A sky-ground integrated data fusion management system and method thereof

Abstract

The present invention discloses a sky-ground integrated data fusion management system and method thereof, which relates to the field of data processing technology. By efficiently integrating and processing multi-source data, high-precision virtual space data is generated, and real-time decision support and data sharing functions are provided for users. The system includes a data access module, a data preprocessing module, a space generation module, a data adaptation module, a spatiotemporal alignment module, a spatiotemporal coordinate embedding module, a data fusion module, a global supervision module and a data service module. Through a multi-head attention mechanism and spatiotemporal synchronization technology, the present invention can ensure the precise alignment and efficient fusion of data from different sources in time and space. The system utilizes dynamic segmentation and cross-modal data association technology to optimize data fusion accuracy. At the same time, a global supervision mechanism is used to monitor the operating status of each module in real time and perform dynamic optimization to ensure efficient and stable operation of the system.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a sky-ground integrated data fusion management system and method thereof. Background Art

[0002] With the rapid development of remote sensing technology, drones, ground sensors and other technologies, a wide variety of environmental data are being obtained, covering multiple sources such as satellite, aerial and ground data. However, due to the temporal and spatial differences of these data sources and the diversity of data formats, how to efficiently and accurately integrate them and apply them to decision support has become a challenge for current technology.

[0003] In the existing technology, Chinese invention patent CN110068655B proposes an integrated sky-ground atmospheric monitoring system, which includes data acquisition, storage, processing and monitoring modules. The accuracy of atmospheric monitoring is improved by multi-source data fusion. The system realizes large-scale, all-weather environmental monitoring. However, the lack of an efficient spatiotemporal alignment mechanism in the data processing process may lead to insufficient accuracy of the fused data, especially when processing dynamically changing high-frequency data, which may cause errors.

[0004] In addition, Chinese invention patent application CN111028096A discloses a system for integrated data fusion of space, air, and ground. It uses satellite, air, and ground monitoring data combined with Internet of Things technology to achieve a balance between large-scale and small-scale data. However, although its method has advantages in agricultural monitoring, the processing of spatiotemporal alignment and fusion accuracy of multi-source data is still not sophisticated enough, especially when involving dynamic scenes and highly dynamic areas. It may not be possible to guarantee high-precision spatiotemporal consistency and data accuracy. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and to propose a sky-ground integrated data fusion management system and method thereof to solve the above-mentioned problems.

[0006] This paper introduces a multi-head attention mechanism, spatiotemporal alignment technology, and a dynamic segmentation algorithm to solve the problems of insufficient data fusion accuracy and spatiotemporal synchronization in the existing technology. Specifically, the system has the following improvements:

[0007] Multi-head attention mechanism: A multi-head attention mechanism is used to perform weighted matching of triangle vertices from data sources such as satellites, drones, and ground sensors to ensure the accuracy of spatiotemporal alignment. This technology makes up for the shortcomings of existing technologies in the spatiotemporal synchronization processing of high-frequency updated data (such as meteorological sensors).

[0008] Spatiotemporal alignment and embedding: After data adaptation and spatiotemporal alignment, the present invention uses the BeiDou satellite positioning system and timestamp to accurately embed the spatiotemporal coordinates of each data point to form a spatiotemporal quadruple (x, y, z, t), providing accurate spatial positioning for subsequent data fusion.

[0009] Dynamic segmentation and cross-modal fusion: Through dynamic segmentation units, high-dynamic areas are carefully processed based on data error, ensuring accurate fusion of data in dynamically changing areas, thereby overcoming the problem of inaccurate processing of high-dynamic scenes in existing technologies.

[0010] The object of the present invention is achieved through the following technical solutions: a sky-ground integrated data fusion management system, characterized by comprising a data access module, a data preprocessing module, a space generation module, a data adaptation module, a time-space alignment module, a time-space coordinate embedding module, a data fusion module and a global supervision module;

[0011] Space generation module: This module establishes a three-dimensional virtual space based on the target area, divides the virtual space into basic grids according to the preset granularity, and uses the Delaunay method to triangulate each basic grid to generate a virtual grid.

[0012] The data adaptation module is used to adapt data from different sources to the virtual grid, specifically including: satellite data adaptation unit, aerial video adaptation unit and ground data adaptation unit;

[0013] The spatiotemporal alignment module aligns the spatiotemporal coordinates of the data. It adjusts the spatiotemporal alignment of the data through the multi-head attention mechanism, spatial attention, and temporal attention submodules. Specifically, it includes: multi-head attention mapping unit, spatial attention unit, and temporal attention unit.

[0014] The spatiotemporal coordinate embedding module is used to embed the spatiotemporal coordinates of each triangle vertex using the BeiDou satellite positioning system and timestamp after completing data adaptation and spatiotemporal alignment, forming a spatiotemporal quadruple (x, y, z, t).

