News Event Spatial Verification and Analysis Method and System Based on Satellite Remote Sensing Images

Abstract

The present invention provides a method and system for spatially validating and analyzing news events based on satellite remote sensing images, which relates to the field of satellite technology. The method includes extracting spatio-temporal information from news texts through natural language processing, obtaining remote sensing images based on multi-source satellite collaborative observation, using a deep learning model to extract ground object elements and change features, constructing an event feature association graph for graph reasoning, and combining multi-dimensional cross-validation and transfer learning to calculate the credibility of events, so as to realize the intelligent validation of the authenticity of news events. The present invention can effectively improve the automation degree and accuracy of news event validation and reduce the manual validation cost.

Description

[0001] News Event Spatial Verification and Analysis Method and System Based on Satellite Remote Sensing Images Technical Field

[0002] The present invention relates to satellite technology, and in particular to a news event spatial verification and analysis method and system based on satellite remote sensing images. Background Art

[0003] With the rapid development of the Internet and social media, the dissemination speed of news event information has been continuously accelerating. The traditional verification of news event authenticity mainly relies on manual verification and text analysis, with low efficiency and difficult to guarantee accuracy. In recent years, with the progress of remote sensing technology, satellite remote sensing images have shown great potential in the field of event verification. At present, some studies have begun to attempt to apply satellite remote sensing data to the verification of news event authenticity, and verify the authenticity of event occurrence by comparing and analyzing the ground object change characteristics in remote sensing images.

[0004] Main problems existing in the prior art:

[0005] Existing satellite remote sensing verification methods often only rely on a single type of satellite data, lacking the collaborative observation ability of multi-source remote sensing data, resulting in incomplete image information obtained and difficult to meet the verification requirements of complex and diverse news events.

[0006] The current remote sensing image analysis mainly adopts traditional image processing methods, lacking intelligent analysis models for different types of events, unable to effectively extract key features related to events, and affecting the accuracy of verification results.

[0007] Existing event verification methods generally lack a knowledge accumulation and experience inheritance mechanism, and fail to make full use of the experience knowledge in historical verification cases, resulting in low verification efficiency and difficult to meet the verification requirements of new types of events. Summary of the Invention

[0008] The embodiments of the present invention provide a news event spatial verification and analysis method and system based on satellite remote sensing images, which can solve the problems in the prior art.

[0009] In the first aspect of the embodiments of the present invention,

[0010] A news event spatial verification and analysis method based on satellite remote sensing images is provided, including:

[0011] Receive the news event text information, and use a natural language processing model to extract the event timestamp, geospatial coordinates, and event type identifier from the news event text information; determine the target observation area based on the extracted geospatial coordinates, and calculate the optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into a strategy correction model to generate a satellite constellation collaborative observation strategy, where the satellite constellation collaborative observation strategy defines the collaborative observation sequence of high-resolution satellites, high-temporal satellites, and hyperspectral satellites; issue a scheduling instruction according to the satellite constellation collaborative observation strategy to activate the corresponding satellite constellation to execute the collaborative observation task;

[0012] Receive the multi-source remote sensing image data obtained during the process of the satellite constellation executing the collaborative observation task, perform adaptive correction processing on the multi-source remote sensing image data to generate spatially and temporally registered standardized image data; select the corresponding target detection network and scene segmentation network from the deep learning model library according to the event type identifier, perform intelligent analysis on the standardized image data, and extract the ground object elements and spatio-temporal change features related to the event; input the extracted ground object elements and spatio-temporal change features into a graph neural network, fuse historical verification knowledge to construct an event feature association graph, and obtain the feature association intensity value through graph reasoning calculation;

[0013] Perform multi-dimensional cross-verification on the feature association intensity value, the event timestamp, and the geospatial coordinates to construct an event authenticity evaluation index; based on the event authenticity evaluation index, combine historical cases in the verification knowledge base, and use transfer learning methods to calculate the event credibility score; when the event credibility score exceeds the preset benchmark credibility threshold, confirm the authenticity of the news event, and write the satellite constellation collaborative observation strategy, the event feature association graph, and the event authenticity evaluation index in this verification process into the verification knowledge base.

[0014] Determine the target observation area based on the extracted geospatial coordinates, and calculate the optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into a strategy correction model, and the generated satellite constellation collaborative observation strategy includes:

[0015] Receive the geospatial coordinates of the news event, use the event geospatial coordinates as seed points, perform iterative expansion on the target area based on the adaptive region growing algorithm, obtain the geographical information entropy value and spatial correlation degree during each iterative calculation, determine the boundary of the target observation area when the geographical information entropy value and the spatial correlation degree meet the preset expansion threshold, and assign an observation priority weight to each sub-region within the boundary of the target observation area to generate a weighted target observation area set;

[0016] Obtain the event timestamp and the weighted set of target observation regions, construct the atmospheric visibility function, solar altitude angle function, and satellite observation angle function into an observation suitability evaluation function in the time dimension, perform time optimization on the weighted set of target observation regions based on the observation suitability evaluation function, determine differentiated observation time windows according to the observation priority weights of each sub-region, and generate an optimal observation time series for each sub-region;

[0017] Based on the optimal observation time series for each sub-region, establish a satellite observation ability evaluation matrix. Each element of the satellite observation ability evaluation matrix represents the observation ability score of a single satellite for a specific sub-region within a specified time window. Based on the observation ability score, satellite mission time, energy consumption, and data transmission delay, construct a multi-dimensional optimization objective function, and obtain an initial satellite group collaborative observation strategy through iterative optimization. The initial satellite group collaborative observation strategy includes the observation time sequence, attitude adjustment sequence, and data transmission path of each satellite;

[0018] Monitor the real-time state of the satellites during the execution of the initial satellite group collaborative observation strategy, obtain the satellite position deviation value, observation condition change value, and energy state change value. When any change value exceeds the preset fluctuation threshold, input the deviation between the any change value and the original observation parameter into the strategy correction model. The strategy correction model dynamically optimizes the observation time sequence, attitude adjustment sequence, and data transmission path of the satellites based on the observation priority weights of each sub-region, and generates an updated satellite group collaborative observation strategy.

[0019] Input the extracted ground object elements and spatio-temporal change characteristics into a graph neural network, fuse historical verification knowledge to construct an event feature association graph, and obtain the feature association intensity value through graph reasoning calculation, including:

[0020] Receive the ground object elements and spatio-temporal change characteristics, construct the ground object elements into a feature vector including category attributes, spatial attributes, and temporal attributes, construct the spatio-temporal change characteristics into a change vector describing the evolution of the ground object, and construct an initial feature graph with the feature vector and the change vector as nodes;

[0021] Based on the current event type, extract historical verification cases from the verification knowledge base to construct a knowledge sub-graph, calculate the cosine similarity of the feature vectors and the spatial position distance between each node in the initial feature graph and the nodes in the knowledge sub-graph, and generate a node similarity matrix;

[0022] Input the node similarity matrix into the knowledge transfer mapping model. The knowledge transfer mapping model projects the verification experience in the knowledge sub-graph onto the corresponding nodes in the initial feature graph, and generates an enhanced feature graph that fuses the verification experience;

[0023] Input the enhanced feature map into the graph attention network. The graph attention network aggregates and updates node features based on the attention coefficients between nodes, calculates edge feature values according to the updated node features, adjusts the connection relationships between nodes using the edge feature values, and outputs a feature association graph. For each pair of connected nodes in the feature association graph, input the node features and edge features into a multi-layer perceptron to calculate the feature association intensity value.

[0024] Perform multi-dimensional cross-validation on the feature association intensity value, the event timestamp, and the geospatial coordinates to construct an event authenticity evaluation index. Based on the event authenticity evaluation index, combined with historical cases in the verification knowledge base, use transfer learning methods to calculate the event credibility score, including:

[0025] Calculate the rationality score of the event occurrence time sequence based on the event timestamp, calculate the relevance score of the geographical location based on the geospatial coordinates, calculate the feature consistency score based on the feature association intensity value, and fuse the rationality score of the time sequence, the relevance score of the geographical location, and the feature consistency score to construct an event authenticity evaluation index.

[0026] Retrieve historical verification cases from the verification knowledge base, construct the historical verification cases into a source domain feature set, and construct the current event to be evaluated into a target domain feature set. Input the source domain feature set and the target domain feature set into a feature encoder for feature mapping. The feature encoder maps the features to a common feature space to generate source domain encoded features and target domain encoded features.

[0027] Input the source domain encoded features and the target domain encoded features into a domain discriminator. The domain discriminator calculates the difference degree between the source domain features and the target domain features distribution, optimizes the feature encoder based on the difference degree, and makes the source domain features and the target domain features distribution tend to be consistent. Input the optimized target domain encoded features into a authenticity evaluator, and the authenticity evaluator calculates the event credibility score based on the source domain verification experience.

