Virtual space-time environment automatic completion method and system based on data density dynamic perception and medium

Abstract

The invention discloses a virtual space-time environment automatic completion method and system based on data density dynamic perception and a medium, and relates to the field of computer data processing. The method comprises the following steps: monitoring a data activeness index (such as spatio-temporal information entropy) of a spatio-temporal grid in real time in a distributed index database of a virtual spatio-temporal environment; when the index meets a preset data sparseness condition, generating a completion request; analyzing the space-time attribute tag of the target grid and acquiring associated multi-modal reference data from an external heterogeneous data source; carrying out feature recombination on the reference data by utilizing a generative neural network model, and generating discrete spatio-temporal data slices containing spatio-temporal anchor points and scene content attributes (such as environment illumination and material mapping); and finally, injecting the slices into the database to fill data holes. According to the method, a passive trigger mechanism and a structured reconstruction technology of unstructured data are introduced, so that the computing power cost of a large-scale virtual environment is effectively reduced, the problems of data sparseness and cold start of a long-tail region are solved, and the historical authenticity and time-space consistency of a virtual scene are ensured.

Description

Technical Field

[0001] This invention relates to the fields of computer data processing, artificial intelligence content generation (AIGC), and virtual reality technology, and particularly to a method, system, electronic device, and computer-readable storage medium for automated completion of virtual spatiotemporal environments based on dynamic data density perception. Background Technology

[0002] With the development of digital twins, augmented reality (AR), and next-generation immersive internet technologies, building large-scale virtual spatiotemporal environments that run parallel to the physical world has become a significant industry trend. These systems aim to reconstruct the geographical features and historical memories of the physical world through digital technology, providing users with immersive interactive experiences that transcend the limitations of time and space. In these systems, the core asset is data slices with precise four-dimensional labels (three-dimensional spatial coordinates X / Y / Z + time dimension T).

[0003] However, existing technologies face significant challenges in constructing and maintaining such global-scale virtual spatiotemporal environments, primarily in the following three aspects:

[0004] 1. The "Long-Tail Sparsity" and "Cold Start" Problems of Spatiotemporal Data: The physical world is extremely vast, and existing user-generated content (UGC) or professionally produced content (PGC) is often highly concentrated in popular landmarks (such as city centers and famous tourist attractions). For the vast long-tail areas (such as ordinary streets, rural areas, and lesser-known historical sites), the virtual environment often lacks effective interactive data, forming a large number of "data voids." When users wander into these areas in the virtual environment or try to trace back the historical timeline of these areas, the system cannot provide effective content feedback, resulting in a break in spatiotemporal continuity and severely impacting the user's immersion.

[0005] 2. The Cost-Quality Contradiction of Traditional Filling Methods To fill the aforementioned data gaps, existing technologies mainly employ two methods:

[0006] Manual editing: This method relies on professional teams to manually collect data and create models. It is extremely costly and cannot meet the needs of global-scale geographic grid coverage.

[0007] Procedural Content Generation (PCG) uses algorithms to automatically generate terrain, vegetation, or generic buildings. However, PCG-generated content is often based on "pseudo-data" from random seeds, lacking historical authenticity and semantic relevance. For example, an algorithm might generate a tree at a given coordinate, but it cannot generate information about "specific historical events that occurred at that location 50 years ago" or "unique old shop signs at that location."

[0008] 3. The disconnect between heterogeneous data sources and virtual environment rendering engines: The internet contains massive amounts of historical data (such as encyclopedia entries, news archives, and old photos), which are potential resources for solving the "data void." However, this data is usually unstructured or semi-structured and distributed across heterogeneous external data sources.

[0009] Existing search engine technologies (such as Google / Baidu) can only return webpage links or text summaries and cannot understand the spatiotemporal indexing architecture of virtual environments.

[0010] Virtual environment rendering engines require structured assets (such as 3D slices with precise XYZT coordinates, ambient lighting parameters, and material properties). Currently, there is a lack of technical solutions that can automatically convert "unstructured information from the Internet" into "structured spatiotemporal slices that can be recognized by the rendering engine" and automatically inject them into a spatiotemporal index database.

