Intelligent livestock breeding full-chain visualization method and system based on GIS

Abstract

The invention provides a GIS-based intelligent livestock breeding full-chain visualization method and system, and the method comprises the steps: collecting the basic information of a farm, livestock information, production process data and other multi-source data through a standardized data interface, and carrying out the correlation mapping of business data and GIS spatial data based on a geographic coordinate system; constructing a three-dimensional data model containing spatial information, attribute information and relation information; three-dimensional scene rendering is realized by utilizing a WebGL technology, a macroscopic visual engine, a mesoscopic visual engine and a microcosmic visual engine are constructed, and multi-scale visual display is realized; constructing a full-life-cycle data model of the livestock products, and realizing visual display of full-chain traceability of the livestock products; based on a visual data analysis method, a decision support system for industrial layout analysis, breeding scale prediction, epidemic disease risk assessment and market supply and demand analysis is constructed. Visual management of the whole industry chain of livestock breeding is achieved, the industry management efficiency is improved, the epidemic disease risk prevention and control capacity is enhanced, and data support is provided for decision making of the livestock industry.

Description

Technical Field

[0001] The present invention relates to the technical field of geographic information systems, and in particular to a GIS-based smart livestock breeding full-chain visualization method and system, which is applied to the intelligent management and visual supervision of the livestock breeding industry. Background Art

[0002] Animal husbandry is a vital component of the national economy. With socioeconomic development and rising consumer demand, the industry faces significant opportunities for transformation and upgrading. However, traditional animal husbandry faces numerous challenges in production management, quality and safety, disease prevention and control, and traceability, hindering its high-quality development.

[0003] At present, livestock breeding management mainly presents the following characteristics: First, data collection is decentralized and lacks unified standards. Different farms use different information systems to record production data, resulting in inconsistent data formats and incompatible interfaces, making it difficult to form industrial big data; second, there is a lack of integration in the spatial dimension, which cannot reflect the regional distribution and spatial relationships of the breeding industry. Traditional systems mainly focus on the internal management of individual farms, ignoring the distribution characteristics and regional connections of the breeding industry in geographical space; third, the level of visualization is low, making it difficult to intuitively display complex information. Existing systems mostly use two-dimensional tables and charts to display data, which cannot effectively express multidimensional data relationships and spatial hierarchies; fourth, the product traceability system is fragmented, making it difficult to achieve full-chain traceability. The traceability systems at different links are independent of each other and cannot coherently display the complete information chain from breeding to sales.

[0004] Therefore, there is an urgent need to develop an intelligent livestock farming full-chain visualization system that integrates GIS spatial technology and modern visualization technology. By integrating spatial data with business data, multi-scale visualization from macro-regions to micro-farms can be achieved, providing full-chain traceability from breeding to sales, and providing data support for industry supervision and decision-making. Summary of the Invention

[0005] The purpose of the present invention is to provide a GIS-based intelligent livestock breeding full-chain visualization method and system. By collecting and processing farm basic information, livestock information, production process data, transaction data and supervision data, business data is integrated with GIS spatial data to construct a three-dimensional data model, realize multi-scale visualization and full-chain traceability, provide industry analysis and decision support services, and solve the problems of data dispersion, lack of spatial dimension, low visualization and fragmented traceability system in traditional livestock breeding management.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A GIS-based visualization method for the entire chain of intelligent animal husbandry, including:

[0008] Collect basic farm information, livestock information, production process data, transaction data, and regulatory data through standardized data interfaces, and clean, convert, and normalize the collected data to form a unified data resource pool;

[0009] Based on the geographic coordinate system, the business data in the data resource pool is associated and mapped with the GIS spatial data to construct a three-dimensional data model containing spatial information, attribute information and relationship information;

[0010] Using the three-dimensional data model and WebGL technology, a visualization engine with three levels, macro, meso, and micro, was designed. Through layered design, multi-scale visualization from regional overview to individual farm details was achieved.

[0011] Based on the multi-scale visualization, a data model covering the entire life cycle of livestock products, from breeding and farming to slaughter and sales, is constructed. By combining a dynamic timeline with a flow chart, a visualization of the entire chain of livestock product traceability is achieved.

[0012] Based on the above-mentioned visual display and full-chain traceability data, a decision support system is constructed that includes industrial layout analysis, breeding scale forecast, disease risk assessment and market supply and demand analysis to generate a decision support report.

[0013] Basic farm information, livestock information, production process data, transaction data and regulatory data are collected through standardized data interfaces, and the collected data are cleaned, converted and normalized to form a unified data resource pool, including: checking the integrity, consistency and timeliness of the collected basic farm information, livestock information, production process data, transaction data and regulatory data, identifying and marking abnormal data according to preset data verification rules, performing data error correction and compensation, and outputting data streams; using a distributed database to store the data streams, and establishing a hybrid storage system including relational databases, time series databases and spatial databases to achieve unified management of structured, semi-structured and unstructured data in the data streams.

[0014] Based on the geographic coordinate system, the business data is associated and mapped with GIS spatial data to construct a three-dimensional data model containing spatial information, attribute information and relationship information, including: obtaining and processing basic geographic information data including administrative boundaries, road networks, water systems and terrain, and constructing a standardized spatial reference system; based on the address information and GPS coordinates of the farm, the location of the farm is located in the spatial reference system through geocoding, and a spatial index of the farm is established; the spatial index of the farm is associated with business data including livestock quantity, breed structure and production status, and a spatial attribute association table is constructed, and multi-level LOD technology is used to process data expression at different scales.

[0015] Utilizing the three-dimensional data model and WebGL technology, a visualization engine with three levels, namely macro, meso and micro, is designed, including: generating heat maps and bar charts of regional livestock breeding resource distribution based on regional administrative boundaries and statistical data, displaying macro indicators including the total number of farms, total livestock and breed distribution; for farm groups in a specific area, generating three-dimensional scenes of the spatial distribution, scale structure and production capacity of the farms using custom icons, color coding and shape transformation; for the interior of a single farm, generating a three-dimensional visualization interface of building layout, livestock distribution and equipment configuration, and displaying equipment operation status, environmental parameters and individual livestock information through click interaction.

[0016] Construct a data model for the entire life cycle of livestock products from breeding, farming to slaughtering and sales, including: designing a traceability data model covering breeding, farming, slaughtering, processing and sales links, and defining the data structure of key information points and data association relationships in each link; based on the data structure and GIS map, generate a dynamic path map of the spatial movement trajectory of products from farms to slaughterhouses, processing plants, and sales terminals; use a force-directed graph algorithm to calculate and display the association relationship between livestock products and related entities, and generate a visual view of the responsible entities and handover relationships in the product flow process through nodes and lines.

[0017] The method also includes: based on farm distribution and livestock density data, using spatial statistical methods to calculate the number of susceptible animals and contact risk coefficients in a unit area, and constructing a disease risk assessment model; based on the risk assessment model and combined with historical disease records, using network analysis methods to calculate the probability of disease transmission and identify high-risk areas; performing spatial cluster analysis on the high-risk areas to generate a disease risk distribution map, and generating early warning information for areas where the risk index exceeds a preset threshold.

[0018] The network analysis method is used to calculate the probability of disease transmission and identify high-risk areas, including: collecting personnel flow records, vehicle transportation records, material exchange records and livestock transaction records between farms to construct a farm association network; converting the farm association network into an adjacency matrix, using the Strassen algorithm to calculate the fourth power of the adjacency matrix, and calculating the number of four-ring roads involved by each farm. The four-ring road density is defined as the ratio of the number of four-ring roads involved by a farm to the number of its directly connected neighboring farms; based on the four-ring road density, the DBSCAN algorithm is used to spatially cluster the farms, and areas in the clustering results with a density greater than a preset threshold are marked as transmission risk hotspots.

[0019] After collecting basic farm information, livestock information, production process data, transaction data and regulatory data through a standardized data interface, the method also includes: performing differential privacy protection on the collected data, including: classifying livestock breeding data into four levels according to their sensitivity: public data, internal data, sensitive data and highly sensitive data, and establishing access rights and protection strategies for data of different levels; constructing a linear Markov decision process model and designing an adaptive differential privacy parameter optimization algorithm, wherein the decision regret is defined as the difference in expected loss between the actual decision and the ideal decision, and the decision regret is controlled within a preset range through iterative optimization; and implementing a Laplace noise addition mechanism according to the data sensitivity level and query type, adding random noise that meets the differential privacy requirements to the query results to ensure that the privacy of individual data is not leaked while providing analysis results.

[0020] When constructing the decision support system, the method also includes: abstracting the entities and relationships in the breeding industry chain into nodes and edges of a generalized graph model, wherein the nodes represent entities including farms, slaughterhouses, and feed mills, and the edges represent relationships including supply relationships, capital flows, and logistics paths; optimizing the generalized graph model using a counterexample-guided abstract refinement algorithm, determining the graph area that needs to be refined by calculating the deviation between the inference results and the actual observation data, and dynamically adjusting the degree of abstraction of the nodes and edges; constructing a finite field continuous logic reasoning engine based on the optimized graph model, mapping continuous variables and fuzzy concepts in breeding management to a truth space in the interval [0,1], and generating optimized decision information for breeding management using fuzzy reasoning rules.

[0021] A GIS-based smart livestock farming full-chain visualization system includes: a data acquisition module for collecting basic farm information, livestock information, production process data, transaction data and supervision data through a standardized data interface, and cleaning, converting and normalizing the collected data to form a unified data resource pool; a data association module for associating and mapping the business data in the data resource pool with GIS spatial data based on a geographic coordinate system, and constructing a three-dimensional data model containing spatial information, attribute information and relationship information; a visualization engine module for using the three-dimensional data model and WebGL technology to design The system is designed to include three levels of visualization engines: macro, meso and micro. Through layered design, it realizes multi-scale visualization display from regional overview to details of a single farm; the traceability visualization module is used to build a data model of the entire life cycle of livestock products from breeding, farming to slaughter and sales based on the multi-scale visualization display, and realize the visualization display of the entire chain traceability of livestock products through the combination of dynamic timeline and flow chart; the decision support module is used to build a decision support system including industrial layout analysis, breeding scale prediction, disease risk assessment and market supply and demand analysis based on the visualization display and full-chain traceability data, and generate decision support reports.

