Adaptive TIN digital twin terrain grid density adjustment method and device based on machine learning

Abstract

The embodiment of the invention discloses a machine learning-based adaptive TIN (Triangulated Irregular Network) digital twin terrain grid density adjustment method and device. The method comprises the following steps: acquiring topographic data; cleaning and preprocessing the topographic data to obtain a preprocessing result; extracting features corresponding to the preprocessing result; performing region division and clustering analysis on the features to obtain a processing result; generating a TIN grid according to the processing result; inputting the TIN grid into a grid optimization model to adjust the optimal grid density of different areas so as to obtain a final TIN grid; and outputting the final TIN grid. By implementing the method provided by the embodiment of the invention, the grid density can be increased in places needing details, and the grid density can be properly reduced in flat areas, so that the modeling precision is improved, the calculation efficiency is optimized, and the three-dimensional terrain model is more real and effective.

Description

Technical Field

[0001] The present invention relates to a TIN digital twin terrain system, and more specifically to an adaptive TIN digital twin terrain grid density adjustment method and device based on machine learning. Background Art

[0002] When building a 3D terrain model, it is crucial to accurately reflect the terrain features. Although the traditional mesh subdivision method is simple, it often causes discontinuities when dealing with terrain undulations because it relies on a fixed mesh edge length or resolution. For example, flat areas may be too dense, resulting in unnecessary calculations, while complex areas may lose details due to sparse meshes.

[0003] In contrast, the TIN (Triangulated Irregular Network) model better captures the local features of the terrain by connecting discrete terrain points into triangles to form a non-overlapping triangulated network. This structure allows for more flexible adaptation to different terrain changes, allowing the model to maintain high accuracy in complex areas while reducing details in simple areas to improve efficiency. However, the traditional TIN modeling method lacks adaptability when subdividing the grid and cannot be flexibly adjusted according to the actual undulations of the terrain, resulting in grids that are too sparse in some areas and too dense in other areas.

[0004] Therefore, it is necessary to design a new method to increase the mesh density where details are needed and appropriately reduce it in flat areas, so as to improve the modeling accuracy, optimize the calculation efficiency, and make the three-dimensional terrain model more realistic and effective. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and provide a method and device for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: an adaptive TIN digital twin terrain grid density adjustment method based on machine learning, comprising: Get terrain data; Cleaning and preprocessing the terrain data to obtain a preprocessing result; Extracting features corresponding to the preprocessing results; Performing regional division and cluster analysis on the features to obtain processing results; Generate a TIN grid according to the processing result; Inputting the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; Output the final TIN grid.

[0007] A further technical solution is: the cleaning and preprocessing of the terrain data to obtain a preprocessing result includes: Performing noise removal and precision balancing on the terrain data to obtain a first processing result; Performing data difference and oversampling processing on the first processing result to obtain a second processing result; The second processing result is standardized and optimized to obtain a preprocessing result.

[0008] A further technical solution is: extracting the features corresponding to the preprocessing results includes: The terrain features and geometric features in the preprocessing result are extracted.

[0009] A further technical solution is: performing regional division and cluster analysis on the features to obtain processing results, including: Extracting contour lines from the terrain data and dividing the area by a set elevation threshold to obtain different elevation zones; Based on the terrain features, the regions are divided from different slopes, slope aspects and terrain classification results to obtain mountain ranges and river valley regions; Cluster analysis is performed on different elevation zones as well as mountain ranges and river valley areas to obtain the analysis results.

[0010] A further technical solution is: generating a TIN grid according to the processing result includes: constructing a digital elevation model based on the terrain data; Generate a TIN grid using a Delaunay triangulation algorithm according to the digital elevation model and the analysis results; Perform preliminary optimization on the generated TIN mesh.

[0011] The further technical solution is: the preliminary optimization of the generated TIN grid includes: The generated TIN mesh is subjected to node position adjustment, local smoothing, triangle quality optimization, and reduction of redundant nodes and triangle numbers.

[0012] Its further technical solution is: the grid optimization model is a model obtained by generating a TIN grid through cleaning and preprocessing, feature extraction, regional division and cluster analysis of the acquired historical terrain data, and performing preliminary optimization, dynamic grid refinement and grid shape optimization on the generated TIN grid, and using the manually assisted adjusted TIN grid as a sample set to train convolutional neural networks and graph neural networks.

[0013] The adaptive TIN digital twin terrain grid density adjustment device based on machine learning includes: A terrain data acquisition unit, used for acquiring terrain data; A preprocessing unit, used for cleaning and preprocessing the terrain data to obtain a preprocessing result; A feature extraction unit, used to extract features corresponding to the preprocessing result; A division and analysis unit, used for performing regional division and cluster analysis on the features to obtain processing results; A grid generation unit, used for generating a TIN grid according to the processing result; An optimization unit, used for inputting the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; Output unit, used to output the final TIN grid.

[0014] The present invention further provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor implements the above method when executing the computer program.

[0015] The present invention also provides a storage medium, wherein the storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0016] Compared with the prior art, the present invention has the following beneficial effects: after acquiring terrain data, the present invention performs cleaning and preprocessing to extract relevant features and perform regional division and cluster analysis, generates a TIN grid according to the processing results, and inputs it into a grid optimization model to adjust the optimal grid density of different regions, and finally outputs the optimized TIN grid; the grid density is increased where details are required, and appropriately reduced in flat areas, thereby improving the accuracy of modeling, optimizing calculation efficiency, and making the three-dimensional terrain model more realistic and effective.

