A GIS 3D Dynamic Modeling and Analysis System and Method Integrating Big Data and Cloud Platform Networks

Abstract

This invention relates to the field of data processing technology, specifically to a GIS 3D dynamic modeling and analysis system and method integrating big data and cloud platform networks. It includes: a multi-source geographic data acquisition unit; a distributed data processing unit; a cloud platform support unit; a 3D dynamic modeling unit for constructing a 3D dynamic model of a geographic scene, achieving real-time updates of the 3D dynamic model based on multi-source data fusion technology; an intelligent analysis unit for performing multi-dimensional geospatial analysis of the 3D model, outputting pattern mining and trend prediction results through improved spatiotemporal analysis algorithms; and a visualization and interaction unit. This invention constructs a dynamic access channel and automated fusion mechanism for multi-source heterogeneous geographic data, integrates a big data distributed parallel processing framework and the elastic resource scheduling capabilities of a cloud platform, breaking the closed-loop limitations of traditional BIM+GIS static data integration, and achieving efficient incremental iteration of large-scale geographic scene 3D models.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a GIS three-dimensional dynamic modeling and analysis system and method that integrates big data and cloud platform networks. Background Technology

[0002] With the development of Geographic Information System (GIS) technology, 3D modeling is gradually upgrading from static scene construction to multi-source data fusion, dynamic updating and intelligent analysis. In particular, after integrating big data and cloud platform technologies, it is necessary to achieve efficient aggregation, parallel processing, real-time modeling and in-depth analysis of geographic data to meet the dynamic management needs of complex geographic scenes.

[0003] Chinese patent CN202111170773.0 discloses a method and application for establishing a temporary construction model based on the integration of 3D GIS and BIM. It establishes a BIM model, collects 3D GIS information, imports the 3D GIS information into a BIM platform using a classification and transformation method, integrates the BIM model data with the transformed GIS information to establish a 3D integrated temporary construction model, and combines construction information with a BIM5D platform to simulate the construction process and construct a 5D temporary construction model. Chinese patent CN202310366333.5 discloses a method and system for automatic digital twin modeling based on satellite imagery. This includes acquiring satellite imagery of the target area and performing GIS modeling to generate a digital twin model. After transmitting the model to a data backend for processing, it is displayed in various formats, while aerial photography data is used to supplement details and improve the model's completeness.

[0004] Despite the design advantages of the aforementioned technical solutions, they also exhibit significant technical shortcomings: First, static integration is limited, lacking dynamic response and cloud adaptation: CN202111170773.0's model construction relies on a BIM platform, and data integration is limited to the static integration of BIM and GIS. It lacks a dynamic access and automated fusion mechanism for multi-source heterogeneous geographic data (such as LiDAR point clouds and real-time UAV imagery), and it does not involve distributed big data processing and elastic resource scheduling on cloud platforms, making it difficult to cope with the high-frequency update requirements of large-scale geographic scenarios. Second, data sources are singular, lacking multi-source fusion and real-time update mechanisms: CN202310366333.5's modeling is centered on satellite images, with a singular data source. It fails to address the issues of spatiotemporal alignment and deviation correction of multi-source data, and model processing relies on centralized processing in the data backend, lacking dynamic load balancing and cross-regional data synchronization capabilities. It cannot achieve real-time updates and intelligent spatiotemporal analysis of the 3D model, making it difficult to meet the actual requirements of complex geographic scenarios for model dynamism and analytical depth. In view of this, we propose a GIS 3D dynamic modeling and analysis system and method that integrates big data and cloud platform networks. Summary of the Invention

[0005] The purpose of this invention is to provide a GIS 3D dynamic modeling and analysis system and method that integrates big data and cloud platform networks to solve the problems mentioned in the background art.

[0006] To address the aforementioned technical problems, one objective of this invention is to provide a GIS 3D dynamic modeling and analysis system that integrates big data and cloud platform networks, comprising: The multi-source geographic data acquisition unit is used to achieve the aggregation and access of multiple types of geospatial data, and ensures data validity through standardized preprocessing procedures and quality control mechanisms; The distributed data processing unit is used to perform parallel feature extraction and spatiotemporal correlation calculation on geographic data, and improves data processing efficiency based on a dynamic load balancing architecture. The cloud platform support unit is used to provide elastic resource scheduling and cross-regional data storage services, and ensures high system availability through a hybrid cloud collaboration mechanism; The 3D dynamic modeling unit is used to construct a 3D dynamic model of a geographic scene and achieves real-time updates of the 3D dynamic model based on multi-source data fusion technology. The intelligent analysis unit is used to perform multi-dimensional geospatial analysis on the 3D model and outputs pattern mining and trend prediction results through improved spatiotemporal analysis algorithms. The visualization and interaction unit is used to realize the multi-terminal presentation and interactive operation of 3D scenes, and improves multi-device adaptability based on adaptive rendering technology.

[0007] As a further improvement to this technical solution, the multi-source geographic data acquisition unit includes a data protocol adaptation module, a data transmission control module, a metadata extraction module, and a data source management module, wherein: The data protocol adaptation module integrates a multi-protocol conversion component to achieve standardized conversion of protocols from different data sources; The data transmission control module is used to dynamically adjust the transmission rate to adapt to the network bandwidth and supports the recovery of transmission from the point of data interruption. The metadata extraction module is used to automatically parse the header information of various types of data and extract metadata fields including collection time, spatial coordinate range, and data precision level. The data source management module is used to provide data source registration, status monitoring and abnormal alarm functions. When the data source connection is abnormal, an alarm mechanism is triggered.

[0008] As a further improvement to this technical solution, in the data transmission control module, resuming transmission from the point of data interruption includes the following steps: S120.1 Monitoring link connectivity via periodic heartbeat packets: The number of consecutive unresponsive heartbeat cycles ( (The default value is 3) The threshold is reached, and the duration of a single heartbeat cycle is... If there is no response within 1–10 seconds (configurable range), the transmission is considered interrupted. Synchronously record the absolute byte position of the currently transmitted data. This serves as a reference benchmark for the starting point of recovery. S120.2, Divide the transmitted data into fixed block sizes. Divided into data blocks and execute: Calculate the SHA-256 hash value for each data block. (Unique identifier for data block content); Generate block sequence number SHA-256 hash value Block start byte position Interrupt status table For subsequent comparison; S120.3 Send a recovery request to the data sender, carrying the interrupt status table. ; The sender compares the hash table of the local data blocks. The formula is used to locate the sequence number of the first block with a hash mismatch. ; ; in, The hash value of the local data block at the sending end; Sequential numbering of data blocks; The total number of blocks into which the transmitted data has been divided; Determine the starting byte position for breakpoint recovery ; ; in, For the first The starting byte position of the block data; Fixed block size; S120.4. A sliding window mechanism is used to retransmit missing data blocks in parallel, with a window size of [missing information]. Dynamically adjusted to: ; in, For network round-trip latency; This represents the currently available bandwidth. This represents the maximum number of parallel windows. This is a floor function; S120.5 After the resume download is complete, execute: Calculate the MD5 hash of the complete data (Globally unique identifier for data integrity); The checksum preset at the sending end Comparison: like The transmission was deemed complete. If there is a mismatch, backtrack to step S120.3, relocate the local mismatch block, and repeatedly execute the resume-verification until the hash matches.

[0009] As a further improvement to this technical solution, the distributed data processing unit includes a data sharding and scheduling module, a feature extraction module, a spatiotemporal correlation calculation module, a load balancing control module, and a communication module, wherein: The data sharding scheduling module divides geographic data into spatial shards with fixed granularity based on spatial grid index and associates them with unique spatial identifiers. It maintains a shard-node mapping table to support dynamic shard migration and load balancing, and realizes version control and incremental updates of sharded data. The feature extraction module integrates a terrain feature extractor and a land feature extractor to identify features of terrain parameters, buildings, roads, water bodies, and vegetation. It supports multi-scale feature extraction and allows configuration of extraction accuracy. A feature caching mechanism avoids redundant calculations. The spatiotemporal correlation calculation module is used to construct a spatiotemporal index structure to support millisecond-level spatiotemporal range queries, and uses spatial proximity analysis and time series analysis algorithms to realize multi-source data correlation and fusion. The load balancing control module is used to monitor the CPU, memory, and IO utilization of computing nodes in real time, and dynamically adjust task allocation and trigger task migration based on the node load status. The communication module uses a publish-subscribe message queue to achieve loosely coupled communication between modules, supporting the transmission of processing status and result data.

[0010] As a further improvement to this technical solution, the cloud platform support unit includes an elastic resource scheduling module and a hybrid cloud collaboration module, wherein: The elastic resource scheduling module dynamically allocates computing resources based on container orchestration technology, supports horizontal and vertical scaling of service instances, builds a resource demand prediction model, predicts resource usage trends through historical load data and business cycle characteristics, and enables cross-availability zone failover. Furthermore, cross-availability zone failover includes the following steps: First, an HTTP heartbeat request is sent to the service instance every 10 seconds. If the request times out three times in a row, the instance is considered to be faulty. Then, in the event of a failure in the primary availability zone, the service instances are automatically migrated to the backup availability zone via Kubernetes cluster federation. Subsequently, update the service's DNS resolution or load balancer endpoint to ensure that traffic is switched to the backup instance; Finally, after the primary availability zone is restored, the resources of the backup instance are automatically reclaimed.

[0011] The hybrid cloud collaboration module is used to build a two-way data synchronization channel between private cloud and public cloud, supports encrypted transmission and breakpoint resume, realizes unified management of multi-cloud resources, shields the differences in APIs of different cloud vendors through an abstraction layer, intelligently schedules multi-cloud resources based on business priorities, deploys critical tasks on private cloud, and allocates elastic tasks to idle resources on public cloud.

