Three-dimensional model data management system and method for land surveying and mapping

Abstract

The invention relates to the technical field of data management, in particular to a three-dimensional model data management system and method for land surveying and mapping. The method comprises the following steps: acquiring original three-dimensional point cloud data of land surveying and mapping; performing terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model; performing terrain feature extraction on the terrain semantic segmentation model, and performing semantic feature extraction to obtain terrain feature data and semantic feature data; performing terrain complexity analysis according to the terrain feature data and the semantic feature data to obtain a terrain complexity map; performing self-adaptive partitioning according to the terrain complexity map to obtain multi-resolution partitioning data; and performing multi-resolution index construction according to the multi-resolution block data to obtain a multi-resolution block index database. According to the invention, through efficient, flexible and intelligent management of the land surveying and mapping three-dimensional model data, the defects of the existing method in the aspects of data management efficiency and terrain adaptability are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and particularly to a three-dimensional model data management system and method for land surveying and mapping. Background Art

[0002] In the field of land surveying and mapping, a vast amount of three-dimensional point cloud data has been accumulated. How to efficiently manage and utilize this data has become an extremely urgent problem to be solved. In the early data management methods, the data was stored in the form of files locally or on the server, lacking organization and indexing, with low retrieval efficiency and being difficult to perform spatial queries and analyses. Using relational databases or spatial databases to manage three-dimensional model data allows for structured queries and management, but the processing efficiency for massive point cloud data is still limited. Especially when performing spatial analysis and visualization, the performance bottleneck is obvious. The LOD technology renders models with different precisions according to different viewing distances and perspectives, which can improve the rendering efficiency, but the data organization and management still rely on traditional databases or file systems, making it difficult to achieve efficient access and management. Using spatial indexing technologies such as Octree and k-d tree can speed up spatial queries and retrieval, but for complex terrain areas, the cost of constructing and maintaining the index is relatively high, and it is difficult to adapt to terrain changes.

[0003] However, the existing three-dimensional model data management methods for land surveying and mapping have low data management efficiency and poor terrain adaptability. The existing land surveying and mapping three-dimensional model data management methods often store and manage data with a single resolution. For large-scale and high-precision data, single high-resolution storage will occupy a large amount of storage space, increasing the storage cost. When performing data retrieval and access, all data needs to be loaded, resulting in a long loading time and being difficult to meet the requirements of real-time applications. In existing methods, in complex terrain areas, the data organization and management are not fine enough. In flat areas, using high-resolution storage will cause data redundancy and waste storage space. In complex terrain areas, using low-resolution storage will lose important terrain details, affecting data accuracy and application effects. A single resolution is difficult to accurately represent the characteristics of complex terrains, such as valleys, ridges, steep slopes, etc. Summary of the Invention

[0004] Based on this, it is necessary to provide a three-dimensional model data management system and method for land surveying and mapping to solve at least one of the above technical problems.

[0005] To achieve the above object, a three-dimensional model data management method for land surveying and mapping includes the following steps:

[0006] Step S1: Obtain the original three-dimensional point cloud data of land surveying and mapping; perform terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model;

[0007] Step S2: Extract terrain features and semantic features from the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and semantic feature data to obtain a terrain complexity map; perform adaptive chunking based on the terrain complexity map to obtain multi-resolution chunk data; construct a multi-resolution index based on the multi-resolution chunk data to obtain a multi-resolution chunk index database;

[0008] Step S3: Develop an encryption policy based on the multi-resolution chunk index database to obtain an encryption policy table; encrypt the multi-resolution chunk data according to the encryption policy table and update the encryption index to obtain encrypted chunk data and an updated multi-resolution chunk index database; obtain system user role data; perform role- and attribute-based access control based on the system user role data to obtain a role- and attribute-based access control list;

[0009] Step S4: Extract classified user access sequence data according to the role- and attribute-based access control list; extract node feature data according to the updated multi-resolution chunk index database to obtain node feature data; construct an access probability prediction model based on the classified user access sequence data and node feature data to obtain an access probability prediction model; use the access probability prediction model to predict the data block access probability to obtain a predicted access probability table; perform spatial-aware cache management based on the predicted access probability table, encrypted chunk data, and role- and attribute-based access control list to obtain spatial-aware cache data, so as to implement the management of 3D model data for land surveying and mapping.

[0010] Through point cloud preprocessing, feature extraction, and training and inference of the ground object point cloud recognition model, the present invention realizes high-precision and high-efficiency terrain semantic segmentation, providing a reliable data basis for subsequent terrain analysis and data management. Denoising, outlier removal, and data completion operations improve data quality. Feature extraction effectively expresses the geometric features of the point cloud, and the ground object point cloud recognition model realizes automated and intelligent terrain classification. Through terrain feature and semantic feature extraction, terrain complexity analysis, adaptive partitioning, and construction of a multi-resolution index, multi-scale expression and efficient management of terrain data are achieved. The extracted terrain and semantic features comprehensively describe terrain information. The terrain complexity map guides adaptive partitioning, and the multi-resolution index database supports fast retrieval and access to data blocks of different resolutions. Through attribute sensitivity analysis, formulation of a differential encryption strategy, data encryption, and generation of an access control list, secure management and access control of multi-resolution partitioned data are realized. Sensitivity analysis and the differential encryption strategy ensure data confidentiality, and the access control list realizes fine-grained permission management, preventing unauthorized access and data leakage. The application of homomorphic encryption supports the analysis requirements for encrypted data while protecting data security. Through user access history records, spatial proximity relationship analysis, extraction of user access sequences, node feature extraction, construction of an access probability prediction model, and spatial awareness cache management, an intelligent caching mechanism based on user behavior prediction is realized, improving the cache hit rate, reducing data access latency, and enhancing the user experience and system efficiency. The application of the graph neural network model effectively utilizes spatial and semantic information to achieve more accurate access prediction. Therefore, the present invention provides a method for managing three-dimensional model data for land surveying and mapping, realizing efficient, flexible, and intelligent management of three-dimensional model data for land surveying and mapping, and solving the deficiencies of existing methods in terms of data management efficiency and terrain adaptability.

[0011] Preferably, step S1 includes the following steps:

[0012] Step S11: Obtain the original three-dimensional point cloud data of land surveying and mapping; perform point cloud data preprocessing on the original three-dimensional point cloud data to obtain preprocessed point cloud data;

[0013] Step S12: Perform point cloud feature extraction on the preprocessed point cloud data to obtain point cloud feature data;

[0014] Step S13: Construct a ground object point cloud recognition model based on the point cloud feature data and preset ground object category semantic labels to obtain a ground object point cloud recognition model;

[0015] Step S14: Use the ground object point cloud recognition model to perform terrain semantic segmentation inference on the preprocessed point cloud data to obtain a terrain semantic segmentation model.

[0016] The present invention preprocesses the acquired original 3D point cloud data, effectively removing noise and outliers in the data and filling in the missing data, thereby improving the accuracy and reliability of subsequent point cloud feature extraction and semantic segmentation, and providing a high-quality data basis for subsequent processing steps. By extracting features from the preprocessed point cloud data, feature information such as the normal vector, curvature, and FPFH descriptor of each point is obtained, effectively expressing the local geometric features of the point cloud and providing effective input data for the training of subsequent ground object point cloud recognition models, which helps improve the learning ability and generalization ability of the models. By using the extracted point cloud feature data and the preset semantic labels of ground object categories to train the ground object point cloud recognition model, a ground object point cloud recognition model capable of effectively identifying different terrain ground objects is obtained. This model can learn complex patterns and features in the point cloud data, providing strong technical support for subsequent terrain semantic segmentation inference, thereby achieving high-precision terrain classification. Using the trained ground object point cloud recognition model to perform terrain semantic segmentation inference on the preprocessed point cloud data, a terrain semantic segmentation model with semantic labels is obtained. This model can accurately divide each point in the point cloud data into different ground object categories, providing important basic data for subsequent terrain feature extraction, semantic feature extraction, and adaptive chunking, etc., and realizing automated and efficient terrain semantic segmentation.

[0017] Preferably, step S2 includes the following steps:

[0018] Step S21: Extract terrain features from the terrain semantic segmentation model to obtain terrain feature data;

[0019] Step S22: Extract semantic features according to the terrain semantic segmentation model to obtain semantic feature data;

[0020] Step S23: Construct a terrain complexity index based on the terrain feature data and the semantic feature data to obtain terrain complexity index data; perform raster data conversion on the terrain complexity index data to obtain a terrain complexity map;

[0021] Step S24: Perform adaptive chunking according to the terrain complexity map and the terrain semantic segmentation model to obtain multi-resolution chunk data;

[0022] Step S25: Calculate the block attributes of the multi-resolution chunk data to obtain block attribute data;

[0023] Step S26: Construct a multi-resolution index for the multi-resolution chunk data and the block attribute data to obtain a multi-resolution chunk index database.

