A modeling method, system and storage medium based on semantic segmentation model

Abstract

The present invention discloses a modeling method, system and storage medium based on a semantic segmentation model, which belongs to the field of three-dimensional modeling technology. The method includes: S1: remote sensing data preprocessing stage; S2: semantic segmentation model training and tuning stage; S3: forest area semantic segmentation recognition stage, using the trained semantic segmentation model to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results, and performing vectorization conversion to obtain the Thiessen polygon area of ​​the effective dense forest area; S4: modeling scene construction stage; S5: scene asset generation and output stage, based on the digital elevation model and the vegetation spatial coordinate point set, a dense forest environment model is generated; based on the CGA rule, the vegetation monomer model is geometrically randomized to obtain a digital twin scene asset and output. The spatial distribution characteristics of trees based on satellite observations are restored more accurately, so that the accuracy of environmental modeling is improved within a low cost range.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a modeling method, system and storage medium based on a semantic segmentation model. Background Art

[0002] Digital twin technology, as the top-level enabling technology for digital new infrastructure, is widely used in scenarios such as digital infrastructure construction and operations, digital industrial production, smart cities, and smart transportation. As part of the scene construction process, vegetation modeling typically uses surveying and mapping technologies such as drone-based aerial photography or laser point cloud to reconstruct a real-world vegetation model. This technology is mature and can guarantee a certain degree of authenticity and accuracy. However, disadvantages such as high computing and storage costs, poor data compatibility, poor close-range rendering, and poor connectivity hinder the further development and application of this environmental modeling technology in digital twin scenarios. Vegetation models constructed using 3D modeling software such as 3DSMAX and MAYA are more intuitive and realistic. However, most 3D modeling software lacks data communication interfaces with geographic information system (GIS) software, including domestic SuperMap platforms and international platforms such as ArcGIS. This makes it difficult to quickly deploy the created real-world models to local or cloud-based 3D GIS scenarios for basic calculations or rendering. At the same time, its modeling under large-scale scene conditions is time-consuming and labor-intensive, and the cost becomes uncontrollable, so it is only suitable for small-scale digital twin scene vegetation modeling. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a modeling method, system and storage medium based on a semantic segmentation model.

[0004] The object of the present invention is achieved through the following technical solutions: In a first aspect, the present invention provides: a modeling method based on a semantic segmentation model, comprising the following steps:

[0005] S1: Remote sensing data preprocessing stage, remote sensing data is acquired, then preprocessed, and finally the preprocessed remote sensing data is exported as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI);

[0006] S2: Semantic segmentation model training and tuning phase: building a semantic segmentation model and performing model training and model tuning to obtain a trained semantic segmentation model;

[0007] S3: In the forest area semantic segmentation and recognition stage, the trained semantic segmentation model is used to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results. The semantic segmentation results are then vectorized to obtain the Thiessen polygon area of ​​the effective dense forest area.

[0008] S4: In the modeling scenario construction phase, the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area are imported into the modeling software to construct the digital elevation model; the statistical numerical method in the geographic information system software is used to generate vegetation distribution points within the Thiessen polygon area of ​​the effective dense forest area and import them into the modeling software; finally, the vegetation distribution points are interpolated to assign three-dimensional height parameters to the vegetation distribution points to obtain the vegetation spatial coordinate point set;

[0009] S5: In the scene asset generation and output stage, a dense forest environment model is generated based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model; based on the CGA rules, the vegetation monomer model is geometrically randomly transformed to obtain the digital twin scene asset and output it.

[0010] Preferably, the remote sensing data preprocessing step S1 further includes the following steps:

[0011] S11: Acquire remote sensing data from open source data sources, multispectral radar data from the Sentinel-2 data source, digital elevation model data from the ALOS platform 12.5m resolution data source, and orthochromatic image data from satellite image data sources;

[0012] S12: Import the remote sensing data into remote sensing image processing software and use seamless mosaicking tools to perform mosaicking operations on the scattered remote sensing data to obtain first remote sensing data; perform radiometric calibration on the first remote sensing data to obtain second remote sensing data; perform atmospheric correction on the second remote sensing data to obtain third remote sensing data; perform orthorectification on the third remote sensing data, and determine the atmospheric model, aerosol model, and ground elevation parameters of the third remote sensing data based on the acquisition time, satellite attitude, orbital position, and data set range of the third remote sensing data to obtain fourth remote sensing data; calculate the enhanced vegetation index (EVI) of the fourth remote sensing data to obtain remote sensing data after data preprocessing; the enhanced vegetation index (EVI) is calculated using the following formula:

[0013]

[0014] Where NIR is the near-infrared band reflectivity, which is the 8th band data of the Sentinel-2 radar; R is the red band reflectivity, which is the 4th band data of the Sentinel-2 radar; B is the blue band reflectivity, which is the 2nd band data of the Sentinel-2 radar; G is the gain factor, which is 2.5; C1 and C2 are the first and second atmospheric correction coefficients, which are 6 and 7.5 respectively; L is the soil adjustment parameter, which is 1;

[0015] S13: Load the pre-processed multispectral radar data and orthophoto panchromatic image data into the geographic information system software, and perform mask mapping on the target data set to be processed; for the pre-processed orthophoto panchromatic image data, perform mask mapping on its outline; for the pre-processed multispectral radar data, refer to the pre-processed orthophoto panchromatic image data and use continuous and closed polygons to perform mask mapping; finally, export the masked data as training data.

[0016] Preferably, the S2: semantic segmentation model training and tuning stage further includes the following steps:

[0017] S21: Build a semantic segmentation model using the PyTorch framework, U-Net architecture, and DenseNet backbone network;

[0018] S22: Perform model training on the semantic segmentation model and adjust the hyperparameters of the semantic segmentation model to reach a first preset quantitative index; test the semantic segmentation model, and tune the model according to the test results until the semantic segmentation model reaches a second preset quantitative index to obtain a trained semantic segmentation model.