[0028] Calculate the rationality score of the event occurrence time sequence based on the timestamp, calculate the relevance score of the geographical location based on the geospatial coordinates, calculate the feature consistency score based on the feature association intensity value, and fuse the rationality score, the relevance score, and the feature consistency score to construct an event authenticity evaluation index, including:

[0029] Construct a time reference sequence model based on the timestamp. The time reference sequence model uses an adaptive weight Gaussian kernel function to calculate the time difference degree between the current event timestamp and the historical associated event time points, calculates the state transition probability matrix of the event sequence through a time Markov chain model, and generates a time sequence rationality score based on the time difference degree and the state transition probability matrix.

[0030] Taking the temporal rationality score as a constraint condition, a spatial feature description model is constructed in combination with the geospatial coordinates. The spatial feature description model calculates the spatial distance matrix between the event location and the set of georeference points using the Hausdorff distance, extracts the spatial distribution features according to the spatial distance matrix, and calculates the matching degree between the spatial distribution features and the spatial constraint rules in the geographic information system to generate a geographical location relevance score;

[0031] Based on the temporal rationality score and the geographical location relevance score, a feature verification model is constructed. The feature verification model normalizes the feature association strength value, extracts the deep spatio-temporal association pattern between features using a multi-layer perceptron, and performs consistency verification on the deep spatio-temporal association pattern with a preset set of feature constraint rules to generate a feature association consistency score;

[0032] An adaptive weight allocation network including a feature extraction layer, an attention mechanism layer, and a normalization output layer is constructed. The temporal rationality score, the geographical location relevance score, and the feature association consistency score are input into the adaptive weight allocation network. The adaptive weight allocation network generates a dynamic fusion weight based on the event type feature vector, and performs weighted fusion on the three scores based on the dynamic fusion weight to generate an event authenticity evaluation index.

[0033] Inputting the source domain encoded features and the target domain encoded features into a domain discriminator, the domain discriminator calculates the difference degree between the source domain feature distribution and the target domain feature distribution, and optimizes the feature encoder based on the difference degree to make the source domain and the target domain feature distributions tend to be consistent; Inputting the optimized target domain encoded features into a authenticity evaluator, the authenticity evaluator calculates the event credibility score based on the source domain verification experience, including:

[0034] Inputting the source domain encoded features and the target domain encoded features into a domain discriminator to generate domain encoded features. The domain discriminator calculates the first-order statistic and the second-order statistic of the domain encoded features, constructs a feature distribution representation based on the first-order statistic and the second-order statistic, obtains the first-order statistic feature distribution representation and the second-order statistic feature distribution representation, and calculates the difference degree between the first-order statistic feature distribution representation and the second-order statistic feature distribution representation using the Wasserstein distance;

[0035] Construct a domain adversarial loss function including an adversarial loss term, a gradient penalty term, and a consistency constraint term according to the difference degree, where the gradient penalty term restricts the feature gradient norm, and the consistency constraint term preserves the feature semantic information, and updates the network parameters of the feature encoder through backpropagation;

[0036] Input the optimized target domain encoded features into the authenticity evaluator. The authenticity evaluator first calculates the semantic similarity between the target features and the source domain verification samples. The semantic similarity calculates the distance between feature vectors based on the cosine metric function, and at the same time considers the temporal correlation and spatial proximity of the features;

[0037] Construct an attention weight matrix based on the semantic similarity, perform weighted aggregation on the source domain verification samples, and generate comprehensive evaluation features of the target event. The comprehensive evaluation features include event attribute features, spatio-temporal association features, and context features;

[0038] Input the comprehensive evaluation features into a multi-layer perceptron. Each layer of the multi-layer perceptron learns evaluation rules at different abstraction levels, and the final output layer maps the features to an event credibility score through the softmax function.

[0039] Input the source domain encoded features and the target domain encoded features into a domain discriminator to generate domain encoded features. The domain discriminator calculates the first-order statistics and second-order statistics of the domain encoded features, constructs a feature distribution representation based on the first-order statistics and the second-order statistics, obtains a first-order statistics feature distribution representation and a second-order statistics feature distribution representation, and calculates the difference degree between the first-order statistics feature distribution representation and the second-order statistics feature distribution representation using the Wasserstein distance, including:

[0040] Input the source domain encoded features and the target domain encoded features into the feature fusion layer of the domain discriminator. The feature fusion layer uses the attention mechanism to calculate the feature weight matrix, and performs weighted fusion on the source domain encoded features and the target domain encoded features based on the feature weight matrix to generate domain encoded features;

[0041] Calculate the weighted mean vector and median vector of the domain encoded features as the first-order statistics. The weighted mean vector is obtained by performing sample weighted averaging on each dimension of the domain encoded features, and the median vector is obtained by performing sorting statistics on each dimension of the domain encoded features;

[0042] Calculate the covariance matrix and correlation coefficient matrix of the domain encoded features as the second-order statistics. The covariance matrix characterizes the mutual relationship between different dimensions of the domain encoded features, and the correlation coefficient matrix is obtained by performing standardization processing on the covariance matrix;

[0043] Combine the weighted mean vector and the median vector to construct a first-order statistics feature distribution representation, and combine the eigenvalues and eigenvectors of the covariance matrix to construct a second-order statistics feature distribution representation;

[0044] The Wasserstein distance represented by the first-order statistic feature distribution is calculated using the Sinkhorn algorithm, and the Wasserstein distance represented by the second-order statistic feature distribution is calculated using matrix factorization techniques;

[0045] Based on cross-validation, the optimal weight coefficient is determined, and the Wasserstein distance represented by the first-order statistic feature distribution and the Wasserstein distance represented by the second-order statistic feature distribution are weighted and combined to generate the inter-domain distribution difference degree.

[0046] In the second aspect of the embodiments of the present invention,

[0047] A news event spatialization verification analysis system based on satellite remote sensing images is provided, including:

[0048] A first unit for receiving news event text information, extracting an event timestamp, a geospatial coordinate, and an event type identifier from the news event text information using a natural language processing model; determining a target observation area based on the extracted geospatial coordinate, and calculating an optimal satellite observation time window according to the event timestamp; inputting the target observation area and the optimal satellite observation time window into a strategy correction model to generate a satellite group collaborative observation strategy, where the satellite group collaborative observation strategy defines the collaborative observation sequence of high-resolution satellites, high-temporal satellites, and hyperspectral satellites; issuing a scheduling instruction according to the satellite group collaborative observation strategy to activate the corresponding satellite group to perform a collaborative observation task;

[0049] A second unit for receiving multi-source remote sensing image data obtained during the process of the satellite group performing the collaborative observation task, performing adaptive correction processing on the multi-source remote sensing image data to generate spatio-temporally registered standardized image data; selecting a corresponding target detection network and scene segmentation network from a deep learning model library according to the event type identifier, performing intelligent analysis on the standardized image data, and extracting ground object elements and spatio-temporal change features related to the event; inputting the extracted ground object elements and spatio-temporal change features into a graph neural network, fusing historical verification knowledge to construct an event feature association graph, and obtaining a feature association strength value through graph reasoning calculation;

[0050] A third unit for performing multi-dimensional cross-validation on the feature association strength value, the event timestamp, and the geospatial coordinate to construct an event authenticity evaluation index; based on the event authenticity evaluation index, combining historical cases in the verification knowledge base, and using a transfer learning method to calculate an event credibility score; when the event credibility score exceeds a preset benchmark credible threshold, confirming the authenticity of the news event, and writing the satellite group collaborative observation strategy, the event feature association graph, and the event authenticity evaluation index in the current verification process into the verification knowledge base.

[0051] In the third aspect of the embodiments of the present invention,

[0052] Provided is an electronic device, including:

[0053] a processor;

[0054] a memory for storing instructions executable by the processor;

[0055] Wherein, the processor is configured to call the instructions stored in the memory to execute the foregoing method.

[0056] In a fourth aspect of the embodiments of the present invention,

[0057] Provided is a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.

[0058] The beneficial effects of this application are as follows:

[0059] By extracting key information of news events through a natural language processing model and formulating a cooperative observation strategy for a satellite constellation based on this information, intelligent scheduling and cooperative observation of multiple types of satellites are achieved, the spatial verification efficiency of news events is improved, and the verification process is made more systematic and automated.

[0060] An intelligent analysis of multi-source remote sensing image data is performed using a deep learning model, an event feature association graph is constructed in combination with a graph neural network, and accurate identification and association analysis of event elements are achieved through the calculation of feature association strength values, greatly improving the accuracy and reliability of news event verification.