[0015] Data fusion module: uses a multi-head attention mechanism to perform weighted fusion of data to achieve deep fusion and real-time update. Specifically, it includes: dynamic segmentation unit and cross-modal data association unit;

[0016] The global supervision module is used to monitor the operating status of each module in real time and dynamically optimize the system. It specifically includes: a global multi-head attention supervision unit and a feedback control unit;

[0017] The data service module is used to display and share integrated data to support decision-making and collaborative operations.

[0018] The data access module is used to receive and integrate multi-source data from satellites, drones, ground sensors, and point clouds, and to centrally manage each data source through the data transmission unit and achieve real-time and efficient access.

[0019] The data preprocessing module is used to process the access data to ensure data quality and spatiotemporal consistency. Specifically, it includes: a spatiotemporal synchronization correction unit for performing spatial and temporal correction on the access data to ensure spatiotemporal consistency of multi-source data; a data format standardization unit for converting various types of data into a unified format to ensure the compatibility of subsequent processing; and a noise suppression unit for filtering noise from the original data, eliminating invalid information, and improving data quality.

[0020] The satellite data adaptation unit cuts the satellite linear scan image according to the virtual grid size, triangulates each strip, and dynamically calibrates the spatiotemporal coordinates of the triangle vertices through satellite orbit parameters; the aerial video adaptation unit splits the video frame into continuous strips, triangulates them according to the same rules, and performs adaptive secondary subdivision on high dynamic areas; the ground data adaptation unit maps the point cloud or sensor data to the virtual grid, constructs a local triangulated network, and performs secondary subdivision on the local network based on curvature or density thresholds.

[0021] The multi-head attention mapping unit uses a multi-head attention mechanism to perform weighted matching of triangle vertices from different data sources to ensure the alignment of spatiotemporal coordinates; the spatial attention unit adjusts the triangulation density according to terrain features (such as ridges and rivers) to ensure accurate alignment of data in highly dynamic areas; the temporal attention unit analyzes the timing characteristics of high-frequency update data through a self-attention mechanism, automatically adjusts the time weight, and suppresses historical data noise.

[0022] The dynamic subdivision unit recursively subdivides the areas exceeding the threshold according to the regional errors in data fusion, thereby achieving refined processing of high-dynamic areas; the cross-modal data association unit optimizes the information complementarity between different data sources through cross-modal data association technology, thereby improving the accuracy of data fusion.

[0023] The global multi-head attention supervision unit monitors the status and quality of each module including data acquisition, preprocessing, virtual space modeling and data fusion in real time; the feedback control unit dynamically adjusts the operating parameters and processing strategies of each module according to the global supervision feedback information to ensure efficient coordination of the overall system.

[0024] The data service module includes: a data visualization unit, which is used to display the fused data in a graphical form and supports the intuitive display of spatial information; a decision support unit, which is used to provide users with real-time decision support based on data analysis; and a data sharing unit, which is used to provide a standardized data interface and support cross-platform data sharing and collaboration.

[0025] A method for integrated sky-ground data fusion management includes the following steps:

[0026] S1: Receives and integrates multi-source data from satellites, drones, ground sensors, and point clouds, and implements unified data management and real-time efficient access through data transmission units;

[0027] S2: Preprocess the incoming data, including spatiotemporal synchronization correction, data format standardization, and noise suppression, to ensure data quality and spatiotemporal consistency;

[0028] S3: Establish a three-dimensional virtual space based on the target area, divide the virtual space into basic grids according to the preset granularity, and use the Delaunay method to triangulate the basic grid to generate a virtual grid;

[0029] S4: Adapting data from different sources to the virtual grid, including satellite data adaptation, aerial video adaptation, and ground data adaptation;

[0030] S5: Use multi-head attention mechanism, spatial attention and temporal attention to align the spatiotemporal coordinates of sub-data to ensure the spatiotemporal consistency of multi-source data;

[0031] S6: After completing data adaptation and spatiotemporal alignment, the spatiotemporal coordinates of each triangle vertex are embedded using the BeiDou satellite positioning system and timestamp to form a spatiotemporal quadruple (x, y, z, t).

[0032] S7: Perform weighted fusion on the data after spatiotemporal coordinate embedding, including dynamic segmentation and cross-modal data association, to achieve deep data fusion and real-time update;

[0033] S8: Monitors the operating status of each module in real time, and adjusts the operating parameters of each module through a global multi-head attention supervision mechanism and feedback control to ensure efficient coordination of the entire system;

[0034] S9: The integrated data is visualized, decision-supported, and shared for users to make real-time decisions and collaborate.

[0035] In the process of receiving and integrating multi-source data in step S1, a data transmission unit is used to access satellite, UAV, ground sensor and point cloud data in real time, and synchronous access of each data source is achieved through a unified interface protocol.