[0022] The beneficial effects of the present invention are:

[0023] This system achieves unified collection and management of livestock breeding data, resolves data silos, and provides a data foundation for full-chain analysis and supervision. The use of standardized data interfaces and a hybrid storage system enables unified management of structured, semi-structured, and unstructured data, improving data quality and utilization efficiency.

[0024] By mapping business data with GIS spatial data, a three-dimensional data model was constructed, achieving spatial representation of the livestock farming industry. By integrating spatial location, attribute characteristics, and relationship networks, a complete livestock farming spatial information system was formed.

[0025] We developed visualization engines at the macro, meso, and micro levels, enabling multi-scale visualization from regional overviews to individual farm details. Three-dimensional rendering based on WebGL technology provides a more intuitive and interactive data visualization experience.

[0026] A data model for the entire life cycle of livestock products has been constructed, enabling a visual display of the entire livestock product traceability chain. By combining a dynamic timeline with a flow chart, it comprehensively displays complete information about the product, from breeding and farming to slaughter and sales, improving the traceability of product quality and safety.

[0027] Based on spatial statistics and network analysis methods, a disease risk assessment model was developed to improve the scientificity and accuracy of disease prevention and control. Through four-ring density analysis and spatial clustering algorithms, hotspots of disease transmission risk were identified to support epidemic prevention decision-making.

[0028] Differential privacy protection is implemented, effectively protecting the security of sensitive data while ensuring data analysis capabilities. Through the Markov decision process model and adaptive differential privacy parameter optimization, it balances data availability and privacy protection requirements.

[0029] A decision support system based on a generalized graph model was constructed to provide data support for livestock industry supervision and decision-making. Through a counterexample-guided abstract refinement algorithm and a finite field continuous logic reasoning engine, the scientific nature and accuracy of decision-making were improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the embodiments. The drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and, together with the specification, are used to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without inventive effort.

[0031] Figure 1 This is the overall flow chart of the GIS-based visualization method for the entire chain of intelligent animal husbandry.

[0032] Figure 2 This is a flow chart of data collection and processing of the present invention;

[0033] Figure 3 This is a flowchart of the association mapping between business data and GIS spatial data of the present invention;

[0034] Figure 4 This is a multi-scale visualization flowchart of the present invention;

[0035] Figure 5 This is a visual flow chart of the whole chain traceability of livestock products of the present invention;

[0036] Figure 6 This is a structural diagram of the GIS-based visualization system for the entire chain of intelligent animal husbandry in the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the disclosure for which protection is sought, but merely represents selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present disclosure.

[0038] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0039] The term "and / or" herein simply describes an association relationship, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, the simultaneous existence of A and B, and the existence of B alone. In addition, the term "at least one" herein refers to any combination of at least two of any one or more of a plurality of items. For example, "at least one of A, B, and C" can represent any one or more elements selected from the set consisting of A, B, and C.

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0041] The GIS-based intelligent livestock farming full-chain visualization method provided by this paper implements data collection, processing, visualization, traceability, and decision support for the entire livestock farming chain through five key steps. Each step is explained in detail below.

[0042] Example 1

[0043] See also Figure 1 The present invention provides a GIS-based visualization method for the entire chain of intelligent animal husbandry, comprising the following steps:

[0044] Step S1: Collect basic farm information, livestock information, production process data, transaction data, and regulatory data through a standardized data interface, and clean, convert, and normalize the collected data to form a unified data resource pool;

[0045] First, by establishing standardized data interfaces, comprehensive data collection is achieved for basic farm information, livestock information, production process data, transaction data, and regulatory data. Basic farm information includes the farm's location, site size, infrastructure configuration, and operating entity; livestock information includes breed, number, age structure, and health status; production process data includes feeding management, vaccinations, and growth monitoring; transaction data includes purchase and sales records, pricing information, and transaction entities; and regulatory data includes quarantine inspection, quality safety, and environmental assessments. The system collects raw data through various methods, including REST APIs, direct database connections, IoT device interfaces, and file import interfaces, ensuring the diversity and integrity of data sources. The collected raw data undergoes integrity, consistency, and timeliness checks. Pre-set data validation rules identify anomalies, and data correction and compensation are implemented to produce a high-quality data stream. This processed data is stored in a distributed database architecture, establishing a hybrid storage system combining relational, time-series, and spatial databases. This enables unified management of structured, semi-structured, and unstructured data, ultimately forming a dynamically updated livestock data resource pool.

[0046] Step S2: Based on the geographic coordinate system, the business data in the data resource pool is associated and mapped with the GIS spatial data to construct a three-dimensional data model containing spatial information, attribute information and relationship information;

[0047] Based on a geographic coordinate system, the system maps and associates business data with GIS spatial data to construct a three-dimensional data model. First, the system acquires and processes basic geographic information such as administrative boundaries, road networks, water systems, and topography to establish a standardized spatial reference system, which serves as the foundation for spatial data association. Then, based on the farm's address information and GPS coordinates, the system uses geocoding technology to accurately locate the farm's location within the spatial reference system. This creates a spatial index for the farm, achieving precise representation of the farm's location. Next, the system associates the farm's spatial index with business data, including livestock numbers, breed structure, and production status, to construct a spatial attribute association table, transforming non-spatial business data into geographic information with spatial attributes. Finally, the system utilizes multi-level-of-detail (LOD) technology to process data representation at different scales. The system automatically adjusts the level of detail of displayed content based on viewing distance and viewing angle, displaying a simplified model at a distance and a detailed model at close range. This results in a complete three-dimensional data model that incorporates spatial, attribute, and relationship information. This model provides the data foundation and spatial framework for subsequent visualization.

[0048] Step S3: Using the three-dimensional data model and WebGL technology, a visualization engine including macro, meso, and micro levels is designed, and a multi-scale visualization display from regional overview to individual farm details is achieved through layered design;

[0049] Utilizing 3D data models and WebGL technology, a visualization engine encompassing macro, meso, and micro levels was designed, enabling multi-scale visualization, from regional overviews to detailed individual farm details. At the macro level, the system generates heat maps and bar charts of regional livestock resource distribution based on regional administrative boundaries and statistical data. These charts visually display macro indicators such as the total number of farms, total livestock population, and breed distribution, helping users understand the overall status of regional livestock industry development. At the meso level, the system uses customizable visual elements such as icons, color coding, and shape transformations to generate 3D scenes depicting the spatial distribution, scale structure, and production capacity of farms within specific regions or farming clusters, showcasing the detailed status and spatial relationships of the livestock industry within the region. At the micro level, the system focuses on individual farms, generating 3D visualizations of building layout, livestock distribution, and equipment configuration. Users can interactively view equipment operating status, environmental parameters, and individual livestock information through clicks, enabling precise on-farm management. The system also incorporates spatial scale switching and data linkage mechanisms to seamlessly transition from macro to micro perspectives and display correlated data, providing users with a complete and coherent visualization experience.

[0050] Step S4: Based on the multi-scale visualization, a data model of the entire life cycle of livestock products from breeding and farming to slaughter and sales is constructed, and a visualization display of the entire chain traceability of livestock products is achieved by combining a dynamic timeline and a flow chart;

[0051] In step S4, the system builds a data model covering the entire lifecycle of livestock products, from breeding and farming to slaughter and sales, building on a multi-scale visualization. This model enables a visual display of the entire livestock product traceability chain. First, the system designs a traceability data model encompassing the breeding, farming, slaughtering, processing, and sales stages. This model defines the data structure for key information points and data relationships within each stage, explicitly recording key information for each stage. This includes information about breeding stock and breeding methods in the breeding stage; feeding cycles, feed composition, and medication records in the breeding stage; slaughtering times and inspection results; processing methods and additive use in the processing stage; and sales channels and timing. The system then designs an interactive timeline control to chronologically arrange traceability data, allowing users to browse traceability data in a time-series manner by dragging and dropping on the timeline. Furthermore, the system visualizes the spatial flow of livestock products based on GIS maps, using dynamic paths and markers to represent the spatial movement of products from farms to slaughterhouses, processing plants, and sales outlets. The system also uses a force-directed graph algorithm to calculate and display the relationships between livestock products and related entities such as farms, slaughterhouses, and distributors. This creates a visual representation of the responsible entities and handover relationships during the product flow process through nodes and connections. By comprehensively representing information across time, space, and relationship networks, the system intuitively displays traceability information across the entire livestock product supply chain, enhancing the traceability of product quality and safety.

[0052] Step S5: Based on the visual display and full-chain traceability data, a decision support system is constructed that includes industrial layout analysis, breeding scale forecast, disease risk assessment and market supply and demand analysis, and a decision support report is generated.

[0053] In step S5, the system builds a decision support system based on visualization and full-chain traceability data, encompassing industrial layout analysis, livestock scale forecasting, disease risk assessment, and market supply and demand analysis, and generates a decision support report. Regarding industrial layout analysis, the system constructs an industrial layout assessment model based on farm distribution and scale data, incorporating factors such as land use, environmental capacity, and transportation conditions. Using spatial analysis methods, it identifies industrial clusters and areas of development potential, providing a basis for industrial planning. Regarding livestock scale forecasting, the system uses a time series analysis algorithm based on historical data and growth trends to construct a livestock scale forecasting model. This model enables scientific predictions of future livestock production, livestock output, and market supply, providing a reference for production planning. Regarding disease risk assessment, the system integrates farm distribution, livestock density, and historical disease records to construct a disease risk assessment model. Using network analysis, it calculates the probability of disease transmission, identifies high-risk areas, and provides early warning for disease prevention and control. Regarding market supply and demand analysis, the system integrates livestock production data, slaughter and processing data, and market sales data to construct a market supply and demand balance analysis model. This model predicts market supply and demand relationships and price trends, supporting market decision-making. Based on the analysis results, the system also designs standardized decision-making report templates and generates decision support reports that include data analysis charts, evaluation conclusions and recommended measures, providing comprehensive and scientific decision-making basis for industry managers and decision makers.