[0017] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0019] Figure 1 A schematic diagram of an application scenario of the method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 2 A schematic flow chart of a method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 3 A schematic diagram of a sub-process of a method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 4 A schematic diagram of a sub-process of a method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 5 A schematic diagram of a sub-process of a method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 6 A schematic block diagram of a device for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning provided in an embodiment of the present invention; Figure 7 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0022] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0024] See also Figure 1 and Figure 2 , Figure 1 Schematic diagram of application scenario of the adaptive TIN digital twin terrain grid density adjustment method based on machine learning provided in an embodiment of the present invention. Figure 2 A schematic flow chart of an adaptive TIN digital twin terrain grid density adjustment method based on machine learning provided in an embodiment of the present invention. The adaptive TIN digital twin terrain grid density adjustment method based on machine learning is applied to a server. The server interacts with the terminal, analyzes the terrain data, and dynamically adjusts the density and shape of the grid using a machine learning algorithm. A denser grid can be used in areas with large terrain undulations, while a sparser grid can be used in relatively flat areas. This adaptive subdivision strategy not only improves the accuracy and detail of the model, but also effectively avoids the surface discontinuity phenomenon that occurs in traditional methods. This approach better meets the needs of modern geographic information systems, environmental monitoring, and digital twin technologies, and lays a solid foundation for terrain analysis and application.

[0025] Figure 2 Schematic diagram of the process of adjusting the grid density of the adaptive TIN digital twin terrain based on machine learning provided by an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110 to S170.

[0026] S110: Acquire terrain data.

[0027] In this embodiment, the terrain data is the elevation data corresponding to the actual terrain state of the area where the digital twin terrain model is to be generated. The source of these elevation data is not limited to a specific measurement method and can be flexibly selected according to actual needs. The following is a detailed introduction to the source of elevation data: Remote sensing satellite data: uses sensors on satellites (such as synthetic aperture radar or lidar) to obtain large-scale elevation data, which is suitable for analysis and monitoring of wide areas and has the advantages of high efficiency and low cost.

[0028] Aerial survey data: High-density terrain data obtained through high-precision equipment equipped with aircraft or drones (such as lidar or photogrammetry technology). It is particularly suitable for rapid and detailed analysis of cities or complex terrains. It is usually combined with ground surveys to improve measurement accuracy.

[0029] Ground survey data: High-precision surveying using traditional surveying equipment (such as total stations or GPS) is suitable for projects with high precision requirements within a small area, such as architectural design. This data is often used to verify and supplement data from other sources.

[0030] By comprehensively using the above multiple data sources, we can ensure that the generated target terrain model accurately reflects the actual terrain status to meet the needs of different application scenarios.

[0031] S120: Clean and preprocess the terrain data to obtain a preprocessing result.

[0032] In this embodiment, the preprocessing result refers to the result formed after cleaning and preprocessing the terrain data.

[0033] In one embodiment, see Figure 3 , the above-mentioned step S120 may include steps S121~S123.

[0034] S121. Perform noise removal and precision balancing on the terrain data to obtain a first processing result.

[0035] In this embodiment, noise removal is to process data using statistical analysis and filtering techniques. Common filtering methods include: Median filtering: By taking the median of the data in the neighborhood to replace the value of the center point, spike noise can be effectively removed.

[0036] Mean filtering: Smooths data by calculating the average value of data in the neighborhood, which is suitable for removing general noise.

[0037] Outlier detection is to combine machine learning algorithms to process outliers, such as: Isolation Forest: Identifies abnormal data points by building random trees, suitable for processing high-dimensional data.

[0038] Outlier detection: Identify and remove outliers using clustering algorithms or density-based methods such as DBSCAN.

[0039] Supersampling: The density of data points is increased through interpolation methods to ensure that the data can fully reflect the terrain features. In this process, it is necessary to carefully balance the accuracy and noise of the data to avoid introducing false terrain features.

[0040] S122: Perform data difference and oversampling processing on the first processing result to obtain a second processing result.

[0041] In this embodiment, the interpolation method is to fill in the missing data points. Common interpolation methods include: Kriging interpolation: A model based on spatial statistics, suitable for processing data with spatial correlation and can provide more accurate interpolation results.

[0042] Spline interpolation: Use low-order polynomials to smoothly connect data points, suitable for handling smooth curves.

[0043] Supersampling: Generate additional data points between existing data points to increase the resolution and detail of the model. Common interpolation algorithms include: Nearest Neighbor Interpolation: A fast and simple interpolation method suitable for handling categorical data.

[0044] Bilinear interpolation: Based on the linear weighting of four adjacent data points, it is suitable for processing uniformly distributed data.

[0045] Bicubic interpolation: Smoother than bilinear interpolation and suitable for processing high-resolution data.

[0046] S123: Standardize and optimize the second processing result to obtain a preprocessing result.

