Multi-source heterogeneous data fusion sky-ground intelligent monitoring system

Abstract

The invention relates to the technical field of data processing, and provides a multi-source heterogeneous data fusion sky-ground intelligent monitoring system which obtains ground sensor network data and satellite remote sensing data, and carries out space-time calibration on the obtained ground sensor network data and satellite remote sensing data; a multi-dimensional space-time correlation model is constructed, the data calibrated by the space-time dynamic calibration module is further analyzed, and the fusion weight of a space and time correlation matrix is optimized; collaborative correction of multi-source data is implemented, the space-time joint incidence matrix is optimized, and space and time deviations are eliminated; and finally fusing and calibrating the processed multi-source heterogeneous data, and providing data output. According to the method, the multi-source data can be fused under the same precision, the dynamic response capability of the system to the credibility of different data sources is improved, errors caused by data deviation and inconsistency are effectively reduced, and the precision and reliability of a monitoring result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a multi-source heterogeneous data fusion sky-ground intelligent monitoring system. Background Art

[0002] With the development of remote sensing technology and ground sensor technology, the integrated sky-ground intelligent monitoring system has gradually become an important application means in the fields of environmental monitoring, weather forecasting, disaster warning, etc. Especially in large-scale and multi-scale environmental monitoring tasks, the combination of ground sensors and satellite remote sensing data can provide more comprehensive and accurate information. At present, ground sensor networks are mainly used to provide high-precision, real-time local environmental data, covering a variety of environmental parameters such as temperature, humidity, and air pressure; while satellite remote sensing technology can cover a large area and obtain important environmental information such as atmospheric composition, cloud cover, and surface temperature. With the continuous advancement of sensor data acquisition technology and satellite remote sensing technology, multi-source heterogeneous data fusion technology has gradually become a key technology to improve monitoring accuracy and timeliness. This technology makes up for the limitations of a single data source by uniformly modeling and fusing the spatiotemporal characteristics of different data sources, providing a more comprehensive and efficient monitoring method.

[0003] However, existing technologies still face many challenges in practical applications. For example, the spatiotemporal inconsistency between ground sensors and satellite remote sensing data is an issue that cannot be ignored. Ground sensors are limited in spatial distribution and have a high data update frequency, while satellite remote sensing data is subject to satellite orbits, observation angles, and imaging frequencies, resulting in large differences in temporal and spatial resolution between the two. Existing technologies mostly use static correction or linear interpolation methods to correct spatiotemporal deviations, but this method often ignores the complex nonlinear relationship between data, resulting in insufficient correction effects; existing multi-source data fusion methods usually assume that each data source has the same credibility or fusion weight, and fail to fully consider the reliability differences between different data sources, especially in abnormal weather conditions or high-density monitoring areas, which may result in large fusion errors, affecting the accuracy of the final data fusion. Summary of the invention

[0004] The embodiments of the present invention provide a multi-source heterogeneous data fusion sky-ground intelligent monitoring system, which can reduce errors caused by data deviation and inconsistency at least to a certain extent, and improve the accuracy and reliability of monitoring results.

[0005] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0006] According to one aspect of the present invention, there is provided a sky-ground intelligent monitoring system for multi-source heterogeneous data fusion, comprising: a data acquisition module for acquiring ground sensor network data and satellite remote sensing data; a spatiotemporal dynamic calibration module for performing spatiotemporal calibration on the acquired ground sensor network data and satellite remote sensing data to ensure the spatial and temporal consistency of the ground sensor network data and satellite remote sensing data; a spatiotemporal correlation matrix analysis module for constructing a multi-dimensional spatiotemporal correlation model, further analyzing the data calibrated by the spatiotemporal dynamic calibration module, optimizing the fusion weights of the spatial and temporal correlation matrices, and generating a spatiotemporal joint correlation matrix; an intelligent collaborative correction module for implementing collaborative correction of multi-source data, optimizing the spatiotemporal joint correlation matrix, and eliminating spatial and temporal deviations; and a data fusion and calibration output module for performing final fusion and calibration on the processed multi-source heterogeneous data to provide data output.

[0007] In the present invention, based on the aforementioned scheme, the data acquisition module includes: constructing the network data collected by distributed sensors into a ground sensor matrix, wherein the spatial dimension maps the sensor geographic coordinates and the time dimension records the time series changes of parameters; decomposing the satellite remote sensing data into an atmospheric composition matrix, a cloud cover matrix and a surface temperature matrix.

[0008] In the present invention, based on the above-mentioned scheme, the spatiotemporal dynamic calibration module includes: matching the spatial coordinates of the ground sensor network data with the spatial coordinates of the satellite remote sensing data, aligning each position in the ground sensor matrix with the data of the corresponding position in the atmospheric composition matrix, cloud cover matrix and surface temperature matrix; mutual verification of the ground sensor network data and the satellite remote sensing data, and when it is monitored that the spatial coordinate deviation exceeds a preset threshold, automatically triggering a coordinate correction strategy to adjust the coordinate information of the ground or satellite data to ensure spatial alignment of the two; when the time step difference between the ground sensor network data and the satellite remote sensing data exceeds the set tolerance range, the interpolation compensation algorithm will be activated, and the missing data will be supplemented by interpolation or extrapolation methods.

[0009] In the present invention, based on the above-mentioned scheme, the spatiotemporal correlation matrix analysis module includes: a spatial correlation calculation unit, which is used to analyze the spatial distribution correlation between the ground sensor matrix and the satellite data matrix to generate a spatial correlation matrix; a time dynamic coupling unit, which is used to extract the time series characteristics of each data source through a sliding window Fourier transform to construct a time dynamic coupling matrix; and a correlation matrix optimization unit, which is used to iteratively optimize the fusion weights of the spatial correlation matrix and the time coupling matrix to generate a spatiotemporal joint correlation matrix.

[0010] In the present invention, based on the above scheme, the generation of the spatial association matrix includes the following steps: converting the ground sensor network data into a spatiotemporal feature matrix according to a geographic coordinate mapping algorithm, the rows of the matrix represent sensor nodes, the columns represent time steps, and each element in the matrix corresponds to an environmental parameter at a certain moment; decomposing the satellite remote sensing data into a grid matrix aligned with the ground sensor network data space, each grid unit represents a fixed geographical area, and the time dimension is synchronized with the ground sensor network number; averaging the time series data of each sensor node to obtain the spatial distribution characteristics of the ground data; averaging the time series data of each satellite grid unit to reflect the spatial distribution characteristics of satellite observations; for ground node i and satellite grid j, the covariance value is the average of the product of the deviation of the time series data of the two from their respective means; and normalizing the covariance matrix to obtain the spatial association matrix.

