A three-dimensional ground model data generation and rapid reading method, medium and system

Abstract

The application provides a three-dimensional surface model data generation and rapid reading method, medium and system, belongs to the three-dimensional surface model data processing technical field, first extracts the surface features through wavelet transform and singular value decomposition, then carries out adaptive partitioning based on spatial correlation analysis, and realizes the accurate expression of data by using multi-scale grid division and morphological feature extraction. On this basis, hierarchical compression storage and quadtree index structure are used to organize data, the importance of the data block is determined through topological correlation analysis and viewpoint correlation evaluation, and finally the rapid data reading based on the progressive loading strategy is realized. The method can ensure the data integrity while significantly improving the data processing and display efficiency. The technical problems that the existing technology is difficult to realize the rapid reading and efficient rendering of massive three-dimensional surface model data while ensuring the data integrity and display effect are solved.

Description

Technical Field

[0001] This invention belongs to the field of three-dimensional surface model data processing technology, and specifically relates to a method, medium and system for generating and rapidly reading three-dimensional surface model data. Background Technology

[0002] 3D land surface model data has wide applications in geographic information systems, urban planning, and military simulation. Traditional methods for processing 3D land surface model data mainly include regular grid methods, irregular triangular mesh methods, and hybrid grid methods. Regular grid methods use a uniform grid size to divide the land surface, which is simple to implement but difficult to adapt to complex terrains; irregular triangular mesh methods can adaptively divide the surface according to terrain features, but have high computational complexity; hybrid grid methods combine the advantages of both methods, but still have efficiency issues in data organization and management. With the development of remote sensing technology, the amount of 3D land surface model data is growing exponentially. Existing technologies face several challenges when processing massive amounts of data: First, data compression methods often use a uniform compression ratio, making it difficult to balance data volume and accuracy; second, traditional data organization structures (such as octrees, R-trees, etc.) suffer from low retrieval efficiency when processing data at different scales; third, existing data loading strategies mostly use fixed-size data blocks, which are difficult to adapt to dynamically changing display requirements. These problems often lead to slow loading, display lag, or excessive memory consumption of 3D land surface models in practical applications. Especially when processing high-resolution, large-scale surface data, how to achieve fast data reading and smooth display under limited computing and storage resources has become a core technical problem that urgently needs to be solved.

[0003] In summary, existing technologies present the technical challenge of achieving rapid reading and smooth display of massive 3D surface model data under limited computing and storage resources. Summary of the Invention

[0004] In view of this, the present invention provides a method, medium, and system for generating and rapidly reading three-dimensional surface model data, which can solve the technical problem of how to achieve rapid reading and smooth display of massive three-dimensional surface model data under limited computing and storage resources.

[0005] The present invention is implemented as follows: The first aspect of the present invention provides a method for generating and rapidly reading three-dimensional surface model data, comprising the following steps: acquiring raw surface elevation data and performing wavelet multi-scale decomposition to obtain multi-frequency surface feature data; performing singular value decomposition on the multi-frequency surface feature data to obtain a surface feature vector matrix; constructing a surface spatial correlation matrix using the surface feature vector matrix and performing regional division; generating multi-scale grid data based on the regional division results; constructing a quadtree structure to store the multi-scale grid data; calculating the topological correlation matrix between nodes in the quadtree structure; constructing a data reading evaluation function based on the topological correlation matrix; and employing a progressive loading strategy to achieve rapid data reading.

[0006] The steps of acquiring the original surface elevation data and performing wavelet multi-scale decomposition specifically include: performing a data integrity check on the original surface elevation data and removing outliers; performing a 4-level decomposition on the original surface elevation data using Haar wavelet transform; normalizing the decomposed multi-frequency surface feature data; and evaluating the multi-frequency surface feature data using a surface texture evaluation function.

[0007] The steps for performing singular value decomposition on multi-frequency surface feature data specifically include: constructing a covariance matrix; performing eigenvalue decomposition on the covariance matrix using the Jacobi iteration method; selecting eigenvectors with a cumulative contribution rate of 90% to form a surface feature vector matrix; and evaluating the surface feature vector matrix using a feature importance evaluation function.

[0008] The steps of constructing the surface spatial correlation matrix and dividing the region specifically include: constructing the surface spatial correlation matrix using the Moran index method; calculating the spatial autocorrelation coefficient of the local region using the sliding window method; optimizing the spatial autocorrelation coefficient using the spatial correlation optimization function; dividing the region using the spectral clustering algorithm; and constructing the surface partition weight matrix.

[0009] The steps for generating multi-scale grid data specifically include: dividing each partition into grids using an adaptive quadtree partitioning strategy; calculating the surface feature complexity of each grid cell; generating multi-scale grid data of the surface through recursive subdivision; and evaluating the multi-scale grid data of the surface using a grid structure evaluation function.

[0010] The steps for constructing a quadtree structure to store multi-scale grid data specifically include: calculating the curvature matrix of grid nodes; extracting surface feature lines and feature points to form a surface morphology matrix; calculating the data compression factor; performing hierarchical compression and storage of the multi-scale grid data; and organizing the compressed multi-scale grid data using a quadtree structure. The steps for calculating the topological association matrix between nodes in the quadtree structure specifically include: calculating the spatial relationships between nodes; constructing the topological association matrix; calculating the connection strength using a topological evaluation function; and optimizing the node connection method to ensure the rationality of the topological structure.

[0011] The steps of constructing the data reading evaluation function specifically include: calculating the importance value of each data block; constructing a viewpoint association matrix; performing quantitative analysis on the viewpoint association matrix using the viewpoint importance function; and organizing the data into blocks according to the viewpoint association matrix.

[0012] The steps for achieving fast data reading using a progressive loading strategy include: prioritizing the loading of data blocks with high importance; monitoring the loading process using a loading performance function; preloading data blocks through a preloading mechanism; and dynamically adjusting the batch size of data blocks.

[0013] A second aspect of the present invention provides a computer-readable storage medium storing program instructions, which, when executed in a computer, are used to perform the above-described method for generating and rapidly reading three-dimensional surface model data.

[0014] A third aspect of the present invention provides a system for generating and rapidly reading three-dimensional surface model data, comprising the aforementioned computer-readable storage medium. The system can be any one of a computer, a server, or a microcontroller. The computer-readable storage medium is disposed within the system, and the system is provided with a microprocessor that executes the program instructions stored in the computer-readable storage medium.

