A Method and System for Fusion of 3D Earth Data Based on Multi-Source Spatiotemporal Imagery

Abstract

This invention discloses a method and system for fusion of 3D Earth data based on multi-source spatiotemporal imagery. The method relates to the field of 3D Earth data fusion technology and includes the following steps: quantification of visualization efficiency, adaptive adjustment of visualization efficiency, quantification of fusion quality, and adaptive adjustment of fusion accuracy. This invention acquires multi-source data in parallel and extracts visualization efficiency parameters, determining whether to perform visualization adjustment; it performs data fusion and evaluates the fusion quality, determining whether to perform fusion accuracy adjustment. While ensuring output availability, it achieves intelligent resource scheduling and accurate fusion results, improving the timeliness and compatibility of 3D Earth data fusion. It solves the problem in existing technologies where delays in the integration and synchronization of time sensor data and static 3D models lead to frequent attribute loss and positional shifts during data migration between different GIS platforms, resulting in low timeliness and compatibility of 3D Earth data fusion.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional earth data fusion technology, and in particular to a method and system for three-dimensional earth data fusion based on multi-source spatiotemporal imagery. Background Technology

[0002] First, multi-source data acquisition and preprocessing are carried out. Three-dimensional Earth data are acquired through satellite remote sensing (such as optical and radar images from Landsat and Sentinel series), aerial photogrammetry, lidar, ground mapping, IoT sensors, and existing vector maps and digital elevation models. Geometric correction and registration are performed on data from different coordinate systems. For example, the GDAL (Geospatial Data Abstraction Library) library is used for reprojection, radiometric calibration and atmospheric correction are performed on the images, noise reduction and classification are performed on the point cloud data, and data format and accuracy standards are unified to lay the foundation for subsequent fusion.

[0003] Secondly, the data fusion and integrated modeling stage is entered. Using a spatial data engine or a 3D geographic information system platform as support, a basic terrain scene is constructed. Multi-resolution DEM (Digital Elevation Model) data is mosaicked and edge-joined to generate a continuous terrain surface. Imagery and terrain are fused, with orthorectified high-resolution remote sensing images precisely overlaid onto the terrain model. Pyramid tile technology is often used to achieve multi-scale LOD (Level of Detail) scheduling. 3D solid models are integrated, including OSGB (Open Scene Graph Binary) models generated by oblique photogrammetry and BIM (Building Information Modeling) models, which are precisely embedded into the terrain scene through coordinate transformation. Point cloud data can be directly used as a high-precision surface or fused with the model for reconstruction. Socioeconomic, environmental monitoring, and other attribute information from the database are linked to the 3D model through spatial location.

[0004] Finally, visualization rendering and application deployment are performed, utilizing WebGL (Web Graphics Library) technology for high-performance 3D rendering, supporting dynamic lighting, shadows, water effects, and providing interactive functions such as multi-dimensional query, spatial analysis, and simulation.

[0005] For example, Chinese invention patent CN119202324B discloses a method, device, and medium for data compression and fusion in digital earth space, including: extracting features, classifying, and encoding bit vectors stored in the cloud to obtain an index directory set; reading each data unit in a multi-source heterogeneous data stream and searching for the longest matching string in the index directory set; outputting the encoding in the index directory set when the longest matching string is found to obtain a compressed encoded stream; reading the encoded stream and searching for corresponding geospatial data, meteorological data, and interaction records in the index directory set; and fusing the geospatial data, meteorological data, and interaction records according to timestamps to obtain multi-source fused data.

[0006] For example, Chinese invention patent CN103425801B discloses a data fusion method and apparatus for digital earth, comprising: fusing a first data LOD layer and a second data LOD layer with the same level of detail (LOD) to obtain a fusion region; extracting the boundary of the second data within the fusion region, expanding it with the boundary, and constructing a second data buffer; obtaining feature points within the second data buffer according to a first preset rule, and performing weighted processing and linear interpolation to obtain smoothed data within the buffer; geometrically joining the high- and low-precision fusion data blocks simplified by boundary points according to a second preset rule to obtain joined data blocks; constructing a triangular mesh within the joined data blocks according to a third preset rule, and outputting the fusion result; and repeating the above steps until all LOD layers have obtained the fusion result.

[0007] However, in the process of implementing the inventive technical solution in the embodiments of this application, it was found that the above-mentioned technology has at least the following technical problems:

[0008] In existing technologies, firstly, the total amount of data in a single city-level 3D scene is increasing, with single-square-kilometer oblique photogrammetry models often reaching GB levels. The data growth rate far exceeds the upgrade speed of traditional single-machine and serial computing architectures. At the same time, in the visualization stage, the existing technical architecture struggles to balance the demands of high-precision rendering and efficient transmission: complex 3D models contain billions of triangular faces and high-resolution textures, which can easily cause GPU memory overflow on the client side; while mainstream rendering engines on the web are limited by the browser's single-threaded mechanism and network bandwidth, frequently experiencing stuttering and latency when loading and scheduling massive amounts of data, making it difficult to guarantee a user experience.

[0009] Secondly, the Earth's surface and subsurface structures are highly complex and uncertain. When sparse borehole data is fused with terrain data, existing interpolation algorithms are prone to errors in estimating stratigraphic boundaries, leading to blurred subsurface models. The inherent speckle noise in radar data and the interference from lighting and weather in optical images directly cause geometric distortion and feature loss. The fundamental reason is that current data completion, denoising, and fusion algorithms lack robustness and have limited ability to filter noise and reconstruct sparse information, making it impossible to accurately restore complex real-world geographical features.

[0010] Finally, the 3D Earth needs to be dynamically updated quarterly according to a certain ratio. However, the rapid integration and synchronous calibration between real-time sensor data streams and static 3D models are challenging. Different GIS (Geographic Information System) platforms (such as ArcGIS and SuperMap) have differences in coordinate systems and data organization methods. Due to the delay in the integration and synchronization of real-time sensor data and static 3D models, the data frequently suffers attribute loss and positional shifts when migrating between different GIS platforms, resulting in low timeliness and compatibility of 3D Earth data fusion. Summary of the Invention

[0011] To address the technical problems of low timeliness and compatibility in 3D Earth data fusion caused by the delay in the synchronization of time sensor data and static 3D models during data migration between different GIS platforms, this invention provides a method and system for 3D Earth data fusion based on multi-source spatiotemporal imagery. The technical solution is as follows:

[0012] On the one hand, a method for fusion of 3D Earth data based on multi-source spatiotemporal imagery is provided. This method includes: Step 1, acquiring 3D Earth data from multi-source spatiotemporal imagery in parallel; obtaining visualization parameters during the conversion from raw 3D Earth data to an interactive 3D scene; and obtaining data processing and visualization efficiency based on these parameters to quantify the effectiveness of 3D Earth data processing and visualization; Step 2, determining whether to perform adaptive adjustment of visualization efficiency based on the data processing and visualization efficiency, avoiding unnecessary adjustment overhead and strategy oscillations, and achieving precise and intelligent resource scheduling. If yes, then the 3D Earth data fusion stage is performed after adaptive adjustment of visualization efficiency; otherwise, it proceeds directly. The three-dimensional earth data fusion process includes: Step 3: Obtaining fusion quality parameters in the three-dimensional earth data fusion process, and obtaining the fusion quality based on the fusion quality parameters. This is used to quantify the success rate of automatic matching after the fusion of multi-source heterogeneous three-dimensional earth data during the fusion process; Step 4: Determining whether to perform adaptive adjustment of fusion accuracy based on the fusion quality, thereby constructing an event-driven precision optimization closed loop guided by the output of fusion quality while ensuring the availability of 3D earth data. If yes, the fusion result is output after adaptive adjustment of fusion accuracy; otherwise, the fusion result is output directly.