[0011] In the present invention, based on the above solution, the correlation matrix optimization unit specifically includes: defining a spatial correlation matrix and the time coupling tensor ,in, is the number of sensor nodes, is the number of satellite observation grids, the spatial correlation matrix is the spatial correlation coefficient matrix between ground sensors and satellite data, with the dimension of number of sensors × number of satellite grids; the temporal coupling tensor Include A set of dynamic coupling matrices of frequency components, each matrix dimension is the number of sensors × the number of satellite grids; error calculation is performed to obtain the total optimization target value, and the momentum-accelerated adaptive gradient descent algorithm is used to accelerate convergence; the adaptive gradient descent algorithm includes: the current iteration number is marked as k, and Represent the current spatial weight matrix and temporal weight tensor respectively; and is the learning rate of spatial and temporal weights, and the initial value is set based on experience; for each sensor-grid pair (i, j), the gradient value is the product of the spatial correlation coefficient R(i, j) of the corresponding position and the element of the error matrix (i, j), multiplied by 2; the objective function pair is calculated The partial derivative of :For each frequency component w and sensor-grid pair (i, j), the gradient value is the product of the time coupling coefficient D(i, j, w) and the element of the error matrix (i, j), multiplied by 2; update the spatial weights; fuse the updated weights with the original associated features to generate a spatiotemporal joint matrix.

[0012] In the present invention, based on the above-mentioned scheme, the error calculation includes: element-by-element multiplication of the spatial weight matrix α and the spatial correlation matrix R, that is, multiplication of the elements at the same position of the two matrices to obtain a weighted spatial correlation contribution matrix; for each frequency component w, element-by-element multiplication of the w-th layer slice of the time weight tensor β and the w-th layer slice of the time coupling tensor D, and then adding the results of all frequency components to obtain a weighted time coupling contribution matrix; adding the spatial correlation contribution matrix and the time coupling contribution matrix, subtracting the ground truth matrix, to obtain an error matrix.

[0013] In the present invention, based on the aforementioned scheme, the intelligent collaborative correction module includes: a deviation propagation analysis unit, which is used to calculate the propagation path of the spatial deviation through the Jacobian matrix; a dynamic weight matrix generation unit, which constructs a dynamic credibility weight matrix based on the spatiotemporal joint correlation matrix, wherein the weight value of each data source is dynamically updated according to the real-time verification result; a collaborative correction execution unit, which uses matrix decomposition technology to project high-dimensional data into a unified feature space, and realizes collaborative correction of multi-source data through weighted least squares method.

[0014] In the present invention, based on the above scheme, the deviation propagation analysis unit includes: constructing a two-way deviation propagation equation; initializing the Jacobian matrix from the ground to the satellite based on historical data statistics and the Jacobian matrix from satellite to ground Normalize the spatial correlation matrix; use the least squares method to iteratively optimize to minimize the mean square error between the model prediction deviation and the actual deviation, and obtain the optimized Jacobian matrix; obtain the deviation propagation path based on the optimized Jacobian matrix.

[0015] In the present invention, based on the aforementioned scheme, the collaborative correction execution unit includes: performing low-rank matrix decomposition to obtain a low-dimensional representation; after obtaining the low-dimensional representation, applying the weighted least squares method to perform data correction; the application of the weighted least squares method to perform data correction includes: giving different weights to each data point according to the dynamic credibility weight matrix.

[0016] According to one aspect of the present invention, there is provided an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement a sky-ground intelligent monitoring system with multi-source heterogeneous data fusion as described in the above-mentioned embodiments.

[0017] According to one aspect of the present invention, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the multi-source heterogeneous data fusion sky-ground intelligent monitoring system provided in the above various optional implementations.

[0018] In the technical solution of the present invention, the coordinates and time information of different data sources are compared and aligned, which can eliminate the errors caused by spatiotemporal deviations, ensure that data can be compared and fused in the same reference frame, improve the accuracy and reliability of data integration, avoid monitoring errors caused by data deviations, and ensure the basic accuracy of data fusion; effectively couple the spatial distribution of data sources with temporal dynamics. This stage can provide quantitative support for the association between different data sources by calculating the fusion of spatial correlation and temporal dynamic features. The process of optimizing the spatiotemporal correlation matrix enables the correlation between different data sources to be accurately quantified, thereby avoiding overly simple linear relationship assumptions and improving the complexity and adaptability of data fusion.

[0019] By calculating the deviation propagation path and dynamically adjusting the weights of data sources, the system can accurately propagate and eliminate deviations between multi-source data. This not only ensures that multi-source data can be integrated with the same accuracy, but also improves the system's dynamic response capability to the credibility of different data sources. The correction process effectively reduces errors caused by data deviations and inconsistencies, and improves the accuracy and reliability of monitoring results.

[0020] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present invention, and together with the specification are used to explain the principles of the present invention. Obviously, the accompanying drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0022] Figure 1 The flowchart of a sky-ground intelligent monitoring system for multi-source heterogeneous data fusion in one embodiment of the present invention is schematically shown.

[0023] Figure 2 The flowchart of a sky-ground intelligent monitoring solution for multi-source heterogeneous data fusion in one embodiment of the present invention is schematically shown.

[0024] Figure 3 The block diagram of a spatiotemporal correlation matrix analysis module of a sky-ground intelligent monitoring system for multi-source heterogeneous data fusion in one embodiment of the present invention is schematically shown.

[0025] Figure 4 A block diagram of an intelligent collaborative correction module of a sky-to-ground intelligent monitoring system for multi-source heterogeneous data fusion in one embodiment of the present invention is schematically shown. DETAILED DESCRIPTION

[0026] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present invention will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0027] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solution of the present invention can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present invention.

