A method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy

Abstract

This invention discloses a dynamic hierarchical storage and visualization method for geospatial data based on spatiotemporal entropy, comprising the following steps: First, through a data access layer, static attribute data and spatial neighborhood data of the target geographic entity are concurrently read from a heterogeneous database cluster to initialize the computational data. Second, a dynamic quantization model of spatiotemporal entropy is constructed to calculate the real-time spatiotemporal entropy value of the geographic entity. Based on the comparison result between the spatiotemporal entropy value and a preset threshold, the geographic entity data is dynamically scheduled to a hot data layer, a warm data layer, or a cold data layer. Third, visualization loading and rendering are performed based on the storage hierarchy. The front-end rendering engine performs differentiated geometric resource loading and visual shading according to the storage hierarchy. This invention ensures the visibility of key information in emergency or high-concurrency access scenarios while reducing the overall system hardware cost requirements.

Description

Technical Field

[0001] This invention relates to the field of geospatial technology, specifically to a method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy. Background Technology

[0002] In the field of urban housing safety lifecycle management, 3D spatial data is massive (TB / PB level) and updated frequently. Geographic Information Systems (GIS) not only need to render massive housing spatial grid objects (such as B3DM format), but also need to handle complex related attributes (such as construction year and structural level). Especially when dealing with sudden public safety events such as typhoons and earthquakes, computer systems not only need to maintain the smooth display of large-scale static map scenes, but also need to handle massive dynamic and changing emergency data. This poses a severe challenge to the I / O resource scheduling strategy and data throughput capacity of the underlying geospatial database. The existing mainstream geographic information systems (GIS) and spatial database architectures have the following deep-seated computer technology defects when dealing with large-scale, highly dynamic data application scenarios: (1) Static storage cannot evolve automatically over time and cannot perceive changes in data popularity: In existing spatial databases (such as PostgreSQL / PostGIS), the stored geographic entity attributes are usually static records. For example, the "time attribute" of the data is stored as a regular text field. Unless the external system explicitly initiates an UPDATE command, the database kernel cannot perceive the weight change (entropy increase) of the data as the system clock progresses. This passive storage mechanism leads to a serious "spatiotemporal decoupling" between the index priority inside the database and the actual business value of the data (such as urgency and risk). In high-concurrency scenarios where the system needs to retrieve the "most critical data at the current moment", the database cannot dynamically adjust the retrieval weight based on time, resulting in a low query hit rate and a long retrieval response time. (2) Indiscriminate resource loading leads to slow response of critical data: Existing 3D rendering engines (such as Cesium and Unreal Engine) generally use the LOD algorithm, which is driven solely by "visual distance", resulting in the computer system being in a "blind" state in terms of the calculation priority of data content. In emergency scenarios such as typhoon landfall, emergency commanders may urgently need to view the detailed condition of a distant cluster of dilapidated buildings. However, because these buildings are far from the virtual camera, the LOD algorithm mechanically assigns them low priority, resulting in slow loading or even failure to load. Conversely, numerous nearby low-risk buildings (such as sturdy new landmarks and residential buildings) occupy valuable server I / O bandwidth and memory resources due to their large visual footprint. This "egalitarian" allocation of resources prevents the system from concentrating its limited computing power on high-risk building data at critical moments, causing a severe data throughput bottleneck. Summary of the Invention

[0003] Purpose of the Invention: The main purpose of this invention is to provide a dynamic hierarchical storage and visualization method for geospatial data based on spatiotemporal entropy. By constructing a mapping model between physical and digital space, it drives the dynamic flow of data between the underlying storage medium and the top-level rendering pipeline. This solves the problems of static storage failing to evolve automatically over time and failing to detect changes in data popularity, as well as the problem of indiscriminate resource loading leading to slow response times for critical data.