[0015] Compared with existing technologies, this invention provides a method, medium, and system for generating and rapidly reading 3D surface model data. The proposed method establishes a complete data processing and management framework through the organic combination of multi-scale decomposition, feature extraction, adaptive partitioning, and progressive loading techniques. This method can automatically adjust the data compression ratio and grid density according to the complexity of surface features, optimizing the utilization of data storage space. In terms of data organization, this invention adopts a multi-level index structure based on quadtrees and introduces spatial correlation analysis and viewpoint correlation evaluation mechanisms, enabling the data retrieval and loading process to be dynamically adjusted according to actual display needs. By constructing topological relationships between data blocks and optimizing data block size, this method significantly improves data reading efficiency and rendering performance. Simultaneously, the introduction of a progressive loading strategy ensures the continuity and smoothness of the display process. This invention successfully solves the core problem in processing massive 3D land surface model data. Through precise extraction and multi-level organization of data features, it achieves an optimized balance between data storage and retrieval efficiency. In other words, this invention solves the technical problem of how to achieve rapid retrieval and smooth display of massive 3D land surface model data under limited computing and storage resources, providing reliable technical support for the real-time display of large-scale land surface data. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention;

[0017] Figure 2 This is a map showing the surface texture distribution characteristics of different landform types in Example 2;

[0018] Figure 3 This is a graph showing the cumulative contribution rate of eigenvalue decomposition in Example 2. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0020] like Figure 1 The diagram shown is a flowchart of a method for generating and quickly reading three-dimensional surface model data according to the first aspect of this invention. This method includes the following steps:

[0021] S01. Obtain the original surface elevation data, perform wavelet multi-scale decomposition on the original surface elevation data to obtain multi-frequency surface feature data, and evaluate the multi-frequency surface feature data using the surface texture evaluation function.

[0022] S02. Perform singular value decomposition on the multi-frequency surface feature data to obtain the surface feature vector matrix, and evaluate the surface feature vector matrix using the feature importance evaluation function.

[0023] S03. Construct a surface spatial correlation matrix using the surface feature vector matrix, calculate the surface spatial correlation coefficient, and normalize the surface spatial correlation coefficient using a spatial correlation optimization function.

[0024] S04. Based on the surface spatial correlation coefficient, perform adaptive zoning of the surface region to obtain the surface zoning weight matrix, and evaluate the surface zoning weight matrix using the regional distribution evaluation function.

[0025] S05. Calculate the grid density of each zone using the aforementioned surface partition weight matrix to generate multi-scale surface grid data, and optimize the multi-scale surface grid data using the grid structure evaluation function.

[0026] S06. Calculate the grid node curvature matrix based on the multi-scale grid data of the land surface, construct the land surface morphology matrix, and use the morphology evaluation function to perform quantitative analysis on the land surface morphology matrix.

[0027] S07. Calculate the data compression factor using the landform matrix, construct a compression optimization function to determine the compression threshold, and perform hierarchical compression storage on the multi-scale grid data of the land surface.

[0028] S08. Construct a quadtree structure, store the compressed multi-scale grid data of the land surface in the quadtree structure, and evaluate the retrieval efficiency of the quadtree structure using an index performance function.

[0029] S09. Calculate the topological association matrix between nodes in the quadtree structure, and use the topological evaluation function to calculate the connection strength of the topological association matrix.

[0030] S10. Construct a data reading evaluation function based on the topological correlation matrix. The data reading evaluation function is used to calculate the importance value of the data block.

[0031] S11. Calculate the viewpoint correlation matrix using the data reading evaluation function, and perform quantitative analysis on the viewpoint correlation matrix using the viewpoint importance function;

[0032] S12. Organize the data into blocks according to the viewpoint association matrix, and optimize the data block size using a storage evaluation function;

[0033] S13. The data blocks are read using a progressive loading strategy, and the progressive loading strategy is dynamically adjusted using a loading performance function.

[0034] The specific implementation methods of the above steps are described in detail below. The specific implementation method of step S01 is as follows: First, the acquired raw surface elevation data undergoes a data integrity check to remove outliers. The outlier criterion is that the deviation from surrounding data exceeds three times the standard deviation. Next, the raw surface elevation data is decomposed into multiple scales using Haar wavelet transform. The decomposition level is set to four levels, and each level decomposes the corresponding high-frequency and low-frequency components. During the decomposition process, orthogonal basis functions are introduced to ensure the completeness and orthogonality of the decomposition results. Then, the decomposed multi-frequency surface feature data is normalized to ensure that the data distribution falls within the range of 0 to 1. Finally, the processing results are evaluated using a surface texture evaluation function. This evaluation function measures the complexity of the surface texture by calculating the variance and entropy of pixel values ​​within a local area; a higher evaluation value indicates a more complex texture. The purpose of this step is to extract multi-scale features of the surface, laying the foundation for subsequent processing.

[0035] The specific implementation of step S02 is as follows: A covariance matrix is ​​constructed from the multi-frequency surface feature data, the size of which is determined by the data dimensionality. Then, the Jacobi iteration method is used to decompose the covariance matrix into eigenvalues, with an iteration convergence threshold of 0.0001. Based on the decomposed eigenvalues, they are arranged in descending order, and eigenvectors with a cumulative contribution rate reaching 90% are selected to form the surface feature vector matrix. The feature vector matrix is ​​evaluated using a feature importance evaluation function, which comprehensively considers the magnitude of the eigenvalues ​​and the spatial distribution characteristics of the eigenvectors; eigenvectors with an evaluation value greater than 0.8 are retained. The purpose of this step is to reduce the data dimensionality and extract the main feature information.

[0036] The specific implementation of step S03 is as follows: Based on the surface feature vector matrix, a surface spatial correlation matrix is ​​constructed using the Moran index method, where matrix elements represent the degree of correlation between different locations. The spatial autocorrelation coefficient of a local area is calculated using the sliding window method, with the window size set to one-tenth of the data range. The calculated surface spatial correlation coefficients are processed using the min-max normalization method to ensure that the values ​​are distributed between 0 and 1. Then, the normalized coefficients are optimized using a spatial correlation optimization function, which employs an exponential decay model to describe the distance decay characteristics of spatial correlation strength, with a decay coefficient set to 0.3. The purpose of this step is to quantify the spatial distribution patterns of surface features.