[0019] Preferably, the S3: forest area semantic segmentation and recognition stage further includes the following steps:

[0020] S31: Use the trained semantic segmentation model to segment the forest area of ​​the pre-processed multispectral radar data into multiple sub-areas and perform data prediction on each sub-area to obtain local segmentation maps. Then, perform data mosaicking on the local segmentation maps to merge them to obtain a complete segmentation map. The input multispectral radar data is segmented using the parameters in the following formula:

[0021]

[0022] Where, C x is the cutting step size for the input multispectral radar data in the horizontal direction; I x is the number of samples cut out in the horizontal direction of the input multispectral radar data; T x is the number of pixels in the horizontal direction of the input multispectral radar data, in pixels; Sx The number of pixels in the horizontal direction of a single sample obtained by cutting, in pixels; G x is the number of pixels in the overlapping area of ​​two adjacent samples in the horizontal direction, in pixels; C y The cutting step size for the input multispectral radar data in the vertical direction; I y is the number of samples cut out in the vertical direction of the input multispectral radar data; T y is the number of pixels in the vertical direction of the input multispectral radar data, in pixels; S y The number of pixels in the vertical direction of a single sample obtained by cutting, in pixels; G y The number of pixels in the overlapping area of ​​two adjacent samples in the vertical direction, in pixels; “⌈...⌉” means rounding up; C x 、 I x 、 T x 、 S x 、 G x 、 C y 、 I y 、 T y 、 S y 、 G y are all positive integers;

[0023] S32: Load the complete segmentation map, set the judgment condition to clear the background to obtain the first segmentation map; use the raster to vector surface conversion function to convert the effective dense forest area in the raster of the first segmentation map into a vector surface to obtain the first raster data; use the first raster data as a mask and clear the non-forest area of ​​the multispectral radar data after the data preprocessing initially input in S31 to obtain the second raster data; convert the second raster data into a grid point set, and the data field inherited by the grid point set is the value of each pixel of the input raster; input the grid point set into the create Thiessen polygon function to obtain multiple Thiessen polygon areas with the points in the grid point set as the center and the sides adjacent to each other; use the geometric intersection calculation function to calculate the intersection of the first raster data and the Thiessen polygon area to obtain the Thiessen polygon area of ​​the effective dense forest area.

[0024] Preferably, the S4: modeling scenario construction phase further includes the following steps:

[0025] S41: Import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software, use the terrain import mode to input the terrain data of the same range as the multispectral radar data of the target area, and perform seamless mosaicking and clipping operations on the terrain data to obtain the digital elevation model;

[0026] S42: Calculate the vegetation quantity per unit area TD based on any rectangular unit in the Thiessen polygon area of ​​the effective dense forest area:

[0027]

[0028] Where A is the area of ​​the current measurement area; is the average canopy area of ​​vegetation in the current measurement area; EVI is the EVI value of the current rectangular unit in the Thiessen polygon area of ​​the effective dense forest area; “⌈...⌉” is the rounding operation;

[0029] After obtaining the number of vegetation per unit area TD, calculate the vegetation distribution points:

[0030]

[0031] in, is the easting of the current t-th random coordinate point, and satisfies t≤TD; is the northing distance of the current t-th random coordinate point, and satisfies t≤TD; D east is the westernmost east distance of the current rectangular unit; D west is the easternmost east distance of the current rectangular unit; D north is the northernmost northing distance of the current rectangular unit; D south is the southernmost north distance of the current rectangular unit; N r is a random number that obeys Gaussian distribution; n The number of samples generated to satisfy the uniform distribution condition; X i is the i-th sample value that satisfies the uniform distribution;

[0032] S43: Import the vegetation distribution points into the modeling software, use the digital elevation model as the elevation data source, and match the vegetation distribution points with the elevation data source to obtain a vegetation spatial coordinate point set.

[0033] Preferably, the S5: scene asset generation and output stage further includes the following steps:

[0034] S51: Generate a vegetation monomer model and initialize the attributes of the vegetation monomer model to obtain a rule file, select all coordinate points in the vegetation spatial coordinate point set and bind them to the rule file, and then start the model generation instruction to obtain a dense forest environment model;

[0035] S52: Perform geometric random transformation on the vegetation monomer model based on CGA rules, and then export the vegetation monomer model and dense woodland environment model as digital twin scene assets and output them.

[0036] Preferably, the semantic segmentation model is a U-Net architecture.

[0037] Preferably, the modeling software is CityEngine and the geographic information system software is ArcGIS Pro.

[0038] A second aspect of the present invention provides: a modeling system based on a semantic segmentation model, for implementing any of the above-mentioned modeling methods based on a semantic segmentation model, comprising:

[0039] A remote sensing data preprocessing module is used to acquire remote sensing data, then preprocess the remote sensing data, and finally export the preprocessed remote sensing data as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI);

[0040] The semantic segmentation model training and tuning module is used to build a semantic segmentation model and perform model training and model tuning to obtain a trained semantic segmentation model;

[0041] The forest area semantic segmentation and recognition module is used to use the trained semantic segmentation model to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results, and then vectorize the semantic segmentation results to obtain the Thiessen polygon area of ​​the effective dense forest area;

[0042] The modeling scenario construction module is used to import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software to construct the digital elevation model; use the statistical numerical method in the geographic information system software to generate vegetation distribution points under the limitation of the Thiessen polygon area of ​​the effective dense forest area and import them into the modeling software; finally, perform elevation interpolation on the vegetation distribution points to assign three-dimensional height parameters to the vegetation distribution points to obtain a vegetation spatial coordinate point set;

[0043] The scene asset generation and output module is used to generate a dense forest environment model based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model; based on the CGA rules, the vegetation monomer model is geometrically randomized to obtain the digital twin scene asset and output it.

[0044] The third aspect of the present invention provides: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned modeling methods based on the semantic segmentation model is implemented.

[0045] The beneficial effects of the present invention are:

[0046] 1) The training data annotation process based on orthophoto panchromatic imagery data adopts the idea of ​​visualizing masks based on orthophoto panchromatic imagery data and applying them to multispectral radar data for output. This circumvents the problem of multispectral radar data being indiscernible to the human eye and indirectly improves the quality of semantic segmentation training data.

[0047] 2) By utilizing the EVI index, with the support of workflows such as Thiessen polygons and raster-to-point conversion, we can obtain fine-grained vegetation distribution feature vector data. This plays a key role in the subsequent workflow of describing the spatial distribution characteristics of vegetation distribution and performing random vector point distribution of dense woodland vegetation. It accurately restores the spatial distribution characteristics of trees based on satellite observations, making it possible to achieve the highest possible accuracy of environmental modeling under the direction of drone aerial survey technology at a low cost.