[0004] Technical solution: The present invention provides a method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy, comprising the following steps:

[0005] S1: Through the data access layer, static attribute data and spatial neighborhood data of the target geographic entity are read concurrently from the heterogeneous database cluster to complete the initialization of the calculation data;

[0006] S2: Construct a dynamic quantification model of spatiotemporal entropy to calculate the real-time spatiotemporal entropy value of geographic entities, using the following formula:

[0007] ;

[0008] in, Characterizing geographic entities exist The dimensionless spatiotemporal entropy value at any given moment; Normalized weighting coefficients representing the contributions of physical aging, intrinsic structure, and environmental neighborhood factors to the overall spatiotemporal entropy; and satisfying... ; Characterizes the current lifespan of the target entity; Characterizes the design lifespan of the target entity; Characterizing aging-accelerating factors; Characterizing structural vulnerability parameters (e.g., basic risk values ​​determined by structural types such as timber, brick-concrete, and frame structures). Representation by geographic entities The set of effective spatial neighborhoods within a radius R centered at R; The first in the neighborhood set Discrete risk status values ​​of adjacent entities (1 for high risk, 0 for normal); Characterizing the first The size of each neighbor's influence factor; Characterizing geographic entities With the Spatial distance between neighboring entities; Characterizing the trust decay constant; It represents the time interval since the last data verification or on-site inspection of a geographic entity (i.e., the inspection vacuum period).

[0009] S3: Based on the comparison results between the spatiotemporal entropy value and the preset threshold, the geographic entity data is dynamically scheduled to the hot data layer, warm data layer or cold data layer.

[0010] S4: Visual loading and rendering are based on storage hierarchy. The front-end rendering engine performs differentiated geometric resource loading and visual shading according to the storage hierarchy.

[0011] Furthermore, in step S2, the model coefficients The determination method is as follows: a dataset containing historical geographical entity safety accident samples is established, and machine learning algorithms are used to assess the importance of features, calculate the contribution rate of service life, structure type and environmental neighborhood risk to the occurrence of accidents, and perform normalization processing to obtain the initial weights.

[0012] Furthermore, in step S2, the aging acceleration factor The determination method is as follows: extract the service life-failure data of historical geographical entities, use Weibull distribution for survival analysis fitting, and solve for the curvature parameter of the material performance degradation curve.

[0013] Furthermore, in step S2, the size influence factor The determination method is as follows: based on the actual physical scale parameters of the geographic entity, the neighboring nodes are divided into multiple contribution levels and assigned different... value.

[0014] Furthermore, in step S3, the scheduling logic for the hot data layer, warm data layer, and cold data layer includes: when >= At that time, the data resides in an in-memory database, and an active push channel is established; when < < When the LRU caching strategy is used, memory is loaded from disk on demand; when... <= At that time, only the index is retained in the relational database, and the model file is stored in a cold storage medium; among which Tcritical and T normal This refers to the spatiotemporal entropy value corresponding to the quantile, dynamically calculated based on the histogram of entropy distribution of the full data.

[0015] Furthermore, in step S4, the differentiated loading includes: for the hot data layer, loading the 3D model data directly from memory; for the warm data layer, loading from disk on demand based on the line-of-sight drive; for the cold data layer, only loading the 2D outline by default, and triggering the 3D model loading only when the user interacts.

[0016] Furthermore, the visual shading described in step S4 includes: for the hot data layer, overlaying a red semi-transparent highlight style on the 3D model; for the warm data layer, rendering realistic photographic textures; and for the cold data layer, drawing only the edge outlines by default.

[0017] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention maps the “entity aging” of three-dimensional geospatial data in the physical world, the “neighborhood risk” of geospatial space and the “confidence decay” in information theory into a dimensionless value that can be recognized by computers—spatiotemporal entropy, and drives the underlying data governance and final rendering display. The invention overcomes the performance bottleneck and logical defects of the prior art in processing massive three-dimensional geographic information data: (1) Constructing a spatiotemporal entropy model eliminates the spatiotemporal lag between static data and dynamic entities; The present invention constructs a model containing nonlinear functions, which enables the computer system to automatically and dynamically calculate and update the risk weight of three-dimensional geographic entities in real time according to the system clock without relying on external manual update instructions. This realizes the active approximation of the state of physical entities by digital objects at the data structure level, and eliminates the “spatiotemporal lag” caused by static data storage. (2) It breaks through the I / O throughput bottleneck of massive spatial data in high-concurrency disaster scenarios. In response to the high-concurrency reading pressure faced by the database during disasters such as typhoons, this invention abandons the traditional "line-of-sight priority" scheduling strategy of graphics and implements dynamic storage migration based on statistical quantiles. This method forces the calculated "high-entropy data" to reside in the memory cluster (such as Redis), so that the reading of core data is no longer limited by the I / O bandwidth and seek time of mechanical hard disk. Experiments show that under high-concurrency stress test with tens of millions of data, this method can significantly improve the data throughput performance of the system. (3) It realizes visualization loading and rendering based on entropy value, concentrates the limited GPU computing power and video memory resources to high-risk targets, and realizes "using the best steel on the blade". It not only ensures the visibility of key information in emergency or high-concurrency access scenarios, but also reduces the hardware cost requirements of the overall system. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention;