[0037] The specific implementation of step S04 is as follows: Based on the normalized surface spatial correlation coefficient, a spectral clustering algorithm is used to divide the region, and the number of clusters is adaptively determined by the silhouette coefficient. The centroid location and boundary features of each cluster region are calculated to construct a surface partition weight matrix, where the matrix elements represent the relative importance of each region. The partitioning results are evaluated using a regional distribution evaluation function, which considers two indicators: regional area balance and boundary smoothness. An evaluation value greater than 0.7 indicates that the partitioning results are reasonable. The partition boundaries are adjusted through iterative optimization until the evaluation value meets the requirements. The purpose of this step is to achieve a reasonable division of surface regions.

[0038] The specific implementation of step S05 is as follows: Based on the surface partition weight matrix, an adaptive quadtree partitioning strategy is used to divide each partition into grids, with the initial grid size set to 4 times the data resolution. The surface feature complexity is calculated for each grid cell, and areas with a complexity exceeding a threshold of 0.6 are further subdivided. Multi-scale surface grid data is generated through recursive subdivision, with a maximum subdivision layer of 6 layers. The grid quality is evaluated using a grid structure evaluation function, which comprehensively considers grid shape regularity and size variability. Grids with an evaluation value greater than 0.8 remain unchanged; otherwise, they are optimized and adjusted. The purpose of this step is to generate a multi-scale grid structure adapted to surface features.

[0039] The specific implementation of step S06 is as follows: For each node in the multi-scale grid data of the land surface, the principal curvature and Gaussian curvature are calculated using the discrete differential geometry method to construct a grid node curvature matrix. Then, based on the curvature information, surface feature lines and feature points are extracted to form a land surface morphology matrix. The feature extraction results are quantitatively analyzed using a morphological evaluation function. This function evaluates the saliency of land surface morphological features by calculating the continuity of feature lines and the distribution density of feature points; an evaluation value greater than 0.75 indicates that the feature extraction is effective. The purpose of this step is to achieve an accurate expression of land surface morphological features.

[0040] The specific implementation of step S07 is as follows: Utilizing the feature information in the landform matrix, an adaptive compression algorithm is used to calculate the data compression factor, with the factor ranging from 0.1 to 0.9. A compression optimization function is constructed to determine the optimal compression threshold. This function balances two indicators: compression ratio and reconstruction error. A suitable compression threshold is considered to be achieved when the reconstruction error is less than 0.05 and the compression ratio is greater than 80%. Based on the determined compression threshold, the multi-scale grid data of the land surface is compressed and stored in layers, with low compression ratios used for important feature areas and high compression ratios used for flat areas. The purpose of this step is to achieve efficient compressed storage of the data.

[0041] The specific implementation of step S08 is as follows: The compressed multi-scale grid data of the Earth's surface is organized using a quadtree structure, with the tree depth determined by the grid hierarchy. During construction, the root node range is first determined, and then child nodes are recursively divided until the finest level is reached. A storage location index and adjacency information are appended to each node. The retrieval efficiency of the quadtree structure is evaluated using an index performance function, which calculates the average retrieval time and memory usage. A reasonable index structure is indicated when the retrieval time is less than 10 milliseconds and the memory usage is less than a preset value. The purpose of this step is to establish an efficient data management structure.

[0042] The specific implementation of step S09 is as follows: Based on the quadtree structure, the spatial relationships between nodes are calculated, and a topological association matrix is ​​constructed. Matrix elements represent the connection and hierarchical relationships between nodes. Graph theory methods are used to analyze the connectivity and reachability of nodes. A topological evaluation function is used to calculate the connection strength, which considers the distance and hierarchical differences between nodes. Node pairs with a connection strength greater than 0.6 maintain their topological relationship. By optimizing the node connection method, the rationality of the topological structure is ensured. The purpose of this step is to establish the association relationships between data blocks.

[0043] The specific implementation of step S10 is as follows: A data reading evaluation function is constructed based on the topological association matrix. This function comprehensively considers the size, hierarchy, and access frequency of data blocks. An importance value is calculated for each data block using a weighted summation method, with the weighting coefficients dynamically adjusted based on historical access records. When the importance value of a data block is greater than 0.7, it is marked as a priority loading object. The purpose of this step is to achieve a reasonable allocation of data block loading priorities.

[0044] The specific implementation of step S11 is as follows: The correlation between the viewpoint position and each data block is calculated using a data reading evaluation function, constructing a viewpoint correlation matrix. The correlation calculation considers three factors: viewpoint distance, viewing direction, and occlusion relationship. The correlation matrix is ​​then quantitatively analyzed using a viewpoint importance function, which is based on the frustum clipping principle and calculates the contribution of each data block to the visual effect. When the contribution is greater than 0.65, the corresponding data block is included in the rendering sequence. The purpose of this step is to optimize the loading order of data blocks.

[0045] The specific implementation of step S12 is as follows: Data is organized into blocks based on the viewpoint association matrix, with the size of each data block dynamically determined according to a storage evaluation function. This function comprehensively considers data access efficiency and storage space utilization; optimal storage efficiency is achieved when the data block size is between 32 kilobytes and 256 kilobytes. An index table is then created for the organized data blocks, recording their location and attribute information. The purpose of this step is to achieve reasonable block storage of data.

[0046] The specific implementation of step S13 is as follows: A progressive loading strategy is adopted to achieve fast data reading. First, high-importance data blocks are loaded, and then less important data blocks are loaded gradually. A loading performance function is used to monitor and adjust the loading process. This function records the loading time and display effect of data blocks. When the loading time exceeds 50 milliseconds, the batch size is automatically adjusted. A preloading mechanism is used to preload potentially needed data blocks, with the preloaded quantity not exceeding 30% of the currently displayed data volume. The purpose of this step is to ensure the real-time and continuous nature of data loading.

[0047] The functions or calculation processes involved in this invention will be described in detail below.

[0048] 1. The specific expression of the surface texture evaluation function is as follows:

[0049] ;

[0050] In the formula, For surface texture; These are elements of the spatial weight matrix; The elevation values ​​of adjacent grid points; This represents the average elevation of a local area. This represents the number of grid points in a local area. This represents the standard deviation of elevation in a local area. This represents the local area's elevation entropy value. These are the weighting coefficients.

[0051] 2. The feature importance evaluation function is specifically expressed as follows:

[0052] ;

[0053] In the formula, For feature importance; For the first One eigenvalue; The distance between feature vectors in the space; These are the parameters of the Gaussian kernel; Correlation in the feature space; The characteristic clustering degree; These are the weighting coefficients.