[0011] In summary, there is an urgent need for a technology that can automatically detect data sparsity and intelligently utilize external multimodal data to generate structured data slices with historical realism and adaptable to distributed index databases, so as to achieve low-cost and automated completion of virtual spatiotemporal environments. Summary of the Invention

[0012] (I) Purpose of the Invention

[0013] The main objective of this invention is to provide an automated completion method, system, electronic device, and computer-readable storage medium for virtual spatiotemporal environments based on dynamic data density perception. This invention aims to solve the technical problems existing in the construction of large-scale virtual spatiotemporal environments, such as data sparsity in long-tail regions, high costs of manual filling, and the difficulty for rendering engines to directly access unstructured Internet data.

[0014] Specifically, this invention aims to achieve the following technical objectives:

[0015] 1. Enable automatic detection and on-demand generation of data gaps in the virtual environment, avoiding the waste of computing power in full generation.

[0016] 2. Establish an automated conversion link from multimodal Internet reference data to structured spatiotemporal slices to solve the "cold start" problem.

[0017] 3. Generate structured assets with spatiotemporal anchor points and scene content attributes (such as lighting and materials) to ensure the spatiotemporal consistency between automatically populated content and the virtual environment.

[0018] (II) Technical Solution

[0019] To achieve the above objectives, this invention provides an automated virtual spatiotemporal environment completion method based on dynamic data density perception, comprising the following steps:

[0020] State awareness and triggering steps: In the distributed index database of the virtual spatiotemporal environment, the data activity index is monitored in real time in units of preset spatiotemporal grids; the data activity index is used to characterize the richness of interactive data within the grid; when the data activity index of the target spatiotemporal grid meets the preset data sparsity condition, the system automatically generates a data completion request for the grid.

[0021] Heterogeneous data acquisition steps: In response to the data completion request, the system parses the spatiotemporal attribute labels (including geographic location codes and time windows) of the target spatiotemporal grid, and constructs a retrieval vector based on the labels to obtain multimodal reference data associated with the target area from external heterogeneous data sources; the external heterogeneous data sources include but are not limited to publicly available Internet indexes, existing data from geographic information systems (GIS), or IoT sensor logs.

[0022] The structured slice generation steps are as follows: a pre-trained generative neural network model is used to perform feature cleaning, denoising, and recombination on the multimodal reference data to generate discrete spatiotemporal data slices adapted to the virtual spatiotemporal environment; the discrete spatiotemporal data slice is a structured data package that encapsulates spatiotemporal anchor data and scene content attribute data.

[0023] Closed-loop injection step: The generated discrete spatiotemporal data slices are injected into the distributed index database through the consistency interface to fill the data holes of the target spatiotemporal grid, so that end users can obtain continuous virtual scene feedback when accessing the grid.

[0024] Furthermore, the data activity index is preferably spatiotemporal information entropy; the state perception and triggering steps specifically include: calculating the spatiotemporal information entropy based on the distribution probability of historical interaction data within the target spatiotemporal grid and combining it with a time decay function; when the calculated entropy value is lower than a preset threshold, it is determined that the data sparsity condition is met.

[0025] Furthermore, the scene content attribute data includes at least one of the following or a combination thereof:

[0026] Semantic descriptive features: Text summary data that describes historical events or geographic information of the target area, generated based on multimodal reference data;

[0027] Ambient lighting parameters: The main light source vector, light intensity (Lux) and color temperature data are derived from the image information and stored in JSONB format to drive real-time lighting and shadow rendering of the virtual scene;

[0028] Material mapping data: Surface roughness and reflectivity parameters of objects inferred from the texture features of retrieved images;

[0029] Meteorological status data: Historical weather characteristics corresponding to the timestamps of the spatiotemporal anchor data.

[0030] Furthermore, the discrete spatiotemporal data slices employ the following encoding logic to adapt to the distributed index database: extracting the spatial coordinate features of each slice. With time anchor features Using a preset hash operation or spatial curve filling algorithm, the spatial coordinate features and time anchor features are fused to generate a unique spatiotemporal key, which serves as the primary key of the slice in the database.

[0031] Furthermore, the present invention also provides an automated virtual spatiotemporal environment completion system, comprising:

[0032] The density monitoring module is used to calculate the data activity index of the target spatiotemporal grid in real time and trigger a completion request when the data sparsity condition is met.

