A method for generating city-level real-scene 3D geographic scene data

Abstract

This invention provides a method for generating city-level real-scene 3D geographic scene data, relating to the field of 3D modeling and analysis technology. The method includes: collecting multiple sets of historical defect sample records; identifying multiple modeling defect areas; extracting multiple structural deviation features of the modeling defect areas related to different model construction stages to construct a defect structure association model for each modeling defect type; extracting scene stage association data and constructing a scene stage association list; after determining the target defect area of ​​the target scene model, identifying the target defect type and target scene type of the target defect area; constructing scene association tracing paths and stage association tracing paths for the target defect area; fusing the scene association tracing paths and stage association tracing paths to perform source tracing optimization of the target scene model, generating real-scene 3D geographic scene data corresponding to the target scene model. This invention improves the construction quality and efficiency of city-level real-scene 3D data.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling and analysis technology, and in particular to a method for generating city-level real-scene 3D geographic scene data. Background Technology

[0002] City-level 3D geographic scene models built based on multi-source imagery have become an important data foundation for constructing digital twin cities due to their high accuracy and strong reconstruction capabilities. They are currently widely used in various fields such as urban planning, emergency command, and intelligent transportation. The generation of such models typically involves multiple consecutive modeling stages, such as aerial triangulation, point cloud reconstruction, and mesh reconstruction.

[0003] During the construction of reality models, although multiple structural indicators exist at each stage for quality assessment, some minor anomalies are insufficient to trigger traditional threshold mechanisms. This often leads to potential quality issues going undetected in the early stages, ultimately manifesting as various defects in the reality model, such as texture misalignment, mesh tearing, and structural blurring. Relying on professionals to analyze and troubleshoot stage by stage and region by region is not only inefficient but also difficult to achieve comprehensive coverage. Furthermore, if the differences in modeling sensitivity of different spatial structures in a 3D scene at different stages are not fully utilized, it can further exacerbate the difficulty of anomaly tracing, impacting the efficiency of quality control and repair in the construction of 3D reality models. Summary of the Invention

[0004] To address the aforementioned technical challenges, this invention proposes a method for generating city-level real-scene 3D geographic scene data. Based on historical data, an analysis mechanism is established to identify potential correlations between defects and the modeling stage. By combining the multidimensional features of scene types and structural indicators, the method enables accurate and efficient source tracing analysis of abnormal areas in the final scene model. This method helps improve the efficiency of anomaly tracing and automatic identification capabilities, thereby providing higher-quality support for the efficient construction of large-scale city-level real-scene 3D data.

[0005] This invention provides a method for generating city-level real-scene 3D geographic scene data, including:

[0006] Collect multiple sets of historical defect sample records related to the construction of 3D geographic scene models, and determine the modeling defect areas of the 3D geographic scene models in each set of historical defect sample records;

[0007] Extract the defect type information, region type information, and target defect stage of each modeling defect region. Perform modeling deviation analysis on each modeling defect region and extract multiple structural deviation features of the modeling defect region with respect to different model construction stages.

[0008] Based on the defect type information of the modeled defect area, the target defect stage, and multiple structural deviation characteristics, defect structure correlation analysis is performed to construct a defect structure correlation model for each modeled defect type.

[0009] Extract scene stage association data related to the construction of a 3D geographic scene model, and construct a scene stage association list based on the scene stage association data;

[0010] After determining the target defect area of ​​the target scenario model, identify the target defect type and target scenario type of the target defect area, and determine the scenario association tracing path based on the scenario stage association list and the target scenario type;

[0011] The target deviation features of the target area are extracted, and the stage-related tracing path is determined based on the defect structure association model and the target deviation features. The scene-related tracing path and the stage-related tracing path are integrated to perform tracing optimization of the target scene model and generate real-scene 3D geographic scene data corresponding to the target scene model.

[0012] Preferably, based on the defect type information of the modeled defect region, the target defect stage, and multiple structural deviation characteristics, defect structure correlation analysis is performed to construct a defect structure correlation model for each modeled defect type, including:

[0013] Based on the defect type information of the modeled defect regions and the target defect stage, multiple modeled defect regions are divided into defect type groups to construct multiple defect type clusters.