[0054] 3. The spatial correlation optimization function is specifically expressed as follows:

[0055] ;

[0056] In the formula, Spatial correlation coefficient; Spatial weights; For spatial location The attribute value at the location; The average value of the attribute; This represents the number of sampling points; This is the spatial distance attenuation term; These are the weighting coefficients.

[0057] 4. The specific expression of the regional distribution evaluation function is as follows:

[0058] ;

[0059] In the formula, For regional distribution uniformity; For the first Area of ​​each region; The average area of ​​the region; Total number of regions; For the smoothness of the region boundary; For the regularity of the region shape; These are the weighting coefficients.

[0060] 5. The specific expression of the grid structure evaluation function is as follows:

[0061] ;

[0062] In the formula, For mesh quality; Area of ​​the grid cell; The lengths of the three sides of the grid cell; For grid gradient; Uniformity of grid vertex distribution; These are the weighting coefficients.

[0063] 6. The morphological evaluation function is specifically expressed as follows:

[0064] ;

[0065] In the formula, These are morphological characteristic values; Principal curvature; The density of feature points; For the continuity of characteristic lines; These are the weighting coefficients.

[0066] 7. The compression optimization function is specifically expressed as follows:

[0067] ;

[0068] In the formula, For compression efficiency; This represents the original data size. This refers to the size of the compressed data. This is for reconstruction error; For data redundancy; These are the weighting coefficients.

[0069] 8. The viewpoint importance function is specifically expressed as follows:

[0070] ;

[0071] In the formula, Importance of viewpoint; The angle between the lines of sight; Viewpoint distance; For occlusion degree; Visual attention value; These are the weighting coefficients.

[0072] 9. The index performance function is specifically expressed as follows:

[0073] ;

[0074] In the formula, This is an index performance value; This represents the average retrieval time. Total memory usage; Memory already used; Tree height variation; These are the weighting coefficients.

[0075] 10. The topology evaluation function is specifically expressed as follows:

[0076] ;

[0077] In the formula, For topology connection strength; Node connectivity; The distance between nodes; For hierarchical connectivity; The node degree distribution; These are the weighting coefficients.

[0078] 11. The data reading evaluation function is specifically expressed as follows:

[0079] ;

[0080] In the formula, Prioritize data block reads; This refers to the data block size. For data block hierarchy; Access frequency; These are the weighting coefficients; This is a random disturbance term.

[0081] 12. The storage evaluation function is specifically expressed as follows:

[0082] ;

[0083] In the formula, For storage efficiency; This refers to the data block size. This represents the average access time. For space utilization; For data dispersion; These are the weighting coefficients.

[0084] 13. The loading performance function is specifically represented as follows:

[0085] ;

[0086] In the formula, To load performance values; This represents the number of data blocks that have been loaded. Loading time; To display quality; Memory usage; These are the weighting coefficients.

[0087] The specific representation of the surface feature vector matrix is ​​as follows:

[0088] ;

[0089] In the formula, For the first The first sample point One eigenvalue; This represents the number of sample points. For feature dimensions.

[0090] The specific representation of the surface spatial correlation matrix is ​​as follows:

[0091] ;

[0092] In the formula, For position With position Spatial correlation coefficient; This represents the number of spatial locations.

[0093] The topological correlation matrix is ​​specifically represented as follows:

[0094] ;

[0095] In the formula, For nodes With nodes The strength of topological association; This represents the total number of nodes.

[0096] The parameters for each function are obtained as follows: spatial weight matrix elements The inverse distance weighted method is used for calculation, specifically involving the following steps: calculating the distance between spatial point pairs, constructing a distance matrix, and normalizing the matrix; elevation values... Eigenvalues ​​are obtained directly from the raw data. Gaussian kernel parameters are obtained through singular value decomposition. The values ​​were determined through cross-validation and range from 0.1 to 1.0; the weight coefficients were all optimized using a grid search method and range from 0 to 1; the curvature values... Calculated using discrete differential geometry methods; viewpoint parameters are calculated based on viewpoint position and orientation, distance... Normalized to the range of 0 to 1; retrieval time The average value was obtained through multiple tests; memory usage was monitored in real time via system calls; node connectivity was also measured. The calculations are performed using graph theory algorithms; data block parameters are obtained through real-time statistics; and weight coefficients are determined through optimization using machine learning methods.

[0097] The principles behind the construction of these functions are as follows: The surface texture evaluation function combines spatial autocorrelation and information entropy, considering local elevation changes and complexity; the feature importance evaluation function combines feature value magnitude and spatial distribution characteristics, using an exponential decay model to describe spatial correlation; the spatial association optimization function is based on an improvement of the Moran index, adding a distance decay term to better describe spatial relationships; the regional distribution evaluation function considers area uniformity and shape regularity, using the squared difference form to measure the degree of non-uniformity; the grid structure evaluation function is based on grid shape quality measurement, combining the area and side length ratio; the morphology evaluation function integrates local curvature features and... The system comprehensively describes the terrain morphology based on global distribution characteristics. A compression optimization function balances compression ratio and reconstruction quality, introducing redundancy to assess data characteristics. A viewpoint importance function employs a distance decay model, comprehensively considering line-of-sight direction and occlusion relationships. An indexing performance function considers retrieval speed and memory efficiency, introducing tree structure balance assessment. A topology evaluation function is based on graph theory, considering node connection density and hierarchical relationships. A data reading evaluation function uses a weighted summation model, introducing random perturbations to increase adaptability. A storage evaluation function balances storage space and access efficiency, considering data distribution characteristics. A loading performance function combines loading speed and display effects to optimize user experience.

[0098] The derivation process for each function is described in detail below.

[0099] 1. The process of deriving and establishing the surface texture evaluation function:

[0100] First, the basic model is constructed using the principle of spatial autocorrelation. This model only considers the spatial distribution characteristics of elevation values; then a standard deviation term is introduced. To describe the degree of data dispersion, an entropy term is introduced. Describing data complexity and obtaining an improved model Weighting coefficient The function is obtained through optimization using the least squares method. It can simultaneously reflect both the local variation characteristics and the overall complexity of the Earth's surface.

[0101] 2. The process of deriving and establishing the feature importance evaluation function:

[0102] Based on eigenvalue decomposition theory, the initial model is: Only the magnitude of the eigenvalues ​​is considered; considering the spatial distribution of the eigenvalues, a Gaussian kernel function is introduced. ,get Further introduce feature correlation and clustering To obtain the final model .parameter It was determined through cross-validation. Optimize through grid search.

