Tree monomer modeling method for urban twinborn scene construction

Abstract

The invention provides a tree monomer modeling method for urban twinborn scene construction, which comprises the following steps: respectively generating a color point cloud and a true orthophoto based on inclined model data, segmenting a ground point cloud and a tree point cloud from the color point cloud, and respectively interpolating to generate terrain raster data and tree raster data; extracting single tree crown vector data from the true orthophoto based on a network model, instantiating tree point clouds, extracting single tree point clouds, performing mesh networking on the single tree point clouds, and generating a single tree reconstruction model; a single tree fine model object is generated through matching in a mode of performing three-dimensional shape similarity measurement on tree single reconstruction and a tree three-dimensional fine model, and model instance zooming and scene position placement are performed in combination with height and crown breadth information of a single tree. And the high-efficiency monomer modeling and rendering requirements of the large-range city twinborn scene tree model are met.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional scene modeling, and more specifically, to a tree monomer modeling method for constructing urban twin scenes. Background Art

[0002] As a crucial component of urban 3D scenes, detailed tree modeling and efficient, large-scale rendering are key technologies for constructing simulation-grade urban twin scenes. Currently, the mainstream approaches for modeling urban trees include detailed individual tree modeling based on terrestrial LiDAR point clouds and random point generation and planting of trees based on vegetation areas in imagery.

[0003] LiDAR devices are typically unable to directly capture the color and texture of tree branches and leaves. Currently, they can only reconstruct the three-dimensional geometry of trees, but are unable to extract the texture of branches and leaves. The segmentation of individual tree point clouds also relies primarily on manual labor. Furthermore, LiDAR point cloud data acquisition is more expensive than oblique photography and remote sensing technologies, making it primarily suitable for detailed modeling of small-scale trees.

[0004] Image-based random planting modeling of vegetation areas typically involves extracting vector surfaces of vegetation coverage areas using DOM (Domain Mapping), then randomly planting individual tree models within the vegetation coverage areas according to a specific grid density. This grid-based random planting modeling method has low accuracy in reconstructing tree positions, heights, and species, making it difficult to meet the requirements for detailed tree restoration in simulated urban scenes. Summary of the Invention

[0005] This paper addresses the technical problems existing in the existing technology and provides a method for individualized tree modeling for urban twin scene construction. This method can address the current problems of low efficiency in individualized urban tree modeling and slow rendering speed of large-scale scenes. First, it implements a method for individualized tree reconstruction based on a tilted model, reducing the time and cost of individualized tree modeling when constructing urban scenes.

[0006] The present invention provides a tree monomer modeling method for constructing urban twin scenes, comprising: Step S1, performing three-dimensional node sampling on the tilt model data of the designated area to generate color point cloud data, and performing vertical sampling to generate a true orthogonal image TDOM; Step S2, performing semantic segmentation on the color point cloud data based on a deep learning model to obtain ground point cloud, grass point cloud, and tree point cloud; Step S3, merging the ground point cloud and the grassland point cloud, and generating DEM terrain raster data based on Kriging interpolation; Step S4, merging the ground point cloud and the tree point cloud, and generating DTSM tree surface raster data expressing the undulations of the ground trees based on Kriging interpolation; Step 5: extracting the circumscribed rectangle vector data of the crown of a single tree from the true orthogonal image TDOM based on a preset network model; Step S6, performing instance segmentation on the tree point cloud in step S2 based on the circumscribed rectangle vector data of the individual tree crown, wherein each point cloud cluster after segmentation is a single tree point cloud; Step S7, performing mesh processing on each individual tree point cloud to obtain a simplified individual tree model structure; Step S8, performing similarity matching between the simplified single-tree model structure and each tree fine model in the tree three-dimensional model asset library, and matching the tree fine model with the highest similarity as the single-tree model expressing the specified area; Step S9, performing spatial overlay analysis on the individual tree crown circumscribed rectangular vector data extracted in step S5, the DEM terrain raster data generated in step S3, and the DTSM tree surface raster data generated in step S4, calculating the bottom elevation, top elevation, and crown width of each individual tree to obtain the outer bounding box of the individual tree; In step S10, the single tree model obtained by matching in step S8 is scaled according to the outer bounding box of the single tree, and translated to the center point of the single tree, thereby completing the single tree modeling and instantiation rendering of the urban twin scene trees.