[0014] Based on the target defect stage of the modeling defect area and the construction stage sequence of the 3D geographic scene model, defect type feature data are extracted from multiple modeling defect areas contained in the defect type cluster, and a defect type feature dataset for each defect type cluster is constructed, which includes defect feature sample data related to each model construction stage.

[0015] Defect structure association analysis is performed on multiple sets of defect feature sample data in the defect type feature dataset. The defect feature sample data at each model construction stage is divided into samples to generate multiple sets of normal sample data and abnormal sample data. Multiple normal sample feature vectors and abnormal sample feature vectors corresponding to the defect feature sample data are constructed. Reference feature vectors for the model construction stage are constructed based on multiple normal sample feature vectors. The association calculation is performed on each abnormal sample feature vector and the reference feature vector to calculate multiple local association indices and generate global defect structure association parameters corresponding to multiple local association indices. A defect structure association model for modeling defect types is constructed, including reference feature vectors and global defect structure association parameters for modeling defect types at different model construction stages.

[0016] Preferably, a scenario stage association list is constructed based on scenario stage association data, including:

[0017] The scenario stage association data includes model building stage distribution data corresponding to various regional scenarios. The defect distribution frequency of each regional scenario in each model building stage is extracted from the model building stage distribution data. The multiple defect distribution frequencies are normalized to generate a scenario stage association list corresponding to each regional scenario, including the defect association parameters of the regional scenario in each model building stage.

[0018] Preferably, for scene-related tracing paths and stage-related tracing paths, the following are also included:

[0019] Based on the target defect type of the target defect area, determine the scene stage association list corresponding to the target scene model. Based on the defect association parameters corresponding to different model construction stages in the scene stage association list, sort multiple model construction stages based on the defect association parameters and generate scene association tracing paths.

[0020] Based on the target deviation characteristics of the target region, generate the stage feature vector corresponding to the target region in each model construction stage. Based on the target defect type of the target defect region, determine the reference feature vector and global defect structure association parameters corresponding to the target scene model in different model construction stages. Perform matching analysis on the stage feature vector and the reference feature vector to calculate the stage defect structure association parameters corresponding to the target defect region in each model construction stage. Based on the global defect structure association parameters and the stage defect structure association parameters, calculate the stage association parameters of the target region in each model construction stage. Based on the stage association parameters, sort multiple model construction stages and generate stage association tracing paths.

[0021] Preferably, the target scene model is optimized by integrating scene-related tracing paths and stage-related tracing paths, including:

[0022] Based on the preset fusion weight coefficient, the defect association parameters and stage association parameters of each model construction stage are weighted and fused to calculate the structural anomaly index of each model construction stage. A fusion tracing path is constructed based on multiple structural anomaly indices, and the target scene model is optimized based on the fusion tracing path.

[0023] Preferably, a defect type feature dataset for each defect type cluster is constructed, including:

[0024] The system identifies multiple sets of structural deviation features associated with each defect type cluster. Based on the model building stage to which the structural deviation features belong, the system divides the multiple sets of structural deviation features associated with each defect type cluster into stages, obtaining multiple sets of structural deviation features corresponding to each model building stage. Based on the modeling order data of the model building stage in the construction stage sequence of the 3D geographic scene model, the system performs feature filtering on the multiple sets of structural deviation features corresponding to each model building stage, obtaining defect feature sample data for each model building stage. This generates a defect type feature dataset for each defect type cluster, which includes defect feature sample data corresponding to each model building stage.

[0025] Preferably, a matching analysis is performed on the stage feature vector and the reference feature vector to calculate the distance between the stage feature vector and the reference feature vector, thereby obtaining the stage defect structure association parameters corresponding to the target defect region in each model construction stage.

[0026] Preferably, the stage association parameters of the target region at each model construction stage are calculated based on the global defect structure association parameters and the stage defect structure association parameters, including using the ratio between the global defect structure association parameters and the stage defect structure association parameters as the stage association parameters of the model construction stage.