[0019] Figure 2 This is a schematic diagram of the calculation logic of the secure spatiotemporal entropy model of the present invention (corresponding to step S2);

[0020] Figure 3 This is a diagram of the entropy-driven dynamic scheduling architecture for storage media of the present invention (corresponding to step S3);

[0021] Figure 4 This is a visual loading and rendering diagram of the present invention (corresponding to step S4). Detailed Implementation

[0022] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0023] Step S1: Multi-source data reading and initialization

[0024] This step is the data preparation phase. The system concurrently reads the target geographic entities (buildings) from the heterogeneous database cluster through the data access layer. The static attribute space neighborhood data is used to initialize the computational data.

[0025] Static attribute data reading: The system reads the structural type of the target geographic entity through the data interface provided by the backend database. Current target entity's service life Design life ( Data such as historical patrol logs, etc.

[0026] Spatial neighborhood data retrieval: Utilizing the R-Tree spatial index of the database, quickly retrieve a set of a finite number of neighborhood objects within a radius R centered on the current target geographic entity. For each neighbor in the set Read static attribute data for use in subsequent model calculations.

[0027] Step S2: Construct a dynamic quantification model of "spatiotemporal entropy" (constructing an entropy model)

[0028] like Figure 2 As shown, a real-time computing engine is deployed on the server side. Through a multi-dimensional coupling model, the aging characteristics of 3D geospatial data in the physical dimension, the correlation effects in the geographical dimension, and the uncertainty in the information dimension are mapped into a dimensionless low-level computer control signal—spatiotemporal entropy. This value, from a computer science perspective, represents the priority weight of the geographic entity data object in system resource scheduling. Its detailed construction process is as follows:

[0029] Model building logic

[0030] a. Physically Intrinsic Weight Dimension (Intrinsic Risk Item): The model is based on the inherent physical properties of the entity (the target entity's lifespan). Structure type (etc.), introducing an exponential function to simulate the performance degradation characteristics of data-related entities. This design can accurately characterize the geometric growth of entity data weights over time, enabling the scheduling priority of aging data to automatically jump at the algorithm level.

[0031] b. Spatial Neighborhood Convolution Dimension (Neighborhood Risk Field): Inspired by spatial interaction models (such as the gravitational model), it calculates the set of neighborhoods within a specific radius R centered on the target object. Impact on the current object. The model propagates the discrete state values ​​of surrounding neighbors and their volume factors inversely proportional to the square of the distance towards the center. This allows the system to keenly capture the chain reaction of weight increases caused by surrounding geological subsidence, construction, or failure of adjacent nodes, realizing the weight radiation of "critical high-weight data" to the surrounding areas.

[0032] c. Information Entropy Correction Dimension (Reliability Decay Term): This term, based on information theory principles, characterizes the uncertainty of the system's perception of the state of geographical objects. It affects the period of data not being verified / updated (i.e., the patrol vacuum period). The extension of timeliness of stored data in the system leads to a monotonically increasing cognitive uncertainty (entropy value), which in turn amplifies the overall weight in the calculation.

[0033] The dynamic evolution calculation model of "spatiotemporal entropy" constructed in this invention has the following formula:

[0034] ;

[0035] in: For geographical entities (such as houses) )exist The dimensionless spatiotemporal entropy value at any given moment; These are normalized weight coefficients, and satisfy... The meanings of each part of the model are explained below:

[0036] Physical endogenous risk item In this section, It is a nonlinear aging function that simulates the physical properties of the target entity whose material properties degrade exponentially over time. As an aging acceleration factor, For the target entity's service life, Designed for service life; The structural fragility parameter, representing the inherent properties of an entity (e.g., timber structure > brick-concrete structure > frame structure), serves as a basic weighting term. This part transforms the linear passage of time into a geometric progression of data weights, automatically increasing the priority of "older" data at the computational level.