[0103] 3. The process of deriving and establishing the spatial correlation optimization function:

[0104] Starting with the Moran index, a basic model is constructed. Introducing a distance attenuation term Taking into account the influence of spatial location, an optimization model is obtained. ;in Calculated using an exponential decay function, Determined by maximum likelihood estimation.

[0105] 4. The process of deriving and establishing the regional distribution evaluation function:

[0106] Initially, area difference was used as a metric. Introducing boundary smoothness and shape regularity To obtain the complete model ;in Calculated by boundary curvature The weighting coefficients are optimized using a simulated annealing algorithm after being calculated based on the roundness index.

[0107] 5. The process of deriving and establishing the evaluation function for the mesh structure:

[0108] Based on grid quality evaluation theory, the initial model is: Introducing grid gradients and vertex distribution uniformity , thus obtaining the optimized model ;in Calculated using the mesh side length ratio Calculated using the area variance of the Thiessen polygon.

[0109] 6. The process of deriving and establishing the morphological evaluation function:

[0110] Starting from differential geometry theory, the initial model is: Introducing feature point density and continuity of characteristic lines To obtain the complete model The principal curvatures were calculated using least-squares fitting. Obtained through kernel density estimation. Evaluation was conducted using curve fitting error assessment.

[0111] 7. The process of deriving and establishing the compression optimization function:

[0112] The basic model uses compression ratio representation Introducing reconstruction error and data redundancy , thus obtaining the optimized model ;in Calculated using root mean square error. The weighting coefficients are optimized using a genetic algorithm based on information entropy evaluation.

[0113] 8. The process of deriving and establishing the viewpoint importance function:

[0114] Initially, a line-of-sight attenuation model was used. Introducing occlusion and visual attention value To obtain the complete model ;in Calculations based on ray tracing Obtained through significance testing.

[0115] 9. The process of deriving and establishing the index performance function:

[0116] The basic model takes retrieval time into account. By incorporating memory utilization and tree structure balance, an optimized model is obtained. The parameters are obtained through performance testing and dynamically adjusted.

[0117] 10. The process of deriving and establishing the topological evaluation function:

[0118] Starting from graph theory, the initial model is: By introducing hierarchical connectivity and node degree distribution, a complete model is obtained. The parameters are determined through graph structure analysis.

[0119] A second aspect of the present invention provides a computer-readable storage medium storing program instructions, which, when executed in a computer, are used to perform the above-described method for generating and rapidly reading three-dimensional surface model data.

[0120] A third aspect of the present invention provides a system for generating and rapidly reading three-dimensional surface model data, comprising the aforementioned computer-readable storage medium. The system can be any one of a computer, a server, or a microcontroller. The computer-readable storage medium is disposed within the system, and the system is provided with a microprocessor that executes the program instructions stored in the computer-readable storage medium.

[0121] Specifically, the principle of this invention is as follows: The technical solution of this invention is based on the principles of data feature analysis and multi-level organization. First, multi-frequency features of the land surface are extracted through wavelet multi-scale decomposition. This method can effectively separate surface details at different scales, providing a theoretical basis for subsequent feature extraction and data compression. The introduction of singular value decomposition further reduces the data dimensionality, retaining key feature information while reducing redundant data. Spatial correlation analysis is another important theoretical basis of this invention. By constructing a spatial correlation matrix and calculating correlation coefficients, the distribution pattern of surface features can be accurately described, providing a basis for regional division. The implementation of adaptive partitioning relies on a spectral clustering algorithm, which can automatically determine the optimal number of partitions and boundary positions based on data features. In terms of data organization, this invention adopts a quadtree structure because of its good spatial partitioning characteristics and hierarchical management capabilities. By establishing a topological correlation matrix between nodes, the spatial and hierarchical relationships between data blocks can be effectively described, providing support for rapid data retrieval. The progressive loading strategy is designed based on viewpoint correlation analysis. By calculating the contribution of data blocks to the display effect, a reasonable allocation of loading priorities is achieved.

[0122] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0123] The specific implementation of step S01 is as follows: First, the acquired raw surface elevation data undergoes a data integrity check. Anomalies are identified by constructing a local elevation statistical model, calculating the elevation difference between each data point and its neighboring points. Anomalies are identified when the difference exceeds three times the local standard deviation. Then, Haar wavelet transform is used to decompose the raw surface elevation data into a multi-scale decomposition. The decomposition process is recursive, and each decomposition yields one low-frequency component and three high-frequency components, corresponding to detailed features in the horizontal, vertical, and diagonal directions, respectively. The decomposed multi-frequency surface feature data is evaluated using a surface texture evaluation function, the expression of which is: Spatial weights Obtained through inverse distance weighted calculation , For spatial point-to-point distance, The distance decay index ranges from 2 to 3; the local elevation average. The calculation is performed using a sliding window, with the window size set to one-tenth of the data range; the standard deviation term... An improved local variance calculation method is adopted, and directional weights are introduced to enhance the ability to extract edge features; the entropy term... The degree of uncertainty in the data is reflected by calculating the elevation histogram using normalized methods. Weighting coefficients. The least squares method is used to fit a large amount of experimental data, generally... The value ranges from 0.3 to 0.5. The value ranges from 0.2 to 0.4. The main purpose of this step is to extract multi-scale features of the land surface, providing a basis for subsequent feature analysis and regional division.

[0124] The specific implementation of step S02 is as follows: A covariance matrix is ​​constructed from the multi-frequency surface feature data, and the matrix elements are calculated using the following formula: ,in Indicates the first The first sample 1 eigenvalue, Indicates the first The mean of each feature, The sample size is given. The Jacobi iterative method is used to perform eigenvalue decomposition on the covariance matrix, and the rotation matrix is ​​used during the iteration process. Off-diagonal elements are eliminated, and the iteration terminates when the sum of squares of the off-diagonal elements is less than a set threshold of 0.0001. The eigenvectors obtained from eigenvalue decomposition are sorted according to their corresponding eigenvalues, and the eigenvectors with a cumulative contribution rate reaching 90% are selected to form the surface feature vector matrix. A feature importance evaluation function is introduced. To evaluate the importance of feature vectors, where For the first 1 eigenvalue, spatial distance The Gaussian kernel parameters are calculated by the Euclidean distance between the eigenvectors in the feature space. Cross-validation determined that the value range is from 0.1 to 1.0, and the correlation in the feature space is... The characteristic clustering degree was calculated using the Pearson correlation coefficient. Obtained through local density estimation. Weighting coefficients. The values ​​are determined through a grid search method, ranging from 0 to 1. The main purpose of this step is to reduce the dimensionality of the data and extract key feature information.