[0027] The present invention has the following beneficial effects:

[0028] This invention analyzes historical defect sample records to perform attribution analysis on structural micro-anomalies in the multi-stage modeling process. By constructing a defect structure association model, it quantifies the abnormal characteristics of different modeling stages. Combining the defect susceptibility of scene types in different modeling stages, it constructs a scene stage association list and effectively integrates structural feature deviations and scene-sensitive information to build a multi-source driven modeling anomaly identification path. This allows for targeted optimization and repair of areas in the model that may have hidden modeling defects, achieving efficient and accurate source tracing analysis of defective areas in the final 3D scene model. This improves the modeling consistency and visual quality of the final generated real-world 3D geographic scene data. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a method for generating city-level real-scene 3D geographic scene data, as provided in an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0031] like Figure 1 As shown in the figure, an embodiment of the present invention provides a method for generating city-level real-scene 3D geographic scene data, which specifically includes the following steps:

[0032] Step S1: Collect multiple sets of historical defect sample records for constructing a 3D geographic scene model, and determine the modeling defect areas of the 3D geographic scene model in each set of historical defect sample records.

[0033] Specifically, multiple sets of historical defect sample records can be collected from past modeling tasks involving the construction of 3D geographic scene models. These records include information related to modeling defects in the initially constructed 3D geographic scene model. For example, for defects such as texture misalignment, mesh tearing, and structural blurring in the scene model, this includes the location of the corresponding defect areas and the structural feature information at each stage. The data can also include information on different modeling defect areas from different historical defect sample records annotated manually. This information may also include, for example, the target defect stage that caused the anomaly in the area, and the distribution data of defects that are prone to occur in modeling under different scene types. This provides foundational data support for subsequent feature extraction and attribution modeling.

[0034] Step S2: Extract the defect type information, region type information and target defect stage of each modeling defect region, perform modeling deviation analysis on each modeling defect region, and extract multiple structural deviation features of the modeling defect region with respect to different model construction stages.

[0035] Specifically, the defect type information includes the specific modeling defect type corresponding to the modeling defect area in the 3D geographic scene model, and the region type information includes the scene type corresponding to the modeling defect area, such as eaves, curtain walls, grassland, roads, etc. The target defect stage is used to indicate the type of defect appearing in the modeling defect area in the 3D geographic scene model, and which key stage it was mainly caused by. Specifically, it can be relevant information obtained through manual analysis and annotation. For example, for a mesh tear defect in a certain area, after investigation, it was found that it was caused by a minor anomaly in the point cloud reconstruction stage.

[0036] It's worth noting that constructing a city-level 3D geographic scene model based on multi-source imagery involves multiple processing steps after data acquisition. Depending on the specific construction task, this can be divided into several model construction stages. These include aerial triangulation (calculating the relative position and orientation of photos), dense point cloud reconstruction (recovering spatial points from photos), mesh reconstruction (connecting different points in the point cloud data to form a mesh model), and texture mapping (applying different photos to the mesh model to create a rich 3D structure). Furthermore, some modeling defects in the final scene model are related to these earlier stages. For example, texture mapping distortion may be caused by mesh surface distortion during mesh reconstruction, and local breaks or holes in the model may be due to insufficient overlap of regional images during aerial triangulation.

[0037] To address this issue, we further explored the structural deviation characteristics of the modeling defect region across different model building stages, which will be used for subsequent correlation analysis between modeling defects and structural features. Specifically, representative structural feature parameters can be extracted for different model building stages. For example, in the aerial triangulation reconstruction stage, after calculating camera pose and sparse point positions, the 3D points are reprojected back onto the image, and the pixel distance difference between the projected points and the actual image points is measured to obtain the reprojection residual characteristics of the aerial triangulation reconstruction stage. Here, the reprojection residual = These are the pixel coordinates of the projection point and the actual image point, respectively. This represents the number of pixels.

[0038] In the point cloud reconstruction stage, the ratio of the number of points to the area in a local region is calculated to obtain the local density feature of the point cloud, representing the number of points per unit area. In the mesh reconstruction stage, the degree of abrupt change in the angle between the normals of adjacent faces is calculated, i.e., the mesh normal variation rate, representing the angle between adjacent faces. The mesh normal variation rate = These are the normal vectors of adjacent faces. This represents the number of dough pieces.