[0037] Environmental neighborhood risk field In this section, The project was constructed with the inspiration of the universal gravitation model, in which For the target entity (such as a house) Let R be the set of neighborhoods within the center and radius R. For the neighbors The discrete value of risk status (1 for high risk, 0 for normal). For the neighbors The size of the risk is a factor that influences risk, representing the proximity transmission effect of risk in geographic space. Characteristic Entity and neighboring entities Spatial distance. This part can quantify the correlation and influence between data objects in geospatial space, and realize the weight transmission of "high-risk data" to surrounding "ordinary data".

[0038] Information entropy correction term In this section, For the inspection vacuum period (number of days). This represents the trust decay constant. This part characterizes how the uncertainty of the system's perception of the state of geographical entities increases monotonically over time as information acquisition lags, thus amplifying the risk.

[0039] Among them, the spatiotemporal entropy model coefficients in step S2 Method for determining:

[0040] Determination of (normalized weight coefficients): Establish a dataset containing safety accident samples and health samples of historical geographical entities (such as buildings); use machine learning algorithms (such as random forest, XGBoost, etc.) to evaluate the importance of features and calculate the contribution rate of age, structure type and environmental neighborhood risk to the reduction of accident occurrence; normalize the above contribution rates to obtain the initial weights.

[0041] Determination of (Aging Acceleration Factor): Extract "service life-failure" data of historical geographic entities (such as a building), perform survival analysis fitting using a statistical distribution model (e.g., Weibull distribution), and inversely solve for the curvature parameter of the material performance degradation curve of the geographic entity. .

[0042] (Neighbor Determination of the size influence factor: A hierarchical discrete assignment method based on the magnitude attributes of geographic entity data is adopted. The system assigns neighboring nodes according to the actual physical scale parameters (such as height attribute field or spatial area component) corresponding to the geographic entity data object. Contribution levels are divided into three priority levels: low contribution level (such as low-rise residential buildings and small geographical facilities) =0.5; Baseline contribution level (e.g., ordinary multi-story buildings): Set =1.0; High contribution level (such as high-rise landmarks or large complexes): Set =2.0. This assignment logic simulates the characteristic of a large mass object with a strong energy field in the universal gravitation model, and represents that the larger the size of a geographical entity, the more significant its ripple effects on the surrounding area (such as the collapse range or foundation disturbance) when a risk occurs.

[0043] Determination of (trust decay constant): Based on the maximum permissible patrol interval stipulated in the management regulations. and preset risk multiplication factors Through formula The calculation shows that, among which The value of is determined based on practical experience. This ensures that the risk value automatically jumps to the warning range when the patrol vacuum period reaches its limit.

[0044] It should be noted that the above formula is merely a preferred calculation model provided by an embodiment of the present invention. Those skilled in the art will understand that the formula structure can be modified without departing from the core logic of the present invention. For example, a nonlinear aging function... It can be replaced with other monotonically increasing nonlinear functions (such as logarithmic or power functions); the distance decay factor 1 / d in the neighborhood risk field. 2 It can also be adjusted to 1 / d or 1 / d based on the actual city density. 3 All coupled computational models based on the three dimensions of physical aging, neighborhood transmission, and information entropy increase should fall within the protection scope of this invention.

[0045] This value is not only a mathematical representation of the current risk, but also the only underlying computer control signal for subsequent steps S3 to allocate storage resources and S4 to output visualizations.