[0061] Through multi-dimensional cross-verification and transfer learning methods, a scientific event authenticity evaluation system is established, and key information in the verification process is stored in a knowledge base, forming a continuously optimizable verification mechanism, making the spatial verification of news events more objective and credible. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic flowchart of a method for spatial verification and analysis of news events based on satellite remote sensing images according to an embodiment of the present invention;

[0063] Figure 2 is a schematic diagram of an event authenticity evaluation system according to an embodiment of the present invention;

[0064] Figure 3 is a schematic diagram for comparing comprehensive evaluation indexes of different methods according to an embodiment of the present invention;

[0065] Figure 4 is a schematic structural diagram of a system for spatial verification and analysis of news events based on satellite remote sensing images according to an embodiment of the present invention. DETAILED DESCRIPTION

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0067] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0068] Figure 1 It is a schematic flowchart of a method for spatial verification and analysis of news events based on satellite remote sensing images in an embodiment of the present invention, as Figure 1 shown. The method includes:

[0069] Receiving news event text information, and using a natural language processing model to extract an event timestamp, a geospatial coordinate, and an event type identifier from the news event text information; determining a target observation area based on the extracted geospatial coordinate, and calculating an optimal satellite observation time window according to the event timestamp; inputting the target observation area and the optimal satellite observation time window into a strategy correction model to generate a satellite group collaborative observation strategy, where the satellite group collaborative observation strategy defines a collaborative observation sequence of high-resolution satellites, high-temporal-phase satellites, and hyperspectral satellites; issuing a scheduling instruction according to the satellite group collaborative observation strategy to activate the corresponding satellite group to execute a collaborative observation task;

[0070] Receiving multi-source remote sensing image data obtained during the process of the satellite group executing the collaborative observation task, performing adaptive correction processing on the multi-source remote sensing image data to generate spatially and temporally registered standardized image data; selecting a corresponding target detection network and scene segmentation network from a deep learning model library according to the event type identifier, performing intelligent analysis on the standardized image data to extract ground object elements and spatio-temporal change characteristics related to the event; inputting the extracted ground object elements and spatio-temporal change characteristics into a graph neural network, fusing historical verification knowledge to construct an event feature association graph, and obtaining a feature association intensity value through graph reasoning calculation;

[0071] Perform multi-dimensional cross-validation on the feature correlation strength value, the event timestamp, and the geospatial coordinates to construct an event authenticity evaluation index; based on the event authenticity evaluation index, combined with historical cases in the verification knowledge base, use transfer learning methods to calculate the event credibility score; when the event credibility score exceeds the preset benchmark credibility threshold, confirm the authenticity of the news event, and write the satellite constellation collaborative observation strategy, the event feature correlation graph, and the event authenticity evaluation index in the current verification process into the verification knowledge base.

[0072] In an alternative implementation, determine the target observation area based on the extracted geospatial coordinates, and calculate the optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into the strategy correction model to generate a satellite constellation collaborative observation strategy, including:

[0073] Receive the geospatial coordinates of the news event, use the event geospatial coordinates as seed points, and iteratively expand the target area based on the adaptive region growing algorithm. Obtain the geographical information entropy value and the spatial correlation degree during each iterative calculation. When the geographical information entropy value and the spatial correlation degree meet the preset expansion threshold, determine the boundary of the target observation area, and assign an observation priority weight to each sub-region within the boundary of the target observation area to generate a weighted set of target observation areas;

[0074] Obtain the event timestamp and the weighted set of target observation areas, construct an observation suitability evaluation function in the time dimension with the atmospheric visibility function, the solar altitude angle function, and the satellite observation angle function, optimize the time of the weighted set of target observation areas based on the observation suitability evaluation function, and determine the differentiated observation time windows according to the observation priority weights of each sub-region to generate an optimal observation time series for sub-regions;

[0075] Establish a satellite observation ability evaluation matrix based on the optimal observation time series for sub-regions. Each element of the satellite observation ability evaluation matrix represents the observation ability score of a single satellite for a specific sub-region within a specified time window. Construct a multi-dimensional optimization objective function based on the observation ability score, the satellite mission time, the energy consumption, and the data transmission delay, and obtain the initial satellite constellation collaborative observation strategy through iterative optimization. The initial satellite constellation collaborative observation strategy includes the observation time sequence, the attitude adjustment sequence, and the data transmission path of each satellite;

[0076] Monitor the real-time status of satellites during the execution of the initial satellite cluster collaborative observation strategy, obtain the satellite position deviation value, the observation condition change value, and the energy state change value. When any change value exceeds the preset fluctuation threshold, input the deviation between the any change value and the original observation parameters into the strategy correction model. The strategy correction model dynamically optimizes the observation time sequence, the attitude adjustment sequence, and the data transmission path of the satellites based on the observation priority weights of each sub-region, and generates an updated satellite cluster collaborative observation strategy.

[0077] In the process of determining the target observation area based on geospatial coordinates, first receive the longitude and latitude coordinate information of the event, and use this coordinate as the initial seed point. The adaptive region growing algorithm is used for region expansion, and the geographical information entropy value and the spatial correlation degree are calculated during the expansion process. The geographical information entropy value is measured by analyzing the complexity of geographical elements such as terrain features and ground object distributions in the target area, and the spatial correlation degree is quantified based on the spatial dependence between geographical elements. When these two indicators reach the preset thresholds (for example, the geographical information entropy value is greater than 0.8 and the spatial correlation degree is greater than 0.7), the boundary range of the target observation area can be determined.

[0078] Divide the determined target area into sub-regions, and assign observation weights to each sub-region based on the regional importance assessment. The evaluation factors include population density, economic activity intensity, environmental sensitivity, etc., and the weight value range is 0-1. For example, the weight of the area with intensive economic activities can be set to 0.9, the weight of the environmentally sensitive area can be set to 0.8, and the weight of the general area can be set to 0.5.

[0079] The calculation of the optimal observation time window first constructs an observation suitability evaluation system. The atmospheric visibility function considers factors such as atmospheric transparency and cloud cover, and has the highest score when the visibility is greater than 10 kilometers and the cloud cover is less than 30%. The solar altitude angle function evaluates the lighting conditions based on the solar incident angle, and the angle in the range of 45-75 degrees is most suitable for observation. The satellite observation angle function evaluates the imaging quality according to the angle between the satellite and the target, and has the highest score when observing vertically.

[0080] Integrate the above three evaluation functions to form an observation suitability index in the time dimension, and perform optimization calculations in combination with the weights of each sub-region. High-weight regions are preferentially arranged in the time window with the best observation conditions, and low-weight regions can be arranged in the sub-optimal time window, and finally generate the optimal observation time sequence for each sub-region.

[0081] The construction of the satellite observation ability evaluation matrix needs to consider technical indicators such as the spatial resolution, spectral resolution, and time resolution of the satellite. For example, the observation ability score of a certain satellite for a specific sub-region can be expressed as: 10 points for a spatial resolution of 2 meters, 8 points for more than 4 spectral channels, and 7 points for a revisit period of less than 3 days. The final score is obtained by weighted averaging of each score.

[0082] During the strategy optimization process, the task time is calculated based on the satellite overpass time, the energy consumption is evaluated based on the attitude maneuver and data transmission power consumption, and the data transmission delay takes into account the ground station visibility time and data volume. An optimal solution that meets multi-dimensional constraints is obtained through iterative calculations to form an initial observation strategy.

[0083] During the strategy correction process, the satellite status changes are monitored in real time. When the position deviation exceeds 100 meters, or the score drops by 20% due to changes in the observation conditions, or the remaining energy is less than 30%, the strategy correction is triggered. The correction model re-optimizes the calculation based on the deviation values of the real-time status and the original parameters, combined with the sub-region weights, to ensure that the observation tasks in high-weight regions are prioritized.

[0084] By determining the optimal observation area and time window through adaptive region growth and multi-dimensional evaluation system, the scientificity and accuracy of the observation plan are improved, and the precise coverage and efficient observation of the target area are realized. Based on the real-time status monitoring and dynamic strategy correction mechanism, the adaptability of the satellite observation system to uncertain factors is enhanced, the reliable execution and continuity of the observation tasks are ensured, and the stability of the system operation is improved. By adopting the sub-region differential observation strategy and multi-dimensional constraint optimization method, the reasonable scheduling and efficient utilization of satellite resources are realized, the system operation cost is reduced, and the observation efficiency and timeliness of data acquisition are improved.

[0085] In an alternative embodiment, the extracted ground object elements and spatio-temporal change features are input into a graph neural network, and an event feature correlation graph is constructed by integrating historical verification knowledge. The feature correlation intensity values obtained through graph reasoning calculation include:

[0086] Receiving the ground object elements and spatio-temporal change features, constructing the ground object elements into a feature vector including category attributes, spatial attributes, and temporal attributes, constructing the spatio-temporal change features into a change vector describing the evolution of the ground object, and using the feature vector and the change vector as nodes to construct an initial feature graph;

[0087] Based on the current event type, extracting historical verification cases from the verification knowledge base to construct a knowledge sub-graph, calculating the cosine similarity of the feature vectors and the spatial position distance between each node in the initial feature graph and the nodes in the knowledge sub-graph, and generating a node similarity matrix;

[0088] Inputting the node similarity matrix into a knowledge transfer mapping model, and the knowledge transfer mapping model projects the verification experience in the knowledge sub-graph onto the corresponding nodes of the initial feature graph to generate an enhanced feature graph integrating the verification experience;

[0089] Input the enhanced feature map into the graph attention network. The graph attention network aggregates and updates node features based on the attention coefficients between nodes, calculates edge feature values according to the updated node features, adjusts the connection relationships between nodes using the edge feature values, and outputs a feature correlation graph. For each pair of connected nodes in the feature correlation graph, input the node features and edge features into a multi-layer perceptron to calculate the feature correlation strength value.