[0012] As a further improvement to this technical solution, the cloud platform support unit also includes a hierarchical storage module and a cross-region synchronization module, wherein: The tiered storage module adopts a tiered storage architecture; the tiered storage architecture specifically includes a hot data layer, a warm data layer, and a cold data layer, wherein: The hot data layer stores frequently accessed data through a distributed in-memory database; The warm data layer stores recent data through an object storage system; The cold data layer stores historical archived data through a tape library, supporting retrieval and batch recovery by timestamp.

[0013] The cross-regional synchronization module ensures data consistency among multiple geographical regions based on a consistency algorithm, constructs a data synchronization priority queue, configures the synchronization cycle according to data timeliness, implements data compression and encrypted transmission, and feeds back the data synchronization status to the elastic resource scheduling module.

[0014] As a further improvement to this technical solution, the three-dimensional dynamic modeling unit includes a multi-source data fusion module, a three-dimensional model construction module, and a model dynamic update module, wherein: The multi-source data fusion module is used to access multiple types of geographic data (such as LiDAR point clouds, UAV imagery, GIS vectors, etc.). Through format parsing, coordinate calibration and feature matching algorithms, it spatially aligns and semantically associates the multi-source data to achieve fusion. If there is a spatial deviation in the data, the deviation is corrected through feature matching. The 3D model building module constructs a terrain triangular mesh and a parametric model of ground features based on the fused data, and adapts to different rendering accuracies through a LOD hierarchical strategy, thereby generating a 3D scene with semantic information. The model dynamic update module is used to compare geographic data at different timestamps, detect areas of scene change, calculate incremental update data, and inject it into the terrain triangulation mesh and the parametric model of ground features to refresh the status of the terrain triangulation mesh and the parametric model of ground features in real time.

[0015] As a further improvement to this technical solution, the intelligent analysis unit outputs pattern mining and trend prediction results through an improved spatiotemporal analysis algorithm, including the following steps: S500.1, Spatiotemporal Feature Extraction: Spatiotemporal representations of 3D geographic entities are constructed using a spatiotemporal cube model, and spatiotemporal feature vectors are extracted. : ,in, For spatial feature vectors; This is a time feature vector; For attribute feature vectors; S500.2, Spatiotemporal correlation modeling: Constructing the spatiotemporal correlation matrix Quantifying the spatiotemporal dependencies between geographic entities: ; in, The spatial distance function can be Euclidean distance or network distance. The time distance function uses timestamp differences or time interval ratios. The attribute similarity function can be cosine similarity or edit distance. This is the combined distance function; For the first , The spatiotemporal correlation of an entity; For the spatial feature vector of a geographic entity; The temporal feature vector of a geographic entity; S500.3, Spatiotemporal Pattern Recognition: Based on spatiotemporal correlation matrix An improved spatiotemporal clustering algorithm is used to identify pattern clusters. : ; in, This is a penalty coefficient that controls the balance between cluster compactness and separation. The penalty function is defined based on the cluster diameter or density. For the first A spatiotemporal pattern cluster; To find the cluster that maximizes the expression within the parentheses ; For pattern clusters Geographic entity pairs within; S500.4 Trend Prediction Model: Constructing a spatiotemporal regression model to predict future states ; ; in, This represents the spatiotemporal feature vector at the current moment; The weight matrix is ​​obtained through training with historical data; As a bias vector, it captures the inherent trend of the system; This is the random error term; S500.5 Result Verification and Output: The prediction accuracy is evaluated through cross-validation, and the mean squared error is calculated. : ; in, To verify the sample size; for The real spatiotemporal feature vector at any given moment; for The predicted spatiotemporal feature vector at any given time; when When the threshold is reached, the prediction result is output. and confidence interval ,in The confidence interval is half-width.

[0016] As a further improvement to this technical solution, the visualization interaction unit includes a multi-terminal rendering module, an interaction control module, a data visualization module, a performance optimization module, and an event bus module, wherein: The multi-terminal rendering module is used to achieve cross-terminal adaptive rendering of 3D scenes, and adjusts display parameters according to device characteristics based on adaptive layout technology; The interactive control module is used to implement multimodal interactive operations and respond to user selections and operation commands based on the interactive event processing mechanism; The data visualization module is used to visualize thematic maps and feature plotting, and converts geographic information into graphical representations based on data mapping technology. The performance optimization module is used to improve rendering efficiency and optimize the display effect of complex models based on scene simplification strategies; The event bus module is used to realize communication and collaboration between modules, and to distribute and process interactive events based on the message passing mechanism.

[0017] The second objective of this invention is to provide a GIS 3D dynamic modeling and analysis method that integrates big data and cloud platform networks. Based on the aforementioned GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks, the method includes the following steps: S100 Multi-source geographic data aggregation and access: Aggregates and accesses multiple types of geospatial data, adapts to different data sources through standardized protocol conversion, dynamically adjusts the transmission rate and supports breakpoint recovery of transmission, automatically extracts metadata containing information such as collection time and spatial coordinates, registers data sources, monitors their status and provides anomaly alarms, and ensures data validity through standardized preprocessing and quality control. S200, Distributed Geographic Data Processing: Parallel feature extraction and spatiotemporal correlation calculation of geographic data; spatial partitioning based on spatial grid index and associated with unique identifiers; extraction of multiple types of features such as terrain and buildings and support for multi-scale configuration; construction of spatiotemporal index to support millisecond-level queries; real-time monitoring of computing node load and dynamic adjustment of task allocation; and inter-module communication through message queues to improve data processing efficiency. S300, cloud platform resources and storage support: provides elastic resource scheduling and cross-regional data storage services, dynamically allocates computing resources based on container orchestration and supports service expansion, builds a private cloud-public cloud bidirectional synchronization channel and schedules multi-cloud resources according to business priority, and adopts a layered storage architecture to store high-frequency access, recent and historical data respectively. S400, 3D dynamic model construction and updating: Constructing terrain triangular mesh and parametric model of geographic scene, performing format parsing, coordinate calibration and feature matching on multiple types of geographic data to achieve fusion and correct spatial deviation, constructing terrain and feature models based on fused data and adapting to different rendering precision through hierarchical strategy, detecting scene changes by comparing data with different timestamps, calculating incremental data and injecting it into the model to refresh the status in real time. S500, 3D model intelligent analysis: performs multi-dimensional geospatial analysis on terrain triangular mesh and parametric model of land features, extracts spatial, temporal and attribute feature vectors through spatiotemporal cube model, constructs spatiotemporal correlation matrix to quantify the dependency relationship between geographic entities, adopts improved clustering algorithm based on correlation matrix to identify spatiotemporal pattern clusters, constructs regression model to predict future state, evaluates accuracy through cross-validation and outputs prediction results and confidence interval, realizing pattern mining and trend prediction; S600, 3D Scene Visualization and Interaction: Enables multi-terminal presentation and interactive operation of 3D scenes. Based on adaptive rendering technology, it adjusts display parameters according to device characteristics to adapt to multiple terminals. It supports multi-modal interactive operation to respond to user commands, converts geographic information into graphic representation to realize thematic maps and feature plotting display, optimizes the display effect of complex models to improve rendering efficiency, and realizes inter-module communication and collaboration through message passing mechanism.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention constructs a dynamic access channel and automated fusion mechanism for multi-source heterogeneous geographic data, integrates a big data distributed parallel processing framework and cloud platform elastic resource scheduling capabilities, breaks the closed-loop limitation of traditional BIM+GIS static data integration, supports real-time processing of frequently updated geographic information, and realizes efficient incremental iteration of large-scale geographic scene 3D models. 2. This invention establishes a spatiotemporal alignment and deviation correction system for multi-source data such as satellite remote sensing, GIS vectors, and aerial imagery, and combines dynamic load balancing and cross-regional data synchronization technologies to replace the centralized data processing mode; it also embeds intelligent analysis modules such as spatiotemporal correlation mining and trend prediction to achieve real-time status updates and in-depth geographic pattern analysis of the 3D model, meeting the dual requirements of dynamism and analytical dimensions in complex geographic scenarios. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system framework of the present invention; Figure 2 This is a schematic diagram of the method steps of the present invention; The meanings of the labels in the diagram are as follows: 100. Multi-source geographic data acquisition unit; 110. Data protocol adaptation module; 120. Data transmission control module; 130. Metadata extraction module; 140. Data source management module; 200. Distributed data processing unit; 210. Data sharding and scheduling module; 220. Feature extraction module; 230. Spatiotemporal correlation calculation module; 240. Load balancing control module; 250. Communication module; 300. Cloud Platform Support Unit; 310. Elastic Resource Scheduling Module; 320. Hybrid Cloud Collaboration Module; 330. Tiered Storage Module; 340. Cross-Region Synchronization Module; 400. 3D dynamic modeling unit; 410. Multi-source data fusion module; 420. 3D model construction module; 430. Model dynamic update module; 500. Intelligent Analysis Unit; 600. Visualization and Interaction Unit; 610. Multi-Terminal Rendering Module; 620. Interactive Control Module; 630. Data Visualization Module; 640. Performance Optimization Module; 650. Event Bus Module. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 like Figure 1 As shown, this embodiment provides a GIS 3D dynamic modeling and analysis system that integrates big data and cloud platform networks, including: The multi-source geographic data acquisition unit 100 is used to realize the aggregation and access of multiple types of geospatial data, and ensures the validity of data through standardized preprocessing procedures and quality control mechanisms. In this embodiment, the multi-source geographic data acquisition unit 100 includes a data protocol adaptation module 110, a data transmission control module 120, a metadata extraction module 130, and a data source management module 140, wherein: The data protocol adaptation module 110 integrates a multi-protocol conversion component to achieve standardized conversion of protocols from different data sources; As a further explanation of this embodiment, the data protocol adaptation module 110 in this embodiment integrates a multi-protocol conversion component, and adopts differentiated protocol processing logic for different types of geographic data, as follows: First, for LiDAR point cloud data, the LAS protocol is used for parsing. By parsing the point cloud quantity, coordinate reference system, and other identifiers in the file header, the X / Y / Z coordinate fields of the point cloud are converted into the system's internal standard "absolute 3D coordinate" format. For UAV imagery data, spatial reference information (such as projected coordinate system and resolution) is extracted based on the GeoTIFF protocol, and the image pixel values ​​are reassembled into a unified raster data structure according to band order. For GIS vector data, feature information is obtained through the WFS protocol interface, and the attribute fields of vector features (such as "feature type" and "area") are mapped to the system's preset key-value pair format (such as "feature_type" and "area"). For real-time monitoring data (such as dynamic terrain data), the MQTT protocol is used to achieve lightweight transmission, and data types are distinguished by subscribing to topics to ensure real-time performance. Secondly, during the protocol conversion process, non-standard fields are standardized through a pre-defined "protocol-field mapping table": for example, the "gps_time" field in the LAS protocol and the "datetime" field in the GeoTIFF protocol are both mapped to the "collect_time" field inside the system to ensure the consistency of time information; for coordinate fields, regardless of whether the original protocol uses relative or absolute coordinates, they are converted to absolute coordinates (X / Y / Z) in the WGS84 coordinate system. The conversion process is achieved by calling the Proj.4 library to adapt the projection parameters.