[0007] The present invention provides a tree monomer modeling method for the construction of urban twin scenes, which realizes urban monomer tree modeling and efficient scene rendering based on tilted model data. This method can fully automatically construct monomer tree models and adopt instance scene rendering to reduce the cost of monomer modeling of urban scene trees and shorten the modeling cycle. At the same time, it can realize efficient rendering of massive tree models in large-scale urban scenes, providing basic support for the construction of city-level twin scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flowchart of a tree monomer modeling method for building urban twin scenes provided by the present invention; Figure 2 This is a schematic diagram of calculating the height and crown width parameters of a single tree in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, the technical features in the various embodiments or single embodiments provided by the present invention can be arbitrarily combined with each other to form a feasible technical solution. This combination is not restricted by the sequence of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0010] In view of the current difficulties in urban twin scene construction, such as large scope, high requirements for model simulation, difficulty in individualizing trees, long modeling cycle and high cost, the present invention proposes a method for automatic individual modeling of trees based on oblique model data (OSGB) and instantiation rendering of highly simulated tree models in urban twin scenes. The method of the present invention first derives multi-source data such as color point cloud, true orthogonal image TDOM, terrain raster DEM, etc. from the oblique model data, and then extracts the individual tree structure in the scene based on the multi-source data. The shape similarity measurement algorithm based on the three-dimensional Fourier shape descriptor is used to match the tree fine models in the three-dimensional model asset library to obtain a highly simulated tree model. Finally, the tree model is planted, placed and instantiated in the scene based on the calculated parameters of the individual tree crown width, bottom elevation and top elevation. The following is a detailed introduction to the implementation steps. The overall process is as follows: Figure 1 shown.

[0011] Figure 1 The flowchart of the tree monomer modeling method for building urban twin scenes provided by the present invention is shown as follows: Figure 1 As shown, the method includes the following steps: In step S1 , the tilt model data of the designated area is subjected to three-dimensional node sampling to generate color point cloud data, and vertical sampling is performed to generate a true orthogonal image TDOM.

[0012] It is understood that the geometry traversal of the oblique model data (OSGB) is performed, the mesh data vertices are read, and grid sampling and texture color extraction are performed to obtain color point cloud data. Vertical projection sampling is performed to generate true orthogonal image data (TDOM) with a resolution of 0.1m.

[0013] Step S2: semantically segment the color point cloud data based on a deep learning model to obtain ground point cloud, grass point cloud, and tree point cloud.

[0014] It is understandable that semantic segmentation is performed on the color point cloud generated in step S1 based on the improved RandLA-Net deep learning model to obtain classified point cloud data such as trees, grass, and ground. RandLA-Net is an efficient and lightweight deep learning model for point cloud semantic segmentation. To further improve the accuracy and robustness of point cloud classification, the present invention uses a grid sampling method to replace RandLA-Net's default random sampling method to resample the color point cloud, but still uses RandLA-Net as the point cloud semantic segmentation framework. The improved RandLA-Net not only ensures the accuracy and efficiency of large-scale point cloud semantic segmentation, but also improves the adaptability and generalization of point clouds with different densities.

[0015] Step S3: Merge the ground point cloud and the grassland point cloud, and generate DEM terrain raster data based on Kriging interpolation.

[0016] It is understood that the ground and grassland point clouds derived from the point cloud classification results in step S2 were subjected to point cloud denoising using the DBSCAN clustering algorithm. Because grassland points are close to actual ground points, the confirmed ground and grassland points were merged, and Kriging interpolation was performed on the discrete X, Y, and Z coordinates to generate a continuous DEM terrain raster. This DEM terrain raster data was used in subsequent steps to calculate and correct the base elevation of individual trees.