[0024] Through terrain feature extraction from the terrain semantic segmentation model, terrain feature data such as slope, curvature, and elevation change rate of each point are obtained, which can more comprehensively describe the geometric shape of the terrain and provide an important data basis for subsequent terrain complexity analysis. Through semantic feature extraction according to the terrain semantic segmentation model, semantic feature data of different ground object categories are obtained, such as vegetation density, building height and volume, etc., enriching the expression of terrain information and making the subsequent terrain complexity analysis more comprehensive and representative. By combining terrain feature data and semantic feature data to construct a terrain complexity index and converting it into raster data to generate a terrain complexity map, the visual expression of terrain complexity is realized, providing a basis for subsequent adaptive chunking and making the chunking strategy more targeted. By performing adaptive chunking according to the terrain complexity map and the terrain semantic segmentation model, the resolution of different terrain areas is adaptively adjusted. On the premise of ensuring data accuracy, the data storage volume and calculation amount are effectively reduced, and the data processing efficiency is improved. By calculating the block attributes of the multi-resolution chunked data, attribute information such as average elevation, average slope, main ground object type, and point cloud number of each chunk is obtained, providing richer index information for the construction of subsequent multi-resolution indexes and data retrieval. By using the Octree structure to construct a multi-resolution index for the multi-resolution chunked data and block attribute data, an efficient multi-resolution chunked index database is formed, which can quickly retrieve and access data blocks of different resolutions, significantly improving the data access efficiency.

[0025] Preferably, step S21 includes the following steps:

[0026] Step S211: Determine the neighborhood radius according to the terrain semantic segmentation model to obtain neighborhood radius data;

[0027] Step S212: Perform a neighborhood search according to the neighborhood radius data and the terrain semantic segmentation model to obtain neighboring point data;

[0028] Step S213: Calculate the slope according to the neighboring point data and the terrain semantic segmentation model to obtain slope data;

[0029] Step S214: Calculate the curvature according to the neighboring point data and the terrain semantic segmentation model to obtain curvature data;

[0030] Step S215: Calculate the elevation change rate according to the neighboring point data and the terrain semantic segmentation model to obtain elevation change rate data;

[0031] Step S216: Integrate the slope data, curvature data, and elevation change rate data to obtain terrain feature data.

[0032] By determining the neighborhood radius according to the terrain semantic segmentation model, the present invention can adaptively adjust the size of the neighborhood radius, so as to better adapt to point cloud data with different density distributions, ensure obtaining appropriate neighboring points in subsequent neighboring point searches, and improve the accuracy of terrain feature extraction. By performing neighboring point searches according to the neighborhood radius data and the terrain semantic segmentation model, neighboring point data within a certain range around each point is obtained, providing a necessary data basis for the subsequent calculations of slope, curvature, and elevation change rate. By calculating the slope according to the neighboring point data and the terrain semantic segmentation model, the slope data of each point is obtained, which can effectively reflect the inclination degree of the terrain and is one of the important terrain features. By calculating the curvature according to the neighboring point data and the terrain semantic segmentation model, the curvature data of each point is obtained, which can effectively reflect the bending degree of the terrain and is one of the important terrain features. By calculating the elevation change rate according to the neighboring point data and the terrain semantic segmentation model, the elevation change rate data of each point is obtained, which can effectively reflect the severity of terrain undulation and is one of the important terrain features. By integrating the slope data, curvature data, and elevation change rate data, complete terrain feature data is formed, which can more comprehensively describe the geometric shape of the terrain and provide richer information for subsequent terrain complexity analysis.

[0033] Preferably, step S22 includes the following steps:

[0034] Step S221: Classify the point cloud data according to the semantic labels of each point in the terrain semantic segmentation model to obtain classified point cloud data;

[0035] Step S222: Calculate the vegetation density of the classified point cloud data to obtain vegetation density data;

[0036] Step S223: Calculate the building height and volume of the classified point cloud data to obtain building height and volume data;

[0037] Step S224: Calculate the road width and direction of the classified point cloud data to obtain road width and direction data;

[0038] Step S225: Calculate the water area and average depth of the classified point cloud data to obtain water area and average depth data;

[0039] Step S226: Integrate the semantic features of the vegetation density data, building height and volume data, road width and direction data, and water area and average depth data to obtain semantic feature data.

[0040] The present invention classifies point cloud data according to semantic tags, divides point cloud data of different ground object types into different sets, facilitates subsequent extraction of specific semantic features for different ground object types, and improves the efficiency and accuracy of feature extraction. By calculating the vegetation density of the classified vegetation point cloud data, vegetation density data reflecting the vegetation coverage degree is obtained, providing important vegetation information for subsequent terrain analysis and applications. By calculating the height and volume of the classified building point cloud data, the height and volume data of the buildings are obtained, providing important three-dimensional information of the buildings, which can be used in applications such as urban planning and 3D modeling. By calculating the width and direction of the classified road point cloud data, the width and direction data of the roads are obtained, providing important road information, which can be used in applications such as traffic planning and navigation. By calculating the area and average depth of the classified water area point cloud data, the area and average depth data of the water area are obtained, providing important hydrological information, which can be used in applications such as water resource management and environmental monitoring. By integrating the vegetation density data, building height and volume data, road width and direction data, and water area area and average depth data, semantic feature data containing semantic features of multiple ground object types is formed, which can more comprehensively describe the ground object features and provide richer data support for subsequent applications.

[0041] Preferably, step S3 includes the following steps:

[0042] Step S31: Perform attribute sensitivity analysis according to the multi-resolution block index database to obtain an attribute sensitivity level table;

[0043] Step S32: Develop an encryption policy according to the attribute sensitivity level table and the multi-resolution block index database to obtain an encryption policy table;

[0044] Step S33: Encrypt the multi-resolution block data according to the encryption policy table and the multi-resolution block index database to obtain encrypted block data;

[0045] Step S34: Update the encrypted index of the multi-resolution block index database according to the encrypted block data to obtain an updated multi-resolution block index database;

[0046] Step S35: Obtain system user role data; define role permissions according to the system user role data to obtain a role permission table;

[0047] Step S36: Develop an attribute access policy according to the attribute sensitivity level table and the role permission table to obtain an attribute access policy table;

[0048] Step S37: Generate an access control list based on the attribute access policy table, the updated multi-resolution block index database, and the encryption policy table to obtain a role- and attribute-based access control list.

[0049] Through the attribute sensitivity analysis of the multi-resolution block index database, the present invention obtains an attribute sensitivity level table, clarifies the sensitivity levels of different attributes, provides a basis for formulating differential encryption policies subsequently, and thus can better protect sensitive data. By formulating an encryption policy according to the attribute sensitivity level table and the multi-resolution block index database, an encryption policy table is obtained, realizing differential encryption with different encryption algorithms and key lengths for attributes with different sensitivities. While ensuring data security, the computational overhead brought by encryption is minimized as much as possible. By encrypting the multi-resolution block data according to the encryption policy table, encrypted block data is obtained, effectively protecting the confidentiality of the data and preventing unauthorized access and data leakage. By updating the multi-resolution block index database according to the encrypted block data, an updated multi-resolution block index database is obtained, ensuring the consistency between the index database and the encrypted data and facilitating subsequent access control and data retrieval. By obtaining the system user role data and defining role permissions, a role permission table is obtained, clarifying the data and operation permissions that users with different roles can access, providing a basis for subsequent access control. By formulating an attribute access policy according to the attribute sensitivity level table and the role permission table, an attribute access policy table is obtained, realizing fine-grained access control and ensuring that only authorized users can access sensitive data. By generating an access control list based on the attribute access policy table, the updated multi-resolution block index database, and the encryption policy table, a role- and attribute-based access control list is obtained, realizing a flexible access control mechanism, which can effectively control the access permissions of users to different data blocks and attributes, and ensure the security and integrity of the data.

[0050] Preferably, step S32 includes the following steps:

[0051] Step S321: Determine the encryption algorithm and key length according to the attribute sensitivity level table to obtain a preliminary encryption policy;

[0052] Step S322: Conduct a data analysis requirement analysis on the multi-resolution block index database to obtain data analysis requirement data;

[0053] Step S323: Select a homomorphic encryption algorithm according to the data analysis requirement data and the preliminary encryption policy to obtain homomorphic encryption algorithm selection data;

[0054] Step S324: Generate an encryption policy table according to the preliminary encryption policy and the homomorphic encryption algorithm selection data to obtain an encryption policy table.

[0055] The present invention determines the encryption algorithm and key length according to the attribute sensitivity level table, obtaining a preliminary encryption strategy, providing a basic encryption solution for data with different sensitivity levels, and ensuring that sensitive data is protected as necessary. By analyzing the data analysis requirements of the multi-resolution block index database, the data analysis requirement data is obtained, clarifying the tasks and requirements of data analysis, providing guidance for subsequent selection of appropriate homomorphic encryption algorithms, and avoiding unnecessary computational overhead. By selecting a homomorphic encryption algorithm according to the data analysis requirement data and the preliminary encryption strategy, the homomorphic encryption algorithm selection data is obtained, ensuring that while meeting the data analysis requirements, operations can be performed on the encrypted data, achieving a balance between data security and data availability. By generating an encryption strategy table according to the preliminary encryption strategy and the homomorphic encryption algorithm selection data, the final encryption strategy table is obtained. This strategy table comprehensively considers data sensitivity, data analysis requirements, and homomorphic encryption algorithms, providing complete guidance for subsequent data encryption and ensuring the best balance between data security and data availability.

[0056] Preferably, step S4 includes the following steps:

[0057] Step S41: Perform user access history recording according to the role- and attribute-based access control list to obtain a user access history database;

[0058] Step S42: Perform spatial proximity relationship analysis according to the updated multi-resolution block index database to obtain a spatial proximity relationship graph; construct a spatial proximity graph from the spatial proximity relationship graph to obtain spatial proximity graph data;

[0059] Step S43: Extract classified user access sequences from the user access history database to obtain classified user access sequence data; extract node feature data from the updated multi-resolution block index database and the classified user access sequence data to obtain node feature data;

[0060] Step S44: Construct an access probability prediction model based on a graph neural network model according to the spatial proximity graph data, the classified user access sequence data, and the node feature data to obtain an access probability prediction model;

[0061] Step S45: Obtain the current user access data block ID; use the current user access data block ID as the starting node and utilize the access probability prediction model to predict the data block access probability to obtain a predicted access probability table;

[0062] Step S46: Perform spatial awareness cache management according to the predicted access probability table, the encrypted block data, and the role- and attribute-based access control list to obtain spatial awareness cache data.