[0022] The data transmission control module 120 is used to dynamically adjust the transmission rate to adapt to the network bandwidth and supports the recovery of transmission from the point of data interruption. As a further explanation of this embodiment, the data transmission control module 120 in this embodiment monitors the network status in real time and adjusts the transmission rate, specifically including: Network status acquisition: The network round-trip time (RTT) and current available bandwidth (BW) are acquired every second by sending a 1KB probe packet from the sender to the receiver. RTT is the time difference between sending the probe packet and receiving the response, and BW is the number of bytes successfully transmitted per unit time / time interval (the time interval is set to 1 second). Rate adjustment rules: When BW utilization (actual transmission rate / BW) > 80%, reduce the transmission rate by 20%; when BW utilization < 30%, increase the transmission rate by 10% to ensure that transmission efficiency matches network load.

[0023] The metadata extraction module 130 is used to automatically parse the header information of various types of data and extract metadata fields including collection time, spatial coordinate range, and data precision level; As a further explanation of this embodiment, the metadata extraction module 130 in this embodiment automatically parses the header information of different types of data and extracts key metadata fields: First, for LiDAR point cloud data, parse the "acquisition time" (taken from GPS timestamp), "spatial coordinate range" (calculated by traversing the X / Y / Z coordinates of the point cloud), and "data accuracy level" in the file header; Secondly, for UAV image data, extract "shooting time" and "sensor model" from EXIF ​​information, calculate "spatial coordinate range" by combining the four corner coordinates of the image, and determine "data accuracy level" based on the ground resolution of the image (e.g., 1 pixel corresponds to the actual distance). Finally, all extracted metadata is stored in JSON format, containing fixed fields such as "data ID" and "source type" to facilitate quick querying of basic data attributes by subsequent modules.

[0024] The data source management module 140 is used to provide data source registration, status monitoring and abnormal alarm functions, and triggers an alarm mechanism when the data source connection is abnormal.

[0025] As a further explanation of this embodiment, the data source management module 140 in this embodiment realizes full lifecycle management of the data source through registration, monitoring, and alarms, specifically including: Data source registration: Users enter the data source name, protocol type, access address and authentication information (such as username and password) through the system interface. The system automatically sends a test request (attempting to obtain 1KB of test data). After the test is passed, a unique data source ID is assigned and the data source information is stored in a relational database (such as MySQL) to support subsequent queries and modifications. Status monitoring: Real-time collection of data source connection status (online / offline), data transmission rate (average over the past minute), and data integrity (percentage of successfully parsed data), with a monitoring frequency set to once every 30 seconds; Anomaly Alarm: When the system detects that the data source has been offline for more than 5 minutes or the data integrity is less than 90%, the system triggers an alarm mechanism—recording alarm details (including data source ID, anomaly type, and occurrence time) in the local log database, and sending a notification to the preset administrator email address. The notification includes an anomaly description and preliminary troubleshooting suggestions (such as "Please check the data source network connection").

[0026] In this embodiment, the data transmission control module 120 includes the following steps for resuming transmission from the point of data interruption: S120.1 Monitoring link connectivity via periodic heartbeat packets: The number of consecutive unresponsive heartbeat cycles ( (The default value is 3) The threshold is reached, and the duration of a single heartbeat cycle is... If there is no response within 1–10 seconds (configurable range), the transmission is considered interrupted. Synchronously record the absolute byte position of the currently transmitted data. This serves as a reference benchmark for the starting point of recovery. S120.2, Divide the transmitted data into fixed block sizes. Divided into data blocks and execute: Calculate the SHA-256 hash value for each data block. (Unique identifier for data block content); Generate block sequence number SHA-256 hash value Block start byte position Interrupt status table For subsequent comparison; S120.3 Send a recovery request to the data sender, carrying the interrupt status table. ; The sender compares the hash table of the local data blocks. The formula is used to locate the sequence number of the first block with a hash mismatch. ; ; in, The hash value of the local data block at the sending end; Sequential numbering of data blocks; The total number of blocks into which the transmitted data has been divided; Determine the starting byte position for breakpoint recovery ; ; in, For the first The starting byte position of the block data; Fixed block size; For example, when hour, byte.

[0027] S120.4. A sliding window mechanism is used to retransmit missing data blocks in parallel, with a window size of [missing information]. Dynamically adjusted to: ; in, For network round-trip latency; This represents the currently available bandwidth. This represents the maximum number of parallel windows. This is a floor function; S120.5 After the resume download is complete, execute: Calculate the MD5 hash of the complete data (Globally unique identifier for data integrity); The checksum preset at the sending end Comparison: like The transmission was deemed complete. If there is a mismatch, backtrack to step S120.3, relocate the local mismatch block, and repeatedly execute the resume-verification until the hash matches.

[0028] Furthermore, the internal collaborative process of the multi-source geographic data acquisition unit 100 in this embodiment is as follows: The data source management module 140 first completes the registration of the data source and monitors its status, and only sends acquisition instructions to data sources that are "online and in normal status"; the data protocol adaptation module 110 performs protocol conversion on the acquired raw data, and the converted standard data is transmitted to the data transmission control module 120; during the transmission process, the data transmission control module 120 synchronizes information such as the sequence number and starting position of the data block to the metadata extraction module 130, so that it can associate the "transmission status" (such as "completely transmitted" or "partially transmitted") when extracting metadata; the operation logs of all modules are uniformly stored in the unit-level log library to provide traceability basis for data quality control.

[0029] The distributed data processing unit 200 is used to perform parallel feature extraction and spatiotemporal correlation calculation on geographic data, and improves data processing efficiency based on a dynamic load balancing architecture. In this embodiment, the distributed data processing unit 200 includes a data sharding scheduling module 210, a feature extraction module 220, a spatiotemporal correlation calculation module 230, a load balancing control module 240, and a communication module 250, wherein: The data sharding scheduling module 210 divides geographic data into spatial shards with fixed granularity based on the spatial grid index and associates them with unique spatial identifiers. It maintains a shard-node mapping table to support dynamic migration and load balancing of shards and realizes version control and incremental updates of sharded data. As a further explanation of this embodiment, the data sharding scheduling module 210 in this embodiment divides geographic data into spatial shards of fixed granularity based on the spatial grid index. The specific implementation process is as follows: S210.1 Spatial grid generation and unique identifier generation: According to the formula The grid granularity is dynamically determined; among which... Minimum grid size (preset 50 meters). The total spatial extent of the geographic data. To calculate the number of nodes, The maximum number of shards that a single node can carry (preset 20); Taking the lower left corner of the study area as the origin The geographic space is divided into a two-dimensional grid, and each grid generates a unique identifier. ;in, ; In the formula: The grid number in the X direction; X and Y are the grid number in the Y direction; X and Y are the planar coordinates of the target point in geographic space. , The coordinates of the origin at the lower left corner of the study area are given. This represents the side length of the spatial grid. For cross-grid geographic entities, a "primary fragment + reference" mechanism is adopted: the grid where the entity's center of gravity is located is the primary fragment, and other intersecting grids store the entity ID and reference pointer.

[0030] S210.2, Sharding-Node Mapping and Dynamic Migration: Maintaining a distributed mapping table Stored in Redis, supporting dynamic migration of shards; During migration, the mapping table status is first updated to "migration in progress", and then updated to the new node ID after data replication is complete to ensure read-write consistency. S210.3, Version Control and Incremental Updates: Assign a version number to each shard =Time stamp + Number of updates; Incremental data is identified by comparing spatial range intersection and feature hash, and only the differences are transmitted. The receiving end merges the increments according to the version number.

[0031] The feature extraction module 220 integrates a terrain feature extractor and a land feature extractor to identify features of terrain parameters, buildings, roads, water bodies, and vegetation. It supports multi-scale feature extraction and allows configuration of extraction accuracy. It also avoids redundant calculations through a feature caching mechanism. As a further explanation of this embodiment, the feature extraction module 220 in this embodiment integrates a terrain and feature extractor, supports multi-scale configuration and cache optimization, specifically including: Terrain feature extraction algorithm: Slope calculation: based on elevation values ​​within a 3×3 window The calculation formula is: ; in: ; ; In the formula: The slope of the terrain; The elevation gradient in the X direction; The elevation gradient in the Y direction; For grid points Elevation value at the location; For grid points Elevation value at the location; For grid points Elevation value at the location; For grid points Elevation value at the location; For grid points Elevation value at the location; For grid points Elevation value at the location; For grid points Elevation value at the location; This represents the side length of the spatial grid. Slope aspect calculation: The value range is 0°-360° (true north is 0°); where: The slope aspect of the terrain; It is a two-parameter arctangent function; Ground feature extraction process: Building: Extract the outline through edge detection and morphological operations, and calculate the floor area and height (elevation difference between the highest and lowest points). Roads: The centerline is extracted using a skeletonization algorithm, the width is calculated perpendicular to the centerline, and the material is classified by combining image spectral features; Water areas: Boundaries are determined by threshold segmentation and void filling, and the area and perimeter are calculated. Vegetation: Coverage was extracted based on NDVI index (≥0.3), and average height was calculated using point cloud elevation.