[0039] In the process of extracting structural features, those skilled in the art can extract different types of representative feature information for different model building stages, including structural feature parameters of different dimensions such as residual index, density index, normal jump rate, and texture projection offset angle, thereby obtaining multiple structural deviation features for different model building stages. This embodiment does not specifically limit them.

[0040] Step S3: Based on the defect type information of the modeled defect area, the target defect stage, and multiple structural deviation characteristics, perform defect structure correlation analysis to construct a defect structure correlation model for each modeled defect type.

[0041] Specifically, based on the aforementioned historical samples with clear defect attribution, namely the target defect stage and multiple structural deviation characteristics of different defect types, the correlation characteristics between defect types and model structural features at different modeling stages are analyzed, and a defect structure correlation model is constructed to represent the implicit correlation information between modeling defect types and model structural features.

[0042] In one implementation process, the process of constructing a defect structure association model for each modeled defect type specifically includes:

[0043] Based on the defect type information of the modeling defect region and the target defect stage, multiple modeling defect regions are divided into defect types to construct multiple defect type clusters, where each defect type cluster represents one type of modeling defect.

[0044] Based on the target defect stage of the modeling defect area and the construction stage sequence of the 3D geographic scene model, defect type feature data are extracted from multiple modeling defect areas contained in the defect type cluster, and a defect type feature dataset for each defect type cluster is constructed.

[0045] In this embodiment, the construction stage sequence of the 3D geographic scene model is used to indicate the construction order of the scene model. For example, aerial triangulation is performed first, followed by point cloud reconstruction and mesh reconstruction, indicating the modeling order relationship corresponding to each model construction stage. For each defect type cluster, based on the defect type information of the modeling defect area, multiple sets of structural deviation features associated with each defect type cluster are first determined, that is, the structural feature information corresponding to different modeling defect areas of the modeling defect type and defect type cluster. Furthermore, based on the model construction stage to which the structural deviation features belong, the multiple sets of structural deviation features associated with the defect type cluster are further divided into stages to obtain multiple sets of structural deviation features corresponding to each model construction stage, which serve as the initial sample data for each model construction stage.

[0046] Furthermore, considering that anomalies in certain stages during the sequential modeling process can lead to anomalies in subsequent stages, feature information is filtered out based on the modeling order data in the construction stage sequence to prevent contamination of samples from later modeling stages. For example, in the initial sample data of the point cloud reconstruction stage, feature information of the target defect stage located between stages is filtered out. For instance, if a set of structural deviation features corresponds to the aerial triangulation calculation stage in the initial sample data of the point cloud reconstruction stage, this data is removed. This ensures that the target defect stage corresponding to each set of structural deviation features does not precede the point cloud reconstruction stage in the scene model construction order. In this way, defect feature sample data for each model construction stage is constructed. In this case, structural deviation feature data where the target defect stage is the point cloud reconstruction stage can be considered abnormal sample data, while structural deviation feature data in other stages after the point cloud reconstruction stage can be considered normal sample data, because the target defect stage causing the modeling defect occurs in a subsequent modeling stage, thus achieving state identification of different sample data. This method yields the defect type feature dataset for each defect type cluster, which includes defect feature sample data corresponding to each model building stage.

[0047] Defect structure association analysis is performed on multiple sets of defect feature sample data in the defect type feature dataset to construct a defect structure association model for modeling defect types, including reference feature vectors and global defect structure association parameters for modeling defect types at different model construction stages.

[0048] In this embodiment, the defect feature sample data of each model construction stage are divided into multiple sets of normal sample data and abnormal sample data, i.e., the sample state identification method based on the construction order of the scene model is used for the division. For normal sample data and abnormal sample data, multiple normal sample feature vectors and abnormal sample feature vectors are constructed according to structural deviation characteristics. Then, using multiple normal sample feature vectors as a reference, a reference feature vector for the model construction stage is determined based on their average level. For example, the average value of multiple normal sample feature vectors in the same model construction stage is taken as the reference feature vector for that stage. Based on this reference feature vector, the correlation calculation is performed on multiple abnormal sample feature vectors in that stage. For example, the distance between two vectors is calculated to obtain the corresponding local correlation index, where: Local correlation index = It is the k-th dimension of the feature vector of the abnormal sample. It is the k-th dimension of the reference eigenvector. It is the dimension of the feature vector.