[0054] Through the organic combination of the above five steps, the present invention realizes data collection, analysis, visualization, traceability and decision support for the entire chain of animal husbandry, provides a powerful tool for intelligent management and precise decision-making in the animal husbandry industry, and promotes the transformation, upgrading and high-quality development of the animal husbandry industry.

[0055] See also Figure 2 In step S1, basic farm information, livestock information, production process data, transaction data, and regulatory data are collected through standardized data interfaces, and the collected data are cleaned, converted, and standardized to form a unified data resource pool, including:

[0056] Step S1.1: Check the integrity, consistency, and timeliness of the collected farm basic information, livestock information, production process data, transaction data, and regulatory data, identify and mark abnormal data according to preset data verification rules, perform data error correction and compensation, and output the data stream;

[0057] Step S1.2: Use a distributed database to store the data stream, and establish a hybrid storage system including a relational database, a time series database, and a spatial database to achieve unified management of structured, semi-structured, and unstructured data in the data stream.

[0058] This includes the collection of basic farm information (such as site location, scale, construction time, affiliated company, main facilities, staffing, etc.), livestock information (such as breed, quantity, growth stage, health status, pedigree relationship, etc.), production process data (such as feeding, immunization, reproduction, environmental monitoring, disease prevention, etc.), transaction data (such as purchase and sales records, price information, supply and demand relationship, etc.) and regulatory data (such as quarantine inspection, quality safety, environmental protection assessment, etc.). The collection methods include RESTAPI interface, database direct connection interface, IoT device interface and file import interface.

[0059] See also Figure 3 In step S2, the business data is associated and mapped with the GIS spatial data based on the geographic coordinate system to construct a three-dimensional data model containing spatial information, attribute information and relationship information, including:

[0060] Step S2.1: Acquire and process basic geographic information data including administrative boundaries, road networks, water systems, and topography, and construct a standardized spatial reference system;

[0061] Step S2.2: Based on the address information and GPS coordinates of the farm, locate the farm location in the spatial reference system through geocoding, and establish a spatial index of the farm;

[0062] The basic geographic information data obtained in step S2.1 can be obtained from the National Basic Geographic Information Center, provincial and municipal natural resources departments, or third-party map services. The spatial reference system uses the WGS84 or CGCS2000 coordinate system. The geocoding process in step S2.2 combines address matching and coordinate correction to ensure the accurate positioning of the farm.

[0063] Step S2.3: Associating the farm spatial index with the business data including livestock quantity, breed structure and production status, constructing a spatial attribute association table, and using multi-level LOD technology to process data expressions at different scales.

[0064] Step S2.3 is the key to achieving the spatial expression of business data. It associates and integrates the spatial location information of the farm with rich business data to construct a complete spatial attribute relationship model. In this step, the system first needs to establish a mapping relationship between the farm spatial index and the business data. The farm spatial index contains spatial information such as the farm's geographic coordinates, spatial range, and geographic location description, while the business data contains non-spatial attribute information such as livestock quantity, breed structure, and production status. The system uses the farm's unique identifier (such as the farm ID) as the association key to associate these two types of data, forming a one-to-one correspondence between "spatial location-business attributes". This association ensures that each farm not only has a clear spatial location but also carries rich business attribute information, realizing the spatial expression of business data.

[0065] In the process of constructing the spatial attribute association table, the system adopts a standardized data structure design. The main structure of the spatial attribute association table consists of three major parts: spatial index fields, basic attribute fields, and extended attribute fields. The spatial index field stores information such as the spatial coordinates, area range, and spatial relationships of the farm; the basic attribute field stores basic information such as the farm's name, scale, establishment time, and operating entity; and the extended attribute field stores detailed business-related information such as the total number of livestock, breed composition, production stage distribution, stocking density, and environmental parameters. This structural design not only ensures the integrity and standardization of the data, but also provides sufficient flexibility to adapt to the data characteristics of different types of farms. At the same time, the system also sets metadata information such as data type, value range, and update frequency for each attribute field to facilitate subsequent data management and update maintenance.

[0066] To address the need for data representation at different scales, the system employs multi-level LOD (Level of Detail) technology. The core concept of LOD technology is to dynamically adjust the level of detail of displayed content based on viewing distance and scene importance, optimizing computing resource usage while ensuring visual quality. In this system, LOD technology is primarily embodied in three aspects: spatial geometry representation, attribute information aggregation, and visual symbol simplification. For spatial geometry representation, the system automatically adjusts the level of detail of farm boundaries based on viewing distance, using simplified polygonal representations for distant observations and displaying detailed boundary lines for close-up observations. For attribute information aggregation, the system statistically aggregates attribute data from multiple farms in large-scale views, displaying only statistical indicators such as regional totals and averages, while displaying detailed attribute data for each farm in small-scale views. For visual symbol simplification, the system dynamically adjusts icon size, color richness, and shape complexity based on viewing scale, ensuring a clear and intuitive visual experience at any scale.

[0067] The implementation of LOD technology is based on a complete hierarchical grading mechanism. The system divides the observation scale into six levels: national, provincial, municipal, county, township and farm levels, and each level corresponds to a different data expression strategy. For example, at the national level, the system only displays the statistical data of provincial administrative divisions and the total number of livestock in each province, using heat maps or color scale maps for expression; at the provincial level, the system displays information such as the density and scale distribution of farms at the municipal administrative divisions and each city; at the municipal and county levels, the system begins to display the location and basic attributes of specific farms, such as scale grade, main breeds, etc.; at the township and farm levels, the system displays the detailed boundaries, internal structure and all business attribute data of the farms. The system automatically loads the data of the corresponding level according to the current view scale, and achieves a smooth transition when switching levels to avoid user experience problems caused by visual jumps.

[0068] In order to improve the efficiency of data loading and rendering, the system adopts a pre-calculation and caching strategy. During the system initialization phase, the data expression at different levels is pre-calculated and the results are stored in the cache. When the user zooms or pans the view, the system can quickly obtain the data of the corresponding level from the cache without the need for real-time calculation, which greatly improves the system response speed. At the same time, the system also adopts an asynchronous loading mechanism, which prioritizes loading high-precision data in the central area of ​​the view, and then gradually loads peripheral areas and more detailed data to ensure that users can have a smooth interactive experience during any operation. In addition, the system also implements a dynamic update mechanism. When business data changes, only the data expression of the relevant farms is updated without the need to recalculate the entire scene, further improving the system's operating efficiency and data timeliness.

[0069] Through the above design and implementation, step S2.3 successfully integrated the farm spatial index with business data, constructing a complete three-dimensional data model encompassing spatial location, business attributes, and relationship networks. Multi-level-of-distinction (LOD) technology enabled efficient data representation at different scales, laying a solid data foundation for subsequent multi-scale visualization. This deep integration of spatial and business data transcends the limitations of traditional livestock management systems, which lack spatial dimensions and fragmented data. It enables spatialized and visualized management of the entire livestock industry chain, providing a new perspective and powerful tools for industry decision-making and regulation.

[0070] See also Figure 4 In step S3, using the three-dimensional data model and WebGL technology, a visualization engine with three levels of macro, meso and micro is designed, including:

[0071] Step S3.1: Generate heat maps and bar charts of regional livestock breeding resource distribution based on regional administrative boundaries and statistical data, displaying macro indicators including the total number of farms, total livestock population, and breed distribution;

[0072] Step S3.2: For farm groups in a specific area, a three-dimensional scene of the farm spatial distribution, scale structure, and production capacity is generated using custom icons, color coding, and shape transformation;

[0073] Step S3.3: For a single farm, a three-dimensional visualization interface is generated for the building layout, livestock distribution, and equipment configuration, and the equipment operating status, environmental parameters, and individual livestock information are displayed through click interaction.

[0074] The design of the visualization engine adopts a three-layer architecture consisting of a data access layer, a rendering core layer, and an interactive control layer. The data access layer is responsible for processing data obtained from the data resource pool and converting it into the format required for visualization; the rendering core layer implements underlying rendering based on WebGL technology, and combines SVG and Canvas technologies to achieve two- and three-dimensional hybrid rendering; the interactive control layer implements user interaction processing and interface control, supporting operations such as zooming, panning, rotating, clicking, and querying.

[0075] See also Figure 5 In step S4, a data model of the entire life cycle of livestock products from breeding and farming to slaughter and sales is constructed, including:

[0076] Step S4.1: Design a traceability data model covering the breeding, farming, slaughtering, processing, and sales stages, and define the data structure of key information points and data association relationships in each stage;

[0077] In achieving a visual display of the entire livestock product traceability chain, the core task of step S4 is to build a complete data model of the livestock product lifecycle and, through multi-dimensional visualization, to intuitively display the complete product flow from source to destination. This process involves three key steps: data model design, spatiotemporal trajectory generation, and relationship network construction. These steps are explained in detail below.

[0078] Step S4.2: Based on the data structure and GIS map, a dynamic path map of the spatial movement trajectory of the product from the farm to the slaughterhouse, processing plant, and sales terminal is generated;

[0079] Step S4.3: Use the force-directed graph algorithm to calculate and display the relationship between livestock products and related entities, and generate a visual view of the responsible entities and handover relationships in the product flow process through nodes and lines.

[0080] Among them, the traceability data model in step S4.1 defines the key information of each link, such as the breeding link records the breeding stock information, breeding time, breeding method, etc.; the breeding link records the feeding cycle, feed composition, drug use, health monitoring, etc.; the slaughter link records the slaughter time, inspection results, etc.; the processing link records the processing method, additive use, etc.; the sales link records the sales channel, sales time, price, etc. The spatial movement trajectory in step S4.2 uses animation to display the product flow process, including information such as the movement path, residence time, and handover node. The force-directed graph algorithm in step S4.3 is based on the physical gravity and repulsion model, and automatically layouts the nodes and lines, making the relationship graph structure clear and intuitive.