[0013] On the other hand, a 3D Earth data fusion system based on multi-source spatiotemporal imagery is provided. This system includes: a visualization efficiency quantification module, a visualization efficiency adaptive control module, a fusion quality quantification module, and a fusion accuracy adaptive control module. Specifically, the visualization efficiency quantification module acquires 3D Earth data from multi-source spatiotemporal imagery in parallel. During the conversion from raw 3D Earth data to an interactive 3D scene, visualization parameters are obtained. Based on these parameters, data processing and visualization efficiency are determined, quantifying the effectiveness of 3D Earth data processing and visualization. The visualization efficiency adaptive control module determines whether to implement adaptive control based on data processing and visualization efficiency, avoiding unnecessary control overhead and strategy oscillations, achieving precise and intelligent resource scheduling. If so, the visualization efficiency is automatically adjusted. After adaptive adjustment, the 3D Earth data fusion stage is performed; otherwise, the 3D Earth data fusion stage is performed directly. The fusion quality quantification module is used to acquire fusion quality parameters in the 3D Earth data fusion stage. Based on these parameters, the 3D Earth data fusion quality is obtained, which is used to quantify the success rate of automated matching after the fusion of multi-source heterogeneous 3D Earth data. The fusion accuracy adaptive adjustment module is used to determine whether to perform fusion accuracy adaptive adjustment based on the 3D Earth data fusion quality. This ensures the usability of the 3D Earth data and constructs an event-driven, precise optimization closed loop guided by the output of 3D Earth data fusion quality. If yes, the 3D Earth data fusion result is output after fusion accuracy adaptive adjustment; otherwise, the 3D Earth data fusion result is output directly.

[0014] Beneficial effects

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0016] 1. By acquiring 3D Earth data from multi-source spatiotemporal imagery in parallel, visualization parameters are obtained during the conversion of raw 3D Earth data into an interactive 3D scene. Based on these parameters, data processing and visualization efficiency are determined to quantify the effectiveness of 3D Earth data processing and visualization. Adaptive adjustment of visualization efficiency is then implemented based on the data processing and visualization efficiency, thus constructing a dynamic decision-making mechanism based on efficiency perception. This mechanism precisely triggers targeted adaptive adjustment strategies only when the system detects that the collaborative efficiency of the data processing and visualization pipeline is lower than a preset threshold or shows an imbalance trend. This achieves a shift from continuous passive response to precise proactive intervention, ensuring the stable operation of the system's core processes while completely avoiding the additional computational overhead and parameter oscillation risks caused by indiscriminate, all-time control. Ultimately, it achieves optimal dynamic allocation of system resources and intelligent maximization of overall energy efficiency; [This is part of a larger discussion on the fusion of 3D Earth data.] The fusion quality parameters, based on which the fusion quality quality of 3D Earth data is obtained, are used to quantify the success rate of automated matching after the fusion of multi-source heterogeneous 3D Earth data during the fusion process. The fusion quality determines whether to perform adaptive adjustment of fusion accuracy, thus establishing an intelligent arbitration mechanism driven by output quality. This mechanism evaluates key quality indicators such as geometric consistency, texture matching degree, and spatiotemporal alignment accuracy after the fusion of multi-source heterogeneous data in real time. Only when the confidence level of the fusion result is detected to be lower than the dynamic threshold or a systematic accuracy deviation exists, will the corresponding parameter optimization, algorithm switching, or iterative correction process be activated. This achieves a paradigm shift from constant process production to defect-aware self-evolution. While ensuring the basic reliability of 3D geospatial products, it dynamically improves the absolute accuracy and engineering usability of the fusion output with minimal intervention costs, thereby improving the timeliness and compatibility of 3D Earth data fusion.

[0017] 2. By dynamically adjusting the rendering engine loading time based on the rendering command submission rate, a negative feedback coordination mechanism is established between CPU-side command generation and GPU-side resource readiness. This mechanism proactively reduces the time budget for asynchronous resource loading when the submission rate exceeds the system's stable processing threshold, prioritizing the assembly and submission of core rendering commands with the freed-up CPU computing cycles and memory bandwidth. This effectively avoids rendering pipeline instruction starvation and frame generation interval fluctuations caused by excessive competition for system resources from background loading tasks. Ultimately, it maintains the determinism of end-to-end processing latency and the stability of frame rate output in complex scenes. LOD tiles are also dynamically adjusted based on the visualization interaction response latency. This mechanism improves efficiency and establishes a feedback control loop based on end-user experience. By monitoring the end-to-end latency from user interaction to visual feedback in real time, it immediately reduces the thread priority, computational granularity, or geometric precision of background LOD tile generation tasks when the latency exceeds the perceived smoothness threshold. The saved CPU / GPU computing resources are then forcibly allocated to the interactive response and real-time rendering pipeline, ensuring the immediacy and visual continuity of user operation responses. This fundamentally avoids interactive stuttering, screen lag, and system response instability caused by the disorderly contention of system resources by large-scale background tile generation tasks. It achieves absolute priority for interactive smoothness under limited computing power and improves the timeliness and compatibility of 3D earth data fusion.

[0018] 3. By dynamically adjusting the dynamic rendering update frame rate threshold based on the cross-source data matching time, a feedforward-feedback collaborative control mechanism is constructed between multi-source data fusion and real-time rendering. This mechanism monitors the processing time of multi-source heterogeneous data in key stages such as spatiotemporal alignment, feature association, and semantic matching in real time. When the matching time increases, indicating a data readiness delay, the frame rate threshold for triggering global dynamic rendering updates is proactively and non-linearly increased. This forces the system to perform high-cost global visual updates based on the latest fusion results only when there is sufficient rendering performance margin (i.e., the actual frame rate is consistently higher than the higher threshold). This effectively prevents the vicious cycle of meaningless repeated refreshes of the downstream rendering pipeline and system load avalanche caused by upstream data processing delays, ensuring the effectiveness of rendering updates and the overall stability of the system under uncertain data flow conditions. By dynamically adjusting the width of the time-series data sliding window to reflect the verification delay, an adaptive data ingestion control mechanism with feedback from the quality verification process is constructed in the data stream processing system. This mechanism monitors the computation time of the confidence verification process of the fusion results in real time. When the verification delay increases, indicating an increase in system verification load or data complexity, it proactively and precisely shrinks the width of the time-series data sliding window or reduces its internal data sampling density. This directly reduces the scale of the original data to be processed, fused, and verified, suppressing the risk of queue backlog and latency accumulation in the data processing pipeline from the source. It ensures that the system can prioritize the fusion and verification of high-confidence core data under limited real-time constraints, achieving an optimal balance between processing throughput and result reliability under dynamic load, and improving the timeliness and compatibility of 3D Earth data fusion. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of a three-dimensional Earth data fusion method based on multi-source spatiotemporal imagery provided in this application embodiment;

[0021] Figure 2 A flowchart illustrating the intelligent adjustment process for tile generation efficiency in a 3D Earth data fusion method based on multi-source spatiotemporal imagery, provided in an embodiment of this application.