[0047] In this embodiment, the elevation data is standardized to ensure that different features are on the same scale, which facilitates subsequent analysis and modeling. Common standardization methods include: Normalization: Scaling the data to a specific interval (such as 0 to 1) so that the data can be compared within the same range.

[0048] Z-score standardization: Make the data have zero mean and unit variance by subtracting the mean and dividing by the standard deviation.

[0049] Data feature optimization: Combine machine learning technology to further improve the quality of data features. Common methods include: Principal Component Analysis (PCA): used for dimensionality reduction, projecting data into a new feature space, extracting main features, and reducing redundant information.

[0050] Feature selection: Use algorithms (such as LASSO regression) to select features that have the greatest impact on model performance, thereby improving the accuracy and efficiency of the model.

[0051] Through the above steps, the elevation data after data cleaning and preprocessing can more accurately reflect the actual terrain status and lay a solid foundation for the subsequent terrain model generation.

[0052] S130: Extract features corresponding to the preprocessing results.

[0053] In this embodiment, the features include terrain features and geometric features.

[0054] Specifically, terrain features and geometric features in the preprocessing result are extracted.

[0055] After data preprocessing is completed, feature extraction is the key step in generating an adaptive TIN digital twin terrain model, which can significantly improve the performance and accuracy of the model. Feature extraction mainly includes two aspects: Terrain feature extraction: Elevation Features: Ridgelines and Valleys: Identify ridgelines and valley locations to reveal major terrain structures.

[0056] Slope and Aspect: Calculate the slope and aspect to analyze the tilted characteristics of the terrain.

[0057] Elevation Change Rate: Evaluate the rate of elevation change and identify areas of abrupt changes such as cliffs and mountains.

[0058] Terrain classification: Terrain type: The terrain is divided into different types such as mountains and plains through machine learning algorithms.

[0059] Vegetation coverage: Combine remote sensing data to extract vegetation information and enrich terrain description.

[0060] Geometric feature extraction: TIN generation: Triangle features: Extract geometric properties of each triangle in the TIN, such as area and side length.

[0061] Mesh Density: Analyzes the mesh density to assess the level of terrain detail.

[0062] Boundary Features: Terrain Boundaries: Extract natural and artificial boundaries, such as mountains and river directions.

[0063] Characteristic lines: Identify important characteristic lines, such as fault lines and ridge lines, to help understand the terrain structure.

[0064] S140: performing region division and cluster analysis on the features to obtain processing results.

[0065] In this embodiment, the processing result refers to the characteristics of dividing areas such as mountains and plains, mountain ranges, and river valleys.

[0066] In one embodiment, see Figure 4 , the above-mentioned step S140 may include steps S141~S143.

[0067] S141, extracting contour lines from the terrain data, and dividing the area by a set elevation threshold to obtain different elevation zones.

[0068] In this embodiment, contour lines are extracted from a digital elevation model (DEM), and interpolation can be used to convert discrete elevation points into continuous elevation lines. These contour lines help identify the boundaries and shapes of different elevation areas.

[0069] Contour lines provide a visual representation of terrain undulations, showing the distribution of ridges, valleys, and slopes, and are useful for terrain analysis and mapping.

[0070] Set one or more elevation thresholds to divide elevation data into different areas, such as plains below a certain elevation and mountains above that elevation. You can determine the appropriate threshold by analyzing the elevation distribution histogram.

[0071] This method helps to quickly delineate terrain areas and lays the foundation for subsequent feature analysis.

[0072] S142. Divide the area based on the terrain features from different slopes, slope directions and terrain classification results to obtain mountain ranges and river valley areas.

[0073] In this embodiment, the slope (degree of inclination) and aspect (direction of the slope) of each pixel are calculated, which is usually achieved by calculating the elevation change of neighboring pixels. Terrain analysis tools can be used to obtain these features.

[0074] Slope and aspect zoning can be used to assess water flow paths, soil erosion risk, suitability for vegetation growth, etc.

[0075] Use machine learning or rule-based methods to classify terrain based on labeled training data and identify different terrain types (such as mountains, valleys, rivers, etc.).

[0076] This segmentation provides important information for landscape management, development planning, and ecological research.

[0077] S143. Cluster analysis is performed on different elevation zones, mountain ranges, and river valley areas to obtain analysis results.

[0078] In this embodiment, cluster analysis is performed in the following manner: Cluster analysis based on data: K-means clustering: Select a K value, randomly initialize K cluster centers, and then iteratively adjust the cluster assignment of each point and update the cluster center until convergence (the cluster center no longer changes). K-means clustering can be used to identify major terrain patterns, such as mountain distribution, valley characteristics, etc., which helps to simplify data and pattern recognition.

[0079] Hierarchical clustering: Choose a bottom-up or top-down clustering method to build a hierarchical clustering tree (dendrogram). Each data point starts as a separate cluster, and then gradually merges based on similarity, or splits into smaller clusters from a large cluster. Hierarchical clustering can reveal the hierarchical relationship of terrain data and the connection between regions, which helps to understand the complex structure of the terrain.

[0080] Cluster analysis based on machine learning: DBSCAN: Set the distance threshold and the minimum number of samples, and the algorithm clusters by identifying high-density areas and adjacent areas. Noise points are not included in any clusters, which is suitable for processing non-uniformly distributed data. DBSCAN is particularly suitable for complex terrain data, and can effectively identify characteristic areas such as ridges and valleys, while filtering out isolated points.