[0125] The specific implementation of step S03 is as follows: Based on the surface feature vector matrix, a surface spatial correlation matrix is ​​constructed using the improved Moran's index method. The expression for the spatial correlation optimization function is: Spatial weights An adaptive kernel function is used for calculation, and the kernel function bandwidth is determined through cross-validation. Distance attenuation term. Using an exponential decay model ,in For spatial distance, The feature correlation length is determined through variogram analysis. Weighting coefficients. Maximum likelihood estimation is used to calculate the value, typically ranging from 0.2 to 0.4. The spatial autocorrelation coefficient for a local area is calculated using the sliding window method, with the window size set to one-tenth of the data range and the window overlap rate set to 50%. The calculated spatial autocorrelation coefficient is then processed using the min-max normalization method to ensure that the values ​​are distributed between 0 and 1. The main purpose of this step is to quantify the spatial distribution pattern of surface features.

[0126] The specific implementation of step S04 is as follows: Based on the normalized spatial autocorrelation coefficient, a spectral clustering algorithm is used to divide the region. First, a similarity matrix is ​​constructed. The formula for calculating matrix elements is: ,in The parameters are Gaussian kernel parameters. Then, the Laplacian matrix is ​​calculated. ,in For a degree matrix, the diagonal elements Solving the generalized eigenvalue problem Before selection A new feature space is constructed from the feature vectors. The mean clustering algorithm performs clustering in the new feature space, and the number of clusters... The region distribution evaluation function is determined adaptively using the silhouette coefficient. The area of ​​the region Boundary smoothness is obtained through the polygon area calculation formula. Shape regularity is calculated using boundary curvature. Calculated using the roundness index. Weighting coefficients. The value was determined through simulated annealing algorithm optimization, ranging from 0 to 1. The main purpose of this step is to achieve a reasonable division of the land surface area.

[0127] The specific implementation of step S05 is as follows: Based on the surface partition weight matrix, an adaptive quadtree partitioning strategy is used to perform grid subdivision on each partition. The initial grid size is set to 4 times the data resolution. The surface feature complexity is calculated for each grid cell, and the complexity calculation uses a grid structure evaluation function. ,in For the area of ​​the grid cell, The three sides of the grid cell are the grid gradient. Calculated by side length ratio , These represent the maximum and minimum side lengths, and the vertex distribution uniformity, respectively. The area variance of the Thiessen polygon is calculated. When the complexity exceeds a threshold of 0.6, the mesh cells are subdivided recursively, with a maximum subdivision level of 6 layers. Weighting coefficients are used. The values ​​are determined through optimization using a genetic algorithm, ranging from 0 to 1. The main purpose of this step is to generate a multi-scale grid structure that adapts to the surface features.

[0128] The specific implementation of step S06 is as follows: For each node in the multi-scale grid data of the Earth's surface, the principal curvature and Gaussian curvature are calculated using the discrete differential geometry method. First, a local quadratic surface fitting model is constructed. The coefficients are solved using the least squares method. Calculate Gaussian curvature based on fitting coefficients. and mean curvature Then, the principal curvature is obtained. The morphological evaluation function expression is as follows: The feature point density Calculated using kernel density estimation , For bandwidth parameters, Kernel function; continuity of characteristic lines The continuity parameters of the characteristic lines are calculated using spline interpolation based on curve fitting error evaluation. Weighting coefficients are also used. Cross-validation determined that the value ranged from 0.2 to 0.5. The main purpose of this step is to achieve an accurate representation of the landform features.

[0129] The specific implementation of step S07 is as follows: Utilizing the feature information in the landform matrix, an adaptive compression algorithm is used to calculate the data compression factor. The compression optimization function expression is: ,in Storage sizes for the original data and compressed data, respectively, and reconstruction error. Calculated using root mean square error , These are the original value and the reconstructed value, respectively, and the data redundancy is... Calculation by information entropy , This represents the probability of a data value occurring. The compression factor ranges from 0.1 to 0.9, and the weighting coefficient... The genetic algorithm is used for optimization. The main purpose of this step is to achieve efficient compressed storage of the data.

[0130] The specific implementation of step S08 is as follows: The compressed multi-scale grid data of the land surface is organized using a quadtree structure. The index performance function expression is: retrieval time Memory usage was calculated by averaging multiple tests. Real-time monitoring of tree height differences via system calls. Calculation of variance through hierarchical distribution , This represents the node hierarchy depth. Weight coefficients. The gradient descent method is used for optimization, with values ​​ranging from 0.1 to 0.3. The purpose of this step is to establish an efficient data management structure.

[0131] The specific implementation of step S09 is as follows: Based on the quadtree structure, calculate the spatial relationships between nodes and construct a topological association matrix. The expression for the topological evaluation function is: , where node connectivity The distance between nodes is calculated using graph theory algorithms. Using Euclidean distance, hierarchical connectivity The node degree distribution was evaluated using a depth-first search algorithm. Calculated by degree distribution entropy , For degree The percentage of nodes. Weighting coefficient. This is determined using a grid search method. The main purpose of this step is to establish the relationships between data blocks.

[0132] The specific implementation of step S10 is as follows: A data reading evaluation function is constructed based on the topological correlation matrix, and the function expression is: The size of the data block Data block hierarchy is obtained through byte counting. The depth of the node in the tree, and the frequency of access. Statistics are based on historical access records. Random disturbance term. It follows a normal distribution with a mean of 0 and a standard deviation of 0.1. Weighting coefficients. The priority of data block loading is determined through optimization using machine learning methods. This step aims to achieve a reasonable allocation of data block loading priorities.

[0133] The specific implementation of step S11 is as follows: The correlation between the viewpoint position and each data block is calculated using a data reading evaluation function, and a viewpoint correlation matrix is ​​constructed. First, the viewpoint position coordinates are determined. and line-of-sight vector Calculate the distance and angle between the viewpoint and the center point of the data block. The viewpoint importance function expression is: Among them, the angle of view Calculated by vector dot product , The coordinates of the data block center; the line-of-sight distance. The Euclidean distance from the viewpoint to the center of the data block; occlusion degree. The visual attention value is calculated using a ray tracing algorithm, taking into account the occlusion relationships between data blocks. The weighting coefficients are calculated using a saliency detection algorithm and a multi-scale contrast feature extraction method. The optimization is achieved using reinforcement learning methods, with values ​​ranging from 0.2 to 0.4. The main purpose of this step is to optimize the loading order of data blocks.