[0090] Introduce the processing methods for ground feature elements and spatio-temporal change features. For the input ground feature elements, it is necessary to construct a feature vector containing three dimensions: category, space, and time series. The category attributes include information such as the type identifier and physical properties of the ground feature; the space attributes include geographical coordinates, area range, shape features, etc.; the time series attributes record the state information of the ground feature at different time points. For example, for a certain building, its feature vector may include: the building type is a residential house, the location coordinates are 116 degrees east longitude and 40 degrees north latitude, the building area is 800 square meters, and the completion time is 2010, etc. The spatio-temporal change features describe the evolution process of the ground feature, such as the area expansion rate, height growth trend, etc., and these change features are also transformed into vector form.

[0091] Construct an initial feature map. Use the above-mentioned feature vectors and change vectors as nodes in the graph respectively, and establish connections between nodes based on spatial proximity relationships and change correlations. For example, an edge connection will be established between two adjacent building nodes, indicating that they may be related.

[0092] Extract historical cases from the verification knowledge base. For the current event type of concern, select similar historical verification cases, extract the ground feature and change patterns therein to construct a knowledge sub-graph. Calculate the similarity between each node in the initial feature map and the nodes in the knowledge sub-graph, including the cosine similarity of the feature vectors and the distance of the actual geographical locations, and generate a similarity matrix between nodes.

[0093] In the knowledge transfer mapping stage, input the similarity matrix into a pre-trained mapping model. This model can project the verification experience in the knowledge sub-graph onto the corresponding nodes in the initial feature map. For example, if the change pattern of a certain type of building in the historical case is of great value for event judgment, this experience will be transferred to the building node with similar features in the current situation.

[0094] In the graph attention network processing stage, the network aggregates features based on the attention coefficients between nodes. The attention coefficients reflect the importance weights between different nodes. The network updates the node features through multiple rounds of iteration and calculates the edge feature values accordingly. The edge feature values are used to adjust the connection relationships between nodes, and finally output a feature correlation graph reflecting the association relationships of ground feature elements.

[0095] For each pair of connected nodes in the feature association graph, their node features and edge features are input into a multi-layer perceptron network to calculate the feature association strength value representing the degree of association. The larger this strength value, the closer the association between the two geographical feature elements.

[0096] By constructing multi-dimensional feature vectors and change vectors, the spatio-temporal features of geographical feature elements are comprehensively expressed, improving the integrity and accuracy of feature expression and providing a reliable data basis for subsequent analysis. Feature association analysis by integrating historical verification knowledge makes full use of existing experience, improves the reliability and persuasiveness of analysis results, and avoids the one-sidedness that may occur by relying solely on current data. Using a graph neural network for feature association analysis can automatically learn the complex relationships between nodes, highlight important features through the attention mechanism, and improve the accuracy and interpretability of association analysis.

[0097] In an optional implementation, the feature association strength value is cross-validated in multiple dimensions with the event timestamp and the geospatial coordinates to construct an event authenticity evaluation index; based on the event authenticity evaluation index, combined with historical cases in the verification knowledge base, a transfer learning method is used to calculate the event credibility score, including:

[0098] Calculate the rationality score of the event occurrence time sequence based on the event timestamp, calculate the relevance score of the geographical location based on the geospatial coordinates, calculate the feature consistency score based on the feature association strength value, and fuse the rationality score of the time sequence, the relevance score of the geographical location, and the feature consistency score to construct an event authenticity evaluation index;

[0099] Retrieve historical verification cases from the verification knowledge base, construct the source domain feature set with the historical verification cases, and construct the target domain feature set with the current event to be evaluated; input the source domain feature set and the target domain feature set into a feature encoder for feature mapping, and the feature encoder maps the features to a common feature space to generate source domain encoded features and target domain encoded features;

[0100] Input the source domain encoded features and the target domain encoded features into a domain discriminator, and the domain discriminator calculates the difference degree between the source domain feature and target domain feature distributions, optimizes the feature encoder based on the difference degree, and makes the source domain feature and target domain feature distributions tend to be consistent; input the optimized target domain encoded features into a authenticity evaluator, and the authenticity evaluator calculates the event credibility score based on the source domain verification experience.

[0101] This implementation provides an event authenticity evaluation and credibility calculation method, which realizes accurate evaluation of events through multi-dimensional cross-validation and transfer learning.

[0102] Calculate the rationality score of the event occurrence sequence based on the event timestamp. In this step, the system analyzes whether the occurrence time of the event conforms to the logical order. For example, if an event claims to occur before a certain historical event, but the timestamp shows that it actually occurs after, then the sequence rationality score will be lower. The system may use a time window to evaluate the rationality of the event. For example, given a 30-day window, if the event occurs within 15 days before or after the expected time, a higher rationality score will be given.

[0103] Calculate the relevance score of the geographical location based on the geospatial coordinates. The system checks whether the location where the event occurs matches the claimed location. For example, if an event claims to occur in a desert area, but the geographical coordinates show that it is located in the ocean, then the geographical location relevance score will be very low. The system may use Geographic Information System (GIS) data to verify the rationality of the location, such as checking whether the coordinates fall within the expected geographical area.

[0104] Calculate the feature consistency score based on the feature association intensity value. This step analyzes whether there are contradictions or inconsistencies among the various features of the event. For example, if an event claims to be a large-scale gathering, but there is little relevant social media activity, then the feature consistency score will be lower. The system may use predefined feature weights to calculate the overall consistency. For example, assign a weight of 0.3 to social media activity, 0.5 to on-site photos, and 0.2 to eyewitness testimony.

[0105] Fuse the sequence rationality score, geographical location relevance score, and feature consistency score to construct an event authenticity evaluation metric. This fusion process may adopt a weighted average method. For example, assign a weight of 0.3 to sequence rationality, 0.3 to geographical location relevance, and 0.4 to feature consistency, so as to obtain a comprehensive event authenticity evaluation metric.

[0106] The system retrieves historical verification cases from the verification knowledge base, constructs these cases into a source domain feature set, and constructs the current event to be evaluated into a target domain feature set. For example, if the current event is about a natural disaster, the system retrieves past similar natural disaster events from the knowledge base as the source domain feature set.

[0107] Input the source domain feature set and the target domain feature set into the feature encoder for feature mapping. The feature encoder may adopt a deep neural network structure, such as a multi-layer perceptron or a convolutional neural network, to map the input features to a common feature space, generating source domain encoded features and target domain encoded features. This process can be understood as converting features from different domains into a general representation form.

[0108] Input the source domain encoded features and the target domain encoded features into the domain discriminator. The role of the domain discriminator is to calculate the difference degree between the source domain features and the target domain features distribution. This may be achieved by calculating the distance between the two distributions, such as using KL divergence or Wasserstein distance. Based on the calculated difference degree, the system will optimize the feature encoder, with the goal of making the source domain features and the target domain features distribution tend to be consistent. This optimization process may involve adjusting the parameters of the feature encoder so that it can generate more similar source domain and target domain encoded features.

[0109] Input the optimized target domain encoded features into the authenticity evaluator. The authenticity evaluator will calculate the credibility score of the event based on the verification experience of the source domain. This process may involve using machine learning models, such as support vector machines or random forests, which are trained on the source domain data and then applied to the target domain data. For example, if in the source domain, certain specific feature combinations are usually associated with high credibility, then when similar feature combinations are observed in the target domain, the system will also give a higher credibility score.

[0110] This method improves the accuracy and comprehensiveness of event authenticity evaluation through multi-dimensional cross-validation, comprehensively considering multiple aspects such as time, geographical location, and feature consistency, effectively reducing the bias and errors that may be brought by single-dimensional evaluation. The transfer learning method is used to achieve the effective transfer of historical experience to the evaluation of new events, overcoming the evaluation difficulties that may occur in traditional methods when facing new or rare events, and improving the adaptability and evaluation accuracy of the system to various types of events. Through the optimization process of the feature encoder and the domain discriminator, the alignment of the source domain and target domain features distribution is achieved, effectively reducing the impact of domain differences on the evaluation results, and improving the reliability and generalization ability of the evaluation results.

[0111] In an alternative embodiment, calculate the rationality score of the event occurrence sequence based on the timestamp, calculate the relevance score of the geographical location based on the geospatial coordinates, calculate the feature consistency score based on the feature association strength value, and the construction of the event authenticity evaluation index by fusing the rationality score, the relevance score, and the feature consistency score includes:

[0112] Construct a time reference sequence model based on the timestamp. The time reference sequence model uses a Gaussian kernel function with adaptive weights to calculate the time difference degree between the current event timestamp and the time points of historical associated events, calculates the state transition probability matrix of the event sequence through a time Markov chain model, and generates a temporal rationality score based on the time difference degree and the state transition probability matrix;

[0113] Taking the temporal rationality score as a constraint condition, a spatial feature description model is constructed in combination with the geospatial coordinates. The spatial feature description model calculates the spatial distance matrix between the event location and the set of georeference points using the Hausdorff distance, extracts the spatial distribution features according to the spatial distance matrix, and calculates the matching degree between the spatial distribution features and the spatial constraint rules in the geographic information system to generate a geographical location relevance score;

[0114] A feature verification model is constructed based on the temporal rationality score and the geographical location relevance score. The feature verification model normalizes the feature association strength value, extracts the deep spatio-temporal association pattern between features using a multi-layer perceptron, and performs consistency verification on the deep spatio-temporal association pattern with a preset set of feature constraint rules to generate a feature association consistency score;

[0115] An adaptive weight allocation network including a feature extraction layer, an attention mechanism layer, and a normalization output layer is constructed. The temporal rationality score, the geographical location relevance score, and the feature association consistency score are input into the adaptive weight allocation network. The adaptive weight allocation network generates a dynamic fusion weight based on the event type feature vector, and performs weighted fusion on the three scores based on the dynamic fusion weight to generate an event authenticity evaluation index.