[0081] Spectral clustering: First, a similarity matrix is ​​constructed, and then the data is mapped to a low-dimensional space through spectral decomposition. K-means clustering is performed in the new space to identify complex terrain patterns. Spectral clustering is suitable for capturing nonlinear features and can effectively identify complex terrain structures, such as diverse landform features and regional changes.

[0082] These steps provide a systematic basis for generating an adaptive TIN digital twin terrain model, enabling the model to accurately reflect the real terrain characteristics.

[0083] S150: Generate a TIN grid according to the processing result.

[0084] In one embodiment, see Figure 5 , the above-mentioned step S150 may include steps S151~S153.

[0085] S151, constructing a digital elevation model according to the terrain data.

[0086] In this embodiment, the process of constructing a digital elevation model (DEM) can be divided into the following steps in detail: First, discrete elevation points need to be interpolated to generate a continuous terrain surface. Common interpolation methods include: Kriging interpolation: This is an interpolation technique based on spatial statistics. It takes into account the spatial correlation between elevation data points. It is suitable for processing terrain data with randomness and unevenness and can provide higher prediction accuracy.

[0087] Spline interpolation: Use piecewise polynomials to connect data points to generate smooth curves. This method is suitable for terrain with smoother changes and can better capture subtle changes in terrain.

[0088] After interpolation, the elevation points are converted into irregular triangulated networks (TIN) using the Delaunay triangulation algorithm. The advantages of this algorithm are: Triangle quality: Ensure that the generated triangles have no irregularities in long and short sides, thus ensuring the stability of the network and the effectiveness of the calculation.

[0089] S152, generating a TIN grid using a Delaunay triangulation algorithm according to the digital elevation model and the analysis results.

[0090] In this embodiment, the generation of the TIN grid generally includes the following steps: Select the initial point pair: Select the two points with the closest distance from all elevation points to form the initial edge.

[0091] Construct the first triangle: Search for the third point on the right side of the initial edge to form the first triangle. Use the Delaunay criterion to ensure that the new point and the endpoints of the initial edge form a valid triangle.

[0092] Extended triangulation: Take the three sides of the existing triangle as the baseline and find the extension point that forms a triangle with the baseline. Make sure that the third point of the triangle formed by the extension point and the baseline is located on both sides of the baseline.

[0093] Repeat the expansion process until all elevation points are included in the TIN grid.

[0094] Through these steps, an accurate digital elevation model can be effectively generated to reflect the real terrain characteristics and lay the foundation for subsequent terrain analysis and application.

[0095] S153. Perform preliminary optimization on the generated TIN grid.

[0096] In this embodiment, the generated TIN mesh is subjected to node position adjustment, local smoothing, triangle quality optimization, and reduction of redundant nodes and triangle numbers.

[0097] Specifically, after generating the preliminary TIN (Triangulated Irregular Network) mesh, preliminary automatic optimization is an important step to improve computational efficiency and accuracy. The following is a detailed description of each optimization step: Node position adjustment aims to improve the shape and distribution of triangles. The specific steps include: Calculate triangle side lengths and angles: Analyze the side lengths and angles of each triangle to identify unqualified triangles. Common criteria include: Side length ratio: Calculate the ratio of the longest side to the shortest side of a triangle. If it exceeds the set threshold, the triangle is considered unqualified.

[0098] Angle conditions: Make sure all angles are within a certain range (e.g. no less than 30 degrees) to avoid overly distorted triangles.

[0099] Fine-tune vertex positions: For identified unqualified triangles, adjust the positions of their vertices. The adjustment usually uses optimization algorithms such as the least squares method to achieve local optimization and ensure that the geometric quality of the triangle is improved.

[0100] Local smoothing aims to remove noise and irregularities in the mesh. The main steps include: Apply smoothing algorithm: You can use the local average method to average the elevation value of each node to reduce the change of local elevation. For example, calculate the average elevation of nodes within a certain range around each node, and update the elevation value of the node to the average value.

[0101] Iterative Smoothing: You can choose to perform smoothing for multiple iterations until the desired smoothing effect is achieved, but be careful to avoid over-smoothing to avoid losing important terrain details.

[0102] The goal of triangle quality optimization is to ensure uniform distribution of triangles and avoid excessive elongation or distortion. The specific method is as follows: Apply Delaunay triangulation: This helps improve the geometric quality of the triangles by recalculating the positions of the mesh nodes to ensure that the resulting triangles satisfy the Delaunay condition (i.e. no node lies within the circumcircle of any triangle).

[0103] Inspection and optimization: For unqualified triangles, node rearrangement is performed to select appropriate points to generate better triangles. This process may involve multiple iterations until the quality requirements of the triangles are met.

[0104] Reducing the number of redundant nodes and triangles is a key step in simplifying the mesh. Specific methods include: Set Simplification Threshold: Define a simplification threshold to determine which triangles or nodes are redundant. The threshold can be based on the area of ​​the triangle, the distance to surrounding nodes, etc.

[0105] Apply mesh simplification techniques: Edge Collapse: Select an edge and collapse it to its adjacent nodes to form a new node. This method can effectively reduce the number of triangles and simplify the mesh structure.