[0081] Specifically, in step S4.1, the system designed a traceability data model encompassing the breeding, farming, slaughtering, processing, and sales stages, defining the data structure of key information points and data associations within each stage. This model utilizes a multi-level, multi-dimensional structural design to ensure that it can fully record and express key information throughout the entire life cycle of livestock products. In the breeding stage, the model records basic information about breeding stock (such as breed, age, and pedigree), breeding methods (such as natural mating and artificial insemination), breeding time, conception rate, and litter size. In the farming stage, the model records information such as the feeding cycle, feed composition (including feed type, nutrients, and additive use), vaccination records (including vaccine type, vaccination time, and vaccination dosage), and growth monitoring data (including weight changes, health status, and abnormal records). In the slaughtering stage, the model records information such as slaughter time, slaughterhouse information, inspection and quarantine results, and meat grading. In the processing stage, the model records information such as processing time, processing methods, processing parameters, additive use, and quality inspection results. In the sales stage, the model records information such as sales channels, sales time, sales regions, and pricing strategies.

[0082] The core feature of the traceability data model is that it establishes logical connections between various links and realizes the vertical connection of data. The system assigns a unique identifier to each individual or batch of livestock products, and uses this identifier to associate the data of different links to form a complete data chain. In addition, the system has also designed a handover record structure between links, which records in detail the time, responsible parties, handover quantity, quality status and other information of the product flow between different links to ensure the complete traceability of the product flow process. This detailed record of the connection between links is the key foundation for achieving seamless traceability of the entire chain. The traceability data model adopts an object-oriented design method, abstracting each link into an independent data object, and forming a complete object system through inheritance, combination and other relationships. It not only ensures the clear structure of the model, but also provides sufficient flexibility to adapt to the traceability needs of different types of products.

[0083] In step S4.2, the system generates a dynamic path map of the spatial movement trajectory of the product from the farm to the slaughterhouse, processing plant, and sales terminal based on the designed traceability data structure and GIS map. This step first extracts the spatial location information of the product in each link from the traceability data model, including the geographic coordinates of the farm, slaughterhouse, processing plant, and sales terminal. Then, according to the time series of product flow, the system connects these spatial location points in chronological order to form the spatial movement path of the product. Considering that actual transportation usually takes place along traffic roads, the system combines road network data and calculates the actual movement route of the product between each node through a path optimization algorithm to generate a path expression that is more in line with the actual situation.

[0084] In terms of the visual expression of the path, the system adopts a combination of dynamic streamlines and process animations. Dynamic streamlines use lines of different thicknesses to indicate the quantity and scale of product circulation, different colors to indicate product types or quality grades, and changes in the transparency of lines to indicate the passage of time, forming an intuitive spatial flow effect. Process animations use moving icons or particle effects to show the actual movement of products in space. Users can adjust the animation speed or pause to view the status at a specific time point using the time controller. The system also sets interactive points at key nodes of the path. Users can click to view detailed information about the node, such as dwell time, processing methods, quality inspection results, etc. In addition, the system supports multi-granularity path viewing. Users can choose to view the precise path of a single product, or view the comprehensive path of product batches or specific types of products to meet the needs of traceability analysis at different levels.

[0085] Step S4.3 is a crucial step in visualizing the entire traceability chain. The system uses a force-directed graph algorithm to calculate and display the relationships between livestock products and related entities. This algorithm uses nodes and connections to generate a visualization of the responsible entities and handover relationships during the product flow process. A force-directed graph is a graph layout algorithm based on a physical model. It treats nodes in a graph as charged particles, with repulsive forces between them, and edges as springs, which generate attractive forces. By simulating this physical system to achieve energy equilibrium, a graph layout with evenly distributed nodes, consistent edge lengths, and minimal crossovers is achieved, making it particularly suitable for displaying complex network structures.

[0086] During the construction of a force-directed graph, the system first defines how nodes and edges are represented. Nodes are categorized into four types: product nodes (representing individual livestock products or batches), subject nodes (representing the various responsible entities involved in the product flow, such as farms, slaughterhouses, processing plants, logistics companies, and distributors), link nodes (representing the various processing steps a product undergoes), and regulatory nodes (representing regulatory activities such as quality inspection and quarantine testing). Different node types are distinguished by distinct shapes and colors, and the size of nodes is dynamically adjusted based on their importance or data volume. Edges represent relationships between nodes, primarily including product flow relationships (products flow from one entity to another), processing relationships (a subject performs specific processing on a product), and regulatory relationships (regulatory activities target specific products or entities). The thickness of an edge indicates the strength or frequency of the relationship, the color of the edge indicates the type of relationship, and the arrows on the edge indicate the direction of the relationship.

[0087] The core of the force-directed graph algorithm is to define appropriate mechanical models and parameters. In this system, the repulsive force between nodes uses a modified Coulomb's law, where the magnitude of the repulsive force is inversely proportional to the square of the distance between the nodes, and the force coefficient is adjusted according to the node type. The gravitational force of the edge uses a spring model, where the magnitude of the gravitational force is proportional to the difference between the edge length and the ideal edge length, and the spring coefficient is adjusted according to the importance of the edge. In addition, the system introduces clustering gravity, which makes similar nodes or closely related nodes tend to cluster together, enhancing the readability of the graph. In force calculations, the system uses the Barnes-Hut algorithm to optimize, simplifying the force calculation of distant nodes to the overall calculation of the node group, greatly improving computational efficiency and enabling the system to handle large-scale network relationships containing thousands of nodes.

[0088] In order to improve the interactivity and information expression capabilities of force-directed graphs, the system has implemented a series of advanced functions. The first is a multi-level expansion mechanism. In the initial state, only core nodes and relationships are displayed. Users can gradually expand more relationships around the nodes of interest by clicking, avoiding visual confusion caused by displaying too much information at one time. The second is the relationship filtering function. Users can choose to display only specific types of nodes or relationships to facilitate targeted analysis. The third is the time evolution display. The system can display the network status of different periods according to the timeline, showing the dynamic changes in industrial relations. The fourth is the path tracking function. After the user selects the starting point and end point, the system can highlight all possible paths between the two points and provide path comparison analysis to help users identify key flow paths or risk points.

[0089] The implementation of force-directed graphs uses WebGL technology for efficient rendering, supporting the smooth display of large-scale nodes and edges. The system uses shaders to achieve advanced visual effects for nodes and edges, such as the flowing texture of the edges to indicate the direction of material flow, and the luminous effect of the nodes to indicate the level of activity. In terms of interaction, the system supports basic operations such as zooming, panning, and rotation. It also supports advanced interactive functions such as node dragging (users can manually adjust the node position), node freezing (fixing the position of a specific node), and graph capture (selecting a local area of ​​the graph for zoomed analysis). In addition, when the user clicks or hovers over nodes and edges, the system displays a detailed information panel that provides all relevant data and statistical information for the node or relationship.

[0090] Through the visualization of force-directed graphs, the system successfully presents complex product flow relationships and responsible entity networks. Users can clearly see the complete product flow path from source to destination, identify the responsible parties at each stage, understand the coverage of quality and safety supervision, and discover potential risk points or optimization opportunities. This relationship-based traceability visualization transcends the limitations of traditional traceability systems that focus solely on one-way traceability chains. It provides a more comprehensive and systematic perspective on industrial relationships, providing a powerful analytical tool for product quality management, supply chain optimization, and risk prevention and control.

[0091] By implementing steps S4.1 through S4.3, the system successfully constructed a multi-dimensional visualization of the entire livestock product traceability chain, transforming abstract data into an intuitive visual representation, making the product traceability process clear, understandable, and interactive. This innovative traceability visualization method not only meets the oversight needs of regulatory authorities and the management requirements of manufacturers, but also provides consumers with a convenient product quality and safety query experience, achieving comprehensive value that benefits all parties.

[0092] In step S5, the method further includes:

[0093] Step S5.1: Based on farm distribution and livestock density data, spatial statistical methods are used to calculate the number of susceptible animals and the exposure risk coefficient per unit area, and a disease risk assessment model is constructed;

[0094] The number of susceptible animals in step S5.1 refers to the number of animals susceptible to a specific disease, and the exposure risk coefficient takes into account factors such as the distance between farms, transportation convenience, and the frequency of personnel and material exchanges;

[0095] In step S5.1, the system uses spatial statistical methods to calculate the number of susceptible animals and the exposure risk coefficient within a unit area based on farm distribution and livestock density data, and constructs a disease risk assessment model. The first task in model construction is to clarify the scope and characteristics of susceptible animals. Different diseases have different host specificities. Based on the characteristics of the target disease, the system extracts information on eligible susceptible animals from the livestock database, including animal species, age structure, immune status, etc. For example, for foot-and-mouth disease, the system selects all even-toed ungulates as susceptible animals; for avian influenza, poultry is selected as susceptible animals. This targeted screening of susceptible animals ensures the accuracy and pertinence of risk assessment.

[0096] After susceptible animals are identified, the system further calculates the density of susceptible animals within a given area. This calculation takes into account not only the number of animals but also factors such as their immune status, age structure, and breed characteristics. The system uses a weighted approach, assigning different weights to animals with different immune statuses. For example, unimmunized animals are weighted higher than immunized animals, and animals with a longer immunization history are weighted higher than animals with more recent immunizations. The system also considers the animal's age, as young and old animals are generally more susceptible to disease. Through this comprehensive weighted approach, the system generates a more accurate map of susceptible animal density, reflecting the degree of disease susceptibility in different areas.