[0116] Evaluate the temporal rationality of events. By constructing a time reference sequence model, which uses a Gaussian kernel function with adaptive weights to process time information. Specifically, for the timestamp of the current event, the time point information of historical associated events is extracted, and the time difference degree is calculated. For example, for a commercial activity event occurring in a certain business district in Beijing, its timestamp is 3 pm on a certain day. By querying historical data, it is found that the commercial activities in this business district are usually concentrated between 2 pm and 5 pm. The time difference degree calculated using the Gaussian kernel function is 0.85, indicating that this time point has a high degree of rationality. Then, a state transition probability matrix is constructed using the time Markov chain model to analyze the temporal law of the event sequence. Based on the time difference degree and the state transition probability, a temporal rationality score of 0.9 is finally generated.

[0117] Perform a geographical location relevance assessment. Use the time-series rationality score obtained above as a constraint condition, and combine the geospatial coordinate information of the event to construct a spatial feature description model. This model uses the Hausdorff distance to calculate the spatial distance matrix between the event location and the geographical reference point. Taking the above commercial activity as an example, by calculating the distance between the event occurrence location and the main commercial facilities in the business district, a spatial distance matrix is obtained. Extract spatial distribution features such as location density and aggregation degree from this matrix. Calculate the matching degree between these features and the spatial constraint rules in the geographical information system, such as considering factors like building distribution and road network, and finally generate a geographical location relevance score of 0.85.

[0118] Construct a feature verification model to evaluate the feature association consistency. This model first normalizes the feature association strength values, unifying feature values with different dimensions into the range of zero to one. Then use a multi-layer perceptron to extract the deep spatio-temporal association patterns between features. In the example, extract features such as the number of people flow, consumption level, and surrounding facilities of the commercial activity, and analyze the association patterns between these features and time and location through a multi-layer perceptron. Verify the consistency of the extracted association patterns with the preset set of feature constraint rules, and generate a feature association consistency score of 0.88.

[0119] Construct an adaptive weight assignment network for score fusion. This network includes a feature extraction layer, an attention mechanism layer, and a normalized output layer. Input the above three scores into the network, and generate dynamic fusion weights based on the event type feature vector. For commercial activity events, the weights of time-series rationality, geographical location relevance, and feature consistency calculated according to their feature vectors are 0.3, 0.35, and 0.35 respectively. The final event authenticity evaluation index is obtained as 0.87 through weighted fusion.

[0120] Figure 2 Schematic diagram of the event authenticity evaluation and analysis for the embodiment of the present invention:

[0121] This figure shows a complete event authenticity evaluation and analysis platform. At the top of the interface, the title "Event Authenticity Evaluation and Analysis" and the event ID (EV-20250407-001) are displayed. The main part below consists of three parallel analysis cards and a comprehensive evaluation area at the bottom. The left card shows the temporal rationality analysis. Five key time points from 08:15 to 14:05 and their Markov transition probabilities (0.87, 0.75, 0.62, 0.91) are visually presented through a timeline graph. The Gaussian kernel time difference is 0.23, and finally, the temporal rationality score is 0.79. The middle card shows the geographical location relevance analysis. The current event location and multiple reference points are marked on the map, along with their Hausdorff distance (42.8) and the spatial constraint area. The spatial constraint matching degree is 0.83, and the geographical location relevance score is 0.83. The right card shows the feature association consistency analysis, which presents feature normalization (7 feature bar charts), deep association patterns (multi-layer perceptron network structure diagrams), and consistency verification (a table of 5 constraint rules with consistency values ranging from 0.68 to 0.92) in the form of tabs. The feature association consistency score is 0.82. The comprehensive evaluation area at the bottom shows the final score of 0.82 in a pie chart and lists three sub-indicators (temporal 0.79, geographical 0.83, feature 0.82) and their corresponding weights (0.32, 0.36, 0.32). The entire interface comprehensively displays a multi-dimensional evaluation system for event authenticity based on an adaptive weight allocation network through progress bars, charts, and data visualization, providing users with an intuitive and scientific basis for determining event authenticity.

[0122] By analyzing the temporal features through an adaptive weight Gaussian kernel function and a Markov chain model, the accurate evaluation of the event occurrence time sequence is realized, the accuracy of the temporal rationality judgment is improved, and the evaluation results are more objective and credible. By adopting a spatial feature description model and a Hausdorff distance calculation method, combined with the constraint rules of a geographic information system, the comprehensive evaluation of the event geographical location relevance is realized, and the reliability and adaptability of the spatial relevance analysis are enhanced. Based on the fusion method of a multi-layer perceptron and an adaptive weight allocation network, the in-depth mining of multi-dimensional features and dynamic weight allocation are realized, the accuracy of the feature relevance analysis is improved, and the authenticity evaluation results are more comprehensive and reliable.

[0123] In an optional implementation manner, the source domain encoded features and the target domain encoded features are input into a domain discriminator. The domain discriminator calculates the difference degree between the source domain features and the target domain features distribution, and optimizes the feature encoder based on the difference degree to make the source domain and the target domain features distribution tend to be consistent. The optimized target domain encoded features are input into an authenticity evaluator. The authenticity evaluator calculates the event credibility score based on the source domain verification experience, including:

[0124] Input the source domain encoded features and the target domain encoded features into the domain discriminator to generate domain encoded features. The domain discriminator calculates the first-order statistics and second-order statistics of the domain encoded features, constructs a feature distribution representation based on the first-order statistics and the second-order statistics, obtains the first-order statistics feature distribution representation and the second-order statistics feature distribution representation, and calculates the difference degree between the first-order statistics feature distribution representation and the second-order statistics feature distribution representation using the Wasserstein distance;

[0125] Construct a domain adversarial loss function including an adversarial loss term, a gradient penalty term, and a consistency constraint term according to the difference degree, where the gradient penalty term restricts the feature gradient norm, and the consistency constraint term preserves the feature semantic information, and update the network parameters of the feature encoder through backpropagation;

[0126] Input the optimized target domain encoded features into the authenticity evaluator. The authenticity evaluator first calculates the semantic similarity between the target features and the source domain verification samples. The semantic similarity calculates the distance between feature vectors based on the cosine metric function, and also considers the temporal correlation and spatial proximity of the features;

[0127] Construct an attention weight matrix based on the semantic similarity, perform weighted aggregation on the source domain verification samples, and generate a comprehensive evaluation feature of the target event. The comprehensive evaluation feature includes event attribute features, spatio-temporal association features, and context features;

[0128] Input the comprehensive evaluation feature into a multi-layer perceptron. Each layer of the multi-layer perceptron learns evaluation rules at different abstraction levels, and the final output layer maps the features to an event credibility score through the softmax function.

[0129] The feature encoder encodes the source domain data and the target domain data respectively to generate corresponding feature vectors. The source domain data comes from a verified real event dataset, and the target domain data comes from the event data to be verified. The feature encoding adopts a deep neural network structure, including multiple convolutional layers and fully connected layers, and is used to extract the deep semantic features of the data.

[0130] The domain discriminator receives the encoded features of the source domain and the target domain as inputs. The domain discriminator first calculates the first-order statistics of the features, that is, the mean of the feature vectors; at the same time, it calculates the second-order statistics, that is, the covariance matrix of the feature vectors. Based on these statistics, a feature distribution representation is constructed. Specifically, the feature vectors can be divided into multiple subspaces, and the local statistics are calculated respectively, and then they are combined to form a global distribution representation. For example, for a 128-dimensional feature vector, it can be divided into 8 16-dimensional subspaces.

[0131] The domain discriminator uses the Wasserstein distance to measure the difference in the feature distributions between the source domain and the target domain. The greater the difference, the more significant the difference in the feature distributions of the two domains. Based on the calculated difference, a domain adversarial loss function is constructed. The loss function consists of three parts: the adversarial loss term is used to minimize the domain difference; the gradient penalty term ensures the model stability by restricting the feature gradient norm; and the consistency constraint term preserves the semantic information of the features from being destroyed. The parameters of the feature encoder are optimized through the backpropagation algorithm to gradually make the feature distributions of the source domain and the target domain tend to be consistent.

