Geographic information data analysis method and system based on blockchain and federated learning, electronic device and storage medium

Abstract

The application provides a geographic information data analysis method and system based on a blockchain and federated learning, an electronic device and a storage medium, relates to the technical field of geographic information intelligent analysis, and the geographic information data analysis method and system based on the blockchain and federated learning are characterized in that geographic spatial data of a target monitoring area is acquired and converted into structured data with geographic coordinates, a geographic partition topology structure is constructed by setting an edge node according to the coordinates, nodes are grouped into blockchain fragments by generating geographic grid codes, an edge node in a fragment performs federated learning to generate an update amount, and the update amount is integrated into a fragment update amount after interaction verification; a secure multi-party computation is used to aggregate the fragment update amount to generate a global model through a preset program; the model is segmented and then synchronized to an edge node for reorganization, and new data is processed to output an analysis result, so that the geographic spatial data can be effectively analyzed and the result can be output by combining the blockchain and federated learning, through a series of data processing, node interaction and model operation.

Description

Technical Field

[0001] This application relates to the field of intelligent geographic information analysis technology, and in particular to a geographic information data analysis method, system, electronic device and storage medium based on blockchain and federated learning. Background Technology

[0002] In cross-provincial and municipal river pollution joint analysis and other wide-area environmental monitoring applications involving multiple agencies, it is necessary to acquire geographic information data in real time, comprehensively, and accurately to understand the spatial distribution, propagation paths, and changing trends of river pollution. This requires the efficient integration of multi-source data, including water quality monitoring data, geomorphological data, and meteorological data from different regions. Through in-depth analysis of geographic information, scientific basis for pollution control can be provided, enabling collaborative operations and data sharing among multiple agencies, and the development of effective cross-regional pollution prevention and control strategies.

[0003] Currently, to address these technological needs, some regions have adopted an integrated "sky-ground" monitoring system. This system integrates satellite remote sensing, UAV hyperspectral imaging, and unmanned surface vessel (USV) mobile monitoring technologies to construct a monitoring network that coordinates space, air, and ground operations. Satellite remote sensing conducts general surveys and screenings from high altitudes to promptly identify suspected pollution areas; UAVs equipped with devices conduct further detailed investigations of suspected problems; and USVs and mobile monitoring vehicles perform rapid in-situ detection on the ground or water surface to accurately verify pollution issues, thereby gaining a comprehensive and detailed understanding of the ecological environment.

[0004] However, this existing solution has significant drawbacks. Firstly, the lack of standardized data across different institutions makes integrating multi-source data extremely difficult, hindering effective data fusion and analysis and reducing its utilization value. For example, differences in the definitions, testing methods, and accuracy requirements of water quality monitoring indicators across provinces and cities lead to confusion during cross-regional data exchange. Secondly, the equipment maintenance costs in the integrated "sky-ground" monitoring system are high, and data transmission and interaction between different devices suffer from delays and interruptions, affecting the real-time performance and continuity of monitoring data and failing to provide timely support for pollution emergency response. Furthermore, this solution has limited monitoring capabilities when dealing with complex geographical environments and special pollution situations, making it difficult to meet the complex and ever-changing technical requirements of multi-institutional collaborative scenarios in wide-area environmental monitoring. Summary of the Invention

[0005] The purpose of this application is to provide a geographic information data analysis method, system, electronic device and storage medium based on blockchain and federated learning, so as to solve the problems of inconsistent data standards among multiple agencies, difficulty in integration, data transmission delay and limited monitoring capabilities in the existing integrated monitoring system.

[0006] To address the aforementioned technical problems, firstly, this application provides a geographic information data analysis method based on blockchain and federated learning, comprising:

[0007] Acquire geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates;

[0008] Based on the geographic coordinates in the structured data, multiple edge nodes are set up within the target monitoring area to construct a geographic partition topology. Geographic grid codes are generated synchronously using the geographic coordinates. The edge nodes in the geographic partition topology are grouped into blockchain shards based on the geographic grid codes.

[0009] Within the blockchain shard, each edge node performs federated learning computation to generate node update volume. At the same time, the node update volume is cross-confirmed through the interactive verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update volume. The cross-confirmed node update volume is then integrated to form the shard update volume.

[0010] An encrypted communication channel between edge nodes is established through a pre-set automatic execution program on the blockchain shard, and a secure aggregation method based on secure multi-party computation is adopted to merge the shard update amounts across multiple blockchain shards to generate a global learning model.

[0011] The global learning model is divided into multiple parts, synchronized to all edge nodes through the data transmission mechanism of the blockchain sharding, and the global learning model is recombined on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

[0012] Optionally, the step of establishing an encrypted communication channel between edge nodes through a preset automatic execution program on the blockchain shard, and using a secure aggregation method based on secure multi-party computation to fuse the shard update amounts across multiple blockchain shards to generate a global learning model, includes:

[0013] Activate the pre-set automatic execution program on the blockchain shard, generate encryption key pairs according to the network topology of the edge nodes, and establish an encrypted communication channel between the edge nodes based on the encryption key pairs;

[0014] The encrypted communication channel initiates a shard update transmission request to all blockchain shards and receives shard update data packets returned by each blockchain shard. The shard update data packets contain encrypted model parameter change data.

[0015] A secure computing group is formed by representative nodes elected by each blockchain shard. The shard update data packet is transmitted to the secure computing group through the encrypted communication channel. Within the secure computing group, the shard update data packet is split into multiple parameter fragments.

[0016] The parameter fragments are distributed among representative nodes through a parameter distribution storage mechanism, and local fusion computation is performed based on the distributed parameter fragments to generate a global learning model.

[0017] Optionally, the local fusion computation based on the allocated parameter fragments to generate a global learning model includes:

[0018] Each representative node performs a local fusion calculation operation based on the assigned parameter fragment, generating calculation results that include parameter changes;

[0019] After each round of local fusion computation is completed, the computation results of all representative nodes are collected, parameter change data are extracted, and the global change magnitude value is calculated.

[0020] The local fusion calculation and change magnitude calculation process is executed iteratively. After each iteration, a new global change magnitude value is generated. When the global change magnitude value generated three times in a row is less than a preset critical threshold, the parameter changes generated by all representative nodes are weighted and merged, and the weighted merging result is set as the final parameter of the global learning model.

[0021] Optionally, the step of dividing the global learning model into multiple parts, synchronizing it to all edge nodes through the blockchain sharding data transmission mechanism, and recombining the global learning model on the edge nodes, and processing newly acquired geospatial data based on the combined global learning model and geographic coordinates to generate geospatial distribution analysis results, includes:

[0022] The global learning model is divided into multiple model blocks, and each model block is assigned a partitioning identifier and a position sequence identifier.

[0023] Based on the node connection relationship of the geographic partition topology, the model block is sent to the blockchain shard specified by the shard ownership identifier through data relay transmission between edge nodes;

[0024] The model blocks are received by the edge nodes within the blockchain shard, and the model blocks are arranged and combined according to the numerical order of the position sequence identifiers to form a complete global learning model.

[0025] The newly acquired geospatial data is converted into structured data with geographic coordinates and input into the reorganized global learning model. Based on the geographic coordinates, the corresponding geographic grid is determined and spatial constraint calculations are performed.

[0026] The calculation results and geographic coordinates of each geographic grid are collected to generate the geospatial distribution analysis results of the target monitoring area.