[0106] Threshold method: Remove triangles whose area is smaller than the set threshold. This step ensures that the amount of calculation is reduced while maintaining the accuracy of the model.

[0107] Merge adjacent edges: Reduce the number of triangles by merging adjacent edges to maintain the overall terrain characteristics.

[0108] Through the automatic optimization of the above steps, the TIN grid can be significantly improved, which not only improves the calculation efficiency, but also ensures the accuracy and details of the terrain model. This optimization process is the basis for generating high-quality digital elevation models, providing reliable support for subsequent terrain analysis and applications.

[0109] S160: Input the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid.

[0110] In this embodiment, the final TIN network refers to a TIN network in which each region has an optimal grid density.

[0111] Specifically, the grid optimization model is a model obtained by generating a TIN grid through cleaning and preprocessing, feature extraction, area division and cluster analysis of the acquired historical terrain data, and performing preliminary optimization, dynamic grid refinement and grid shape optimization on the generated TIN grid, as well as using the manually assisted adjusted TIN grid as a sample set to train convolutional neural networks and graph neural networks.

[0112] The input of the grid optimization model is the TIN network after preliminary optimization, and the output is the optimal TIN grid after dynamic adjustment.

[0113] In this embodiment, the above-mentioned dynamic mesh refinement and mesh shape optimization and manual auxiliary adjustment include the following: Dynamic Mesh Refinement: Mesh density adjustment: Dynamically adjust the density of the mesh based on terrain features (such as slope, curvature, and change range). Increase the mesh density in areas with significant terrain changes, such as ridges or cliffs, to capture more details. Reduce the density in flat areas to improve computational efficiency. By setting a threshold, ensure that sufficient details are maintained in key areas while avoiding unnecessary computational overhead, achieving a balance between accuracy and efficiency.

[0114] Adaptive refinement: Quadtree refinement algorithm: Recursively subdivides the quadrilaterals in the mesh into four smaller quadrilaterals to adapt to uniform terrain changes and gradually increase details. This method allows the refinement process to be flexible and adapt to different terrain characteristics.

[0115] Triangulation Refinement Algorithm: Adapts to irregular terrain changes by adding new nodes and adjusting the boundaries of triangles. Increases the number of triangles in key areas to improve model accuracy. Automatically adjusts mesh density based on local features or user-defined accuracy requirements to ensure accurate terrain representation.

[0116] Mesh shape optimization includes: Triangle quality optimization: Geometry optimization: Optimize the shape of triangles, reduce the number of long-side and short-side triangles, and avoid overly elongated or distorted triangles. This can be achieved by applying the Delaunay triangulation algorithm, which rearranges the node positions to ensure that the side lengths and angles of the triangles are evenly distributed, thereby improving the stability and computational efficiency of the mesh and avoiding the generation of irregularly shaped triangles.

[0117] Smoothing: Smoothing algorithm: Use algorithms such as Laplace smoothing and Gaussian smoothing to adjust the position of mesh nodes, smooth the terrain surface, and eliminate small-scale bumps and depressions. This processing can improve the visual effect and continuity of the terrain model.

[0118] Local vs. global smoothing: Choose a smoothing strategy based on specific needs. In urban planning, local smoothing is applied to remove detail noise around buildings, while in environmental monitoring, global smoothing is used to maintain the consistency of the overall terrain.

[0119] Optimization goal: Reduce redundant nodes: Improve computational efficiency through mesh simplification techniques (such as threshold method and edge collapse method). These methods remove unnecessary triangles or nodes by setting simplification thresholds, or simplify the mesh structure by merging adjacent edges, thereby reducing computational overhead and improving model accuracy and practicality.

[0120] Dynamic adjustment of TIN grids is the key to generating efficient and accurate digital twin terrain models. By adaptively refining, optimizing grid shapes, and applying smoothing, the model's detail capture, computational efficiency, and visual performance can be effectively improved. The combination of these technologies can meet the needs of different application scenarios and ensure the effectiveness of the model in various environments.

[0121] The above content can be automatically executed and can also be adjusted with manual assistance.

[0122] Specifically, the training process of the grid optimization model is as follows: In model training, it is crucial to optimize the construction of the dataset. First, the dynamically adjusted TIN grid is used to generate optimized data to ensure that the dataset covers a variety of terrain features, such as ridges, cliffs, and plains, to improve the generalization ability of the model. This requires that the data is not only high-quality, but also diverse in order to effectively capture the details of different terrain patterns.

[0123] Next, comprehensive data preprocessing is performed, including data cleaning, denoising, and filling missing values ​​to remove abnormal data and ensure data integrity. At the same time, through standardization and data scaling, the data is kept on the same scale to improve the performance of the machine learning algorithm. Feature engineering is also very important. By extracting and selecting key features, the data set can be further optimized to make it more suitable for model training.

[0124] When selecting a model, it is necessary to consider the characteristics of the optimized TIN grid data. Considering the complexity of the data and the expected output type, convolutional neural networks (CNNs) are suitable for processing data with regular grid structures, while graph neural networks (GNNs) are good at capturing complex terrain relationships.