[0063] The present invention records the user access history and constructs a user access history database, providing a data basis for subsequent user behavior analysis and access prediction, and helping to understand the user's data access patterns and habits. By performing spatial proximity relationship analysis on the updated multi-resolution block index database and constructing spatial proximity graph data, the spatial correlation between data blocks is captured, providing important graph structure information for subsequent access probability prediction based on graph neural networks. By extracting classified user access sequences and node feature data from the user access history database, classified user access sequence data and node feature data are obtained, providing training data and feature inputs for subsequent construction of an access probability prediction model, enabling the model to learn the spatial and semantic features of user access behavior. By constructing an access probability prediction model based on graph neural networks using the spatial proximity graph data, classified user access sequence data, and node feature data, an access probability prediction model capable of predicting the user's future access behavior is obtained, providing a prediction basis for realizing spatial-aware cache management. By using the access probability prediction model to predict the access probability of the data blocks currently accessed by the user, a predicted access probability table is obtained, clarifying the data blocks that the user may access next and their probabilities, providing guidance for cache preloading. By performing spatial-aware cache management according to the predicted access probability table, encrypted block data, and role- and attribute-based access control lists, intelligent cache preloading and replacement strategies are realized, improving the cache hit rate, reducing data access latency, and thus enhancing the user experience and system efficiency.

[0064] Preferably, step S43 includes the following steps:

[0065] Step S431: Group users according to the user access history database to obtain user group data; perform time sorting on the user group data to obtain user time sorting data;

[0066] Step S432: Extract access sequences from the user time sorting data to obtain user access sequence data; perform role and purpose classification according to the user access sequence data and the user time sorting data to obtain classified user access sequence data;

[0067] Step S433: Extract data block attributes from the updated multi-resolution block index database to obtain data block attribute data; perform normalization processing on the numerical type attributes of the data block attribute data to obtain normalized data;

[0068] Step S434: Perform one-hot encoding on the categorical type attributes of the data block attribute data to obtain one-hot encoded data; perform missing value processing on the data block attribute data to obtain missing value processed data;

[0069] Step S435: Conduct access frequency statistics on the classified user access sequence data to obtain data block access frequency data; construct node feature vectors from the normalized data, one-hot encoded data, missing value processed data, and data block access frequency data to obtain node feature data.

[0070] Through user grouping and time sorting of the user access history database, the present invention obtains user time-sorted data, providing a well-organized data basis for subsequent extraction of user access sequences and user behavior analysis. Through access sequence extraction and role-purpose classification of the user time-sorted data, classified user access sequence data is obtained, which divides the user's access behavior in detail according to time sequence, role, and purpose, providing high-quality training data for subsequent construction of a more accurate access prediction model. Through data block attribute extraction and numerical type attribute normalization processing of the updated multi-resolution block index database, normalized data is obtained, making attributes with different numerical ranges comparable and improving the efficiency and accuracy of subsequent model training. Through one-hot encoding of category type attributes and missing value processing of the data block attribute data, one-hot encoded data and missing value processed data are obtained, converting category type attributes into numerical representations and effectively handling missing values, ensuring data integrity and providing reliable input features for subsequent model training. Through access frequency statistics on the classified user access sequence data and constructing node feature vectors in combination with the normalized data, one-hot encoded data, and missing value processed data, node feature data is obtained, effectively integrating the attribute information of data blocks and the user access behavior features to form a more comprehensive node feature representation, which helps to improve the performance of the access probability prediction model.

[0071] Preferably, the present invention also provides a three-dimensional model data management system for land surveying and mapping, which is used to execute the three-dimensional model data management method for land surveying and mapping as described above. The three-dimensional model data management system for land surveying and mapping includes:

[0072] A terrain semantic segmentation module, which is used to obtain the original three-dimensional point cloud data of land surveying and mapping; conduct terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model;

[0073] A multi-resolution block index module, which is used to extract terrain features and conduct semantic feature extraction on the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; conduct terrain complexity analysis based on the terrain feature data and semantic feature data to obtain a terrain complexity map; conduct adaptive blocking based on the terrain complexity map to obtain multi-resolution block data; construct a multi-resolution block index based on the multi-resolution block data to obtain a multi-resolution block index database;

[0074] An encryption storage and access control module is used to formulate an encryption policy according to a multi - resolution block index database to obtain an encryption policy table; encrypt block data of multi - resolution block data according to the encryption policy table and update the encryption index to obtain encrypted block data and an updated multi - resolution block index database; obtain system user role data; perform role - and - attribute - based access control according to the system user role data to obtain a role - and - attribute - based access control list.

[0075] A space - aware caching module is used to extract classified user access sequences according to a role - and - attribute - based access control list to obtain classified user access sequence data; extract node feature data according to the updated multi - resolution block index database to obtain node feature data; construct an access probability prediction model according to the classified user access sequence data and the node feature data to obtain an access probability prediction model; predict the access probability of data blocks using the access probability prediction model to obtain a predicted access probability table; perform space - aware caching management according to the predicted access probability table, the encrypted block data, and the role - and - attribute - based access control list to obtain space - aware cached data. Brief Description of the Drawings

[0076] Figure 1 It is a schematic diagram of the step - by - step process of a three - dimensional model data management method for land surveying and mapping.

[0077] Figure 2 It is Figure 1 a detailed schematic diagram of the implementation steps of step S2 in

[0078] Figure 3 It is Figure 1 a detailed schematic diagram of the implementation steps of step S3 in

[0079] Figure 4 It is a schematic diagram of the process of a land surveying and mapping three - dimensional model data management system.

[0080] The realization of the purpose, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0081] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0082] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0083] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0084] To achieve the above object, please refer to Figures 1 to 4 , a three-dimensional model data management method for land surveying and mapping, comprising the following steps:

[0085] Step S1: Obtain the original three-dimensional point cloud data of land surveying and mapping; perform terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model;

[0086] Step S2: Extract terrain features and semantic features from the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and the semantic feature data to obtain a terrain complexity map; perform adaptive partitioning according to the terrain complexity map to obtain multi-resolution partitioned data; construct a multi-resolution index based on the multi-resolution partitioned data to obtain a multi-resolution partition index database;

[0087] Step S3: Formulate an encryption policy according to the multi-resolution partition index database to obtain an encryption policy table; encrypt the multi-resolution partitioned data according to the encryption policy table and update the encryption index to obtain encrypted partitioned data and an updated multi-resolution partition index database; obtain system user role data; perform role- and attribute-based access control according to the system user role data to obtain a role- and attribute-based access control list;

[0088] Step S4: Extract classified user access sequences based on the role- and attribute-based access control list to obtain classified user access sequence data; extract node feature data based on the updated multi-resolution block index database to obtain node feature data; construct an access probability prediction model based on the classified user access sequence data and the node feature data to obtain an access probability prediction model; use the access probability prediction model to predict the data block access probability to obtain a predicted access probability table; perform spatial-aware cache management based on the predicted access probability table, the encrypted block data, and the role- and attribute-based access control list to obtain spatial-aware cache data, so as to implement the three-dimensional model data management for land surveying and mapping.

[0089] In the embodiment of the present invention, refer to Figure 1 As shown, it is a schematic flowchart of the steps of the three-dimensional model data management method for land surveying and mapping according to the present invention. In this example, the three-dimensional model data management method for land surveying and mapping includes the following steps:

[0090] Step S1: Obtain the original three-dimensional point cloud data of land surveying and mapping; perform terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model;

[0091] In the embodiment of the present invention, the original three-dimensional point cloud data is obtained, for example, using lidar. The original data is preprocessed, including statistical filtering to remove noise points, radius filtering to remove outliers, and using a local surface reconstruction method to fill in missing data. Then, point cloud features are extracted, such as normal vectors, curvatures, and FPFH descriptors. Using the extracted features and preset semantic labels of ground object categories, a ground object point cloud recognition model (such as PointNet++) is trained. Finally, the trained model is used to perform terrain semantic segmentation inference on the preprocessed point cloud data to obtain a terrain semantic segmentation model.

[0092] Step S2: Extract terrain features from the terrain semantic segmentation model and perform semantic feature extraction to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and the semantic feature data to obtain a terrain complexity map; perform adaptive partitioning based on the terrain complexity map to obtain multi-resolution block data; construct a multi-resolution index based on the multi-resolution block data to obtain a multi-resolution block index database;

[0093] In the embodiments of the present invention, terrain feature extraction is performed on the terrain semantic segmentation model, including calculating the slope, curvature, and elevation change rate of each point. At the same time, semantic feature extraction is carried out, and corresponding features are calculated according to different ground object categories (vegetation, buildings, roads, water areas), such as vegetation density, building height and volume, road width and direction, water area and average depth. Then, combining the terrain features and semantic features, a terrain complexity index is constructed, and it is converted into raster data to generate a terrain complexity map. According to the terrain complexity map, the point cloud data is adaptively segmented, and the areas with higher complexity are segmented smaller. Finally, the attributes of each segment (average elevation, average slope, main ground object type, number of point clouds) are calculated, and a multi-resolution segment index database is constructed using the Octree structure.