[0022] Figure 3 A flowchart illustrating the intelligent adjustment of the update frame rate threshold for a three-dimensional Earth data fusion method based on multi-source spatiotemporal imagery provided in this application embodiment;

[0023] Figure 4This is a schematic diagram of the structure of a three-dimensional earth data fusion system based on multi-source spatiotemporal images provided in an embodiment of this application. Detailed Implementation

[0024] The technical solution provided in this application will now be described with reference to the accompanying drawings.

[0025] To facilitate understanding of the embodiments of this application, the following points will be explained first:

[0026] First, in this application, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A alone, A and B simultaneously, or B alone, where A and B can be singular or plural. The character " / " generally indicates an "or" relationship between the preceding and following related objects, but it does not exclude the possibility of indicating an "and" relationship; the specific meaning can be understood in context. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c; a and b; a and c; b and c; or a and b and c. Here, a, b, and c can be single or multiple.

[0027] Second, the use of prefixes such as "first" and "second" in this application is solely for the purpose of distinguishing and describing different things belonging to the same category, and does not constrain the order, size, or quantity of things. For example, "first message" and "second message" are simply different messages, and there is no chronological, size, or priority relationship between them.

[0028] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0029] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0030] like Figure 1 The diagram shown is a flowchart of a three-dimensional Earth data fusion method based on multi-source spatiotemporal imagery provided in an embodiment of this application. The method includes the following steps:

[0031] Step 1: Acquire 3D Earth data from multi-source spatiotemporal imagery in parallel. During the conversion process from raw 3D Earth data to an interactive 3D scene, obtain visualization parameters. Based on the visualization parameters, obtain the data processing and visualization efficiency to quantify the effectiveness of 3D Earth data processing and visualization.

[0032] It should be noted that the process of determining whether to implement adaptive efficiency control based on visualization previously included:

[0033] Visualization parameters include parallel computing speedup, first-screen loading time, and LOD tile scheduling hit rate. Among them, the parallel computing speedup ratio refers to the ratio of the execution time of the same 3D earth data processing task under a single-machine serial computing architecture to the execution time under a parallel computing architecture. The total time from task start to completion is recorded by performance monitoring tools to obtain the serial time. In the parallel computing cluster, the total time from task distribution to completion of all nodes is recorded by cluster scheduling and management tools. The ratio of serial time to the total execution time is recorded as the parallel computing speedup ratio. The first screen loading time refers to the time interval from when the user triggers the 3D scene loading request (such as clicking the 3D map entry on the web, entering the target area and submitting) to when the first complete and interactive 3D scene is rendered on the client screen. The first screen loading time is recorded as the difference between the rendering completion timestamp of the first complete 3D scene captured by the front-end rendering monitoring tool and the instantaneous timestamp of the user clicking the loading button recorded by the front-end tracking technology. The LOD tile scheduling hit rate refers to the proportion of data tiles that the client can directly obtain from the local cache when the view changes. By setting a statistical period (such as a single view roaming session), the LOD tile scheduling hit rate is automatically calculated by the code after the period ends and the result is output.

[0034] The speedup ratio is adjusted by a speedup adjustment factor to compensate for the speedup ratio against the speedup reference value, resulting in a corrected speedup ratio value. The constraint expression for the corrected speedup ratio value is as follows:

[0035] ;

[0036] In the formula, D represents the speedup correction value; v1 represents the speedup adjustment factor obtained from the 3D Earth data fusion database; A0 represents the speedup reference value obtained from the 3D Earth data fusion database; and A represents the parallel computing speedup.

[0037] The loading time correction value is obtained by compensating the loading time reference value and the first screen loading time using a loading time adjustment factor. The constraint expression for the loading time correction value is as follows:

[0038] ;

[0039] In the formula, E represents the loading time correction value; v2 represents the loading time adjustment factor obtained from the 3D Earth data fusion database; H0 represents the loading time reference value obtained from the 3D Earth data fusion database; and H represents the first screen loading time.

[0040] The hit rate correction value is obtained by compensating the LOD tile scheduling hit rate against the hit rate reference value using a hit rate adjustment factor. The constraint expression for the hit rate correction value is as follows:

[0041] ;

[0042] In the formula, G represents the hit rate correction value; v3 represents the hit rate adjustment factor obtained from the 3D Earth data fusion database; B0 represents the hit rate reference value obtained from the 3D Earth data fusion database; and B represents the LOD tile scheduling hit rate.

[0043] By coupling the speedup correction value, loading time correction value, and hit rate correction value, the data processing and visualization efficiency is obtained. The constraint expression for data processing and visualization efficiency is as follows:

[0044] ;

[0045] In the formula, P represents the efficiency of data processing and visualization.

[0046] It should be explained that a higher parallel computing speedup means that the backend server can complete heavy tasks such as processing massive amounts of raw data (such as oblique photography), building LOD tile pyramids, and data compression in a shorter time, resulting in a shorter first-screen loading time; a higher LOD tile scheduling hit rate means fewer network requests are required, resulting in a shorter first-screen loading time; and a higher parallel computing speedup enables the backend to execute more complex and refined preprocessing tasks at a lower cost and faster speed, resulting in a higher LOD tile scheduling hit rate. Meanwhile, there is a positive correlation between the parallel computing speedup ratio and data processing and visualization efficiency. The higher the parallel computing speedup ratio, the greater the speed improvement when using distributed computing resources (such as CPU / GPU clusters) compared to traditional single-machine serial processing, in completing tasks such as LOD tile generation, data compression, and format conversion, resulting in higher data processing and visualization efficiency. On the other hand, there is a negative correlation between the first-screen loading time and data processing and visualization efficiency. The longer the first-screen loading time, the more likely it is to cause front-end loading and decoding lag, prolonging the loading time and resulting in lower data processing and visualization efficiency. On the other hand, there is a positive correlation between the LOD tile scheduling hit rate and data processing and visualization efficiency. The higher the LOD tile scheduling hit rate, the more likely it is that when users are panning, zooming, or rotating scenes, most of the required data comes from the local cache, without waiting for the network, resulting in higher data processing and visualization efficiency.

[0047] Step two: Based on the data processing and visualization efficiency, determine whether to perform adaptive adjustment of visualization efficiency to avoid unnecessary adjustment costs and strategy oscillations, and achieve precise and intelligent resource scheduling. If yes, proceed to the 3D earth data fusion stage after the adaptive adjustment of visualization efficiency; otherwise, proceed directly to the 3D earth data fusion stage.

[0048] Furthermore, the specific steps for determining whether to implement visual efficiency adaptive control are as follows:

[0049] If the data processing and visualization efficiency is higher than or equal to the visualization efficiency reference value, then the visualization efficiency adaptive adjustment will not be performed. Otherwise, the engine loading time and tile generation efficiency will be intelligently adjusted based on the visualization efficiency offset. The visualization efficiency offset represents the negative difference between the data processing and visualization efficiency and the visualization efficiency reference value.

[0050] It should be further explained that the specific steps for intelligently adjusting the loading time of the execution engine are as follows:

[0051] Monitoring the submission rate of the rendering command queue and intelligently adjusting the engine loading time provides real-time and accurate pressure feedback for regulation, ensuring that regulation decisions are highly matched with the actual system operating status. This is achieved based on a preset submission rate-loading budget ratio mapping relationship, which is configured as an inversely correlated adaptive adjustment relationship, specifically as follows:

[0052] When the rendering instruction submission rate is lower than the first threshold, the first level of mapping is performed. This logic dynamically increases the loading budget by combining GPU rendering efficiency feedback (reflected by offset) when the CPU instruction submission pressure is low. This ensures the integrity of data loading and avoids GPU resource idleness, thus achieving collaborative optimization of CPU and GPU resources.

