Forestry information sharing method and system based on multi-source data fusion and storage medium

Abstract

The invention relates to the field of forestry information management, in particular to a forestry information sharing method and system based on multi-source data fusion and a storage medium, and the method comprises the following steps: classifying multi-source data according to data types and importance, compressing raster data, fragmenting and structuring pattern spot data according to administrative divisions; associating the sample plot data with the pattern spot identification code and removing redundancy; a hybrid architecture is established, a core data master node adopts three-copy backup, general data slave nodes are subjected to fragmentation storage according to types, and a multi-dimensional index is established; data transmission is carried out by adopting a multiplexing network transmission protocol, data is updated based on differential transmission, and a transmission bandwidth is dynamically allocated; dividing a basic role and a special role, setting an authorized IP white list, and encrypting and transmitting secret-related data; and the RESTfulAPI shared by multiple departments is published. The data fusion precision and sharing convenience are synchronously improved, and efficient and safe technical support is provided for integrated general survey.

Description

Technical Field

[0001] This invention belongs to the field of forestry information management, specifically relating to a method, system, and storage medium for sharing forestry information through multi-source data fusion. Background Technology

[0002] In the current integrated forestry census, multi-source data fusion and sharing have become crucial for improving census efficiency, but they face multiple technical bottlenecks. The sheer volume of data, including high-resolution remote sensing imagery, millions of map patches, and tens of thousands of sample plots, overwhelms the storage and computational demands of conventional hardware and software, leading to low efficiency in data processing and sharing. Raster data lacks efficient compression formats, and outdated overlay and transmission technologies result in slow transmission speeds when sharing across regions, severely hindering collaborative census efforts across multiple departments. Furthermore, the lack of technical support for defining data access permissions across departments creates a significant conflict between ensuring core data security and meeting cross-departmental data access needs. These issues restrict the timeliness and security of forestry information sharing, necessitating the construction of an efficient and secure multi-source data fusion and sharing system to support the integrated census. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and storage medium for forestry information sharing based on multi-source data fusion, so as to solve the problems mentioned in the background art.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A method for forestry information sharing through multi-source data fusion, comprising the following steps: Multi-source data are classified into levels according to data type and importance; raster data is compressed; patch data is structured into plots according to administrative regions; and sample plot data is associated with patch identification codes and redundancy is removed. A hybrid architecture is built that combines distributed storage with structured data. The core data master node is backed up with three replicas, while general data is stored in shards by type on the slave nodes, and multi-dimensional indexes are established. Data transmission is performed using a multiplexed network transmission protocol, data is updated based on differential transmission, high-frequency data is cached at edge nodes, and transmission bandwidth is dynamically allocated. Define basic and special roles, assign corresponding permissions to departments, set up an authorized IP whitelist, and encrypt the transmission of classified data. Data fusion and remote sensing inversion verification were completed using Spark parallel computing, and a RESTful API was released for sharing by multiple departments.

[0005] Furthermore, the aforementioned hybrid architecture for building distributed storage and structured data association specifically involves building a storage architecture that combines HDFS and PostgreSQL. HDFS carries large-capacity raster data and structured patch and sample plot data files, while PostgreSQL stores structured data with spatial attributes. HDFS and PostgreSQL achieve metadata synchronization and association calls through an interface. A master-slave node cluster and regional node hierarchical deployment are adopted, with the master node cluster ensuring the core data read and write speed, and the slave node cluster being deployed in a distributed manner according to administrative divisions.

[0006] Furthermore, the general data is stored in fragments by type from the nodes. Specifically, the general data is stored in three levels of general shared data from the nodes. First, storage units are defined according to administrative divisions, and then further subdivided into raster, vector, attribute, and document data fragments according to data types. Each fragment has corresponding storage content and format specifications. The fragments are associated with unique identifiers containing administrative region code information, an association mapping table is established, and the data is synchronized to the master node database for traceability and association with core data.

[0007] Furthermore, the establishment of a multi-dimensional index specifically involves constructing a spatial, temporal, keyword, and administrative affiliation index system based on the spatial characteristics, temporal attributes, and business needs of forestry data. The spatial index is constructed for data with spatial coordinates, the temporal index is centered on the data generation and update time, the keyword index focuses on high-frequency retrieval fields, and the administrative affiliation index is associated with administrative divisions and data fragments. Data changes trigger index updates, and index fragments are regularly organized and the structure is optimized.

[0008] Furthermore, the differential transmission-based data update specifically involves the following: the differential transmission update object is the pre-processed full multi-source forestry data; a full initialization and incremental update mode is adopted; the update content is identified by data type differentiation comparison; after extraction, an incremental data packet containing update type and associated identification information is encapsulated; during transmission, the data packet enjoys bandwidth priority; after receiving, the edge node synchronously updates the cache; the receiving end verifies the integrity of the data packet; if it passes, it is automatically merged; if it fails, it triggers breakpoint resume transmission.

[0009] Furthermore, the dynamic allocation of transmission bandwidth specifically involves constructing a dynamic bandwidth allocation mechanism based on data priority and network status, setting high, medium, and low transmission priorities according to data levels, using a bandwidth scheduling engine to collect network indicators in real time to build a supply and demand model, allocating bandwidth differently according to bandwidth status, delineating dedicated channels for edge nodes, supporting temporary expansion, establishing a bandwidth preemption and release mechanism for high-priority tasks to low-priority tasks, and synchronizing the allocation results to the platform.

[0010] Furthermore, the division of basic and special roles specifically involves using the RBAC model to create a two-tiered system of basic and special roles. Permissions are bound to data hierarchy and departmental responsibilities. Basic roles include read-only roles, modify roles, and download roles; special roles include census administrators and security auditors. Roles are matched with departments, and permissions are managed through PostgreSQL and synchronized to the shared platform.

[0011] Furthermore, the data fusion and remote sensing inversion verification completed through Spark parallel computing specifically involves adopting a master node scheduling and slave node distributed deployment mode, connecting to the storage architecture to obtain preprocessed multi-source data, splitting tasks according to data type and administrative region for parallel execution, constructing an integrated dataset by associating sample plot measured data and sample plot attributes with plot identification codes, removing mismatched data after verification, correcting remote sensing inversion results using sample plot measured data, marking anomalies and triggering a review.

[0012] Furthermore, the encrypted transmission of the classified data specifically involves the following: the classified data is core information related to national ecological security, and a dynamically updated catalog is established by the census administrator and security auditor according to the plan; the data is statically encrypted during preprocessing, dynamically encrypted using the AES-256 algorithm before cross-departmental transmission, and marked with a "classified identifier" during transmission; the key is generated and distributed hierarchically and updated regularly.

[0013] This application also discloses a forestry information sharing system for multi-source data fusion, comprising: a data classification and processing module, used to classify multi-source data according to data type and importance, compress raster data using a high-fidelity, high-compression-ratio image compression format, structure map patch data by administrative region, and associate map patch identification codes with sample plot data and perform redundancy removal; a hybrid storage management module, used to build a hybrid architecture of distributed storage and structured data association, implement three-copy backup for the core data master node, store general data in fragments by type on the slave nodes, and establish a multi-dimensional data index; a data transmission optimization module, used to perform data transmission using a multiplexed network transmission protocol, update data based on a differential transmission mechanism, deploy edge nodes to cache high-frequency data, and realize dynamic allocation of transmission bandwidth; a permission security control module, used to classify basic roles and special roles, assign corresponding operation permissions by department, set an authorized IP whitelist, and implement encrypted transmission protection for confidential data; and a parallel computing publishing module, used to complete data fusion and remote sensing inversion verification through Spark parallel computing, and publish a RESTful API that can be shared by multiple departments.

[0014] This application also discloses an electronic device, including: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the forestry information sharing method based on multi-source data fusion of the present invention.