[0027] Optionally, within the blockchain shard, each edge node performs federated learning computation to generate node update quantities. Simultaneously, an interactive verification mechanism among the edge nodes cross-verifies the node update quantities to confirm their integrity and source authenticity. The cross-verified node update quantities are then integrated to form shard update quantities, including:

[0028] Within the blockchain shard, each edge node performs federated learning computation based on locally stored structured data to generate a node update quantity containing model parameter change data, synchronously creates a digest identifier corresponding to the node update quantity, and broadcasts the digest identifier to all other edge nodes within the blockchain shard.

[0029] After receiving the digest identifier, the edge nodes within the blockchain shard verify the data matching between the received digest identifier and the corresponding node update amount, and at the same time verify the digital identity certificate of the sending edge node to confirm the integrity and authenticity of the node update amount.

[0030] When more than half of the edge nodes in the blockchain shard have completed the verification and confirmation of the same node update, the node update is marked as a valid update;

[0031] The update amounts of all nodes marked as valid updates within the blockchain shard are integrated to generate a shard update amount that includes the merged model parameter change data.

[0032] Optionally, the step of setting multiple edge nodes within the target monitoring area based on the geographic coordinates in the structured data to construct a geographic partitioning topology, synchronously generating geographic grid codes using the geographic coordinates, and grouping the edge nodes in the geographic partitioning topology into blockchain shards based on the geographic grid codes includes:

[0033] Based on the geographic coordinate distribution density in the structured data, multiple edge nodes are dynamically deployed within the target monitoring area, and each edge node manages a subset of data containing at least one geographic coordinate.

[0034] Establish direct communication links between the edge nodes to form a geographic partitioning topology composed of the edge nodes and the communication links;

[0035] Using the longitude, latitude, and elevation values ​​of geographic coordinates, the target monitoring area is divided into geographic grids of uniform size, and a unique geographic grid code is generated for each geographic grid. The geographic grid code includes longitude segment identifiers, latitude segment identifiers, and hierarchical identifiers.

[0036] Based on the numerical continuity of the geographic grid encoding, the edge nodes that manage adjacent geographic grids are grouped into the same node set, and each node set constitutes a blockchain shard.

[0037] Optionally, acquiring geospatial data of the target monitoring area and converting the geospatial data into structured data with geographic coordinates includes:

[0038] Acquire multi-source geospatial data of the target monitoring area, wherein the multi-source geospatial data contains location description information from at least two different sources;

[0039] The location description information is converted into numerical geographic coordinates in a unified format, which include longitude, latitude, and elevation values.

[0040] According to the preset planar division rules, the target monitoring area is divided into continuous and seamlessly covered rectangular units, and each rectangular unit has a unique unit identifier;

[0041] Each geographic coordinate is mapped to a corresponding rectangular cell, generating structured data containing the correspondence between the geographic coordinates and cell identifiers.

[0042] Secondly, this application provides a geographic information data analysis system based on blockchain and federated learning, including:

[0043] The acquisition module is used to acquire geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates;

[0044] The grouping module is used to set up multiple edge nodes within the target monitoring area based on the geographic coordinates in the structured data, construct a geographic partitioning topology, synchronously generate geographic grid codes using the geographic coordinates, and group the edge nodes in the geographic partitioning topology into blockchain shards based on the geographic grid codes.

[0045] The confirmation module is used to generate node update quantities by each edge node performing federated learning computation within the blockchain shard, and to cross-confirm the node update quantities through the interactive verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update quantities, and to integrate the cross-confirmed node update quantities to form shard update quantities.

[0046] The generation module is used to establish an encrypted communication channel between edge nodes through a preset automatic execution program on the blockchain shard, and to merge the shard update volume across multiple blockchain shards using a secure aggregation method based on secure multi-party computation to generate a global learning model.

[0047] The output module is used to divide the global learning model into multiple parts, synchronize them to all edge nodes through the data transmission mechanism of the blockchain sharding, and reorganize the global learning model on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

[0048] Thirdly, this application provides an electronic device, comprising:

[0049] Memory, used to store computer programs;

[0050] A processor, configured to execute the computer program to implement the steps of the geographic information data analysis method based on blockchain and federated learning as described in the first aspect above.

[0051] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the geographic information data analysis method based on blockchain and federated learning as described in the first aspect above.

[0052] The geographic information data analysis method based on blockchain and federated learning provided in this application lays a unified and accurate data foundation for subsequent analysis by converting geospatial data into structured data with geographic coordinates. By setting edge nodes, constructing a topology, and grouping them into blockchain shards based on geographic coordinates, it achieves refined partitioning management of target monitoring areas, improving the targeting and efficiency of data processing. Edge nodes within shards execute federated learning to generate update quantities, which are then integrated through interactive verification. This ensures the integrity and authenticity of node update quantities while fully utilizing the computing power of distributed nodes. An encrypted communication channel is established through a pre-set program, and secure multi-party computation is used to aggregate shard update quantities to generate a global model. This ensures data transmission security while achieving effective fusion of multi-shard data, improving the model's globality and accuracy. The global model is then segmented and synchronized to edge nodes for reorganization and processing of new data. This not only enhances the model's reusability and response speed but also allows for accurate output of geospatial distribution analysis results combined with geographic coordinates. Ultimately, this achieves the technical effect of efficient, secure, and accurate geographic information data analysis in multi-agency collaborative scenarios such as wide-area environmental monitoring.

[0053] The method further activates the automatic execution program on the blockchain shards, generates encrypted key pairs based on the network topology of the edge nodes, and establishes an encrypted communication channel between the edge nodes. Then, it initiates a shard update transmission request through this channel, receiving data packets containing encrypted model parameter change data returned by each shard. Next, a secure computing group is formed by representative nodes elected by each shard, and the shard update data packets are transmitted to this group and split into multiple parameter fragments. Finally, parameter fragments are distributed among the representative nodes through a parameter distributed storage mechanism, and each node performs local fusion computation based on the allocated fragments to generate a global learning model. This method ensures the security of shard update transmission by generating encrypted key pairs to establish an encrypted communication channel; the distributed processing of shard update data is achieved by representative nodes forming a secure computing group and splitting and distributing parameter fragments, improving the efficiency and security of data fusion; and the local fusion computation performed by each representative node to generate a global learning model further ensures the security and accuracy of the model generation process, effectively solving the problems of insecure data transmission, low fusion efficiency, and insufficient model generation reliability in multi-institutional collaborative scenarios. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a geographic information data analysis method based on blockchain and federated learning, provided for an embodiment of this application;

[0056] Figure 2 A schematic diagram illustrating the specific implementation process of a geographic information data analysis method based on blockchain and federated learning, provided for an embodiment of this application;

[0057] Figure 3 A system architecture diagram illustrating a specific implementation of a geographic information data analysis method based on blockchain and federated learning, provided in this application embodiment;

[0058] Figure 4 This is a schematic diagram of the structure of a geographic information data analysis system based on blockchain and federated learning, provided as an embodiment of this application. Detailed Implementation

[0059] In collaborative scenarios involving multiple agencies for wide-area environmental monitoring, such as joint analysis of river pollution across provinces and cities, the existing integrated "sky-ground" monitoring system has significant shortcomings. Inconsistent data standards among different agencies make it difficult to integrate multi-source data and fully realize its value; high equipment maintenance costs and data transmission delays and interruptions affect real-time monitoring; furthermore, its monitoring capabilities for complex geographical environments and special pollution situations are limited, failing to meet the needs of collaborative analysis.