[0053] When the rendering command submission rate exceeds the second threshold, a second-level mapping is performed. When the CPU command submission pressure is overloaded, the loading budget is predictively tightened to prioritize the deterministic latency of the rendering command execution pipeline, avoiding scene stuttering and interaction delays caused by excessive loading time, and significantly improving the smoothness of visualization in high-pressure scenarios.

[0054] When the rendering command submission rate is within the submission rate threshold range, intelligent adjustment of engine loading time is not performed, maintaining the current loading time as a percentage of the frame time budget. The submission rate threshold range represents the closed interval formed by the first and second thresholds. This avoids frequent adjustments under moderate pressure, reduces system adjustment overhead, and maintains the stability of the visualization effect.

[0055] Perform the first-level mapping as follows:

[0056] The submission rate and visualization efficiency offset are input into the submission rate-load budget ratio mapping relationship, which is mapped to a gain target load time ratio. The gain target load time ratio is combined with the current target total time budget for each frame, and the target gain absolute load time for the current frame is obtained through linear mapping. Thus, based on the baseline adjustment according to the CPU instruction submission pressure, the load budget ratio is further dynamically compensated and corrected by combining the real-time feedback of GPU rendering execution efficiency.

[0057] Perform the second-level mapping, specifically:

[0058] The submission rate and the visual efficiency offset are input into the submission rate-load budget ratio mapping relationship, which is mapped to a reduction target load time ratio. The reduction target load time ratio is combined with the current target total time budget per frame, and the target reduction absolute load time of the current frame is obtained through linear mapping. Thus, when the CPU instruction submission pressure reaches the threshold, the load budget is further predictively tightened and controlled in combination with the actual rendering load of the GPU, so as to prioritize the deterministic latency of the rendering instruction execution pipeline.

[0059] In this embodiment, collaborative optimization of the entire data processing-visualization chain is achieved through dual-path parallel perception (efficiency offset, interaction latency) and dual-track adaptive intervention (loading time, generation efficiency). Its core effect is that a precise, multi-objective collaborative resource reallocation strategy is only initiated when the overall system efficiency falls short of standards. This avoids unnecessary regulatory overhead and, through joint CPU-GPU load perception and forward-looking budget adjustment, intelligently balances the competitive relationship between resource loading, geometric calculation, and real-time rendering under dynamic load, ultimately ensuring stable frame rate, smooth interaction, and visual continuity in complex 3D scenes.

[0060] It needs to be explained that, such as Figure 2 The diagram shows a flowchart of the intelligent adjustment process for tile generation efficiency in a 3D Earth data fusion method based on multi-source spatiotemporal imagery provided in this application embodiment. The specific process is as follows: First, through inter-component collaborative monitoring, the system continuously monitors the user interaction response latency in real time and determines whether a response latency issue exists based on a preset threshold. When the interaction response is normal, the system maintains the execution state of the current background task (such as tile generation) and enters the next monitoring cycle. When the interaction response latency exceeds the threshold, the system triggers an exception handling process. At this time, the system combines the interaction response latency curve data and the visualized efficiency offset, inputting them together into a preset latency-generation efficiency mapping relationship to calculate the generation efficiency attenuation coefficient, and then obtains the target generation efficiency through corresponding processing.

[0061] It should be further explained that the specific steps for intelligent adjustment of tile generation efficiency are as follows:

[0062] The system monitors the interaction response latency of user interaction operations and continuously collects the interaction response latency of user interaction operations (such as dragging, scaling, and rotating the 3D scene). This indicator directly reflects the user's perception of the smoothness of scene operation and provides front-end user experience feedback for tile generation efficiency control. It makes up for the limitations of control based solely on back-end indicators and dynamically adjusts the generation efficiency of LOD tile generation thread based on the interaction response latency.

[0063] Step S21 defines the inverse control relationship between interaction response latency and generation efficiency. Specifically, it establishes an inverse mapping relationship between interaction response latency and LOD tile generation efficiency. This inverse mapping relationship is configured as follows: when the interaction response latency is lower than or equal to a latency threshold, intelligent adjustment of tile generation efficiency is not performed, maintaining the current LOD tile generation efficiency and avoiding over-regulation that could affect the integrity of tile loading, thus balancing interaction smoothness and data loading efficiency; when the interaction response latency is higher than the latency threshold, the LOD tile generation efficiency is reduced. Under CPU instruction submission pressure overload, the loading budget is preemptively compressed, prioritizing the deterministic latency of the rendering instruction execution pipeline, avoiding scene stuttering and interaction delays caused by excessive loading time, and significantly improving visualization smoothness under high-pressure scenarios.

[0064] Step S22, reduce LOD tile generation efficiency, specifically: input the interaction response latency offset and the visualization efficiency offset into the latency-generation efficiency inverse mapping relationship to obtain the generation efficiency attenuation coefficient, multiply the generation efficiency attenuation coefficient by the current LOD tile generation efficiency to obtain the target LOD tile generation efficiency for the next round, thereby merging the front-end interaction latency index that characterizes the user's perceived smoothness with the back-end rendering efficiency index that reflects the GPU execution load, prioritizing the deterministic latency of the user interaction response link and the stability of the rendering pipeline, the interaction response latency offset represents the positive difference between the interaction response latency and the latency threshold.

[0065] In this embodiment, by simultaneously sensing both user-side interaction smoothness and hardware-side rendering load, precise and coordinated suppression of the intensity of background tile generation tasks is achieved. When the system faces the dual pressures of increased interaction latency and increased GPU load (i.e., both offsets are positive), this mechanism calculates a larger attenuation coefficient, imposing more severe suppression on the tile generation task, thereby strongly releasing CPU and GPU resources. This ensures that during the most critical period of system resource competition, the user interaction response link and real-time rendering pipeline receive absolute priority, effectively preventing system-level performance degradation and interaction experience collapse caused by unrestrained competition for resources by background computing tasks. Ultimately, it achieves the intelligent scheduling goal of prioritizing interaction smoothness over visual completeness under limited computing power.

[0066] Step 3: Obtain the fusion quality parameters in the 3D Earth data fusion process. Based on the fusion quality parameters, obtain the 3D Earth data fusion quality, which is used to quantify the success rate of automatic matching after the fusion of multi-source heterogeneous 3D Earth data during the 3D Earth data fusion process.

[0067] It should be noted that the specific steps to obtain the quality of 3D Earth data fusion are as follows:

[0068] The fusion quality parameters include the visualization efficiency to be compared, the 3D earth data fusion latency, and the automated synchronization trigger response time. Specifically, the visualization efficiency to be compared refers to the efficiency of the newly acquired data processing and visualization if adaptive adjustment of visualization efficiency has been implemented; otherwise, the current data processing and visualization efficiency is used. The 3D earth data fusion latency refers to the total time interval between fusing the collected dynamic geographic data into the static 3D earth model and completing consistency verification. This latency is obtained through the sensor's built-in time synchronization module. The automated synchronization trigger response time refers to the time interval from receiving geographic feature change data pushed by the sensor (such as terrain subsidence, newly added building outlines, etc.) to automatically starting the entire process of data preprocessing, matching, and model updating. This time interval is obtained by batch parsing the synchronization task's startup logs using log analysis tools (such as ELK and Splunk), automatically extracting timestamps, and calculating response times.