[0015] Beneficial Effects: This application significantly reduces data storage footprint and optimizes computational resource allocation through hierarchical data processing and efficient compression, combined with a hybrid storage architecture and multi-dimensional indexing. Data read / write and retrieval efficiency are improved, and conventional hardware and software can support rapid processing of multi-source data. Cross-regional data transmission efficiency is significantly improved. Advanced transmission protocols and differential update technology reduce redundant transmission, edge node caching shortens access distances, dynamic bandwidth allocation ensures priority for core data transmission, and transmission latency is significantly reduced, effectively supporting multi-departmental collaborative census. A balance is achieved between data security and sharing needs. The RBAC role system and IP whitelist precisely control access permissions, encrypted transmission of classified data strengthens security, and full-process audit logs ensure traceability, preventing data risks while meeting cross-departmental data access requirements. Data fusion accuracy and sharing convenience are simultaneously improved, providing efficient and secure technical support for integrated census.

[0016] In practice, raster data is compressed using H.265 / AV1 format, a storage architecture combining HDFS and PostgreSQL is built, and HTTP / 3 and QUIC protocols are used for transmission. Through Spark parallel computing, there is a close synergy between them, forming a closed-loop support for the entire process of "data preprocessing-storage-transmission-computation-sharing". After efficient compression, raster data significantly reduces storage footprint and transmission load, alleviating pressure on the hybrid storage architecture combining HDFS and structured databases. It also adapts to the low-latency transmission requirements of HTTP / 3 and QUIC protocols, enabling rapid transfer of multi-source data to parallel computing nodes. The hybrid storage architecture, through distributed file storage and structured data association management, provides precise and efficient data retrieval support for Spark parallel computing, ensuring rapid coupling of multiple data types during computation. The highly reliable transmission characteristics of HTTP / 3 and QUIC protocols guarantee real-time synchronization of multi-source data required for parallel computing, avoiding computational interruptions or delays. After Spark parallel computing completes data coupling and verification, it generates a high-quality fused dataset, which is made available to multiple departments through a standardized RESTful API. The high-frequency data demands generated by API calls can, in turn, leverage the sharding caching of the storage architecture and the dynamic bandwidth allocation of the transmission protocol to optimize response efficiency, achieving efficient end-to-end collaboration in data processing, storage, transmission, computation, and sharing. Attached Figure Description

[0017] Figure 1 This is a flowchart of a forestry information sharing method based on multi-source data fusion according to the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides a forestry information sharing method based on multi-source data fusion, such as... Figure 1 As shown, the steps include: Multi-source data are classified into levels according to data type and importance; raster data is compressed; patch data is structured into plots according to administrative regions; and sample plot data is associated with patch identification codes and redundancy is removed. A hybrid architecture is built that combines distributed storage with structured data. The core data master node is backed up with three replicas, while general data is stored in shards by type on the slave nodes, and multi-dimensional indexes are established. Data transmission is performed using a multiplexed network transmission protocol, data is updated based on differential transmission, high-frequency data is cached at edge nodes, and transmission bandwidth is dynamically allocated. Define basic and special roles, assign corresponding permissions to departments, set up an authorized IP whitelist, and encrypt the transmission of classified data. Data fusion and remote sensing inversion verification were completed using Spark parallel computing, and a RESTful API was released for sharing by multiple departments.

[0020] The proposed classification of multi-source data according to data type and importance is specifically implemented as follows: Data is clearly categorized into three main types based on data type: raster data, patch data, and sample plot data. Simultaneously, it is classified into three levels based on importance: Level 1 consists of core survey data, including key data directly affecting the accuracy of the survey results, such as measured volume, biomass, and carbon storage data from sample plots; core fields for land type, ownership, and vegetation cover type of patches; and high-resolution remote sensing orthophotos used for base map production. Level 2 consists of important auxiliary data, including supplementary soil texture and slope aspect attributes for patches; intermediate remote sensing inversion data for grassland yield and vegetation cover; and monitoring data for wetland dissolved oxygen and soil moisture content. Level 3 consists of general shared data, including historical survey comparison data, auxiliary explanatory documents required for inter-departmental collaboration, and redundant supplementary data for non-core attributes.

[0021] The raster data is compressed, specifically using the H.265 / AV1 format. The raster data includes high-resolution aerospace remote sensing imagery, UAV aerial survey imagery, and core raster data from digital orthophoto (DOM) surveys. High-resolution remote sensing imagery must meet requirements of spatial resolution better than 1m and at least six multispectral bands, including near-infrared and short-wave infrared bands. UAV aerial survey imagery must meet positioning accuracy requirements of spatial resolution better than 0.05m. The compression process strictly adheres to a compression ratio range of 10:1-20:1. For core survey data (such as DOM for base map production and images required for forest stock volume remote sensing inversion), a conservative compression ratio of 10:1-15:1 is used to prioritize the integrity of land cover identification, patch boundary extraction, and core pixel information on vegetation cover features. For general shared data (such as historical comparative remote sensing imagery and auxiliary explanatory raster data), a high-efficiency compression ratio of 15:1-20:1 is used to control data volume while ensuring that cross-departmental collaborative viewing needs are not affected.

[0022] The map patch data is structured by administrative division. Specifically, it is divided into three levels: province, county, and township. The province serves as the first-level unit, the county as the second-level unit, and the township as the third-level unit. Map patch division strictly adheres to the principle of not crossing administrative boundaries, ensuring that all map patches within a single patch belong to the same administrative jurisdiction. During the structuring process, the map patch data is first standardized into a vector format, uniformly adopting the CGCS2000 National Geodetic Coordinate System and the 1985 National Height Datum, and processed according to the 3° zone Gauss-Kruger projection to ensure spatial coordinate consistency. Then, the attribute fields are standardized. Required fields include the map patch identifier code, land type code and name, property rights nature, vegetation cover type, vegetation structure, area, and the core information of the administrative region code. Optional fields include soil type, slope aspect, slope position, and supplementary information related to desertification, sandification, and rocky desertification. For small, fragmented patches with an area of ​​less than 400 square meters, they are merged according to the principle of proximity. Isolated forest and grassland wetland patches with an area of ​​less than 400 square meters in the annual land change survey are retained, and forest belts are represented by area patches. At the same time, the boundaries of forestry and grassland management units (state-owned forest farms, pastures, and nature reserves) are integrated. When the deviation between the boundary of the management unit and the boundary of the patch does not exceed 5 meters, the boundary of the patch in the forest and grassland resource map is used.

[0023] The process of associating sample plot data with map patch identifiers and removing redundancy involves the following steps: Sample plot data includes measured data from surveys of forests, grasslands, wetlands, desertification, and rocky desertification. This measured data includes the sample plot's location coordinates, site conditions, vegetation growth status, species composition, and core information such as measured volume, biomass, and cover. Data sources must be linked to corresponding field survey records and instrument testing reports. When associating map patch identifiers, the coordinates of the sample plot's center point are used as a reference. Spatial overlay analysis is used to match the nearest neighboring map patch. If a sample plot spans multiple map patches, the identifier of the patch with the largest area percentage is used as the primary identifier, along with secondary map patch identifiers and spatial overlap information. After association, a correspondence table is established: "Sample Plot Identifier - Primary Map Patch Identifier - Secondary Map Patch Identifier - Association Confidence," ensuring the spatial attribution of sample plots and map patches is traceable. During the redundancy removal process, firstly, attribute fields that are duplicated with the map patch data are removed, such as the administrative region code, land category name, and ownership information, retaining only the measured data fields specific to the sample plot; secondly, duplicate records with identical content are merged, and data with different field names due to different survey batches (such as "vegetation cover" and "vegetation density") are standardized and unified; finally, invalid fields and redundant fields with more than 80% null values ​​are deleted to form a simplified sample plot dataset.

[0024] The proposed hybrid architecture for distributed storage and structured data association is implemented by building a storage architecture that combines HDFS and PostgreSQL, constructing a dual-core architecture of "file storage + database". The functional division of labor between the Hadoop Distributed File System (HDFS) and the PostgreSQL spatial database is clearly defined. HDFS mainly carries high-resolution remote sensing imagery, large-capacity raster data of UAV aerial survey data, and structured patch and sample plot data files, while the PostgreSQL spatial database focuses on storing structured data with spatial attributes, such as patch identification codes, sample plot association information, and administrative division codes. The two achieve metadata synchronization and association calls through a data interface.