[0017] Step S4: merging the ground point cloud and the tree point cloud, and generating DTSM tree surface raster data expressing the undulations of trees on the ground surface based on Kriging interpolation.

[0018] It is understood that the ground point cloud and tree point cloud from step S2 are derived and Kriging interpolation is performed based on the XYZ coordinates of the ground and tree points to generate DTSM raster data that represents the elevation of the trees on the ground surface. The DTSM raster data is used to subsequently calculate and correct the top elevation of individual trees.

[0019] Step 5: extracting the circumscribed rectangle vector data of the crown of a single tree from the true orthogonal image TDOM based on a preset network model.

[0020] It is understandable that a network model is pre-trained, and the circumscribed rectangle vector data of the crown of a single tree is extracted from the true orthogonal image TDOM based on the trained network model.

[0021] The process of pre-training a network model involves collecting image data from different cities and regions to create a training set of labeled individual trees. The labeled training set consists of the circumscribed rectangles of the crowns of individual trees within the image range. Taking into account the requirements for the recognition accuracy of individual trees in the image and the efficiency of large-scale scene extraction, Yolov8 was selected as the deep learning network model for image recognition and monitoring of individual tree targets. The Yolov8 network model inherits the high efficiency characteristics of the Yolo series models and can achieve fast detection speeds while maintaining high accuracy. Thanks to rich data augmentation and regularization strategies, as well as improved label allocation and loss functions, it has strong adaptability to image data under different environmental, lighting, and seasonal conditions, and can stably identify individual tree targets. The use of technologies such as dynamically scaled feature pyramid networks can effectively handle individual tree targets of different scales, and can well detect and identify both tall trees and small shrubs. After completing the production of the monomer tree annotation training set, the Yolov8 network model is used for model training. The model accuracy is evaluated and verified on the validation set and test set respectively. The learning rate and regularization parameters are adjusted according to the test results to optimize the model performance, and the final image monomer tree extraction network model program and weight file are output.

[0022] The Yolov8 network model trained in step S5 is used to automatically extract individual trees from the true orthogonal image TDOM data obtained in step S1, and the circumscribed rectangular vector data of the individual tree crowns is output. Each rectangular box represents an individual tree object in the urban scene.

[0023] Step S6: performing instance segmentation on the tree point cloud in step S2 based on the circumscribed rectangle vector data of the individual tree crowns, and each point cloud cluster after segmentation is a single tree point cloud.

[0024] It is understandable that the tree point cloud obtained in step S2 is segmented according to the circumscribed rectangular range of the individual tree crown obtained in step S5, and each point cloud cluster after segmentation is a single tree point cloud.

[0025] Step S7: Mesh the point cloud of each individual tree to obtain a simplified model structure of the individual tree.

[0026] It is understandable that each single tree point cloud cluster obtained in step S6 is meshed. Since the amount of mesh data directly constructed based on the point cloud cluster is large, in order to improve the efficiency of the subsequent tree model geometric shape similarity calculation, the embodiment of the present invention uses the quadratic error mesh simplification algorithm based on edge contraction to mesh each single tree point cloud cluster. The core idea of the quadratic error metric (QEM) mesh simplification algorithm based on edge contraction is to reduce the number of vertices by iteratively contracting the edges in the mesh while minimizing the geometric error introduced in the simplification process. The algorithm defines the error by the sum of the squares of the distances from the vertex to the adjacent faces, constructs an error matrix, and then adopts a greedy strategy to contract the edge with the smallest error increment each time to ensure that the simplified geometric shape is as close as possible to the original mesh. At the same time, the priority queue is used to dynamically maintain the contraction order of the edges, and the optimal operation is quickly selected to achieve efficient three-dimensional mesh simplification. The quadratic error mesh simplification algorithm based on edge contraction performs mesh simplification on each single tree point cloud cluster to obtain a simplified single model structure of a single tree.

[0027] Step S8: performing similarity matching between the simplified single tree model structure and each tree fine model in the tree three-dimensional model asset library, and matching the tree fine model with the highest similarity as the single tree model expressing the specified area.