[0069] The visualization efficiency adjustment factor is used to compensate for the visualization efficiency of the comparison object and the visualization efficiency reference value, resulting in a revised visualization efficiency value. The constraint expression for the revised visualization efficiency value is as follows:

[0070] ;

[0071] In the formula, Q represents the visualization efficiency correction value; s1 represents the visualization efficiency adjustment factor obtained from the 3D Earth data fusion database; W0 represents the visualization efficiency reference value obtained from the 3D Earth data fusion database; and W represents the visualization efficiency to be compared.

[0072] The fusion delay correction value is obtained by compensating for the fusion delay reference value and the fusion delay of 3D Earth data through a fusion delay adjustment factor. The constraint expression for the fusion delay correction value is as follows:

[0073] ;

[0074] In the formula, Y represents the fusion delay correction value; s2 represents the fusion delay adjustment factor obtained from the 3D Earth data fusion database; U0 represents the fusion delay reference value obtained from the 3D Earth data fusion database; and U represents the 3D Earth data fusion delay.

[0075] The response time correction value is obtained by compensating the automated synchronization trigger response time against the response time reference value using a response time adjustment factor. The constraint expression for the response time correction value is as follows:

[0076] ;

[0077] In the formula, L represents the response time correction value; s3 represents the response time adjustment factor obtained from the 3D Earth data fusion database; K0 represents the response time reference value obtained from the 3D Earth data fusion database; and K represents the automatic synchronization trigger response time.

[0078] The visualization efficiency correction value, fusion latency correction value, and response time correction value are coupled to obtain the 3D Earth data fusion quality. The constraint expression for the 3D Earth data fusion quality is as follows:

[0079] ;

[0080] In the formula, X represents the quality of 3D Earth data fusion.

[0081] It's important to understand that the longer the latency of 3D Earth data fusion, the less timely the static 3D model can synchronize with the latest geographic feature data (such as newly added buildings and terrain deformations). The visualization stage can only load the old version of the model data, resulting in lower visualization efficiency for comparison. The longer the automated synchronization trigger response time, the more redundant the waiting time is in the system's perception and initiation stage of sensor data changes. Even if the subsequent preprocessing and matching stages are highly efficient, the overall fusion latency will be lengthened due to the time spent in the preceding stages. The longer the 3D Earth data fusion latency, the more likely the visualization stage will be unable to obtain the updated complete model data in a timely manner, causing problems such as loading lag and interaction delays, resulting in lower visualization efficiency for comparison. Meanwhile, there is a positive correlation between the comparison visualization efficiency and the quality of 3D Earth data fusion. The higher the comparison visualization efficiency, the more accurate the geometric alignment, semantic labeling, and attribute information mapping of multi-source heterogeneous data (LiDAR point cloud, oblique photogrammetry, borehole data, and sensor dynamic data) have been, resulting in higher 3D Earth data fusion quality. There is a negative correlation between 3D Earth data fusion latency and 3D Earth data fusion quality. The longer the 3D Earth data fusion latency, the more defects such as geometric distortion, attribute loss, and positional offset occur in the fusion model, resulting in lower 3D Earth data fusion quality. There is also a negative correlation between the automatic synchronization trigger response time and 3D Earth data fusion quality. The longer the automatic synchronization trigger response time, the more invalid triggers consume a lot of computing resources, reducing the processing accuracy of effective tasks, ultimately leading to poor dynamic timeliness and low semantic matching accuracy of the fusion model, resulting in lower 3D Earth data fusion quality.

[0082] Step four: Determine whether to perform adaptive adjustment of fusion accuracy based on the quality of 3D Earth data fusion. In this way, while ensuring the usability of 3D Earth data, construct an event-driven precision optimization closed loop that is guided by the output of 3D Earth data fusion quality. If yes, output the 3D Earth data fusion result after adaptive adjustment of fusion accuracy; otherwise, output the 3D Earth data fusion result directly.

[0083] Furthermore, the specific steps for determining whether to perform adaptive adjustment of fusion accuracy are as follows:

[0084] If the fusion quality of the 3D Earth data is higher than or equal to the fusion quality reference value, then adaptive adjustment of fusion accuracy will not be performed. Otherwise, intelligent adjustment of the update frame rate threshold and intelligent adjustment of the sliding window width will be performed based on the fusion quality offset. The fusion quality offset represents the negative difference between the fusion quality of the 3D Earth data and the fusion quality reference value.

[0085] It needs to be explained that, such as Figure 3 The diagram shows the intelligent adjustment flowchart of the update frame rate threshold for the 3D Earth data fusion method based on multi-source spatiotemporal imagery provided in this application embodiment. The specific process is as follows: Starting from the start node, the system continuously monitors the cross-source data matching time and compares it with a preset matching time reference value to determine the real-time status of data preparation. If the matching time does not exceed the reference value, the system maintains the current dynamic rendering update frame rate threshold, preserving the original rendering update rhythm and ensuring the timeliness of visual updates when data is ready. If the matching time exceeds the reference value, the system triggers the intelligent frame rate threshold adjustment process. This process first calculates the matching time offset (reflecting the degree of data latency) and simultaneously obtains the fusion quality offset (evaluating the quality of the data fusion result). These two key indicators are jointly input into the time-frame rate threshold inverse elastic mapping relationship to calculate the update frame rate threshold attenuation coefficient. Subsequently, the system interacts with this attenuation coefficient and the current frame rate threshold to obtain a new, higher rendering update frame rate threshold. This adjustment directly leads to an increase in the rendering update trigger threshold. The update strategy is as follows: global rendering updates are only triggered when the actual rendering frame rate is consistently higher than the new threshold; at the same time, invalid rendering based on low-quality or unready data is actively suppressed to avoid unnecessary computational load.

[0086] It should be further explained that the specific steps for performing intelligent adjustment of the frame rate threshold update are as follows:

[0087] Monitor the cross-source data matching time within the current data fusion cycle, dynamically adjust the frame rate threshold that triggers global dynamic rendering updates based on the cross-source data matching time, define the time-frame rate threshold inverse elastic mapping relationship between the cross-source data matching time and the dynamic rendering update trigger frame rate threshold, and configure the time-frame rate threshold inverse elastic mapping relationship to have decreasing monotonicity, that is, when the cross-source data matching time increases, the dynamic rendering update frame rate threshold decreases accordingly;

[0088] If the cross-source data matching time is less than or equal to the matching time reference value, the intelligent adjustment of the frame rate threshold will not be performed, and the current dynamic rendering frame rate threshold will be maintained. This avoids frequent adjustments under moderate pressure, reduces system adjustment overhead, and maintains the stability of the visualization effect.

[0089] If the cross-source data matching time exceeds the reference value for matching time, the matching time offset and the fusion quality offset are input into the time-frame rate threshold inverse elastic mapping relationship to obtain the updated frame rate threshold decay coefficient. This updated frame rate threshold decay coefficient is then interacted with the dynamic rendering updated frame rate threshold to obtain the next round of dynamic rendering updated frame rate threshold. This prioritizes effective rendering updates based on high-quality fused data when delays occur in the data matching process, while suppressing invalid rendering computational load caused by low-quality or unready data. The matching time offset represents the positive difference between the cross-source data matching time and the reference value for matching time. When CPU instruction submission pressure is overloaded, the loading budget is preemptively compressed to prioritize the deterministic latency of the rendering instruction execution pipeline, avoiding scene stuttering and interaction delays caused by excessive loading time, significantly improving the smoothness of visualization in high-pressure scenarios.