[0060] To address the aforementioned issues, this application proposes a geographic information data analysis method based on blockchain and federated learning. This method first converts geospatial data into structured data with geographic coordinates, unifying the data foundation. Then, it establishes edge nodes based on coordinates, constructs a topology structure, and groups data into blockchain shards for refined management. Within each shard, updates are generated through federated learning and integrated through interactive verification, ensuring data authenticity and utilization efficiency. Encrypted communication and secure multi-party computation are used to aggregate shard updates to generate a global model, ensuring data security and fusion effectiveness. Finally, the global model is segmented and synchronized to edge nodes for reorganization and processing of new data, improving response speed and analytical accuracy. This solution effectively solves the problems of inconsistent data standards, transmission delays, and limited monitoring capabilities in existing systems, meeting the collaborative analysis needs of multiple agencies in wide-area environmental monitoring.

[0061] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0062] The core of this application is to provide a geographic information data analysis method based on blockchain and federated learning. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0063] S101. Obtain geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates;

[0064] Optionally, step S101 may specifically include the following steps:

[0065] S1011. Obtain multi-source geospatial data of the target monitoring area, wherein the multi-source geospatial data contains location description information from at least two different sources;

[0066] S1012. Convert the location description information into a unified format of numerical geographic coordinates, wherein the geographic coordinates include longitude, latitude and elevation values.

[0067] S1013. According to the preset planar division rules, the target monitoring area is divided into continuous and seamlessly covered rectangular units, and each rectangular unit has a unique unit identifier.

[0068] S1014. Map each geographic coordinate to the corresponding rectangular cell to generate structured data containing the correspondence between the geographic coordinates and the cell identifier.

[0069] In the above scheme, the target monitoring area refers to a specific large-scale area requiring geographic information data analysis. Geospatial data describes the location, shape, etc., of geographic features within this area. Structured data is data organized in a specific format for easy computer processing; it includes the correspondence between geographic coordinates and cell identifiers. Multi-source geospatial data is geographic information from at least two different sources, such as satellite remote sensing data and ground sensor data. Location description information indicates the location of geographic features and may include textual descriptions or coordinates in different formats. Geographic coordinates are location information expressed using longitude, latitude, and elevation values. A rectangular cell is a small, continuous, and seamless rectangular block that is divided according to rules and covers the target area. A cell identifier is a unique mark for each rectangular cell, used to distinguish different cells.

[0070] In this embodiment, multi-source geospatial data of the target monitoring area is first acquired through step S1011. This data comes from at least two different channels, such as satellite remote sensing systems, ground-deployed sensor networks, and drone aerial photography equipment. During the acquisition process, data is collected through the corresponding receiving devices or interfaces of each channel. For example, remote sensing image data of the target monitoring area is acquired from a satellite data receiving station, and the location and related environmental data of the monitoring points are received from the wireless transmission module of the ground sensor. This data contains information that can describe the location of geographic elements, providing the raw data foundation for subsequent processing.

[0071] Next, in step S1012, a coordinate transformation algorithm is used to convert various location description information into numerical geographic coordinates in a unified format. Specifically, for textual location description information such as "3 kilometers northeast of XX area", geocoding technology is first used to map it to an approximate coordinate range, and then it is corrected by combining it with known coordinate points in the surrounding area. For the pixel position of remote sensing image, the corresponding longitude, latitude, and elevation values ​​are calculated based on the image resolution, satellite attitude parameters at the time of shooting, etc., using the conversion formula from pixel coordinates to geographic coordinates. For coordinate data in different formats, such as UTM coordinates, they are converted into a unified longitude, latitude, and elevation format using preset conversion rules. Finally, all location description information becomes geographic coordinates containing longitude, latitude, and elevation values, achieving data format unification.

[0072] Then, in step S1013, the target monitoring area is divided into continuous and seamlessly overlapping rectangular units according to preset planar division rules. Each unit has a unique unit identifier. During division, the latitude and longitude range of the target monitoring area is first determined, and then the side length of the rectangular units is set according to actual needs. Subsequently, starting from the initial latitude and longitude of the target monitoring area, rectangular areas are divided sequentially in a southeast direction. These rectangular areas do not overlap and completely cover the entire target monitoring area. At the same time, a unique identifier is assigned to each rectangular unit according to its spatial location, using the format of "area code + row number + column number" to ensure that each unit can be accurately distinguished, preparing for the subsequent mapping of coordinates to units.

[0073] Finally, in step S1014, the longitude and latitude values ​​of each geographic coordinate are extracted and compared with the longitude and latitude ranges of each rectangular unit to determine which rectangular unit the coordinate falls within. For example, if a geographic coordinate has a longitude between 118.500° and 118.501° and a latitude between 32.300° and 32.301°, and the rectangular unit numbered R-003 covers this range, then the geographic coordinate is associated with R-003. Subsequently, this correspondence is organized according to a fixed format, with each record containing longitude, latitude, elevation values, and the corresponding unit identifier, forming structured data for easy subsequent storage and processing.

[0074] In practical applications, for a cross-regional ecological monitoring project, the target monitoring area is a vast area covering multiple cities and counties. First, multi-source geospatial data for the region is acquired, including remote sensing images from meteorological satellites, location information of monitoring points recorded by ground ecological monitoring stations distributed across various locations, and regional topographic data obtained from drone aerial photography. Then, the location description information in this data is transformed. For example, the pixel coordinates of a specific ecological patch in the remote sensing image, the "500 meters east of XX village" recorded by the ground monitoring station, and the relative coordinates of topographic points in the aerial photography data are all converted into unified longitude, latitude, and elevation numerical forms of geographic coordinates, i.e., (118.52°E, 32.33°N, elevation 20 meters). Next, according to a 200m × 200m planar division rule, the entire target monitoring area is divided into multiple continuous and seamless rectangular units, each assigned a unique unit identifier such as R-001, R-002, etc. Finally, for each geographic coordinate obtained earlier, such as (118.52°E, 32.33°N, elevation 20 meters), its corresponding rectangular cell is identified as R-005. Structured data containing the correspondence between the geographic coordinate and R-005 is generated, such as a record as "118.52°E, 32.33°N, elevation 20 meters; R-005".

[0075] The overall solution of S101 described above enriches the sources and types of data by acquiring multi-source geospatial data, providing a comprehensive foundation for subsequent analysis; it converts location description information into geographic coordinates in a unified format, eliminating format differences between data from different sources and achieving data standardization; it divides the vast target monitoring area into rectangular units and generates unit identifiers, enabling orderly division and facilitating management and positioning; and it generates structured data containing the correspondence between geographic coordinates and unit identifiers, making the data more organized and easier for computer processing and analysis, thus laying a solid data foundation for subsequent geographic information data analysis.

[0076] S102. Based on the geographical coordinates in the structured data, set up multiple edge nodes within the target monitoring area to construct a geographical partition topology. Simultaneously generate a geographical grid code using the geographical coordinates. Group the edge nodes in the geographical partition topology into blockchain shards based on the geographical grid code.

[0077] Optionally, step S102 may specifically include the following steps:

[0078] S1021. Based on the geographic coordinate distribution density in the structured data, multiple edge nodes are dynamically deployed within the target monitoring area, and each edge node manages a subset of data containing at least one geographic coordinate.