[0097] The contact risk coefficient is another key indicator in the disease risk assessment model. It reflects the probability of contact between animals and the subsequent transmission of disease. The system calculates the contact risk coefficient from three dimensions: intra-farm contact, inter-farm contact, and contact between the farm and the external environment. The intra-farm contact risk is primarily related to stocking density, feeding practices, and management level. The system calculates the internal contact risk index based on the farm's stocking density data, building layout information, and biosafety measures rating. The inter-farm contact risk is primarily related to the spatial distance between farms, accessibility, and frequency of business transactions. The system calculates the inter-farm contact risk index through spatial buffer zone analysis, road network analysis, and transaction record analysis. The farm-to-external environment contact risk takes into account external risk factors such as wildlife activity areas, migratory bird migration routes, and densely populated areas. The system calculates the external contact risk index by overlaying these risk layers. Finally, the system combines the risk indices of these three dimensions through a weighted average to obtain a comprehensive contact risk coefficient.

[0098] Based on the calculation of susceptible animal density and contact risk coefficient, a unified disease risk assessment model was systematically constructed. The model uses a spatial statistical method based on geographically weighted regression (GWR) and takes into account the changes in the impact of spatial location on risk factors. The basic form of the model is: Risk i=β0(ui,vi)+β1(ui,vi)×SD i +β2(ui,vi)×CR i +β3(ui,vi)×EF i +ε i , where Risk i Represents the disease risk index at location i, SD i represents the density of susceptible animals, CR i Indicates the exposure risk factor, EF i represents environmental factors (such as climate conditions and topographical features), (ui,vi) represents location coordinates, β represents the regression coefficient, and ε represents the error term. Through statistical analysis of historical epidemic data and risk factor data, the system calibrated the model parameters, established a quantitative relationship between epidemic risk and various factors, and formed a complete epidemic risk assessment model.

[0099] Step S5.2: Based on the risk assessment model and in combination with historical disease records, a network analysis method is used to calculate the probability of disease transmission and identify high-risk areas;

[0100] In step S5.2, the system uses network analysis methods based on the risk assessment model and historical disease records to calculate the probability of disease transmission and identify high-risk areas. First, the system collects and organizes historical disease records, including information such as disease type, time of occurrence, location, spread, duration, and control measures. This historical data is an important basis for validating and optimizing risk assessment models and also provides valuable empirical evidence for analyzing disease transmission characteristics. The system uses spatiotemporal clustering analysis to identify the spatiotemporal distribution patterns of historical epidemics, discovering seasonal patterns and spatial clustering characteristics of disease transmission, further enriching the dimensions of risk assessment.

[0101] Based on historical data analysis, the system constructed a farm-related network and used network analysis methods to calculate the probability of disease transmission. The farm-related network is a complex network structure with farms as nodes and connections between farms as edges. The system collects data on personnel movement, vehicle transportation, material exchanges, and livestock transactions between farms, quantifying the type, frequency, and intensity of connections between farms. This creates a multi-level, multi-attribute farm-related network. During network construction, the system assigns different weights to different types of connections. For example, direct livestock transactions typically carry a higher risk of disease transmission and are therefore given a higher weight, while one-time material exchanges, which carry a lower risk, are given a lower weight.

[0102] Based on the constructed farm association network, the system uses a network propagation model to simulate the spread of the disease in the network. The system implements a variety of propagation models, including the SI (susceptible-infected) model, the SIR (susceptible-infected-recovered) model, and the SEIR (susceptible-exposed-infected-recovered) model, etc., and selects the appropriate model according to the characteristics of different diseases. In the propagation simulation, the system takes into account the susceptibility of the node (related to the previously calculated density of susceptible animals), the propagation probability of the edge (related to the connection strength and type), and the influence of external infection sources (such as wild animals, environmental factors, etc.). Through the Monte Carlo simulation method, the system generates a large number of propagation scenarios, calculates the probability of each farm being infected during the outbreak, and the distribution characteristics of the infection time.

[0103] In addition to the traditional network transmission model, the system also innovatively applies the loop analysis method in the network. The system pays special attention to the four-loop structure, that is, the closed loop path formed by four farms. The four-loop is of special significance in the spread of diseases. It represents the path of possible cyclical transmission of the disease, increasing the possibility of the disease persisting and repeatedly breaking out. The system uses the Strassen algorithm to efficiently calculate the fourth power of the adjacency matrix, identify all four-loop structures in the network, and calculate the number of four-loops and the four-loop density of each farm (defined as the ratio of the number of four-loops a farm participates in to the number of directly connected neighboring farms). Farms with a high four-loop density are usually located in important positions in the network and play a key role in the spread of the disease.

[0104] By combining the static risk scores of the risk assessment model with the dynamic transmission probabilities of the network transmission model, the system ultimately determines the regional disease risk level. Risk levels are divided into five levels, ranging from low risk to extremely high risk, each corresponding to different risk characteristics and prevention and control requirements. The system generates a regional disease risk distribution map, visually illustrating the spatial distribution of risk levels, helping managers quickly identify high-risk areas and supporting prevention and control decisions.

[0105] Step S5.3: Perform spatial cluster analysis on the high-risk areas to generate a disease risk distribution map, and generate early warning information for areas where the risk index exceeds a preset threshold.

[0106] In step S5.3, the system performs spatial cluster analysis on the identified high-risk areas, generating a disease risk distribution map and warning information for areas where the risk index exceeds a preset threshold. The purpose of spatial cluster analysis is to group spatially close farms with similar risk levels into risk blocks for unified management and prevention. The system uses DBSCAN (density-based spatial clustering algorithm) for cluster analysis, which is particularly suitable for discovering clusters of arbitrary shapes, can identify noise points, and does not require a pre-specified number of clusters.

[0107] In the application of the DBSCAN algorithm, the system defines two key parameters: ε (neighborhood radius) and MinPts (minimum number of points). ε determines the maximum distance for a point to be considered a neighbor, and MinPts determines the minimum number of points required to form a cluster core. The settings of these two parameters directly affect the results and quality of clustering. The system uses an adaptive parameter setting method to dynamically adjust parameter values ​​based on the density distribution and risk level distribution of regional farms. In dense areas, a smaller ε value is used to identify fine-grained risk clusters; in sparse areas, a larger ε value is used to ensure that there are enough points to form clusters. At the same time, the system takes into account the differences in farm size and importance, assigning different weights to farms of different sizes, so that large farms have greater influence in cluster formation.

[0108] In addition to spatial location, the clustering process also considers the similarity of risk attributes. The system employs a multidimensional clustering approach, combining spatial distance and risk attribute distance into a comprehensive distance metric. Risk attributes include the previously calculated density of susceptible animals, contact risk coefficients, and network location importance (such as the density of the Fourth Ring Road). This multidimensional clustering approach ensures that clustering results are not only spatially continuous but also similar in risk characteristics, improving the practicality and interpretability of the clustering.

[0109] After clustering is complete, the system performs statistical analysis on each cluster block, calculating indicators such as the average risk index, maximum risk index, risk variance, number of farms, and total livestock population, and generates a block-level risk assessment report. The system also analyzes the spatial morphological characteristics of the cluster blocks, such as area, perimeter, and shape complexity. These characteristics are valuable for formulating disease prevention and control strategies. For example, blocks with complex shapes and long borders generally indicate that effective quarantine measures are more difficult to implement and may require more stringent prevention and control measures.

[0110] Based on the results of cluster analysis, the system generates a precise disease risk distribution map. This map uses a hierarchical coloring scheme, with a gradient from green to red representing low-risk to high-risk areas, visually illustrating the spatial distribution of risk. The system also annotates key risk clusters on the map, displaying the risk level, number of farms, and key risk characteristics within each cluster, providing users with a wealth of information. Users can interactively view detailed information about each cluster, including a list of the specific farms that comprise the cluster, a detailed risk score for the cluster, and historical epidemic records.

[0111] For areas where the risk index exceeds the preset threshold, the system automatically generates warning information. Warning information includes the warning level, warning area, risk type, risk description, warning time, and recommended measures. Warning levels are divided into three levels: Level 1 warning (red) indicates extremely high risk, requiring immediate emergency prevention and control measures; Level 2 warning (orange) indicates high risk, requiring enhanced monitoring and preparation of prevention and control measures; Level 3 warning (yellow) indicates moderate risk, requiring vigilance and increased monitoring frequency. Warning information is sent to relevant responsible persons and management departments through the system's built-in message push mechanism, ensuring that risk information is transmitted in a timely manner and promoting rapid response.

[0112] In order to enhance the accuracy and operability of early warnings, the system also provides customized prevention and control recommendations for each early warning message. These recommendations are based on regional characteristics, risk types, and historical experience, and cover specific measures in multiple aspects such as enhanced monitoring, disinfection and isolation, movement restrictions, and vaccination. For example, for high-risk areas caused by the high correlation of farms, the system may recommend temporarily restricting the flow of people and vehicles between farms in the area, and implementing enhanced disinfection of farms at key nodes; for risks caused by excessive density of susceptible animals, it may be recommended to speed up the vaccination process and increase the frequency of sampling and testing. These targeted recommendations greatly enhance the practical value of early warning information and help managers quickly formulate effective prevention and control strategies.

[0113] The system also implements a tracking and evaluation mechanism for early warning effectiveness. After an alert is issued, the system continuously monitors changes in the alert area, including changes in the risk index, the implementation of prevention and control measures, and the actual outbreak. By comparing changes before and after the alert, the system evaluates the accuracy of the alert and the effectiveness of the prevention and control measures, providing feedback for optimizing the early warning model and prevention and control strategy. This closed-loop early warning evaluation mechanism enables the system to continuously learn and improve, thereby enhancing the accuracy and practicality of the alerts.

[0114] Through the implementation of these three steps, a comprehensive disease risk assessment and early warning mechanism has been systematically established, enabling full-process management from risk factor analysis to precise early warning. This mechanism organically combines spatial statistical methods, network analysis methods, and spatial cluster analysis, fully leveraging the spatial analysis capabilities of geographic information systems (GIS). It provides a scientific and efficient disease risk management solution, offering strong technical support for the healthy development of the livestock industry and public health security.