[0090] In this embodiment, a joint intelligent gating system based on efficiency and quality is implemented. It not only responds to the speed of data processing but is also sensitive to data quality: when matching time increases but fusion quality remains high, the decay is relatively mild, allowing the system to moderately raise the update threshold while waiting for high-quality data; when matching time increases and fusion quality also decreases, the decay is more significant, strongly suppressing invalid rendering updates based on low-quality data. This ensures that rendering computing resources are prioritized for visualizing truly ready and reliable high-quality fusion results. In environments with unstable data flows, it intelligently balances the timeliness of visual updates, the stability of system load, and the value density of the final rendered content, thereby significantly improving the overall effectiveness and efficiency of complex multi-source data fusion rendering systems.

[0091] It should be further explained that the specific steps for intelligently adjusting the width of the sliding window are as follows:

[0092] The system monitors the verification delay generated during the fusion result verification process, dynamically adjusts the width of the time-series data sliding window used for data fusion based on the verification delay, and defines a delay-window width mapping relationship between the verification delay of the fusion result and the target width of the time-series data sliding window. The delay-window width mapping relationship is configured such that when the verification delay increases, the target width of the time-series data sliding window decreases. The delay-window width mapping relationship is further configured such that within the time delay critical interval, the reduction rate of the time-series data sliding window width as the verification delay increases is higher than the reduction rate outside the time delay critical interval.

[0093] Intelligent adjustment of sliding window width also includes:

[0094] S100: If the verification latency of the fusion result is lower than the critical lower limit of the verification latency, the current verification latency of the fusion result and the fusion quality offset are input into the latency-window width mapping relationship to obtain the window width adjustment coefficient. The window width adjustment coefficient is then interacted with the current time-series data sliding window width to obtain the target width of the time-series data sliding window to be used in the next fusion cycle. This target width is then allocated to fusion tasks that require higher data quality and deeper historical context. By actively expanding the sliding window width, the integrity of data fusion is improved. Under low CPU instruction submission pressure, the loading budget is dynamically increased in conjunction with GPU rendering efficiency feedback (reflected by the offset), which not only ensures the integrity of data loading but also avoids GPU resource idleness, achieving collaborative optimization of CPU and GPU resources.

[0095] S101, if the fusion result verification latency is within the critical interval, the current fusion result verification latency and fusion quality offset are input into the latency-window width mapping relationship to obtain the window width reduction coefficient. This coefficient is then interacted with the current time-series data sliding window width to obtain the target width of the time-series data sliding window for the next fusion cycle. This mitigates potential overload risks and ensures the stability and timeliness of the data processing pipeline. When CPU instruction submission pressure is overloaded, the loading budget is preemptively compressed, prioritizing the deterministic latency of the rendering instruction execution pipeline. This avoids scene stuttering and interaction delays caused by excessive loading time, significantly improving visualization smoothness under high-pressure scenarios.

[0096] S102, if the verification delay of the fusion result is higher than the critical upper limit of the verification delay, an early warning message is sent to the preset maintenance personnel.

[0097] In this embodiment, by sensing the efficiency of the verification process in real time and translating it into precise and hierarchical control of the data input scale, intelligent load management is achieved across the entire data processing chain. The system proactively pursues better output quality when resources are plentiful, sensitively defends against increased load to maintain stability, and issues safety alarms when approaching overload. Thus, in a dynamic and complex data environment, it consistently ensures that the fusion processing operates within controllable load boundaries, achieving optimal synergy between processing efficiency, result quality, and system robustness.

[0098] like Figure 4 The diagram shown is a structural schematic of a 3D Earth data fusion system based on multi-source spatiotemporal imagery provided in this application embodiment. The 3D Earth data fusion system based on multi-source spatiotemporal imagery provided in this application embodiment includes: a processing visualization efficiency quantification module, a visualization efficiency adaptive control module, a fusion quality quantification module, and a fusion accuracy adaptive control module. Specifically, the processing visualization efficiency quantification module is used to acquire 3D Earth data from multi-source spatiotemporal imagery in parallel. During the conversion process from raw 3D Earth data to an interactive 3D scene, visualization parameters are obtained. Based on these visualization parameters, data processing and visualization efficiency are obtained, which is used to quantify the effectiveness of 3D Earth data processing and visualization. The visualization efficiency adaptive control module is used to determine whether to execute adaptive control based on data processing and visualization efficiency, avoiding unnecessary control overhead and strategy oscillations, and achieving precise resource scheduling. The system includes several modules: Standardization and Intelligence. If both are present, the 3D Earth data fusion process proceeds after adaptive adjustment of visualization efficiency; otherwise, it proceeds directly. A fusion quality quantification module is used to acquire fusion quality parameters during the 3D Earth data fusion process. Based on these parameters, the 3D Earth data fusion quality is determined, quantifying the success rate of automated matching after fusion of multi-source heterogeneous 3D Earth data. A fusion accuracy adaptive adjustment module determines whether to perform fusion accuracy adaptive adjustment based on the 3D Earth data fusion quality. This ensures the usability of the 3D Earth data and constructs an event-driven, precise optimization closed loop guided by the output of 3D Earth data fusion quality. If the fusion accuracy adaptive adjustment is performed, the 3D Earth data fusion result is output; otherwise, the 3D Earth data fusion result is output directly.

[0099] In this embodiment, the visualization efficiency quantification module enables parallel acquisition and visualization parameter quantification of multi-source spatiotemporal imagery 3D Earth data, accurately assessing the effectiveness of data processing and visualization, and providing a scientific basis for subsequent regulation. The visualization efficiency adaptive regulation module intelligently decides whether to implement regulation based on the quantified efficiency results, avoiding unnecessary regulation overhead and strategy oscillations from the source. Furthermore, through multi-dimensional collaborative strategies such as adjusting engine loading time, optimizing LOD tile generation efficiency, and adapting dynamic rendering update frame rate thresholds, it achieves precise and intelligent resource scheduling, ensuring the smoothness and stability of the visualization process. Finally, the fusion quality quantification module comprehensively collects key parameters such as matching class, completeness, and consistency during the fusion process. This system accurately quantifies the success rate of automated matching after the fusion of multi-source heterogeneous data, providing data support for optimizing fusion quality. Relying on the adaptive control module for fusion accuracy, it constructs an event-driven precision optimization closed loop guided by fusion quality. Based on the quality assessment results, it targets and executes control measures such as feature point extraction algorithm optimization, registration parameter adjustment, and fusion weight model improvement. At the same time, it dynamically adjusts the sliding window width of time-series data in conjunction with the fusion result verification delay. Under the premise of ensuring the availability of 3D earth data, it achieves a dynamic balance between data fusion quality, processing efficiency, and system stability, effectively improving the accuracy, efficiency, and reliability of multi-source spatiotemporal image 3D earth data fusion, and adapting to the diversified needs for data quality and interactive experience in different scenarios.

[0100] The various features and processes described above can be used independently of each other or can be combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure. Furthermore, certain method or process blocks may be omitted in some embodiments. The methods and processes described herein are not limited to any particular order, and the blocks or states associated with them may be performed in other suitable orders. For example, the described blocks or states may be performed in an order different from the order specifically disclosed, or multiple blocks or states may be combined in a single block or state. Example blocks or states may be performed serially, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed example embodiments. The exemplary systems and components described herein may be configured differently from those described. For example, elements may be added to, removed from, or rearranged compared to the disclosed example embodiments.