[0079] S1022. Establish direct communication links between the edge nodes to form a geographic partitioning topology structure composed of the edge nodes and the communication links;

[0080] S1023. Using the longitude, latitude and elevation values ​​of the geographic coordinates, the target monitoring area is divided into geographic grids of uniform size, and a unique geographic grid code is generated for each geographic grid. The geographic grid code includes longitude segment identifier, latitude segment identifier and hierarchical identifier.

[0081] S1024. Based on the numerical continuity of the geographic grid code, the edge nodes of adjacent geographic grids are grouped into the same node set, and each node set constitutes a blockchain shard.

[0082] In the above scheme, edge nodes are small computing nodes deployed within the target area to process and manage partial data. The geographic partitioning topology is a structure reflecting the spatial distribution and connectivity of nodes, composed of edge nodes and communication links between them. Geographic grid coding is a unique identifier generated for each geographic grid, including longitude segment identifiers, latitude segment identifiers, and hierarchical identifiers, used to distinguish different grids. Blockchain shards are sets of nodes composed of edge nodes managing adjacent geographic grids, facilitating partitioned data processing and management.

[0083] In this embodiment, step S1021 dynamically deploys edge nodes based on the distribution density of geographic coordinates in structured data. First, a density clustering algorithm is used to analyze the distribution of geographic coordinates, calculating the number of coordinates per unit area in different regions to determine density levels. Then, based on the density results, the deployment number and location of edge nodes are planned, increasing the number of nodes in high-density areas and decreasing the number of nodes in low-density areas. Each edge node is assigned to manage a subset of data containing at least one geographic coordinate. These data subsets are divided into geographic ranges from the structured data, preparing for subsequent node communication and data processing. For example, analysis using the DBSCAN algorithm reveals a dense distribution of geographic coordinates in the northern part of the target area and a sparse distribution in the southern part. Therefore, 10 edge nodes are deployed in the north and 5 in the south, with each node managing the coordinate data within its respective range.

[0084] Secondly, after deploying the edge nodes in step S1021, step S1022 utilizes wireless ad hoc network technology to allow each edge node to search for other nodes in its vicinity. By detecting signal strength, it determines the reachable neighboring nodes and then establishes wireless communication links between these reachable nodes, enabling nodes to directly send and receive data. The resulting geographically partitioned topology structure intuitively reflects the spatial distribution and connection relationships of each edge node, providing a channel for data transmission between nodes.

[0085] Next, in step S1023, geographic grids are divided using the longitude, latitude, and elevation values ​​of the geographic coordinates, and geographic grid codes are generated. Based on the geographic range and analysis accuracy requirements of the target monitoring area, a uniform grid size is set, such as setting each grid to span 0.02 degrees in the longitude direction, 0.02 degrees in the latitude direction, and 10 meters in the elevation direction. Then, the target monitoring area is divided into multiple non-overlapping and continuous geographic grids according to this size. Next, a unique code is generated for each grid, using the format of "longitude segment identifier + latitude segment identifier + level identifier." The longitude segment identifier consists of the longitude interval number of the grid, the latitude segment identifier consists of the latitude interval number of the grid, and the level identifier consists of the elevation interval number of the grid, providing a basis for subsequent grouping of edge nodes. For example, a grid is divided according to the dimensions of 0.02 degrees longitude, 0.02 degrees latitude, and 10 meters elevation. A code "J08W15L03" is generated for a certain grid, where J08 represents the longitude segment, W15 represents the latitude segment, and L03 represents the level.

[0086] Finally, step S1024 determines the geographic grid to which the geographic coordinates of the data subset managed by each edge node belong, thus clarifying the geographic grid code associated with each node. Then, by comparing the codes associated with each node, nodes whose longitude and latitude segment identifiers are numerically consecutive are grouped into the same node set. Each such node set constitutes a blockchain shard. In this way, edge nodes managing adjacent geographic grids can collaborate within the same shard, improving the collaboration and efficiency of data processing. For example, edge nodes associated with codes "J08W15L03", "J08W16L03", and "J09W15L03" are grouped into the same blockchain shard.

[0087] In practical applications, when analyzing a cross-regional ecological monitoring zone, the DBSCAN algorithm is first used to analyze the distribution of geographic coordinates in the structured data. It was found that the eastern part of the monitoring zone has a dense distribution of coordinates due to the presence of more wetland monitoring points, while the western mountainous area has fewer monitoring points and sparser coordinates. Therefore, eight edge nodes were deployed in the east and three in the west, with each node managing a subset of geographic coordinate data within its respective region. Next, AdHoc network technology is used to enable these edge nodes to search for reachable neighboring nodes. After signal strength detection, wireless communication links are established, forming a geographic partition topology that reflects the spatial distribution and connectivity of the nodes, ensuring direct data transmission between nodes. Subsequently, based on the required monitoring accuracy, each geographic grid was set to span 0.03 degrees in longitude, 0.03 degrees in latitude, and 15 meters in elevation. The monitoring zone was divided into multiple continuous and non-overlapping geographic grids according to these dimensions, and a unique code was generated for each grid, such as "J12W20L05," where J12 is the longitude segment identifier, W20 is the latitude segment identifier, and L05 is the elevation level identifier. Finally, the geographic grid and code corresponding to the data managed by each edge node were determined. Nodes with consecutive longitude and latitude segment identifier values ​​in the associated codes were grouped into the same set. For example, nodes associated with "J12W20L05," "J12W21L05," and "J13W20L05" formed a blockchain shard, enabling collaborative processing of edge nodes in adjacent areas.

[0088] The overall solution of S102 described above dynamically deploys edge nodes based on the distribution density of geographic coordinates, which makes the node distribution more reasonable and improves the efficiency of data processing; it establishes a geographic partition topology structure, which makes communication between edge nodes smoother and facilitates data interaction and sharing; it divides geographic grids and generates codes, which makes the management of target areas more orderly and provides a clear basis for node grouping; and it groups the edge nodes of adjacent grids into blockchain shards, which is conducive to realizing partitioned data processing and collaborative computing, and improves the overall effect of geographic information data analysis.

[0089] S103. Within the blockchain shard, each edge node performs federated learning computation to generate node update volume. At the same time, the node update volume is cross-confirmed through the interaction verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update volume. The cross-confirmed node update volume is then integrated to form the shard update volume.

[0090] Optionally, step S103 may specifically include the following steps:

[0091] S1031. Within the blockchain shard, each edge node performs federated learning computation based on locally stored structured data to generate a node update quantity containing model parameter change data, synchronously creates a digest identifier corresponding to the node update quantity, and broadcasts the digest identifier to all other edge nodes within the blockchain shard.

[0092] S1032. After receiving the digest identifier, the edge node within the blockchain shard verifies the data matching between the received digest identifier and the corresponding node update amount, and at the same time verifies the digital identity certificate of the sending edge node to confirm the integrity and authenticity of the node update amount.

[0093] S1033. When more than half of the edge nodes in the blockchain shard have completed the verification and confirmation of the same node update, the node update is marked as a valid update.

[0094] S1034. Integrate the update amounts of all nodes marked as valid updates within the blockchain shard to generate a shard update amount that includes the merged model parameter change data.

[0095] In the above scheme, federated learning computation is a machine learning approach where each node processes data locally, sharing only model parameter updates and not the original data. Node update quantity refers to the model parameter changes obtained by edge nodes through federated learning computation. The interactive verification mechanism is a mechanism for edge nodes to mutually check and confirm data. A digest identifier is a unique identifier for node update quantities, used for matching and verification. Digital identity credentials are the identity proof of the edge node, used to verify its origin. Valid updates are node update quantities verified and confirmed by more than half of the nodes. Shard update quantity is the merged model parameter change data obtained by aggregating the node update quantities of all valid updates in a shard.