[0125] After selecting a suitable model, it is necessary to design a suitable data processing architecture, including the selection of network hierarchy, activation function, and loss function, to ensure that the model can accurately learn and predict terrain features. At the same time, attention should be paid to the computing resource requirements of the model to ensure that it runs efficiently while ensuring accuracy.

[0126] During model training, key training parameters such as learning rate, batch size, and training rounds need to be set. Choosing a suitable optimization algorithm (such as Adam or SGD) can improve training efficiency. In addition, introducing regularization techniques (such as Dropout and L2 regularization) can prevent overfitting.

[0127] After the model training is completed, comprehensive model evaluation and tuning are required. The performance of the model on different data subsets is evaluated through cross-validation, and the accuracy of the model is measured using performance indicators (such as precision, recall, etc.). Hyperparameters are tuned based on the evaluation results.

[0128] In the model deployment phase, the trained model is integrated into the actual application environment to ensure that it can efficiently process real-time or offline data. In this process, attention should be paid to the computing resource requirements of the model and its operating efficiency should be optimized to meet the requirements of real-time or offline applications.

[0129] To ensure the effectiveness of the model, it is necessary to establish a real-time update mechanism. This includes regularly updating the model to adapt to new data or environmental changes to maintain its predictive ability and analytical effect.

[0130] During the model training phase, a validation set needs to be separated from the training data to evaluate the model performance. The stability and generalization ability of the model can be further verified through the cross-validation method. The final performance of the model is evaluated using an independent test set, and appropriate performance indicators are selected to fully reflect the performance of the model.

[0131] In the validation phase, the model is applied in a simulated environment to test its performance under real conditions. A feedback mechanism is established to collect performance data from the actual application and make necessary improvements based on the feedback.

[0132] In the actual application stage, the trained model is integrated into the target application system to achieve efficient data processing and prediction functions. Through integrated applications, the computing performance of the model is optimized to meet actual operational needs. At the same time, the effect is monitored to continuously track the accuracy and stability of the model.

[0133] In order to maintain the long-term applicability of the model, it is necessary to regularly update the model based on new data or environmental changes, collect user feedback, and continuously optimize model performance to address problems encountered in actual use.

[0134] Through these steps, the efficiency and accuracy of the model in practical applications can be ensured.

[0135] First, a validation set is split from the training data to evaluate model performance and perform hyperparameter tuning. The validation set should be similar to the data distribution and characteristics of the actual application scenario to ensure the effectiveness of the evaluation. Implementing k-fold cross-validation can further improve the stability and generalization ability of the model by dividing the data into multiple subsets to reduce evaluation bias and obtain more reliable results.

[0136] At the end of model training, use an independent test set for performance evaluation. The test set should contain unseen data to objectively evaluate the generalization ability of the model. Choose appropriate performance indicators such as accuracy, precision, recall, and F1 score to fully reflect the performance of the model in practical applications.

[0137] The model needs to be simulated in actual application scenarios to verify its effect under real operating conditions. The model's applicability and practicality can be tested by simulating the data flow in the real environment. At the same time, a feedback loop mechanism is established to collect actual performance data and optimize the model based on the feedback. This helps to adjust the model parameters and improve its performance and adaptability.

[0138] S170. Output the final TIN grid.

[0139] The TIN grid is adjusted through machine learning to form the final TIN grid, which is then output to different task systems for the execution of corresponding tasks.

[0140] The method of this embodiment can achieve higher modeling accuracy and computational efficiency by adaptively adjusting the mesh density. Increasing the mesh density in areas where details are required and reducing the density in flat areas helps to generate a more realistic and effective three-dimensional terrain model. Specifically, combining machine learning and cluster analysis can effectively identify complex terrain features that require high-density meshes, thereby optimizing the performance of the overall model.

[0141] The above-mentioned adaptive TIN digital twin terrain grid density adjustment method based on machine learning obtains terrain data, cleans and preprocesses it to extract relevant features and perform regional division and cluster analysis, generates a TIN grid based on the processing results, and inputs it into the grid optimization model to adjust the optimal grid density in different areas. Finally, the optimized TIN grid is output; the grid density is increased where details are required, and appropriately reduced in flat areas, thereby improving the modeling accuracy, optimizing the calculation efficiency, and making the three-dimensional terrain model more realistic and effective.

[0142] Figure 6 is a schematic block diagram of an adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning provided by an embodiment of the present invention. Figure 6As shown, corresponding to the above-mentioned adaptive TIN digital twin terrain grid density adjustment method based on machine learning, the present invention also provides an adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning. The adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning includes a unit for executing the above-mentioned adaptive TIN digital twin terrain grid density adjustment method based on machine learning, and the device can be configured in a server. Specifically, please refer to Figure 6 The adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning includes a terrain data acquisition unit 301, a preprocessing unit 302, a feature extraction unit 303, a division analysis unit 304, a grid generation unit 305, an optimization unit 306 and an output unit 307.

[0143] A terrain data acquisition unit 301 is used to acquire terrain data; a preprocessing unit 302 is used to clean and preprocess the terrain data to obtain a preprocessing result; a feature extraction unit 303 is used to extract features corresponding to the preprocessing result; a division analysis unit 304 is used to perform regional division and cluster analysis on the features to obtain a processing result; a grid generation unit 305 is used to generate a TIN grid according to the processing result; an optimization unit 306 is used to input the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; and an output unit 307 is used to output the final TIN grid.

