, and missing semantic information in the field of three-dimensional digital twins by constructing an inter-entity asset set, a color and texture asset set, and a ground entity asset set to form a three-dimensional asset library; this three-dimensional asset library is extremely valuable from macroscopic planning to microscopic risk monitoring and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is a flowchart of a method for constructing a three-dimensional asset library in an embodiment of the present invention;

[0037] Figure 2 This is a framework diagram of a three-dimensional asset library construction system according to an embodiment of the present invention;

[0038] Figure 3 Schematic diagram of the structure of a three-dimensional asset library in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. It should be noted that as long as no conflict arises, the various embodiments and features of each embodiment of the present invention can be combined with each other, and the resulting technical solutions are all within the scope of protection of the present invention.

[0040] In an embodiment of the present invention, a method for constructing a three-dimensional asset library is provided. Figure 1 , Figure 1 This is a flowchart of a method for constructing a three-dimensional asset library according to an embodiment of the present invention. The method includes the following steps:

[0041] S1 acquires primary multi-source datasets by collecting multiple remote sensing images and aerial images, and performs super-resolution processing on the multi-source datasets to obtain advanced multi-source datasets.

[0042] In an embodiment of the present invention, the advanced multi-source data set in step S1 includes a multispectral information set, an original camera position set, and an advanced multi-source image set.

[0043] In an embodiment of the present invention, the super-resolution processing is achieved by a trained super-resolution model, which includes a residual hybrid attention network and an edge enhancement network; wherein the residual hybrid attention group includes m cascaded hybrid attention convolution blocks, overlapping cross attention blocks and a second convolution layer with residual connections, where m is a positive integer; wherein the hybrid attention convolution block includes a multi-head self-attention block and a channel attention block arranged in parallel, and a deep convolution layer connected after the multi-head self-attention block and the channel attention block; wherein the edge enhancement network includes a Laplacian operator, strided convolution, a residual dense subnetwork and sub-pixel convolution, wherein the residual dense subnetwork includes multiple residual-in-residual dense blocks, the residual-in-residual dense blocks include multiple residual dense blocks, the residual dense blocks are connected as residual connections, and each residual operation of the residual-in-residual dense blocks is set with a weighting coefficient. The residual hybrid attention network is used to super-reconstruct each primary multi-source image to obtain multiple intermediate multi-source images. The edge enhancement network is used to optimize the edges of each multi-source image to obtain multiple high-level multi-source images, forming a high-level multi-source dataset. The residual hybrid attention network includes a shallow feature extraction module, a deep feature extraction module, and an image reconstruction module. The deep feature extraction module includes n cascaded residual hybrid attention groups and the first convolutional layer.

[0044] S2 uses the SfM algorithm and 3D Gaussian sputtering model to obtain a triangular mesh dataset and multiple 3D Gaussian images corresponding to an advanced multi-source dataset; and obtains multiple rendered images based on the multiple 3D Gaussian images.

[0045] In this embodiment of the present invention, step S2 further includes:

[0046] S21 uses the SfM algorithm to generate sparse point cloud sets, point cloud boundary sets, and camera internal and external parameter sets for advanced multi-source raw data sets;

[0047] S22 obtains a triangular mesh dataset according to the point cloud boundary set and the original camera position set corresponding to the advanced multi-source raw data.

[0048] In an embodiment of the present invention, a Structure from Motion (SfM) algorithm is used to estimate and generate a sparse point cloud set and a point cloud boundary set from a high-level multi-source raw data set. Based on the point cloud boundary set and the sparse point cloud set, DPGrid software is used to perform aerial triangulation on the pre-high-level multi-source raw data set to obtain a camera intrinsic and extrinsic parameter set and a triangular mesh data set.

[0049] S23 uses a 3D Gaussian sputtering model based on the camera's internal and external parameter sets to model each sparse point cloud in the sparse point cloud set into a 3D Gaussian image;

[0050] S24 maps and renders each 3D Gaussian image onto a 2D image plane corresponding to the advanced multi-source original dataset to obtain a plurality of rendered images.

[0051] In an embodiment of the present invention, the 3D Gaussian sputtering model includes a Gaussian point cloud distribution conversion unit, a differentiable Gaussian rasterization unit, and a decoupled appearance modeling unit.