[0096] In this embodiment of the application, firstly, in step S1031, within the blockchain shard, each edge node calls the locally stored structured data and performs federated learning computation using a federated averaging algorithm. That is, each node trains the model with its own data and obtains the change value of the model parameters, which is the node update amount. The calculation method is to subtract the parameters of the previous round from the parameters of the current round. The calculation formula is as follows: ,in Indicates the node update amount. This is the parameter for the current round. This refers to the parameters from the previous round. Simultaneously, the SHA-256 hash algorithm is used to encrypt and calculate the node's update, generating a unique string as a digest identifier. The formula is... H is the digest identifier, which is then broadcast to all other edge nodes in the shard via the communication link in the shard, so that they know that there is a new node update that needs to be verified.

[0097] Secondly, in step S1032, after other edge nodes in the shard receive the broadcast digest identifier, they first obtain the corresponding node update amount from the sending node, and then recalculate the update amount using the same SHA-256 hash algorithm to obtain a new digest identifier. The formula is as follows: ,in This is the amount of node updates received. This is the newly calculated digest identifier, which is compared with the received digest identifier H. If they match, it means that the node update has not been tampered with during transmission, ensuring integrity. Simultaneously, the digital certificate of the sending node is checked to verify its validity and whether it was issued by a trusted authority, thereby confirming the authenticity of the sending node's identity and ensuring the reliability of the node update source.

[0098] Next, in step S1033, after a node completes the above verification, the verifying node will report the verification result to other nodes in the shard. The system will count the verification results for the node's update. If there are a total of N edge nodes in the shard, when the number of nodes k that have passed verification satisfies... When the update is considered to have been approved by a majority of nodes, it is marked as a valid update. This avoids inaccurate verification results due to errors or malicious interference from individual nodes.

[0099] Finally, after determining all valid updates in the shards through step S1034, the system collects these node update amounts and integrates them using a weighted average method. First, corresponding weights are assigned based on the local data volume or computing power of each edge node; nodes with larger data volumes and stronger computing power may have higher weights. Then, the update amount of each valid update is multiplied by its corresponding weight, all products are summed, and then divided by the sum of the weights. The calculation formula is as follows: ,in For the amount of data updated per shard, It is the weight of the i-th valid update. It is the update amount of the i-th valid update node. The merged model parameter change data is also the shard update amount, which prepares for the subsequent model fusion between multiple shards.

[0100] In practical applications, a blockchain shard contains four edge nodes. Node 1 accesses local structured data and performs federated learning computation using a federated averaging algorithm, retrieving the model parameters from the previous round. The value is 0.2, representing the current round parameter. It is 0.5, according to the formula Calculate the node update amount Simultaneously, the digest identifier "EF78" is generated from 0.3 using the SHA-256 hash algorithm and broadcast. Other nodes, upon receiving this, obtain the update value 0.3 from node 1 and calculate the result using the same algorithm. The result matches "EF78", and the digital certificate of node 1 is verified to be valid. Nodes 2 and 3 also completed similar calculations. Node 2's... Node 3 Node 4 did not generate a valid update; according to the system's statistical verification results, 3 out of 4 nodes passed the verification, satisfying the requirements. Therefore Updates are marked as valid; finally, they are integrated using a weighted average method, with weights assigned to the three valid updates. Both are 1, according to the formula Calculate the amount of data updated per shard. .

[0101] The overall scheme of S103 described above allows each edge node to perform federated learning computation locally, avoiding direct sharing of raw data and protecting data privacy; it generates summary identifiers and performs interactive verification, ensuring the integrity of node update volume and the authenticity of its source; it uses verification by more than half of the nodes as the standard for valid updates, improving the reliability of verification results; and it integrates valid updates to form shard update volume, realizing the effective fusion of data in shards and providing accurate and reliable basic data for the subsequent generation of the global model.

[0102] S104. An encrypted communication channel between edge nodes is established through the preset automatic execution program on the blockchain shard, and a secure aggregation method based on secure multi-party computation is adopted to merge the shard update volume among multiple blockchain shards to generate a global learning model.

[0103] Optionally, step S104 may specifically include the following steps:

[0104] S1041. Activate the pre-set automatic execution program on the blockchain shard, generate encryption key pairs according to the network topology of the edge nodes, and establish an encrypted communication channel between the edge nodes based on the encryption key pairs.

[0105] S1042. Initiate a shard update transmission request to all blockchain shards through the encrypted communication channel, and receive the shard update data packets returned by each blockchain shard, wherein the shard update data packets contain encrypted model parameter change data.

[0106] S1043. A secure computing group is formed by representative nodes elected by each blockchain shard. The shard update data packet is transmitted to the secure computing group through the encrypted communication channel. The shard update data packet is split into multiple parameter fragments within the secure computing group.

[0107] S1044. The parameter fragments are allocated among the representative nodes through a parameter distribution storage mechanism, and local fusion calculations are performed based on the allocated parameter fragments to generate a global learning model.

[0108] Specifically, step S1044 includes the following process: each representative node performs a local fusion calculation operation based on the assigned parameter fragment to generate a calculation result containing parameter changes; after each round of local fusion calculation, the calculation results of all representative nodes are collected, parameter change data is extracted, and the global change magnitude value is calculated; the local fusion calculation and change magnitude calculation process is iteratively executed, and a new global change magnitude value is generated after each iteration; when the global change magnitude value generated three consecutive times is less than a preset critical threshold, the parameter changes generated by all representative nodes are weighted and merged, and the weighted merging result is set as the final parameter of the global learning model.

[0109] In the above scheme, the automatic execution program is a program pre-set on the blockchain shard that can automatically trigger and execute specific operations. An encryption key pair is a combination of public and private keys used for encryption and decryption; the public key can be publicly used for encryption, while the private key is kept by the holder for decryption. An encrypted communication channel is a secure data transmission path established between edge nodes using encryption technology. A shard update data packet is a data carrier containing encrypted shard update values. A representative node is an edge node elected by each blockchain shard to represent that shard in global computation. A secure computation group is a group composed of representative nodes from each shard, responsible for performing secure aggregate computation. A parameter fragment is a partial parameter data obtained by splitting the shard update data packet. A parameter distributed storage mechanism is a method of distributing parameter fragments to different representative nodes for storage. A global learning model is a model obtained by fusing multiple shard updates and can be used for global analysis. Parameter change is the change in model parameters during fusion computation. The global change magnitude is a value reflecting the overall degree of change of all parameter changes. A preset critical threshold is a reference value for determining whether the model has converged.

[0110] In the embodiments of this application, such as Figure 2As shown, step S1041 first activates the pre-set automatic execution program on the blockchain shard, scans the network topology of the edge nodes, clarifies the connection status and location distribution of each node, and then calls the RSA asymmetric encryption algorithm to generate encryption key pairs. Each edge node receives a public key and a private key pair. The public key is publicly available, while the private key is kept secret by the node. Subsequently, the program constructs encrypted communication rules based on these key pairs. That is, when a node sends data, it must encrypt it with the public key of the receiving node, and the receiving node decrypts it with its own private key. In this way, an encrypted communication channel is established between all edge nodes, ensuring that subsequent data transmission will not be stolen or tampered with by unauthorized parties.