[0144] In one embodiment, the pre-processing unit 302 includes a first processing sub-unit, a second processing sub-unit and a third processing sub-unit.

[0145] The first processing subunit is used to remove noise and balance accuracy of the terrain data to obtain a first processing result; the second processing subunit is used to perform data difference and oversampling processing on the first processing result to obtain a second processing result; the third processing subunit is used to standardize and optimize the second processing result to obtain a preprocessing result.

[0146] In one embodiment, the feature extraction unit is used to extract terrain features and geometric features from the preprocessing results.

[0147] In one embodiment, the partitioning and analyzing unit 304 includes: The first division subunit is used to extract contour lines from the terrain data and divide the area by a set elevation threshold to obtain different elevation bands; the second division subunit is used to divide the area based on the terrain characteristics from different slopes, slope directions and terrain classification results to obtain mountain ranges and river valley areas; the cluster analysis subunit is used to perform cluster analysis on different elevation bands and mountain ranges and river valley areas to obtain analysis results.

[0148] In one embodiment, the grid generation unit 305 includes: The model building subunit is used to build a digital elevation model according to the terrain data; the generation subunit is used to generate a TIN grid using the Delaunay triangulation algorithm according to the digital elevation model and the analysis results; the preliminary optimization subunit is used to perform preliminary optimization on the generated TIN grid.

[0149] In one embodiment, the preliminary optimization subunit is used to adjust the node positions, perform local smoothing, optimize the triangle quality, and reduce the number of redundant nodes and triangles on the generated TIN mesh.

[0150] It should be noted that technicians in the relevant field can clearly understand that the specific implementation process of the above-mentioned adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning and each unit can refer to the corresponding description in the aforementioned method embodiment. For the convenience and conciseness of the description, it will not be repeated here.

[0151] The above-mentioned adaptive TIN digital twin terrain grid density adjustment device 300 based on machine learning can be implemented in the form of a computer program. The computer program can be used in Figure 7 Runs on the computer device shown.

[0152] See also Figure 7 , Figure 7 5 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device 500 may be a server, wherein the server may be an independent server or a server cluster composed of multiple servers.

[0153] See also Figure 7 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0154] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions, which, when executed, can enable the processor 502 to execute a method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning.

[0155] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0156] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an adaptive TIN digital twin terrain grid density adjustment method based on machine learning.

[0157] The network interface 505 is used to communicate with other devices over the network. Figure 7 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device 500 to which the solution of the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0158] The processor 502 is used to run the computer program 5032 stored in the memory to implement the following steps: Acquire terrain data; clean and preprocess the terrain data to obtain preprocessing results; extract features corresponding to the preprocessing results; perform regional division and cluster analysis on the features to obtain processing results; generate a TIN grid according to the processing results; input the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; and output the final TIN grid.

[0159] Among them, the grid optimization model is a model obtained by generating a TIN grid through cleaning and preprocessing, feature extraction, regional division and cluster analysis of the acquired historical terrain data, and performing preliminary optimization, dynamic grid refinement and grid shape optimization on the generated TIN grid, and using the manually assisted adjusted TIN grid as a sample set to train convolutional neural networks and graph neural networks.

[0160] In one embodiment, when the processor 502 implements the step of cleaning and preprocessing the terrain data to obtain a preprocessing result, the processor 502 specifically implements the following steps: The terrain data is subjected to noise removal and precision balancing to obtain a first processing result; the first processing result is subjected to data difference and oversampling processing to obtain a second processing result; the second processing result is standardized and optimized to obtain a preprocessing result.

[0161] In one embodiment, when the processor 502 implements the step of extracting the feature corresponding to the preprocessing result, the processor 502 specifically implements the following steps: The terrain features and geometric features in the preprocessing result are extracted.

[0162] In one embodiment, when the processor 502 implements the step of performing region division and cluster analysis on the features to obtain processing results, the processor 502 specifically implements the following steps: Contour lines are extracted from the terrain data, and regions are divided by a set elevation threshold to obtain different elevation zones; regions are divided from different slopes, slope directions and terrain classification results based on the terrain characteristics to obtain mountain ranges and river valley regions; cluster analysis is performed on different elevation zones and mountain ranges and river valley regions to obtain analysis results.

[0163] In one embodiment, when the processor 502 implements the step of generating a TIN grid according to the processing result, the processor 502 specifically implements the following steps: A digital elevation model is constructed according to the terrain data; a TIN grid is generated using a Delaunay triangulation algorithm according to the digital elevation model and the analysis results; and the generated TIN grid is preliminarily optimized.

[0164] In one embodiment, when the processor 502 implements the step of performing preliminary optimization on the generated TIN mesh, the processor 502 specifically implements the following steps: The generated TIN mesh is subjected to node position adjustment, local smoothing, triangle quality optimization, and reduction of redundant nodes and triangle numbers.

[0165] It should be understood that in the embodiment of the present application, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0166] It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0167] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein when the computer program is executed by a processor, the processor executes the following steps: Acquire terrain data; clean and preprocess the terrain data to obtain preprocessing results; extract features corresponding to the preprocessing results; perform regional division and cluster analysis on the features to obtain processing results; generate a TIN grid according to the processing results; input the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; and output the final TIN grid.