[0032] Multi-scale configuration and caching mechanism: It supports "low / medium / high" precision configuration. Low precision extracts terrain features and only identifies the outline of ground features using a 100m grid. High precision calculates terrain parameters and extracts detailed attributes of ground features using a 10m grid. An LRU caching strategy is adopted, using "spatial shard ID + feature type + precision level" as the key (e.g., "shard_123_terrain_high") to store feature data. The cache capacity is 30% of the node memory, and the corresponding cache is automatically invalidated when the shard version is updated.

[0033] The spatiotemporal correlation calculation module 230 is used to construct a spatiotemporal index structure to support millisecond-level spatiotemporal range queries, and uses spatial proximity analysis and time series analysis algorithms to achieve multi-source data correlation and fusion; As a further explanation of this embodiment, the spatiotemporal correlation calculation module 230 in this embodiment realizes multi-source data fusion through spatiotemporal indexing and correlation algorithms, specifically including: Spatiotemporal index structure construction: A composite index of "spatial R-tree + temporal B-tree" is adopted: the spatial R-tree organizes the hierarchical structure according to the entity MBR, and the leaf nodes store the entity ID and coordinates; the temporal B-tree uses the timestamp as the key, and the nodes are associated with the spatial R-tree pointers. The index file is stored in HDFS and supports parallel loading. During construction, an R-tree is first built according to the spatial MBR, and then a B-tree is built according to the timestamp.

[0034] Optimization of spatiotemporal range query: Input the spatiotemporal range conditions (such as "2024-07-10 to 2024-07-15, 116°-116.1° East longitude, 39.9°-40° North latitude"), first filter the entity IDs that match the time through a time B-tree, and then filter the entities that match the space through a spatial R-tree. The average query time is ≤50ms.

[0035] Spatial proximity and time series analysis: Spatial proximity analysis: The KNN algorithm is used to calculate the Euclidean distance between the target entity and surrounding entities. Candidate entities with a distance ≤ 500m are pre-filtered using a spatial R-tree. Then, the exact distance is calculated and the K nearest entities are returned (K defaults to 5). Time series analysis: For multiple periods of data for the same entity, a sliding window (window size 30 days) is used to calculate trend values, such as height growth rate; Multi-source data association: Data from different sources (such as drone imagery outlines and LiDAR altitude) are fused by the condition of "spatial intersection ≥70% and time difference ≤1 hour".

[0036] The load balancing control module 240 is used to monitor the CPU, memory, and IO utilization of computing nodes in real time, and dynamically adjust task allocation and trigger task migration based on the node load status. As a further explanation of this embodiment, the load balancing control module 240 achieves node load balancing through real-time monitoring and dynamic adjustment, specifically including: Load monitoring and threshold setting: Collect node CPU utilization every 5 seconds Memory utilization IO utilization ; High load threshold: or or ; Low load threshold: and and .

[0037] Task allocation and migration algorithms: Task allocation: based on node load rate Assign new tasks, select The smallest node; Task migration: When a node is under continuous high load (≥30 seconds) and there are low-load nodes, prioritize migrating compute-intensive tasks to the node with the lowest load rate. The original node task is paused, intermediate results are copied, and execution is resumed. At the same time, the task-node mapping table is updated.

[0038] The communication module 250 uses a publish-subscribe message queue to achieve loosely coupled communication between modules, supporting the transmission of processing status and result data.

[0039] Furthermore, the collaborative process of the distributed data processing unit 200 in this embodiment is as follows: the data sharding scheduling module 210 divides the geographic data into spatial shards and publishes the shard-node mapping through a message queue; the feature extraction module 220 receives the sharded data, extracts features according to the configuration and caches the results, and transmits them to the spatiotemporal correlation calculation module 230 through a message queue; the spatiotemporal correlation calculation module 230 constructs a spatiotemporal index, performs spatial proximity and time series analysis, and outputs the correlation results; the load balancing control module 240 monitors the node load in real time, and when a high-load node is detected, it publishes a migration command through a message queue; the communication module 250 ensures that all operation statuses, such as shard migration progress and feature extraction status, are synchronized through a message queue to achieve full-process traceability.

[0040] The cloud platform support unit 300 is used to provide elastic resource scheduling and cross-regional data storage services, and ensures high system availability through a hybrid cloud collaboration mechanism; In this embodiment, the cloud platform support unit 300 includes an elastic resource scheduling module 310 and a hybrid cloud collaboration module 320, wherein: The Elastic Resource Scheduling Module 310 dynamically allocates computing resources based on container orchestration technology, supports horizontal and vertical scaling of service instances, builds a resource demand prediction model, predicts resource usage trends through historical load data and business cycle characteristics, and enables cross-availability zone failover. As a further explanation of this embodiment, the construction of the resource demand forecasting model includes the following steps: First, historical load data is collected hourly via KubernetesMetricsServer. And obtain business cycle characteristics from calendar service. ; Then, based on historical data from the past 30 days, a linear regression model was fitted using the least squares method. Calculate model parameters , , ; Then, input the current information in real time. and Predicting resource demand for the next hour ; in, For the future Resource requirements; Historical load weighting This is the business cycle impact coefficient. The basic resource requirements were obtained by fitting historical data from the past 30 days using the least squares method. Next, based on the prediction results, drive the container orchestration system to scale horizontally or vertically.

[0041] Furthermore, the cross-availability zone failover in this embodiment includes the following steps: First, an HTTP heartbeat request is sent to the service instance every 10 seconds. If the request times out three times in a row, the instance is considered to be faulty. Then, in the event of a failure in the primary availability zone, the service instances are automatically migrated to the backup availability zone via Kubernetes cluster federation. Subsequently, update the service's DNS resolution or load balancer endpoint to ensure that traffic is switched to the backup instance; Finally, after the primary availability zone is restored, the resources of the backup instance are automatically reclaimed.

[0042] It is worth noting that the historical load data in this embodiment The collection specifically includes: First, collect historical data on the entire GIS 3D modeling process over the past 3 years (or a specific timeframe), including but not limited to: Hardware resource dimensions: server CPU utilization (collected in 1-second increments, accurate to 0.1%), memory usage (distinguishing between cache and actual usage), and GPU computing load (categorized by modeling task type, such as terrain modeling and detailed building modeling). Task characteristic dimensions: model data volume (graded by point cloud / vector data size, such as <1GB, 1-10GB, >10GB), modeling complexity (distinguishing between simple geometric shapes and complex urban building clusters through a combination of manual annotation and algorithm recognition), and task concurrency (recording the number of parallel modeling tasks and type conflicts during the same period). Time dimension: Categorized by time tags such as weekdays / holidays, daytime / nighttime, etc., covering different business peak periods.

[0043] Then, the collected data is filtered for outliers (such as false alarms where the CPU utilization rate is 100% instantaneously and lasts for less than 1 second) and missing values ​​are filled (using the average interpolation of the same type of task over a period of time). Finally, a standardized historical dataset is constructed according to the "resource-task-time" relationship for model training and prediction.

[0044] The Hybrid Cloud Collaboration Module 320 is used to build a two-way data synchronization channel between private cloud and public cloud, supporting encrypted transmission and breakpoint resume, realizing unified management of multi-cloud resources, shielding the API differences of different cloud vendors through an abstraction layer, intelligently scheduling multi-cloud resources based on business priorities, deploying critical tasks on the private cloud, and allocating elastic tasks to idle resources on the public cloud.

[0045] As a further explanation of this embodiment, the construction of the private cloud-public cloud bidirectional data synchronization channel in this embodiment includes the following steps: First, an encrypted channel is established using the SSL / TLS 1.3 protocol, and a certificate issued by a private CA is used to ensure transmission security. Then, the data is divided into 100MB segments, and the SHA-256 hash value of each segment is calculated. Subsequently, a hash table of the transmitted segments is recorded, and completed segments are skipped when resuming transmission. Finally, the APIs of different cloud vendors are encapsulated through an API abstraction layer to provide a unified interface to the outside world.

[0046] Furthermore, in this embodiment, intelligent scheduling of multi-cloud resources based on business priorities includes the following steps: First, prioritize tasks based on their type; then, select cloud resources using a greedy algorithm. and ,in The final target cloud platform; To select the cloud platform with the lowest unit resource cost from the cloud platform collection Clouds; The unit resource cost of cloud platform c; Clouds is the collection of available cloud platforms; The remaining available resources for cloud platform C; The resource requirements for the current task are determined; then, critical tasks are deployed to the private cloud, and elastic tasks are allocated to available resources on the public cloud; finally, data between the private cloud and the public cloud is synchronized through an encrypted channel.

[0047] In this embodiment, the cloud platform support unit 300 further includes a hierarchical storage module 330 and a cross-region synchronization module 340, wherein: The tiered storage module 330 adopts a tiered storage architecture; the tiered storage architecture specifically includes a hot data layer, a warm data layer, and a cold data layer, wherein: The hot data layer stores frequently accessed data through a distributed in-memory database; The warm data layer stores recent data through an object storage system; The cold data layer stores historical archived data through a tape library, supporting retrieval and batch recovery by timestamp.