[0028] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0029] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0030] With the development of remote sensing technology and ground sensor technology, the integrated sky-ground intelligent monitoring system has gradually become an important application means in the fields of environmental monitoring, weather forecasting, disaster warning, etc. Especially in large-scale and multi-scale environmental monitoring tasks, the combination of ground sensors and satellite remote sensing data can provide more comprehensive and accurate information. At present, ground sensor networks are mainly used to provide high-precision, real-time local environmental data, covering a variety of environmental parameters such as temperature, humidity, and air pressure; while satellite remote sensing technology can cover a large area and obtain important environmental information such as atmospheric composition, cloud cover, and surface temperature. With the continuous advancement of sensor data acquisition technology and satellite remote sensing technology, multi-source heterogeneous data fusion technology has gradually become a key technology to improve monitoring accuracy and timeliness. This technology makes up for the limitations of a single data source by uniformly modeling and fusing the spatiotemporal characteristics of different data sources, providing a more comprehensive and efficient monitoring method.

[0031] However, existing technologies still face many challenges in practical applications. For example, the spatiotemporal inconsistency between ground sensors and satellite remote sensing data is an issue that cannot be ignored. Ground sensors are limited in spatial distribution and have a high data update frequency, while satellite remote sensing data is subject to satellite orbits, observation angles and imaging frequencies, resulting in large differences in temporal and spatial resolution between the two. Existing technologies often use static correction or linear interpolation methods to correct spatiotemporal deviations, but this method often ignores the complex nonlinear relationship between data, resulting in insufficient correction effect; existing multi-source data fusion methods usually assume that each data source has the same credibility or fusion weight, and fail to fully consider the reliability differences of different data sources, especially in abnormal weather conditions or high-density monitoring areas, where large fusion errors may occur; furthermore, spatiotemporal correlation analysis in existing technologies mostly relies on simple spatial distance calculations or time synchronization strategies, which makes it difficult to effectively extract deep spatiotemporal features between data, resulting in the construction of spatiotemporal correlation models that are insufficient to fully reflect the dynamic coupling relationship between data sources, affecting the accuracy of the final data fusion.

[0032] Therefore, in response to the above problems, the technical concept of the present invention is to achieve precise spatiotemporal alignment of ground sensor network data and satellite remote sensing data through a spatiotemporal dynamic calibration module, using geographic coordinate mapping and timestamp interpolation algorithms, and eliminate spatiotemporal deviations between different data sources. In particular, the two-way verification control unit further improves the accuracy of spatiotemporal correction through the mutual verification mechanism of ground sensors and satellite remote sensing data; through the spatiotemporal correlation matrix analysis module, the covariance matrix analysis and sliding window Fourier transform and other methods are used to comprehensively extract the spatial correlation and time dynamic coupling characteristics between data sources, and the fusion weights of the correlation matrix are optimized through an adaptive learning algorithm to improve the accuracy and reliability of data fusion; in addition, the intelligent collaborative correction module generates a dynamic weight matrix based on the spatiotemporal joint correlation matrix, combines matrix decomposition technology and weighted least squares method, and realizes the collaborative correction of multi-source data, thereby ensuring the accuracy and credibility of the final fusion result.

[0033] Figure 1 The block diagram of a sky-ground intelligent monitoring system for multi-source heterogeneous data fusion according to an embodiment of the present invention. Figure 2 FIG. 1 is a data flow diagram of a sky-ground intelligent monitoring system for multi-source heterogeneous data fusion according to an embodiment of the present invention. Figure 1 and Figure 2 As shown, a multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to an embodiment of the present invention includes: The data acquisition module 100 is used to obtain ground sensor network data and satellite remote sensing data; the spatiotemporal dynamic calibration module 200 is used to establish a spatiotemporal reference framework for multi-source data, and to perform spatiotemporal calibration on the ground sensor network data and satellite remote sensing data obtained by the data acquisition module to ensure the consistency of the ground sensor network data and the satellite remote sensing data in space and time; the spatiotemporal correlation matrix analysis module 300 is used to construct a multi-dimensional spatiotemporal correlation model, further analyze the data calibrated by the spatiotemporal dynamic calibration module, optimize the fusion weights of the spatial and temporal correlation matrices, and improve the spatiotemporal consistency of the data; the intelligent collaborative correction module 400 is used to implement collaborative correction of multi-source data, optimize the spatiotemporal joint correlation matrix generated by the spatiotemporal correlation matrix analysis module, eliminate spatial and temporal deviations, and ensure the accuracy of the data fusion results; the data fusion and calibration output module 500 is used to perform the final fusion and calibration of multi-source heterogeneous data and provide data output.

[0034] In the embodiment of the present invention, the data acquisition module is used to acquire ground sensor network data and satellite remote sensing data, ensuring synchronous acquisition and spatiotemporal alignment of multi-source heterogeneous data for subsequent analysis and optimization.

[0035] Specifically, the network data of temperature, humidity, air pressure, etc. collected by distributed sensors are constructed into a ground sensor matrix, in which the spatial dimension maps the sensor geographic coordinates, and the time dimension records the time series changes of parameters. By organizing the sensor network data in the form of geographic coordinates and time series, the data of environmental parameters not only have spatial correlation, but also reflect their trend over time. This method provides an effective data structure for spatiotemporal correlation analysis and subsequent processing; the satellite remote sensing data is decomposed into an atmospheric composition matrix, a cloud cover matrix, and a surface temperature matrix, and each matrix dimension is spatially aligned with the ground sensor matrix. By parsing the satellite remote sensing data into different matrices, each data source can independently provide specific environmental information that matches the ground sensor data. In addition, maintaining spatial alignment is the key to ensuring that ground and satellite data are processed in the same spatial coordinate system, which helps to improve the accuracy of data fusion.

[0036] It should be understood that the ground sensor network and satellite remote sensing data in the data acquisition module 100 are organized and analyzed by means of a spatiotemporal feature matrix, which ensures the spatial and temporal consistency between multi-source data. First, the ground sensor network data can be spatially mapped according to geographic coordinates through the construction of a ground sensor matrix, and environmental parameters such as temperature, humidity, and air pressure can be visualized in the spatial dimension, laying the foundation for subsequent data analysis. By recording the time dimension, the changing trend of environmental parameters over time can be clearly presented.

[0037] In terms of remote sensing data analysis, satellite remote sensing data is deconstructed into multiple matrices, including the atmospheric composition matrix, cloud cover matrix and surface temperature matrix. These matrices provide independent data dimensions for analyzing changes in different atmospheric layers and surface temperature. In this way, the relationship and influence between the ground and air environments can be more intuitively reflected, providing an accurate data basis for spatiotemporal correlation analysis and subsequent data fusion.