[0111] Secondly, step S1042 utilizes the established encrypted communication channel to send shard update transmission requests to all blockchain shards. The request information is transmitted after being signed by the sending node's private key and encrypted by the receiving shard's public key. Upon receiving the request, each blockchain shard extracts the shard update amount locally, encrypts it using the requester's public key, attaches its own digital signature, packages it into a shard update amount data packet containing encrypted model parameter change data, and then returns it to the requester through the encrypted communication channel. After receiving the data packet, the requester decrypts it with its own private key and verifies the shard's digital signature, confirming that the data packet's source is reliable and has not been tampered with, thus completing the collection of shard update amounts.

[0112] Then, in step S1043, each blockchain shard elects 1-2 representative nodes through democratic voting. These representative nodes need to have high computing power and good communication capabilities, and they together form a secure computing group. Each shard transmits the received shard update data packets to the secure computing group through an encrypted communication channel. Within the computing group, a secret sharing algorithm is used to split each data packet into multiple parameter fragments. During the splitting, it is ensured that each fragment alone cannot reconstruct the complete shard update amount, and that all fragments combined can accurately recover the original data. For example, the shard update amount of 0.6 is split into three parameter fragments: 0.2, 0.3, and 0.1. A single fragment cannot reveal the original update amount, but combining them yields the accurate value.

[0113] Finally, in step S1044, a parameter distributed storage mechanism is used to randomly allocate the split parameter fragments to different representative nodes of the secure computing group. Each node only stores a portion of the fragments and cannot access the complete data. Each representative node performs local fusion calculations based on its assigned parameter fragments, such as adding its fragments to related fragments transmitted from other nodes via encrypted channels, to obtain a calculation result that includes parameter changes. After each round of calculation, all parameter changes are collected and processed according to the formula. Calculate the global change magnitude, where G is the global change magnitude and n is the number of parameter changes. For the i-th change, repeat this process until the G value is less than the preset critical threshold for three consecutive times. At this point, weights are assigned according to the amount of data in the partition where each representative node is located, using the formula... We weight and combine all parameter changes to generate a global learning model, where F is the final parameter. As weight, This represents the change in parameters.

[0114] In a practical application, a cross-regional environmental monitoring system has three blockchain shards. Activating the automatic execution program in each shard allows the program to generate key pairs using the RSA algorithm based on node connectivity, establishing encrypted communication channels between edge nodes. Requests are sent to the three shards through these encrypted channels. Shard 1 returns an encrypted shard update data packet containing... Fragment 2 returns containing The data packet, fragment 3, returns containing The data packets; then each fragment elects one representative node to form a secure computing group, and each representative node splits the data packets, fragment 1... Split into 2-shard Split into 3-part Split into The parameters are allocated to three representative nodes for storage; each representative node calculates its assigned parameter fragment, and the parameter changes obtained in the first round are 0.7, 0.8, and 0.9 respectively. The global change magnitude is then calculated. The second round of iterations The third round The fourth round If the G value is less than 0.01 three times consecutively, then the values ​​are weighted and merged with a weight of 1 each to obtain the final parameters. , which serves as the final parameter of the global learning model.

[0115] The overall solution of S104 described above ensures the security of data transmission and prevents information leakage or tampering by automatically generating encryption key pairs and establishing encrypted communication channels. It requests and receives data packets of shard update quantities from each shard, realizing the aggregation of multi-shard data. It elects representative nodes to form a secure computing group and splits parameter fragments, reducing the risk of a single node possessing complete data and enhancing data security. It adopts parameter distributed storage and local fusion computing, combined with iterative judgment of model convergence, to ensure the accuracy and reliability of the global learning model. At the same time, through secure multi-party computation, it achieves model fusion without disclosing the specific data of each shard, meeting the data security and sharing requirements in multi-organization collaborative scenarios.

[0116] S105. The global learning model is divided into multiple parts, synchronized to all edge nodes through the data transmission mechanism of the blockchain sharding, and the global learning model is recombined on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

[0117] Optionally, step S105 may specifically include the following steps:

[0118] S1051. Divide the global learning model into multiple model blocks, and attach a partitioning identifier and a position sequence identifier to each model block;

[0119] S1052. Based on the node connection relationship of the geographic partition topology, the model block is sent to the blockchain shard specified by the shard ownership identifier through data relay transmission between edge nodes;

[0120] S1053. The model blocks are received by the edge nodes within the blockchain shard, and the model blocks are arranged and combined according to the numerical order of the position sequence identifiers to form a complete global learning model.

[0121] S1054. Convert the newly acquired geospatial data into structured data with geographic coordinates, and input it into the reorganized global learning model. Determine the corresponding geographic grid based on the geographic coordinates and perform spatial constraint calculations.

[0122] S1055. Collect the calculation results and geographic coordinates of each geographic grid to generate the geospatial distribution analysis results of the target monitoring area.

[0123] In the above scheme, a model block is a portion of the model data obtained after segmenting the global learning model. A shard attribution identifier is a mark attached to the model block to indicate the blockchain shard to which it belongs. A location sequence identifier is a mark attached to the model block to indicate its positional order within the global learning model. Data relay transmission is a method of transmitting data sequentially between edge nodes. The reconstructed global learning model is the complete global learning model restored by splicing and combining the model blocks according to their location sequence identifiers. Newly acquired geospatial data is the geographic information collected subsequently for analysis. Spatial constraint calculation is a calculation performed on the data in accordance with spatial rules, combining geographic coordinates. The geospatial distribution analysis results are analytical conclusions reflecting the distribution of geographic elements in the target monitoring area.

[0124] In this embodiment, step S1051 first uses a model segmentation algorithm to divide the global learning model into multiple model blocks. During segmentation, the hierarchical structure and parameter relationships of the model are considered to ensure that each model block is functionally independent and of balanced size. After segmentation, a sharding identifier and a location sequence identifier are added to each model block. The sharding identifier is determined based on the processing capacity and scope of each blockchain shard. For example, shard 1, which processes data from the northern region, is responsible for receiving model blocks related to the north. The location sequence identifier is numbered according to the logical order of the model blocks in the global learning model, such as "1", "2", "3", etc., to ensure accurate subsequent splicing.

[0125] Secondly, step S1052, based on the connection relationship of edge nodes in the geographic partitioning topology, first determines the transmission path of each model block with a sharding ownership identifier. Path selection is based on the communication quality and distance between nodes, prioritizing nodes with stable connections and closer distances as relay nodes. Then, through data relay transmission, the model block starts from the initial node and passes through the intermediate nodes on the path in sequence. After receiving the model block, each intermediate node verifies whether its sharding ownership identifier is consistent with the target shard. If it is correct, it is then passed to the next node until it reaches the blockchain shard specified by the sharding ownership identifier.

[0126] Next, in step S1053, after the edge node in the blockchain shard receives the model block, it checks the integrity of the model block to see if there is any missing or corrupted data. If a problem is found, it requests retransmission. After the check is successful, the position sequence identifier on the model block is extracted, and all received model blocks are sorted in ascending order of identifier value. Then, the sorted model blocks are connected sequentially using a model splicing algorithm to restore a complete global learning model. During the splicing process, the interfaces of adjacent model blocks are verified to ensure that the recombined model functions normally.

[0127] Then, in step S1054, the newly acquired geospatial data is transformed using the same method as in step S101, extracting its location information and converting it into geographic coordinates containing longitude, latitude, and elevation, thus organizing it into structured data. This structured data is then input into the reorganized global learning model. The model searches for the corresponding geographic grid based on the geographic coordinates in the data, and then performs spatial constraint calculations within the spatial range of that grid, such as calculating the maximum, minimum, or probability distribution of the data within the grid, ensuring that the calculation results conform to the spatial characteristics of the geographic grid.