[0115] Example 2

[0116] Based on Example 1, the disease risk analysis is further described in detail. The network analysis method is used to calculate the probability of disease transmission and identify high-risk areas, including:

[0117] Step S6.1: Collect records of personnel flow, vehicle transportation, material exchange, and livestock transactions between farms to build a farm association network;

[0118] Step S6.2: Convert the farm association network into an adjacency matrix, calculate the fourth power of the adjacency matrix using the Strassen algorithm, and calculate the number of four-loops that each farm participates in. The four-loop density is defined as the ratio of the number of four-loops that a farm participates in to the number of its directly connected neighboring farms.

[0119] In this analysis, a four-ring path refers to a closed path encompassing four farms within a farm network. The importance of a four-ring path lies in its ability to represent a potential path for disease transmission, allowing the disease to continue spreading within this closed loop. Farms with a high density of four-ring paths are therefore more likely to participate in the circulation and occupy a key position in the disease transmission network.

[0120] Strassen's algorithm is an efficient matrix multiplication algorithm with a time complexity of O(n 2.807 ), which is faster than the traditional matrix multiplication O(n 3 ) is more efficient. In this embodiment, by calculating the fourth power of the adjacency matrix, the number of four-loops in the network can be quickly counted. The fourth power of the adjacency matrix A 4 Element A 4 [i,j] represents the number of paths of length 4 from node i to node j, and the diagonal elements A 4 [i,i] represents the number of loops of length 4 starting from node i, that is, the number of four loops in which node i participates.

[0121] Step S6.3: Based on the density of the four ring roads, the DBSCAN algorithm is used to spatially cluster the farms, and areas in the clustering results with a density greater than a preset threshold are marked as transmission risk hotspots.

[0122] Based on the calculated four-ring density, the DBSCAN algorithm was used for spatial clustering. DBSCAN is a density-based spatial clustering algorithm suitable for discovering clusters of arbitrary shapes and capable of identifying noise points. In disease risk analysis, the DBSCAN algorithm can identify clusters of farms that are spatially concentrated and have high four-ring density based on their spatial location and four-ring density. These clusters are labeled as transmission risk hotspots.

[0123] The identification process of transmission risk hotspots adopts a hierarchical approach. First, an initial abstract representation of the farm association network is performed, and entities and relationships with similar characteristics are aggregated to obtain a highly abstract graph model. Then, impact propagation analysis is performed on the abstract graph. By comparing the analysis results with the actual observation data, key deviation data are extracted as counterexamples to guide the optimization of the graph model. Then, a hierarchical decomposition strategy is adopted to prioritize the refinement of the graph areas related to the counterexample set, decomposing them into finer-grained entities and recalculating their relationship attributes. Finally, an information gain evaluation is performed on each refinement operation. When the information gain value falls below the preset threshold, the refinement is stopped, and the final transmission risk hotspot identification result that balances analysis accuracy and computational efficiency is obtained.

[0124] Based on the identified transmission risk hotspots, targeted transmission cutting-off strategies can be generated: each transmission risk hotspot is regarded as a subgraph, and the Karger algorithm is applied to calculate the minimum set of edges required to isolate the subgraph from the external network, thereby obtaining the set of connections between key farms that need to be cut off; the cutting-off strategies are optimized and sorted based on the weight evaluation of the edges, and the farms that need to be monitored, the traffic routes that need to be temporarily restricted, and the key points for enhanced disinfection are identified to obtain a graded list of prevention and control measures; the changes in the number of the four ring roads in the network and the degree of reduction in transmission risk after the implementation of the cutting-off strategy are evaluated through simulation calculations to evaluate the effectiveness of the strategy; based on the real-time update of farm-related data, the network graph model is dynamically updated, and the statistical results of the four ring roads are updated using incremental calculation to form a dynamic monitoring plan for transmission risk hotspots.

[0125] In one embodiment, steps 6.2-6.3 are described in detail as follows:

[0126] Step S6.2 is the core of the disease transmission network analysis. The system first converts the farm-to-farm network into a standard adjacency matrix. During this conversion, the system constructs an n×n matrix A, where n is the total number of farms. Matrix elements A[i,j] represent the strength of the connection between farm i and farm j. Connection strength is a comprehensive indicator reflecting the likelihood of disease transmission between two farms. It is determined by multiple factors: the frequency of personnel movement (such as regular visits by technicians and veterinarians), the frequency of vehicle traffic (such as feed and live animal transportation), material exchange records (such as equipment sharing and feed supply), and livestock transaction records (such as piglet purchases and fattening animal transactions). The system assigns different weights to these factors. For example, direct livestock transactions are given the highest weight, followed by personnel movement, and material exchanges and indirect connections are given lower weights. By combining these weighted factors, the system calculates a quantitative value for the connection strength between farms and populates the adjacency matrix.

[0127] In certain application scenarios, the system also supports simplifying the adjacency matrix into a binary matrix. That is, when the strength of the association between two farms exceeds a certain threshold, the corresponding matrix element is set to 1, otherwise it is set to 0. This simplification can improve computational efficiency in some network analysis algorithms, but it will result in a certain loss of information accuracy. The system flexibly chooses to use a weighted matrix or a binary matrix based on actual needs and data scale. In large-scale network analysis, the system also uses sparse matrix storage technology, storing only the positions and values ​​of non-zero elements, significantly reducing storage space requirements and computational complexity.

[0128] After the adjacency matrix is ​​constructed, the system uses the Strassen algorithm to calculate the fourth power of the adjacency matrix. The Strassen algorithm is an efficient matrix multiplication algorithm with a time complexity of O(n log(2) )≈O(n 2.807 ), compared with the traditional matrix multiplication algorithm O(n 3 ) has significant performance advantages, especially when processing large matrices. The core idea of ​​the Strassen algorithm is to use a divide-and-conquer strategy to convert large matrix multiplication into submatrix operations, reducing the total number of multiplication operations. Specifically, for two 2×2 matrices, the traditional algorithm requires 8 multiplications and 4 additions, while the Strassen algorithm only requires 7 multiplications and 18 additions. Although the number of additions increases, the performance improvement brought by the reduction in multiplication operations far outweighs the increased overhead of addition operations in large matrix calculations.

[0129] In this system, calculating the fourth power of the adjacency matrix has special significance. 4 Element A in 4 [i,j] represents the number of paths of length 4 from farm i to farm j. In particular, the diagonal elements A 4 [i,i] represents the number of paths of length 4 starting and ending at farm i, that is, the number of four-loop paths in which farm i participates. A four-loop path is a closed loop formed by four nodes (farms) in a network. It is of great significance in disease transmission networks: first, a four-loop path represents a path for the disease to circulate, increasing the likelihood of the disease persisting within a region; second, the existence of a four-loop path complicates disease control within a region, because even if a transmission path is cut off, the disease may continue to spread through other paths within the loop; finally, farms participating in a large number of four-loop paths are often at the core of the network, acting as "super spreaders" in disease transmission.

[0130] The system not only calculates the total number of four ring roads each farm participates in, but also further calculates the four ring road density indicator. The four ring road density is defined as the ratio of the number of four ring roads a farm participates in to the number of its directly connected neighboring farms. This definition takes into account the differences in farm size and connectivity, providing a fairer comparison benchmark. For example, a large farm may participate in a large number of four ring roads due to its extensive business relationships, but if it also has a large number of neighboring farms, the four ring road density may not be high; on the contrary, a small or medium-sized farm may not participate in a large number of four ring roads in total, but if it has a small number of neighbors, the four ring road density may be high, indicating that it occupies a key position in the local network. By calculating the four ring road density, the system can more accurately identify key nodes in the network without being biased by differences in farm size.

[0131] In the actual implementation of the four-loop calculation, the system also adopts a series of optimization measures. For large-scale networks, directly calculating A 4 The system may face the problem of high computational complexity. The system adopts an incremental computation strategy, firstly calculating A 2 , then based on A 2 Calculate A 4 , reducing repeated calculations. Furthermore, the system implements a parallel computing framework, dividing matrices into multiple sub-blocks and assigning them to multiple computing cores for simultaneous processing. This fully leverages the parallel computing capabilities of modern multi-core processors, further improving computational efficiency. For ultra-large-scale networks, the system also supports distributed computing, assigning computational tasks to multiple servers for collaborative completion, thus breaking through the limitations of single-machine computing capabilities.

[0132] In step S6.3, the system uses the DBSCAN algorithm to spatially cluster the farms based on the calculated four-loop density and identify transmission risk hotspots. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based spatial clustering algorithm that can discover clusters of any shape and process noise points. It is particularly suitable for identifying areas of density variation in spatial data. Compared with traditional clustering algorithms (such as K-means), DBSCAN does not require the pre-specification of the number of clusters and can automatically discover natural clusters in the data, which is more in line with the actual distribution of risk hotspots in the disease transmission network.

[0133] The core idea of ​​DBSCAN algorithm is to group data points based on density connectivity. The algorithm defines two key parameters: ε (neighborhood radius) and MinPts (minimum number of points). For a data point p, its ε-neighborhood N ε(p) is defined as the set of all points within a distance of p of no more than ε. Based on the number of points in the ε-neighborhood, DBSCAN divides data points into three categories: core points (the ε-neighborhood contains at least MinPts points), border points (non-core points but within the ε-neighborhood of a core point), and noise points (neither core points nor border points). The algorithm expands the density connectivity of core points and groups all densely connected points into a single cluster.

[0134] In this system, the DBSCAN algorithm uses a set of spatial points of farms with four-way density attributes. Each farm is represented as a multidimensional data point, including its geographic coordinates (longitude and latitude) and four-way density values. The system uses a hybrid distance metric that comprehensively considers spatial distance and attribute similarity. Specifically, the distance d between two farms is calculated as: d = α·d geo +(1-α)·d attr , where d geo is the geographic distance (such as Euclidean distance or Haversin distance), d attr is a measure of the difference in the density of the four-loops, and α is a balancing factor (0 ≤ α ≤ 1) that adjusts the relative importance of spatial distance and attribute distance in clustering. This hybrid distance metric ensures that the clustering results take into account both the spatial proximity of farms and their similar roles in the transmission network.