[0038] Specifically, remote sensing data preprocessing technology (such as principal component analysis technology such as PCA, ICA, SVD, etc.) is used to decompose and reduce the dimension of the remote sensing data to obtain the corresponding matrix.

[0039] For example, suppose the original remote sensing data has multiple bands, such as infrared, visible light, near infrared, etc. Each pixel observation has a corresponding value in different bands. Organize these observations into a matrix, such as ,in, For the raw data of each band, is the number of pixels; calculate the covariance matrix C of the matrix to describe the relationship between different bands; perform eigenvalue decomposition on the covariance matrix C to obtain eigenvalues ​​and eigenvectors: , in, is the characteristic value, is the eigenvector; sort the eigenvalues ​​and select the eigenvectors corresponding to the largest eigenvalues. These eigenvectors constitute the principal components.

[0040] The original data is projected onto the selected principal component to obtain the reduced-dimensional matrix Z, where the moment Z after principal component analysis is a low-dimensional matrix that represents the result of the reduced-dimensional information of different bands. Each element represents the principal component value of a pixel, and these data must be mapped back to the geographic coordinate system so that they can be connected to the actual geographic data. Remote sensing data are usually aligned using geographic references (such as longitude and latitude coordinates or projected coordinate systems). It can be seen that this technical solution of organizing and parsing multi-source data in the form of a spatiotemporal matrix avoids problems such as data inconsistency that may occur in traditional methods, so that subsequent steps such as spatiotemporal dynamic calibration and association analysis can be carried out under a unified data framework, thereby improving the accuracy and efficiency of the entire system.

[0041] The present invention solves the problem of spatiotemporal alignment of multi-source data by proposing a method for constructing a spatiotemporal feature matrix. Specifically, the ground sensor network data can effectively map environmental parameters at different locations and times to a unified coordinate system through the construction of a spatiotemporal feature matrix, thus avoiding spatiotemporal mismatches between data. At the same time, the decomposition and spatiotemporal alignment of satellite remote sensing data further enhance the comparability and fusion between different data sources. Through these optimization measures, the problem of low data fusion accuracy in the prior art is solved, laying a solid foundation for subsequent data analysis and optimization.

[0042] In the embodiment of the present invention, the spatiotemporal dynamic calibration module 200 is mainly used to effectively calibrate the ground sensor matrix, atmospheric composition matrix, cloud cover matrix, and surface temperature matrix generated by the data acquisition module. This module ensures the consistency of ground sensor network data and satellite remote sensing data in space and time dimensions, and provides a basis for subsequent data analysis, correction, and optimization. The specific implementation steps of the spatiotemporal calibration module 200 are as follows: The spatiotemporal dynamic calibration module 200 includes: To ensure that ground data and satellite data are spatially aligned, the spatiotemporal calibration module matches the spatial coordinates (i.e., geographic location) of the ground sensor network data with the spatial coordinates of the satellite remote sensing data. By using technologies such as geographic information systems (GIS), ground and satellite data of different resolutions can be mapped to the same spatial coordinate system. In this way, each location in the ground sensing matrix is ​​aligned with the data of the corresponding location in the atmospheric composition matrix, cloud cover matrix, and surface temperature matrix.

[0043] Specifically, the spatial coordinates of the ground sensor network data are aligned with the spatial coordinates of the satellite remote sensing data through a geographic coordinate mapping algorithm.

[0044] For example, by using Geographic Information System (GIS) technology, accurate mapping between ground and satellite data of different spatial resolutions can be achieved.

[0045] At the same time, since the time of acquisition of ground sensor network data and satellite remote sensing data is usually inconsistent, a timestamp interpolation algorithm is needed to align them. The timestamp interpolation algorithm ensures the synchronization of the timestamps of ground sensor network data and satellite remote sensing data. If the ground sensor and satellite remote sensing data are collected at different times, the interpolation algorithm can supplement or interpolate the inconsistent time data according to the sampling time point, thereby eliminating the impact caused by the time difference. This operation can ensure that all data sources are consistent on the same time scale and avoid analysis errors caused by time asynchrony.

[0046] The spatiotemporal dynamic calibration module 200 also introduces a two-way verification control mechanism for mutual verification of ground sensor network data and satellite remote sensing data. When the system detects that the spatial coordinate deviation exceeds a preset threshold, it will automatically trigger a coordinate correction strategy to adjust the coordinate information of the ground or satellite data to ensure that the two are spatially aligned.

[0047] Specifically, the coordinate correction strategy calculates the error and compares the spatial position difference between the ground sensor and satellite data. When the coordinate error is detected to exceed the set tolerance range, the error is corrected based on the known geographic information. This process can eliminate the spatial mismatch caused by factors such as insufficient position accuracy of ground sensors and deviations in satellite data, thereby improving the accuracy of data fusion.

[0048] The error calculation includes: using the square difference of geographic coordinates to calculate the error; the error correction includes: correcting the error through an optimization algorithm (such as the least squares method or the interpolation method).

[0049] In addition to spatial alignment, the interpolation compensation algorithm for time step differences is also applied in this module. When the time step difference between the ground sensor network data and the satellite remote sensing data exceeds the set tolerance range, the interpolation compensation algorithm will be activated, and the missing data will be supplemented by interpolation or extrapolation methods to ensure that all data have the same time resolution. Among them, the interpolation compensation algorithm includes linear interpolation, spline interpolation, Lagrange interpolation, etc. The present invention uses linear interpolation to process time differences. This measure can eliminate the impact of different time steps on data accuracy, so that all data can be analyzed and fused on a unified time scale.

[0050] It should be noted that traditional spatiotemporal calibration technology usually only focuses on the alignment of spatial coordinates, or simple time interpolation methods, but the present invention achieves precise synchronization in space and time by introducing a space-time dual alignment and correction mechanism, thereby improving the accuracy of data fusion and spatiotemporal analysis.

[0051] In addition, the two-way verification control mechanism not only ensures the spatial consistency of ground data and satellite remote sensing data, but also takes into account the data deviation that may be caused by external environmental factors (such as satellite position, sensor error, etc.). Through continuous two-way verification and dynamic adjustment, the system can correct the data deviation in real time according to the actual situation, avoiding the error accumulation problem that is not considered in traditional methods.