[0048] As a further explanation of this embodiment, the hierarchical storage module 330 in this embodiment adopts a hierarchical storage architecture of hot data layer, warm data layer, and cold data layer to achieve hierarchical data management: First, the positioning and technology selection of each layer are clarified. The hot data layer stores real-time data with high frequency access in the past 7 days through RedisCluster to ensure efficient access; the warm data layer stores recent data within 30 days through the MinIO object storage system to balance storage performance and cost; the cold data layer stores historical archived data through a tape library to support batch recovery of historical data. At the same time, data flow rules are constructed: when hot data flows to warm data, the access frequency of data in Redis is monitored. When the access frequency continues to decrease, the data is serialized and uploaded to MinIO, the original data in Redis is deleted, and metadata such as storage path and timestamp are retained; when warm data flows to cold data, it relies on the lifecycle management rules of MinIO. When the data storage duration meets the archiving conditions, it is automatically packaged into an archive file, written to the tape library, and indexed by timestamp; when cold data is recovered, the target tape is first located by searching the tape index by timestamp, the data is read in parallel through the tape drive, and then the recovered data is migrated back to the warm data layer for business access.

[0049] The cross-regional synchronization module 340 ensures data consistency between multiple geographical regions based on a consistency algorithm, constructs a data synchronization priority queue, configures the synchronization cycle according to data timeliness, implements data compression and encrypted transmission, and feeds back the data synchronization status to the elastic resource scheduling module 310.

[0050] As a further explanation of this embodiment, the cross-regional synchronization module 340 in this embodiment constructs a guarantee mechanism around the consistency of multi-regional data: First, Raft nodes are deployed in each geographical region, and a leader node is elected to uniformly coordinate data synchronization tasks; then, a synchronization priority queue is constructed based on data type (hot, warm, cold) and timeliness (such as real-time monitoring data marked as high timeliness), configuring a shorter synchronization cycle for high timeliness data and extending the synchronization interval for low timeliness data; subsequently, the data to be synchronized is compressed using the Snappy algorithm to reduce bandwidth consumption, and then encrypted using AES-256 to ensure transmission security; next, the leader node broadcasts the encrypted synchronization log to the follower nodes in the region, and following the Raft protocol, completes the data submission after more than half of the nodes confirm receipt, ensuring data consistency across multiple regions; finally, the data synchronization status is continuously monitored, and when the synchronization delay (the difference between the actual and planned synchronization time) exceeds a preset threshold, an alarm is fed back to the elastic resource scheduling module 310, driving the dynamic adjustment of resources of the synchronization nodes.

[0051] Furthermore, the collaborative workflow of multiple modules in the cloud platform support unit 300 in this embodiment is as follows: First, the elastic resource scheduling module 310 calculates resource requirements through a resource demand prediction model, drives container scaling, and allocates computing resources for task deployment by the hybrid cloud collaboration module 320, data reading and writing by the hierarchical storage module 330, and transmission operations by the cross-region synchronization module 340. Then, the hybrid cloud collaboration module 320 deploys critical tasks to the private cloud and allocates elastic tasks to the public cloud based on task priorities, synchronizing multi-cloud data through encrypted channels. Simultaneously, the hierarchical storage module 330 automatically transfers data between hot, warm, and cold storage layers based on data access frequency. Subsequently, the cross-region synchronization module 340 ensures data consistency across multiple regions based on the Raft algorithm, schedules synchronization tasks according to priority queues, and transmits data after compression and encryption. Finally, the cross-region synchronization module 340 feeds back data synchronization delay and other statuses to the elastic resource scheduling module 310, triggering dynamic resource adjustments and forming a collaborative closed loop of "resource allocation → task execution → data transfer → synchronization guarantee → resource feedback".

[0052] The 3D dynamic modeling unit 400 is used to construct a 3D dynamic model of a geographic scene and realizes real-time updating of the 3D dynamic model based on multi-source data fusion technology. In this embodiment, the 3D dynamic modeling unit 400 includes a multi-source data fusion module 410, a 3D model construction module 420, and a model dynamic update module 430, wherein: The multi-source data fusion module 410 is used to access multiple types of geographic data (such as LiDAR point clouds, UAV imagery, GIS vectors, etc.). Through format parsing, coordinate calibration and feature matching algorithms, it spatially aligns and semantically associates the multi-source data to achieve fusion. If there is spatial deviation in the data, the deviation is corrected through feature matching. As a further explanation of this embodiment, the multi-source data fusion module 410 in this embodiment, for various types of geographic data such as LiDAR point clouds, UAV imagery, and GIS vectors, sequentially achieves spatial alignment and semantic association through format parsing, coordinate calibration, and feature matching: First, the data formats are parsed. The 3D coordinates and classification codes of LiDAR point clouds are extracted from the LAS format file header. The image projection relationship and band data are extracted from the GeoTransform parameters of the GeoTIFF format. The geometric contours and attributes of GIS vectors are parsed from the SHP or GeoJSON format files. Then, the coordinates are calibrated using a "coarse registration + fine registration" process. During coarse registration, ground vertex, SIFT feature points, and FPFH keypoints are extracted from vectors, images, and point clouds to construct corresponding point pairs, which are then processed using affine transformation formulas. The initial transformation parameters are solved by prototyping cognite pairs using the least squares method; where The target's three-dimensional coordinates in the reference coordinate system; It is a rotation matrix; These are the original three-dimensional coordinates of the source data; It is a translation vector; During fine registration, the ICP algorithm is iteratively executed on the coarsely registered data. First, the nearest point in the reference point cloud is matched for each point in the source point cloud. The error of the sum of squared distances between point pairs is calculated and minimized. Then, the rotation matrix and translation vector are updated through singular value decomposition until the iteration converges (e.g., 30 iterations or the error change meets the threshold). Finally, SIFT features of the image, FPFH features of the point cloud, and vector geometric parameters are extracted. When the spatial distance between the image and the point cloud features is ≤0.5m, and the vector land cover attributes match the point cloud classification code and the image spectral features, geometric and semantic associations are established to complete the fusion of multi-source data.

[0053] The 3D model building module 420 constructs terrain triangular meshes and parametric models of ground features based on the fused data, and adapts to different rendering accuracies through a LOD grading strategy, thereby generating a 3D scene with semantic information. As a further explanation of this embodiment, the three-dimensional model construction module 420 in this embodiment first constructs a terrain triangular mesh based on the fused LiDAR point cloud elevation data: taking the Delaunay triangulation as the core, dividing the point cloud into 100m×100m tiles according to the point cloud density, and constructing a triangular mesh that satisfies the "empty circle characteristic" for each tile to maximize the regularity of the triangles; for the edge tiles, adjacent triangular meshes are merged by sharing boundary points to ensure the continuity of the terrain surface.

[0054] Furthermore, for different terrain features such as buildings, roads, and vegetation, the 3D model building module 420 adopts a differentiated parametric modeling strategy: the building model extracts vector contours and point cloud elevations, and generates a parametric model with modifiable height by stretching the contour polygons; the road model generates a strip surface along the vector centerline, and the width is determined by vector attributes or point cloud cross-section calculations; the vegetation model generates a Billboard model for "vegetation" type point clouds, and simulates vegetation morphology with single-layer texture mapping to balance detail and performance.

[0055] In addition, to adapt to different rendering precisions, the 3D model building module 420 introduces a Level of Detail (LOD) strategy: preset viewing distance thresholds (such as 50m, 200m, 500m), when the distance to the camera is ≤50m, the complete triangular mesh and texture details are loaded (LOD0); between 50m and 200m, the triangular faces are simplified by 50% through an edge folding algorithm (LOD1); between 200m and 500m, only the boundary lines of the terrain are retained (LOD2); and the model is hidden when the distance exceeds 500m (LOD3), dynamically reducing the rendering load and ensuring scene smoothness.

[0056] The model dynamic update module 430 is used to compare geographic data at different timestamps, detect areas of scene change, calculate incremental update data and inject it into the terrain triangulation mesh and the parametric model of ground features, so as to refresh the status of the terrain triangulation mesh and the parametric model of ground features in real time.

[0057] As a further explanation of this embodiment, the model dynamic update module 430 in this embodiment first accesses fused data from different timestamps, and calculates the elevation difference between the two point clouds in response to terrain changes. ,in: for The elevation value of the point cloud at any given time. for The elevation value of the point cloud at any given time; when Real-time determination of terrain changes; for changes in ground features, comprehensive detection is performed from geometric (vector ground feature outline area change rate > 10%), attribute (ground feature attribute field update, such as change in building number of floors) and spectral (vegetation NDVI index change > 0.2) dimensions to accurately locate the area of ​​scene change.

[0058] Furthermore, for the detected change areas, the model dynamic update module 430 then extracts incremental data: it filters change points with elevation differences exceeding a threshold based on terrain dimensions, constructs a minimum bounding rectangle (MBR) to locate the change areas, and outputs data containing... The coordinates and point cloud subset marked with "new / modified / deleted" are used to identify the IDs of newly added, modified, and deleted features in the feature dimension. Their geometric parameters (such as the outline vertices of newly added buildings) or attribute data (such as the building height changing from 10m to 15m) are extracted and encapsulated into JSON format instructions to standardize the incremental data format to support the subsequent injection of terrain triangulation mesh and feature parametric model.

[0059] In addition, the model dynamic update module 430 injects incremental data into the terrain triangular mesh and the parametric model of the ground features: the terrain mesh deletes the original triangular mesh of the changed area, and re-performs Delaunay triangulation on the incremental point cloud. If the changed area overlaps with the adjacent tile (such as a 100m×100m terrain block), the edge triangular mesh connection relationship is corrected simultaneously to ensure terrain continuity. As for the ground feature model, the added ground features generate parametric models based on the incremental data (such as stretching the building outline to generate a three-dimensional body), and the modified ground features update parameters such as height and texture. The deleted ground features remove the corresponding model instance, and finally realize the real-time status update of the terrain and ground features.

[0060] It should be added that the collaborative process of the multi-module 3D dynamic modeling unit 400 in this embodiment is as follows: the multi-source data fusion module 410 accesses and aligns LiDAR, image, and vector data, and outputs spatially aligned fused data; the 3D model construction module 420 constructs a terrain triangulation network and a parametric model of ground features based on the fused data, and optimizes rendering through LOD hierarchical optimization; the model dynamic update module 430 compares multiple timestamp data at regular intervals to detect changes in terrain and ground features; calculates incremental data of the changed areas, injects it into the terrain mesh and ground feature model, and completes real-time updates.