[0033] The data acquisition proxy is used to call external data source interfaces to obtain multimodal reference data associated with the target area;

[0034] The slice generation engine is used to run generative neural network models and output discrete spatiotemporal data slices containing spatiotemporal anchor points and scene content attributes.

[0035] The database write interface is used to perform index key calculation and closed-loop injection of discrete spatiotemporal data slices.

[0036] In addition, the present invention also provides an electronic device and a computer-readable storage medium for implementing the above method.

[0037] (III) Beneficial Effects

[0038] Compared with the prior art, the present invention has the following significant advantages:

[0039] 1. On-demand allocation and dynamic optimization of computing resources

[0040] This invention achieves precise quantification of data density in virtual environments by introducing "spatiotemporal information entropy" as a monitoring indicator. The system only triggers a high-cost AI generation task when it detects a "data sparsity" state (i.e., a cold start region), rather than blindly generating a full global map. This "passive triggering" mechanism significantly reduces server computing power and storage costs, achieving optimal resource allocation.

[0041] 2. Enabled the automated conversion of unstructured information into structured 3D assets.

[0042] Existing technologies typically only display webpage snapshots or plain text returned by search engines, failing to integrate them into 3D scenes. This invention, through a generative model, can extract rendering attributes such as "ambient lighting parameters" and "material mapping data" from messy internet data, and encapsulate this information into standard XYZT discrete slices. This makes the generated completion content more than just a text pop-up; it becomes a virtual asset with immersive lighting and shadows that can be directly called by the rendering engine, significantly enhancing the user's time-travel experience.

[0043] 3. It ensures the historical authenticity and logical consistency of the virtual spatiotemporal environment.

[0044] This invention does not rely on randomly generated fictional scenes (as in traditional procedural generation techniques), but rather on reconstruction based on "multimodal reference data" (such as real historical photographs and documentary records). By combining an index key generation mechanism with spatiotemporal anchor features, it ensures that each generated data slice is precisely "anchored" to the correct historical time and geographical coordinates, thereby constructing a virtual parallel world with a sense of historical weight and authentic semantics.

[0045] 4. A data closed loop and ecosystem self-healing mechanism have been constructed.

[0046] Through a "closed-loop injection" and optional "observer consensus verification" mechanism, this invention transforms AI-generated data into permanent assets in a database, and continuously verifies and optimizes data quality through subsequent user interactions (such as user dwell time and feedback). This mechanism endows the virtual spatiotemporal environment with the ability to self-repair data gaps, enabling it to automatically grow and improve as the user's exploration scope expands. Attached Figure Description

[0047] Figure 1 is a schematic diagram of the overall process of an automated virtual spatiotemporal environment completion method based on dynamic data density perception provided in an embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of the data activity index monitoring and cold start triggering logic based on spatiotemporal grid in an embodiment of the present invention;

[0049] Figure 3 is a schematic diagram of the data structure and encoding logic of the discrete spatiotemporal data slice (XYZT Slice) generated in an embodiment of the present invention;

[0050] Figure 4 is a functional module architecture diagram of the virtual spatiotemporal environment automatic completion system provided in an embodiment of the present invention;

[0051] Figure 5 is a schematic diagram of the data closed-loop injection and observer consensus verification process in an embodiment of the present invention;

[0052] Figure 6 is a timing diagram of the data activity decaying over time in an embodiment of the present invention, which triggers the completion operation. Detailed Implementation

[0053] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments.

[0054] 1. Cross-Reference

[0055] In the description of this specification, the underlying storage and retrieval logic of the "virtual spatiotemporal environment", "distributed spatiotemporal index database" and "spatiotemporal index key" can be referred to the applicant's previous patent application with application number 2026100229634 entitled "A data processing method and system based on spatiotemporal index".

[0056] Specifically, after the "discrete spatiotemporal data slice" described in this invention is generated, it is preferably stored using the XYZT spatiotemporal discrete slice reconstruction technology described in the prior application; the "spatiotemporal index key" described in this invention is preferably using the spatial coordinate encoding described in the prior application. With time characteristics The XOR hash value is used as a unique identifier. This invention focuses on the automated generation source and construction process of the above-mentioned data slices, and is a further evolution and improvement of the prior application technical solution in the data production stage.