[0036] The beneficial effects of the present invention are:

[0037] 1. Through the multi-source data access module, it can access data from different sources such as satellites, drones, ground sensors, point clouds, etc. in real time and efficiently. Through the unified management of the data transmission unit, the synchronous access and stable transmission of various data sources are ensured. The system can automatically adapt according to different data types and sources, greatly improving the data integration efficiency. For application scenarios with high real-time requirements (such as disaster warning, traffic management, etc.), the system can provide real-time data streams to ensure rapid response and processing, and can ensure real-time access to multiple data sources to meet the real-time data requirements of different fields. Through the unified management of the data transmission unit, the data access efficiency is improved and the integration process of multi-source data is simplified.

[0038] 2. The data preprocessing module ensures that data from different data sources can be integrated in the same time and space framework through steps such as time-space synchronization correction, data format standardization and noise suppression. Time-space synchronization correction can eliminate the time error of data from different sources, data format standardization ensures that the data can be seamlessly connected, and noise suppression improves data quality, removes invalid information, and ensures data accuracy in subsequent processing. Through time-space synchronization correction, the consistency of data in time and space is ensured, providing a reliable data foundation for subsequent analysis.

[0039] 3. The spatial generation module employed in this invention meticulously divides the virtual space into basic grids using the Delaunay triangulation method, and the data adaptation module enables precise adaptation of different data sources. This process ensures that different data types (such as satellite imagery, aerial video, and ground sensor data) can be accurately mapped into a unified virtual space, ensuring the consistency and high precision of the data within the virtual grid. The virtual grid generated by the Delaunay triangulation method ensures a refined representation of the virtual space and is suitable for geographic information systems (GIS) and spatial analysis.

[0040] 4. The present invention realizes the efficient operation of the spatiotemporal alignment module through multi-head attention mechanism, spatial attention, temporal attention and other technologies. The system can accurately align the spatiotemporal coordinates from different data sources, and perform weighted fusion of multi-source data through the data fusion module. This process can eliminate the errors between data from different sources, ensure the accurate fusion of multi-source data in the same spatiotemporal framework, and optimize the complementarity between different data sources through cross-modal data association technology, further improving the fusion accuracy. Through weighted fusion and dynamic segmentation processing, the system can achieve high-precision data fusion and adapt to complex dynamic environments.

[0041] 5. The system's global supervision module can monitor the operating status of modules such as data acquisition, preprocessing, virtual space modeling and data fusion in real time. It can evaluate the status and quality of each module in real time through the global multi-head attention supervision mechanism, and dynamically adjust the system's operating parameters based on feedback information to ensure that each module works together. Through this adaptive optimization, the system can cope with changing environments and needs, ensure overall efficiency and accuracy, and the real-time monitoring and feedback mechanism ensures that the system remains efficient and stable during long-term operation, reducing possible errors and delays.

[0042] 6. The data service module provides data visualization, decision support and data sharing functions. It can display the fused data in graphical form to help users intuitively understand complex data. In addition, the decision support unit provides real-time decision suggestions based on the fused data to assist users in making scientific decisions. The data sharing function supports cross-platform data sharing through standardized interfaces, promoting data circulation and collaboration.

[0043] 7. This invention uses multiple technologies, including spatiotemporal synchronization, data adaptation, spatiotemporal alignment, and high-precision fusion, to ensure that the system can process massive amounts of data from multiple platforms and devices in real time, providing timely decision support and real-time feedback. The system is particularly suitable for application scenarios that require real-time response and processing, such as intelligent traffic management, environmental monitoring, and urban planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a system architecture diagram of the present invention;

[0045] Figure 2 System interaction for the present invention Figure 1 ;

[0046] Figure 3 System interaction for the present invention Figure 2 . DETAILED DESCRIPTION

[0047] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0048] It is to be noted that the directions of "left", "right", "up", "down", "front", "back", "inside" and "outside" in the following schemes are all relative directions and are not listed here one by one.

[0049] Example 1:

[0050] like Figures 1 to 3As shown, this embodiment provides an integrated sky-ground data fusion management system, which aims to generate high-precision virtual space data by integrating and processing multi-source data, and provide users with decision support and data sharing functions. The system architecture includes a data access module, a data preprocessing module, a space generation module, a data adaptation module, a space-time alignment module, a space-time coordinate embedding module, a data fusion module, a global supervision module and a data service module.

[0051] The data access module is responsible for receiving data from multiple sources such as satellites, drones, ground sensors, point clouds, etc., and uniformly managing them through the data transmission unit to ensure real-time and efficient access to the data. During this process, the system can perform appropriate preprocessing based on the characteristics and data formats of different data sources to adapt to subsequent processing modules.

[0052] The incoming data first passes through the data preprocessing module for spatiotemporal synchronization correction, data format standardization and noise suppression. The spatiotemporal synchronization correction unit ensures the consistency of multi-source data in the spatiotemporal dimensions by using the spatiotemporal calibration algorithm. The data format standardization unit converts the formats of different sources into a unified standard format to ensure seamless connection of data during processing. The noise suppression unit filters the original data, removes invalid information, and improves the quality of the data.