[0094] Step S3: Formulate an encryption policy according to the multi-resolution segment index database to obtain an encryption policy table; encrypt the segment data according to the encryption policy table and update the encryption index to obtain encrypted segment data and the updated multi-resolution segment index database; obtain the system user role data; perform role- and attribute-based access control according to the system user role data to obtain a role- and attribute-based access control list;

[0095] In the embodiments of the present invention, sensitivity analysis is performed according to the block attribute data in the multi-resolution segment index database, and the sensitivity levels (high, medium, low) of different attributes are divided. Then, an encryption policy is formulated according to the sensitivity level, and different encryption algorithms (AES-256, AES-128, XOR) and key lengths are selected. The segment data is encrypted, and the storage location and encryption information of the encrypted data are updated to the multi-resolution segment index database. At the same time, the system user role data is obtained, the access permissions of different roles are defined, and an attribute access policy is formulated in combination with the attribute sensitivity level. Finally, according to the attribute access policy, the updated index database, and the encryption policy, a role- and attribute-based access control list (ACL) is generated, specifying the data blocks and operation permissions that each role can access. During the formulation of the encryption policy, the data analysis requirements are analyzed, and if calculations need to be performed on the encrypted data, a suitable homomorphic encryption algorithm is selected.

[0096] Step S4: Extract classified user access sequences based on the role- and attribute-based access control list to obtain classified user access sequence data; extract node feature data based on the updated multi-resolution block index database to obtain node feature data; construct an access probability prediction model based on the classified user access sequence data and the node feature data to obtain an access probability prediction model; use the access probability prediction model to predict the data block access probability to obtain a predicted access probability table; perform spatial awareness cache management based on the predicted access probability table, encrypted block data, and the role- and attribute-based access control list to obtain spatial awareness cache data, so as to implement the three-dimensional model data management work for land surveying and mapping;

[0097] In the embodiment of the present invention, the user access history is recorded, including the access time, user role, accessed data block ID, accessed attributes, and operation type, and stored in the user access history database. Analyze the updated multi-resolution block index database to construct a spatial proximity graph representing the spatial proximity relationship between data blocks. Extract the user access sequence from the user access history database and classify it according to the user role and access purpose. Extract the data block attributes from the updated index database, normalize the numerical type attributes, perform one-hot encoding on the categorical type attributes, and handle missing values. Count the access frequency of the data block in the classified user access sequence and combine all features into a node feature vector. Use a graph neural network model (such as GraphSAGE), combined with the spatial proximity graph, node features, and classified user access sequence, to construct an access probability prediction model. Obtain the data block ID accessed by the current user, use the prediction model to predict the data blocks that the user may access next and their probabilities, and generate a predicted access probability table. Finally, according to the predicted access probability table, encrypted block data, and access control list, perform spatial awareness cache management, preload the data blocks with high access probabilities into the cache, and perform cache replacement according to the LRU algorithm.

[0098] Preferably, step S1 includes the following steps:

[0099] Step S11: Obtain the original three-dimensional point cloud data of land surveying and mapping; perform point cloud data preprocessing on the original three-dimensional point cloud data to obtain preprocessed point cloud data;

[0100] Step S12: Extract point cloud features from the preprocessed point cloud data to obtain point cloud feature data;

[0101] Step S13: Construct a ground object point cloud recognition model based on the point cloud feature data and the preset ground object category semantic labels to obtain a ground object point cloud recognition model;

[0102] Step S14: Use the ground object point cloud recognition model to perform terrain semantic segmentation inference on the preprocessed point cloud data to obtain a terrain semantic segmentation model.

[0103] In the embodiment of the present invention, the original three-dimensional point cloud data of a certain area of land is obtained by using a lidar scanner. This data contains the three-dimensional coordinate information and reflection intensity information of millions of points. There are problems such as noise points, outliers, and data missing in the original point cloud data. The statistical filtering method is used to remove noise points. A threshold is set, and the number of points within a certain radius around each point is counted. If the number is less than the threshold, it is considered a noise point and removed. Then, the radius filtering method is used to remove outliers. The average distance from each point to all points within its neighborhood is calculated. If this distance is greater than the preset threshold, the point is considered an outlier and removed. Finally, for the data missing area, the method based on local surface reconstruction is used for interpolation and completion, and the data of the missing points is estimated by fitting the local surface. After the above processing, the preprocessed point cloud data is obtained, which lays the foundation for subsequent feature extraction and semantic segmentation.

[0104] Feature extraction is performed on the preprocessed point cloud data. For each point, its normal vector, curvature, and local linear feature descriptor (such as FPFH descriptor) are calculated. The normal vector is obtained by calculating the covariance matrix within the neighborhood of each point and taking the eigenvector corresponding to the smallest eigenvalue. The curvature is obtained by calculating the surface change rate within the neighborhood of each point. The FPFH descriptor is obtained by statistically analyzing the point cloud distribution characteristics within the neighborhood of each point, and it can effectively describe the local geometry. The calculated feature information such as the normal vector, curvature, and FPFH descriptor is combined into a feature vector as the point cloud feature data of this point. The feature vectors of all points constitute the point cloud feature data set, which is used for the training of the subsequent ground object point cloud recognition model.

[0105] The PointNet++ deep learning network architecture is used for model training. The point cloud feature data extracted in step S12 is used as the input, and the preset semantic labels of ground object categories (such as: vegetation, building, road, water area, ground, etc.) are used as the output labels. A training data set is constructed, which contains a large amount of point cloud data and its corresponding semantic labels. The cross-entropy loss function is used as the loss function for model training, and the Adam optimizer is used to optimize the model parameters. The learning rate is set to 0.001, the batch size is 32, and the number of training epochs is 100. During the training process, the performance of the model is evaluated on the validation set regularly, and the model parameters and training strategies are adjusted according to the performance of the validation set. After the training is completed, the trained ground object point cloud recognition model is saved for subsequent terrain semantic segmentation inference.

[0106] The preprocessed point cloud data is input into the PointNet++ ground object point cloud recognition model for inference. The model classifies each point in the point cloud data and assigns a semantic label to each point. For example, points are classified as vegetation, buildings, roads, water areas, or ground, etc. After the inference is completed, each point and its corresponding semantic label are stored in the terrain semantic segmentation model. The terrain semantic segmentation model is represented in the form of a point cloud, and each point is accompanied by its corresponding semantic category label, thus realizing the semantic segmentation of the terrain. This model will be used for subsequent terrain feature extraction, semantic feature extraction, and adaptive chunking, etc.

[0107] Preferably, step S2 includes the following steps:

[0108] Step S21: Extract terrain features from the terrain semantic segmentation model to obtain terrain feature data;

[0109] Step S22: Extract semantic features according to the terrain semantic segmentation model to obtain semantic feature data;

[0110] Step S23: Construct a terrain complexity index based on the terrain feature data and the semantic feature data to obtain terrain complexity index data; perform raster data conversion on the terrain complexity index data to obtain a terrain complexity map;

[0111] Step S24: Perform adaptive chunking according to the terrain complexity map and the terrain semantic segmentation model to obtain multi-resolution chunk data;

[0112] Step S25: Calculate the block attributes of the multi-resolution chunk data to obtain block attribute data;

[0113] Step S26: Construct a multi-resolution index for the multi-resolution chunk data and the block attribute data to obtain a multi-resolution chunk index database.

[0114] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0115] Step S21: Extract terrain features from the terrain semantic segmentation model to obtain terrain feature data;

[0116] In the embodiments of the present invention, terrain features are extracted from the terrain semantic segmentation model. For each point in the model, its slope, curvature, and elevation change rate are calculated. The slope is obtained by calculating the angle between the normal vector of the point and the vertical direction. The curvature is obtained by calculating the change rate of the elevation change rate within the neighborhood of the point, and the quadratic surface fitting method is used for calculation. The elevation change rate is obtained by calculating the elevation difference between the point and the points in its neighborhood, and the neighborhood radius is set to 1 meter. The slope, curvature, and elevation change rate of each point are stored as a three-dimensional vector, which serves as the terrain feature data of the point. The terrain feature data of all points constitutes the terrain feature dataset.

[0117] Step S22: Extract semantic features according to the terrain semantic segmentation model to obtain semantic feature data;

[0118] In the embodiments of the present invention, semantic features are extracted based on the terrain semantic segmentation model. First, the point cloud data is classified into different categories according to the semantic labels of the points, such as vegetation, buildings, roads, and water areas. Then, specific semantic features are calculated for each category. For vegetation, the vegetation density is calculated, which is the number of vegetation points per unit area. For buildings, the height, volume, and floor area of the buildings are calculated. For roads, the width, length, and direction of the roads are calculated, and the direction is obtained by calculating the main direction of the road point cloud through the principal component analysis method. For water areas, the area and average depth of the water areas are calculated. The semantic features of each point are stored as a vector. For example, the feature vector of vegetation contains the vegetation density, and the feature vector of buildings contains the height, volume, and floor area. The semantic feature data of all points constitutes the semantic feature dataset.

[0119] Step S23: Construct a terrain complexity index according to the terrain feature data and the semantic feature data to obtain terrain complexity index data; perform raster data conversion on the terrain complexity index data to obtain a terrain complexity map;

[0120] In the embodiments of the present invention, a terrain complexity index is constructed by combining the terrain feature data and the semantic feature data. The terrain complexity index is defined as the weighted sum of the slope, curvature, elevation change rate, vegetation density, building height, road width, and water area. The weight values are set according to the influence degree of different land cover types on the terrain complexity. For example, the weight of the slope is set to 0.3, the weight of the curvature is set to 0.2, the weight of the elevation change rate is set to 0.1, the weight of the vegetation density is set to 0.1, the weight of the building height is set to 0.15, the weight of the road width is set to 0.1, and the weight of the water area is set to 0.05. Calculate the terrain complexity index of each point and store the result as terrain complexity index data. Then, convert the terrain complexity index data into raster data, and the raster size is set to 1 meter. The value of each raster is the average of the terrain complexity indices of all points within the raster. Finally, a terrain complexity map is obtained, where the value of each raster represents the terrain complexity of the area.