[0025] The architecture is deployed in a hierarchical manner using a "master-slave node cluster + regional nodes" model. The master node cluster ensures the read and write speed of core data, while the slave node cluster is deployed in a distributed manner according to the "province-county" administrative division to meet the storage needs of general shared data. An internal data synchronization channel is established, where raster data metadata (such as file path, compression format, and resolution) stored in HDFS is synchronized to PostgreSQL in real time. When the attribute data of map features and sample plots in PostgreSQL changes, the association identifier of the corresponding file in HDFS is updated synchronously through a trigger mechanism, ensuring data consistency between the two storage media. Simultaneously, the architecture is compatible with various data formats after preprocessing. HDFS supports direct writing of raster data in H.265 / AV1 compressed format, and PostgreSQL, through the PostGIS spatial extension module, supports vector data storage and spatial query operations in the CGCS2000 coordinate system, allowing for seamless integration with subsequent fusion calculation processes without additional format conversion.

[0026] For example, the block size of the HDFS in the master node cluster is set to 128MB to adapt to the characteristics of large file storage. The PostgreSQL spatial database adopts partitioning table technology, partitioning the map patch data by "year-provincial administrative region", and the data volume of a single partition is controlled within 5 million records. After the core census data is backed up to three copies in HDFS, its storage usage is about 1.2 times that of the original data. The storage usage of a single copy of general shared data is basically the same as the data volume after preprocessing. The architecture supports more than 1,000 concurrent query requests per second. The response time for map patch location retrieval in the PostgreSQL spatial database is no more than 500ms, and the response time for reading raster data fragments in HDFS is no more than 1 second, meeting the data access needs of multi-department collaborative census.

[0027] The core data master node employs a three-replica backup system. Specifically, the backup target of the core data master node is clearly defined as primary core survey data, including key data on measured volume, biomass, and carbon storage from sample plots; core fields such as land type, ownership, and vegetation cover type for map features; and high-resolution remote sensing orthophotos used for base map production. The backup process is synchronized in real-time with the master node data write operation to ensure complete consistency between the replica data and the source data. The three replicas are stored on independent servers in different physical locations within the master node cluster. Each server is configured with 256GB of memory, 20TB or more of SSD storage, and a 10GbE high-speed network card. Furthermore, the replica storage nodes are avoided to be located in the same rack or power supply unit to reduce the risk of multiple replicas failing due to a single point of failure.

[0028] When the backup operation is started, the integrity of the core data is first checked. After the check passes, incremental backup is used to synchronize to the replica nodes, transmitting only the changed data to improve backup efficiency. At the same time, a replica health check mechanism is established to check the data consistency of each replica every hour. The integrity of the replica is confirmed by comparing hash values. If a replica is found to be lost or damaged, an automatic repair process is immediately triggered to reconstruct the failed replica by synchronizing data from the intact replica.

[0029] The general data is stored by type from the nodes. In practice, the general data storage objects of the nodes are three levels of general shared data, including historical census comparison data, auxiliary explanatory documents required for departmental collaboration, and redundant supplementary data of non-core attributes. When storing, the provincial-level nodes are first designated as the first-level storage unit and the county-level nodes as the second-level storage unit according to the administrative division of "province-county". Then, within each storage unit, the data is further subdivided into raster data fragments, vector data fragments, attribute data fragments, and document data fragments according to data type.

[0030] Raster data shards primarily store historical remote sensing imagery and low-precision auxiliary aerial survey data, archived uniformly in TIFF or JPEG format, distinguishing them from the H.265 / AV1 format of the master node's core raster data. Vector data shards carry non-core patch vector data and historical boundary data, using the CGCS2000 coordinate system and Gauss-Kruger projection, named according to "county-level administrative code - vector type". Attribute data shards store non-core supplementary attribute tables and historical survey statistics, stored in structured data table format. Document data shards collect various auxiliary documentation and operation manuals, supporting common formats such as PDF and DOCX. Each shard is associated with a unique storage identifier, including "administrative region code - data type - storage time" information, facilitating data location and retrieval between nodes. A shard data association mapping table is also established and synchronized to the master node's PostgreSQL database, enabling traceability and association with the core data.

[0031] The establishment of a multi-dimensional index, in practice, involves constructing a multi-dimensional index system—combining spatial characteristics, temporal attributes, and business retrieval needs—of multi-source forestry data. This system comprises a spatial index, a time index, a keyword index, and an administrative affiliation index, achieving deep integration between the index and the HDFS distributed file system and the PostgreSQL spatial database. The spatial index is built for high-resolution remote sensing images and vector data with spatial coordinates. In the PostgreSQL database, an R-tree spatial index is established using PostGIS extensions, linking the coordinates of the center point of the map patch, the latitude and longitude range of its boundaries, and the location coordinates of the sample plot. This supports spatial retrieval operations such as querying by spatial range and neighborhood analysis. In HDFS, raster data is indexed based on spatial grid division, with images divided into 1km × 1km grid blocks. The index records the spatial range and storage path of each image block. The time index focuses on the data generation and update time, including key time nodes such as the remote sensing image capture time, sample plot survey time, and map patch attribute update time. It adopts a B+ tree index structure and supports data filtering by time granularity of "year-quarter-month." The time index accuracy for core census data is down to the hour level, while the accuracy for general shared data is down to the day level.

[0032] The keyword index focuses on frequently searched fields, including map patch identifiers, sample plot numbers, land use codes, ownership unit names, and data types. It employs an inverted index mechanism to enable multi-keyword combination queries and fuzzy matching; for example, "map patch identifier + land use name" can be used to quickly locate target data. The administrative affiliation index corresponds logically to the "province-county-township" segmentation of map patches, establishing an association index between administrative region codes and data segments. The index hierarchy is consistent with the administrative division hierarchy, supporting rapid aggregation and retrieval of all data within a corresponding region by administrative unit. Simultaneously, an index collaboration mechanism is established, triggering automatic index updates when data is added, modified, or deleted, ensuring consistency between the index and the data. An index optimization task is set up to clean up index fragments daily at midnight and dynamically adjust the index structure monthly based on high-frequency query scenarios analyzed from retrieval logs, improving retrieval efficiency.

[0033] The data transmission using a multiplexed network transmission protocol is specifically implemented using HTTP / 3 and QUIC protocols. HTTP / 3 serves as the standard protocol for cross-departmental and cross-regional data transmission, while its integrated QUIC protocol achieves 0-RTT (Round-Trip Time) fast handshakes through a UDP-based connection establishment mechanism, reducing reconnection delays caused by network switching or connection interruptions. Simultaneously, multiplexing technology avoids the head-of-line blocking problem of traditional TCP protocols, improving the efficiency of concurrent data transmission. The transmitted data covers the entire pre-processed multi-source forestry data, including H.265 / AV1 compressed raster data, fragmented structured patch data, and lightweight sample plot data. Core survey data transmission uses protocol priority settings to ensure priority access to transmission channels when network resources are scarce.

[0034] The differential transmission of updated data, in practice, specifically involves the differential transmission covering all pre-processed multi-source forestry data, including H.265 / AV1 format compressed raster data, "province-county-township" segmented patch data, and sample plot data with associated patch identification codes. The transmission logic is deeply integrated with the HTTP / 3+QUIC protocol transmission system, forming an efficient transmission mode of "full initialization + incremental update".

[0035] Data update identification employs a "type-differentiated comparison" mechanism: Raster data (such as remote sensing imagery and UAV aerial survey data) is divided into 1km×1km spatial grid blocks. Through dual hash comparison of pixel grayscale values ​​and spectral features, the updated grid blocks are located, with comparison accuracy down to a single pixel. Patch data is identified by verifying the vector boundary coordinates, land use codes, and core fields of ownership information one by one, and new, deleted, and attribute-changed patch records are identified, supporting field-level difference extraction. Sample plot data uses the sample plot identification code as an index to compare the changes in measured data (such as volume and biomass) and related patch information, ensuring that no differential data is missed.

[0036] After the updated data is extracted, it is automatically packaged into an incremental data packet. The packet includes the update type (add / modify / delete), data association identifier (such as plot identifier code, sample plot number), update timestamp and difference data body, and also attaches the association mapping information with the full data to facilitate rapid integration at the receiving end.