[0053] The spatial generation module constructs a three-dimensional virtual space based on a target area (such as a specific city, region, or county). It uses the Delaunay triangulation algorithm to grid the virtual space, generating a base grid at a set granularity (such as 1 km × 1 km × altitude level), and then generating a three-dimensional virtual grid based on this base grid. This process ensures that the virtual space has a detailed and highly adaptable structure and can accurately represent the spatial characteristics of the target area.

[0054] The data adaptation module adapts data from different sources to the generated virtual grid, including:

[0055] Satellite data adaptation unit: cuts the satellite's linear scan image according to the size of the virtual grid, triangulates each strip, and dynamically calibrates it using satellite orbit parameters to ensure accurate data alignment.

[0056] Aerial video adaptation unit: Splits aerial video frames into continuous strips, uses triangulation rules similar to satellite data, and performs adaptive encryption subdivision on high dynamic areas.

[0057] Ground data adaptation unit: maps ground sensor or point cloud data to a virtual grid, constructs a local triangulated network, and performs secondary subdivision on high-density or complex areas to ensure precise data adaptation.

[0058] The spatiotemporal alignment module is responsible for aligning the spatiotemporal coordinates of the data. Through the multi-head attention mechanism, the spatial attention and temporal attention sub-modules are combined to ensure that multi-source data can be aligned in space and time. Specifically, it includes:

[0059] Multi-head attention mapping unit: A multi-head attention mechanism is used to perform weighted matching of triangle vertices from different data sources (such as satellite data, aerial videos, and ground data) to ensure accurate alignment of spatiotemporal coordinates.

[0060] Spatial Attention Unit: Automatically adjusts triangulation density based on terrain features (e.g., ridges, rivers), ensuring precise alignment of data in highly dynamic areas.

[0061] Temporal Attention Unit: Analyzes the temporal characteristics of high-frequency updated data through the self-attention mechanism, automatically adjusts weights, and suppresses historical noise.

[0062] To achieve accurate spatiotemporal alignment of multi-source data, the following spatiotemporal dynamic alignment attention formula is used:

[0063]

[0064] in:

[0065] and are the query vector of data i and the key vector of data j in the h-th attention head respectively;

[0066] and Respectively represent the time and space differences between data i and data j;

[0067] are learnable parameters;

[0068] represents element-wise product, is the L1 norm.

[0069] After completing data adaptation and spatiotemporal alignment, the spatiotemporal coordinate embedding module uses the BeiDou satellite positioning system and precise timestamps to embed the spatiotemporal coordinates of each triangle vertex, generating a spatiotemporal quadruple (x, y, z, t), providing accurate spatial information for subsequent data fusion and precision enhancement.

[0070] In order to achieve accurate spatiotemporal positioning of each triangle vertex, the following embedding formula is proposed:

[0071]

[0072] in:

[0073] (x, y, z) are spatial coordinates, and t is the timestamp;

[0074] W is the weight matrix, b is the bias;

[0075] Λ and are the scale factors for space and time, respectively;

[0076] Sine and cosine functions are used for periodic encoding, and then mapped to the high-order feature space through the nonlinear activation function tanh to achieve the fusion expression of spatiotemporal information;

[0077] The data fusion module uses a multi-head attention mechanism to perform weighted fusion of data, ensuring deep fusion and real-time updating of multi-source data. Specifically, it includes:

[0078] Dynamic subdivision unit: Based on the regional errors that occur during the data fusion process, recursively subdivides the areas that exceed the threshold to ensure that high dynamic areas are processed in detail.

[0079] In order to adaptively subdivide the high dynamic area during the data fusion process, the following recursive subdivision formula is proposed:

[0080]

[0081] in:

[0082] represents a triangular area, For its subdivided sub-areas;

[0083] is the preset error threshold;

[0084] Regional error The following definitions are used:

[0085]

[0086] in:

[0087] is the original value of a data source, is the fused value;

[0088] is a hyperparameter, is the number of data points in the region;

[0089] Cross-modal data association unit: Optimizes information complementarity between different data sources through cross-modal data association technology, further improving the accuracy of data fusion.

[0090] The global supervision module is responsible for real-time monitoring of the operating status of each module and dynamically optimizing system performance based on feedback information, including:

[0091] Global multi-head attention supervision unit: monitors the operating status and quality of data acquisition, preprocessing, virtual space modeling, and data fusion modules to ensure the consistency and efficiency of the system.

[0092] Feedback control unit: adjusts the operating parameters of each module based on global supervision feedback to ensure the coordinated operation and optimization of each part.

[0093] The data service module visualizes the integrated data, provides real-time decision support for users, and supports cross-platform data sharing and collaboration. This module includes:

[0094] Data visualization unit: presents the fused data graphically to help users understand the data intuitively.