[0121] Step S24: Perform adaptive partitioning based on the terrain complexity map and the terrain semantic segmentation model to obtain multi-resolution partitioned data;

[0122] In the embodiment of the present invention, the point cloud data is adaptively partitioned based on the terrain complexity map and the terrain semantic segmentation model. Areas with high terrain complexity are divided into smaller blocks, and areas with low terrain complexity are divided into larger blocks. Specifically, the terrain complexity map is divided into multiple regions, and different partition sizes are set according to the average terrain complexity of each region. For example, for regions with an average terrain complexity greater than 0.8, the partition size is set to 2 meters; for regions with an average terrain complexity between 0.5 and 0.8, the partition size is set to 5 meters; for regions with an average terrain complexity less than 0.5, the partition size is set to 10 meters. The point cloud data in the terrain semantic segmentation model is partitioned according to the set partition size to obtain multi-resolution partitioned data.

[0123] Step S25: Calculate the block attributes of the multi-resolution partitioned data to obtain block attribute data;

[0124] In the embodiment of the present invention, the attributes of each block obtained in step S24 are calculated, including: the average elevation, average slope, average curvature, main ground object type, number of point clouds, etc. within the block. The average elevation is obtained by calculating the average value of the elevations of all points within the block. The average slope and average curvature are obtained by calculating the average values of the slopes and curvatures of all points within the block, respectively. The main ground object type is obtained by counting the number of points of various ground object types within the block and selecting the ground object type with the largest number as the main ground object type of the block. The number of point clouds refers to the total number of points contained within the block. The calculated block attributes are stored as block attribute data.

[0125] Step S26: Construct a multi-resolution index for the multi-resolution partitioned data and the block attribute data to obtain a multi-resolution partitioned index database;

[0126] In the embodiments of the present invention, an Octree data structure is used to construct a multi-resolution block index. The multi-resolution block data obtained in step S24 is organized into an Octree structure according to its spatial position. The root node of the Octree represents the bounding box of the entire point cloud data. Each non-leaf node represents a spatial region and is divided into eight sub-regions corresponding to eight child nodes. The leaf node stores a pointer to the actual block data and the block attribute data calculated in step S25. The depth of the Octree is adaptively adjusted according to the spatial distribution and resolution of the block data to ensure the efficiency of the index. The constructed Octree index is stored in the multi-resolution block index database for subsequent rapid retrieval and access to data. For example, if a region is divided into blocks of sizes 10 meters, 5 meters, and 2 meters, the leaf nodes of the Octree will respectively point to the block data of these different resolutions and store the corresponding block attribute information, such as average elevation, average slope, main land cover type, and number of point clouds. The structure of the Octree and the information of each node are stored in the database, thus enabling rapid access to the multi-resolution block data.

[0127] Preferably, step S21 includes the following steps:

[0128] Step S211: Determine the neighborhood radius according to the terrain semantic segmentation model to obtain neighborhood radius data;

[0129] Step S212: Perform a classic search according to the neighborhood radius data and the terrain semantic segmentation model to obtain neighboring point data;

[0130] Step S213: Calculate the slope according to the neighboring point data and the terrain semantic segmentation model to obtain slope data;

[0131] Step S214: Calculate the curvature according to the neighboring point data and the terrain semantic segmentation model to obtain curvature data;

[0132] Step S215: Calculate the elevation change rate according to the neighboring point data and the terrain semantic segmentation model to obtain elevation change rate data;

[0133] Step S216: Integrate the terrain features of the slope data, curvature data, and elevation change rate data to obtain terrain feature data.

[0134] In the embodiments of the present invention, the density distribution of the point cloud data in the terrain semantic segmentation model is analyzed. The average distance between each point and its nearest k neighbors is calculated, and the value of k is set to 10. Then, the distribution of the average distances of all points is statistically analyzed, for example, calculating the average value and standard deviation of the average distances. According to the statistical results, the neighborhood radius is set to the average value of the average distances plus the standard deviation. For example, if the average value of the average distances is 0.5 meters and the standard deviation is 0.1 meters, the neighborhood radius is set to 0.6 meters. The calculated neighborhood radius is stored as neighborhood radius data for subsequent neighboring point search. This method can adaptively determine the neighborhood radius according to the density distribution of the point cloud data, ensuring that suitable neighboring points can be found in different density regions.

[0135] The KD-Tree data structure is used to organize the point cloud data in the terrain semantic segmentation model. For each point in the model, using the neighborhood radius, all points located within the neighborhood of this point are searched in the KD-Tree. The search method is to perform a spherical range query centered on this point with the neighborhood radius as the radius. The searched points are stored as the neighboring point data of this point. For example, for a point P with a neighborhood radius of 0.6 meters, the KD-Tree search algorithm will return all points whose distance from point P is less than or equal to 0.6 meters, and these points constitute the neighboring point data of point P. The neighboring point data of all points constitute the neighboring point dataset.

[0136] For each point in the terrain semantic segmentation model, a local plane is fitted using the neighboring point data. The least squares method is used for the fitting. The normal vector of the fitted plane is calculated. The slope is defined as the angle between this normal vector and the vertical direction, and is calculated through the arctangent function. The calculated slope value is stored as the slope data of this point. The slope data of all points constitute the slope dataset.

[0137] For each point in the terrain semantic segmentation model, a local quadratic surface is fitted using the neighboring point data. The least squares method is used for the fitting. The curvature is defined as the average value of the maximum principal curvature and the minimum principal curvature of this quadratic surface. The principal curvatures are obtained by calculating the eigenvalues of the Hessian matrix of the quadratic surface. The calculated curvature value is stored as the curvature data of this point. The curvature data of all points constitute the curvature dataset.

[0138] For each point in the terrain semantic segmentation model, the average value of the elevation difference between this point and its neighboring points is calculated. The elevation change rate is defined as the ratio of this average value to the neighborhood radius. The calculated elevation change rate is stored as the elevation change rate data of this point. The elevation change rate data of all points constitute the elevation change rate dataset.

[0139] Combine the slope data, curvature data, and elevation change rate data into a three-dimensional vector. This vector is the terrain feature data. Each point corresponds to a terrain feature vector, and the feature vectors of all points constitute the terrain feature dataset. For example, the terrain feature vector of a point can be expressed as (slope, curvature, elevation change rate).

[0140] Preferably, step S22 includes the following steps:

[0141] Step S221: Classify the point cloud data according to the semantic labels of each point in the terrain semantic segmentation model to obtain classified point cloud data;

[0142] Step S222: Calculate the vegetation density of the classified point cloud data to obtain vegetation density data;

[0143] Step S223: Calculate the building height and volume of the classified point cloud data to obtain building height and volume data;

[0144] Step S224: Calculate the road width and direction of the classified point cloud data to obtain road width and direction data;

[0145] Step S225: Calculate the water area and average depth of the classified point cloud data to obtain water area and average depth data;

[0146] Step S226: Integrate the semantic features of the vegetation density data, building height and volume data, road width and direction data, and water area and average depth data to obtain semantic feature data.

[0147] In the embodiment of the present invention, each point in the terrain semantic segmentation model is traversed, and its semantic label is read. According to the preset semantic label categories (such as vegetation, building, road, water area, ground, etc.), the point cloud data is divided into different category sets. For example, all points with the label "vegetation" are divided into the vegetation point cloud set, all points with the label "building" are divided into the building point cloud set, and so on. Finally, the point cloud data classified according to the semantic labels is obtained, such as vegetation point cloud data, building point cloud data, road point cloud data, water area point cloud data, etc.

[0148] For the vegetation point cloud data, the vegetation density is calculated using a voxel grid-based method. First, the area where the vegetation point cloud data is located is divided into regular voxel grids, and the voxel size is set to 1 cubic meter. Then, the number of vegetation points in each voxel is counted. The vegetation density is defined as the number of vegetation points in each voxel divided by the volume of the voxel. The calculated vegetation density value is stored as the vegetation density data, and each voxel corresponds to a vegetation density value.

[0149] For building point cloud data, first, calculate the building height by obtaining the maximum and minimum values of the building point cloud in the vertical direction. Then, use the convex hull algorithm to calculate the minimum convex hull of the building point cloud, and the volume of the convex hull is the volume of the building. Store the calculated building height and volume as building height and volume data, with each building corresponding to a height value and a volume value.

[0150] For road point cloud data, first, use the principal component analysis (PCA) method to calculate the principal direction of the road point cloud. The principal direction corresponds to the eigenvector corresponding to the maximum eigenvalue and represents the direction of the road. Then, project the road point cloud onto a plane perpendicular to the principal direction. Calculate the width of the minimum bounding rectangle of the projected point cloud on this plane, and this width is the width of the road. Store the calculated road width and direction as road width and direction data, with each section of road corresponding to a width value and a direction vector.

[0151] For water area point cloud data, first, project the water area point cloud onto a horizontal plane. Calculate the polygon area of the minimum convex hull of the projected point cloud, and this area is the area of the water area. Then, calculate the average elevation of the water area point cloud and subtract the elevation of the lowest point in this area to obtain the average depth of the water area. Store the calculated water area and average depth as water area and average depth data, with each water area corresponding to an area value and an average depth value.