[0052] In the embodiment of the present invention, the spatiotemporal correlation matrix analysis module 300 is used to further optimize the data generated by the spatiotemporal dynamic calibration module 200 and analyze the calibrated data through a multi-dimensional spatiotemporal correlation model.

[0053] like Figure 3 As shown, Figure 3 1 is a block diagram of a spatiotemporal correlation matrix analysis module in a multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to an embodiment of the present invention. The following is the detailed content of the spatiotemporal correlation matrix analysis module 300: The spatiotemporal correlation matrix analysis module 300 includes a spatial correlation calculation unit 301, which analyzes the spatial distribution correlation between the ground sensor matrix and the satellite data matrix based on the covariance matrix to generate a spatial correlation matrix; a time dynamic coupling unit 302, which is used to extract the time series characteristics of each data source through a sliding window Fourier transform to construct a time dynamic coupling matrix; and a correlation matrix optimization unit 303, which is used to iteratively optimize the fusion weights of the spatial correlation matrix and the time coupling matrix to generate a spatiotemporal joint correlation matrix.

[0054] Specifically, the spatial correlation calculation unit 301 is used to analyze the spatial distribution correlation of the ground sensor matrix and the satellite remote sensing data matrix by calculating the covariance matrix, and then generate a spatial correlation matrix. The spatial correlation matrix evaluates the spatial similarity and correlation of the two types of data. The specific steps are as follows: First, the ground sensor network data is converted into a spatiotemporal feature matrix according to the geographic coordinate mapping algorithm. The rows of the matrix represent sensor nodes (spatial dimension), the columns represent time steps, and each element in the matrix corresponds to an environmental parameter (such as temperature, humidity, etc.) at a certain moment. In this way, each row represents the change of a sensor node in the time dimension, forming a spatiotemporal representation of the ground sensor network data.

[0055] Satellite remote sensing data (such as surface temperature, cloud cover, etc.) are decomposed into a grid matrix that is spatially aligned with the ground sensor network data. Each grid cell represents a fixed geographic area (e.g., 1 km×1 km), and the time dimension is synchronized with the ground sensor data. This grid representation ensures that the spatial information of the satellite remote sensing data can be accurately matched with the ground sensor network data, facilitating the subsequent calculation of spatial correlation.

[0056] To generate the spatial correlation matrix, for each time window (e.g., 24 hours or other suitable time period), the covariance values ​​between the ground sensor nodes and the satellite grid cells are calculated. The specific steps are: The time series data of each sensor node is averaged to obtain the spatial distribution characteristics of the ground data; the time series data of each satellite grid unit is averaged to reflect the spatial distribution characteristics of satellite observations; for ground node i and satellite grid j, the covariance value is the average of the product of the deviations of the time series data of the two from their respective means; the covariance matrix is ​​normalized to obtain the spatial correlation matrix, and the numerical range is mapped to [0,1]. The larger the value, the stronger the spatial correlation.

[0057] Through the above covariance calculation and normalization process, the spatial correlation matrix finally obtained can effectively describe the correlation between ground sensor network data and satellite remote sensing data in the spatial dimension. Each element of the matrix represents the spatial correlation between the corresponding sensor node and the satellite grid unit.

[0058] Specifically, the time dynamic coupling unit 302 is used to analyze the dynamic correlation pattern of ground and satellite data in the time dimension and construct a coupling matrix reflecting the time series evolution law. The specific implementation steps are as follows: First, the time series of ground and satellite data are divided into sliding windows of fixed length (such as a 7-day window with a step length of 1 day). Fourier transform is applied to the data in each window to extract the main frequency components (such as daily changes and seasonal changes). The frequency characteristics of ground data are represented as the spectral energy distribution of each sensor node, while those of satellite data are represented as the spectral energy distribution of each grid unit.

[0059] For each frequency component (such as daily variation frequency), calculate the spectral similarity between ground and satellite data: The similarity is defined as the cosine similarity of the spectral energy distribution of the two sequences. The value range is [-1,1], and a positive value indicates that the change trend is consistent. The window is sliding along the time axis to generate a dynamic coupling intensity matrix. The rows and columns of the matrix correspond to the ground nodes and the satellite grids respectively, and the element value is the similarity of the current time window.

[0060] Higher weights are given to low-frequency components (such as seasonal changes) because they reflect long-term trends. High-frequency components (such as hourly fluctuations) are subjected to noise reduction filtering to avoid short-term interference. Through the time sliding mechanism, the memory of historical coupling patterns is retained to form a time-evolving coupling tensor.

[0061] For example, in terms of the daily frequency component, the coupling strength between ground sensors and satellite data in urban areas can reach above 0.9, while in suburban areas it may drop to 0.6 due to the influence of local microclimate, suggesting that the fusion weight of the time dimension needs to be reduced in suburban areas. Before a rainstorm event, the coupling strength between ground humidity sensors and satellite cloud data will increase significantly (e.g., from 0.4 to 0.8), and the system can trigger the abnormal weather monitoring mode in advance.

[0062] Specifically, the correlation matrix optimization unit is used to dynamically optimize the fusion weights of spatial correlation and temporal coupling, generate a spatiotemporal joint correlation matrix, and provide high-precision input for data correction and compensation. The specific implementation steps are as follows: Defining the spatial incidence matrix and the time coupling tensor ,in, is the number of sensor nodes, is the number of satellite observation grids, the spatial correlation matrix is the spatial correlation coefficient matrix between ground sensors and satellite data, with the dimension of number of sensors × number of satellite grids; the temporal coupling tensor Include The dynamic coupling matrix of frequency components is a set, and the dimension of each matrix is ​​the number of sensors × the number of satellite grids. The initial value prior knowledge (such as sensor accuracy, satellite resolution) is set.

[0063] Exemplarily, the spatial correlation matrix sets initial values ​​based on sensor accuracy (such as calibration certificate) and satellite resolution (such as grid size), with the weight of the grid where the high-precision sensor is located set to 0.8 and the low-precision area set to 0.5; the time coupling tensor is dynamically initialized according to the data update frequency, with the weight of real-time satellite data (such as updated every hour) set to 0.7 and the ground sensor (updated every 10 minutes) set to 0.9.