[0132] After domain adversarial training, the optimized encoded features of the target domain are input into the authenticity evaluator. The evaluator first calculates the semantic similarity between the target features and the source domain validation samples. The similarity calculation is based on the cosine metric function, taking into account the temporal correlation and spatial proximity of the features. For example, for temporal correlation, the time window can be set to 24 hours, and only the samples within this window are considered; for spatial proximity, the geographical distance threshold can be set to 10 kilometers.

[0133] Based on the calculated semantic similarity, an attention weight matrix is constructed. The weight matrix is used to perform weighted aggregation on the source domain validation samples to generate the comprehensive evaluation features of the target event. The comprehensive evaluation features include three aspects: the event attribute features describe the basic features of the event; the spatio-temporal association features depict the spatio-temporal patterns of the event occurrence; and the context features reflect the association relationships between the event and other related events.

[0134] The comprehensive evaluation features are input into a multi-layer perceptron for credibility evaluation. The multi-layer perceptron contains multiple hidden layers, and each layer learns the evaluation rules at different abstraction levels. For example, the first layer learns the evaluation rules for basic attribute features, the second layer learns the evaluation rules for spatio-temporal features, and the third layer learns the evaluation rules for context features. The final output layer maps the features to an event credibility score between 0 and 1 through the softmax function.

[0135] Figure 3 The following is a schematic diagram for comparing the comprehensive evaluation indicators of different methods in the embodiments of the present invention:

[0136] This multi - indicator radar chart shows the performance comparison of four different technical methods in five key evaluation indicators: accuracy, F1 - score, precision, recall, and feature matching degree. The cross - shaped line in the figure represents "the proposed technical solution", which has achieved the best results in all five dimensions, with values of accuracy 0.95, F1 - score 0.97, precision 0.94, recall 0.94, and feature matching degree 0.96, forming the outermost polygon; the triangular line represents the "DANN method", which is the second - best performing solution, with values of accuracy 0.85, F1 - score 0.86, precision 0.83, recall 0.85, and feature matching degree 0.84; the circular line represents the "first - order statistics only" method, with indicator values of accuracy 0.80, F1 - score 0.79, precision 0.77, recall 0.76, and feature matching degree 0.81; the square line represents the "traditional GAN method", which has the weakest comprehensive performance among the four methods, with values of accuracy 0.75, F1 - score 0.74, precision 0.77, recall 0.73, and feature matching degree 0.71. The proposed technical solution maintains a high level above 0.94 in all indicators, especially performing excellently in F1 - score and feature matching degree, generally improving by 15 - 20 percentage points compared to other methods, demonstrating its excellent performance and robustness in the authenticity assessment task.

[0137] The alignment of the source - domain and target - domain feature distributions is achieved through domain - adversarial training, effectively reducing the negative impact brought by domain shift and improving the generalization performance of the model in the target domain. At the same time, the multi - level statistical - quantity feature - distribution representation can more comprehensively characterize the data - distribution features and enhance the domain - adaptation effect. Introducing a semantic - similarity calculation method with time - correlation and space - proximity constraints can more accurately identify similar events and avoid interference from irrelevant events with large distances or long time spans. The introduction of the attention mechanism realizes the adaptive weighting of source - domain verification samples, highlighting the contribution of important samples. The hierarchical evaluation mechanism of the multi - layer perceptron can learn evaluation rules from different abstraction levels, achieving multi - angle and multi - level evaluation of event credibility. Considering event attributes, spatio - temporal associations, and context features comprehensively ensures the comprehensiveness and reliability of the evaluation results.

[0138] In an optional implementation, the source - domain encoded features and the target - domain encoded features are input into a domain discriminator to generate domain - encoded features. The domain discriminator calculates the first - order statistics and second - order statistics of the domain - encoded features, constructs a feature - distribution representation based on the first - order statistics and the second - order statistics, and obtains a first - order - statistics feature - distribution representation and a second - order - statistics feature - distribution representation. The calculation of the difference degree between the first - order - statistics feature - distribution representation and the second - order - statistics feature - distribution representation using the Wasserstein distance includes:

[0139] Input the source domain encoded features and the target domain encoded features into the feature fusion layer of the domain discriminator. The feature fusion layer uses an attention mechanism to calculate a feature weight matrix, and based on the feature weight matrix, performs weighted fusion on the source domain encoded features and the target domain encoded features to generate domain encoded features;

[0140] Calculate the weighted mean vector and the median vector of the domain encoded features as first-order statistics. The weighted mean vector is obtained by performing sample weighted averaging on each dimension of the domain encoded features, and the median vector is obtained by performing sorting statistics on each dimension of the domain encoded features;

[0141] Calculate the covariance matrix and the correlation coefficient matrix of the domain encoded features as second-order statistics. The covariance matrix characterizes the mutual relationship between different dimensions of the domain encoded features, and the correlation coefficient matrix is obtained by performing normalization processing on the covariance matrix;

[0142] Combine the weighted mean vector and the median vector to construct a first-order statistic feature distribution representation, and combine the eigenvalues and eigenvectors of the covariance matrix to construct a second-order statistic feature distribution representation;

[0143] Use the Sinkhorn algorithm to calculate the Wasserstein distance of the first-order statistic feature distribution representation, and use matrix factorization technology to calculate the Wasserstein distance of the second-order statistic feature distribution representation;

[0144] Determine the optimal weight coefficient based on cross-validation, and perform weighted combination on the Wasserstein distance of the first-order statistic feature distribution representation and the Wasserstein distance of the second-order statistic feature distribution representation to generate the inter-domain distribution difference degree.

[0145] During the domain adaptation process, first obtain the source domain encoded features and the target domain encoded features. Process these two types of features through the feature fusion layer. The feature fusion layer uses an attention mechanism to calculate feature weights. Specifically, calculate attention scores for each feature vector, and the attention scores reflect the importance of the features. For example, for a 128-dimensional feature vector, 128 corresponding weight values are calculated, and the weight values range from 0 to 1. These weight values are formed into a feature weight matrix for subsequent feature fusion.

[0146] During feature fusion, multiply the feature vectors of the source domain and the target domain by the corresponding weight values, perform weighted summation, and obtain the fused domain encoded features. Take a specific example to illustrate. Suppose the source domain feature vector is [0.3, 0.5, 0.7], the target domain feature vector is [0.4, 0.6, 0.8], and the corresponding weight values are [0.4, 0.3, 0.3], then the fused feature vector is [0.35, 0.55, 0.75].

[0147] Calculate the first-order statistics of the computational domain coding features. First, calculate the weighted mean vector by taking the weighted average of each dimension of the features. At the same time, calculate the median vector by sorting the values of each dimension of the features and taking the value at the middle position as the median.

[0148] Calculate the second-order statistics. When calculating the covariance matrix, analyze the correlation relationship between different dimensions of the features. For the eigenvalue sequences of two dimensions of features, calculate the covariance by computing the mean of the product of their deviations. Standardize the covariance matrix to obtain the correlation coefficient matrix, where the values range from -1 to 1.

[0149] When constructing the feature distribution representation, concatenate the weighted mean vector and the median vector to form the first-order statistic feature distribution representation. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues and eigenvectors, and combine them to form the second-order statistic feature distribution representation.

[0150] When calculating the distribution difference degree, use the Sinkhorn iterative algorithm to calculate the Wasserstein distance of the first-order statistics. Through the iterative optimization method, find the optimal transport plan between the two distributions. For the second-order statistics, use the matrix decomposition method to calculate the Wasserstein distance. Finally, determine the weight coefficient through cross-validation, and combine the two distances with weights to obtain the final inter-domain distribution difference degree.

[0151] By calculating the feature weight matrix through the attention mechanism, the adaptive fusion of source domain and target domain features is realized, improving the accuracy and robustness of feature representation. The importance of different features is considered during the feature fusion process, making the fusion result more reasonable. At the same time, use the first-order and second-order statistics to construct the feature distribution representation, comprehensively characterizing the features of the data distribution. The weighted mean and median reflect the central tendency of the data, and the covariance matrix and correlation coefficient matrix reflect the mutual relationship between the features, making the distribution representation more complete. Use the Wasserstein distance to measure the distribution difference and determine the optimal weight coefficient through cross-validation, making the calculation of the inter-domain difference degree more accurate and reliable. This method can effectively capture the distribution difference between the source domain and the target domain, providing an important basis for subsequent domain adaptation tasks.

[0152] The transformation of satellite news topics is the primary link in the production process. The transformation process requires in-depth analysis and decomposition of the topics, decomposing the news topics into multiple specific sub-topics. Taking the report on the changes in the urban ecological environment as an example, it can be decomposed into multiple sub-topics such as urban expansion, vegetation coverage change, and water body change. There are internal connections between the sub-topics, and a complete news report framework is jointly constructed through analysis from different dimensions.