[0168] Among them, the grid optimization model is a model obtained by generating a TIN grid through cleaning and preprocessing, feature extraction, regional division and cluster analysis of the acquired historical terrain data, and performing preliminary optimization, dynamic grid refinement and grid shape optimization on the generated TIN grid, and using the manually assisted adjusted TIN grid as a sample set to train convolutional neural networks and graph neural networks.

[0169] In one embodiment, when the processor executes the computer program to implement the step of cleaning and preprocessing the terrain data to obtain a preprocessing result, the processor specifically implements the following steps: The terrain data is subjected to noise removal and precision balancing to obtain a first processing result; the first processing result is subjected to data difference and oversampling processing to obtain a second processing result; the second processing result is standardized and optimized to obtain a preprocessing result.

[0170] In one embodiment, when the processor executes the computer program to implement the step of extracting the features corresponding to the preprocessing result, the processor specifically implements the following steps: The terrain features and geometric features in the preprocessing result are extracted.

[0171] In one embodiment, when the processor executes the computer program to implement the step of performing region division and cluster analysis on the features to obtain processing results, the processor specifically implements the following steps: Contour lines are extracted from the terrain data, and regions are divided by a set elevation threshold to obtain different elevation zones; regions are divided from different slopes, slope directions and terrain classification results based on the terrain characteristics to obtain mountain ranges and river valley regions; cluster analysis is performed on different elevation zones and mountain ranges and river valley regions to obtain analysis results.

[0172] In one embodiment, when the processor executes the computer program to implement the step of generating a TIN grid according to the processing result, the processor specifically implements the following steps: A digital elevation model is constructed according to the terrain data; a TIN grid is generated using a Delaunay triangulation algorithm according to the digital elevation model and the analysis results; and the generated TIN grid is preliminarily optimized.

[0173] In one embodiment, when the processor executes the computer program to implement the step of performing preliminary optimization on the generated TIN grid, the processor specifically implements the following steps: The generated TIN mesh is subjected to node position adjustment, local smoothing, triangle quality optimization, and reduction of redundant nodes and triangle numbers.

[0174] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0175] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0176] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0177] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0178] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0179] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An adaptive TIN digital twin terrain grid density adjustment method based on machine learning, characterized in that: include: Get terrain data; Cleaning and preprocessing the terrain data to obtain a preprocessing result; Extracting features corresponding to the preprocessing results; Performing regional division and cluster analysis on the features to obtain processing results; Generate a TIN grid according to the processing result; Inputting the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; Output the final TIN grid.

2. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to claim 1 is characterized in that: The cleaning and preprocessing of the terrain data to obtain a preprocessing result includes: Performing noise removal and precision balancing on the terrain data to obtain a first processing result; Performing data difference and oversampling processing on the first processing result to obtain a second processing result; The second processing result is standardized and optimized to obtain a preprocessing result.

3. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to claim 1 is characterized in that: The extracting the features corresponding to the preprocessing results includes: The terrain features and geometric features in the preprocessing result are extracted.

4. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to claim 3 is characterized in that: The performing of regional division and cluster analysis on the features to obtain processing results includes: Extracting contour lines from the terrain data and dividing the area by a set elevation threshold to obtain different elevation zones; Based on the terrain features, the regions are divided from different slopes, slope aspects and terrain classification results to obtain mountain ranges and river valley regions; Cluster analysis is performed on different elevation zones as well as mountain ranges and river valley areas to obtain the analysis results.

5. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to claim 4 is characterized in that: Generating a TIN grid according to the processing result includes: constructing a digital elevation model based on the terrain data; Generate a TIN grid using a Delaunay triangulation algorithm according to the digital elevation model and the analysis results; Perform preliminary optimization on the generated TIN mesh.

6. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to claim 5 is characterized in that: The preliminary optimization of the generated TIN grid includes: The generated TIN mesh is subjected to node position adjustment, local smoothing, triangle quality optimization, and reduction of redundant nodes and triangle numbers.

7. The method for adjusting the grid density of an adaptive TIN digital twin terrain based on machine learning according to any one of claims 1 to 6, characterized in that: The grid optimization model is a model obtained by generating a TIN grid through cleaning and preprocessing, feature extraction, regional division and cluster analysis of the acquired historical terrain data, and performing preliminary optimization, dynamic grid refinement and grid shape optimization on the generated TIN grid, and using the TIN grid after manual auxiliary adjustment as a sample set to train a convolutional neural network and a graph neural network.

8. An adaptive TIN digital twin terrain grid density adjustment device based on machine learning, characterized in that: include: A terrain data acquisition unit, used for acquiring terrain data; A preprocessing unit, used for cleaning and preprocessing the terrain data to obtain a preprocessing result; A feature extraction unit, used to extract features corresponding to the preprocessing result; A division and analysis unit, used for performing regional division and cluster analysis on the features to obtain processing results; A grid generation unit, used for generating a TIN grid according to the processing result; An optimization unit, used for inputting the TIN grid into a grid optimization model to adjust the optimal grid density of different regions to obtain a final TIN grid; Output unit, used to output the final TIN grid.

9. A computer device, characterized in that: The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images