[0052] In an embodiment of the present invention, a Gaussian point cloud distribution conversion unit is used to map each sparse point cloud in the sparse point cloud set to a 3D Gaussian point cloud distribution space to generate multiple sets of 3D Gaussian point cloud distributions; the parameters of each set of 3D Gaussian point cloud distributions include the primary center point coordinates μ, the primary Gaussian ellipsoid covariance matrix ∑, the primary color value Y, and the primary opacity A. A differentiable Gaussian rasterization unit is used to project and render each 3D Gaussian image onto the 2D image plane corresponding to each camera pose, to obtain multiple predicted images and 2D Gaussian point clouds corresponding to different camera poses. The coordinates under the projected camera coordinates are calculated. and the intermediate Gaussian ellipsoid covariance matrix The mid-level opacity, mid-level color value, and mid-level center coordinates of each 2D Gaussian point cloud are also included. Based on the predicted image, a decoupled appearance modeling unit is used to generate multiple primary transformation maps. Each predicted image is then optimized in appearance based on each primary transformation map to obtain multiple rendered images.

[0053] S3 obtains the geometric structure data and rendering parameters of each rendered image to obtain a geometric structure data set, a rendering parameter set, and a three-dimensional dense point cloud set; aligns the rendering parameter set corresponding to the advanced multi-source data set with the multispectral information of the remote sensing image in sequence to obtain a material information set;

[0054] In an embodiment of the present invention, the geometric structure data set in step S3 includes a plurality of high-level center point coordinates and a plurality of high-level Gaussian ellipsoid covariance matrices; the rendering parameter set includes a plurality of high-level color values ​​and a high-level opacity;

[0055] In this embodiment of the present invention, step S3 further includes:

[0056] S31 sequentially acquires corresponding geometric structure data and rendering parameters from a plurality of rendered images to obtain a geometric structure data set and a rendering parameter set.

[0057] Based on the above step S2, when each rendered image is obtained, the optimized high-level Gaussian ellipsoid covariance matrix, high-level color values, high-level opacity parameters and high-level center point coordinates are also obtained, and the geometric structure data set and rendering parameter set are obtained.

[0058] S32 extracts a three-dimensional dense point cloud set from the geometric structure data corresponding to each rendered image.

[0059] S33 sequentially transforms each advanced multi-source image in the advanced multi-source image set into the coordinate system where the rendered image is located, unifies the pixel coordinates, and obtains a transformed advanced image set;

[0060] S34 constructs a mapping relationship between the rendering parameter set and the multispectral information set based on the transformed advanced multi-source image set, and aligns the rendering parameter set with the multispectral information of the corresponding position in turn to obtain the material information set.

[0061] In an embodiment of the present invention, a feature point matching method is used to unify the rendered image and remote sensing image into the same coordinate system, unifying pixel coordinates and obtaining a transformed high-level image set. Furthermore, by establishing an association model between the rendering parameter set and the multispectral band information set, the rendering parameter set is sequentially aligned with the multispectral information at the corresponding position, mapping visual colors to physical spectral properties. The aligned spectral information set is then used to invert material information, such as vegetation inversion using the NDVI index (near-infrared and red bands) and water inversion using the absorption characteristics of the shortwave infrared band.

[0062] S4 uses the geometric structure datasets of multiple rendered images to perform 3D reconstruction to obtain multiple reconstructed scenes; based on the semantic segmentation model, multiple monomers and multiple components in each reconstructed scene are sequentially identified and labeled to obtain a label information set for each reconstructed scene;

[0063] In an embodiment of the present invention, the monomers in step S4 include buildings, bridges, roads, and woodlands; the components include doors, windows, pillars, roofs, steps, lane lines, and zebra crossings; and the label information set includes a monomer-level information set and a component-level information set.