[0048] 3) In creating the model using CGA rules, this invention fully considers the random nature of dense woodland vegetation growth. By employing a randomized vegetation point distribution and randomized model scaling, the resulting dense woodland vegetation features simulate, to a certain extent, the random distribution of vegetation in both spatial and morphological dimensions. Furthermore, the use of code-generated modeling rules significantly reduces the modeling workload, saving the time and effort of technical implementers.

[0049] 4) Except for the data labeling work required for the semantic segmentation task, which is a one-time initial human investment, the rest of the processes can be converted into code programs with low or no human intervention for automatic execution, so that the environmental modeling of dense woodland areas can truly automatically produce dense woodland environment scenes that conform to the laws of remote sensing observations and output scene assets that meet the relevant digital twin standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Flowchart of the modeling method based on the semantic segmentation model. DETAILED DESCRIPTION

[0051] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0052] See Figure 1 The first aspect of the present invention provides: a modeling method based on a semantic segmentation model, comprising the following steps:

[0053] S1: Remote sensing data preprocessing stage, remote sensing data is acquired, then preprocessed, and finally the preprocessed remote sensing data is exported as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI);

[0054] S2: Semantic segmentation model training and tuning phase: building a semantic segmentation model and performing model training and model tuning to obtain a trained semantic segmentation model;

[0055] S3: In the forest area semantic segmentation and recognition stage, the trained semantic segmentation model is used to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results. The semantic segmentation results are then vectorized to obtain the Thiessen polygon area of ​​the effective dense forest area.

[0056] S4: In the modeling scenario construction phase, the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area are imported into the modeling software to construct the digital elevation model; the statistical numerical method in the geographic information system software is used to generate vegetation distribution points within the Thiessen polygon area of ​​the effective dense forest area and import them into the modeling software; finally, the vegetation distribution points are interpolated to assign three-dimensional height parameters to the vegetation distribution points to obtain the vegetation spatial coordinate point set;

[0057] S5: In the scene asset generation and output stage, a dense forest environment model is generated based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model; based on the CGA rules, the vegetation monomer model is geometrically randomly transformed to obtain the digital twin scene asset and output it.

[0058] In this embodiment, S1 first acquires open-source high-resolution remote sensing data, including multispectral radar, orthochromatic imagery, and digital elevation models (DEMs). Data preprocessing then includes data mosaicking, radiometric calibration, atmospheric correction, orthorectification, and quantitative calculation of the vegetation index (EVI). Finally, training data is derived, including target detection mask calibration, semantic segmentation mask calibration, and training data export. In S2, a segmentation model is developed using a U-Net architecture within a deep neural network to classify pixels in dense woodland areas when the NDVI (Natural Diversity Index) is greater than 0. This model is suitable for recognizing red and infrared data from multispectral radar, primarily for converting remote sensing imagery into vegetation density-related parameters. This step primarily involves model code development, model training, and testing. The effective Thiessen polygons of dense woodland areas obtained through vectorization in S3 serve as the basis for modeling point locations in dense woodland scenarios. S4 involves the creation of a real-world digital terrain and the creation of basic spatial coordinates for input into the modeling software. Two-dimensional raster data layers cannot accurately reflect the topographical undulations of the data, nor do they contain the coordinate point information required for vegetation model creation. Therefore, the modeling software system CityEngine (presented in this invention) must first be used to import ALOS open-source terrain data and Thiessen polygons of effective dense woodland areas. A digital elevation model (DEM) is then constructed based on the terrain data. Then, within ArcGIS, consistent statistical numerical methods are used, and within the "Thiessen polygons of effective dense woodland areas," vegetation distribution points are generated and imported into CityEngine. Finally, elevation interpolation is performed on these vegetation distribution points based on the DEM model to assign three-dimensional height parameters. Therefore, this process primarily involves DEM modeling, vegetation distribution point generation, and elevation interpolation. In step S5, based on the elevation modeling and interpolation results, CityEngine's CGA code is written to generate a dense woodland environment model that conforms to certain rules. The vegetation unit model within the dense woodland model uses CityEngine's built-in obj format model resources, but the implementer can also create their own model as needed. Based on CGA rules, geometric random transformations are applied to adjust individual vegetation models within acceptable limits, avoiding visual repetition and enhancing the naturalness of the visual experience. The resulting model scene is then output as a digital twin scene asset that complies with Omniverse-USD and Unreal-DataSmith data standards, resulting in a vegetation environment with a density distribution close to that of real life. This step includes writing and calling CGA code and saving the scene assets.

[0059] In some embodiments, the remote sensing data preprocessing stage S1 further includes the following steps:

[0060] S11: Acquire remote sensing data from open source data sources, multispectral radar data from the Sentinel-2 data source, digital elevation model data from the ALOS platform 12.5m resolution data source, and orthochromatic image data from satellite image data sources;

[0061] S12: Import the remote sensing data into remote sensing image processing software and use seamless mosaicking tools to perform mosaicking operations on the scattered remote sensing data to obtain first remote sensing data; perform radiometric calibration on the first remote sensing data to obtain second remote sensing data; perform atmospheric correction on the second remote sensing data to obtain third remote sensing data; perform orthorectification on the third remote sensing data, and determine the atmospheric model, aerosol model, and ground elevation parameters of the third remote sensing data based on the acquisition time, satellite attitude, orbital position, and data set range of the third remote sensing data to obtain fourth remote sensing data; calculate the enhanced vegetation index (EVI) of the fourth remote sensing data to obtain remote sensing data after data preprocessing; the enhanced vegetation index (EVI) is calculated using the following formula:

[0062]

[0063] Where NIR is the near-infrared band reflectivity, which is the 8th band data of the Sentinel-2 radar; R is the red band reflectivity, which is the 4th band data of the Sentinel-2 radar; B is the blue band reflectivity, which is the 2nd band data of the Sentinel-2 radar; G is the gain factor, which is 2.5; C1 and C2 are the first and second atmospheric correction coefficients, which are 6 and 7.5 respectively; L is the soil adjustment parameter, which is 1;

[0064] S13: Load the pre-processed multispectral radar data and orthophoto panchromatic image data into the geographic information system software, and perform mask mapping on the target data set to be processed; for the pre-processed orthophoto panchromatic image data, perform mask mapping on its outline; for the pre-processed multispectral radar data, refer to the pre-processed orthophoto panchromatic image data and use continuous and closed polygons to perform mask mapping; finally, export the masked data as training data.