[0057] 2. System Architecture and Hardware Environment

[0058] Referring to Figure 4, the virtual spatiotemporal environment automated completion system provided by this invention can run on a distributed server cluster or a cloud computing platform. The system mainly includes the following logical units:

[0059] Density monitoring module: responsible for monitoring the read and write logs of distributed index databases (such as spatiotemporal databases built on PostgreSQL + PostGIS) and calculating the entropy value of each GeoHash grid.

[0060] AI Generation Agent Cluster: Deploys GPU computing nodes with Large Language Model (LLM) and Latent Diffusion Model to perform RAG (Retrieval Enhancement Generation) tasks.

[0061] Data closed-loop interface: Middleware used to connect external Internet data sources (such as search engine APIs, public GIS databases) with internal private spatiotemporal databases.

[0062] 3. Example 1: Overall Flow of the Automated Completion Method

[0063] Referring to Figure 1, the method provided in this embodiment of the invention mainly includes the following steps:

[0064] S101: State Awareness and Triggering

[0065] The system backend uses a preset spatial granularity (e.g., GeoHash level 6 grid, approximately...). The system divides the spatiotemporal grid by time granularity (e.g., "1990s-2000s"). The system monitors the data activity index within each grid in real time or periodically.

[0066] In this embodiment, the data activity index is not just the number of data entries, but the spatiotemporal information entropy based on information theory (see Embodiment 2 for details).

[0067] When the entropy value of a certain grid (e.g., a hutong in Xicheng District, Beijing, during the period of 1980) is monitored When the preset threshold is reached, the system determines that the area is a "digital desert" or meets the data sparsity condition, and automatically puts the XYZT label of the grid into the generation queue to generate a completion request.

[0068] S102: Heterogeneous Data Acquisition

[0069] The data acquisition agent responds to the completion request by parsing the spatiotemporal labels of the target grid (e.g., location = "X Hutong", time = "1980s"). The agent initiates a breadth-first search to external heterogeneous data sources via the API interface.

[0070] Data sources include: publicly available encyclopedia entries on the Internet (for historical events), image search engines (for old photos), and historical meteorological databases (for weather records from that year).

[0071] Cross-modal alignment: The system timestamps the acquired unstructured text and images. For example, if the retrieved text description is "heavy snow in 1980", historical images tagged with "snow scene" will be prioritized.

[0072] S103: Structured Slice Generation

[0073] This is the core step of the invention. Instead of simply storing webpage snapshots, the system uses a generative neural network model to reorganize messy information into renderable 3D asset attributes.

[0074] Scene content attribute extraction: Analyze retrieved historical photos using computer vision algorithms to infer the ambient lighting parameters (such as the direction of the main light source and the color temperature of 5600K) and material mapping data (such as the wall roughness of 0.8).

[0075] Slice encapsulation: The above parameters, along with spatial coordinates and time anchor points, are encapsulated into a standard JSONB object, i.e., discrete spatiotemporal data slice.

[0076] S104: Closed-loop injection

[0077] The generated slice is injected into the distributed index database through the database write interface. During this process, the system calculates the spatiotemporal index key of the slice according to the algorithm described in the applicant's previous patent application No. 2026100229634, entitled "A Data Processing Method and System Based on Spatiotemporal Index", ensuring that it can be quickly retrieved and loaded by the front-end application through LBS logic.

[0078] 4. Example 2: Spatiotemporal Information Entropy Calculation and Cold Start Triggering Algorithm

[0079] Referring to Figure 2, this embodiment details how to determine whether a grid needs to be completed by calculating the spatiotemporal information entropy.

[0080] In this invention, the system does not perform indiscriminate AI generation on all grids, but instead introduces an entropy algorithm with time decay weights. The specific calculation steps are as follows:

[0081] Step S201: Classification and Statistics of Interactive Data

[0082] For the target spatiotemporal grid The system first analyzes the distribution of existing interactive data within the grid. The interactive data is then divided into... Categories (e.g.: =Text message, =Image, =3D model, =Business activities).