[0095] Decision support unit: Provides real-time decision support based on fused data to assist users in making scientific decisions.

[0096] Data sharing unit: provides standardized data interfaces and supports cross-platform data sharing and collaboration.

[0097] The system of this embodiment can generate high-precision virtual spatial data through the precise fusion of multi-source data. It is widely used in fields such as geographic information systems (GIS), intelligent transportation, and environmental monitoring. Through the spatiotemporal alignment and data fusion process, it can effectively reduce errors and improve the accuracy and practicality of data. The system can not only perform efficient data processing and dynamic optimization, but also support users in real-time decision-making and data sharing, enhancing the functions of collaboration and decision support.

[0098] In addition, through the multi-head attention mechanism and spatiotemporal coordinate embedding technology, this embodiment can ensure the consistency of various types of data in multi-dimensional space, providing a reliable data foundation for subsequent data analysis and processing.

[0099] Example 2:

[0100] like Figures 1 to 3 As shown, this embodiment introduces in detail the integrated sky-ground data fusion management system based on Example 1, and further emphasizes the specific implementation methods of the data access module, data preprocessing module and data adaptation module. The functions of these modules are to ensure efficient access and high-quality processing of multi-source data, and provide accurate input for subsequent data fusion and spatiotemporal alignment. The system is also suitable for integrated processing of multi-source data such as satellites, aerial photography, and ground sensors to construct accurate virtual space models and provide decision support and data sharing functions.

[0101] The main task of the data access module is to receive data from different data sources, including satellite data, drone data, ground sensor data, and point cloud data. The working process of this module is as follows:

[0102] Data collection: Obtain multi-source data, including images, videos, sensor data, etc., from satellites, drones and other devices through wireless communication technology or other communication means.

[0103] Data integration: All data sources (satellites, drones, ground sensors, and point cloud data) are centrally managed through the data transmission unit to ensure real-time access and efficient synchronization of various data sources.

[0104] Real-time processing: This module ensures that the real-time collected data can quickly enter the pre-processing module, providing a fast response for subsequent data processing.

[0105] This module ensures efficient and stable access to data, provides the system with real-time and complete multi-source data input, adapts to the access requirements of different types of data, and ensures efficient operation of the system.

[0106] The data preprocessing module is used to process the incoming multi-source data to ensure data quality and temporal and spatial consistency. The specific implementation process includes the following submodules:

[0107] Spatiotemporal synchronization correction unit: This unit ensures the spatiotemporal consistency of different data sources by comparing the timestamps and geographic coordinate information of multiple data sources. Through the spatiotemporal calibration algorithm, all accessed data can be aligned within the same time and space framework.

[0108] Data format standardization unit: This unit converts data in different formats into a unified standard format to ensure that the system can efficiently process multiple data from different sources.

[0109] Noise suppression unit: Filters the raw data to remove invalid or erroneous information to improve data quality and provide a clean data source for subsequent data fusion and analysis.

[0110] By synchronizing the time and space of data, unifying the format, and suppressing noise, this module significantly improves the quality of data, ensures that the data can be seamlessly connected in the subsequent processing process, and reduces calculation errors caused by data problems.

[0111] The main function of the data adaptation module is to adapt data from different sources to the virtual grid. This module specifically includes:

[0112] Satellite data adaptation unit: The satellite data adaptation unit is responsible for cutting the linear scan images acquired by the satellite according to the size of the virtual grid and triangulating each strip. During this process, the time and space coordinates of each triangle vertex are dynamically calibrated through satellite orbit parameters to ensure accurate data alignment.

[0113] The satellite data adaptation unit formula is as follows:

[0114] Assume that the satellite scan image is , the default virtual grid size is and Cut the image into strips s according to the grid size, and triangulate the strips. Suppose a triangle T in a strip is obtained = { }. For each vertex , using the satellite orbit parameter function (in Perform dynamic calibration (at the acquisition moment) and calculate the vertex coordinates after calibration:

[0115] ,i=1,2,3

[0116] in, is the correction factor, This formula combines satellite orbit information with vertex distance to achieve dynamic calibration.

[0117] Aerial Video Adaptation Unit: This unit splits each frame of aerial video into continuous strips, triangulates them according to rules similar to satellite data, and performs adaptive secondary subdivision on high-dynamic areas to ensure that the data in these areas can be accurately adapted to the virtual grid.

[0118] The aerial video adaptation unit formula is as follows:

[0119] Assume that the aerial video frame is V(t,x,y), and split each frame into continuous strips Then, a triangulation method similar to that used for satellite data was used to obtain the initial grid. To adapt to high dynamic areas, the regional adaptation factor D is defined:

[0120]

[0121] in, is the local pixel change within the strip (or motion estimation value), is the threshold, is the scaling parameter, is the basic subdivision factor, and then the preliminary grid Recursive subdivision:

[0122]

[0123] This formula adaptively increases the grid density in high dynamic areas to achieve refined adaptation.