[0065] In this embodiment, S11 first downloads open-source commercial remote sensing data from open-source data sources such as the Geospatial Data Cloud or the European Space Agency's Copernicus data platform. Note that the present invention requires a large amount of high-resolution data for model training and tuning. Therefore, in addition to elevation data, the Sentinel-2 data source is selected as the primary multispectral radar data source, the ALOS platform's 12.5m resolution data source is selected as the primary digital elevation data source, and mainstream satellite imagery products such as Google or Bing are selected as the primary orthophotod full-color satellite imagery data source. These raster data are generally stored in the GeoTIFF format with the suffix ".tif." In S12, the source data is imported using ENVI software, and the seamless mosaicking tool is used to mosaic the previously separate data files. Based on remote sensing principles, the source data for each band is raw brightness or DN (Digital Number) values, so radiometric calibration is required to convert them into quantifiable physical quantities. First, use radiometric calibration, using the calibration type corresponding to the dataset, to calibrate the data. Tools are all automated for specific satellite types, and no parameter tuning is required. Next, use the Atmospheric Correction tool to reprocess the radiometric calibration data to eliminate the effects of atmospheric factors and other factors on the data. Select the appropriate parameters in the tool tab. For example, when using Sentinel-2, the FLAASH algorithm is recommended for correction. Parameters such as the atmospheric model, aerosol model, and ground elevation are determined based on the dataset's acquisition time, satellite attitude, orbital position, and dataset range. Finally, the Enhanced Vegetation Index (EVI) is calculated to assess vegetation density per unit area when vegetation density is high. While S13 typically utilizes general data label masking tools for deep learning machine vision applications, such as LabelAnything and LabelMe, this method utilizes ArcGIS Pro's native label annotation tools for mask annotation to ensure compatibility with image data containing geographic location information. First, load the ENVI-preprocessed multispectral radar data and orthophoto panchromatic image data into ArcGIS Pro, ensuring the target dataset is selected. Second, use the "Label Objects for Deep Learning" tool in the Classification toolbar to mask the dataset. Multispectral radar data is primarily used in dense vegetation conditions, such as plantations and natural forests, where dense vegetation cover and continuous tree canopy distribution are indistinguishable to the naked eye. Therefore, it is necessary to mask the dense woodland type area data using continuous, closed polygons to delineate the boundaries of the dense woodland.Because multispectral radar data is expressed numerically and cannot generally be directly interpreted by humans, the mask is drawn based on the orthophoto panchromatic imagery. Based on the fact that the boundaries of prominent dense forest features in the orthophoto panchromatic imagery are discernible to the human eye, their outlines are masked and applied to the multispectral radar data for export. This export method is still through the "Export Dataset for Deep Learning" tool, ensuring that the "Metadata Format" for this data is set to "Panoramic Segmentation." Ultimately, this export step forms two branches: one for single, non-dense vegetation using the "RCNN" metadata format, and the other for dense, continuous vegetation using the "Panoramic Segmentation" metadata format.

[0066] In some embodiments, the S2: semantic segmentation model training and tuning stage further includes the following steps:

[0067] S21: Build a semantic segmentation model using the PyTorch framework, U-Net architecture, and DenseNet backbone network;

[0068] S22: Perform model training on the semantic segmentation model and adjust the hyperparameters of the semantic segmentation model to reach a first preset quantitative index; test the semantic segmentation model, and tune the model according to the test results until the semantic segmentation model reaches a second preset quantitative index to obtain a trained semantic segmentation model.

[0069] In this embodiment, the semantic segmentation model in S21 utilizes the PyTorch framework and the DenseNet backbone network as the core algorithm components for image processing. The main algorithm employs the U-Net architecture, employing an encoding-decoding approach that uses convolution and pooling for downsampling and transposed convolution and channel splicing for upsampling to perform pixel-level segmentation of remote sensing multispectral data. S22 first involves the semantic segmentation model training and tuning process, which requires tuning hyperparameters to ensure that performance meets the corresponding quantitative indicators. Next, the semantic segmentation model is tested, combining the trained semantic segmentation model results with quantitative indicators. The semantic segmentation model training and tuning process primarily focuses on whether loss iterations exhibit vanishing and exploding gradients. Dynamic tuning is performed on non-descriptive model structure hyperparameters, such as the learning rate, batch size, and number of iterations. The model is refined until its final performance meets the required quantitative indicators, such as a Dice coefficient greater than 0.75 for the semantic segmentation model. The selection of one or more quantitative indicators is at the discretion of the technician and is not limited by this invention. The final trained data is stored in ONNX, a universal deep learning model data format, to ensure its portability and versatility. The semantic segmentation model result test is a key step in using the reserved validation set data to verify the performance of the model. Generally, 10% of the random data pre-selected and separated from the original training set is used for the final test, and the final test index results of the semantic segmentation model of the present invention meet the same trend as the relevant performance index obtained after the model training, such as the Dice coefficient of the semantic segmentation model is greater than 0.75, otherwise it means that the overall performance of the model is poor and the model needs to be re-tuned.

[0070] In some embodiments, the S3: forest area semantic segmentation and recognition stage further includes the following steps:

[0071] S31: Use the trained semantic segmentation model to segment the forest area of ​​the pre-processed multispectral radar data into multiple sub-areas and perform data prediction on each sub-area to obtain local segmentation maps. Then, perform data mosaicking on the local segmentation maps to merge them to obtain a complete segmentation map. The input multispectral radar data is segmented using the parameters in the following formula:

[0072]

[0073] Where, C x is the cutting step size for the input multispectral radar data in the horizontal direction; I x is the number of samples cut out in the horizontal direction of the input multispectral radar data; T x is the number of pixels in the horizontal direction of the input multispectral radar data, in pixels;S x The number of pixels in the horizontal direction of a single sample obtained by cutting, in pixels; G x is the number of pixels in the overlapping area of ​​two adjacent samples in the horizontal direction, in pixels; C y The cutting step size for the input multispectral radar data in the vertical direction; I y is the number of samples cut out in the vertical direction of the input multispectral radar data; T y is the number of pixels in the vertical direction of the input multispectral radar data, in pixels; S y The number of pixels in the vertical direction of a single sample obtained by cutting, in pixels; G y The number of pixels in the overlapping area of ​​two adjacent samples in the vertical direction, in pixels; “⌈...⌉” means rounding up; C x 、 I x 、 T x 、 S x 、 G x 、 C y 、 I y 、 T y 、 S y 、 G y are all positive integers;

[0074] S32: Load the complete segmentation map, set the judgment condition to clear the background to obtain the first segmentation map; use the raster to vector surface conversion function to convert the effective dense forest area in the raster of the first segmentation map into a vector surface to obtain the first raster data; use the first raster data as a mask and clear the non-forest area of ​​the multispectral radar data after the data preprocessing initially input in S31 to obtain the second raster data; convert the second raster data into a grid point set, and the data field inherited by the grid point set is the value of each pixel of the input raster; input the grid point set into the create Thiessen polygon function to obtain multiple Thiessen polygon areas with the points in the grid point set as the center and the sides adjacent to each other; use the geometric intersection calculation function to calculate the intersection of the first raster data and the Thiessen polygon area to obtain the Thiessen polygon area of ​​the effective dense forest area.