[0049] Then, the global defect structure association parameters in the model building stage are calculated based on multiple local association indices. For example, the mean of multiple local association indices is used as the global defect structure association parameters. Finally, based on the reference feature vectors of the modeled defect type in different model building stages and the global defect structure association parameters, a defect structure association model of the modeled defect type is constructed as a key basis for anomaly tracing analysis when a certain defect type appears in a specific area.

[0050] Step S4: Extract scene stage association data related to the construction of the 3D geographic scene model, and construct a scene stage association list based on the scene stage association data.

[0051] Specifically, in actual modeling, different geometric structures can affect the modeling error variability at different stages. For example, in eaves scenes, due to significant image occlusion and small angles, anomalies are prone to occur in the aerial triangulation stage. In building facade scenes, due to large projection angles and susceptibility to misalignment and stretching, texture mapping is affected. To address this, a correlation analysis is performed between the scene and the modeling stage. Scene stage correlation data includes distribution data of model building stages corresponding to various regional scenes, such as eaves, glass curtain walls, rooftop structures, and road edges. This data examines the defect occurrence stages in historical modeling tasks. Specifically, it can refer to situations where modeling defects occurred in that area during the modeling process, or, more specifically, the dominant modeling stage confirmed after the area was marked as a modeling defect area, i.e., the aforementioned target defect stage.

[0052] Statistical analysis was performed on the distribution data of the model building stages to extract the frequency of defect distribution for each regional scenario at each model building stage, such as the number of samples appearing in a specific stage as the defect-dominant stage. Considering the differences in sample size under different scenario types, the stage defect distribution of each scenario type was normalized to achieve comparability across scenario types. Finally, the defect correlation parameters of different regional scenarios at each model building stage were determined, and a scenario stage correlation list corresponding to each regional scenario was constructed to reflect the error-prone tendencies of different types of scenarios at each modeling stage. This helps to improve the ability to distinguish the source and sort when the abnormal scores are similar in multiple stages.

[0053] Step S5: After determining the target defect area of ​​the target scene model, identify the target defect type and target scene type of the target defect area, and determine the scene association tracing path based on the scene stage association list and the target scene type.

[0054] In this embodiment, for the modeling process of a city-level real-world 3D geographic scene, such as a target scene model, when a new target defect area is identified—that is, when different model building stages are all identified as normal through simple threshold judgments during the modeling process, but a certain area in the final scene model has a modeling defect—the process first identifies the target defect type and target scene type of the area, corresponding to the actual defect manifestation and scene type manifestation of the area, respectively. Then, based on the aforementioned scene stage association list, according to the target defect type, the scene stage association list corresponding to the target scene model is determined. Based on the defect association parameters corresponding to different model building stages in the scene stage association list, multiple model building stages are sorted in descending order of defect association parameters as the priority inspection path for the defect area in the scene type dimension, i.e., the scene association tracing path. This process uses historical statistical results to model the impact of scene structure on modeling anomalies, thereby providing an anomaly analysis path oriented towards spatial structure differences.

[0055] Step S6: Extract the target deviation features of the target area, determine the stage-related tracing path based on the defect structure association model and the target deviation features, integrate the scene-related tracing path and the stage-related tracing path to perform tracing optimization on the target scene model, and generate real-scene 3D geographic scene data corresponding to the target scene model.

[0056] In this embodiment, to achieve attribution analysis of the target defect region in the modeling process, the aforementioned defect structure association model is used as a benchmark. Based on the target deviation characteristics of the target region, the potential anomaly degree of the region at each modeling stage is evaluated, and a stage-related tracing path is ultimately generated. In this process, firstly, based on the target deviation characteristics of the target region at different model building stages, a stage feature vector corresponding to the target region at each model building stage is constructed. Then, combined with the target defect type of the target defect region, the reference feature vector and global defect structure association parameters corresponding to the target defect region at different model building stages are determined from the defect structure association model, and attribution matching analysis is performed on the target region based on this information.