[0028] It is understandable that a tree three-dimensional model asset library is manually established in advance, and the model library includes precise models of each tree type of common urban tree species such as trees and shrubs.

[0029] Since the amount of geometric data of the data-refined model is large, the quadratic error mesh simplification algorithm based on edge contraction is also used to simplify the mesh of each tree fine model to obtain a simplified tree fine model structure, and the amount of model geometric data is simplified as much as possible while keeping the basic shape of the tree model unchanged.

[0030] Using a shape similarity measurement algorithm based on 3D shape Fourier descriptors, we measure the shape similarity of the simplified single-body model structure of a single tree and each tree model in the tree 3D model asset library. The tree model with the highest similarity is matched to the single-body tree model representing that area. The algorithm encodes the 3D structure into frequency domain coefficients through a 3D Fourier transform, extracts its spectral features as a representation of the shape features, and then measures the shape similarity by comparing the spectral features of the two 3D shapes. The steps for implementing the shape similarity measurement algorithm based on 3D Fourier shape descriptors are as follows: (1) 3D shape voxelization and normalization processing: First, voxelization is performed to convert the tree 3D model (including the extracted simplified single tree model structure and each tree fine model in the tree 3D asset library) into a binary voxel grid. The voxel value is 1 for the inside of the body and 0 for the outside of the body. If the 3D model surface is S , then the voxel grid V ( x , y , z ) is defined as: ; in,( x , y , z ) represents the original x, y, and z axis coordinates in the 3D model. The binarized voxel grid of the 3D model is then normalized, the center of mass of the 3D body is moved to the origin of the coordinate system, the principal axes are aligned through principal component analysis to eliminate the influence of rotation, and the model is scaled to a unit cube with a side length of 1 to ensure translation, rotation, and scaling invariance.

[0031] (2) Three-dimensional discrete Fourier transform: Perform a 3D Fourier transform on the normalized voxel grid to obtain the frequency domain coefficients: ; in kx,ky,kz is the frequency domain coordinate, N is the voxel grid size.

[0032] (3) Fourier shape descriptor extraction: The amplitude of the spectrum is retained while the phase is ignored, the low-frequency part is truncated to retain the main shape features, and the three-dimensional amplitude spectrum is flattened into a one-dimensional vector V as a three-dimensional shape descriptor, namely the spectrum feature.

[0033] (4) Similarity metric calculation: Calculate the Euclidean distance between two 3D shape descriptor vectors V1 and V2 d : ; in, Simplify the spectrum characteristics of the monomer model structure for the monomer trees, is the spectrum feature of the tree model, i is the i-th element in the spectrum feature, and the Euclidean distance d The closer the value is to 0, the more similar the two 3D shapes are. This shape similarity measurement algorithm, based on 3D Fourier shape descriptors, can match a tree model with the closest geometric shape in the 3D tree model asset library, providing the foundation for the subsequent restoration, placement, and rendering of highly realistic tree models in urban twin scenes.

[0034] Step S9, performing spatial overlay analysis on the individual tree crown circumscribed rectangular vector data extracted in step S5, the DEM terrain raster data generated in step S3, and the DTSM tree surface raster data generated in step S4, calculating the bottom elevation, top elevation, and crown width of each individual tree, and obtaining the outer bounding box of the individual tree.

[0035] It is understandable that the bounding rectangle of the crown of the individual tree obtained in step S5 is spatially overlaid with the DEM terrain grid and DTSM tree surface grid obtained in steps S3 and S4, and the bottom elevation, top elevation and crown width of each individual tree are calculated, and the outer bounding box of the individual tree is obtained, such as Figure 2 The outer bounding box of a single tree represents the spatial position of the single tree model in the 3D scene, providing the prerequisite for the next step of planting and placing the model and 3D space transformation.

[0036] In step S10, the single tree model obtained by matching in step S8 is scaled according to the outer bounding box of the single tree, and translated to the center point of the single tree, thereby completing the single tree modeling and instantiation rendering of the urban twin scene trees.