[0046] Step S3: Perform dynamic scheduling of multi-level storage media based on entropy (entropy-driven underlying computer storage method)

[0047] like Figure 3 As shown, the real-time entropy signal output in step S2 The system uses this as the sole control parameter for underlying physical resource scheduling, completing the mapping from "logical weight" to "physical medium." This step reconstructs the physical storage logic of data within the computer, abandoning the indiscriminate loading strategy and establishing a mapping relationship between entropy values ​​and physical media. This is achieved through directly calculated risk entropy values. A three-level closed loop of "hot, warm, and cold" dynamic flow was constructed, and dynamic upgrade and downgrade scheduling of each layer based on real-time entropy values ​​was implemented:

[0048] (31) Hot data layer ( >= ): When the spatiotemporal entropy value of a geographic entity Greater than the set threshold When the system is in a high-risk or extremely active state, the backend triggers a "memory-resident" strategy. It serializes and writes data such as the spatial matrix, business status, and 3D model binary stream into memory (e.g., Redis), while simultaneously establishing a WebSocket channel for proactive pushing, ensuring the data remains resident in memory and achieving millisecond-level latency responses under any high concurrency.

[0049] (32) Temperature data layer ( < < For regular geographic entities, an LRU (Least Recently Used) caching strategy is adopted. Data is loaded from disk into memory on demand. If there are no access requests within a set time, the system will automatically release it back to disk, achieving a dynamic balance between response speed and memory usage.

[0050] (33) Cold data layer ( <= For massive amounts of low-risk geographic entities, a "index-resident, entity-cold storage" strategy is adopted. The system only retains lightweight index pointers in the relational database, while the massive model physical files are kept in a cold disk storage state, without consuming any memory resources. I / O reads are only triggered when the user explicitly initiates a request, thus supporting the operation of massive city-level data with limited hardware costs.

[0051] The system dynamically schedules geographic 3D spatial data across different levels based on the real-time entropy values ​​of geographic entities.

[0052] Where T critical and T normal This refers to the spatiotemporal entropy value corresponding to the quantile based on the histogram of entropy distribution of the full data. For example, if we take the 95th and 40th quantiles, which correspond to entropy values ​​of 2.5 and 1.0 respectively, then Tcritical=2.5 and Tnormal=1.0. The system updates this threshold periodically through a background scheduled task to adapt to the dynamic changes in the data scale.

[0053] This strategy, based on an entropy model, achieves "absolute tilting" of computing resources towards high-risk targets and "intelligent load distribution" of system load. In high-concurrency access scenarios caused by sudden disasters such as typhoons and rainstorms, high-risk geographic entity data (hot data) is directly distributed through memory, achieving millisecond-level latency response; at the same time, massive historical data (cold data) does not occupy memory, effectively solving the throughput bottleneck between limited computing power and massive data, enabling the system to support city-level data scales even on low-cost hardware.

[0054] Step S4: Perform visual loading and rendering

[0055] like Figure 4 As shown, the front-end rendering engine no longer relies solely on LOD (Level of Detail) distance, but instead acts as the execution terminal for step S3, performing differentiated loading and shading, and ultimately rendering a visualized 3D view on the computer screen.

[0056] Phase 1: Differentiated loading of geometric resources (receiving output from S3)

[0057] The rendering engine first acquires the geometric model data from the scene. For hot data (high risk) in step S3: the engine directly reads the model data from memory at high speed, ensuring that the 3D model of high-risk targets appears instantly on the map. For warm data (medium risk) in step S3: the engine uses a standard disk / SSD on-demand read strategy, loading 3D model data only when the virtual camera approaches and unloading it when it moves out of view, balancing performance. For cold data (low risk) in step S3: the engine by default only loads lightweight indexes and displays only the 2D outlines of geographic entities as placeholders on the map, without loading the 3D model. I / O requests are only initiated and 3D model data is loaded when the user clicks on the 2D outline.

[0058] Phase Two: Visual Coloring

[0059] The rendering engine reads the real-time entropy value calculated in step S3 and uses it as the sole control variable to input into the fragment shader, calculating the pixel color of the final 3D model: For hot data (high risk) in step S3: a red semi-transparent highlight style is forcibly overlaid on the 3D model. For warm data (medium risk) in step S3: the original real photographic texture of the geographic entity is loaded and rendered. For cold data (low risk) in step S3: if the user does not click, only the edge outline of the geographic entity is drawn, and the complex 3D model is not loaded; an I / O request is initiated and the 3D model data is loaded only when the user clicks on the 2D outline.