[0037] During transmission, incremental data packets are bound to the core data transmission channel, enjoy bandwidth allocation priority, and are linked to the edge node caching mechanism—after receiving incremental data, the edge node synchronously updates the corresponding dataset in its local cache, avoiding repeated cross-regional transmissions. The data receiving end verifies the integrity of incremental data packets using SHA-256 hash values. If the verification passes, the packets are automatically merged into the local full data based on the associated mapping information, without manual intervention. If the verification fails, a breakpoint resumption mechanism is triggered, retransmitting only the missing or erroneous difference segments.

[0038] The deployment of edge nodes to cache high-frequency data is implemented by combining the regional distribution characteristics of the national forestry census data and the cross-departmental access needs. Edge node caching servers are deployed according to the geographical division principle of "seven major regions + three special regions." Each of the seven major regions (Northeast, North China, East China, Central China, South China, Southwest, and Northwest) has one core edge node. Due to their vast areas and large data demands, Xinjiang, Tibet, and Inner Mongolia each have an additional node. Node locations are prioritized for proximity to the data centers of the provincial forestry and grassland authorities within the region to ensure network link stability. Each edge node is configured with differentiated hardware resources and supports direct caching of multiple data formats, including H.265 / AV1 and TIFF. The selection of cached data follows the dual criteria of "high-frequency access + core needs". By analyzing the access logs of national census data, remote sensing image base maps, provincial administrative region standard map patch datasets, and key area sample plot core data that have been accessed more than 500 times in the past 3 months are listed as high-frequency core cache objects. Historical census comparison data and departmental collaborative auxiliary documents that have been accessed 200-500 times are listed as high-frequency general cache objects. Among them, the core cache data accounts for no less than 60% of the node cache capacity.

[0039] A dual mechanism of "popularity update + trigger update" is established. Every day at midnight, cached data is sorted based on the access popularity of the previous 24 hours, and inefficient data with less than 50 accesses is removed. When the core census data of the main node (such as the update of remote sensing images of key areas and the supplement of sample plot measured data) changes, the cache update of the corresponding edge node is triggered in real time through the data synchronization interface to ensure the consistency between cached data and source data.

[0040] Each edge node establishes a collaborative caching network with the main node and surrounding nodes. Core edge nodes can share cache resources. For example, forestry data of the Yangtze River Delta region cached by the East China node can be synchronized to the Central China node for backup. When cached data is missing at a certain node, it is automatically retrieved from the nearest collaborative node to avoid cross-regional back-to-source main node.

[0041] The dynamic allocation of transmission bandwidth, in practice, aims to ensure the efficiency of core data transmission for the integrated census. It combines the HTTP / 3 and QUIC protocol transmission systems to construct a dynamic bandwidth allocation mechanism based on data priority and real-time network status. First, the data transmission priority hierarchy is clearly defined, corresponding to the multi-source data grading system: Level 1 core census data (key data measured in sample plots, core patch fields, high-resolution base map images) has the highest priority; Level 2 important auxiliary data (intermediate data from remote sensing inversion, wetland monitoring data) has medium priority; and Level 3 general shared data (historical comparison data, auxiliary documents) has low priority. A bandwidth scheduling engine is built, using the SNMP protocol to collect bandwidth utilization, network latency, and packet loss rate indicators of cross-regional transmission links in real time, with a collection frequency set to once every 10 seconds, establishing a dynamic bandwidth supply and demand model.

[0042] A differentiated allocation strategy is implemented based on the model output: When network bandwidth is sufficient (utilization rate below 60%), bandwidth is allocated according to the ratio of "high priority 30%, medium priority 40%, low priority 30%" to ensure synchronous transmission of data at all levels; when network bandwidth is strained during peak survey periods (such as quarterly data aggregation, cross-departmental joint verification) (utilization rate above 80%), priority scheduling is automatically triggered, increasing the bandwidth share of primary core data to 50%-70%, compressing the share of secondary data to 20%-30%, and controlling the share of tertiary data to within 10%, and suspending low-priority data transmission if necessary. For transmission needs within the coverage area of ​​edge nodes, dedicated bandwidth channels are separately allocated between nodes, with channel bandwidth accounting for no less than 20% of the total regional bandwidth, ensuring rapid synchronization and retrieval of cached data at edge nodes. When the proportion of cached data transmission tasks at edge nodes exceeds 50% of the regional bandwidth, temporary bandwidth expansion is automatically enabled for them.

[0043] Simultaneously, a bandwidth preemption and release mechanism is established. In the event of a sudden core data transmission demand (such as emergency sample plot verification data reporting), high-priority tasks can preempt bandwidth from low-priority tasks. The preemption ratio will not exceed 50% of the current bandwidth of the low-priority task. Once the high-priority task has completed its transmission, the preempted bandwidth will be automatically returned to the low-priority task. The bandwidth allocation results are synchronized to the census management platform in real time, and the bandwidth usage of each priority data is displayed through a visual interface. Administrators can manually intervene to adjust the allocation ratio in special scenarios.

[0044] The current method of allocating bandwidth using a fixed threshold ratio fails to consider the differences in the urgency of data at the same level. For example, while both burst sample verification data and regular summary data belong to the first-level core data category, their urgency levels are drastically different. Furthermore, network status assessment relies solely on bandwidth utilization, which cannot accurately reflect complex characteristics such as network fluctuations and latency jitter, easily leading to a disconnect between the allocation strategy and actual transmission needs. Therefore, this application further proposes intelligent bandwidth allocation based on service urgency and network entropy. This involves constructing a dynamic model by integrating data service value (urgency coefficient) and network operating status (network entropy value) to achieve bidirectional adaptation between the demand and supply sides. The specific implementation steps are as follows: The input data includes basic data information, business urgency indicators, and network status parameters. The basic data information includes multi-source data classification results (i), data size, transmission time limit requirement (T), and the business scenario (e.g., sudden verification / routine summary / historical query). The business urgency indicators include sudden task identifier (I) and data impact range (S). The network status parameters include transmission link bandwidth utilization (B), network latency (D), latency jitter (J), and packet loss rate (L). The data is collected jointly via NetFlow and SNMP protocols, with a collection frequency of 5 seconds per collection.

[0045] Define the hierarchical results of multi-source data as follows: i=1 is the first-level core, i=2 is the second-level auxiliary, and i=3 is the third-level shared; define the sudden task identifier I, i=1 is sudden and i=0 is normal; define the data impact range S, which involves the provincial level as level 5, the city level as level 4, the county level as level 3, the township level as level 2, and the village level as level 1.

[0046] First, calculate the business urgency coefficient. Use the analytic hierarchy process (AHP) to determine the weight of each indicator. Normalize the scope of influence S first, then quantify the business urgency and output the urgency coefficient E, where E∈[0,1]. The larger the value of E, the higher the urgency.

[0047] The formula for normalizing the scope of influence is: ; The formula for combining the urgency coefficient is: ; In the formula, w1 is the suddenness indicator weight, w2 is the remaining time limit weight, and w3 is the impact range weight; I i This is used to identify sudden tasks; 1 indicates a sudden task, and 0 indicates a regular task, which directly participates in the calculation and does not require normalization; T max,i The maximum allowed time for data transmission at level i (in minutes) is preset in the business rules: Level i core T max,1 =30, Level 2 Auxiliary T max,2 =60, Level 3 Shared T max,3 =120; T rem,iThe remaining time limit for data transmission at level i (in minutes) is T. max,i Matching usage; S norm,i This is the normalized value of the influence range of the i-th level data, which is derived from S (levels 1-5).

[0048] Then, the network entropy value H is calculated. The information entropy theory is introduced. First, the four network parameters are normalized to [0.01, 0.99] respectively (to avoid ln(0) being meaningless), and then converted into a probability distribution. Finally, the network entropy H is calculated (the larger the entropy value, the more unstable the network state).

[0049] For each network parameter, upper and lower limits are determined based on the service threshold. The normalization formula for a single parameter is: ; Converting normalized values ​​into probabilities that satisfy Σp=1, the probability distribution transformation formula is: ; The formula for calculating network entropy is: ; In the formula, k=1,2,3,4: correspond to network parameters (1=bandwidth utilization B, 2=network latency D, 3=latency jitter J, 4=packet loss rate L), and each parameter needs to be normalized separately; x k : The original value of the k-th network parameter (in units of %, ms, ms, %); x min,k / x max,k x is the business threshold for the k-th parameter. min,k / x max,k The determination is based on the forestry transport scenario, and is referenced in Table 1 below; Let be the normalized probability of the k-th network parameter, satisfying ; The normalization factor is used because k = 4 parameters. The entropy value is constrained to the interval [0,1] (H=0 when the four parameters are completely stable, and H=1 when they are completely random).