[0153] For the decomposed sub-topics, detailed data requirement analysis is needed. Data requirements include multiple dimensions such as time scale, spatial scale, data type, and observation indicators. Among them, the time scale determines the required observation period and time resolution; the spatial scale clarifies the observation area range and spatial resolution requirements; the data type includes data from different sensors such as optical and radar; the observation indicators involve professional parameters such as vegetation index and land surface temperature. For example, when monitoring the construction progress of large infrastructure, high-resolution optical satellite data with a spatial resolution better than 2 meters needs to be selected, and a combination of panchromatic satellite data with a resolution of 0.5 meters and multispectral satellite data with a resolution of 2 meters can be used.

[0154] The selection of data sources needs to be evaluated through feasibility analysis. The evaluation content includes the integrity of data coverage, whether the time resolution meets the requirements of dynamic monitoring, whether the spatial resolution is sufficient to identify target features, the influence degree of interference factors such as cloud amount, and the cost of data acquisition and processing. Only after comprehensive evaluation to ensure that the data source meets the reporting requirements can the next step be carried out.

[0155] In the data acquisition and preprocessing link, first, data needs to be downloaded according to the determined time range and spatial range. Quality inspection should be carried out on the downloaded data to eliminate data with excessive cloud amount and unqualified quality, and a standardized data management directory should be established. Subsequently, radiometric correction and atmospheric correction are carried out to eliminate the system errors of the sensor and the influence of the atmosphere on the reflection characteristics of ground objects, and the data is converted into surface reflectance. In addition, geometric correction and registration are also required to eliminate the geometric distortion caused by terrain undulation and satellite attitude changes, unify the data of different time phases into the same coordinate system, and achieve precise registration of multi-source data. Finally, through data mosaicking and cropping, a dataset that completely covers the study area is generated.

[0156] Satellite data analysis and visualization are the core links of news production. Through feature extraction and classification, the land cover types and change information in the target area are identified. Change detection is carried out based on multi-temporal images, the area and change rate of the changed area are calculated, and the change trend and pattern are analyzed. In terms of visual expression, appropriate presentation forms need to be selected according to the news content, such as rolling comparison, dynamic graphs, etc., and a reasonable color scheme and legend should be designed, and necessary geographical elements and thematic information should be added. Finally, satellite data is fused and displayed with multi-source data such as ground data and statistical information to form a comprehensive visual news product.

[0157] Figure 4 This is a schematic structural diagram of the news event spatialization verification and analysis system based on satellite remote sensing images according to an embodiment of the present invention, as Figure 4 shown, the system includes:

[0158] The first unit is used to receive news event text information, extract an event timestamp, geospatial coordinates, and an event type identifier from the news event text information using a natural language processing model; determine a target observation area based on the extracted geospatial coordinates, and calculate an optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into a strategy correction model to generate a satellite group collaborative observation strategy, where the satellite group collaborative observation strategy defines the collaborative observation sequences of high-resolution satellites, high-temporal-phase satellites, and hyperspectral satellites; issue a scheduling instruction according to the satellite group collaborative observation strategy to activate the corresponding satellite group to perform a collaborative observation task;

[0159] The second unit is used to receive multi-source remote sensing image data obtained during the process of the satellite group performing the collaborative observation task, perform adaptive correction processing on the multi-source remote sensing image data to generate spatio-temporally registered standardized image data; select a corresponding target detection network and scene segmentation network from a deep learning model library according to the event type identifier, perform intelligent analysis on the standardized image data to extract ground object elements and spatio-temporal change features related to the event; input the extracted ground object elements and spatio-temporal change features into a graph neural network, fuse historical verification knowledge to construct an event feature association graph, and obtain a feature association intensity value through graph reasoning calculation;

[0160] The third unit is used to perform multi-dimensional cross-verification on the feature association intensity value, the event timestamp, and the geospatial coordinates to construct an event authenticity evaluation index; based on the event authenticity evaluation index, combined with historical cases in the verification knowledge base, use transfer learning methods to calculate an event credibility score; when the event credibility score exceeds a preset benchmark credible threshold, confirm the authenticity of the news event, and write the satellite group collaborative observation strategy, the event feature association graph, and the event authenticity evaluation index in the current verification process into the verification knowledge base.

[0161] In the third aspect of the embodiments of the present invention,

[0162] There is provided an electronic device, including:

[0163] A processor;

[0164] A memory for storing instructions executable by the processor;

[0165] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0166] In the fourth aspect of the embodiments of the present invention,

[0167] There is provided a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0168] The present invention may be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A spatial verification and analysis method for news events based on satellite remote sensing images, characterized in that: include: Receiving news event text information, and extracting event timestamps, geospatial coordinates, and event type identifiers from the news event text information using a natural language processing model; Determine the target observation area based on the extracted geospatial coordinates, and calculate the optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into the strategy correction model to generate a satellite cluster collaborative observation strategy, wherein the satellite cluster collaborative observation strategy defines a collaborative observation sequence of high-resolution satellites, high-phase satellites, and hyperspectral satellites; issue a scheduling instruction according to the satellite cluster collaborative observation strategy to activate the corresponding satellite cluster to perform the collaborative observation task; Receive multi-source remote sensing image data acquired during the satellite group's collaborative observation mission, perform adaptive correction processing on the multi-source remote sensing image data, and generate standardized image data with spatiotemporal registration; select a corresponding target detection network and scene segmentation network from a deep learning model library according to the event type identifier, perform intelligent analysis on the standardized image data, and extract event-related ground features and spatiotemporal change characteristics; The extracted ground features and spatiotemporal change characteristics are input into the graph neural network, and the event feature association graph is constructed by integrating historical verification knowledge. The feature association strength value is obtained through graph reasoning calculation. The feature association strength value is cross-validated with the event timestamp and the geographic space coordinates in multiple dimensions to construct an event authenticity evaluation index; based on the event authenticity evaluation index and in combination with historical cases in the verification knowledge base, a transfer learning method is used to calculate the event credibility score; When the event credibility score exceeds the preset benchmark credibility threshold, the authenticity of the news event is confirmed, and the satellite group collaborative observation strategy, the event feature association map, and the event authenticity evaluation index in this verification process are written into the verification knowledge base.

2. The method according to claim 1, characterized in that determining a target observation area based on the extracted geospatial coordinates, and calculating an optimal satellite observation time window according to the event timestamp; Inputting the target observation area and the optimal satellite observation time window into the strategy correction model to generate a satellite cluster collaborative observation strategy includes: Receive the geospatial coordinates of a news event, use the geospatial coordinates of the event as a seed point, iteratively expand the target area based on an adaptive region growing algorithm, obtain the geographic information entropy value and the spatial correlation in each iterative calculation process, determine the boundary of the target observation area when the geographic information entropy value and the spatial correlation meet a preset expansion threshold, and assign an observation priority weight to each sub-area within the boundary of the target observation area, so as to generate a weighted target observation area set; Acquire the event timestamp and the weighted target observation area set, construct the atmospheric visibility function, the solar altitude angle function, and the satellite observation angle function as an observation suitability evaluation function in the time dimension, perform time optimization on the weighted target observation area set based on the observation suitability evaluation function, determine differentiated observation time windows according to the observation priority weights of each sub-area, and generate the optimal observation time series for each sub-area; A satellite observation capability evaluation matrix is ​​established based on the optimal observation time sequence of the sub-region, wherein each element of the satellite observation capability evaluation matrix represents the observation capability score of a single satellite for a specific sub-region within a specified time window, a multi-dimensional optimization objective function is constructed based on the observation capability score and satellite mission time, energy consumption, and data transmission delay, and an initial satellite group collaborative observation strategy is obtained through iterative optimization and solving, wherein the initial satellite group collaborative observation strategy includes the observation timing, attitude adjustment sequence, and data transmission path of each satellite; The real-time status of the satellites during the execution of the initial satellite cluster collaborative observation strategy is monitored to obtain satellite position deviation values, observation condition change values, and energy status change values. When any change value exceeds a preset fluctuation threshold, the deviation between any change value and the original observation parameter is input into a strategy correction model. The strategy correction model dynamically optimizes the observation timing of the satellite, the attitude adjustment sequence, and the data transmission path based on the observation priority weights of each sub-area to generate an updated satellite cluster collaborative observation strategy.

3. The method according to claim 1, characterized in that: The extracted ground features and spatiotemporal change characteristics are input into the graph neural network, and the historical verification knowledge is integrated to construct the event feature association graph. The feature association strength values ​​obtained through graph reasoning calculation include: Receiving ground feature elements and spatiotemporal change features, constructing the ground feature elements into feature vectors including category attributes, spatial attributes, and temporal attributes, constructing the spatiotemporal change features into change vectors describing the evolution of ground features, and constructing an initial feature graph using the feature vectors and the change vectors as nodes; Based on the current event type, historical verification cases are extracted from the verification knowledge base to construct a knowledge subgraph, and the cosine similarity and spatial position distance of the feature vectors of each node in the initial feature graph and the nodes in the knowledge subgraph are calculated to generate a node similarity matrix; The node similarity matrix is ​​input into a knowledge transfer mapping model, and the knowledge transfer mapping model projects the verification experience in the knowledge subgraph to the corresponding node of the initial feature graph to generate an enhanced feature graph integrating the verification experience; The enhanced feature graph is input into a graph attention network, which aggregates and updates node features based on attention coefficients between nodes, calculates edge feature values ​​according to the updated node features, uses the edge feature values ​​to adjust the connection relationship between nodes, and outputs a feature association graph; for each pair of connected nodes in the feature association graph, the node features and edge features are input into a multilayer perceptron to calculate feature association strength values.