[0134] The specific implementation of step S12 is as follows: The data is organized into blocks based on the viewpoint correlation matrix. The storage evaluation function expression is: The size of the data block The value ranges from 32 to 256 kilobytes, with an average access time of... Space utilization rate obtained through performance testing Data dispersion is calculated by the ratio of actual storage space to theoretical storage space. Calculated by the distribution entropy of data blocks on the storage medium. , For the data block in the first The probability of each memory cell. Weighting coefficients. The optimal value is determined using a Bayesian optimization method, ranging from 0.1 to 0.3. An index table is created for the organized data blocks, recording information including data block identifier, spatial range, compression ratio, and storage location. The main purpose of this step is to achieve rational data block storage.

[0135] The specific implementation of step S13 is as follows: a progressive loading strategy is adopted to achieve fast data reading. The loading performance function expression is: The number of data blocks that have been loaded Loading time is measured in real time by a counter. Quality is measured and displayed using a high-precision timer. Image quality assessment algorithms are used to calculate metrics including resolution integrity and texture sharpness, as well as memory usage. Data is obtained through the system monitoring interface. When the loading time exceeds 50 milliseconds, optimization is achieved by adjusting the batch size. The batch size adjustment employs an adaptive control algorithm, dynamically changing based on system load and network status. The preloading mechanism uses a predictive model, based on user interaction behavior and viewpoint change trends, to predict the data blocks that may be needed in the next moment. The preloaded amount does not exceed 30% of the currently displayed data volume. Weighting coefficients. The system is optimized through online learning methods and dynamically adjusted according to its operational status. The main purpose of this step is to ensure the real-time and continuous nature of data loading.

[0136] The optimization algorithms involved in the above steps, including genetic algorithms, simulated annealing, and gradient descent, have the following parameter settings: For the genetic algorithm, the population size is set to 100, the crossover probability to 0.8, the mutation probability to 0.1, and the maximum number of iterations to 1000; for the simulated annealing algorithm, the initial temperature is set to 100, the cooling coefficient to 0.95, and the termination temperature to 0.01; for the gradient descent algorithm, the learning rate is set to 0.01, and the convergence threshold to 0.0001. These parameter values ​​are optimal configurations obtained through extensive experimental verification and can be adjusted appropriately according to specific application scenarios.

[0137] The optimization process for weight coefficients in all evaluation functions employs cross-validation, dividing the dataset into training and validation sets in an 8:2 ratio. The optimal weight values ​​are determined by minimizing the error on the validation set. The evaluation function design considers a balance between the diversity of data features and computational efficiency, achieving a comprehensive evaluation of different features by introducing multiple evaluation metrics and weight coefficients.

[0138] To better understand and implement this invention, Example 2 of a specific application scenario is provided below: A research institute uses the three-dimensional surface model data generation and rapid reading method of this invention to process high-resolution terrain data. The total area covered by the study area is approximately 28,600 square kilometers, and the original data comes from multi-temporal airborne lidar scanning, containing approximately 28 billion point cloud data points. The research team carried out data processing work based on the method of this invention, and the specific implementation process is as follows.

[0139] Phase 1: Data Preprocessing and Feature Extraction. First, the raw point cloud data underwent quality checks and outlier identification. Local statistical feature analysis was used to calculate the local elevation standard deviation for each point, with a detection window size of 200m × 200m and an outlier threshold of 3.5 times the standard deviation. The detection results are shown in Table 1 below:

[0140] Table 1 Test Results

[0141]

[0142] An improved Haar wavelet transform was used for a four-level multi-scale decomposition. The feature data and statistical characteristics obtained at each level are shown in Table 2 below:

[0143] Table 2 Feature Data Table

[0144]

[0145] The surface texture calculation uses an improved spatial autocorrelation model, with the following parameters set: spatial weight decay index. Texture weighting coefficient Entropy weighting coefficient The calculation results show that the surface texture features of the study area exhibit significant spatial heterogeneity, as illustrated in Table 3 below:

[0146] Table 3 Statistical Data

[0147]

[0148] Figure 2 It shows the distribution characteristics of surface texture for different landform types, including mean, standard deviation, and range of values.

[0149] Phase Two: Feature Decomposition and Data Dimensionality Reduction. A covariance matrix is ​​constructed from the multi-frequency surface feature data, and eigenvalue decomposition is performed. Gaussian kernel parameters are set. Feature space correlation weights Feature clustering weight The feature importance evaluation results are shown in the table below:

[0150] Table 4 Feature Evaluation Table

[0151]

[0152] Figure 3 The cumulative contribution rate curve of eigenvalue decomposition is displayed, reflecting the basis for feature selection.

[0153] Phase 3: Spatial Association Analysis and Region Division. Spatial association features are calculated based on the eigenvector matrix. An improved Moran's index method is used to construct a spatial association optimization function, with the following parameters set: distance decay coefficient. The spatial association threshold is 0.65. The calculated spatial association analysis results are shown in Table 5 below:

[0154] Table 5 Spatial Association Analysis Table

[0155]

[0156] Based on the spatial correlation analysis results, spectral clustering algorithm was used for region partitioning, with the following parameters set: Gaussian kernel bandwidth 0.5, and adaptive threshold for the number of clusters 0.85. The evaluation of the region partitioning results is shown in Table 6 below.