[0050] ; Table 1x min,k / x max,k Value Reference Table Then construct an intelligent bandwidth allocation model. , ( and ), using the urgency coefficient of the i-th level data and network entropy As the core variable, combined with the basic proportion of data classification. Output the real-time bandwidth allocation ratio of data at each level. (i=1,2,3 correspond to level one, two, and three data points, respectively). The model introduces a network entropy adjustment term to avoid drastic changes in bandwidth allocation during network fluctuations. Calculation formula: ; In the formula, The basic bandwidth ratio of the i-th level data ( ), default value: (Level 1 Core) (Level 2 Auxiliary) (Three-level sharing) reflects the basic priority of data classification; The urgency adjustment factor is 1.2-1.5. When the network is stable (H≤0.3), it is set to 1.5 (to amplify the urgent data demand), and when the network fluctuates (H>0.3), it is set to 1.2 (to avoid excessive allocation bias). The network entropy adjustment factor is set to 1.0 when the network entropy is high (H > 0.8) (to suppress bandwidth allocation fluctuations) and 0.8 when the network entropy is low (H ≤ 0.8) (to enhance the urgency response sensitivity). Denominator: Normalization term, ensuring R1+R2+R3=1, bandwidth allocation logic closed loop; Key constraints: >0 (This is ensured by the range of values ​​for β and γ, avoiding a negative numerator; the actual E) i When >0, the molecule is always positive.

[0051] Then, dynamic adjustments and feedback are made, and quantitative standards for triggering conditions are supplemented. Specifically, the transmission progress of data at each level is monitored in real time (C). i The change in R (number of bytes transmitted / total number of bytes) and network entropy H is recalculated every 5 seconds. i The adjustment logic is as follows: When C i When ≥90%, the remaining bandwidth ΔR of this level will be... i =R i ×(1-C i Proportionally allocate incomplete high-urgency data (only allocated to E) j >0.7 (j level). When H > 0.95 (network is extremely unstable), a conservative allocation strategy is triggered: the primary core data R1 is fixed at 50%, and the remaining bandwidth is allocated to the secondary and tertiary data in the ratio α2:α3. When a certain level E i When the urgency level is >0.9 (extremely high urgency), it is allowed to preempt low urgency data (E j For bandwidth of <0.3), the preemption rate is ≤30%.

[0052] In a clearly defined application scenario, the final output includes: the real-time bandwidth allocation ratio of each level of data (R1, R2, R3), which is directly used by the bandwidth scheduling engine; Edge node dedicated channel bandwidth threshold B edge B edge Based on dynamic setting of network entropy: when H < 0.3, B edge =25% (total regional bandwidth), B when 0.3≤H<0.8 edge =27%, when H>0.8 B edge =30%, ensuring cache synchronization stability; Bandwidth adjustment trigger condition (quantitative indicator): E i >0.9 triggers high-priority preemption, H>0.95 triggers conservative allocation, C i ≥90% triggers bandwidth reclamation, all of which are clearly executable thresholds.

[0053] For example, during the data compilation period of a certain quarterly census, the cross-departmental transmission link in East China simultaneously carried three types of data transmission needs: First, primary core data—sample plot data from a sudden supplementary survey in a rocky desertification area in Southwest China (12GB, i=1, sudden task I=1, provincial impact range S=5, transmission time limit T). max =30min, remaining time T rem =8min); Second, secondary auxiliary data—intermediate data for remote sensing inversion of grassland yield in six provinces of East China (20GB, i=2, routine task I=0, city-level influence range S=4, T max =60min, T rem =25min); Third, three-level shared data—comparative documents of regional forestry historical surveys from 2018 to 2022 (15GB, i=3, routine task I=0, county-level impact range S=3 levels, T max =120min, T rem =60min). At this time, the link monitoring data is: bandwidth utilization B=85%, network latency D=450ms, latency jitter J=45ms, packet loss rate L=4.5%, and the bandwidth scheduling engine immediately starts the intelligent allocation process.

[0054] The first step is to calculate the business urgency coefficient E: first normalize the impact range S (S norm = (S-1) / (5-1)), then calculate the weights w1=0.4, w2=0.35, and w3=0.25 using the analytic hierarchy process. First-level data S norm = (5-1) / 4 = 1, E i =0.4×1+0.35×(30-8) / 30+0.25×1=0.4+0.257+0.25=0.907 (Extremely high urgency); Level 2 data S norm= (4-1) / 4 = 0.75, E i =0.4×0+0.35×(60-25) / 60+0.25×0.75=0+0.204+0.188=0.392 (Medium urgency); Level 3 data S norm = (3-1) / 4 = 0.5, E i =0.4×0+0.35×(120-60) / 120+0.25×0.5=0+0.175+0.125=0.3 (low urgency).

[0055] The second step is to calculate the network entropy value H: normalize the network parameters according to the service threshold (B: 10%-90%, D: 10-500ms, etc.) to obtain x. norm(B) =0.882, x norm(D) =0.892, x norm(J) =0.901, x norm(L) =0.898; after the conversion probabilities, p1=0.252, p2=0.255, p3=0.258, p4=0.235; substituting into the entropy formula, we get H=-(0.252×ln0.252+0.255×ln0.255+0.258×ln0.258+0.235×ln0.235) / ln4≈0.95 (the network is extremely unstable), so we set β=1.2 (network fluctuation) and γ=1.0 (high entropy).

[0056] The third step is to calculate the bandwidth allocation ratio Ri: the basic ratios are α1=0.4, α2=0.35, and α3=0.25.

[0057] The numerator of the first-level data is 0.4 × (1 + 1.2 × 0.907 - 1.0 × 0.95) = 0.4 × 1.138 = 0.455; Secondary molecule = 0.35 × (1 + 1.2 × 0.392 - 1.0 × 0.95) = 0.35 × 0.420 = 0.147; Tertiary molecule = 0.25 × (1 + 1.2 × 0.3 - 1.0 × 0.95) = 0.25 × 0.31 = 0.078; Denominator = 0.455 + 0.147 + 0.078 = 0.68; The final values ​​are R1 = 0.455 / 0.68 ≈ 66.9%, R2 = 0.147 / 0.68 ≈ 21.6%, and R3 = 0.078 / 0.68 ≈ 11.5%. Simultaneously, a dedicated channel bandwidth B for edge nodes is allocated. edge =30% (H>0.8).

[0058] In actual transmission, the 12GB primary core data, relying on a high bandwidth allocation of 66.9%, reduced transmission time from the traditionally fixed 35 minutes to 9 minutes; the secondary data transmission rate remained stable at 25Mbps, and although its proportion decreased, it did not affect the aggregation progress; the tertiary data was temporarily compressed to a low proportion, pending the completion of the core data transmission (C... i After reaching ≥90%, the system automatically allocates the remaining bandwidth of Level 1 data proportionally to Level 2 and Level 3 data, keeping the total transmission time of Level 3 data within 40 minutes. When the Southwest emergency data transmission reached 92%, a county-level forestry and grassland department urgently reported 1.2GB of sample plot verification data (E=0.91). The system immediately preemptively occupied 30% of the bandwidth for Level 3 data (approximately 3% of the total bandwidth) according to the rules, enabling the emergency data to be transmitted within 2 minutes. The bandwidth was automatically returned after 30 seconds, without causing significant delays to the subsequent transmission of Level 3 data. When edge nodes synchronize provincial high-frequency remote sensing base maps, the 30% dedicated channel bandwidth ensures that its transmission rate remains stable at over 110Mbps, which is 3.2 times more efficient than shared links, fully demonstrating the advantages of the intelligent allocation model in adapting to demand and network conditions.

[0059] By capturing data service attributes and network operating status in real time, a dynamically adapted bandwidth allocation scheme is output, ensuring both priority transmission of core and urgent data and efficient utilization of network resources.