4. The method according to claim 1, characterized in that: Perform multi-dimensional cross-validation on the feature association strength value, the event timestamp, and the geographic space coordinates to construct an event authenticity assessment index; Based on the event authenticity evaluation indicators and combined with historical cases in the verification knowledge base, the event credibility score is calculated using the transfer learning method, including: Calculating the rationality score of the event occurrence sequence based on the event timestamp, calculating the relevance score of the geographic location based on the geographic spatial coordinates, calculating the feature consistency score based on the feature association strength value, and fusing the rationality score of the sequence, the relevance score of the geographic location and the feature consistency score to construct an event authenticity evaluation index; Retrieving historical verification cases from a verification knowledge base, constructing the historical verification cases into a source domain feature set, and constructing the current event to be evaluated into a target domain feature set; inputting the source domain feature set and the target domain feature set into a feature encoder for feature mapping, and the feature encoder maps the features to a common feature space to generate source domain encoding features and target domain encoding features; The source domain encoding features and the target domain encoding features are input into a domain discriminator, the domain discriminator calculates the difference between the distribution of source domain features and target domain features, and optimizes the feature encoder based on the difference to make the distribution of the source domain features and the target domain features tend to be consistent; the optimized target domain encoding features are input into an authenticity evaluator, and the authenticity evaluator calculates the event credibility score based on the source domain verification experience.

5. The method according to claim 4, characterized in that Calculating the rationality score of the event occurrence sequence based on the timestamp, calculating the relevance score of the geographic location based on the geographic space coordinates, calculating the feature consistency score based on the feature association strength value, and fusing the rationality score, the relevance score and the feature consistency score to construct an event authenticity evaluation index includes: A time reference sequence model is constructed based on the timestamp, the time reference sequence model uses a Gaussian kernel function with adaptive weights to calculate the time difference between the current event timestamp and the time point of the historical associated event, calculates the state transition probability matrix of the event sequence through a time Markov chain model, and generates a time series rationality score based on the time difference and the state transition probability matrix; The temporal rationality score is used as a constraint condition and combined with the geographic spatial coordinates to construct a spatial feature description model, wherein the spatial feature description model uses Hausdorff distance to calculate a spatial distance matrix between event locations and a set of geographic reference points, extracts spatial distribution features based on the spatial distance matrix, calculates the matching degree between the spatial distribution features and the spatial constraint rules in the geographic information system, and generates a geographic location relevance score; A feature verification model is constructed based on the temporal rationality score and the geographic location correlation score, the feature verification model normalizes the feature correlation strength value, uses a multi-layer perceptron to extract the deep spatiotemporal correlation pattern between features, and performs consistency verification between the deep spatiotemporal correlation pattern and a preset feature constraint rule set to generate a feature correlation consistency score; An adaptive weight allocation network including a feature extraction layer, an attention mechanism layer and a normalized output layer is constructed, and the rationality score of the time series, the correlation score of the geographic location and the feature correlation consistency score are input into the adaptive weight allocation network. The adaptive weight allocation network generates a dynamic fusion weight based on the event type feature vector, and the three scores are weightedly fused based on the dynamic fusion weight to generate an event authenticity evaluation index.

6. The method according to claim 4, characterized in that Inputting the source domain encoding features and the target domain encoding features into a domain discriminator, the domain discriminator calculating the difference between the source domain features and the target domain features, and optimizing the feature encoder based on the difference so that the source domain features and the target domain features are distributed in a consistent manner; The optimized target domain encoding features are input into the authenticity evaluator, and the authenticity evaluator calculates the event credibility score based on the source domain verification experience, including: Inputting the source domain coding feature and the target domain coding feature into a domain discriminator to generate a domain coding feature, the domain discriminator calculating a first-order statistic and a second-order statistic of the domain coding feature, constructing a feature distribution representation based on the first-order statistic and the second-order statistic to obtain a first-order statistic feature distribution representation and a second-order statistic feature distribution representation, and using Wasserstein distance to calculate the difference between the first-order statistic feature distribution representation and the second-order statistic feature distribution representation; Constructing a domain adversarial loss function including an adversarial loss term, a gradient penalty term and a consistency constraint term according to the difference, wherein the gradient penalty term limits the feature gradient norm, the consistency constraint term maintains feature semantic information, and updating the network parameters of the feature encoder through back propagation; The optimized target domain encoding features are input into the authenticity evaluator, which first calculates the semantic similarity between the target features and the source domain verification samples. The semantic similarity is based on the cosine metric function to calculate the distance between feature vectors, while considering the temporal correlation and spatial proximity of the features; An attention weight matrix is ​​constructed based on the semantic similarity, and the source domain verification samples are weighted and aggregated to generate a comprehensive evaluation feature of the target event, wherein the comprehensive evaluation feature includes event attribute features, spatiotemporal correlation features, and context features; The comprehensive evaluation features are input into a multi-layer perceptron, each layer of the multi-layer perceptron learns evaluation rules at different abstract levels respectively, and the final output layer maps the features into event credibility scores through a normalized exponential function.

7. The method according to claim 6, characterized in that Inputting the source domain coding feature and the target domain coding feature into a domain discriminator to generate a domain coding feature, the domain discriminator calculating the first-order statistics and the second-order statistics of the domain coding feature, constructing a feature distribution representation based on the first-order statistics and the second-order statistics to obtain a first-order statistic feature distribution representation and a second-order statistic feature distribution representation, and using the Wasserstein distance to calculate the difference between the first-order statistic feature distribution representation and the second-order statistic feature distribution representation includes: Inputting the source domain encoding feature and the target domain encoding feature into a feature fusion layer of a domain discriminator, wherein the feature fusion layer uses an attention mechanism to calculate a feature weight matrix, and performs weighted fusion on the source domain encoding feature and the target domain encoding feature based on the feature weight matrix to generate a domain encoding feature; Calculating a weighted mean vector and a median vector of the domain coding feature as first-order statistics, wherein the weighted mean vector is obtained by performing sample weighted averaging on each dimension of the domain coding feature, and the median vector is obtained by performing sorting statistics on each dimension of the domain coding feature; Calculating a covariance matrix and a correlation coefficient matrix of the domain coding features as second-order statistics, wherein the covariance matrix represents the relationship between different dimensions of the domain coding features, and the correlation coefficient matrix is ​​obtained by normalizing the covariance matrix; Combining the weighted mean vector and the median vector to construct a first-order statistical feature distribution representation, and combining the eigenvalues ​​and eigenvectors of the covariance matrix to construct a second-order statistical feature distribution representation; The Wasserstein distance represented by the characteristic distribution of the first-order statistics is calculated by using the Sinkhorn algorithm, and the Wasserstein distance represented by the characteristic distribution of the second-order statistics is calculated by using the matrix decomposition technique; The optimal weight coefficient is determined based on cross-validation, and the Wasserstein distance represented by the first-order statistic feature distribution and the Wasserstein distance represented by the second-order statistic feature distribution are weightedly combined to generate the inter-domain distribution difference.

8. A news event spatial verification and analysis system based on satellite remote sensing images, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to receive news event text information and extract event timestamps, geospatial coordinates, and event type identifiers from the news event text information using a natural language processing model; Determine the target observation area based on the extracted geospatial coordinates, and calculate the optimal satellite observation time window according to the event timestamp; input the target observation area and the optimal satellite observation time window into the strategy correction model to generate a satellite cluster collaborative observation strategy, wherein the satellite cluster collaborative observation strategy defines a collaborative observation sequence of high-resolution satellites, high-phase satellites, and hyperspectral satellites; issue a scheduling instruction according to the satellite cluster collaborative observation strategy to activate the corresponding satellite cluster to perform the collaborative observation task; The second unit is used to receive multi-source remote sensing image data acquired during the satellite group's collaborative observation mission, perform adaptive correction processing on the multi-source remote sensing image data, and generate standardized image data with spatiotemporal registration; select a corresponding target detection network and scene segmentation network from a deep learning model library according to the event type identifier, perform intelligent analysis on the standardized image data, and extract event-related ground features and spatiotemporal change characteristics; The extracted ground features and spatiotemporal change characteristics are input into the graph neural network, and the event feature association graph is constructed by integrating historical verification knowledge. The feature association strength value is obtained through graph reasoning calculation. The third unit is used to perform multi-dimensional cross-validation on the feature association strength value, the event timestamp, and the geographic space coordinates to construct an event authenticity evaluation index; Based on the event authenticity evaluation index and combined with historical cases in the verification knowledge base, a transfer learning method is used to calculate the event credibility score; When the event credibility score exceeds the preset benchmark credibility threshold, the authenticity of the news event is confirmed, and the satellite group collaborative observation strategy, the event feature association map, and the event authenticity evaluation index in this verification process are written into the verification knowledge base.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images