[0061] The intelligent analysis unit 500 is used to perform multi-dimensional geospatial analysis on the 3D model and outputs pattern mining and trend prediction results through improved spatiotemporal analysis algorithms. In this embodiment, the intelligent analysis unit 500 outputs pattern mining and trend prediction results through an improved spatiotemporal analysis algorithm, including the following steps: S500.1, Spatiotemporal Feature Extraction: Spatiotemporal representations of 3D geographic entities are constructed using a spatiotemporal cube model, and spatiotemporal feature vectors are extracted. : , where G is the spatial feature vector; This is a time feature vector; For attribute feature vectors; In this step, the spatial feature vector G: extracts the spatial geometric parameters of geographic entities from the 3D model, including: Point entities (such as streetlights): 3D coordinates ; Linear entities (such as roads): centerline length, mean curvature; Surface / volume entity (e.g., building): Length / width / height and centroid coordinates of the minimum bounding rectangle (MBR) , contour polygon vertex set.

[0062] Furthermore, the time feature vector Based on the data collection timestamp and update records, extract time attributes, including: Absolute Time: Timestamp of the first data collection Recent update timestamp ; Relative time: duration of existence Update frequency (e.g., once a month is counted as 1 / 30); Periodic characteristics: If the data is updated periodically (such as seasonal vegetation), add periodic markers (such as "quarter" or "year").

[0063] In addition, attribute feature vectors Integrating semantic attributes of ground features, including: Type attribute: Feature classification label (e.g., "Residential Buildings" or "Main Roads", encoded with integers, such as 1=Building, 2=Road); Quantitative attributes: number of building floors, road width, vegetation coverage (values ​​between 0 and 1); Status attribute: Current status of the feature (e.g., "normal" or "under construction", represented by Boolean or enumerated values).

[0064] Furthermore, the feature vectors are uniformly formatted as numerical arrays (non-numerical attributes are converted through encoding), for example, the feature vectors of a certain building. It can be represented as: =[ [Length, width, height]; =[ [Update frequency]; (1 = building, 10 floors, coverage 0.8, normal condition).

[0065] S500.2, Spatiotemporal correlation modeling: Constructing the spatiotemporal correlation matrix Quantifying the spatiotemporal dependencies between geographic entities: ; in, The spatial distance function can be Euclidean distance or network distance. The time distance function uses timestamp differences or time interval ratios. The attribute similarity function can be cosine similarity or edit distance. This is the combined distance function; For the first , The spatiotemporal correlation of an entity; For the spatial feature vector of a geographic entity; The temporal feature vector of a geographic entity; In this step, the spatial distance in this embodiment Differentiated calculations are used for different entity types, including: Point entities using Euclidean distance ;in: , , For entities 3D coordinates; , , For entities 3D coordinates; Use the hausdorff distance for line / surface / volume entities ;in: Representing entities All point-to-entity The minimum distance; Time Distance Timestamp difference normalization If entities have consistent periodic characteristics, the superimposed periodic phase difference ;in: For entities The most recent update timestamp; For entities The most recent update timestamp; , These are the global maximum and minimum timestamps, respectively. , For entities , Phase within the period; The weighting coefficient for the periodic phase difference is set manually to balance the influence of timestamp difference and phase difference; Attribute similarity Combining type and quantification attributes, the type attribute uses Jaccard similarity. ,in: For entities and Type attribute similarity; , For entities A collection of type attributes; This is the intersection operation of sets; This is the union operation for sets; The number of elements in the set. Quantification attribute using cosine similarity. ,in: For entities and Quantitative attribute similarity; , For entities , Quantized attribute vector; This is the dot product operation for vectors; Let be the magnitude of the vector. Combined as... (Weights are set according to attribute importance), where: For entities and The overall attribute similarity; 0.6 and 0.4 are the weight coefficients of type / quantitative attributes, which are manually set to reflect the importance of the two types of attributes to the business. For example, if type is more important, the proportion of 0.6 is increased.

[0066] Furthermore, the integrated distance function A weighted normalization formula is used to integrate spatial distance, temporal distance, and attribute similarity into a correlation score (the larger the value, the stronger the correlation): ; in, These are the weighting coefficients, and , This is the global maximum spatial distance (used to normalize dist_s to 0-1); Furthermore, if slight spatial discrepancies still exist after multi-source data fusion, the correlation can be fine-tuned through feature matching—for known entities with the same name (such as records from different data sources for the same building), the correlation can be forced to... (Set a lower limit) and correct its spatial distance calculation value (e.g., replace the original calculated value with a known real distance).

[0067] S500.3, Spatiotemporal Pattern Recognition: Based on spatiotemporal correlation matrix An improved spatiotemporal clustering algorithm is used to identify pattern clusters. : ; in, This is a penalty coefficient that controls the balance between cluster compactness and separation. The penalty function is defined based on the cluster diameter or density. For the first A spatiotemporal pattern cluster; To find the cluster that maximizes the expression within the parentheses ; For pattern clusters Geographic entity pairs within; As a further explanation of this embodiment, the spatiotemporal pattern recognition in this embodiment is based on the spatiotemporal correlation matrix. An improved density clustering algorithm is used to identify pattern clusters. : First, randomly select unassigned entities as initial cluster centers, and then calculate their association with other entities. (The larger the value, the stronger the correlation), ( for Entities whose mean (using the mean as the association threshold) are assigned to the initial cluster. This process is repeated until all entities are assigned (if an entity is associated with all cluster centers). (If it is a single cluster, then it forms a new cluster). Redefine the objective function Optimize the cluster structure, among which ( The maximum spatial distance within the cluster. The maximum time distance is used to characterize the degree of cluster dispersion; The penalty coefficient, determined in the range of 0.1–1.0 through grid search, maximizes the average intra-cluster correlation and the most significant inter-cluster differences. Subsequently, iterative updates are performed, attempting to reallocate each entity. If the objective function increases, the cluster to which it belongs is updated, until the rate of change is less than 1% for three consecutive iterations, at which point a stable pattern cluster is output. ; Finally, semantic labels are added to the clusters, such as summarizing features like "5-10 story residential buildings distributed along main roads, newly built in the last 3 years," to explain the meaning of the pattern.

[0068] S500.4 Trend Prediction Model: Constructing a spatiotemporal regression model to predict future states ; ; in, This represents the spatiotemporal feature vector at the current moment; The weight matrix is ​​obtained through training with historical data; As a bias vector, it captures the inherent trend of the system; This is the random error term; As a further explanation of this embodiment, trend prediction constructs a linear spatiotemporal regression model. Predicting future feature vectors ( (Including spatial, temporal, and attribute characteristics, such as the building's three-dimensional coordinates, update time, number of floors, etc.) First, collect historical time series data. The model is divided into an 80% training set and a 20% validation set; The weight matrix is ​​a feature dimension × feature dimension (e.g.) If it is 10-dimensional, then (10×10, depicting the impact of current features on the future). It is a 10-dimensional bias vector. The random error (assuming it follows a normal distribution with a mean of 0) is minimized using the least squares method. Solve and (This can be implemented using Python); if multicollinearity exists among features, for Applying L2 regularization (ridge regression), the objective function is adjusted to ,in, The regularization coefficient is determined by minimizing the validation set error; if the data contains time-series characteristics (such as seasonal variations), in Adding a lag term (e.g.) The model is expanded to By leveraging historical lag characteristics, prediction accuracy can be enhanced.

[0069] S500.5 Result Verification and Output: The prediction accuracy is evaluated through cross-validation, and the mean squared error is calculated. : ; in, To verify the sample size; for The real spatiotemporal feature vector at any given moment; for The predicted spatiotemporal feature vector at any given time; when When the threshold is reached, the prediction result is output. and confidence interval ,in The confidence interval is half-width.

[0070] As a further explanation of this embodiment, the reliability of the model is evaluated through 5-fold cross-validation: historical data is divided into 5 parts, and 4 parts are used for training and 1 part for validation in turn. The mean squared error of the validation set is calculated. If the average MSE is less than or equal to the preset threshold (the threshold is the square of the standard deviation of the feature vector, set according to business requirements), the model is considered effective; otherwise, return to S500.4 for retraining (e.g., adjusting the regularization coefficient). When outputting the prediction results, in addition to the predicted values ​​of the feature vectors at future times... It also includes confidence intervals for each dimension. ,in For feature dimension, the confidence interval half-width ( For the first cross-validation The standard deviation of the prediction error for the dimensional feature (1.96 corresponds to a 95% confidence level); and the output pattern cluster. The core features, such as "cluster" The average annual height of buildings in China is 2.3 meters, distributed along subway lines. This analysis summarizes the spatial distribution and temporal evolution patterns, providing an interpretable basis for decision-making.

[0071] Furthermore, the collaborative process of the intelligent analysis unit 500 in this embodiment is as follows: First, it receives the three-dimensional model data and extracts the spatiotemporal feature vectors. (Including spatial geometry, temporal attributes, and semantic attributes, such as the building's center of gravity coordinates, update time, number of floors, etc.); then, based on F, a spatiotemporal relationship matrix between entities is constructed. By weighted fusion of spatial distance, temporal distance, and attribute similarity, entity dependencies are quantified; subsequently, an improved density clustering algorithm is used to... Identifying spatiotemporal pattern clusters Summarize the patterns in the scenarios; then combine them with history. Train a trend prediction model, incorporate optimization strategies such as lag terms and regularization, to predict future feature vectors. Finally, the model is evaluated through 5-fold cross-validation, and the predicted results with confidence intervals and pattern cluster interpretations are output, forming a closed loop of "feature extraction - association modeling - pattern recognition - trend prediction - validation output", providing intelligent decision support for scenarios such as urban planning and disaster monitoring.