[0064] Considering the complexity of spatiotemporal data and the high requirements for accuracy, in order to solve the existence of errors and inaccuracies, error calculation is first required. This step is essential because the deviation can be effectively measured by quantifying the error. The specific operations are as follows: The spatial weight matrix α is multiplied element by element with the spatial correlation matrix R, that is, the elements in the same position of the two matrices are multiplied to obtain the weighted spatial correlation contribution matrix; for each frequency component w, the w-th slice of the time weight tensor β is multiplied element by element with the w-th slice of the time coupling tensor D, and then the results of all frequency components are added to obtain the weighted time coupling contribution matrix; the above spatial correlation contribution matrix is ​​added to the time coupling contribution matrix, and then the ground truth matrix is ​​subtracted to obtain the error matrix. The Frobenius norm square of the error matrix (that is, the sum of the squares of all elements of the matrix) is calculated, and the errors of all time steps are summed to obtain the total optimization target value.

[0065] Among them, the ground truth matrix is ​​high-precision weather station data.

[0066] It should be noted that the optimal spatiotemporal weights are found through gradual weighting and optimization to minimize the prediction error. By calculating the total optimization target value, the error situation of the entire spatiotemporal data model can be reflected in a centralized manner.

[0067] Once the total optimization target value is calculated, the parameters can be adjusted through the adaptive optimization algorithm. Specifically, the adaptive gradient descent algorithm with momentum acceleration is used to accelerate convergence. The algorithm adjusts the learning rate of spatial and temporal weights and uses the momentum factor to ensure the validity of the historical update direction, thereby avoiding the overfitting or underfitting problem caused by a single update. The specific steps are: Use the adaptive gradient descent algorithm with momentum acceleration to adjust the weights. The specific process is as follows: The current iteration number is marked as k, and Represent the current spatial weight matrix and temporal weight tensor respectively; and is the learning rate of space and time weights, and the initial value is set according to experience (such as =0.01, =0.005), γ is the momentum factor (usually set to 0.9), which is used to preserve the historical update direction.

[0068] Calculate the objective function The partial derivative of The specific method is: for each sensor-grid pair (i, j), the gradient value is the product of the spatial correlation coefficient R(i, j) at the corresponding position and the element of the error matrix (i, j), multiplied by 2 (from the derivative of the squared error).

[0069] Calculate the objective function The partial derivative of : For each frequency component w and sensor-grid pair (i,j), the gradient value is the product of the time coupling coefficient D(i,j,w) and the element of the error matrix (i,j), multiplied by 2.

[0070] The spatial weight update is expressed as: , in, and is the dynamically adjusted learning rate, is the momentum factor.

[0071] Fusion of updated weights and original correlation features to generate a spatiotemporal joint matrix , is the time step, and the operation is as follows: The optimized spatial weight matrix and time weight tensor are weightedly fused: the spatial correlation coefficient matrix is ​​multiplied element by element with the spatial weight matrix to highlight the high confidence area; the coupling matrix of each frequency component is multiplied with the corresponding time weight to strengthen the contribution of long-term trends; the spatial and temporal fusion results are superimposed along the time axis to generate a spatiotemporal joint correlation matrix, in which each element of the matrix comprehensively reflects the data reliability of a specific location at a specific moment.

[0072] like Figure 4 As shown, Figure 4 4 is a block diagram of an intelligent collaborative correction module in a multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to an embodiment of the present invention. The following is the details of the intelligent collaborative correction module 400: In an embodiment of the present invention, the intelligent collaborative correction module 400 includes a deviation propagation analysis unit 401, which is used to calculate the propagation path of the spatial deviation through the Jacobian matrix; a dynamic weight matrix generation unit 402, which constructs a dynamic credibility weight matrix based on the spatiotemporal joint association matrix, wherein the weight value of each data source is dynamically updated according to the real-time verification result; and a collaborative correction execution unit 403, which uses matrix decomposition technology to project high-dimensional data into a unified feature space, and realizes collaborative correction of multi-source data through weighted least squares method.

[0073] Specifically, the deviation propagation analysis unit 401 specifically includes: Assuming that the ground sensor bias will affect the satellite data through spatial correlation, and vice versa, a two-way bias propagation equation is constructed: , in, is the satellite data bias matrix, is the ground sensor bias vector, is the Jacobian matrix from ground to satellite, is the Jacobian matrix from satellite to ground, and In order to obtain the optimized Jacobian matrix, the least squares method is used to iteratively optimize the process to minimize the mean square error between the predicted deviation and the actual deviation. Specifically: Based on historical data statistics, initialize and Normalize the spatial correlation matrix; use the least squares method to iteratively optimize the model prediction deviation Deviation from actual The mean square error is minimized, and the optimized Jacobian matrix is ​​obtained, and finally the prediction result with the minimum mean square error is achieved.

[0074] Similar to the Jacobian matrix from ground to satellite, the Jacobian matrix from satellite to ground can also be optimized by the least squares method to minimize the mean square error between the model's predicted deviation and the actual deviation.

[0075] Through the propagation formula, the Jacobian matrix not only provides the propagation path of the deviation from one domain to another, but also reveals how and in which direction the deviation propagates in space.

[0076] Through data from multiple moments, we observe how the Jacobian matrix is ​​continuously updated in different time steps, thereby affecting the propagation of the deviation; through the Jacobian matrix of different positions (such as the relative position between ground sensors and satellites), we show how the deviation propagates from one spatial position to another.

[0077] In the embodiment of the present invention, the dynamic weight matrix generating unit 402 includes: generating a dynamic credibility weight matrix based on the spatiotemporal joint correlation matrix and the real-time verification result to quantify the reliability of each data source in the spatial and temporal dimensions, and the specific operations are as follows: Based on the spatial and temporal correlations in the spatiotemporal joint correlation matrix, a credibility model for each data source is established. Specifically, by analyzing the weights of each item in the matrix, the system evaluates the influence of each data source at a specific spatial location and time step.

[0078] For ground sensor network data, the model takes into account the spatial distribution characteristics and temporal variability of the data; for satellite remote sensing data, the model mainly considers the data's resolution, coverage, and the strength of its association with ground sensor network data.