[0083] Calculate the probability of each data type within the total data volume of the grid. .

[0084] Step S202: Calculate the basic Shannon entropy

[0085] Calculate data richness using the Shannon entropy formula:

[0086]

[0087] If the data type within the grid is singular (e.g., only a large number of repetitive "check-in" texts), the entropy value is low; if the data type is rich (text, images, and models coexist), the entropy value is high.

[0088] Step S203: Introduce time decay weight

[0089] Considering the time-sensitive nature of virtual environments, older data contributes less to the current experience (i.e., "data staleness"). This invention introduces a time decay function. :

[0090]

[0091] in This is the attenuation coefficient.

[0092] Step S204: Calculate the final spatiotemporal information entropy

[0093]

[0094] when When the (preset sparsity threshold) is reached, the system determines that the grid is in a "data-poor" or "dead" state and triggers a data completion request.

[0095] To illustrate the impact of time on data density monitoring more intuitively, a time series diagram is shown in Figure 6. Figure 6 shows the dynamic process of a target spatiotemporal grid transitioning from a "hot period" to a "cold start period" over time.

[0096] As shown in Figure 6, in Phase 1 (the peak period), for example, a certain landmark area has recently seen a large amount of user activity. The density monitoring module collects a large number of recent interaction records and calculates the spatiotemporal entropy value. If (e.g., 0.8) is significantly higher than the preset sparsity threshold, the system determines that the state is normal and does not trigger the completion operation.

[0097] Subsequently, assuming a relatively long period of time (e.g., months or years) has passed and no new interactive data has been generated in the area, the system enters Phase Two (cold start). At this point, when the density monitoring module collects data again, it finds the data to be outdated. The time decay coefficient in the formula is then used to calculate the entropy value. The significant increase (leading to a decrease in weights) results in a final calculated spatiotemporal entropy value. (For example, it drops to 0.2) falling below the sparsity threshold.

[0098] At this point, the density monitoring module triggers a data sparsity warning, activating the completion system. The system then calls upon an AI generation agent to generate new discrete spatiotemporal slices and injects them into the grid, thereby increasing the data density of the grid and completing an automated ecological maintenance loop.

[0099] 5. Example 3: Structure and Illumination Extraction of Discrete Spatiotemporal Data Slices

[0100] Referring to Figure 3, this embodiment details the internal structure and generation logic of the "discrete spatiotemporal data slice". This is the essential difference between this invention and ordinary web search results.

[0101] 3.1 Definition of Sliced ​​Data Structure

[0102] The generated slice is a standard JSONB object with the following core fields:

[0103] JSON

[0104] {

[0105] "spatiotemporal_key": "0xAB12...", / / Spatiotemporal index key (Hash(Geo) ^ Hash(Time))

[0106] "anchors": {

[0107] "spatial": {"lat": 39.9, "lng": 116.4, "alt": 50}, / / Spatial anchor point

[0108] "time": "1980-05-15T14:00:00Z" / / Time anchor

[0109] },

[0110] "content_attributes": {

[0111] "semantic_summary": "Afternoon sunlight in a Beijing hutong in the 1980s...", / / Semantic description

[0112] "environment_lighting": { / / Ambient lighting parameters (core feature)

[0113] "intensity": 4500, / / Light intensity (Lux)

[0114] "color_temp": 5600, / / Color temperature (K)

[0115] "main_light_vector": [0.5, -0.8, 0.3] / / Main light source direction vector

[0116] },

[0117] "material_map": { / / Material mapping

[0118] "roughness": 0.8, / / Roughness

[0119] "reflectivity": 0.1 / / Reflectivity

[0120] }

[0121] }

[0122] }

[0123] 3.2 Logic for Extracting Ambient Lighting Parameters

[0124] The system uses computer vision models (such as pre-trained Vision Transformers) to process retrieved historical images:

[0125] Main light source estimation: By analyzing the shadow angle of objects in the image, the azimuth angle of the sun at the time of shooting is inferred, and the three-dimensional vector main_light_vector is calculated.

[0126] Color temperature estimation: Analyze the white balance characteristics of the image to infer whether it is "warm morning light" (3000K) or "cool midday light" (6000K).