[0135] The selection of the parameters ε and MinPts has a significant impact on the effectiveness of the DBSCAN algorithm. The system uses an adaptive parameter setting method to dynamically determine parameter values ​​based on data distribution characteristics. The ε value is determined using the k-distance graph method: the distance from each point to its kth nearest neighbor is calculated (k is typically set to dimension + 1 or greater). These distances are then sorted and graphed, and the "inflection point" in the graph is identified as the ε value. This method can adapt to differences in farm density across regions. The MinPts value is set according to the empirical formula MinPts ≥ ln(n), where n is the total number of data points. In practice, the system also allows users to manually adjust these parameters based on their expertise to suit specific analysis needs.

[0136] After executing the DBSCAN algorithm, the system obtained a series of farm clusters, each cluster representing a group of farms that are spatially adjacent and have similar four-ring density. The system further analyzes the characteristics of these clusters and calculates indicators such as the average four-ring density, maximum four-ring density, number of points, and spatial range of each cluster. The average four-ring density of the cluster reflects the overall importance of the farms in the area in the disease transmission network, while the number of points and spatial range reflect the size and scope of the risk area. The system compares the average four-ring density of the cluster with the preset threshold and marks the clusters with a density greater than the threshold as transmission risk hotspots. The determination of the preset threshold is usually based on historical epidemic data and expert experience, and can also be set to a specific percentile (such as the top 10%) based on the overall data distribution.

[0137] For the identified transmission risk hotspots, the system conducted a more in-depth feature analysis and visual display. In terms of feature analysis, the system calculated the morphological characteristics (such as area, perimeter, shape complexity), network characteristics (such as internal connection density, external connection strength) and risk transmission characteristics (such as theoretical transmission speed, impact range) of the hotspot area. These characteristics provide a scientific basis for understanding the nature of risk hotspots and formulating targeted prevention and control strategies. In terms of visual display, the system marks the risk hotspot area with highlighted colors on the map and provides an interactive query function, allowing users to view detailed information on the hotspot, including a list of specific farms that constitute the hotspot, risk indicators of the hotspot, and recommended prevention and control measures.

[0138] The system also implements a dynamic monitoring mechanism for risk hotspots. As farm-related network data is updated in real time, the system regularly recalculates the density of the four ring roads and performs cluster analysis to update the identification of risk hotspots. The system tracks hotspot changes, including the emergence of new hotspots, the expansion or contraction of existing hotspots, and changes in hotspot risk levels. It generates change reports to alert managers to the dynamic changes in the risk landscape. This dynamic monitoring mechanism ensures timely and accurate risk assessments, providing continuous decision-making support for disease prevention and control.

[0139] Based on the results of risk hotspot identification, the system also generates targeted prevention and control recommendations. For each risk hotspot, the system analyzes its formation causes and core characteristics, and proposes customized prevention and control measures. For example, for risk hotspots formed by frequent livestock transactions, the system may recommend strengthening transaction supervision and quarantine inspections; for risk hotspots formed by frequent personnel movement, it may recommend strengthening biosafety measures and personnel disinfection; for risk hotspots occupying key positions in the network, the system may recommend the implementation of stricter isolation measures and monitoring plans. These targeted prevention and control recommendations greatly improve the efficiency and accuracy of risk management, helping managers to concentrate limited prevention and control resources on the most critical areas and links.

[0140] By combining adjacency matrix fourth-power calculations with DBSCAN spatial cluster analysis, the system successfully identified disease transmission risk hotspots based on network structure and spatial distribution. This method, which organically combines graph theory and spatial statistical analysis, considers both the interconnected network structure and the spatial distribution characteristics of farms, providing a more comprehensive and accurate risk assessment. Compared with traditional methods, this four-ring density-based risk hotspot identification method better captures the network effects and spatial clustering characteristics of disease transmission, providing strong scientific support for disease prevention and control decision-making.

[0141] Example 3

[0142] Based on Example 1, the data privacy protection mechanism is further described. After collecting farm basic information, livestock information, production process data, transaction data, and regulatory data through a standardized data interface, the method further includes: performing differential privacy protection on the collected data, including:

[0143] Step S7.1: Classify livestock breeding data into four levels based on their sensitivity: public data, internal data, sensitive data, and highly sensitive data, and establish access rights and protection strategies for data at different levels;

[0144] Among them, the data sensitivity classification standards in step S7.1 are as follows: public data refers to data that is publicly accessible and does not involve commercial secrets and privacy, such as administrative divisions, basic geographic information, etc.; internal data refers to data used within the organization and has little impact after leakage, such as basic information of farms, total output of livestock products, etc.; sensitive data refers to data that involves commercial interests and personal privacy but does not affect national security, such as specific breeding technical parameters, records of epidemic occurrence, etc.; highly sensitive data refers to data involving national security, public safety or major commercial secrets, such as the storage of epidemic prevention materials and emergency plans.

[0145] Step S7.2: Construct a linear Markov decision process model and design an adaptive differential privacy parameter optimization algorithm. The decision regret is defined as the difference between the expected loss of the actual decision and the ideal decision. The decision regret is controlled within a preset range through iterative optimization.

[0146] The linear Markov decision process (MDP) model in step S7.2 abstracts the livestock data analysis process into a series of state transitions and decision choices. In this model, the state represents the current stage of data analysis, the action represents a possible analysis operation, the reward function represents the utility of the operation, and the state transition probability represents the state change caused by different operations. The linear nature of the model is reflected in the fact that the reward function and state transition probability can be expressed as a linear combination of state-action pairs.

[0147] The system constructs a linear Markov decision process model and designs an adaptive differential privacy parameter optimization algorithm, achieving a balance between privacy protection and data availability by controlling decision regret. A Markov decision process (MDP) is a mathematical framework for modeling decision making in situations where the outcomes are partially random and partially controlled by the decision maker. In this system, the MDP is used to simulate the dynamic interaction between data querying and privacy protection. Its core components include a state space, an action space, a transition probability function, a reward function, and a discount factor.

[0148] The state space of a linear MDP model describes the various possible states a system may enter during the data analysis process, including data preparation, query execution, and result analysis. Each state contains multiple attributes, such as the currently consumed privacy budget, the number of historical queries, the query type distribution, and data sensitivity. The action space defines the actions the system can take in each state, primarily including the selection of privacy parameters (such as the value of ε), the choice of noise mechanism (such as the Laplace or Gaussian mechanism), and the privacy budget allocation strategy. The transition probability function describes the probability distribution of the system transitioning from one state to another. These probabilities are influenced by the selected actions, reflecting the stochastic nature of data querying and privacy protection. The reward function quantifies the benefits of taking different actions in each state, consisting of a weighted combination of data availability benefits (the accuracy of query results) and privacy protection benefits (the level of privacy protection achieved). A characteristic of linear MDPs is that the reward function and transition probabilities can be expressed as a linear combination of the feature vectors of state-action pairs, which greatly simplifies the model complexity and improves computational efficiency.

[0149] Based on the linear MDP model, the system designed an adaptive differential privacy parameter optimization algorithm. The core goal of this algorithm is to minimize decision regret, that is, to maximize the accuracy of data analysis results while ensuring privacy protection requirements. Decision regret is defined as the difference in expected loss between the actual decision (after adding privacy protection noise) and the ideal decision (in the absence of noise). It quantifies the degree of reduction in decision quality caused by privacy protection. Specifically, if V*(s) represents the maximum cumulative benefit of the ideal situation in state s, V π (s) represents the expected cumulative benefit of executing strategy π in state s, then the decision regret can be expressed as Regret(π)=V*(s)-V π (s).

[0150] The adaptive differential privacy parameter optimization algorithm uses an iterative learning approach to gradually reduce decision regret by continuously adjusting policy parameters. The algorithm first initializes a base policy, which selects the privacy parameter ε and noise mechanism based on the current state. During each query, the system records state transitions, selected actions, and rewards obtained, forming an experience sample. Based on these samples, the algorithm uses linear function approximation to estimate the state-action value function Q(s,a), which is the expected cumulative reward of performing action a in state s. The function approximator parameters are continuously updated by minimizing the mean squared error between the estimated value and the actual reward. During the policy update phase, the algorithm uses an ε-greedy strategy, selecting the action with the highest estimated value with a certain probability, optimizing the decision-making process in a balanced exploration-exploitation manner.

[0151] A key innovation of the algorithm is the introduction of an adaptive privacy budget allocation mechanism. Based on query history and data characteristics, the system automatically adjusts the privacy budget allocation ratio for different query types and different data components. Specifically, the algorithm tracks the frequency and importance of each query type, as well as the query frequency and sensitivity of each data component, and dynamically adjusts budget allocation based on this information. For high-frequency and important query types, more privacy budget is allocated to ensure accurate results; for highly sensitive but less frequently queried data components, budget allocation is appropriately reduced to conserve the overall budget. This adaptive allocation strategy significantly improves the efficiency of privacy budget utilization, providing higher query accuracy for the same total budget.

[0152] Another innovative aspect of the algorithm is its multi-level privacy assurance mechanism. The system not only considers the privacy protection of a single query but also controls the cumulative privacy leakage of multiple consecutive queries. Based on the combinatorial nature of differential privacy, the system tracks the total consumed privacy budget and sets an upper limit. When the budget approaches the upper limit, the system automatically increases the noise level or rejects certain sensitive queries to ensure that the overall privacy protection level is not significantly reduced by repeated queries. Furthermore, the system implements query sensitivity analysis, automatically identifying query patterns that may lead to serious privacy leakage and applying stricter protection measures for these queries, such as increasing the noise level or limiting the query frequency.