[0072] The Visual Interaction Unit 600 is used to realize the multi-terminal presentation and interactive operation of 3D scenes, and improves multi-device adaptability based on adaptive rendering technology.

[0073] In this embodiment, the visualization interaction unit 600 includes a multi-terminal rendering module 610, an interaction control module 620, a data visualization module 630, a performance optimization module 640, and an event bus module 650, wherein: The multi-terminal rendering module 610 is used to realize cross-terminal adaptive rendering of 3D scenes, and adjusts display parameters according to device characteristics based on adaptive layout technology; As a further explanation of this embodiment, the multi-terminal rendering module 610 in this embodiment first calls the system API to collect the device screen resolution. Based on pixel density and GPU computing power, devices are categorized into high-performance PCs, tablets, and mobile phones, etc.; based on device classification, scaling factors are used to further categorize devices. (e.g., 1.0 for PC, 0.6 for mobile) Dynamically adjust rendering resolution. ), and establish line of sight Mapping rules between (camera distance to scene) and LOD level: ( (Fine level) ( (Simplified level) ( (contour level) For low-performance devices (such as mobile phones), the LOD switching threshold is reduced proportionally (e.g., from 50 to 40), and the scene status (such as viewpoint and selected object) of multiple terminals is synchronized through the event bus.

[0074] The interactive control module 620 is used to implement multimodal interactive operations and respond to user selections and operation commands based on the interactive event handling mechanism; As a further explanation of this embodiment, the interaction control module 620 in this embodiment defines gestures. Click ,voice Set priority for interactive event types. The system captures raw operations through browser event listening and speech recognition API, parses them into semantic commands (such as two-finger distance change corresponding to "zoom", and voice text corresponding to "location coordinates"), schedules execution according to priority queue (high-priority events can interrupt low-priority operations), calls the 3D engine API to respond to interactions (such as zooming the scene, selecting ground features), and notifies other modules through the event bus.

[0075] The data visualization module 630 is used to visualize thematic maps and feature plotting, and converts geographic information into graphical representations based on data mapping technology; As a further explanation of this embodiment, the data visualization module 630 in this embodiment uses hierarchical color schemes for geographic attribute data, according to the formula: Divide the hierarchy and map it to preset color schemes; among them, For attribute values, , It is the global extremum. This represents the number of levels.

[0076] Line-of-sight elements, based on traffic flow Calculate width ,in Based on the width, For coefficients; For user-drawn points, lines, and surfaces, icons are mapped, Bézier curves are fitted, and Delaunay triangulation is performed to generate geometry. At the same time, the data changes of the model dynamic update module are monitored, and the plotting style is automatically refreshed (such as the plotting column elongation corresponding to changes in building height).

[0077] The performance optimization module 640 is used to improve rendering efficiency and optimize the display effect of complex models based on scene simplification strategies; As a further explanation of this embodiment, the performance optimization module 640 in this embodiment combines device-level adjustment of the LOD switching threshold, through coefficients (Use 1.0 for high-performance devices and 0.8 for mobile phones) Scaling view distance threshold; using depth buffering to compare the depth of the current object. With depth buffer value Remove occluded objects; for repetitive features such as roadside trees, use GPU instantiation technology and apply a formula. N =Total quantity / Batch size (Batch size is set to 100 depending on GPU performance) Batch rendering, sharing vertex data reduces transmission overhead.

[0078] The event bus module 650 is used to realize communication and collaboration between modules, and to distribute and process interactive events based on the message passing mechanism.

[0079] As a further explanation of this embodiment, the event bus module 650 in this embodiment is based on the "publish-subscribe" mode, defining message types such as rendering parameters, interactive events, and data changes, and carrying parameters such as resolution and selected object ID; when a module triggers an event, it publishes a message, and other modules subscribe to messages of interest. After the bus forwards the message, the subscribing module parses and executes it (such as the interactive control module 620 publishing a "selected building" event, the data visualization module highlighting the plot, and the multi-terminal rendering module focusing on rendering).

[0080] Furthermore, the multi-module collaborative process of the visualization interaction unit 600 in this embodiment is as follows: the multi-terminal rendering module 610 adapts to device parameters and renders the initial scene; the interaction control module 620 captures operations and publishes messages; the data visualization module 630 responds to messages and updates the plot; the performance optimization module 640 dynamically optimizes rendering based on the viewing distance and device performance; and the event bus module 650 maintains a synchronized state throughout the entire process, ensuring consistent interaction and presentation across devices and multiple modules, and supporting users' visualization analysis and operation of the three-dimensional geographic scene.

[0081] Example 2 like Figure 2As shown, this embodiment also provides a GIS 3D dynamic modeling and analysis method that integrates big data and cloud platform networks. Based on the above-mentioned GIS 3D dynamic modeling and analysis system that integrates big data and cloud platform networks, the method includes the following steps: S100 Multi-source geographic data aggregation and access: Aggregates and accesses multiple types of geospatial data, adapts to different data sources through standardized protocol conversion, dynamically adjusts the transmission rate and supports breakpoint recovery of transmission, automatically extracts metadata containing information such as collection time and spatial coordinates, registers data sources, monitors their status and provides anomaly alarms, and ensures data validity through standardized preprocessing and quality control. S200, Distributed Geographic Data Processing: Parallel feature extraction and spatiotemporal correlation calculation of geographic data; spatial partitioning based on spatial grid index and associated with unique identifiers; extraction of multiple types of features such as terrain and buildings and support for multi-scale configuration; construction of spatiotemporal index to support millisecond-level queries; real-time monitoring of computing node load and dynamic adjustment of task allocation; and inter-module communication through message queues to improve data processing efficiency. S300, cloud platform resources and storage support: provides elastic resource scheduling and cross-regional data storage services, dynamically allocates computing resources based on container orchestration and supports service expansion, builds a private cloud-public cloud bidirectional synchronization channel and schedules multi-cloud resources according to business priority, and adopts a layered storage architecture to store high-frequency access, recent and historical data respectively. S400, 3D dynamic model construction and updating: Constructing terrain triangular mesh and parametric model of geographic scene, performing format parsing, coordinate calibration and feature matching on multiple types of geographic data to achieve fusion and correct spatial deviation, constructing terrain and feature models based on fused data and adapting to different rendering precision through hierarchical strategy, detecting scene changes by comparing data with different timestamps, calculating incremental data and injecting it into the model to refresh the status in real time. S500, 3D model intelligent analysis: performs multi-dimensional geospatial analysis on terrain triangular mesh and parametric model of land features, extracts spatial, temporal and attribute feature vectors through spatiotemporal cube model, constructs spatiotemporal correlation matrix to quantify the dependency relationship between geographic entities, adopts improved clustering algorithm based on correlation matrix to identify spatiotemporal pattern clusters, constructs regression model to predict future state, evaluates accuracy through cross-validation and outputs prediction results and confidence interval, realizing pattern mining and trend prediction; S600, 3D Scene Visualization and Interaction: Enables multi-terminal presentation and interactive operation of 3D scenes. Based on adaptive rendering technology, it adjusts display parameters according to device characteristics to adapt to multiple terminals. It supports multi-modal interactive operation to respond to user commands, converts geographic information into graphic representation to realize thematic maps and feature plotting display, optimizes the display effect of complex models to improve rendering efficiency, and realizes inter-module communication and collaboration through message passing mechanism.

[0082] Those skilled in the art will understand that the process of implementing all or part of the steps of the above embodiments can be carried out by hardware or by a program instructing the relevant hardware.

[0083] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks, characterized in that: include: The multi-source geographic data acquisition unit (100) is used to realize the aggregation and access of multiple types of geospatial data, and ensures the validity of data through standardized preprocessing procedures and quality control mechanisms. The distributed data processing unit (200) is used to perform parallel feature extraction and spatiotemporal correlation calculation on geographic data, and improve data processing efficiency based on a dynamic load balancing architecture. The cloud platform support unit (300) is used to provide elastic resource scheduling and cross-regional data storage services, and ensures high system availability through a hybrid cloud collaboration mechanism; The 3D dynamic modeling unit (400) is used to construct a 3D dynamic model of a geographic scene and realizes real-time updating of the 3D dynamic model based on multi-source data fusion technology. The intelligent analysis unit (500) is used to perform multi-dimensional geospatial analysis on the 3D model and output the results of pattern mining and trend prediction through the improved spatiotemporal analysis algorithm. The visualization interaction unit (600) is used to realize the multi-terminal presentation and interactive operation of the 3D scene, and improves the multi-device adaptability based on adaptive rendering technology.

2. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The multi-source geographic data acquisition unit (100) includes a data protocol adaptation module (110), a data transmission control module (120), a metadata extraction module (130), and a data source management module (140), wherein: The data protocol adaptation module (110) integrates a multi-protocol conversion component to achieve standardized conversion of protocols from different data sources; The data transmission control module (120) is used to dynamically adjust the transmission rate to adapt to the network bandwidth and supports the recovery of transmission from the point of data interruption. The metadata extraction module (130) is used to automatically parse the header information of various types of data and extract metadata fields including collection time, spatial coordinate range, and data precision level; The data source management module (140) is used to provide data source registration, status monitoring and abnormal alarm functions, and triggers an alarm mechanism when the data source connection is abnormal.

3. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 2, characterized in that, In the data transmission control module (120), resuming transmission from the point of data interruption includes the following steps: S120.1 Monitoring link connectivity via periodic heartbeat packets: The number of consecutive unresponsive heartbeat cycles The threshold is reached, and the duration of a single heartbeat cycle is... If there is no response, the transmission is considered interrupted. Synchronously record the absolute byte position of the currently transmitted data. This serves as a reference benchmark for the starting point of recovery. S120.2, Divide the transmitted data into fixed block sizes. Divided into data blocks and execute: Calculate the SHA-256 hash value for each data block. ; Generate block sequence number SHA-256 hash value Block start byte position Interrupt status table For subsequent comparison; S120.3 Send a recovery request to the data sender, carrying the interrupt status table. ; The sender compares the hash table of the local data blocks. The formula is used to locate the sequence number of the first block with a hash mismatch. ; ； in, The hash value of the local data block at the sending end; Sequential numbering of data blocks; The total number of blocks into which the transmitted data has been divided; Determine the starting byte position for breakpoint recovery ; ； in, For the first The starting byte position of the block data; Fixed block size; S120.