[0157] Table 6. Regional Division Results

[0158]

[0159] Phase 4: Mesh Generation and Optimization. An adaptive quadtree partitioning strategy was adopted, with an initial mesh size of 25 meters, and multi-level subdivision was performed based on surface complexity. The mesh quality evaluation results are shown in Table 7 below:

[0160] Table 7 Network Quality Evaluation Table

[0161]

[0162] Phase 5: Data Compression and Index Construction. A morphological feature-adaptive compression algorithm is employed, with the compression optimization function parameters set as follows: reconstruction error weights. Redundancy weight The compression effect evaluation is shown in Table 8 below:

[0163] Table 8 Compression Effect Evaluation Table

[0164]

[0165] Phase 6: Performance Testing and Effect Comparison. Traditional methods generally include: Method 1: The regular grid-based method divides the land surface into fixed-size grids, uses a pyramid hierarchical structure to store data of different resolutions, uses a general compression algorithm for data compression, and loads the display area data step by step in blocks according to the viewpoint position and display ratio. Method 2: The irregular triangular mesh-based method first extracts terrain feature points to construct a Delaunay triangular mesh, generates a multi-level model through a progressive mesh simplification algorithm, organizes the data using a tree structure, and dynamically loads triangular mesh data of different precision according to viewpoint parameters. Method 3: The hybrid grid-based method divides the land surface into partitions and processes them using regular grids and irregular triangular meshes respectively, uses a partition compression strategy and a multi-branch tree structure to manage the data, and combines a preloading mechanism to achieve dynamic reading and display of data. The following performance tests were conducted under a standard test environment (CPU: Intel Xeon 2.8GHz, memory: 256GB, video memory: 32GB), comparing traditional Method 1 and the method of this invention. The results are shown in Table 9:

[0166] Table 9 Performance Test Data

[0167]

[0168] The main advancements of the method described in this invention compared to traditional methods are as follows:

[0169] 1. Through multi-scale decomposition and feature extraction, data processing efficiency is improved by 3.6 times while maintaining a high fidelity of surface features.

[0170] 2. By adopting an adaptive compression algorithm, storage space is reduced by 82.9% while maintaining a reconstruction accuracy greater than 0.98.

[0171] 3. Based on the improved quadtree index structure, the retrieval response time is reduced from 85 milliseconds to 12 milliseconds.

[0172] 4. By employing a viewpoint-dependent progressive loading strategy, the display frame rate was increased to 42.6 frames per second, achieving a smooth interactive experience.

[0173] 5. The overall performance of the system is improved by more than 80% on average compared with existing technologies, meeting the real-time processing requirements of large-scale three-dimensional surface data.

[0174] It should be noted that the variables involved in this invention are explained in detail in Table 10 below.

[0175] Table 10 Variable Explanation Table

[0176]

[0177] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating and rapidly reading three-dimensional surface model data, characterized in that, Includes the following steps: Original surface elevation data were acquired and multi-frequency surface feature data were obtained by wavelet multi-scale decomposition. The surface feature vector matrix is ​​obtained by performing singular value decomposition on the multi-frequency surface feature data. The feature vector matrix is ​​evaluated using a feature importance evaluation function to obtain an evaluation value. Feature vectors with an evaluation value greater than 0.8 are retained. The process involves constructing a surface spatial correlation matrix and dividing the region using the surface feature vector matrix; generating multi-scale grid data based on the region division results; constructing a quadtree structure to store the multi-scale grid data, where the depth of the quadtree is determined by the grid hierarchy. During construction, the root node range is first determined, and then child nodes are recursively divided until the finest level is reached; calculating the topological correlation matrix between nodes in the quadtree structure; constructing a data reading evaluation function based on the topological correlation matrix; and employing a progressive loading strategy to achieve fast data reading. The specific steps for constructing the surface spatial correlation matrix and dividing the region include: constructing the surface spatial correlation matrix using the Moran's index method; calculating the spatial autocorrelation coefficient of local regions using the sliding window method; optimizing the spatial autocorrelation coefficient using a spatial correlation optimization function; dividing the region using a spectral clustering algorithm; and constructing a surface partition weight matrix.

2. The method for generating and rapidly reading three-dimensional surface model data according to claim 1, characterized in that, The steps of acquiring raw surface elevation data and performing wavelet multi-scale decomposition specifically include: performing a data integrity check on the raw surface elevation data and removing outliers; performing a 4-level decomposition on the raw surface elevation data using Haar wavelet transform; normalizing the decomposed multi-frequency surface feature data; and evaluating the multi-frequency surface feature data using a surface texture evaluation function.

3. The method for generating and rapidly reading three-dimensional surface model data according to claim 1, characterized in that, The specific steps for performing singular value decomposition on multi-frequency surface feature data include: constructing a covariance matrix; performing eigenvalue decomposition on the covariance matrix using the Jacobi iteration method; and selecting eigenvectors with a cumulative contribution rate of 90% to form a surface feature vector matrix.

4. The method for generating and rapidly reading three-dimensional surface model data according to claim 1, characterized in that, The specific steps for generating multi-scale grid data include: dividing each partition into grids using an adaptive quadtree partitioning strategy; calculating the surface feature complexity of each grid cell; generating multi-scale grid data of the surface through recursive subdivision; and evaluating the multi-scale grid data of the surface using a grid structure evaluation function.

5. The method for generating and rapidly reading three-dimensional surface model data according to claim 4, characterized in that, The steps for constructing a quadtree structure to store multi-scale grid data specifically include: calculating the curvature matrix of grid nodes; extracting surface feature lines and feature points to form a surface morphology matrix; calculating the data compression factor; performing hierarchical compression and storage of the multi-scale grid data; and organizing the compressed multi-scale grid data using a quadtree structure. The steps for calculating the topological association matrix between nodes in the quadtree structure specifically include: calculating the spatial relationships between nodes; constructing the topological association matrix; calculating the connection strength using a topological evaluation function; and optimizing the node connection method to ensure the rationality of the topological structure.

6. The method for generating and rapidly reading three-dimensional surface model data according to claim 1, characterized in that, The specific steps for constructing the data reading evaluation function include: calculating the importance value of each data block; constructing a viewpoint association matrix; performing quantitative analysis on the viewpoint association matrix using the viewpoint importance function; and organizing the data into blocks based on the viewpoint association matrix.

7. The method for generating and rapidly reading three-dimensional surface model data according to claim 1, characterized in that, The steps to achieve fast data reading using a progressive loading strategy include: prioritizing the loading of data blocks with high importance; monitoring the loading process using a loading performance function; preloading data blocks through a preloading mechanism; and dynamically adjusting the batch size of data blocks.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions, which, when executed in a computer, are used to perform a method for generating and rapidly reading three-dimensional surface model data as described in any one of claims 1-7.

9. A system for generating and rapidly reading three-dimensional surface model data, characterized in that, The system includes the computer-readable storage medium of claim 8, wherein the system is any one of a computer, a server, or a microcontroller, the computer-readable storage medium is disposed within the system, and the system is provided with a microprocessor that executes the program instructions stored in the computer-readable storage medium.

Images