[0057] Specifically, the calculation process for the stage defect structure correlation parameter between the stage feature vector and the reference feature vector is the same as that for the global defect structure correlation parameter. For example, the distance between vectors is represented by Euclidean distance to calculate the stage defect structure correlation parameter. Then, based on the global defect structure correlation parameter and the stage defect structure correlation parameter, the stage correlation parameter for the target region at each model building stage is calculated. Specifically, the stage defect structure correlation parameter is used to characterize the degree of local structural deviation at different stages. By introducing the global defect structure correlation parameter for that stage, the stage defect structure correlation parameter characterizing the degree of deviation is corrected. The ratio between the global defect structure correlation parameter and the stage defect structure correlation parameter is used as the stage correlation parameter for the model building stage. The larger the value, the higher the relative intensity of structural anomalies in the current target region at that stage. Finally, based on the magnitude of the stage correlation parameter values, multiple model building stages are ranked, and a stage correlation tracing path is generated, representing the probability ranking of which modeling stages are most likely to cause the modeling defects in the current region.

[0058] Furthermore, for the stage-related source path reflecting the ranking of structural anomaly intensity in each modeling stage of the target area, and the scene-related source path reflecting the error-prone probability tendency of the scene in the historical data at each stage, in order to give full play to the complementary advantages of the two, a source path fusion mechanism is implemented. This allows for high-confidence modeling stage localization and optimization reconstruction of areas with defects in the target scene model, ultimately outputting quality-controllable city-level real-scene 3D geographic scene data.

[0059] In this process, the defect correlation parameters and stage correlation parameters of each model construction stage can be weighted and fused according to the preset fusion weight coefficients. The structural anomaly index = fusion weight coefficient. ×Defect correlation parameters + Fusion weighting coefficient The specific parameters related to the × stage are the sum of the fusion weight coefficients corresponding to the two, i.e., the fusion weight coefficients. and fusion weight coefficient The sum of these values ​​is 1. The values ​​can be reasonably set according to the actual task requirements or the importance of the scene impact and stage impact as determined by experts. Finally, the structural anomaly index of each model construction stage is calculated, and multiple model construction stages are reordered to obtain a fusion traceability path. Then, for this path, the target scene model can be traced and optimized based on pre-set defect repair strategies for different stages or by using manual intervention repair methods. The target scene model after traceability optimization is used as the result of generating city-level real-scene 3D geographic scene data, thereby improving the consistency of the overall structure and the visualization quality.

[0060] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for generating city-level real-scene 3D geographic scene data, characterized in that, include: Collect multiple sets of historical defect sample records related to the construction of 3D geographic scene models, and determine the modeling defect areas of the 3D geographic scene models in each set of historical defect sample records; Extract the defect type information, region type information and target defect stage of each modeling defect region, perform modeling deviation analysis on each modeling defect region, and extract multiple structural deviation features of the modeling defect region with respect to different model construction stages; Based on the defect type information of the modeled defect regions and the target defect stage, multiple modeled defect regions are divided into defect type groups to construct multiple defect type clusters. Based on the target defect stage of the modeling defect area and the construction stage sequence of the 3D geographic scene model, defect type feature data are extracted from multiple modeling defect areas contained in the defect type cluster, and a defect type feature dataset for each defect type cluster is constructed, which includes defect feature sample data related to each model construction stage. Defect structure association analysis is performed on multiple sets of defect feature sample data in the defect type feature dataset. The defect feature sample data at each model building stage is divided into samples to generate multiple sets of normal sample data and abnormal sample data. Multiple normal sample feature vectors and abnormal sample feature vectors corresponding to the defect feature sample data are constructed. Reference feature vectors for the model building stage are constructed based on multiple normal sample feature vectors. The association calculation is performed on each abnormal sample feature vector and the reference feature vector to calculate multiple local association indices and generate global defect structure association parameters corresponding to multiple local association indices. A defect structure association model for modeling defect types is constructed, including reference feature vectors and global defect structure association parameters for modeling defect types at different model building stages. Extract scene stage association data related to the construction of a 3D geographic scene model, and construct a scene stage association list based on the scene stage association data; After determining the target defect area of ​​the target scenario model, identify the target defect type and target scenario type of the target defect area, and determine the scenario association tracing path based on the scenario stage association list and the target scenario type; The target deviation features of the target area are extracted, and the stage-related tracing path is determined based on the defect structure association model and the target deviation features. The scene-related tracing path and the stage-related tracing path are integrated to perform tracing optimization of the target scene model and generate real-scene 3D geographic scene data corresponding to the target scene model.