[0101] The various operations of the example methods described herein can be performed at least in part by an algorithm. This algorithm can be contained in program code or instructions stored in memory (e.g., the aforementioned non-transitory computer-readable storage medium). Such an algorithm may include a machine learning algorithm. In some embodiments, the machine learning algorithm may not be explicitly programmed into the computer to perform the function, but can learn from training data to create a predictive model that performs the function.

[0102] The various operations of the example methods described herein can be performed, at least in part, by one or more processors that are temporarily configured (e.g., by software) or permanently configured to perform the relevant operations. Whether temporarily or permanently configured, such processors can constitute the engine of a processor implementation that operates to perform one or more of the operations or functions described herein.

[0103] Similarly, the methods described herein can be implemented at least in part by a processor, where one or more specific processors are examples of hardware. For example, at least some operations of a method can be performed by one or more processors or an engine implemented by a processor. Furthermore, one or more processors can also be operated to support the performance of related operations in a “cloud computing” environment or as “Software as a Service” (SaaS). For example, at least some operations can be performed by a set of computers (as an example of a machine including processors), where these operations are accessible via a network (e.g., the Internet) and via one or more suitable interfaces (e.g., application programming interfaces (APIs)).

[0104] The performance of certain operations can be distributed across processors, residing not only within a single machine but also deployed across multiple machines. In some example embodiments, the processor or processor-implemented engine may reside in a single geographic location (e.g., within a home environment, office environment, or server cluster). In other example embodiments, the processor or processor-implemented engine may be distributed across multiple geographic locations.

[0105] In this specification, multiple instances may implement components, operations, or structures described as single instances. Although individual operations of one or more methods are shown and described as separate operations, one or more of the separate operations may be performed simultaneously and do not need to be performed in the order shown. Structures and functions presented as separate components in the example configuration may be implemented as composite structures or components. Similarly, structures and functions presented as single components may be implemented as separate components. These and other variations, modifications, additions, and improvements fall within the scope of this document.

[0106] While an overview of the subject matter has been described with reference to specific example embodiments, various modifications and changes can be made to these embodiments without departing from the broader scope of embodiments of this disclosure. Such embodiments of the subject matter are referred to herein, individually or collectively, by the term "invention," and are used for convenience only and are not intended to limit the scope of this application to any single disclosure or concept, should more than one disclosure or concept be disclosed in fact.

[0107] The embodiments described herein have been described in sufficient detail to enable those skilled in the art to practice the disclosed teachings. Other embodiments may be used and derived therefrom, such that structural and logical substitutions and changes may be made without departing from the scope of this disclosure. Therefore, the detailed description should not be construed as limiting, and the scope of the various embodiments is defined only by the appended claims and the full scope of their equivalents.

Claims

1. A three-dimensional earth data fusion method based on multi-source spatio-temporal images, characterized in that, The method comprises the following steps: Step one, obtaining three-dimensional earth data from multi-source spatio-temporal images in parallel, obtaining visualization parameters in the conversion process from original three-dimensional earth data to an interactive three-dimensional scene, obtaining data processing and visualization efficiency based on the visualization parameters, and quantifying the effectiveness of three-dimensional earth data processing and visualization; Step two, judging whether to perform visualization efficiency adaptive regulation based on the data processing and visualization efficiency, avoiding unnecessary regulation overhead and strategy oscillation, realizing the precision and intelligence of resource scheduling, if yes, performing three-dimensional earth data fusion after visualization efficiency adaptive regulation, if no, directly performing three-dimensional earth data fusion; Step three, obtaining fusion quality parameters in the three-dimensional earth data fusion process, obtaining three-dimensional earth data fusion quality based on the fusion quality parameters, and quantifying the success rate of automatic matching of multi-source heterogeneous three-dimensional earth data after data fusion in the three-dimensional earth data fusion process; Step four, judging whether to perform fusion accuracy adaptive regulation based on the three-dimensional earth data fusion quality, thereby constructing a precise optimization closed loop oriented to output three-dimensional earth data fusion quality and event-driven under the premise of ensuring the availability of three-dimensional earth data, if yes, outputting three-dimensional earth data fusion results after fusion accuracy adaptive regulation, if no, directly outputting three-dimensional earth data fusion results; The specific steps of judging whether to perform fusion accuracy adaptive regulation are as follows: If the three-dimensional earth data fusion quality is higher than or equal to the fusion quality reference value, fusion accuracy adaptive regulation is not performed, if not, intelligent adjustment of the update frame rate threshold and intelligent adjustment of the sliding window width are performed based on the fusion quality offset, the fusion quality offset representing the negative deviation degree of the three-dimensional earth data fusion quality and the fusion quality reference value; Monitoring the cross-source data matching time consumption in the current data fusion period, dynamically adjusting the frame rate threshold triggering global dynamic rendering update according to the cross-source data matching time consumption, defining a time consumption-frame rate threshold inverse elasticity mapping relationship between the cross-source data matching time consumption and the dynamic rendering update triggering frame rate threshold, the time consumption-frame rate threshold inverse elasticity mapping relationship being configured to have decreasing monotonicity, that is, when the cross-source data matching time consumption increases, the dynamic rendering update frame rate threshold correspondingly decreases; The specific steps of the intelligent adjustment of the sliding window width are as follows: Monitoring the verification time delay generated in the fusion result verification process, dynamically adjusting the width of the time series data sliding window used for data fusion according to the verification time delay, defining a time delay-window width mapping relationship between the verification time delay of the fusion result and the target width of the time series data sliding window, the time delay-window width mapping relationship being configured such that when the verification time delay increases, the target width of the time series data sliding window decreases, and the time delay-window width mapping relationship is further configured such that in the time delay critical interval, the reduction rate of the time series data sliding window width with the increase of the verification time delay is higher than that in the non-time delay critical interval.

2. The method of claim 1, wherein the method further comprises: The judgment of whether to perform visualization efficiency adaptive regulation further comprises the following steps: The visualization parameters include parallel computing acceleration ratio, first screen loading time and LOD tile scheduling hit rate; The parallel computing acceleration ratio is compensated by an acceleration ratio adjustment factor and an acceleration ratio reference value to obtain an acceleration ratio correction value; The first screen loading time is compensated by a loading time adjustment factor and a loading time reference value to obtain a loading time correction value; The LOD tile scheduling hit rate is compensated by a hit rate adjustment factor and a hit rate reference value to obtain a hit rate correction value; The acceleration ratio correction value, the loading time correction value and the hit rate correction value are coupled to obtain data processing and visualization efficiency; The specific steps of judging whether to perform the visualization efficiency adaptive regulation and control are: If the data processing and visualization efficiency is higher than or equal to the visualization efficiency reference value, the visualization efficiency adaptive regulation and control is not performed, and if not, the engine loading time consumption intelligent adjustment and the tile generation efficiency intelligent adjustment are performed based on the visualization efficiency offset, the visualization efficiency offset representing the negative deviation degree of the data processing and visualization efficiency from the visualization efficiency reference value.

3. The method of claim 2, wherein the method further comprises: The specific steps of performing the engine loading time consumption intelligent adjustment are: The submission rate of the rendering instruction queue is monitored, and the engine loading time consumption intelligent adjustment is performed based on a preset submission rate-loading budget proportion mapping relationship, wherein the submission rate-loading budget proportion mapping relationship is configured as an inversely related adaptive adjustment relationship, specifically: When the rendering instruction submission rate is lower than a first threshold value, first-level mapping is performed; When the rendering instruction submission rate is higher than a second threshold value, second-level mapping is performed; When the rendering instruction submission rate is within a submission rate threshold interval, the engine loading time consumption intelligent adjustment is not performed, and the current proportion of loading time consumption in frame time budget is maintained, the submission rate threshold interval representing a closed interval formed by the first threshold value and the second threshold value.