[0060] The division of basic and special roles, in practice, is to support the collaborative needs of multi-department integrated census and balance data sharing and security management. Based on the RBAC (role-based access control) model, a two-layer role system of "basic roles + special roles" is constructed. The role permission division is deeply bound to the hierarchical classification of multi-source forestry data (core census data, important auxiliary data, and general shared data) and departmental business responsibilities.

[0061] The basic roles focus on routine data access needs and are clearly divided into three categories: read-only roles, modify roles, and download roles. Read-only roles have access to all shared data, allowing viewing of core information from high-resolution remote sensing images, map features, and sample plot data. However, they lack data modification, download, and attribute editing permissions, only supporting online preview and viewing of statistical results. Modify roles are limited to specific data within their department's responsibilities. They can only update map feature attributes within their administrative region or business area (e.g., forestry and grassland departments can modify the vegetation cover type field of forest map features within their jurisdiction) and supplementary sample plot survey information. They cannot access data from other departments or core census data (e.g., key fields such as measured volume and carbon storage in sample plots), and modification operations must leave a trace. Download roles, in addition to read-only permissions, also grant download permissions for non-confidential data. They can access general shared data and important auxiliary data related to their department. Core census data and confidential data cannot be downloaded, and downloaded files will automatically have a watermark (including the downloader and download time).

[0062] Special roles are set up to meet management and security auditing needs, including two categories: Census Administrator and Security Auditor. Census Administrators have core permissions for role allocation, permission adjustment, and data authorization. They can create temporary roles based on new business needs of departments (such as temporary roles for cross-departmental joint verification), set the validity period of roles (not exceeding 30 days), and are responsible for handling permission application and change processes. When departmental responsibilities are adjusted, the corresponding role permissions must be updated within 24 hours. Security Auditors do not have data access permissions, but can only view the full data operation audit logs. They have log retrieval, abnormal behavior marking, and audit report generation functions. They can track and trace the source of high-frequency access and abnormal time period operations. When violations are found, they have the right to trigger permission freeze and simultaneously notify the Census Administrator.

[0063] The matching of roles and departments follows the principle of "minimum necessity". The land and resources department is mainly configured with read-only roles, with some business positions configured with download roles. The forestry and grassland department is configured with modify roles for core business positions and download roles for management positions. The permission configuration of the wetland department follows the same principle as the forestry and grassland department. Cross-departmental collaborative groups can be configured with composite roles that include permissions from multiple departments by the census administrator.

[0064] The role-based permission association mechanism is implemented through the permission management module of the PostgreSQL database. The correspondence between roles and data levels is synchronized to the shared platform in real time, and the access permissions of the corresponding roles are automatically adjusted when the data level is adjusted.

[0065] For example, in a provincial forestry information sharing system, basic role users account for 92%, including 65% read-only roles, 18% modify roles, and 9% download roles. Among special roles, there is one census administrator at the provincial level and one at the county level, and two safety auditors at the provincial level to achieve dual-person review. When a user with a modify role in the forestry and grassland department attempts to modify map data in other counties, the system immediately pops up an insufficient permission prompt and records the operation. The safety auditor discovers the abnormal behavior and marks it within one hour through log retrieval. When a cross-departmental joint verification task is launched, the census administrator creates a temporary role for the joint team, granting read-only and partial field modification permissions to map data in a specific area. After the task is completed, the system automatically revokes the temporary role permissions within 4 hours to ensure the accuracy of permission control.

[0066] The allocation of permissions by department and the establishment of an authorized IP whitelist are implemented by strictly adhering to the principle of "business compatibility + minimum necessity" when allocating permissions by department. This involves precisely matching permissions with the core responsibilities of each department within the RBAC role system: For the forestry and grassland department, as the main force in the census, core business positions (such as field surveys and internal verification positions) are assigned modification roles, allowing them to update vegetation cover types, management attributes, and supplementary survey data for forestry, grassland, wetland, and desert areas within their jurisdiction. Management positions are configured with download roles, enabling them to access non-confidential regional census summary data. The land and resources department focuses on land ownership verification, primarily using read-only roles, with only some specialized positions configured with download roles to obtain data related to land parcel ownership. Specialized departments for wetland and desertification control are assigned corresponding roles based on their business scope, allowing them to operate only on subsets of data related to their specific census and excluding access to cross-domain core data. After permission allocation, the data is synchronously entered into the PostgreSQL permission management module and linked with data hierarchical permissions. Core census data is only accessible to core positions within the forestry and grassland department for modification, while generally shared data is accessible to all departments with read-only roles.

[0067] When setting up an authorized IP whitelist, a "tiered application + dynamic verification" mechanism is implemented: Each department must submit an IP whitelist application, specifying the IP type (fixed office IP / mobile work IP), usage scenario (daily office / field site access), and corresponding permission role. The application is then entered into the system after being signed by the department head and reviewed by the census administrator. The whitelist is managed in categories of "fixed IP + temporary IP." Fixed IPs are the IPs of each department's office area servers and core office terminals, and are valid indefinitely. Temporary IPs are for mobile scenarios during field surveys, with a maximum validity period of 15 days, after which they automatically expire. A dynamic management ledger for the IP whitelist is established, recording IP addresses, bound roles, authorization periods, applying departments, and reviewer information, and is synchronized to the security audit module. The data sharing platform uses a "IP + account password" dual verification mechanism. When a non-whitelisted IP initiates access, the system immediately intercepts it and generates an abnormal access log, including the accessing IP, access time, and target data identifier. If a whitelisted IP fails password verification three times consecutively, a temporary freeze is automatically triggered (unlocked after one hour), and the corresponding department contact person is notified.

[0068] For example, a provincial forestry and grassland department submitted a whitelist application for 12 fixed office IPs and 5 temporary field IPs. After review by the census administrator, the 12 fixed IPs were bound to the modification / download role, and the 5 temporary IPs were bound to the read-only role with a 10-day validity period. During the monthly IP check, it was found that a fixed IP of a county land and resources department had been deactivated. The census administrator removed the IP from the whitelist within 8 hours. When a field surveyor used a temporary IP to access sample plot data, the IP was frozen due to an incorrect password. The system immediately sent an unfreezing reminder, and the IP regained access after 1 hour. When a non-whitelisted IP attempted to access core sample plot data, the system immediately intercepted the request and generated a log. The security auditor completed the abnormal behavior verification within 2 hours and archived the log after confirming that there was no security risk.

[0069] The encrypted transmission of classified data, in practice, is specifically defined as follows: the scope of classified data is strictly defined as core information involving national ecological security, strategic resources in key forest areas, and the distribution of forests, grasslands, wetlands, and deserts in sensitive areas, including but not limited to high-precision remote sensing images of national nature reserves (spatial resolution better than 0.5m), core data on the ownership and coordinates of map patches within national defense forests and ecological red lines, and precise location and monitoring data of habitat sample plots for key endangered species. The scope is jointly defined by the survey administrator and the security auditor in accordance with the "Law of the People's Republic of China on Guarding State Secrets" and forestry data confidentiality regulations, and a dynamically updated catalog of classified data is established. Changes to the catalog must be reviewed and filed by two people.

[0070] The encrypted transmission process is deeply integrated with the multi-source data hierarchical processing and access control system. Classified data is included in the encryption control from the preprocessing stage. Static encryption is completed when the data is stored on the HDFS master node. The dynamic encryption process is triggered before cross-departmental transmission. The AES-256 algorithm is used to encrypt the data in blocks with a block length of 128 bits. The encryption mode is CBC (Cipher Block Chaining). Combined with a randomly generated initialization vector (IV), it is ensured that the same plaintext data generates different ciphertexts for each transmission. The initialization vector is transmitted synchronously with the ciphertext data but is encapsulated separately to avoid encryption security risks caused by vector leakage.