[0152] Integrate the vegetation density data, building height and volume data, road width and direction data, and water area and average depth data together to form semantic feature data. For each point, according to its semantic category, store the corresponding semantic features as a vector. For example, for a vegetation point, its semantic feature vector only contains the vegetation density; for a building point, its semantic feature vector contains the building height and volume; for a road point, its semantic feature vector contains the road width and direction; for a water area point, its semantic feature vector contains the water area and average depth. The semantic feature vectors of all points constitute the semantic feature data set.

[0153] Preferably, step S3 includes the following steps:

[0154] Step S31: Perform attribute sensitivity analysis according to the multi-resolution block index database to obtain an attribute sensitivity level table;

[0155] Step S32: Develop an encryption policy according to the attribute sensitivity level table and the multi-resolution block index database to obtain an encryption policy table;

[0156] Step S33: Encrypt the multi-resolution block data according to the encryption policy table and the multi-resolution block index database to obtain encrypted block data;

[0157] Step S34: Update the encrypted index of the multi-resolution block index database according to the encrypted block data to obtain the updated multi-resolution block index database;

[0158] Step S35: Obtain the system user role data; Define role permissions according to the system user role data to obtain a role permission table;

[0159] Step S36: Develop an attribute access policy according to the attribute sensitivity level table and the role permission table to obtain an attribute access policy table;

[0160] Step S37: Generate an access control list according to the attribute access policy table, the updated multi-resolution block index database, and the encryption policy table to obtain a role- and attribute-based access control list.

[0161] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0162] Step S31: Perform attribute sensitivity analysis according to the multi-resolution block index database to obtain an attribute sensitivity level table;

[0163] In the embodiment of the present invention, analyze the block attribute data stored in the multi-resolution block index database, such as average elevation, average slope, main land use type, and point cloud quantity, etc. According to the security and confidentiality requirements of land surveying and mapping data, divide the sensitivity levels of different attributes. For example, divide the average elevation into a low sensitivity level, the average slope into a medium sensitivity level, the main land use type into a high sensitivity level, and the point cloud quantity into a low sensitivity level. Store the attributes and their corresponding sensitivity levels in the attribute sensitivity level table. The higher the sensitivity level, the higher the confidentiality requirement for the attribute, and stronger encryption protection is required.

[0164] Step S32: Develop an encryption policy according to the attribute sensitivity level table and the multi-resolution block index database to obtain an encryption policy table;

[0165] In the embodiment of the present invention, according to the attribute sensitivity level table obtained in step S31 and the data characteristics in the multi-resolution block index database, develop different encryption policies. For high-sensitivity attributes, use the AES-256 encryption algorithm for encryption, with a key length of 256 bits; for medium-sensitivity attributes, use the AES-128 encryption algorithm for encryption, with a key length of 128 bits; for low-sensitivity attributes, use a lightweight encryption algorithm for encryption, such as XOR encryption, with a key length of 64 bits. Store the different attributes, corresponding encryption algorithms, and key lengths in the encryption policy table.

[0166] Step S33: Encrypt the multi-resolution block data according to the encryption policy table and the multi-resolution block index database to obtain encrypted block data;

[0167] In an embodiment of the present invention, traverse the multi-resolution block index database to obtain the data of each block. Encrypt the data of each block according to the encryption policy table. For example, for a block containing highly sensitive attributes, use the AES-256 encryption algorithm to encrypt it; for a block containing low-sensitive attributes, use the XOR encryption algorithm to encrypt it. The encrypted data is stored as encrypted block data.

[0168] Step S34: Update the encrypted index of the multi-resolution block index database according to the encrypted block data to obtain an updated multi-resolution block index database;

[0169] In an embodiment of the present invention, update the storage location of the encrypted block data into the multi-resolution block index database. At the same time, according to the encryption policy table, store the encryption algorithm and key information of each block into the index database. The updated multi-resolution block index database contains the access path of the encrypted block data and the information required for decryption.

[0170] Step S35: Obtain the system user role data; define role permissions according to the system user role data to obtain a role permission table;

[0171] In an embodiment of the present invention, obtain the predefined user role data in the system, such as administrator, ordinary user, visitor, etc. Define the data attributes and operation permissions that each role can access according to the responsibilities and permissions of different roles. For example, the administrator has access to all data, ordinary users can access some data attributes, and visitors can only access public data. Store the role and its corresponding access permissions in the role permission table.

[0172] Step S36: Develop an attribute access policy according to the attribute sensitivity level table and the role permission table to obtain an attribute access policy table;

[0173] In an embodiment of the present invention, combine the attribute sensitivity level table and the role permission table to develop an attribute access policy. For example, it is stipulated that only administrators can access highly sensitive attributes, ordinary users can access medium-sensitive and low-sensitive attributes, and visitors can only access low-sensitive attributes. Store the attribute sensitivity levels that different roles can access in the attribute access policy table.

[0174] Step S37: Generate an access control list based on the attribute access policy table, the updated multi-resolution block index database, and the encryption policy table to obtain a role- and attribute-based access control list;

[0175] In the embodiments of the present invention, an access control list (ACL) based on roles and attributes is generated according to the attribute access policy table, the updated multi-resolution block index database, and the encryption policy table. The ACL contains the ID of the block data that each role can access, the permitted operations (such as reading and writing), and the key information required for decryption. For example, for an administrator role, the ACL contains the IDs of all block data, as well as all permitted operations and the corresponding decryption keys; for an ordinary user role, the ACL only contains the IDs of the block data that are permitted to be accessed, as well as the corresponding operations and decryption keys.

[0176] Preferably, step S32 includes the following steps:

[0177] Step S321: Determine the preliminary encryption policy according to the attribute sensitivity level table, encryption algorithm, and key length;

[0178] Step S322: Analyze the data analysis requirements of the multi-resolution block index database to obtain data analysis requirement data;

[0179] Step S323: Select a homomorphic encryption algorithm according to the data analysis requirement data and the preliminary encryption policy to obtain homomorphic encryption algorithm selection data;

[0180] Step S324: Generate an encryption policy table according to the preliminary encryption policy and the homomorphic encryption algorithm selection data to obtain the encryption policy table.

[0181] In the embodiments of the present invention, the attribute sensitivity level table generated in step S31 is read. According to different sensitivity levels, the corresponding encryption algorithm and key length are determined. For example, for high-sensitivity attributes, the AES-256 encryption algorithm is specified, and the key length is 256 bits; for medium-sensitivity attributes, the AES-128 encryption algorithm is specified, and the key length is 128 bits; for low-sensitivity attributes, the XOR encryption algorithm is specified, and the key length is 64 bits. The attributes, sensitivity levels, corresponding encryption algorithms, and key lengths are recorded to form a preliminary encryption policy.

[0182] Analyze the data in the multi-resolution block index database to clarify the requirements for data analysis. For example, it is necessary to statistically analyze the average elevation of different regions, calculate the area of specific land cover types, perform spatial queries on point cloud data, etc. These data analysis requirements are recorded to form data analysis requirement data, including the types of analysis to be performed, the data attributes involved, and the accuracy requirements of the analysis.

[0183] Based on the data for data analysis requirements, determine whether it is necessary to perform calculations on encrypted data. If it is necessary to perform calculations on encrypted data, then a suitable homomorphic encryption algorithm needs to be selected. For example, if it is necessary to calculate the average value of encrypted data, the Paillier homomorphic encryption algorithm can be selected. According to the data analysis requirements and the encryption algorithm specified in the preliminary encryption strategy, select a suitable homomorphic encryption algorithm, and record the selected homomorphic encryption algorithm and the applicable data attributes to form homomorphic encryption algorithm selection data. If it is not necessary to perform calculations on encrypted data, then there is no need to select a homomorphic encryption algorithm, and the homomorphic encryption algorithm selection data is empty.

[0184] Combine the preliminary encryption strategy and the homomorphic encryption algorithm selection data to generate the final encryption policy table. The encryption policy table contains information such as the sensitivity level, encryption algorithm, key length, and whether to use homomorphic encryption for each attribute. For example, for a highly sensitive attribute, if it is necessary to perform a summation operation on its encrypted data, record in the encryption policy table that the encryption algorithm for this attribute is AES-256, the key length is 256 bits, and the Paillier homomorphic encryption algorithm is specified to be used. If it is not necessary to perform calculations on encrypted data, only record the encryption algorithm and the key length, and do not record the homomorphic encryption algorithm.

[0185] Preferably, step S4 includes the following steps:

[0186] Step S41: Perform user access history recording according to the role- and attribute-based access control list to obtain a user access history database;

[0187] Step S42: Perform spatial proximity relationship analysis according to the updated multi-resolution block index database to obtain a spatial proximity relationship graph; perform spatial proximity graph construction on the spatial proximity relationship graph to obtain spatial proximity graph data;

[0188] Step S43: Extract classified user access sequences from the user access history database to obtain classified user access sequence data; extract node feature data from the updated multi-resolution block index database and the classified user access sequence data to obtain node feature data;

[0189] Step S44: Construct an access probability prediction model based on a graph neural network model according to the spatial proximity graph data, the classified user access sequence data, and the node feature data to obtain an access probability prediction model;

[0190] Step S45: Obtain the current user access data block ID; use the current user access data block ID as the starting node, and use the access probability prediction model to predict the data block access probability to obtain a predicted access probability table;

[0191] Step S46: Perform spatial-aware cache management based on the predicted access probability table, encrypted chunk data, and role- and attribute-based access control list to obtain spatial-aware cache data.