2. The method for generating city-level real-scene 3D geographic scene data according to claim 1, characterized in that, Construct a scenario stage association list based on scenario stage association data, including: The scenario stage association data includes model building stage distribution data corresponding to various regional scenarios. The defect distribution frequency of each regional scenario in each model building stage is extracted from the model building stage distribution data. The multiple defect distribution frequencies are normalized to generate a scenario stage association list corresponding to each regional scenario, including the defect association parameters of the regional scenario in each model building stage.

3. The method for generating city-level real-scene 3D geographic scene data according to claim 2, characterized in that, For scenario-related tracing paths and stage-related tracing paths, the following are also included: Based on the target defect type of the target defect area, determine the scene stage association list corresponding to the target scene model. Based on the defect association parameters corresponding to different model construction stages in the scene stage association list, sort multiple model construction stages based on the defect association parameters and generate scene association tracing paths. Based on the target deviation characteristics of the target region, generate the stage feature vector corresponding to the target region in each model construction stage. Based on the target defect type of the target defect region, determine the reference feature vector and global defect structure association parameters corresponding to the target scene model in different model construction stages. Perform matching analysis on the stage feature vector and the reference feature vector to calculate the stage defect structure association parameters corresponding to the target defect region in each model construction stage. Based on the global defect structure association parameters and the stage defect structure association parameters, calculate the stage association parameters of the target region in each model construction stage. Based on the stage association parameters, sort multiple model construction stages and generate stage association tracing paths.

4. The method for generating city-level real-scene 3D geographic scene data according to claim 3, characterized in that, The target scene model is optimized by integrating scene-related tracing paths and stage-related tracing paths, including: Based on the preset fusion weight coefficient, the defect association parameters and stage association parameters of each model construction stage are weighted and fused to calculate the structural anomaly index of each model construction stage. A fusion tracing path is constructed based on multiple structural anomaly indices, and the target scene model is optimized based on the fusion tracing path.

5. The method for generating city-level real-scene 3D geographic scene data according to claim 4, characterized in that, Construct a defect type feature dataset for each defect type cluster, including: The system identifies multiple sets of structural deviation features associated with each defect type cluster. Based on the model building stage to which the structural deviation features belong, the system divides the multiple sets of structural deviation features associated with each defect type cluster into stages, obtaining multiple sets of structural deviation features corresponding to each model building stage. Based on the modeling order data of the model building stage in the construction stage sequence of the 3D geographic scene model, the system performs feature filtering on the multiple sets of structural deviation features corresponding to each model building stage, obtaining defect feature sample data for each model building stage. This generates a defect type feature dataset for each defect type cluster, which includes defect feature sample data corresponding to each model building stage.

6. The method for generating city-level real-scene 3D geographic scene data according to claim 5, characterized in that, Matching analysis is performed on the stage feature vector and the reference feature vector to calculate the distance between the stage feature vector and the reference feature vector, thereby obtaining the stage defect structure association parameters corresponding to the target defect region in each model construction stage.

7. The method for generating city-level real-scene 3D geographic scene data according to claim 3, characterized in that, Based on the global defect structure association parameters and the stage defect structure association parameters, the stage association parameters of the target region in each model construction stage are calculated, including using the ratio between the global defect structure association parameters and the stage defect structure association parameters as the stage association parameters of the model construction stage.

Images