[0037] It is understood that the tree model obtained in step S9 is scaled according to the geometry of the bounding box of the individual tree obtained in step S10 and translated to the center point of the individual tree in the urban twin scene, ultimately completing the individual modeling and efficient instanced rendering of the trees in the urban twin scene. This approach not only ensures the accurate restoration of the position, size, and shape characteristics of individual trees in the urban scene, but also significantly reduces the pressure of model rendering by using 3D model instanced rendering, creating favorable conditions for the construction of ultra-large-scale urban twin scenes.

[0038] The present invention provides a tree monomer modeling method for constructing urban twin scenes, which has the following beneficial effects: 1. Fully automatic individual tree reconstruction based on tilted model data without manual intervention, greatly reducing the time consumption of individual tree modeling and improving the efficiency of individual tree reconstruction in the entire urban scene.

[0039] 2. No need for lidar point clouds and real photos. Tree reconstruction is done using tilt model data, which is currently easier to obtain and more widely used, and has lower requirements for modeling data input.

[0040] 3. First, a single tree model is built based on the tilted 3D model and its derived point cloud and image data. Then, a shape similarity measurement algorithm based on the 3D Fourier shape descriptor is used to perform shape matching retrieval of simulation-level tree models. The individual tree modeling is completed by planting them in 3D space. Only limited manual modeling of high-precision tree models is required to complete the simulation-level rendering modeling of trees in super-large urban scenes.

[0041] 4. Rendering is performed by instantiating single-building models, which significantly reduces video memory usage and improves rendering efficiency, meeting the requirements for tree simulation-level rendering in large-scale urban scenes.

[0042] 5. Tree individualization is based on the currently more mainstream inclined 3D model and its derived 3D color point cloud, TDOM true orthogonal image and DSM digital surface model and other multi-source data. The data is widely available and the comprehensive use of multi-source data also improves the robustness and generalization of the individual tree modeling algorithm.

[0043] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0044] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0046] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0048] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0049] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A tree monomer modeling method for urban twin scene construction, characterized by: include: Step S1, performing three-dimensional node sampling on the tilt model data of the designated area to generate color point cloud data, and performing vertical sampling to generate a true orthogonal image TDOM; Step S2, performing semantic segmentation on the color point cloud data based on a deep learning model to obtain ground point cloud, grass point cloud, and tree point cloud; Step S3, merging the ground point cloud and the grassland point cloud, and generating DEM terrain raster data based on Kriging interpolation; Step S4, merging the ground point cloud and the tree point cloud, and generating DTSM tree surface raster data expressing the undulations of the ground trees based on Kriging interpolation; Step 5: extracting the circumscribed rectangle vector data of the crown of a single tree from the true orthogonal image TDOM based on a preset network model; Step S6, performing instance segmentation on the tree point cloud in step S2 based on the circumscribed rectangle vector data of the individual tree crown, wherein each point cloud cluster after segmentation is a single tree point cloud; Step S7, performing mesh processing on each individual tree point cloud to obtain a simplified individual tree model structure; Step S8, performing similarity matching between the simplified single-tree model structure and each tree fine model in the tree three-dimensional model asset library, and matching the tree fine model with the highest similarity as the single-tree model expressing the specified area; Step S9, performing spatial overlay analysis on the individual tree crown circumscribed rectangular vector data extracted in step S5, the DEM terrain raster data generated in step S3, and the DTSM tree surface raster data generated in step S4, calculating the bottom elevation, top elevation, and crown width of each individual tree to obtain the outer bounding box of the individual tree; In step S10, the single tree model obtained by matching in step S8 is scaled according to the outer bounding box of the single tree, and translated to the center point of the single tree, thereby completing the single tree modeling and instantiation rendering of the urban twin scene trees.

2. The tree monomer modeling method according to claim 1, characterized in that: The step S3, merging the ground point cloud and the grassland point cloud, and generating DEM terrain raster data based on Kriging interpolation, includes: Denoising the ground point cloud and the grassland point cloud respectively based on the DBSCAN clustering algorithm; The denoised ground point cloud and the grassland point cloud are merged, and Kriging interpolation is performed on the discrete XYZ coordinates to generate continuous DEM terrain raster data.