[0124] Ground data adaptation unit: This unit is responsible for mapping ground sensor or point cloud data to the virtual grid, constructing a local triangulated network and subdividing the grid based on curvature or density thresholds to ensure that the ground data can be accurately matched with the virtual space.

[0125] The ground data adaptation unit formula is as follows:

[0126] Assume that the ground sensor or point cloud data is G (x, y, z), which is assigned to the virtual grid through the mapping function:

[0127]

[0128] Perform local triangulation on the data within each grid cell to obtain a preliminary grid To perform secondary subdivision for high-density or complex areas, define the subdivision factor :

[0129]

[0130] in Basic subdivision factor, is the scaling parameter. Then Perform secondary subdivision:

[0131]

[0132] This formula uses local density or curvature information to achieve adaptive encryption and subdivision of complex areas, ensuring fine adaptation of ground data.

[0133] This module enables data from different sources to be seamlessly connected to the virtual space, ensuring that all types of data can be adapted to a unified grid with the same standards. This process greatly improves the accuracy of data integration and provides a reliable foundation for subsequent steps such as spatiotemporal alignment and data fusion.

[0134] The spatiotemporal alignment module achieves precise alignment of different data sources in space and time through a multi-head attention mechanism, spatial attention, and temporal attention sub-modules. Through this precise alignment, the system can ensure that multi-source data are analyzed and fused in the same coordinate system.

[0135] After completing data adaptation and spatiotemporal alignment, the spatiotemporal coordinate embedding module uses the BeiDou satellite positioning system and timestamp to embed the spatiotemporal coordinates of each triangle vertex, providing accurate spatial positioning for subsequent data fusion.

[0136] The data fusion module uses a multi-head attention mechanism combined with dynamic segmentation and cross-modal data association technology to ensure deep fusion and real-time updating of data.

[0137] The global supervision module monitors the operating status of each module in real time, and dynamically optimizes the system through the global multi-head attention supervision unit and feedback control unit to ensure the efficiency and accuracy of data processing.

[0138] The data service module provides data visualization, decision support and data sharing functions, enabling users to intuitively understand data, obtain real-time decision support, and share and collaborate on data with other systems or users.

[0139] During the processes of data access, preprocessing, adaptation, and spatiotemporal alignment, this embodiment ensures that the system can efficiently process data from different sources and provides high-quality input for subsequent data fusion. In particular, during the processes of spatiotemporal synchronization and data adaptation, sophisticated algorithms and techniques are used to ensure that various types of data can be seamlessly integrated, providing strong support for the final spatial generation and decision support.

[0140] In addition, the system's global supervision function ensures the coordinated work of various modules, greatly improving data quality and timeliness. Through these optimizations, the system can be widely used in urban management, environmental monitoring, intelligent transportation and other fields, helping users to conduct efficient data analysis and decision-making.

[0141] Example 3: Sky-ground integrated data fusion management method

[0142] like Figures 1 to 3 As shown, this embodiment describes in detail a method for integrated sky-ground data fusion management based on Embodiments 1 and 2. Through multi-step processing, efficient access, preprocessing, virtual space modeling, data adaptation, spatiotemporal alignment, coordinate embedding, weighted fusion, global supervision, and data services of multi-source data are achieved. The specific steps are as follows:

[0143] S1: Receive and integrate data from multiple sources

[0144] Utilizing high-performance data transmission units, data is collected from satellites, drones, ground sensors, and point cloud devices in real time through a unified interface protocol. Each data source collects information in its own specific format, and the data transmission unit manages it in a unified manner, achieving synchronous data access and stable transmission, laying the foundation for subsequent processing.

[0145] S2: Data Preprocessing

[0146] The incoming raw data is subjected to spatiotemporal synchronization correction, data format standardization, and noise suppression. The spatiotemporal synchronization correction unit eliminates differences in device location and acquisition time by comparing the timestamps and geographic coordinates of different data sources; the format standardization unit converts data in different formats into a unified standard; and the noise suppression unit uses filtering and outlier removal technology to improve data quality and consistency.

[0147] S3: Virtual Space Construction

[0148] A three-dimensional virtual space is constructed based on the target area, and the space is divided into basic grids according to a preset granularity (for example, 1km×1km×altitude layer). The Delaunay triangulation algorithm is used to finely divide each basic grid to generate a finely structured virtual grid, providing an accurate spatial basis for data mapping.

[0149] S4: Data Adaptation

[0150] Adapting data from different sources into the virtual grid:

[0151] The satellite data adaptation unit cuts the satellite linear scan image into grid sizes, triangulates each strip, and dynamically calibrates the time and space coordinates of the triangle vertices using satellite orbit parameters;

[0152] The aerial video adaptation unit splits the video frames into continuous strips, adopts the same segmentation rules, and performs adaptive secondary subdivision on high dynamic areas;

[0153] The ground data adaptation unit maps the point cloud or sensor data to a virtual grid, constructs a local triangulated network, and performs secondary subdivision based on curvature or density thresholds.