[0192] In the embodiments of the present invention, the access history of each user is recorded, including access time, user role, accessed data block ID, accessed attributes, and operation type (such as read, write). These access records are stored in a user access history database, for example, stored using a relational database MySQL, and the database table includes fields such as user ID, access time, role, data block ID, attributes, and operation type. The database is regularly backed up and maintained to ensure data integrity and reliability.

[0193] Based on the updated multi-resolution chunk index database, analyze the spatial proximity relationship between data blocks. For each data block, find other data blocks whose spatial distance is less than or equal to a threshold (such as 10 meters), and regard these data blocks as its neighbors. Represent the relationship between the data block and its neighbors as a spatial proximity graph, where nodes represent data blocks and edges represent the proximity relationship between data blocks. Data structures such as adjacency matrices or adjacency lists can be used to store the spatial proximity graph to form spatial proximity graph data.

[0194] Extract the access sequence of each user from the user access history database, that is, a list of data block IDs sorted by access time. Then, classify the user access sequences according to the user role and access purpose (such as data browsing, data analysis, data editing, etc.). For example, classify the access sequences of users with the administrator role into one category, and classify the access sequences of ordinary users for data browsing into one category. Store the classified user access sequences as classified user access sequence data. Next, extract the attribute information of each data block from the updated multi-resolution chunk index database, such as average elevation, average slope, main land cover type, and point cloud quantity. Use these attribute information and the frequency of the data block appearing in the classified user access sequence data as node features to form node feature data.

[0195] Use a graph neural network model, such as GraphSAGE, to construct an access probability prediction model. Use the spatial proximity graph data obtained in step S42 as the graph structure input, the node feature data as the node feature input, and the classified user access sequence data as the training data. The goal of model training is to predict the data blocks that the user may access next. Use the cross-entropy loss function as the loss function for model training, and use the Adam optimizer to optimize the model parameters. After training is completed, an access probability prediction model is obtained.

[0196] Obtain the ID of the data block that the current user is accessing. Use this data block ID as the starting node and input it into the access probability prediction model. The model will predict other data blocks that the user may access next based on spatial proximity relationships, node characteristics, and the user's historical access behavior, and calculate the access probability for each data block. Store the data block ID and its corresponding access probability in the predicted access probability table.

[0197] According to the predicted access probability table, preload the data blocks with higher access probabilities and that meet the user's access permissions into the cache. The cache size is set according to system resource limitations. The cache replacement policy uses the LRU algorithm, that is, the data block that has been least recently used will be preferentially replaced. When accessing the data in the cache, perform permission verification according to the role- and attribute-based access control list, and decrypt the encrypted chunk data according to the encryption policy table to finally obtain the space-aware cache data that the user can access.

[0198] Preferably, step S43 includes the following steps:

[0199] Step S431: Group the users according to the user access history database to obtain user grouping data; perform time sorting on the user grouping data to obtain user time-sorted data;

[0200] Step S432: Extract the access sequence from the user time-sorted data to obtain user access sequence data; perform role and purpose classification based on the user access sequence data and the user time-sorted data to obtain classified user access sequence data;

[0201] Step S433: Extract the data block attributes from the updated multi-resolution block index database to obtain data block attribute data; perform normalization processing on the numerical type attributes of the data block attribute data to obtain normalized data;

[0202] Step S434: Perform one-hot encoding on the categorical type attributes of the data block attribute data to obtain one-hot encoded data; perform missing value processing on the data block attribute data to obtain missing value processed data;

[0203] Step S435: Count the access frequencies of the classified user access sequence data to obtain data block access frequency data; construct node feature vectors from the normalized data, one-hot encoded data, missing value processed data, and data block access frequency data to obtain node feature data.

[0204] In the embodiments of the present invention, user access records are read from the user access history database. Users are grouped according to the user ID, and each user corresponds to a group. Each group contains all the access records of the user. Then, within each group, the access records are sorted in ascending order according to the access time. The sorted access records are arranged in chronological order, reflecting the user's access history. The grouped and sorted results are stored as user time-sorted data.

[0205] The access sequence of each user is extracted from the user time-sorted data. The access sequence is composed of the data block IDs accessed by the user in chronological order. The extracted access sequences are stored as user access sequence data. Then, the user access sequences are classified according to the user role and access purpose. For example, the access sequences of users with the administrator role are classified into one category, the access sequences of ordinary users for data browsing are classified into one category, and the access sequences of ordinary users for data analysis are classified into another category. The access purpose can be inferred by analyzing the operation types (such as reading, writing) of the user's access and the accessed attributes. The classified user access sequence data is stored as classified user access sequence data.

[0206] The attribute information of each data block, such as average elevation, average slope, main land cover type, and point cloud quantity, is extracted from the updated multi-resolution block index database and stored as data block attribute data. For numerical type attributes, such as average elevation, average slope, and point cloud quantity, normalization is performed. The normalization method uses Min-Max normalization to scale the attribute values to the range of [0, 1]. The normalized numerical type attribute data is stored as normalized data.

[0207] For categorical type attributes, such as the main land cover type, one-hot encoding is used for processing. For example, if there are three main land cover types: vegetation, building, and road, then "vegetation" is encoded as [1, 0, 0], "building" is encoded as [0, 1, 0], and "road" is encoded as [0, 0, 1]. The data after one-hot encoding is stored as one-hot encoded data. For missing attribute values, the mean filling method is used for processing. The average value of the attribute in all data blocks is calculated, and the missing values are filled with the average value. The processed data is stored as missing value processed data.

[0208] Count the frequency of each data block in the classified user access sequence data, and store the result as data block access frequency data. Then, combine the normalized data, one-hot encoded data, missing value processed data, and data block access frequency data into a vector as the node feature vector for each data block. Store the node feature vectors of all data blocks as node feature data. For example, the node feature vector of a data block can consist of features such as normalized average elevation, average slope, number of point clouds, main land cover types after one-hot encoding, and access frequency.

[0209] Preferably, the present invention further provides a three-dimensional model data management system for land surveying and mapping, which is used to execute the three-dimensional model data management method for land surveying and mapping as described above. The three-dimensional model data management system for land surveying and mapping includes:

[0210] A terrain semantic segmentation module, which is used to obtain the original three-dimensional point cloud data of land surveying and mapping; perform terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model;

[0211] A multi-resolution block indexing module, which is used to extract terrain features and semantic features from the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and semantic feature data to obtain a terrain complexity map; perform adaptive block division according to the terrain complexity map to obtain multi-resolution block data; construct a multi-resolution index based on the multi-resolution block data to obtain a multi-resolution block index database;

[0212] An encryption storage and access control module, which is used to formulate an encryption policy based on the multi-resolution block index database to obtain an encryption policy table; encrypt the multi-resolution block data according to the encryption policy table and update the encryption index to obtain encrypted block data and an updated multi-resolution block index database; obtain system user role data; perform role- and attribute-based access control based on the system user role data to obtain a role- and attribute-based access control list;

[0213] A spatial awareness cache module, which is used to extract classified user access sequence data according to the role- and attribute-based access control list to obtain classified user access sequence data; extract node feature data according to the updated multi-resolution block index database to obtain node feature data; construct an access probability prediction model based on the classified user access sequence data and the node feature data to obtain an access probability prediction model; use the access probability prediction model to predict the data block access probability to obtain a predicted access probability table; perform spatial awareness cache management based on the predicted access probability table, the encrypted block data, and the role- and attribute-based access control list to obtain spatial awareness cache data.

[0214] Reference Figure 4 As shown in the figure, the process schematic diagram of the 3D model data management system for land surveying and mapping is described as follows:

[0215] 1. Terrain semantic segmentation module:

[0216] Obtain the original 3D point cloud data: Obtain the 3D point cloud data of the land area from devices such as lidar.

[0217] Semantic segmentation processing: Perform terrain semantic segmentation on the original point cloud data to generate a terrain semantic segmentation model.

[0218] 2. Multi-resolution block indexing module:

[0219] Terrain feature extraction: Extract various terrain features (such as slope, curvature, etc.) from the terrain semantic segmentation model.

[0220] Semantic feature extraction: Extract semantic features related to the land cover categories (such as vegetation density, building height, etc.).

[0221] Terrain complexity analysis: Analyze the extracted features to generate a terrain complexity map.

[0222] Adaptive block partitioning: Based on the complexity map, perform adaptive block partitioning on the point cloud data to generate multi-resolution block data.

[0223] Multi-resolution index construction: Establish a multi-resolution block index database for fast retrieval and access.

[0224] 3. Multi-resolution block index database:

[0225] As the data storage and indexing module, it stores terrain feature data, semantic feature data and their index information.

[0226] 4. Encrypted storage and access control module:

[0227] Formulate an encryption strategy: According to the data characteristics and security requirements, formulate a suitable encryption strategy.

[0228] Data block encryption: Encrypt the multi-resolution block data to protect data security.

[0229] Update the encrypted index: Update the location and related information of the encrypted data block in the multi-resolution block index database.

[0230] Role and attribute access control: According to the user role and data sensitivity, formulate an access control policy to generate an access control list (ACL).

[0231] 5. Spatial awareness cache module:

[0232] Extract the user access sequence: Analyze the user's access history and classify the access sequence data.

[0233] Extract the node feature data: Extract the node features from the updated multi-resolution block index database.

[0234] Construct an access probability prediction model: Construct a model based on the user access sequence and node features to predict the blocks that the user may access.

[0235] Predict the access probability of data blocks: Use the model to predict the access probability of data blocks and generate an access probability table.