[0128] Finally, the spatial constraint calculation results of all geographic grids are collected through step S1055, and the geographic coordinates corresponding to each result are recorded. The data aggregation algorithm is used to integrate these scattered results. For areas without data, reasonable inferences are made based on the results of surrounding grids. Finally, a geospatial distribution analysis result that can intuitively reflect the distribution pattern of geographic elements within the target monitoring area is generated, such as in the form of charts or heat maps.

[0129] In practical applications, a hierarchical segmentation algorithm is used to divide the global learning model into 6 model blocks, each labeled "Segment A + 1", "Segment A + 2", "Segment B + 1", "Segment B + 2", "Segment B + 3", and "Segment C + 1". Based on the geographic partitioning topology, the model blocks of "Segment A" are transmitted via node A → node B → node C of Segment A; the model blocks of "Segment B" are transmitted via node A → node D → node E of Segment B; and the model blocks of "Segment C" are transmitted via node A → node F → node G of Segment C. Then, node C of Segment A is divided into blocks in the order of "1", "2", etc. The model blocks are assembled by piecing together nodes E of segment B in the order of "1", "2", and "3", and reassembling nodes G of segment C in the order of "1", thus restoring a complete global learning model. Next, the newly acquired atmospheric pollutant monitoring data is converted into structured data with geographic coordinates and input into the reassembled model. The model determines the corresponding "grid H", "grid I", and "grid J" based on the coordinates and calculates the pollutant concentration for each grid. The concentration results and corresponding coordinates of these grids are collected and integrated using a spatial interpolation algorithm. Reasonable inferences are made for areas without data, generating an atmospheric pollutant distribution analysis map for that area.

[0130] The overall solution of S105 described above divides the global learning model into model blocks and adds identifiers to facilitate efficient model transmission and accurate attribution; it sends model blocks to designated fragments through data relay transmission, utilizing the connection relationship between nodes to improve transmission efficiency and reliability; it reassembles the model according to the location sequence identifier to ensure the integrity and accuracy of the global learning model; it transforms and performs spatial constraint calculations on new data to enable the data to be adapted to the model for effective analysis; and it aggregates the results to generate geospatial distribution analysis results, providing intuitive and comprehensive conclusions for the geographic information analysis of the target monitoring area, meeting the needs for understanding geospatial distribution in multiple scenarios.

[0131] The following is a complete example for steps 101-105, such as Figure 3As shown, in a cross-regional river pollution monitoring project, water quality monitoring data, satellite remote sensing images, and UAV aerial photographs were acquired from locations A, B, and C, including location descriptions of each monitoring point. These location descriptions were converted into unified latitude and longitude coordinates, such as a monitoring point in location A being located at (116.2°E, 30.5°N, elevation 20 meters). Rectangular units were divided according to longitude and latitude of 0.02°, labeled as G-001, G-002, etc., and the above coordinates were mapped to G-001, forming structured data such as "116.2°, 30.5°, 20 meters; G-001".

[0132] Secondly, density clustering algorithm was used to find that the monitoring points in area A were dense, so 5 edge nodes were deployed, and 3 nodes were deployed in areas B and C respectively. Each node manages the data of the corresponding area. AdHoc technology was used to establish node communication links to form a geographical partition topology. Geographic grids were divided according to longitude 0.03° and latitude 0.03°, and codes were generated, such as “J05W12L01”. Edge nodes of units such as G-001 were grouped into the same blockchain shard to form 3 shards.

[0133] Next, the node update amount is calculated for each shard edge node using the federated averaging algorithm. For example, node 1 of shard 1 calculates... The digest identifier "AB12" is generated and broadcast; other nodes verify the match and digital certificate; shard 1 has 5 nodes, 3 of which pass verification. Marked as valid; then collect 3 valid updates (0.3, 0.4, 0.5), and calculate the fragment update amount according to a weight of 1:1:1. .

[0134] Then, the automatic execution program is activated to generate a key pair using the RSA algorithm to establish an encrypted channel; requests are sent to the three shards, and each shard receives a key pair containing... The encrypted data packet; representatives from each fragment elect a secure computing group to split 0.4 into fragments such as 0.1, 0.1, and 0.2; after allocating the fragments, the representative nodes calculate the parameter changes of 0.6, 0.7, and 0.8, and the first round of global change magnitude values. After iteration, the G values ​​were 0.008, 0.007, and 0.006 for three consecutive times, all of which were less than the threshold of 0.01. The values ​​were then merged according to a weight ratio of 1:1:1 to obtain the final parameter 2.1, thus generating a global learning model.

[0135] Finally, the global model is divided into 4 model blocks, labeled with "segment 1+1" and other identifiers; the data is relayed to the corresponding segments through nodes and the model is reassembled according to the sequence identifier; the new monitoring data is converted into structured data and input into the model to calculate the corresponding grid pollution index; the results are summarized to generate a spatial distribution analysis map of river pollution.

[0136] Figure 4This application provides a schematic diagram illustrating a specific implementation of a geographic information data analysis system based on blockchain and federated learning, as shown in the following embodiments. Figure 4 The system may include:

[0137] The acquisition module 41 is used to acquire geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates;

[0138] Grouping module 42 is used to set up multiple edge nodes in the target monitoring area according to the geographic coordinates in the structured data, construct a geographic partition topology, generate geographic grid codes using the geographic coordinates, and group the edge nodes in the geographic partition topology into blockchain shards according to the geographic grid codes.

[0139] The confirmation module 43 is used to perform federated learning computation to generate node update volume within the blockchain shard, and to cross-confirm the node update volume through the interactive verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update volume, and to integrate the cross-confirmed node update volume to form the shard update volume.

[0140] The generation module 44 is used to establish an encrypted communication channel between edge nodes through a preset automatic execution program on the blockchain shard, and to merge the shard update volume between multiple blockchain shards and generate a global learning model by adopting a secure aggregation method based on secure multi-party computation.

[0141] Output module 45 is used to divide the global learning model into multiple parts, synchronize them to all edge nodes through the data transmission mechanism of the blockchain sharding, and reorganize the global learning model on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

[0142] The geographic information data analysis system based on blockchain and federated learning in this application is used to implement the aforementioned geographic information data analysis method based on blockchain and federated learning. Therefore, the specific implementation of the geographic information data analysis system based on blockchain and federated learning can be found in the embodiment section of the geographic information data analysis method based on blockchain and federated learning mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0143] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described methods for analyzing geographic information data based on blockchain and federated learning.

[0144] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described methods for analyzing geographic information data based on blockchain and federated learning.

[0145] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0146] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the geographic information data analysis method based on blockchain and federated learning.