4. A sliding window mechanism is used to retransmit missing data blocks in parallel, with a window size of [missing information]. Dynamically adjusted to: ； in, For network round-trip latency; This represents the currently available bandwidth. This represents the maximum number of parallel windows. This is a floor function; S120.5 After the resume download is complete, execute: Calculate the MD5 hash of the complete data ; The checksum preset at the sending end Comparison: like The transmission was deemed complete. If there is a mismatch, backtrack to step S120.3, relocate the local mismatch block, and repeatedly execute the resume-verification until the hash matches.

4. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The distributed data processing unit (200) includes a data sharding scheduling module (210), a feature extraction module (220), a spatiotemporal correlation calculation module (230), a load balancing control module (240), and a communication module (250), wherein: The data sharding scheduling module (210) divides geographic data into spatial shards with fixed granularity based on the spatial grid index and associates them with unique spatial identifiers. It maintains a shard-node mapping table to support dynamic migration and load balancing of shards and realizes version control and incremental updates of sharded data. The feature extraction module (220) integrates a terrain feature extractor and a land feature extractor to identify features of terrain parameters, buildings, roads, water bodies, and vegetation. It supports multi-scale feature extraction and can configure the extraction accuracy. It avoids redundant calculations through a feature caching mechanism. The spatiotemporal correlation calculation module (230) is used to construct a spatiotemporal index structure to support millisecond-level spatiotemporal range queries, and to use spatial proximity analysis and time series analysis algorithms to realize multi-source data correlation and fusion. The load balancing control module (240) is used to monitor the CPU, memory, and IO utilization of computing nodes in real time, and dynamically adjust task allocation and trigger task migration based on the node load status. The communication module (250) adopts a publish-subscribe message queue to achieve loosely coupled communication between modules, and supports the transmission of processing status and result data.

5. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The cloud platform support unit (300) includes an elastic resource scheduling module (310) and a hybrid cloud collaboration module (320), wherein: The elastic resource scheduling module (310) dynamically allocates computing resources based on container orchestration technology, supports horizontal and vertical scaling of service instances, constructs a resource demand prediction model, predicts resource usage trends through historical load data and business cycle characteristics, and realizes cross-availability zone failover. The hybrid cloud collaboration module (320) is used to build a two-way data synchronization channel between private cloud and public cloud, support encrypted transmission and breakpoint resume, realize unified management of multi-cloud resources, shield the differences in APIs of different cloud vendors through the abstraction layer, intelligently schedule multi-cloud resources based on business priority, deploy critical tasks on private cloud, and allocate elastic tasks to idle resources in public cloud.

6. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 5, characterized in that, The cloud platform support unit (300) further includes a hierarchical storage module (330) and a cross-region synchronization module (340), wherein: The hierarchical storage module (330) adopts a hierarchical storage architecture; The cross-regional synchronization module (340) ensures data consistency among multiple geographical regions based on a consistency algorithm, constructs a data synchronization priority queue, configures the synchronization cycle according to data timeliness, realizes data compression and encrypted transmission, and feeds back the data synchronization status to the elastic resource scheduling module (310).

7. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The three-dimensional dynamic modeling unit (400) includes a multi-source data fusion module (410), a three-dimensional model construction module (420), and a model dynamic update module (430), wherein: The multi-source data fusion module (410) is used to access multiple types of geographic data. Through format parsing, coordinate calibration and feature matching algorithms, it performs spatial alignment and semantic association of multi-source data to achieve fusion. If there is a spatial deviation in the data, the deviation is corrected through feature matching. The three-dimensional model building module (420) constructs a terrain triangular mesh and a parametric model of ground features based on the fused data, and adapts to different rendering accuracies through a LOD hierarchical strategy, thereby generating a three-dimensional scene with semantic information. The model dynamic update module (430) is used to compare geographic data with different timestamps, detect areas of scene change, calculate incremental update data and inject it into the terrain triangular mesh and the parametric model of ground features, so as to refresh the status of the terrain triangular mesh and the parametric model of ground features in real time.

8. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The intelligent analysis unit (500) outputs pattern mining and trend prediction results through an improved spatiotemporal analysis algorithm, including the following steps: S500.1, Spatiotemporal Feature Extraction: Spatiotemporal representations of 3D geographic entities are constructed using a spatiotemporal cube model, and spatiotemporal feature vectors are extracted. : ,in, For spatial feature vectors; This is a time feature vector; For attribute feature vectors; S500.2, Spatiotemporal correlation modeling: Constructing the spatiotemporal correlation matrix Quantifying the spatiotemporal dependencies between geographic entities: ； in, The spatial distance function can be Euclidean distance or network distance. The time distance function uses timestamp differences or time interval ratios. The attribute similarity function can be cosine similarity or edit distance. This is the combined distance function; For the first , The spatiotemporal correlation of an entity; For the spatial feature vector of a geographic entity; The temporal feature vector of a geographic entity; S500.3, Spatiotemporal Pattern Recognition: Based on spatiotemporal correlation matrix An improved spatiotemporal clustering algorithm is used to identify pattern clusters. : ； in, This is a penalty coefficient that controls the balance between cluster compactness and separation. The penalty function is defined based on the cluster diameter or density. For the first A spatiotemporal pattern cluster; To find the cluster that maximizes the expression within the parentheses ; For pattern clusters Geographic entity pairs within; S500.4 Trend Prediction Model: Constructing a spatiotemporal regression model to predict future states ; S500.5 Result Verification and Output: The prediction accuracy is evaluated through cross-validation, and the mean squared error is calculated. : ； in, To verify the sample size; for The real spatiotemporal feature vector at any given moment; for The predicted spatiotemporal feature vector at any given time; when When the threshold is reached, the prediction result is output. and confidence interval ,in The confidence interval is half-width.

9. The GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks according to claim 1, characterized in that, The visualization interaction unit (600) includes a multi-terminal rendering module (610), an interaction control module (620), a data visualization module (630), a performance optimization module (640), and an event bus module (650), wherein: The multi-terminal rendering module (610) is used to realize cross-terminal adaptive rendering of three-dimensional scenes and adjust the display parameters according to the device characteristics based on adaptive layout technology. The interactive control module (620) is used to implement multimodal interactive operations and respond to user selections and operation commands based on the interactive event processing mechanism; The data visualization module (630) is used to realize the visualization display of thematic maps and feature plotting, and converts geographic information into graphic representation based on data mapping technology; The performance optimization module (640) is used to improve rendering efficiency and optimize the display effect of complex models based on scene simplification strategies; The event bus module (650) is used to realize communication and cooperation between modules, and to distribute and process interactive events based on the message passing mechanism.

10. A GIS 3D dynamic modeling and analysis method integrating big data and cloud platform networks, based on the GIS 3D dynamic modeling and analysis system integrating big data and cloud platform networks as described in any one of claims 1-9, characterized in that, Includes the following steps: S100 Multi-source geographic data aggregation and access: Aggregates and accesses multiple types of geospatial data, adapts to different data sources through standardized protocol conversion, dynamically adjusts the transmission rate and supports breakpoint recovery of transmission, automatically extracts metadata containing information such as collection time and spatial coordinates, registers data sources, monitors their status and provides anomaly alarms, and ensures data validity through standardized preprocessing and quality control. S200, Distributed Geographic Data Processing: Parallel feature extraction and spatiotemporal correlation calculation of geographic data; spatial partitioning based on spatial grid index and associated with unique identifiers; extraction of multiple types of features such as terrain and buildings and support for multi-scale configuration; construction of spatiotemporal index to support millisecond-level queries; real-time monitoring of computing node load and dynamic adjustment of task allocation; and inter-module communication through message queues to improve data processing efficiency. S300, cloud platform resources and storage support: provides elastic resource scheduling and cross-regional data storage services, dynamically allocates computing resources based on container orchestration and supports service expansion, builds a private cloud-public cloud bidirectional synchronization channel and schedules multi-cloud resources according to business priority, and adopts a layered storage architecture to store high-frequency access, recent and historical data respectively. S400, 3D dynamic model construction and updating: Constructing terrain triangular mesh and parametric model of geographic scene, performing format parsing, coordinate calibration and feature matching on multiple types of geographic data to achieve fusion and correct spatial deviation, constructing terrain and feature models based on fused data and adapting to different rendering precision through hierarchical strategy, detecting scene changes by comparing data with different timestamps, calculating incremental data and injecting it into the model to refresh the status in real time. S500, 3D model intelligent analysis: performs multi-dimensional geospatial analysis on terrain triangular mesh and parametric model of land features, extracts spatial, temporal and attribute feature vectors through spatiotemporal cube model, constructs spatiotemporal correlation matrix to quantify the dependency relationship between geographic entities, adopts improved clustering algorithm based on correlation matrix to identify spatiotemporal pattern clusters, constructs regression model to predict future state, evaluates accuracy through cross-validation and outputs prediction results and confidence interval, realizing pattern mining and trend prediction; S600, 3D Scene Visualization and Interaction: Enables multi-terminal presentation and interactive operation of 3D scenes. Based on adaptive rendering technology, it adjusts display parameters according to device characteristics to adapt to multiple terminals. It supports multi-modal interactive operation to respond to user commands, converts geographic information into graphic representation to realize thematic maps and feature plotting display, optimizes the display effect of complex models to improve rendering efficiency, and realizes inter-module communication and collaboration through message passing mechanism.

Images