[0071] The encryption process seamlessly integrates with the HTTP / 3+QUIC protocol transmission system, completing encryption before data is encapsulated into transmission data packets. Encrypted classified data packets have a "classified identifier" field appended to their header, clearly distinguishing them from non-classified data and facilitating priority identification and control of the transmission channel. Key management employs a "hierarchical generation + dynamic distribution" mechanism. The root key is generated by the provincial forestry and grassland administration's security and confidentiality agency and stored in a dedicated hardware encryption machine (HSM). Sub-keys are derived at the "province-county-department" level, with department-level sub-keys bound to corresponding authorized roles and only allocated to core personnel with download roles and above. Sub-keys are distributed via offline key cards or encrypted emails, requiring activation and password setting upon first use. A key lifecycle management rule is established: the root key is replaced every 6 months, and department-level sub-keys are automatically updated every 3 months. When the position of the key holder changes or their permissions are revoked, the system cancels and invalidates the corresponding sub-key within 24 hours to prevent key misuse.

[0072] During the transmission of classified data, in addition to AES-256 encryption, dual protection is simultaneously implemented with transmission link encryption and data integrity verification. The transmission link uses the TLS 1.3 protocol for encapsulation. The data receiving end must simultaneously complete authentication (authorized IP + account password + key triple verification) and ciphertext integrity verification. Data integrity is confirmed through SHA-256 hash value comparison. Only after successful verification can the corresponding subkey be used for decryption. Decrypted data is stored only in an encrypted folder on the authorized terminal, and copying, screenshotting, and secondary transmission are prohibited. All encryption, transmission, and decryption operations of classified data automatically generate detailed audit logs. The log content includes classified data identification, encryption time, IP addresses of both transmitting parties, key usage records, and decryption terminal information. These logs are retained together with other data operation logs for a period extended to 5 years to meet the requirements for confidentiality audit traceability.

[0073] For example, 150GB of high-precision UAV aerial survey data of a national nature reserve, after being encrypted using AES-256, generated approximately 150.2GB of ciphertext data. The encryption process took 8 minutes, and the initialization vector was transmitted to the receiving end 3 seconds in advance as an independent data packet. Core personnel of the forestry and grassland department received the data using an authorized IP address and a bound subkey. After passing triple verification, decryption took 10 minutes. The accuracy error of the map boundary lines in the decrypted image was controlled within 2 meters, fully meeting the requirements for using classified data. When a temporary field worker's access was revoked due to a job adjustment, their subkey was cancelled within 12 hours. When the worker subsequently attempted to decrypt historical classified data using the original key, the system immediately intercepted the operation and generated an advanced anomaly log. The security auditor completed the anomaly verification and archived it within 30 minutes. In a case of classified data transmission, some data packets were lost due to network fluctuations. The receiving end discovered a hash value comparison discrepancy through integrity verification, triggering a breakpoint resume mechanism. Only the missing 3GB of ciphertext data was retransmitted, avoiding the security risks and resource waste caused by full retransmission.

[0074] All data operations generate audit logs and are retained for more than 3 years. The data operation audit logs cover all behaviors throughout the entire lifecycle of multi-source forestry data, including hierarchical classification, format compression, and redundancy removal operations in the data preprocessing stage; copy creation, fragment migration, and expansion operations in the storage stage; request initiation, data reception, and breakpoint resumption operations in the transmission stage; and viewing, modification, download, and encryption / decryption operations in the sharing and use stage, with no operations omitted. The log content adopts a standardized field format, including at least the operator's unique identifier (associated department and position information), operation timestamp (accurate to milliseconds), operation type (add / modify / delete / query), target data identifier (such as plot identification code, sample plot number, file unique hash value), data type and level, operation terminal IP address, operation result (success / failure / abnormal), and abnormal reason (such as insufficient permissions, data corruption). For classified data operation logs, additional key usage records and encryption / decryption status fields are added. The audit logs employ a dual backup mechanism of "local caching + centralized storage on the master node." Logs generated by each node are synchronized to an independent audit table in the master node's PostgreSQL database every 5 minutes. This table uses read-only access control, accessible only to security auditors and designated census administrators. Simultaneously, log files are backed up off-site via HDFS, with the backup frequency matching the master node's synchronization frequency. Log retention follows a "basic retention period + flexible extension" rule. The basic retention period for operation logs of general shared data and important auxiliary data is 3 years, while the retention period for operation logs of core census data is extended to 5 years. Operation logs of classified data are retained for no less than 10 years according to confidentiality requirements. During the retention period, retrieval is supported by multiple conditions including operator, time range, and data identifier. Log data uses blockchain storage technology, ensuring log integrity and traceability, and preventing log tampering or deletion.

[0075] Data fusion and remote sensing inversion verification were accomplished using Spark parallel computing. Specifically, the Spark parallel computing framework adopted a deployment mode of "master node scheduling + slave node distributed computing." The master node seamlessly integrated with HDFS and PostgreSQL storage architectures, acquiring pre-processed multi-source data in real time through data interfaces. This included H.265 / AV1 format raster data, vector data of map features segmented by province, county, and township, and measured sample plot data with associated map feature identifiers. Computational tasks were split according to data type and administrative region, and distributed to slave nodes for parallel execution to improve efficiency. The data fusion process used map feature identifiers as the core association key, accurately matching measured indicators (such as volume, biomass, and cover) in the sample plot data with the corresponding land type, ownership, and spatial coordinate attributes of the map features, constructing an integrated "map feature-sample plot" dataset. During the coupling process, a dual mechanism of spatial overlay analysis and attribute logic verification was used to eliminate mismatched data caused by coordinate deviations, ensuring high coupling accuracy.

[0076] Remote sensing inversion utilizes measured data from sample plots as ground truth to correct errors in vegetation indices, land cover areas, and growth status results obtained from high-resolution remote sensing image inversion. By establishing inversion model error thresholds (e.g., land cover identification error not exceeding 3%, area calculation error not exceeding 5%), inversion results exceeding these thresholds are automatically marked and trigger a manual review process. The corrected data after review is synchronously updated to the calculation result set. During computation, Spark's in-memory computing optimization mechanism is enabled, loading frequently accessed core data into a memory cache to reduce disk I / O time. Simultaneously, a computation task monitoring system is established, with the master node monitoring the computational load of each slave node in real time. When the load of a single node exceeds 70%, unfinished computation tasks are automatically distributed to idle nodes with a load below 50%, ensuring stable and efficient computation. The fused dataset after computation is synchronously stored on the HDFS master node (raster and vector files) and a PostgreSQL database (structured attribute data), generating a data fusion report that clarifies computational accuracy, error range, and data integrity indicators, providing data quality assurance for subsequent shared service releases.

[0077] A RESTful API for multiple departments is released. In practice, the design of the RESTful API strictly follows the principles of standardization and high compatibility, adapts to the business systems and data format requirements of multiple departments, and deeply binds the API functional modules with the RBAC permission system. The data scope and operation permissions of different roles through the API are consistent with the shared platform, ensuring that there are no loopholes in permission control. The APIs are divided into four modules based on business scenarios: data query, statistical analysis, data manipulation, and format conversion. The data query API supports queries based on multiple conditions, including patch identifiers, sample plot numbers, administrative regions, and data time. The returned results include the core attributes of the corresponding data, a thumbnail preview address, and storage location information. For read-only roles, only non-confidential fields are returned; modifying the role allows access to the complete attributes of the data for that department. The statistical analysis API, based on Spark fusion calculation results, provides functions for regional forest, grassland, and wetland area statistics, vegetation cover change analysis, and sample plot data aggregation. It supports outputting statistical reports at the "province-county-township" level, and the results can be directly integrated into the departmental reporting system. The data manipulation API is only accessible to the modification role and supports updating the attributes of patch and sample plot data within the department's scope of responsibility. API calls require the operator's identity token and data modification instructions, and operation records are automatically synchronized to the audit log. The format conversion API supports exporting fused data to commonly used census formats such as Shapefile, GeoTIFF, and Excel. Call permissions are bound to the download role, and exported files are automatically watermarked and format verification codes are added. The API uses HTTPS for encrypted transmission. Each call request must carry a unique identity token, which is valid for 2 hours. After expiration, a new token must be obtained through account and password verification. API call frequency is limited to a maximum of 10 requests per second per account to prevent excessive system load. A shared platform provides an API management backend that supports adding, disabling, and adjusting permissions for API interfaces. It monitors the call count, response time, and anomalies of each API in real time. Alarms are automatically triggered and pushed to technical maintenance personnel when the API response time exceeds 3 seconds or the call failure rate exceeds 5%. Detailed API development documentation is provided to facilitate integration across multiple departments, including interface parameter descriptions, return format examples, error code explanations, and call demos. A dedicated technical support line is also provided to assist departments in API integration and debugging.