[0127] Technical effect: When this slice is loaded into the user's AR / VR terminal, the rendering engine reads these parameters so that the virtual objects in the slice can generate correct lighting and occlusion with the surrounding environment, instead of looking like an abrupt texture.

[0128] 6. Example 4: Observer Consensus Verification and State Transition

[0129] Referring to Figure 5, in order to prevent AI from generating false historical information (AI Hallucination), this invention introduces a post-processing mechanism based on "observer consensus".

[0130] Step S401: Temporary Injection

[0131] When a newly generated slice is injected into the database, the status field is marked as status: "temporary". At this time, it is only visible to a subset of high-level users (or test users).

[0132] Step S402: Interaction Behavior Statistics

[0133] When a real user enters the grid and browses the tile, the system records the user's non-intrusive interaction data:

[0134] Dwell time: The amount of time the user's view stays on this slice.

[0135] Positive feedback: likes, favorites, screenshots and sharing.

[0136] Negative feedback: quickly swipe past, click "not interested" or "report an error".

[0137] Step S403: Consensus Confidence Calculation

[0138] The system calculates the consensus score. :

[0139]

[0140] Step S404: State transition

[0141] like The status is updated to "permanent", making it a formal historical asset of the location.

[0142] like : Triggers the rollback mechanism, softly deletes the slice from the database, and remarks the mesh as pending generation.

[0143] 7. Example 5: Automated Completion System for Virtual Spatiotemporal Environments

[0144] Referring to Figure 4, based on the same inventive concept, this embodiment of the invention also provides an automated virtual spatiotemporal environment completion system. This system can implement all the steps in the above method embodiments. The system includes:

[0145] 7.1 Density Monitoring Module

[0146] This module connects to the log interface of a distributed index database, scanning the target spatiotemporal grid at a preset time frequency. It integrates a spatiotemporal entropy calculation unit to calculate the entropy based on the formula...

[0147]

[0148] Calculate data activity metrics in real time. When the metrics fall below a preset threshold, send a generation command to the task scheduler.

[0149] 7.2 Data Retrieval Agent

[0150] This module acts as a gateway connecting the internal network to the external internet. It is configured with multiple adapters for external APIs, enabling cross-domain searches. Its key functions include:

[0151] Semantic mapping unit: converts spatiotemporal tags (such as "1920 + The Bund") into keyword combinations that search engines can understand.

[0152] Multimodal alignment unit: used to compare the consistency of retrieved text and images in the time dimension and eliminate conflicting data.

[0153] 7.3 Slice Generation Engine

[0154] This module is deployed on high-performance computing nodes (such as GPU clusters) and runs pre-trained generative neural network models (including the RAG architecture and Latent Diffusion Model). Its core sub-modules include:

[0155] Attribute extractor: Used to extract rendering parameters such as ambient lighting and material texture from an image.

[0156] Slice Encapsulator: Used to assemble the extracted parameters and spatiotemporal anchors into XYZT discrete data slices that conform to the system standard.

[0157] 7.4 Database Write Interface

[0158] This interface integrates the spatiotemporal index key calculation logic. Before data is written, it is responsible for extracting the spatial features of each slice. With time characteristics A unique primary key is generated through an XOR hash operation, and data is routed to the corresponding distributed storage node based on this primary key.

[0159] 8. Example 6: Electronic device (hardware entity)

[0160] This invention also provides an electronic device, which may be a server, a server cluster, or a cloud computing node.

[0161] The electronic device includes: a memory, a processor, and a computer program stored in the memory and capable of running on the processor.

[0162] Processor: May include one or more processing cores (such as CPU, GPU, TPU). In this invention, the GPU is used to execute the generative neural network model inference task described in Embodiment 3, and the CPU is used to execute the entropy calculation and logic scheduling task described in Embodiment 2.

[0163] Memory: Used to store non-transitory computer-readable instructions. When the instructions are executed by the processor, they cause the processor to perform the steps of any one of the methods described in Embodiments 1 to 4 above.

[0164] Communication interface: Used to enable data transmission between internal databases and external internet data sources.

[0165] 9. Example 7: Computer-readable storage medium

[0166] This invention also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the automated virtual spatiotemporal environment completion method based on dynamic data density perception described in any of the above embodiments of this specification.