[0060] Taking the S City Housing Safety Management System as an example, the system's underlying layer contains approximately 200,000 three-dimensional building geographic entities (stored in B3DM mesh model format with accompanying JSON attributes). The following demonstrates the system's dynamic governance chain for "Data Node ID-1001 (formerly Building A of Community X)".

[0061] Step S1: Multi-source data reading and initialization

[0062] Through the data access layer interface, read the attribute data of the target object "Building A, Community x" and its spatial neighborhood:

[0063] Suppose the data retrieved is "Building A, Community X", which is a brick-concrete structure. Building age 30 years, the design life of a house =50 years. No dilapidated buildings or large-scale construction within a radius R=100 meters (all houses) 0), and the house has not been inspected for safety for two consecutive years. .

[0064] Step S2: Construct a spatiotemporal entropy model and calculate the real-time entropy value of the building.

[0065]

[0066] (1) Parameter calibration:

[0067] Based on historical housing accident data from City S over the past 10 years, feature importance was determined through machine learning feature importance assessment: Physical Endogenous Risk Weight. Structural vulnerability weight Environmental neighborhood risk weight .

[0068] We extracted the "age-failure" data of 10-year-old failed houses in City S, and used the Weibull distribution for analysis and fitting to solve for the curvature parameter (i.e., aging acceleration factor) of the material performance degradation curve. .

[0069] Assuming the maximum permissible patrol interval Annual risk multiplication factor Through formula The trust decay coefficient was calculated. = 0.002 (meaning that for every additional day without inspection, the uncertainty increases by 0.2%).

[0070] (2) Model calculation:

[0071] Then the spatiotemporal entropy value of building A in community x

[0072]

[0073] Although the building's physical structure is acceptable, the long inspection vacuum period significantly amplifies the information uncertainty (entropy value), resulting in a high calculated spatiotemporal entropy value (2.822).

[0074] Similarly, we can calculate the entropy value of all houses in community x. Let's assume there are 30 houses of varying sizes in this community: 2 houses (including building A) have an entropy value above 2.5 (classified as the first group); 20 houses have an entropy value between 1.0 and 2.5 (classified as the second group); and 8 houses have an entropy value less than 1.0 (classified as the third group). This is to more clearly illustrate the entropy-driven multi-level storage scheduling process in step S3.

[0075] Step S3: Execute entropy-driven multi-level memory scheduling

[0076] Based on the dynamically calculated quantiles (e.g., taking the 95th and 40th percentiles respectively) from the histogram of entropy distribution of the current full housing data in City S, assuming the following... , .

[0077] According to the calculation in step S2, the following was found:

[0078] First group of houses: Spacetime entropy > The backend service immediately triggers the "memory-resident" strategy, serializing the binary stream data of the building's 3D model and writing it to the Redis memory cluster to ensure millisecond-level response speed under high concurrency access.

[0079] Second group of houses: Spacetime entropy value Between , Between. An LRU (Least Recently Used) caching strategy is used, where data is loaded from disk into memory only on demand when accessed.

[0080] The third group of houses: Spacetime entropy < The system employs an "index-resident, entity-cold storage" strategy. Only lightweight indexes are retained in the relational database, while the large physical model files remain in a cold disk storage state, consuming no memory resources.

[0081] Step S4: Perform visual loading and rendering

[0082] Phase 1: Front-end rendering: The engine receives the storage level instructions from step S3, executes differentiated resource loading and shading logic, and displays it on the map: Differentiated loading of geometric resources:

[0083] For the first group (hot data layer, such as Building A): The engine directly retrieves the 3D grid binary stream from the Redis memory cluster at high speed, which can achieve "instant rendering" at the millisecond level, ensuring that critical information is not lost or delayed during emergency response.

[0084] For the second group (warm data layer): The engine executes a standard disk / SSD on-demand read strategy (based on LOD view distance). I / O requests are only initiated to load the model into video memory when the virtual camera approaches the houses in this group and the view frustum is detected as visible; when the view moves out, video memory resources are released, achieving a balance between memory usage and visual continuity.

[0085] For the third group (cold data layer): By default, the engine only reads lightweight 2D contours from the database. Flattened edge contours are drawn as placeholders in the 3D scene, without loading complex B3DM model files, thus minimizing GPU computing resources and system bandwidth. If the user clicks to trigger loading, the 3D model data is loaded into memory.