[0075] In this embodiment, the semantic segmentation model in S31 is loaded using the "torch.load" method in the PyTorch library, and the affine transformation matrix and projection information of the data are respectively obtained and stored based on the "GetGeoTransform" and "GetProjection" methods of the GDAL library. Note that in general, multispectral remote sensing data for this type of task are large-scale data (measured in square kilometers). Therefore, feeding a single large-scale data set into the model for prediction at one time will increase the computer hardware load. Therefore, the present invention reasonably cuts the data and predicts each one in batches before merging the prediction results to avoid such problems. S x 、 S y 、 G x 、 G y The settings of the following four parameters shall comply with the following standards: S x = S y =1024, G x = G y ; and meet G x Can be T x divisible, and G y Can be T yInteger divisibility. Specifically, when the aforementioned divisibility conditions are not met, the final horizontal or vertical step size of the sample needs to be reduced to ensure proper data output. For example, in the Python-based GDAL environment, an if-else statement can be added to the cut output logic to calculate the number of pixels remaining from the last sample to the input data boundary, ultimately determining the final cut size. After the data is segmented, predictions and results are output one by one, forming a separate segmentation map stored in the same format as the input data. After the entire data segmentation and prediction process is complete, the "gdal.Warp" function in GDAL is used to mosaic the results (a raster mosaic operation, merging sub-rasters based on spatial location) to produce a complete segmentation map. The value of each pixel in the map reflects the corresponding area type, predicting dense woodland or other. S32 uses ArcPy, the Python-based ArcGIS Pro geographic data processing library, to vectorize the segmentation results. The complete segmentation map of the predicted results is used as the source data and converted into vector data to define the vegetation modeling range. First, the complete segmentation map was loaded using the ArcPy raster data loading function "arcpy.Raster." The "SetNull" function in the ArcPy library was used to set a condition to set the background value to null, that is, the pixels displayed as "NoData" in the raster data. Based on the resulting segmentation map with the nulled background, the raster-to-vector function "RasterToPolygon" was used to convert the valid dense forest areas in the raster into vector polygons. The third parameter of the function was set to "false" to control the edge simplification of the output vectors, and the remaining parameters were adjusted according to actual needs. The nulled raster was used as a mask, and the "SetNull" function was used again to null the non-forested areas of the original multispectral radar data at the prediction input, resulting in a raster with the non-forested areas nulled. Second, the "RasterToPoint" function was used to convert the nulled raster into a grid point set, which inherited the data field of each pixel in the input raster, namely the EVI value. Subsequently, the "CreateThiessenPolygons" function is used to take the grid point set created above as input. In order to ensure that the output data completely retains the EVI value carried by the point set, the third parameter of the function needs to be set to "ALL". Finally, a number of rectangular areas centered on the point set and adjacent to each other are obtained, and their number is the same as the number of input points.Finally, use the geometric intersection calculation function "Intersect" to calculate the intersection of the aforementioned effective dense forest area vector surface and the Thiessen polygon vector set to obtain the Thiessen polygon area of ​​the effective dense forest area, which is expressed as any effective dense forest area is completely covered by the rectangular surface adjacent to each side. Each rectangle corresponds to a valid raster pixel, and the value inherited by its vector attribute is the same as the EVI value of the raster pixel at the same spatial position. The final processing results need to be stored in a file geodatabase (File GeoDatabase). Use ArcPy's "arcpy.CreateFileGDB_management" to create an empty database, and use the "ExportFeatures" function to output the processing results. The second directory parameter of the output function needs to be set to the storage path of the file geodatabase created above.

[0076] In some embodiments, the S4: modeling scenario construction phase further includes the following steps:

[0077] S41: Import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software, use the terrain import mode to input the terrain data of the same range as the multispectral radar data of the target area, and perform seamless mosaicking and clipping operations on the terrain data to obtain the digital elevation model;

[0078] S42: Calculate the vegetation quantity per unit area TD based on any rectangular unit in the Thiessen polygon area of ​​the effective dense forest area:

[0079]

[0080] Where A is the area of ​​the current measurement area; is the average canopy area of ​​vegetation in the current measurement area; EVI is the EVI value of the current rectangular unit in the Thiessen polygon area of ​​the effective dense forest area; “⌈...⌉” is the rounding operation;

[0081] After obtaining the number of vegetation per unit area TD, calculate the vegetation distribution points:

[0082]

[0083] in, is the easting of the current t-th random coordinate point, and satisfies t≤TD; is the northing distance of the current t-th random coordinate point, and satisfies t≤TD; D east is the westernmost east distance of the current rectangular unit; D west is the easternmost east distance of the current rectangular unit;D north is the northernmost northing distance of the current rectangular unit; D south is the southernmost north distance of the current rectangular unit; N r is a random number that obeys Gaussian distribution; n The number of samples generated to satisfy the uniform distribution condition; X i is the i-th sample value that satisfies the uniform distribution;

[0084] S43: Import the vegetation distribution points into the modeling software, use the digital elevation model as the elevation data source, and match the vegetation distribution points with the elevation data source to obtain a vegetation spatial coordinate point set.