3. The tree monomer modeling method according to claim 1, characterized in that: The preset network model in step S5 is a Yolov8 network model. The training of the preset network model includes: Collect image data from multiple different cities and regions, and annotate the circumscribed rectangle of each individual tree crown in the image data to form a training set; The Yolov8 network model is trained based on the training set.

4. The tree monomer modeling method according to claim 1, characterized in that: The step S7, performing mesh processing on each individual tree point cloud to obtain a simplified individual tree model structure, includes: The quadratic error mesh simplification algorithm based on edge contraction is used to perform mesh processing on each monomer tree point cloud.

5. The tree monomer modeling method according to claim 1, characterized in that: The step S8, performing similarity matching between the simplified single tree model structure and each tree fine model in the tree 3D model asset library, and matching the tree fine model with the highest similarity as the single tree model expressing the specified area, includes: Mesh simplification is performed on each tree fine model in the tree three-dimensional model asset library using a quadratic error mesh simplification algorithm based on edge contraction to obtain a simplified tree fine model structure; Based on the shape similarity measurement algorithm of the three-dimensional shape Fourier descriptor, shape similarity measurement is performed on the simplified monomer model structure of the monomer tree and the simplified tree fine model structure corresponding to each tree fine model in the tree three-dimensional model asset library, and the tree fine model with the highest similarity is matched as the monomer tree model expressing the specified area.

6. The tree monomer modeling method according to claim 5, characterized in that: The shape similarity measurement algorithm based on the three-dimensional shape Fourier descriptor performs shape similarity measurement on the simplified single-unit model structure of a single tree and the simplified tree fine model structure corresponding to each tree fine model in the tree three-dimensional model asset library, including: Performing three-dimensional Fourier transform encoding on the simplified single-tree model structure and the simplified tree fine model structure corresponding to each tree fine model in the tree three-dimensional model asset library to obtain frequency domain coefficients and extract spectrum features; The shape similarity is measured between the frequency spectrum characteristics of the simplified monomer model structure of the monomer tree and the frequency spectrum characteristics of each tree fine model.

7. The tree monomer modeling method according to claim 6, characterized in that: The three-dimensional Fourier transform encoding is performed on the simplified single-tree model structure and the simplified tree fine model structure corresponding to each tree fine model in the tree three-dimensional model asset library to obtain frequency domain coefficients and extract spectrum features, including: The simplified single-body model structure of the single tree and the simplified tree fine model structure corresponding to each tree fine model in the tree 3D model asset library are converted into binary voxel grids, wherein the voxel value of the voxel grid is 1, indicating that the voxel grid is inside the body, and the voxel value of the voxel grid is 0, indicating that the voxel grid is outside the body; wherein, if the 3D model surface is S , voxel grid V ( x , y , z ) is defined as: ； in,( x , y , z ) are the x, y, and z coordinates of the points in the three-dimensional model; Normalize the voxel grid: move the 3D volume centroid to the coordinate origin, align the principal axes through principal component analysis, and scale to a unit cube with a side length of 1; Perform a 3D Fourier transform on the normalized voxel grid to obtain the frequency domain coefficients: ； in, kx,ky,kz is the frequency domain coordinate, N is the voxel grid size; The amplitude of the spectrum is retained while the phase is ignored, the low-frequency part of the frequency domain coefficients is cut off, and the three-dimensional amplitude spectrum is flattened into a one-dimensional vector V as a three-dimensional shape descriptor, which is used as a spectrum feature.

8. The tree monomer modeling method according to claim 6, characterized in that: The shape similarity measurement of the spectral features of the simplified monomer model structure of the monomer tree and the spectral features of each tree fine model includes: ； in, Simplify the spectrum characteristics of the monomer model structure for the monomer trees, is the spectrum feature of the tree model, i is the i-th element in the spectrum feature, and d is the Euclidean distance.

Images