[0078] This application also provides an embodiment of an electronic device. The electronic device is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: one or more processors or processing units, memory, and buses connecting different components (including memory and processing units).

[0079] A bus refers to one or more of several bus architectures, including memory buses or memory controllers, peripheral buses, graphics acceleration ports, processors, or local buses using any of the various bus architectures. Examples of these architectures include, but are not limited to, Industry Standard Architecture (ISA) buses, Micro Channel Architecture (MCA) buses, Enhanced ISA buses, Video Electronics Standards Association (VESA) local buses, and Peripheral Component Interconnect (PCI) buses.

[0080] Electronic devices typically include a variety of computer-readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0081] The memory may include computer-readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. Electronic devices may further include other removable / non-removable, volatile / non-volatile computer device storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media.

[0082] The electronic device can also communicate with one or more external devices (e.g., keyboard, pointing device, camera, etc.), may include a display, and may communicate with one or more devices that enable a user to interact with the electronic device, and / or with any device that enables the electronic device to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via an input / output (I / O) interface. Furthermore, the electronic device can communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN) and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. The processor executes various functional applications and data processing by running programs stored in memory, such as implementing the forestry information sharing method for multi-source data fusion provided in the above embodiments of the present invention.

[0083] This application also discloses a forestry information sharing system for multi-source data fusion, comprising: a data classification and processing module, used to classify multi-source data according to data type and importance, compress raster data using a high-fidelity, high-compression-ratio image compression format, structure map patch data by administrative region, and associate map patch identification codes with sample plot data and perform redundancy removal; a hybrid storage management module, used to build a hybrid architecture of distributed storage and structured data association, implement three-copy backup for the core data master node, store general data in fragments by type on the slave nodes, and establish a multi-dimensional data index; a data transmission optimization module, used to perform data transmission using a multiplexed network transmission protocol, update data based on a differential transmission mechanism, deploy edge nodes to cache high-frequency data, and realize dynamic allocation of transmission bandwidth; a permission security control module, used to classify basic roles and special roles, assign corresponding operation permissions by department, set an authorized IP whitelist, and implement encrypted transmission protection for confidential data; and a parallel computing publishing module, used to complete data fusion and remote sensing inversion verification through Spark parallel computing, and publish a RESTful API that can be shared by multiple departments.

[0084] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for forestry information sharing through multi-source data fusion, characterized in that, Includes the following steps: Multi-source data are classified into levels according to data type and importance; raster data is compressed; patch data is structured into plots according to administrative regions; and sample plot data is associated with patch identification codes and redundancy is removed. A hybrid architecture is built that combines distributed storage with structured data. The core data master node is backed up with three replicas, while general data is stored in shards by type on the slave nodes, and multi-dimensional indexes are established. Data transmission is performed using a multiplexed network transmission protocol, data is updated based on differential transmission, high-frequency data is cached at edge nodes, and transmission bandwidth is dynamically allocated. Define basic and special roles, assign corresponding permissions to departments, set up an authorized IP whitelist, and encrypt the transmission of classified data. Data fusion and remote sensing inversion verification were completed using Spark parallel computing, and a RESTful API was released for sharing by multiple departments.

2. The forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The aforementioned hybrid architecture for building distributed storage and structured data association specifically involves building a storage architecture that combines HDFS and PostgreSQL. HDFS carries large-capacity raster data and structured patch and sample plot data files, while PostgreSQL stores structured data with spatial attributes. HDFS and PostgreSQL achieve metadata synchronization and association calls through an interface. A master-slave node cluster and regional node hierarchical deployment are adopted, with the master node cluster ensuring the core data read and write speed, and the slave node cluster being deployed in a distributed manner according to administrative divisions.

3. The forestry information sharing method based on multi-source data fusion according to claim 2, characterized in that, The general data is stored in fragments by type from the nodes. Specifically, the general data is stored in three levels of general shared data from the nodes. First, storage units are defined according to administrative divisions, and then further subdivided into raster, vector, attribute, and document data fragments according to data types. Each fragment has corresponding storage content and format specifications. The fragments are associated with unique identifiers containing administrative region code information. An association mapping table is established and synchronized to the master node database to trace the source of the core data.

4. The forestry information sharing method based on multi-source data fusion according to claim 3, characterized in that, The establishment of a multi-dimensional index specifically involves constructing a spatial, temporal, keyword, and administrative affiliation index system based on the spatial characteristics, temporal attributes, and business needs of forestry data. The spatial index is constructed for data with spatial coordinates, the temporal index is centered on the data generation and update time, the keyword index focuses on high-frequency retrieval fields, and the administrative affiliation index is associated with administrative divisions and data fragments. Data changes trigger index updates, and index fragments are regularly organized and the structure is optimized.

5. The forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The differential transmission-based data update specifically involves the following: the differential transmission update object is the pre-processed full multi-source forestry data. It adopts a full initialization and incremental update mode, identifies the update content by comparing the data type differences, extracts and encapsulates incremental data packets containing update type and associated identification information, and the data packets enjoy bandwidth priority during transmission. The edge nodes receive the data and synchronously update the cache. The receiving end verifies the integrity of the data packets. If the verification is successful, the data packets are automatically merged. If the verification fails, the interruption is resumed.

6. The forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The dynamic allocation of transmission bandwidth specifically involves constructing a dynamic bandwidth allocation mechanism based on data priority and network status, setting high, medium, and low transmission priorities according to data levels, using a bandwidth scheduling engine to collect network indicators in real time to build a supply and demand model, allocating bandwidth differently according to bandwidth conditions, delineating dedicated channels for edge nodes, supporting temporary expansion, establishing a bandwidth preemption and release mechanism for high-priority tasks to low-priority tasks, and synchronizing the allocation results to the platform.

7. The forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The division of basic roles and special roles specifically involves dividing basic roles and special roles based on the RBAC model, constructing a two-tier system of basic roles and special roles, binding permissions with data hierarchy and departmental responsibilities, and basic roles including read-only roles, modify roles, and download roles. Special roles include census administrators and security auditors. Roles are matched with departments, and permissions are managed through PostgreSQL and synchronized to the shared platform.

8. The forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The process of using Spark parallel computing to complete data fusion and remote sensing inversion verification involves adopting a master node scheduling and slave node distributed deployment mode, connecting to the storage architecture to obtain preprocessed multi-source data, splitting tasks according to data type and administrative region for parallel execution, constructing an integrated dataset by associating sample plot measured data and sample plot attributes with plot identification codes, removing mismatched data after verification, correcting remote sensing inversion results using sample plot measured data, marking anomalies and triggering a review.

9. A forestry information sharing method based on multi-source data fusion according to claim 1, characterized in that, The encrypted transmission of classified data specifically involves the following: classified data is core information related to national ecological security, and a dynamically updated catalog is established by the census administrator and security auditor according to the plan; the data is statically encrypted during preprocessing and dynamically encrypted using the AES-256 algorithm before cross-departmental transmission; the transmitted data is marked with a "classified identifier"; and the keys are generated and distributed hierarchically and updated regularly.

10. A system utilizing the forestry information sharing method of multi-source data fusion as described in claim 1, characterized in that, include: The data classification and processing module is used to classify multi-source data according to data type and importance. It compresses raster data using a high-fidelity, high-compression-ratio image compression format, structures map patch data by administrative region, and associates map patch identification codes with sample plot data and performs redundancy removal. The hybrid storage management module is used to build a hybrid architecture of distributed storage and structured data association. It implements three-replica backups for the core data master node, stores general data in fragments by type on the slave nodes, and establishes a multi-dimensional data index. The data transmission optimization module is used to transmit data using a multiplexed network transmission protocol, update data based on a differential transmission mechanism, deploy edge nodes to cache high-frequency data, and realize dynamic allocation of transmission bandwidth. The access control module is used to classify basic roles and special roles, assign corresponding operation permissions to departments, set up an authorized IP whitelist, and implement encrypted transmission protection for classified data. The parallel computing publishing module is used to complete data fusion and remote sensing inversion verification through Spark parallel computing, and publish a RESTful API that can be shared by multiple departments.

Images