[0085] In this embodiment, S41 imports ALOS open source terrain data and Thiessen polygon areas of effective dense woodland areas based on the modeling software system CityEngine. Using the "Terrain" terrain import mode, input terrain data of the same range as the multispectral radar data of the target area. ArcGIS Pro, QGIS and other geographic information processing tool software or Python's GDAL library can be used to seamlessly mosaic and crop the basic terrain raster data. The final input is a DEM model (digital elevation model) that includes accurate terrain undulations. The biggest feature of CityEngine in S42 is that it can use CGA rule code to generate a three-dimensional model. Therefore, it is necessary to use a program to estimate the unit space coverage of vegetation that conforms to the remote sensing observation law within the target range, and then use Gaussian random numbers to generate the coordinates of spatial points within the calculation range. It is worth noting that the layout of random points is based on the determination of the TD value. The generation of random points requires generating 5,000 to 10,000 random numbers that obey a uniform distribution based on Python's built-in random number generation function "random.uniform", and then calculating the random numbers that obey a Gaussian distribution. N r , based on which random coordinates are calculated. Then, based on ArcPy, the easting and northing value pairs of the generated random coordinates are stored as a data table, and the aforementioned data table is converted into coordinate points using "XYTableToPoint". Finally, "FeatureClassToShapefile" is used to convert it into geographic vector data in shapefile format with the suffix ".shp", thus obtaining vegetation distribution points. In S43, the vegetation distribution points are imported into CityEngine, and the digital elevation model is used as the elevation data source. The "Align shapes to terrain" command is used to match the vegetation distribution point data with the elevation data source to obtain a set of spatial coordinate points that are completely consistent with the actual elevation, i.e., the vegetation spatial coordinate point set.

[0086] In some embodiments, the S5: scene asset generation and output stage further includes the following steps:

[0087] S51: Generate a vegetation monomer model and initialize the attributes of the vegetation monomer model to obtain a rule file, select all coordinate points in the vegetation spatial coordinate point set and bind them to the rule file, and then start the model generation instruction to obtain a dense forest environment model;

[0088] S52: Perform geometric random transformation on the vegetation monomer model based on CGA rules, and then export the vegetation monomer model and dense woodland environment model as digital twin scene assets and output them.

[0089] In this embodiment, the core purpose of S51 is to write CGA code for the creation of a dense woodland environment model, which mainly includes two sections: attribute parameter predefinition and driving rule coding; after the coding is completed, the coding rules are called to generate the model to obtain the result. The attribute parameters are predefined in the form of "attr" keyword + attribute name to initialize the attributes of a single vegetation model. They include: the horizontal and vertical scaling ratios (zoom_x), horizontal and vertical scaling ratios (zoom_y), vertical scaling ratios (zoom_z), rotation angle around the vertical axis (rotate_y), and file path array (model_path). The file path array (such as "array name ['string 1', 'string 2', 'string 3', ...]") needs to initialize all model paths involved in the form of string arrays. The other attribute parameters mentioned above can be initialized to 0. The driving rule encoding adopts the form of main rule + sub-rule. A line "@StartRule" needs to be inserted before the main rule as an entry tag. Its declaration is in the form of "main rule name -->". It only acts on vector elements such as points, lines, and surfaces that are assigned the corresponding initialization rules, allowing the program to perform arbitrary operations based on geometric elements. The main rule is followed by the model instantiation function "i(model_path)", the instance scaling function "s(zoom_x, zoom_z, The instance rotation function "r(rotate_x, rotate_z, rotate_y)" is used. The value of the "model_path" parameter in the instantiation function is "array name[floor(rand(max))]", where floor(...) is the floor function, rand(...) is the uniform random number generator, and max is the total number of tree model types. The instance zoom factor also uses the rand(...) function for restricted random number generation, and its value should refer to the range [0.5, 2.5]. Instance rotation only considers the transformation of the parameter "rotate_y", that is, random rotation around the vertical axis y, and the angle value is generated by the rand(...) function. Based on the above encoding process, a rule file in the ".cga" format is obtained. All the vegetation spatial coordinate point sets are selected. The rule file (Rule file) of the vector point coordinates selected in the "Inspector" window is bound to the rule file in the ".cga" format. The initial rule (StartRule) is set as the main rule in the rule file in the ".cga" format, and the model generation instruction (Generate models of selected shapes) is started to obtain the final scene result, that is, the dense woodland environment model.In S52, the vegetation monomer model and the dense woodland environment model are exported as scene assets, mainly using the Python script (scripting library) export mode of CityEngine. By writing the code "ce.getObjectsFrom(ce.scene)", the output model is defined as all models of the current scene (that is, the model retrieval domain is "ce.scene"), and any of the export parameter modes "USDExportModelSettings()" and "UnrealExportModelSettings()" are used to limit the output parameters. The former corresponds to the USD data exchange standard of Ominiverse, and the latter corresponds to the DataSmith data exchange standard of Unreal. Finally, ce.export({objects},{settings}) is used to output the target model. The {objects} parameter corresponds to the selected model and DEM layer, and {setteing} corresponds to the data output settings.

[0090] In some embodiments, the semantic segmentation model is a U-Net architecture.

[0091] In some embodiments, the modeling software is CityEngine and the geographic information system software is ArcGIS Pro.

[0092] A second aspect of the present invention provides: a modeling system based on a semantic segmentation model, for implementing any of the above-mentioned modeling methods based on a semantic segmentation model, comprising:

[0093] A remote sensing data preprocessing module is used to acquire remote sensing data, then preprocess the remote sensing data, and finally export the preprocessed remote sensing data as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI);

[0094] The semantic segmentation model training and tuning module is used to build a semantic segmentation model and perform model training and model tuning to obtain a trained semantic segmentation model;

[0095] The forest area semantic segmentation and recognition module is used to use the trained semantic segmentation model to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results, and then vectorize the semantic segmentation results to obtain the Thiessen polygon area of ​​the effective dense forest area;

[0096] The modeling scenario construction module is used to import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software to construct the digital elevation model; use the statistical numerical method in the geographic information system software to generate vegetation distribution points under the limitation of the Thiessen polygon area of ​​the effective dense forest area and import them into the modeling software; finally, perform elevation interpolation on the vegetation distribution points to assign three-dimensional height parameters to the vegetation distribution points to obtain a vegetation spatial coordinate point set;

[0097] The scene asset generation and output module is used to generate a dense forest environment model based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model; based on the CGA rules, the vegetation monomer model is geometrically randomized to obtain the digital twin scene asset and output it.

[0098] The third aspect of the present invention provides: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned modeling methods based on the semantic segmentation model is implemented.