[0147] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0148] The foregoing has provided a detailed description of a geographic information data analysis method, system, electronic device, and storage medium based on blockchain and federated learning provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A geographic information data analysis method based on blockchain and federated learning, characterized in that, include: Acquire geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates; Based on the geographic coordinates in the structured data, multiple edge nodes are set up within the target monitoring area to construct a geographic partition topology. Geographic grid codes are generated synchronously using the geographic coordinates. The edge nodes in the geographic partition topology are grouped into blockchain shards based on the geographic grid codes. Within the blockchain shard, each edge node performs federated learning computation to generate node update volume. At the same time, the node update volume is cross-confirmed through the interactive verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update volume. The cross-confirmed node update volume is then integrated to form the shard update volume. An encrypted communication channel between edge nodes is established through a pre-set automatic execution program on the blockchain shard, and a secure aggregation method based on secure multi-party computation is adopted to merge the shard update amounts across multiple blockchain shards to generate a global learning model. The global learning model is divided into multiple parts, synchronized to all edge nodes through the data transmission mechanism of the blockchain sharding, and the global learning model is recombined on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

2. The method according to claim 1, characterized in that, The process involves establishing encrypted communication channels between edge nodes through a pre-set automated execution program on the blockchain shard, and employing a secure aggregation method based on secure multi-party computation to fuse shard update volumes across multiple blockchain shards to generate a global learning model, including: Activate the pre-set automatic execution program on the blockchain shard, generate encryption key pairs according to the network topology of the edge nodes, and establish an encrypted communication channel between the edge nodes based on the encryption key pairs; The encrypted communication channel initiates a shard update transmission request to all blockchain shards and receives shard update data packets returned by each blockchain shard. The shard update data packets contain encrypted model parameter change data. A secure computing group is formed by representative nodes elected by each blockchain shard. The shard update data packet is transmitted to the secure computing group through the encrypted communication channel. Within the secure computing group, the shard update data packet is split into multiple parameter fragments. The parameter fragments are distributed among representative nodes through a parameter distribution storage mechanism, and local fusion computation is performed based on the distributed parameter fragments to generate a global learning model.

3. The method according to claim 2, characterized in that, The local fusion computation based on the assigned parameter fragments is performed to generate a global learning model, including: Each representative node performs a local fusion calculation operation based on the assigned parameter fragment, generating calculation results that include parameter changes; After each round of local fusion computation is completed, the computation results of all representative nodes are collected, parameter change data are extracted, and the global change magnitude value is calculated. The local fusion calculation and change magnitude calculation process is executed iteratively. After each iteration, a new global change magnitude value is generated. When the global change magnitude value generated three times in a row is less than a preset critical threshold, the parameter changes generated by all representative nodes are weighted and merged, and the weighted merging result is set as the final parameter of the global learning model.

4. The method according to claim 1, characterized in that, The process involves dividing the global learning model into multiple parts, synchronizing them to all edge nodes via the blockchain sharding data transmission mechanism, and recombining the global learning model on the edge nodes. Based on the combined global learning model and newly acquired geospatial data processed using geographic coordinates, geospatial distribution analysis results are generated, including: The global learning model is divided into multiple model blocks, and each model block is assigned a partitioning identifier and a position sequence identifier. Based on the node connection relationship of the geographic partition topology, the model block is sent to the blockchain shard specified by the shard ownership identifier through data relay transmission between edge nodes; The model blocks are received by the edge nodes within the blockchain shard, and the model blocks are arranged and combined according to the numerical order of the position sequence identifiers to form a complete global learning model. The newly acquired geospatial data is converted into structured data with geographic coordinates and input into the reorganized global learning model. Based on the geographic coordinates, the corresponding geographic grid is determined and spatial constraint calculations are performed. The calculation results and geographic coordinates of each geographic grid are collected to generate the geospatial distribution analysis results of the target monitoring area.

5. The method according to claim 1, characterized in that, Within the blockchain shard, each edge node performs federated learning computation to generate node update quantities. Simultaneously, an interactive verification mechanism among the edge nodes cross-verifies these node update quantities to confirm their integrity and authenticity. The cross-verified node update quantities are then integrated to form the shard update quantities, including: Within the blockchain shard, each edge node performs federated learning computation based on locally stored structured data to generate a node update quantity containing model parameter change data, synchronously creates a digest identifier corresponding to the node update quantity, and broadcasts the digest identifier to all other edge nodes within the blockchain shard. After receiving the digest identifier, the edge nodes within the blockchain shard verify the data matching between the received digest identifier and the corresponding node update amount, and at the same time verify the digital identity certificate of the sending edge node to confirm the integrity and authenticity of the node update amount. When more than half of the edge nodes in the blockchain shard have completed the verification and confirmation of the same node update, the node update is marked as a valid update; The update amounts of all nodes marked as valid updates within the blockchain shard are integrated to generate a shard update amount that includes the merged model parameter change data.

6. The method according to claim 1, characterized in that, The process involves setting up multiple edge nodes within the target monitoring area based on the geographic coordinates in the structured data, constructing a geographic partitioning topology, simultaneously generating geographic grid codes using the geographic coordinates, and grouping the edge nodes in the geographic partitioning topology into blockchain shards based on the geographic grid codes, including: Based on the geographic coordinate distribution density in the structured data, multiple edge nodes are dynamically deployed within the target monitoring area, and each edge node manages a subset of data containing at least one geographic coordinate. Establish direct communication links between the edge nodes to form a geographic partitioning topology composed of the edge nodes and the communication links; Using the longitude, latitude, and elevation values ​​of geographic coordinates, the target monitoring area is divided into geographic grids of uniform size, and a unique geographic grid code is generated for each geographic grid. The geographic grid code includes longitude segment identifiers, latitude segment identifiers, and hierarchical identifiers. Based on the numerical continuity of the geographic grid encoding, the edge nodes that manage adjacent geographic grids are grouped into the same node set, and each node set constitutes a blockchain shard.

7. The method according to claim 1, characterized in that, The process of acquiring geospatial data of the target monitoring area and converting the geospatial data into structured data with geographic coordinates includes: Acquire multi-source geospatial data of the target monitoring area, wherein the multi-source geospatial data contains location description information from at least two different sources; The location description information is converted into numerical geographic coordinates in a unified format, which include longitude, latitude, and elevation values. According to the preset planar division rules, the target monitoring area is divided into continuous and seamlessly covered rectangular units, and each rectangular unit has a unique unit identifier; Each geographic coordinate is mapped to a corresponding rectangular cell, generating structured data containing the correspondence between the geographic coordinates and cell identifiers.

8. A geographic information data analysis system based on blockchain and federated learning, characterized in that, include: The acquisition module is used to acquire geospatial data of the target monitoring area and convert the geospatial data into structured data with geographic coordinates; The grouping module is used to set up multiple edge nodes within the target monitoring area based on the geographic coordinates in the structured data, construct a geographic partitioning topology, synchronously generate geographic grid codes using the geographic coordinates, and group the edge nodes in the geographic partitioning topology into blockchain shards based on the geographic grid codes. The confirmation module is used to generate node update quantities by each edge node performing federated learning computation within the blockchain shard, and to cross-confirm the node update quantities through the interactive verification mechanism between the edge nodes to confirm the integrity and authenticity of the source of the node update quantities, and to integrate the cross-confirmed node update quantities to form shard update quantities. The generation module is used to establish an encrypted communication channel between edge nodes through a preset automatic execution program on the blockchain shard, and to merge the shard update volume across multiple blockchain shards using a secure aggregation method based on secure multi-party computation to generate a global learning model. The output module is used to divide the global learning model into multiple parts, synchronize them to all edge nodes through the data transmission mechanism of the blockchain sharding, and reorganize the global learning model on the edge nodes. Based on the combined global learning model and geographic coordinates, the newly acquired geospatial data is processed, and the geospatial distribution analysis results are output.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the geographic information data analysis method based on blockchain and federated learning as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the geographic information data analysis method based on blockchain and federated learning as described in any one of claims 1 to 7.

Images