4. The method of claim 3, wherein the method further comprises: The first-level mapping is specifically: The submission rate and the visualization efficiency offset are input into the submission rate-loading budget proportion mapping relationship to map a gain target loading time proportion, the gain target loading time proportion is combined with the current per-frame target total time budget, and a target gain absolute loading time of the current frame is obtained through linear mapping, so that the loading budget proportion is dynamically compensated and corrected based on the real-time feedback of the GPU rendering execution efficiency in addition to the baseline regulation and control according to the CPU instruction submission pressure; The second-level mapping is specifically: The submission rate and the visualization efficiency offset are input into the submission rate-loading budget proportion mapping relationship to map a reduction target loading time proportion, the reduction target loading time proportion is combined with the current per-frame target total time budget, and a target reduction absolute loading time of the current frame is obtained through linear mapping, so that the loading budget predictability is further tightened when the CPU instruction submission pressure reaches the threshold value, and the determinacy delay of the rendering instruction execution pipeline is preferentially guaranteed.

5. The method of claim 2, wherein the method further comprises: The specific steps of the tile generation efficiency intelligent adjustment are: Monitor an interactive response time of a user interaction operation, and dynamically adjust a generation efficiency of a LOD tile generation thread according to the interactive response time; Step S21, define an inverse regulation relationship between the interactive response time and the generation efficiency, specifically: establish a time-delay-generation-efficiency inverse mapping relationship between the interactive response time and the LOD tile generation efficiency, wherein the time-delay-generation-efficiency inverse mapping relationship is configured as: when the interactive response time is lower than or equal to a time-delay threshold, the tile generation efficiency intelligent adjustment is not performed, and the current LOD tile generation efficiency is maintained; when the interactive response time is higher than the time-delay threshold, the LOD tile generation efficiency is reduced; Step S22, the LOD tile generation efficiency is reduced, specifically: input the interactive response time offset and the visualization efficiency offset into the time-delay-generation-efficiency inverse mapping relationship to obtain a generation efficiency attenuation coefficient, multiply the generation efficiency attenuation coefficient with the current LOD tile generation efficiency to obtain a target LOD tile generation efficiency of the next round, so that the front-end interactive time delay index representing the user perception fluency and the back-end rendering efficiency index reflecting the GPU execution load are fused to preferentially guarantee the determinacy time delay of the user interactive response link and the stability of the rendering pipeline, and the interactive response time offset represents a positive deviation degree of the interactive response time from the time-delay threshold.

6. The method of claim 1, wherein the method further comprises: The specific steps of obtaining the three-dimensional earth data fusion quality are: The fusion quality parameter includes to-be-compared visualization efficiency, three-dimensional earth data fusion delay, and automatic synchronization trigger response time; The to-be-compared visualization efficiency is compensated by a visualization efficiency adjustment factor and a visualization efficiency reference value to obtain a visualization efficiency correction value; The fusion delay reference value and the three-dimensional earth data fusion delay are compensated by a fusion delay adjustment factor to obtain a fusion delay correction value; The automatic synchronization trigger response time is compensated by a response time adjustment factor and a response time reference value to obtain a response time correction value; The visualization efficiency correction value, the fusion delay correction value, and the response time correction value are coupled to obtain the three-dimensional earth data fusion quality.

7. The method of claim 6, wherein the method further comprises: The specific steps of performing the intelligent adjustment of the update frame rate threshold are: If the cross-source data matching time consumption is less than or equal to a matching time consumption reference value, the intelligent adjustment of the update frame rate threshold is not performed, and the current dynamic rendering update frame rate threshold is maintained; If the cross-source data matching time consumption is greater than the matching time consumption reference value, a matching time consumption offset and a fusion quality offset are input into a time consumption-frame rate threshold inverse elastic mapping relationship to obtain an update frame rate threshold attenuation coefficient, and the update frame rate threshold attenuation coefficient is interactively processed with the dynamic rendering update frame rate threshold to obtain a dynamic rendering update frame rate threshold of the next round, so that when a delay occurs in the data matching process, effective rendering update based on high-quality fusion data is preferentially triggered, and invalid rendering calculation load caused by low-quality or unready data is inhibited, and the matching time consumption offset represents a positive deviation degree of the cross-source data matching time consumption from the matching time consumption reference value.

8. The method of claim 1, wherein the method further comprises: The sliding window width intelligent adjustment further includes: S100, if the fusion result verification time delay is lower than the lower limit of the verification time delay threshold, input the current fusion result verification time delay and the fusion quality offset to the time delay-window width mapping relationship to obtain a window width up-regulation coefficient, interact the window width up-regulation coefficient with the current time series data sliding window width to obtain a target time series data sliding window width used in the next fusion cycle, to allocate to the fusion task that needs higher data quality and deeper historical context, and improve the integrity of data fusion by actively expanding the sliding window width; S101, if the fusion result verification time delay is within the verification time delay threshold interval, input the current fusion result verification time delay and the fusion quality offset to the time delay-window width mapping relationship to obtain a window width down-regulation coefficient, interact the window width down-regulation coefficient with the current time series data sliding window width to obtain a target time series data sliding window width used in the next fusion cycle, to resist potential overload risk and ensure the stability and timeliness of the data processing pipeline; S102, if the fusion result verification time delay is higher than the upper limit of the verification time delay threshold, send a warning message to the preset operation and maintenance personnel.

9. A system for applying the method for fusing three-dimensional earth data based on multi-source spatio-temporal imagery according to any one of claims 1-8, characterized in that, The processing visualization efficiency quantification module, the visualization efficiency adaptive regulation and control module, the fusion quality quantification module, and the fusion accuracy adaptive regulation and control module are included. The processing visualization efficiency quantification module is configured to obtain three-dimensional earth data from multiple source spatio-temporal images in parallel, obtain visualization parameters in the conversion process from original three-dimensional earth data to an interactive three-dimensional scene, obtain data processing and visualization efficiency based on the visualization parameters, and quantify the effectiveness of three-dimensional earth data processing and visualization. The visualization efficiency adaptive regulation and control module is configured to determine whether to perform visualization efficiency adaptive regulation and control according to the data processing and visualization efficiency, avoid unnecessary regulation and control overhead and strategy oscillation, realize the precision and intelligence of resource scheduling, perform a three-dimensional earth data fusion link after visualization efficiency adaptive regulation and control if yes, and directly perform a three-dimensional earth data fusion link if no. The fusion quality quantification module is configured to obtain fusion quality parameters in the three-dimensional earth data fusion link, obtain three-dimensional earth data fusion quality based on the fusion quality parameters, and quantify the success rate of automatic matching of multi-source heterogeneous three-dimensional earth data after data fusion in the three-dimensional earth data fusion process. The fusion accuracy adaptive regulation and control module is configured to determine whether to perform fusion accuracy adaptive regulation and control according to the three-dimensional earth data fusion quality, thereby constructing a precise optimization closed loop oriented to output three-dimensional earth data fusion quality and event-driven on the premise of ensuring the availability of three-dimensional earth data, output three-dimensional earth data fusion results after fusion accuracy adaptive regulation and control if yes, and directly output three-dimensional earth data fusion results if no.

Images