[0086] Phase Two: Visual Coloring

[0087] For the first group (high-entropy houses, such as Building A): The engine forcibly overlays a layer of red semi-transparent highlight on top of the original texture of the 3D model.

[0088] For the second group (medium-entropy houses): The engine maintains the realistic 3D model photographic texture of the geographic entities for regular rendering, providing an accurate physical appearance reference.

[0089] For the third group (low-entropy houses): If there is no user interaction, it is only displayed as a gray semi-transparent outline; once it is clicked to trigger loading, the 3D model data is loaded and rendered according to the original texture of the 3D model.

Claims

1. A method for dynamic tiered storage and visualization of geospatial data based on spatiotemporal entropy, characterized in that, Includes the following steps: S1: Through the data access layer, static attribute data and spatial neighborhood data of the target geographic entity are read concurrently from the heterogeneous database cluster to complete the initialization of the calculation data; S2: Construct a dynamic quantification model of spatiotemporal entropy to calculate the real-time spatiotemporal entropy value of geographic entities, using the following formula: ； in, Characterizing geographic entities exist The dimensionless spatiotemporal entropy value at any given moment; Normalized weighting coefficients representing the contributions of physical aging, intrinsic structure, and environmental neighborhood factors to the overall spatiotemporal entropy; and satisfying ; Characterizes the current lifespan of the target entity; Characterizes the design lifespan of the target entity; Characterizing aging-accelerating factors; Characterizing structural vulnerability parameters; Representation by geographic entities The set of effective spatial neighborhoods within a radius R centered at R; The first in the neighborhood set Discrete values ​​of the risk status of each adjacent entity; Characterizing the first The size influence factor of each neighboring entity; Characterizing geographic entities With the Spatial distance between neighboring entities; Characterizing the trust decay constant; The time interval representing a geographic entity since the last data verification or on-site inspection; S3: Based on the comparison results between the spatiotemporal entropy value and the preset threshold, the geographic entity data is dynamically scheduled to the hot data layer, warm data layer or cold data layer. S4: Based on storage layer-level progressive visualization loading and rendering, the front-end rendering engine performs differentiated geometric resource loading and visual shading according to the storage layer; the differentiated loading is specifically: for the hot data layer, the 3D model data is loaded directly from memory; For warm data layers, load from disk on demand based on line-of-sight; for cold data layers, only load 2D outlines by default, and 3D model loading is triggered only when the user interacts; the specific process of visual shading is as follows: for hot data layers, overlay a red semi-transparent highlight style on the 3D model; for warm data layers, render realistic photographic textures; for cold data layers, only draw edge outlines by default.

2. The method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy according to claim 1, characterized in that, In step S2, the model coefficients The determination method is as follows: a dataset containing historical geographical entity safety accident samples is established, and machine learning algorithms are used to assess the importance of features, calculate the contribution rate of service life, structure type and environmental neighborhood risk to the occurrence of accidents, and perform normalization processing to obtain the initial weights.

3. The method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy according to claim 1, characterized in that, In step S2, aging acceleration factor The determination method is as follows: extract the service life-failure data of historical geographical entities, use Weibull distribution for survival analysis fitting, and solve for the curvature parameter of the material performance degradation curve.

4. The method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy according to claim 1, characterized in that, In step S2, the body size influence factor The method for determining this is: based on neighboring geographical entities. The actual physical scale parameters are used to divide the neighboring nodes into multiple contribution levels and assign different values. value.

5. The method for dynamic hierarchical storage and visualization of geospatial data based on spatiotemporal entropy according to claim 1, characterized in that, In step S3, the scheduling logic for the hot data layer, warm data layer, and cold data layer is as follows: When >= At that time, the data resides in an in-memory database, and an active push channel is established; when < < When the LRU caching strategy is used, memory is loaded from disk on demand; when... <= At that time, only the index is retained in the relational database, and the model file is stored in a cold storage medium; where T critical and T normal This refers to the spatiotemporal entropy value corresponding to the quantile, dynamically calculated based on the histogram of entropy distribution of the full data.

Images