[0099] The present invention combines mainstream open source multispectral radar remote sensing, orthophoto panchromatic image data, and digital elevation model, with the image segmentation of dense woodland area as the first-level purpose, and initializes the data within the dense woodland area to obtain the EVI index as the measurement value for the subsequent judgment of vegetation density in the unit dense woodland area of ​​this scheme, bypassing the problem of high initial investment in the mainstream oblique photography and laser point cloud for three-dimensional dense woodland scene modeling under conventional circumstances; at the same time, ENVI and ArcGIS Pro software are combined for the preliminary preparation of image segmentation task data, so that the image segmentation results within the dense woodland area can directly carry correct geographic spatial information; and this scheme respectively utilizes the characteristics of multispectral radar remote sensing and orthophoto panchromatic images, that is, the characteristics that orthophoto panchromatic images can be clearly identified by the human eye, and accordingly depicts the mask layer for multispectral radar remote sensing image segmentation, and acts on the multispectral radar data for output to obtain training data, so that the quality of model training data for the dense woodland range segmentation task is improved.

[0100] The division of dense woodland areas mainly adopts the U-Net architecture to train the semantic segmentation model based on the ResNet-50 backbone network, and uses the Dice coefficient to evaluate the performance of the model. The model file with more transferability is output in the ONNX format, ensuring that the model has better performance in the semantic segmentation task of dense woodland areas based on the remote sensing image EVI index.

[0101] The semantic segmentation model combines GDAL and Pytorch libraries to perform miniaturized segmentation operations on the input data, reducing the pressure of data calculations. The two are used to implement complete data segmentation, semantic segmentation, vectorized output and other processes, making the semantic segmentation task of dense woodland areas based on multispectral remote sensing data of the EVI index more standardized and more practical. By using the conversion of raster points and Thiessen polygons, the raster is completely transformed into a vector grid surface based on a single pixel, which provides data support for the subsequent generation of refined coordinates based on EVI.

[0102] Based on the vector grid surface carrying the EVI index, the vegetation modeling base points were generated based on the Gaussian distribution, so that the distribution of the vegetation model is consistent with the results of satellite observations to a certain extent. It is not a mechanical and uniform distribution form, but a random and natural distribution form, which makes the randomness of the modeling results close to the real natural distribution form.

[0103] The model generation of dense woodland vegetation is based on CGA rules. The model creation process is described using a highly descriptive programming language. Based on the alignment of initial rules and three-dimensional point feature rules, vegetation models are generated in large quantities quickly and randomly. The randomness of the individual vegetation models is based on the restrictive random number scaling transformation of the model that controls the insertion of the i function after the main rules. The dense woodland scene assets output by the present invention are based on Python scripts, and the resulting data exchange format complies with the USD data standard of Omniverse or the DataSmith data standard of Unreal, and can be widely and quickly seamlessly connected with mainstream digital twin platforms.

[0104] The foregoing description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the concept described herein through the above teachings or techniques or knowledge in the relevant field. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be protected by the appended claims.

Claims

1. A modeling method based on a semantic segmentation model, characterized by: The following steps are involved: S1: Remote sensing data preprocessing stage, remote sensing data is acquired, then preprocessed, and finally the preprocessed remote sensing data is exported as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI); S2: Semantic segmentation model training and tuning phase: building a semantic segmentation model and performing model training and model tuning to obtain a trained semantic segmentation model; S3: In the forest area semantic segmentation and recognition stage, the trained semantic segmentation model is used to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results. The semantic segmentation results are then vectorized to obtain the Thiessen polygon area of ​​the effective dense forest area. S4: Modeling scenario construction phase, digital elevation model data and Thiessen polygon areas of effective dense forest areas are imported into the modeling software to construct a digital elevation model; Using statistical numerical methods in geographic information system software, vegetation distribution points are generated within the limits of the Thiessen polygon area of ​​the effective dense forest area and imported into the modeling software; finally, elevation interpolation is performed on the vegetation distribution points to assign three-dimensional height parameters to the vegetation distribution points to obtain a vegetation spatial coordinate point set; S5: Scene asset generation and output stage: A dense forest environment model is generated based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model. Based on the CGA rules, the vegetation monomer model is subjected to geometric random transformation to obtain the digital twin scene asset and output it; The S2: semantic segmentation model training and tuning phase also includes the following steps: S21: Build a semantic segmentation model using the PyTorch framework, U-Net architecture, and DenseNet backbone network; S22: training the semantic segmentation model and adjusting the hyperparameters of the semantic segmentation model to reach a first preset quantitative index; testing the semantic segmentation model, and optimizing the model according to the test results until the semantic segmentation model reaches a second preset quantitative index to obtain a trained semantic segmentation model; The S3: forest area semantic segmentation and recognition stage further includes the following steps: S31: Use the trained semantic segmentation model to segment the forest area of ​​the pre-processed multispectral radar data into multiple sub-areas and perform data prediction on each sub-area to obtain local segmentation maps. Then, perform data mosaicking on the local segmentation maps to merge them to obtain a complete segmentation map. The input multispectral radar data is segmented using the parameters in the following formula: Where, C x is the cutting step size for the input multispectral radar data in the horizontal direction; I x is the number of samples cut out in the horizontal direction of the input multispectral radar data; T x is the number of pixels in the horizontal direction of the input multispectral radar data, in pixels; S x The number of pixels in the horizontal direction of a single sample obtained by cutting, in pixels; G x is the number of pixels in the overlapping area of ​​two adjacent samples in the horizontal direction, in pixels; C y The cutting step size for the input multispectral radar data in the vertical direction; I y is the number of samples cut out in the vertical direction of the input multispectral radar data; T y is the number of pixels in the vertical direction of the input multispectral radar data, in pixels; S y The number of pixels in the vertical direction of a single sample obtained by cutting, in pixels; G y The number of pixels in the overlapping area of ​​two adjacent samples in the vertical direction, in pixels; "⌈...⌉" is rounded up; C x 、 I x 、 T x 、 S x 、 G x 、 C y 、 I y 、 T y 、 S y 、 G y are all positive integers; S32: Load the complete segmentation map, set the judgment condition to clear the background to obtain the first segmentation map; use the raster to vector surface conversion function to convert the effective dense forest area in the raster of the first segmentation map into a vector surface to obtain the first raster data; use the first raster data as a mask and clear the non-forest area of ​​the multispectral radar data after the data preprocessing initially input in S31 to obtain the second raster data; convert the second raster data into a grid point set, and the data field inherited by the grid point set is the value of each pixel of the input raster; input the grid point set into the create Thiessen polygon function to obtain multiple Thiessen polygon areas with the points in the grid point set as the center and the sides adjacent to each other; use the geometric intersection calculation function to calculate the intersection of the first raster data and the Thiessen polygon area to obtain the Thiessen polygon area of ​​the effective dense forest area.