[0167] Conclusion

[0168] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system and electronic device embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0169] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for automatically completing a virtual spatiotemporal environment based on dynamic data density perception, characterized in that, Includes the following steps: State awareness and triggering steps: In the distributed index database of the virtual spatiotemporal environment, data activity indicators are monitored in units of preset spatiotemporal grids; When the data activity index of the target spatiotemporal grid meets the preset data sparsity condition, a data completion request is generated; Heterogeneous data acquisition steps: In response to the data completion request, the spatiotemporal attribute labels of the target spatiotemporal grid are parsed, and multimodal reference data associated with the target region is obtained from an external data source based on the labels; Structured slice generation steps: The multimodal reference data is cleaned and reorganized using a generative neural network model to generate discrete spatiotemporal data slices adapted to the virtual spatiotemporal environment; the slices include spatiotemporal anchor data and scene content attribute data; closed-loop injection steps: The generated discrete spatiotemporal data slices are injected into the distributed index database through a consistency interface to fill the data holes in the target spatiotemporal grid.

2. The method according to claim 1, characterized in that, The data activity index is spatiotemporal information entropy; The state perception and triggering steps specifically include: calculating the spatiotemporal information entropy based on the distribution probability of historical interaction data within the target spatiotemporal grid and combining it with a time decay function; The calculation formula is as follows: in, The spatiotemporal information entropy of the target grid. For the first in the grid The probability distribution of interactive data, The weighting coefficient decreases over time; when When the data is below a preset threshold, it is determined that the data sparsity condition is met.

3. The method according to claim 1, characterized in that, The scene content attribute data includes at least one of the following or a combination thereof: Semantic description features: Text summary data that describes historical events or geographic information of the target area, generated based on the multimodal reference data; Ambient lighting parameters: The main light source vector, light intensity, and color temperature data are derived from the image information and stored in JSONB format; Material mapping data: Surface roughness and reflectivity parameters of objects inferred from the texture features of retrieved images; Meteorological status data: Historical weather characteristics corresponding to the timestamps of the spatiotemporal anchor data.

4. The method according to claim 1, characterized in that, In the structured slice generation step, the generated discrete spatiotemporal data slices adopt the following encoding logic to adapt to the distributed index database: Extracting the spatial coordinate features of the slices respectively With time anchor features ; Using a preset hash operation or spatial curve filling algorithm, the spatial coordinate features and time anchor features are fused to generate a composite spatiotemporal hash value, and this hash value is used as the spatiotemporal index key, which is also the primary key of the slice in the database.

5. The method according to claim 1, characterized in that, Before the closed-loop injection step, an "observer consensus verification" step is also included: marking the newly generated discrete spatiotemporal data slice as "to be verified" and injecting it into the database; collecting the interaction behavior data of user terminals that subsequently enter the target spatiotemporal grid on the slice, the interaction behavior data including dwell time, positive evaluation feedback or negative evaluation feedback; calculating the consensus confidence based on the interaction behavior data, when the consensus confidence exceeds a preset threshold, updating the slice to "permanent historical state"; when the consensus confidence is lower than the preset threshold, performing slice rollback or deletion operations.

6. The method according to claim 1, characterized in that, The generative neural network model includes a latent diffusion model or a retrieval-enhanced generative architecture; the multimodal reference data includes text descriptions, image data, video summaries, or historical data from third-party geographic information systems from publicly available internet indexes; the structured tile generation step also includes privacy cleaning processing: identifying facial or sensitive license plate information in the multimodal reference data and performing pixel-based desensitization.

7. A virtual spatiotemporal environment automated completion system, characterized in that, include: The density monitoring module is used to calculate the data activity index or spatiotemporal information entropy of the target spatiotemporal grid in real time and trigger a completion request. The data acquisition proxy is used to call external data source interfaces to obtain multimodal reference data associated with the target area; The slice generation engine is used to run generative neural network models and output discrete spatiotemporal data slices containing spatiotemporal anchor points and scene content attributes. The database write interface is used to perform index key calculation and closed-loop injection of discrete spatiotemporal data slices.

8. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the method as described in any one of claims 1 to 6.

Images