[0154] S5: Space-time alignment

[0155] A multi-head attention mechanism is used to perform weighted matching on the triangle vertices generated by each data source to ensure accurate spatial correspondence of each data; at the same time, the spatial attention unit automatically adjusts the local grid density according to terrain features (such as ridges and rivers), and the temporal attention unit analyzes the timing characteristics of the data, adjusts the time weight, and effectively suppresses historical noise, thereby achieving precise spatiotemporal alignment.

[0156] S6: Spatiotemporal Coordinate Embedding

[0157] After data adaptation and alignment are completed, the Beidou satellite positioning system is used in combination with precise timestamps to embed the spatiotemporal coordinates of each triangle vertex, generating a spatiotemporal quadruple (x, y, z, t), forming an accurate spatiotemporal data mapping and providing a solid foundation for subsequent fusion.

[0158] S7: Data Weighted Fusion

[0159] A multi-head attention mechanism is used to perform weighted fusion on the embedded data to achieve deep data fusion and real-time update. The fusion module is equipped with a dynamic subdivision unit to recursively subdivide high dynamic areas according to regional errors, and uses a cross-modal data association unit to optimize the information complementarity between different data sources to improve fusion accuracy.

[0160] S8: Global Supervision and Feedback Control

[0161] The global multi-head attention supervision unit monitors the operating status of data acquisition, preprocessing, virtual space construction and data fusion modules in real time; the feedback control unit dynamically adjusts the operating parameters of each module according to the monitoring results to ensure the coordinated and efficient operation of all links in the system.

[0162] S9: Data Visualization and Decision Support

[0163] The integrated data is intuitively displayed through a data visualization platform, and a decision support unit is used to provide real-time decision suggestions. Cross-platform data sharing is achieved through standardized data interfaces, supporting user collaborative operations and real-time decision-making.

[0164] S10: Data synchronization access (additional step)

[0165] In step S1, a unified interface protocol is used through the data transmission unit to achieve synchronous access to satellite, UAV, ground sensor and point cloud data, ensuring the real-time and consistency of data access.

[0166] Utilizing a unified interface and high-performance data transmission unit, it ensures real-time collection and synchronous management of multi-source data, making it suitable for dynamic environmental monitoring.

[0167] Time and space synchronization correction, format standardization and noise suppression technology have significantly improved data quality and provided an accurate data basis for subsequent processing.

[0168] The virtual grid constructed by Delaunay triangulation ensures high-precision mapping of spatial data and adapts to the needs of complex terrain.

[0169] The application of multi-head attention mechanism and spatial and temporal attention realizes the accurate alignment and embedding of data in space and time, providing a reliable basis for data fusion.

[0170] Dynamic segmentation and cross-modal data association technologies effectively solve the problems of regional errors and multi-source data complementarity, and achieve deep fusion and real-time updates.

[0171] Real-time monitoring and feedback control mechanisms ensure stable and efficient operation of the system, reduce errors and improve overall processing efficiency.

[0172] The data visualization and decision support module provides users with intuitive data display and real-time decision-making basis, while supporting cross-platform data sharing and promoting collaborative work.

[0173] In summary, this embodiment significantly improves the accuracy and real-time performance of multi-source data fusion, and provides a powerful data analysis and decision support platform for smart city management, environmental monitoring, traffic management and other fields.

[0174] The above description is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be regarded as excluding other embodiments. Instead, it can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or technology or knowledge in related fields. The changes and modifications made by those skilled in the art do not depart from the spirit and scope of the present invention and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A sky-ground integrated data fusion management system, characterized in that: It includes data access module, data preprocessing module, space generation module, data adaptation module, spatiotemporal alignment module, spatiotemporal coordinate embedding module, data fusion module, global supervision module and data service module; The space generation module: establishes a three-dimensional virtual space based on the target area, divides the virtual space into basic grids according to a preset granularity, and triangulates each basic grid using the Delaunay method to generate a virtual grid; The data adaptation module is used to adapt data from different sources to a virtual grid, and specifically includes: a satellite data adaptation unit, an aerial video adaptation unit, and a ground data adaptation unit; the satellite data adaptation unit cuts the satellite linear scan image according to the virtual grid size, triangulates each strip, and dynamically calibrates the spatiotemporal coordinates of the triangle vertices using satellite orbit parameters; the aerial video adaptation unit splits the video frame into continuous strips, triangulates them according to the same rules, and performs adaptive secondary subdivision on high dynamic areas; the ground data adaptation unit maps point clouds or sensor data to the virtual grid, constructs a local triangulated network, and performs secondary subdivision on the local network based on curvature or density thresholds; The spatiotemporal alignment module aligns the spatiotemporal coordinates of the data, and adjusts the spatiotemporal alignment of the data through a multi-head attention mechanism, spatial attention, and temporal attention submodules. Specifically, it includes: a multi-head attention mapping unit, a spatial attention unit, and a temporal attention unit. The multi-head attention mapping unit uses a multi-head attention mechanism to perform weighted matching on triangle vertices from different data sources to ensure spatiotemporal coordinate alignment. The spatial attention unit adjusts the triangulation density according to the terrain characteristics to ensure accurate alignment of data in high-dynamic areas. The temporal attention unit analyzes the temporal characteristics of high-frequency update data through a self-attention mechanism, automatically adjusts the time weight, and suppresses historical data noise. The spatiotemporal coordinate embedding module is used to embed the spatiotemporal coordinates of each triangle vertex using the BeiDou satellite positioning system and timestamp after completing data adaptation and spatiotemporal alignment to form a spatiotemporal quadruple (x, y, z, t); The data fusion module: uses a multi-head attention mechanism to perform weighted fusion of data to achieve deep fusion and real-time update. Specifically, it includes: a dynamic segmentation unit and a cross-modal data association unit; The global supervision module is used to monitor the operating status of each module in real time and dynamically optimize the system. It specifically includes: a global multi-head attention supervision unit and a feedback control unit; The data service module is used to display and share the integrated data to support decision-making and collaborative operations.