[0079] The real-time verification results come from the system's error analysis of the current moment, such as the comparison of ground sensor network data and satellite remote sensing data. If the error of a data source is large, or the verification results show that its reliability is low, the system will reduce the weight of the data source in the credibility weight matrix.

[0080] On the contrary, when the actual error of a data source is small or its spatial and temporal matching degree is high, the system will increase the weight of the data source accordingly.

[0081] As real-time data changes, the dynamic credibility weight matrix is ​​dynamically updated based on the real-time verification results. Each time it is updated, the system re-evaluates the reliability of all data sources and adjusts their weights based on the performance of the current data source in space and time.

[0082] Through an iterative approach, it is ensured that the weight of the data source during the correction process can accurately reflect its true reliability in the time and space dimensions, thereby affecting the subsequent correction process.

[0083] Through the above dynamic adjustment steps, a new dynamic credibility weight matrix is ​​finally generated. Each element in the matrix represents the reliability of a data source at a specific time and space point, and the value range is usually between [0, 1]. The higher the value, the stronger the reliability of the data source.

[0084] In the embodiment of the present invention, the collaborative correction execution unit 403 includes: The collaborative correction execution unit first receives the dynamic credibility weight matrix from the dynamic weight matrix generation unit 402, and the dynamic credibility weight matrix provides weight values ​​for the subsequent weighted least square method.

[0085] It should be noted that matrix decomposition technology usually involves decomposing a high-dimensional data matrix into multiple low-dimensional matrices to facilitate the extraction of important information from a more concise feature space. In the present invention, the goal of decomposition is to project the ground sensor network data matrix and the satellite remote sensing data matrix into a common feature space so that they can be compared and corrected in a unified coordinate system.

[0086] Specifically, through low-rank matrix factorization, the goal is to find a low-dimensional representation that contains the most important features of ground and satellite data to facilitate subsequent correction.

[0087] After obtaining the low-dimensional representation, the weighted least squares method is applied to correct the data. This method minimizes the weighted error function so that the corrected data is as close to the actual value as possible.

[0088] It can be seen that the basic idea of ​​the weighted least squares method is to give different weights to each data point according to the dynamic credibility weight matrix. Data points with higher weight values ​​will have a greater impact on the final result during the correction process.

[0089] Through the weighted least squares method, the correction result can retain the accuracy of the high-credibility data source to the greatest extent, while making appropriate corrections to the low-credibility data source, ultimately achieving the best effect of data fusion.

[0090] Through the weighted least squares optimization process, the final output of the correction data will be a highly consistent result after multi-source data fusion. These results not only combine the advantages of ground sensors and satellite remote sensing data, but also ensure that the influence of each data source is reasonably controlled through dynamic weight adjustment.

[0091] The data fusion and calibration output module 500 is used to finally fuse and calibrate the multi-source heterogeneous data processed by the above modules to provide consistent and accurate data output.

[0092] Specifically, the fused data is output in a unified format to provide a standardized data interface for subsequent applications.

[0093] This interface can be connected to external application systems to achieve seamless data flow.

[0094] In addition, the fusion results can be fed back to the system's monitoring module in real time. If there is a problem with the data quality, the system will automatically trigger the calibration and data update mechanism to ensure the long-term stability and accuracy of the data.

[0095] By simultaneously acquiring environmental data from different sources from ground sensor networks and satellite remote sensing systems, the foundation for subsequent processing is laid. The integration of this multi-source data not only improves the comprehensiveness of monitoring, but also enables the monitoring system to adapt to different environmental monitoring needs by providing data perspectives across time and space.

[0096] Next, the spatial and temporal consistency of ground sensor data and satellite remote sensing data is ensured through the spatiotemporal dynamic calibration step. Comparing and aligning the coordinate and time information of different data sources can eliminate the errors caused by spatiotemporal deviations and ensure that the data can be compared and fused in the same reference frame. The beneficial effect of this step is to improve the accuracy and reliability of data integration, avoid monitoring errors caused by data deviations, and ensure the basic accuracy of data fusion.

[0097] On this basis, the technical solution further effectively couples the spatial distribution of data sources with temporal dynamics through the construction and optimization of the spatiotemporal correlation matrix. This stage can provide quantitative support for the association between different data sources by calculating the fusion of spatial correlation and temporal dynamic features. The process of optimizing the spatiotemporal correlation matrix enables the correlation between different data sources to be accurately quantified, thereby avoiding overly simplistic linear relationship assumptions and improving the complexity and adaptability of data fusion.

[0098] The introduction of the intelligent collaborative correction module further improves the system's ability to correct data deviations. By calculating the deviation propagation path and dynamically adjusting the weight of the data source, the system can accurately propagate and eliminate deviations between multi-source data. This not only ensures that multi-source data can be fused with the same accuracy, but also improves the system's dynamic response ability to the credibility of different data sources. The correction process effectively reduces errors caused by data deviations and inconsistencies, and improves the accuracy and reliability of monitoring results.

[0099] Finally, the data processed by the output module of data fusion and calibration can provide unified and accurate monitoring results. These results are not only highly accurate, but also can adapt to environmental changes in real time and provide more accurate monitoring information. With the cooperation of various steps, the entire technical solution significantly improves the intelligence level and accuracy of data processing, ensures the seamless integration of multi-source heterogeneous data in spatial and temporal dimensions, and greatly improves the comprehensive performance of the sky-ground intelligent monitoring system.

[0100] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable computer program. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0101] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0102] The units involved in the embodiments of the present invention may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, limit the units themselves.

[0103] According to one aspect of the present invention, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above-mentioned various optional implementations.

[0104] As another aspect, the present invention further provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements a multi-source heterogeneous data fusion sky-ground intelligent monitoring system described in the above embodiment.

[0105] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0106] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation of the present invention.

[0107] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the embodiments disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art that are not disclosed in the present invention.

[0108] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A multi-source heterogeneous data fusion sky-ground intelligent monitoring system, characterized in that: include: Data acquisition module, used to obtain ground sensor network data and satellite remote sensing data; A spatiotemporal dynamic calibration module is used to perform spatiotemporal calibration on the acquired ground sensor network data and satellite remote sensing data to ensure the consistency of the ground sensor network data and satellite remote sensing data in space and time; The spatiotemporal correlation matrix analysis module is used to construct a multi-dimensional spatiotemporal correlation model, further analyze the data calibrated by the spatiotemporal dynamic calibration module, optimize the fusion weights of the spatial and temporal correlation matrices, and generate a spatiotemporal joint correlation matrix; An intelligent collaborative correction module, used to implement collaborative correction of multi-source data, optimize the spatiotemporal joint correlation matrix, and eliminate spatial and temporal deviations; The data fusion and calibration output module is used to perform the final fusion and calibration of the processed multi-source heterogeneous data and provide data output.

2. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 1 is characterized in that: The data acquisition module comprises: The network data collected by distributed sensors is constructed into a ground sensor matrix, where the spatial dimension maps the sensor geographic coordinates and the time dimension records the time series changes of parameters; The satellite remote sensing data are decomposed into atmospheric composition matrix, cloud cover matrix and land surface temperature matrix.

3. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 1 is characterized in that: The spatiotemporal dynamic calibration module comprises: Matching the spatial coordinates of the ground sensor network data with the spatial coordinates of the satellite remote sensing data, aligning each position in the ground sensor matrix with the data of the corresponding position in the atmospheric composition matrix, cloud cover matrix and land surface temperature matrix; Mutual verification of ground sensor network data and satellite remote sensing data. When the spatial coordinate deviation exceeds the preset threshold, the coordinate correction strategy is automatically triggered to adjust the coordinate information of the ground or satellite data to ensure spatial alignment of the two. When the time step difference between the ground sensor network data and the satellite remote sensing data exceeds the set tolerance range, the interpolation compensation algorithm will be activated and the missing data will be supplemented by interpolation or extrapolation methods.

4. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 1 is characterized in that: The spatiotemporal correlation matrix analysis module comprises: A spatial correlation calculation unit is used to analyze the spatial distribution correlation between the ground sensor matrix and the satellite data matrix and generate a spatial correlation matrix; A time dynamic coupling unit is used to extract the time series characteristics of each data source through sliding window Fourier transform and construct a time dynamic coupling matrix; The correlation matrix optimization unit is used to iteratively optimize the fusion weights of the spatial correlation matrix and the temporal coupling matrix to generate a spatiotemporal joint correlation matrix.

5. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 1 is characterized in that: The generation of the spatial incidence matrix comprises the following steps: The ground sensor network data is converted into a spatiotemporal feature matrix according to the geographic coordinate mapping algorithm. The rows of the matrix represent sensor nodes, the columns represent time steps, and each element in the matrix corresponds to the environmental parameters at a certain moment. Decompose satellite remote sensing data into a gridded matrix that is spatially aligned with ground sensor network data. Each grid cell represents a fixed geographic area, and the time dimension is synchronized with the ground sensor network data. The time series data of each sensor node is averaged to obtain the spatial distribution characteristics of ground data; the time series data of each satellite grid unit is averaged to reflect the spatial distribution characteristics of satellite observations; For ground node i and satellite grid j, the covariance value is the average of the product of the deviations of their time series data from their respective means; The covariance matrix is ​​normalized to obtain the spatial correlation matrix.

6. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 5 is characterized in that: The association matrix optimization unit specifically includes: Defining the spatial incidence matrix and the time coupling tensor ,in, is the number of sensor nodes, is the number of satellite observation grids, the spatial correlation matrix is the spatial correlation coefficient matrix between ground sensors and satellite data, with the dimension of number of sensors × number of satellite grids; Temporal coupling tensor Include A set of dynamic coupling matrices of frequency components, each of which has the dimension of the number of sensors × the number of satellite grids; Perform error calculations to obtain the total optimization target value, and use the momentum-accelerated adaptive gradient descent algorithm to accelerate convergence; The adaptive gradient descent algorithm includes: The current iteration number is marked as k, and Represent the current spatial weight matrix and temporal weight tensor respectively; and is the learning rate of spatial and temporal weights, and the initial value is set based on experience; For each sensor-grid pair (i, j), the gradient value is the product of the spatial correlation coefficient R(i, j) of the corresponding position and the element of the error matrix (i, j), multiplied by 2; Calculate the objective function The partial derivative of : For each frequency component w and sensor-grid pair (i, j), the gradient value is the product of the time coupling coefficient D(i, j, w) and the element of the error matrix (i, j), multiplied by 2; Update spatial weights; The updated weights are fused with the original correlation features to generate a spatiotemporal joint matrix.

7. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 6 is characterized in that: The error calculation includes: Multiply the spatial weight matrix α and the spatial correlation matrix R element by element, that is, multiply the elements in the same position of the two matrices to obtain the weighted spatial correlation contribution matrix; For each frequency component w, multiply the w-th slice of the time weight tensor β by the w-th slice of the time coupling tensor D element by element, and then add the results of all frequency components to obtain the weighted time coupling contribution matrix; The spatial correlation contribution matrix is ​​added to the temporal coupling contribution matrix, and the ground truth matrix is ​​subtracted to obtain an error matrix.

8. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 1 is characterized in that: The intelligent collaborative correction module includes: A deviation propagation analysis unit, used for calculating the propagation path of the spatial deviation through the Jacobian matrix; A dynamic weight matrix generation unit constructs a dynamic credibility weight matrix based on the spatiotemporal joint association matrix, wherein the weight value of each data source is dynamically updated according to the real-time verification result; The collaborative correction execution unit uses matrix decomposition technology to project high-dimensional data into a unified feature space and realizes collaborative correction of multi-source data through weighted least squares method.

9. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 8 is characterized in that: The deviation propagation analysis unit comprises: Construct a two-way deviation propagation equation; Initialize the Jacobian matrix from ground to satellite based on historical data statistics and the Jacobian matrix from satellite to ground is the normalization of the spatial correlation matrix; The least square method is used for iterative optimization to minimize the mean square error between the predicted deviation and the actual deviation, and the optimized Jacobian matrix is ​​obtained; The deviation propagation path is obtained according to the optimized Jacobian matrix.

10. The multi-source heterogeneous data fusion sky-ground intelligent monitoring system according to claim 9 is characterized in that: The collaborative correction execution unit comprises: Perform low-rank matrix decomposition to obtain a low-dimensional representation; After obtaining the low-dimensional representation, the weighted least squares method is applied for data correction; The method of applying weighted least squares method to perform data correction includes: According to the dynamic credibility weight matrix, different weights are given to each data point.

Images