[0236] Manage the space-aware cache: Perform cache management according to the access probability table and the access control list to ensure that the data blocks with high-frequency access are cached preferentially. Therefore, in any regard, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Thus, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0237] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A three-dimensional model data management method for land surveying and mapping, characterized in that: The following steps are involved: Step S1: obtaining original three-dimensional point cloud data of land surveying and mapping; performing terrain semantic segmentation processing on the original three-dimensional point cloud data to obtain a terrain semantic segmentation model; Step S2: extract terrain features and semantic features from the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and the semantic feature data to obtain a terrain complexity map; perform adaptive block division based on the terrain complexity map to obtain multi-resolution block data; Constructing a multi-resolution index based on the multi-resolution block data to obtain a multi-resolution block index database; Step S3: Formulate an encryption strategy according to the multi-resolution block index database to obtain an encryption strategy table; encrypt the multi-resolution block data according to the encryption strategy table, and update the encryption index to obtain the encrypted block data and the updated multi-resolution block index database; Get system user role data; Perform role-based and attribute-based access control according to system user role data to obtain a role-based and attribute-based access control list; Step S4: extract classified user access sequences according to the access control list based on roles and attributes to obtain classified user access sequence data; extract node feature data according to the updated multi-resolution block index database to obtain node feature data; construct an access probability prediction model according to the classified user access sequence data and the node feature data to obtain an access probability prediction model; predict the access probability of data blocks using the access probability prediction model to obtain a predicted access probability table; perform spatially-aware cache management according to the predicted access probability table, encrypted block data, and the access control list based on roles and attributes to obtain spatially-aware cache data, so as to realize the three-dimensional model data management work for land surveying and mapping.

2. The three-dimensional model data management method for land surveying and mapping according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: obtaining original three-dimensional point cloud data of land surveying and mapping; performing point cloud data preprocessing on the original three-dimensional point cloud data to obtain preprocessed point cloud data; Step S12: extracting point cloud features from the pre-processed point cloud data to obtain point cloud feature data; Step S13: constructing a ground object point cloud recognition model according to the point cloud feature data and preset ground object category semantic labels to obtain a ground object point cloud recognition model; Step S14: Use the terrain point cloud recognition model to perform terrain semantic segmentation reasoning on the pre-processed point cloud data to obtain a terrain semantic segmentation model.

3. The three-dimensional model data management method for land surveying and mapping according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: extracting terrain features from the terrain semantic segmentation model to obtain terrain feature data; Step S22: extracting semantic features according to the terrain semantic segmentation model to obtain semantic feature data; Step S23: constructing a terrain complexity index according to the terrain feature data and the semantic feature data to obtain terrain complexity index data; performing raster data conversion on the terrain complexity index data to obtain a terrain complexity map; Step S24: Adaptively segment the terrain according to the terrain complexity map and the terrain semantic segmentation model to obtain multi-resolution segmented data; Step S25: performing block attribute calculation on the multi-resolution block data to obtain block attribute data; Step S26: construct a multi-resolution index for the multi-resolution block data and the block attribute data to obtain a multi-resolution block index database.

4. The three-dimensional model data management method for land surveying and mapping according to claim 3, characterized in that: Step S21 includes the following steps: Step S211: determining the area radius according to the terrain semantic segmentation model to obtain neighborhood radius data; Step S212: performing a classical search based on the neighborhood radius data and the terrain semantic segmentation model to obtain neighboring point data; Step S213: Calculate the slope based on the neighboring point data and the terrain semantic segmentation model to obtain slope data; Step S214: performing curvature calculation according to the neighboring point data and the terrain semantic segmentation model to obtain curvature data; Step S215: Calculate the elevation change rate based on the neighboring point data and the terrain semantic segmentation model to obtain elevation change rate data; Step S216: performing terrain feature integration on the slope data, the curvature data and the elevation change rate data to obtain terrain feature data.

5. The three-dimensional model data management method for land surveying and mapping according to claim 3, characterized in that: Step S22 includes the following steps: Step S221: classifying the point cloud data according to the semantic label of each point in the terrain semantic segmentation model to obtain classified point cloud data; Step S222: Calculate vegetation density on the classified point cloud data to obtain vegetation density data; Step S223: Calculate the building height and volume of the classified point cloud data to obtain building height and volume data; Step S224: Calculate the road width and direction of the classified point cloud data to obtain road width and direction data; Step S225: performing water area and average depth analysis on the classified point cloud data to obtain water area and average depth data; Step S226: semantic feature integration is performed on the vegetation density data, building height and volume data, road width and direction data, and water area and average depth data to obtain semantic feature data.

6. The three-dimensional model data management method for land surveying and mapping according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing attribute sensitivity analysis according to the multi-resolution block index database to obtain an attribute sensitivity level table; Step S32: Formulate an encryption strategy according to the attribute sensitivity level table and the multi-resolution block index database to obtain an encryption strategy table; Step S33: encrypting the multi-resolution block data according to the encryption strategy table and the multi-resolution block index database to obtain encrypted block data; Step S34: performing encrypted index update on the multi-resolution block index database according to the encrypted block data to obtain an updated multi-resolution block index database; Step S35: Obtain system user role data; define role permissions according to the system user role data to obtain a role permission table; Step S36: Formulate an attribute access policy according to the attribute sensitivity level table and the role authority table to obtain an attribute access policy table; Step S37: Generate an access control list according to the attribute access policy table, the updated multi-resolution block index database and the encryption policy table to obtain an access control list based on roles and attributes.

7. The three-dimensional model data management method for land surveying and mapping according to claim 6, characterized in that: Step S32 includes the following steps: Step S321: Determine the encryption algorithm and key length according to the attribute sensitivity level table to obtain a preliminary encryption strategy; Step S322: performing data analysis requirement analysis on the multi-resolution block index database to obtain data analysis requirement data; Step S323: Select a homomorphic encryption algorithm according to the data analysis requirement data and the preliminary encryption strategy to obtain homomorphic encryption algorithm selection data; Step S324: Select data according to the preliminary encryption strategy and the homomorphic encryption algorithm to generate an encryption strategy table to obtain an encryption strategy table.

8. The three-dimensional model data management method for land surveying and mapping according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Recording user access history according to the access control list based on roles and attributes to obtain a user access history database; Step S42: performing spatial proximity analysis according to the updated multi-resolution block index database to obtain a spatial proximity graph; performing spatial proximity graph construction on the spatial proximity graph to obtain spatial proximity graph data; Step S43: extracting classified user access sequences from the user access history database to obtain classified user access sequence data; extracting node feature data from the updated multi-resolution block index database and the classified user access sequence data to obtain node feature data; Step S44: constructing an access probability prediction model based on a graph neural network model according to the spatial proximity graph data, the classified user access sequence data, and the node feature data to obtain an access probability prediction model; Step S45: obtaining the ID of the data block accessed by the current user; using the ID of the data block accessed by the current user as the starting node, using the access probability prediction model to predict the access probability of the data block, and obtaining a predicted access probability table; Step S46: Perform space-aware cache management according to the predicted access probability table, the encrypted block data, and the access control list based on roles and attributes to obtain space-aware cache data.

9. The three-dimensional model data management method for land surveying and mapping according to claim 8, characterized in that: Step S43 includes the following steps: Step S431: grouping users according to the user access history database to obtain user group data; sorting the user group data by time to obtain user time sorting data; Step S432: extracting the access sequence of the user time-sorted data to obtain user access sequence data; classifying the user access sequence data and the user time-sorted data by role and purpose to obtain classified user access sequence data; Step S433: extracting data block attributes from the updated multi-resolution block index database to obtain data block attribute data; performing normalization processing on the data block attribute data of numerical type attributes to obtain normalized data; Step S434: performing one-hot encoding of the category type attribute on the data block attribute data to obtain one-hot encoded data; performing missing value processing on the data block attribute data to obtain missing value processed data; Step S435: Perform access frequency statistics on the classified user access sequence data to obtain data block access frequency data; construct node feature vectors for the normalized data, one-hot encoded data, missing value processed data, and data block access frequency data to obtain node feature data.

10. A three-dimensional model data management system for land surveying and mapping, characterized in that: Used to execute the three-dimensional model data management method for land surveying as claimed in claim 1, the three-dimensional model data management system for land surveying comprises: The terrain semantic segmentation module is used to obtain the original three-dimensional point cloud data of land surveying and mapping; the original three-dimensional point cloud data is processed for terrain semantic segmentation to obtain a terrain semantic segmentation model; The multi-resolution block indexing module is used to extract terrain features and semantic features from the terrain semantic segmentation model to obtain terrain feature data and semantic feature data; perform terrain complexity analysis based on the terrain feature data and semantic feature data to obtain a terrain complexity map; perform adaptive block segmentation based on the terrain complexity map to obtain multi-resolution block data; perform multi-resolution index construction based on the multi-resolution block data to obtain a multi-resolution block index database; The encryption storage and access control module is used to formulate encryption strategies according to the multi-resolution block index database to obtain an encryption strategy table; encrypt the multi-resolution block data according to the encryption strategy table, and update the encryption index to obtain the encrypted block data and the updated multi-resolution block index database; obtain the system user role data; perform role-based and attribute-based access control according to the system user role data to obtain a role-based and attribute-based access control list; The space-aware cache module is used to extract classified user access sequences according to the access control list based on roles and attributes to obtain classified user access sequence data; extract node feature data according to the updated multi-resolution block index database to obtain node feature data; construct an access probability prediction model according to the classified user access sequence data and the node feature data to obtain an access probability prediction model; use the access probability prediction model to predict the access probability of data blocks to obtain a predicted access probability table; perform space-aware cache management according to the predicted access probability table, encrypted block data and the access control list based on roles and attributes to obtain space-aware cache data.

Images