2. The integrated sky-ground data fusion management system according to claim 1, characterized in that: The data access module is used to receive and integrate multi-source data from satellites, drones, ground sensors, and point clouds, and to uniformly manage each data source through a data transmission unit and achieve real-time and efficient access.

3. The air-ground integrated data fusion management system according to claim 1, characterized in that: The data preprocessing module is used to process the access data to ensure data quality and spatiotemporal consistency, and specifically includes: a spatiotemporal synchronization correction unit for performing spatial and temporal correction on the access data to ensure spatiotemporal consistency of multi-source data; a data format standardization unit for converting various types of data into a unified format to ensure compatibility of subsequent processing; and a noise suppression unit for filtering noise from the original data, eliminating invalid information, and improving data quality.

4. The integrated sky-ground data fusion management system according to claim 1, characterized in that: The dynamic subdivision unit recursively subdivides the areas exceeding the threshold according to the regional errors in data fusion, thereby achieving refined processing of high dynamic areas; the cross-modal data association unit optimizes the information complementarity between different data sources through cross-modal data association technology, thereby improving the accuracy of data fusion.

5. The air-ground integrated data fusion management system according to claim 4, characterized in that: The global multi-head attention supervision unit monitors the status and quality of each module including data acquisition, preprocessing, virtual space modeling and data fusion in real time; the feedback control unit dynamically adjusts the operating parameters and processing strategies of each module according to the global supervision feedback information to ensure efficient coordination of the overall system.

6. The air-ground integrated data fusion management system according to claim 5, characterized in that: The data service module includes: a data visualization unit for displaying the fused data in a graphical form to support the intuitive display of spatial information; a decision support unit for providing users with real-time decision support based on data analysis; and a data sharing unit for providing a standardized data interface to support cross-platform data sharing and collaboration.

7. The system according to any one of claims 1 to 6, characterized in that: The data fusion management method includes the following steps: S1: Receives and integrates multi-source data from satellites, drones, ground sensors, and point clouds, and implements unified data management and real-time efficient access through data transmission units; S2: Preprocess the incoming data, including spatiotemporal synchronization correction, data format standardization, and noise suppression, to ensure data quality and spatiotemporal consistency; S3: Establish a three-dimensional virtual space based on the target area, divide the virtual space into basic grids according to the preset granularity, and use the Delaunay method to triangulate the basic grid to generate a virtual grid; S4: Adapting data from different sources to the virtual grid, including satellite data adaptation, aerial video adaptation, and ground data adaptation; S5: Use multi-head attention mechanism, spatial attention and temporal attention to align the spatiotemporal coordinates of sub-data to ensure the spatiotemporal consistency of multi-source data; S6: After completing data adaptation and spatiotemporal alignment, the spatiotemporal coordinates of each triangle vertex are embedded using the BeiDou satellite positioning system and timestamp to form a spatiotemporal quadruple (x, y, z, t). S7: Perform weighted fusion on the data after spatiotemporal coordinate embedding, including dynamic segmentation and cross-modal data association, to achieve deep data fusion and real-time update; S8: Monitors the operating status of each module in real time, and adjusts the operating parameters of each module through a global multi-head attention supervision mechanism and feedback control to ensure efficient coordination of the entire system; S9: Visualize the integrated data, provide decision support, and share data for users to make real-time decisions and collaborate; In the process of receiving and integrating multi-source data in step S1, a data transmission unit is used to access satellite, UAV, ground sensor and point cloud data in real time, and synchronous access of each data source is achieved through a unified interface protocol.

Images