2. The modeling method based on the semantic segmentation model according to claim 1, characterized in that: The S1: remote sensing data preprocessing stage further includes the following steps: S11: Acquire remote sensing data from open source data sources, multispectral radar data from the Sentinel-2 data source, digital elevation model data from the ALOS platform 12.5m resolution data source, and orthochromatic image data from satellite image data sources; S12: Import the remote sensing data into remote sensing image processing software and use seamless mosaicking tools to perform mosaicking operations on the scattered remote sensing data to obtain first remote sensing data; perform radiometric calibration on the first remote sensing data to obtain second remote sensing data; perform atmospheric correction on the second remote sensing data to obtain third remote sensing data; perform orthorectification on the third remote sensing data, and determine the atmospheric model, aerosol model, and ground elevation parameters of the third remote sensing data based on the acquisition time, satellite attitude, orbital position, and data set range of the third remote sensing data to obtain fourth remote sensing data; calculate the enhanced vegetation index (EVI) of the fourth remote sensing data to obtain remote sensing data after data preprocessing; the enhanced vegetation index (EVI) is calculated using the following formula: Where NIR is the near-infrared band reflectivity, which is the 8th band data of the Sentinel-2 radar; R is the red band reflectivity, which is the 4th band data of the Sentinel-2 radar; B is the blue band reflectivity, which is the 2nd band data of the Sentinel-2 radar; G is the gain factor, which is 2.5; C1 and C2 are the first and second atmospheric correction coefficients, which are 6 and 7.5 respectively; L is the soil adjustment parameter, which is 1; S13: Load the pre-processed multispectral radar data and orthophoto panchromatic image data into the geographic information system software, and perform mask mapping on the target data set to be processed; for the pre-processed orthophoto panchromatic image data, perform mask mapping on its outline; for the pre-processed multispectral radar data, refer to the pre-processed orthophoto panchromatic image data and use continuous and closed polygons to perform mask mapping; finally, export the masked data as training data.

3. The modeling method based on the semantic segmentation model according to claim 1, characterized in that: The S4: modeling scenario construction phase also includes the following steps: S41: Import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software, use the terrain import mode to input the terrain data of the same range as the multispectral radar data of the target area, and perform seamless mosaicking and clipping operations on the terrain data to obtain the digital elevation model; S42: Calculate the vegetation quantity per unit area TD based on any rectangular unit in the Thiessen polygon area of ​​the effective dense forest area: Where A is the area of ​​the current measurement area; is the average canopy area of ​​vegetation in the current measurement area; EVI is the EVI value of the current rectangular unit in the Thiessen polygon area of ​​the effective dense forest area; "⌈...⌉" is the rounding operation; After obtaining the number of vegetation per unit area TD, calculate the vegetation distribution points: in, is the easting of the current t-th random coordinate point, and satisfies t≤TD; is the northing distance of the current t-th random coordinate point, and satisfies t≤TD; D east is the westernmost east distance of the current rectangular unit; D west is the easternmost east distance of the current rectangular unit; D north is the northernmost northing distance of the current rectangular unit; D south is the southernmost north distance of the current rectangular unit; N r is a random number that obeys Gaussian distribution; n The number of samples generated to satisfy the uniform distribution condition; X i is the i-th sample value that satisfies the uniform distribution; S43: Import the vegetation distribution points into the modeling software, use the digital elevation model as the elevation data source, and match the vegetation distribution points with the elevation data source to obtain a vegetation spatial coordinate point set.

4. The modeling method based on the semantic segmentation model according to claim 1, characterized in that: The S5: scene asset generation and output stage further includes the following steps: S51: Generate a vegetation monomer model and initialize the attributes of the vegetation monomer model to obtain a rule file, select all coordinate points in the vegetation spatial coordinate point set and bind them to the rule file, and then start the model generation instruction to obtain a dense forest environment model; S52: Perform geometric random transformation on the vegetation monomer model based on CGA rules, and then export the vegetation monomer model and dense woodland environment model as digital twin scene assets and output them.

5. The modeling method based on the semantic segmentation model according to any one of claims 1 to 4, characterized in that: The semantic segmentation model is a U-Net architecture.

6. The modeling method based on the semantic segmentation model according to claim 1, characterized in that: The modeling software is CityEngine and the geographic information system software is ArcGIS Pro.

7. A modeling system based on a semantic segmentation model, characterized by: A method for implementing a semantic segmentation model according to any one of claims 1 to 6, comprising: A remote sensing data preprocessing module is used to acquire remote sensing data, then preprocess the remote sensing data, and finally export the preprocessed remote sensing data as training data; the remote sensing data includes multispectral radar data, orthophoto panchromatic image data, and digital elevation model data; the data preprocessing includes data mosaicking, radiometric calibration, atmospheric correction, orthophoto correction, and quantitative calculation of the Enhanced Vegetation Index (EVI); The semantic segmentation model training and tuning module is used to build a semantic segmentation model and perform model training and model tuning to obtain a trained semantic segmentation model; The forest area semantic segmentation and recognition module is used to use the trained semantic segmentation model to perform pixel-level classification on the multispectral radar data of the target area to obtain semantic segmentation results, and then vectorize the semantic segmentation results to obtain the Thiessen polygon area of ​​the effective dense forest area; The modeling scenario construction module is used to import the digital elevation model data and the Thiessen polygon area of ​​the effective dense forest area into the modeling software to construct the digital elevation model; use the statistical numerical method in the geographic information system software to generate vegetation distribution points under the limitation of the Thiessen polygon area of ​​the effective dense forest area and import them into the modeling software; finally, perform elevation interpolation on the vegetation distribution points to assign three-dimensional height parameters to the vegetation distribution points to obtain a vegetation spatial coordinate point set; The scene asset generation and output module is used to generate a dense forest environment model based on the digital elevation model and the vegetation spatial coordinate point set. The vegetation monomer model in the dense forest environment model uses the model resources built into the modeling software or a self-built model; based on the CGA rules, the vegetation monomer model is geometrically randomized to obtain the digital twin scene asset and output it.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by the processor, the modeling method based on the semantic segmentation model as described in any one of claims 1 to 6 is implemented.

Images