[0064] In an embodiment of the present invention, a semantic segmentation model is constructed based on the DlinkNet network, which includes an encoder module, a central area module and a decoder module. The encoder module adopts a vector quantization variational autoencoder, which includes two encoders for encoding the input image to obtain a feature map, and using the coding space to perform vector quantization of the feature map in the upper and lower layers, and learn to obtain discrete hidden layer feature representations; the central area module and the decoder module adopt the central part and the decoding part in the DlinkNet network, including a hole convolution layer, a fully connected layer and a classification layer with shortcut connections, which are used to perform hole convolution and fusion on the features to obtain classification features with a sufficiently large receptive field and containing multi-dimensional semantic information, that is, monomer features, including monomer feature information such as buildings, bridges, roads, and woodlands. Afterwards, the classification layer (ReLU layer or Sigmoid layer) is used to classify the monomer features based on semantic information to obtain component features, including component feature information such as doors, windows, pillars, roofs, steps, lane lines, and zebra crossings.

[0065] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of a three-dimensional asset library in an embodiment of the present invention. The asset library is constructed as follows:

[0066] S5 constructs the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; constructs the rendering parameter set and the material information set into a color texture asset set; constructs each reconstructed scene and label information set into a ground object entity asset set; and constructs the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

[0067] In the embodiment of the present invention, a three-dimensional asset library construction system is also provided. Figure 2 , Figure 2 This is a framework diagram of a system for building a three-dimensional asset library according to an embodiment of the present invention. The system 100 includes:

[0068] The multi-source dataset acquisition module 101 collects multiple remote sensing images and aerial images to obtain a primary multi-source dataset;

[0069] A super-resolution processing module 102 is used to perform super-resolution processing on the multi-source dataset to obtain an advanced multi-source dataset;

[0070] The Gaussian rendering module 103 is configured to obtain a triangular mesh dataset and a plurality of 3D Gaussian images corresponding to the advanced multi-source dataset based on the SfM algorithm and the 3D Gaussian sputtering model; and obtain a plurality of rendered images based on the plurality of 3D Gaussian images;

[0071] The material information acquisition module 104 is used to obtain the geometric structure data and rendering parameters of each rendered image to obtain a geometric structure data set, a rendering parameter set, and a three-dimensional dense point cloud set; the rendering parameter set corresponding to the advanced multi-source data set is sequentially aligned with the multispectral information of the remote sensing image to obtain a material information set;

[0072] The label information set acquisition module 105 is configured to reconstruct multiple reconstructed scenes using the geometric structure datasets of the multiple rendered images; identify multiple monomers and multiple components in each reconstructed scene in sequence based on the semantic segmentation model, and perform labeling processing to obtain a label information set for each reconstructed scene;

[0073] The module 106 for constructing a three-dimensional asset library is used to construct the triangular mesh dataset, the geometric structure information set, and the three-dimensional dense point cloud set into a spatial entity asset set; construct the rendering parameter set and the material information set into a color texture asset set; construct each reconstructed scene and the label information set into a ground object entity asset set; and construct the spatial entity asset set, the color texture asset set, and the ground object entity asset set into a three-dimensional asset library.

[0074] The third object of the present invention is to provide an electronic device, comprising: one or more processors; a memory for storing one or more programs, wherein, when the one or more programs are executed by the one or more processors, the one or more processors execute any one of the methods described above.

[0075] In an embodiment of the present invention, a computer-readable storage medium is further provided, on which executable instructions are stored. When the instructions are executed by a processor, the processor executes any one of the methods described above.

[0076] A computer program product is also provided in an embodiment of the present invention, characterized in that it includes a computer program, and the computer program implements any one of the above methods when executed by a processor.

[0077] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0078] 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 that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for constructing a three-dimensional asset library, characterized in that: The construction method comprises the following steps: S1 acquires primary multi-source datasets by collecting multiple remote sensing images and aerial images, and performs super-resolution processing on the multi-source datasets to obtain advanced multi-source datasets. S2 uses the SfM algorithm and 3D Gaussian sputtering model to obtain a triangular mesh dataset and multiple 3D Gaussian images corresponding to the advanced multi-source dataset; and obtains multiple rendered images based on the multiple 3D Gaussian images; S3 obtains the geometric structure data and rendering parameters of each rendered image to obtain a geometric structure data set, a rendering parameter set, and a three-dimensional dense point cloud set; aligns the rendering parameter set corresponding to the advanced multi-source data set with the multispectral information of the remote sensing image in sequence to obtain a material information set; S4 uses the geometric structure datasets of multiple rendered images to perform 3D reconstruction to obtain multiple reconstructed scenes; based on the semantic segmentation model, multiple monomers and multiple components in each reconstructed scene are sequentially identified and labeled to obtain a label information set for each reconstructed scene; S5 constructs the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; constructs the rendering parameter set into a color texture asset set; constructs the multiple reconstructed scenes, label information sets and material information sets into a ground object entity asset set; constructs the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

2. The construction method according to claim 1, characterized in that The advanced multi-source dataset in step S1 includes a multispectral information set, an original camera position set, and an advanced multi-source image set.

3. The construction method according to claim 1, characterized in that The geometric structure data set in step S3 includes multiple high-level center point coordinates and multiple high-level Gaussian ellipsoid covariance matrices; the rendering parameter set includes multiple high-level color values ​​and high-level opacity.

4. The construction method according to claim 2, characterized in that Step S2 further comprises: S21 uses the SfM algorithm to generate sparse point cloud sets, point cloud boundary sets, and camera internal and external parameter sets for advanced multi-source raw data sets; S22 obtains a triangular mesh dataset based on the point cloud boundary set and the original camera position set corresponding to the advanced multi-source raw data; S23 uses a 3D Gaussian sputtering model based on the camera's internal and external parameter sets to model each sparse point cloud in the sparse point cloud set into a 3D Gaussian image; S24 maps and renders each 3D Gaussian image onto a 2D image plane corresponding to the advanced multi-source original dataset to obtain a plurality of rendered images.

5. The construction method according to claim 2, characterized in that: Step S3 further comprises: S31 sequentially acquires corresponding geometric structure data and rendering parameters from a plurality of rendered images to obtain a geometric structure data set and a rendering parameter set; S32 extracts a three-dimensional dense point cloud set from the geometric structure data corresponding to each rendered image; S33 sequentially transforms each advanced multi-source image in the advanced multi-source image set into the coordinate system where the rendered image is located, unifies the pixel coordinates, and obtains a transformed advanced image set; S34 constructs a mapping relationship between the rendering parameter set and the multispectral information set based on the transformed advanced multi-source image set, and aligns the rendering parameter set with the multispectral information of the corresponding position in turn to obtain the material information set.

6. The construction method according to claim 1, characterized in that In step S4, the monomers include buildings, bridges, roads, and woodlands; the components include doors, windows, pillars, roofs, steps, lane lines, and zebra crossings; and the label information set includes a monomer-level information set and a component-level information set.

7. A three-dimensional asset library construction system, characterized in that: The system comprises: The module for acquiring multi-source datasets collects multiple remote sensing images and aerial images to obtain primary multi-source datasets; The super-resolution processing module is used to perform super-resolution processing on multi-source data sets to obtain advanced multi-source data sets; A Gaussian rendering module is used to obtain a triangular mesh dataset and multiple 3D Gaussian images corresponding to an advanced multi-source dataset based on a SfM algorithm and a 3D Gaussian sputtering model; and to obtain multiple rendered images based on the multiple 3D Gaussian images; A material information acquisition module is used to obtain the geometric structure data and rendering parameters of each rendered image to obtain a geometric structure data set, a rendering parameter set, and a three-dimensional dense point cloud set; the rendering parameter set corresponding to the advanced multi-source data set is sequentially aligned with the multispectral information of the remote sensing image to obtain a material information set; A label information set acquisition module is used to reconstruct multiple reconstructed scenes using the geometric structure datasets of multiple rendered images; multiple monomers and multiple components in each reconstructed scene are sequentially identified based on the semantic segmentation model, and labeling is performed to obtain a label information set for each reconstructed scene; A three-dimensional asset library construction module is used to construct the triangular mesh dataset, geometric structure information set and three-dimensional dense point cloud set into a spatial entity asset set; construct the rendering parameter set into a color texture asset set; construct the multiple reconstructed scenes, label information sets and material information sets into a ground object entity asset set; and construct the spatial entity asset set, color texture asset set and ground object entity asset set into a three-dimensional asset library.

8. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to perform the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that Executable instructions are stored thereon, which, when executed by a processor, enable the processor to perform the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that A computer program is included which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images