[0153] During the iterative optimization process, the system monitors the decision regret value to evaluate the optimization effect. When the decision regret drops below a preset threshold, the current strategy is considered to have reached a satisfactory balance point and can be put into use as the official strategy. The preset threshold is typically determined based on the specific needs of the application scenario, with lower thresholds set for scenarios with high security requirements and higher thresholds set for scenarios with high data availability requirements. The system also supports multi-objective optimization, allowing users to adjust the weighting of privacy protection and data availability based on specific needs, flexibly adapting to the needs of different application scenarios.

[0154] Step S7.3: Based on the data sensitivity level and query type, implement the Laplace noise addition mechanism to add random noise that meets the differential privacy requirements to the query results, ensuring that the privacy of individual data is not leaked while providing analysis results.

[0155] The core of the adaptive differential privacy parameter optimization algorithm is to minimize the loss in decision accuracy while ensuring privacy protection. Decision regret is defined as the difference in expected loss between the decision after adding privacy-preserving noise and the ideal (noise-free) decision. Through iterative learning, the algorithm dynamically adjusts the privacy budget allocation and noise parameters to keep decision regret within an acceptable range.

[0156] The Laplace noise addition mechanism in step S7.3 is a specific implementation of differential privacy. For query q, let its sensitivity be Δq and its privacy budget be ε. The probability density function of the Laplace noise added to the query result is: f(x) = (1 / (2b)) exp(-|x| / b), where b = Δq / ε. Sensitivity Δq represents the maximum magnitude of change in the query result that a change in an individual's data can cause. The privacy budget ε controls the strength of privacy protection. A smaller ε results in stronger privacy protection, but also reduces the accuracy of the query result.

[0157] In addition, in order to enhance the transparency and traceability of data privacy protection, the system has also built a blockchain-based privacy protection log system to record the complete process of data access and analysis operations, including access time, access subject, access data type, operation type and privacy protection measures, forming a privacy protection management system that supports audit traceability.

[0158] In step S7.3, the system implements a Laplace noise addition mechanism based on the data sensitivity level and query type, adding random noise that meets the differential privacy requirements to the query results, ensuring that the privacy of individual data is not leaked while providing analysis results. Differential privacy is a rigorous mathematical definition that ensures that the impact of adding or deleting any individual's data in the data set on the statistical query results is within a controllable range, thereby protecting individual privacy. The formal definition is: A random algorithm M satisfies ε-differential privacy if for any two adjacent data sets D1 and D2 (which only differ by one individual's data), and all possible outputs S of the algorithm M, there is: Pr[M(D1)∈S]≤e ε ×Pr[M(D)∈S]. The parameter ε is called the privacy budget, which controls the strength of privacy protection. A smaller ε means stronger privacy protection, but it also means greater noise introduced and lower data availability.

[0159] The Laplace noise addition mechanism is a classic approach to achieving differential privacy, particularly well-suited for protecting the privacy of numerical query results. The mechanism's basic principle is to add random noise that follows a Laplace distribution to the query results. Its probability density function is f(x) = (1 / 2b)exp(-|x| / b), where the parameter b = Δf / ε. Δf is the sensitivity of the query function f, representing the maximum magnitude by which a change in an individual's data can alter the query result, and ε is the privacy budget. By precisely controlling the noise distribution parameters, the Laplace mechanism ensures that the query results meet the ε-differential privacy requirements.

[0160] In the system implementation, the Laplace noise addition process is divided into three key steps: query sensitivity analysis, noise parameter calculation, and random noise generation. First, the system analyzes the sensitivity of each type of query function. For simple statistical queries (such as counting and summing), the sensitivity can usually be calculated directly; for complex queries, the system uses a combination of static analysis and dynamic monitoring to estimate the sensitivity. For example, for counting queries, the sensitivity is usually 1 (an individual affects the counting result by at most 1 unit); for sum queries, the sensitivity is equal to the possible maximum value (for example, in livestock breeding scenarios, it may be the maximum possible number of livestock on a single farm). The system maintains a sensitivity lookup table for common query types and provides a sensitivity analysis tool for custom queries.

[0161] Next, the system calculates the parameter b = Δf / ε of the Laplace distribution based on the query sensitivity and the currently allocated privacy budget ε. The value of ε here is determined by the aforementioned adaptive parameter optimization algorithm and is dynamically adjusted based on the query type, data sensitivity level, and historical query data. The system implements multiple budget allocation strategies, such as uniform allocation (all queries are allocated the same budget), weighted allocation (more budget allocated to important queries), and dynamic allocation (allocation is adjusted based on real-time demand). In practical applications, the system typically adopts a hybrid strategy, combining static rules and dynamic optimization to achieve more flexible budget management.

[0162] Finally, the system generates Laplace random noise that conforms to the computational parameters and adds it to the original query results. The system uses a high-quality random number generator to ensure that the randomness and distribution of the noise meet theoretical requirements. To handle special cases, the system also implements a variety of post-processing techniques, such as range restriction (ensuring that the result after adding noise remains within the valid range), integerization (rounding off queries that require integer results), and consistency adjustment (ensuring logical consistency across multiple related query results). These post-processing techniques improve the practicality of the results without compromising privacy protection.

[0163] The system implements differentiated protection strategies for data of different sensitivity levels and different types of queries. For highly sensitive data (such as disease records on specific farms), the system uses a smaller ε value, introduces greater noise, and provides stronger privacy protection; for less sensitive data (such as regional overall statistical data), a larger ε value is used to reduce the impact of noise and improve data availability. Similarly, for high-risk queries (such as precise queries on small-scale data), the system increases protection strength; for low-risk queries (such as large-scale aggregate queries), the protection strength is appropriately reduced. This differentiated strategy ensures that the system can achieve a reasonable balance between privacy protection and data availability in different scenarios.

[0164] The system also pays special attention to the privacy protection of continuous queries. In real applications, users may execute multiple related queries. Simply allocating the same privacy budget to each query can lead to rapid budget exhaustion or cumulative privacy leaks. To address this issue, the system implements query history analysis and dynamic budget adjustment mechanisms. The system records each user's query history, identifies related query sequences, and issues warnings or takes restrictive measures for query patterns that may lead to excessive privacy leaks. The system also supports smooth budget allocation, evenly distributing the total budget across the expected total number of queries, preventing premature budget exhaustion that could result in subsequent queries being unable to execute or insufficiently protecting the query.

[0165] To evaluate the effectiveness of privacy protection, the system implements comprehensive performance monitoring and security auditing capabilities. The system continuously monitors the noise level, data availability loss, and privacy budget consumption of various queries, generating regular reports to help managers understand the overall status of privacy protection. The system also simulates common privacy attack scenarios, such as repeated query attacks, differential attacks, and association inference attacks, to evaluate the effectiveness of current protection mechanisms and promptly identify and address potential security vulnerabilities. This closed-loop monitoring and optimization mechanism ensures that the system maintains a high level of privacy protection throughout its operation.

[0166] By integrating these technologies, the system has built a complete differential privacy protection framework, enabling the secure sharing and analysis of livestock farming data. This framework not only meets the legal and regulatory requirements for data privacy protection but also promotes the secure circulation and value mining of data, providing a solid data security foundation for the sustainable development of smart livestock farming. While safeguarding the commercial confidentiality and sensitive information of individual farms, the system can still provide accurate industry analysis and decision support, achieving an organic integration of privacy protection and data application.

[0167] Example 4

[0168] Based on the first embodiment, the implementation of the decision support system is further described. When constructing the decision support system, the method further includes:

[0169] Step S8.1: Abstract the entities and relationships in the aquaculture industry chain into nodes and edges of a generalized graph model, where nodes represent entities such as farms, slaughterhouses, and feed mills, and edges represent relationships such as supply relationships, capital flows, and logistics paths.

[0170] Step S8.2: Optimize the generalized graph model using a counterexample-guided abstraction refinement algorithm. By calculating the deviation between the inference results and the actual observed data, the graph regions that need to be refined are determined, and the abstraction levels of nodes and edges are dynamically adjusted.

[0171] In step S8.2, the system optimizes the generalized graph model using the Counterexample-Guided Abstraction Refinement (CEGAR) algorithm. By calculating the deviation between the inference results and the actual observed data, it determines the graph regions that require refinement and dynamically adjusts the abstraction level of nodes and edges. The CEGAR algorithm is an important method in the field of formal verification. It gradually improves the accuracy and efficiency of the model through an iterative abstraction-verification-refinement process. In this system, this algorithm is innovatively applied to the optimization of the livestock breeding knowledge graph, effectively solving the problem of balancing the accuracy of knowledge representation with computational complexity.

[0172] The implementation of the CEGAR algorithm in this system follows a four-step iterative cycle: initial abstraction, model verification, counterexample analysis, and abstract refinement. During the initial abstraction phase, the system simplifies the original livestock breeding knowledge graph to construct a coarse-grained abstract model. This abstraction is primarily manifested in three aspects: node merging (merging semantically similar nodes into a single abstract node, such as combining different pig breeds into a "pig" node); attribute simplification (reducing multidimensional node attributes to key attributes, such as simplifying weight, age, and health status into "growth status"); and relationship abstraction (abstracting multiple fine-grained relationships into a small number of core relationship types, such as abstracting behavioral relationships such as "feeding," "drinking," and "exercise" into "daily activities"). This initial abstraction significantly reduces the complexity of the graph model, making subsequent reasoning and computation efficient and feasible.

[0173] During the model verification phase, the system uses the abstracted graphical model to perform reasoning tasks and generate inference results. These reasoning tasks cover a variety of categories, such as disease diagnosis (inferring possible diseases based on symptoms), growth prediction (predicting future growth trends based on current status), and feed formulation optimization (recommending formulas based on nutritional requirements and ingredient characteristics). For each type of reasoning task, the system predefines formal rules, such as "If body temperature is elevated, appetite is decreased, and breathing is abnormal, then respiratory disease is likely." The system performs reasoning based on these rules and the abstracted graphical model, generating preliminary results.

[0174] Subsequently, the system enters the counterexample analysis phase, comparing the inference results with the actual observation data, and the identified deviations are considered counterexamples. The system collects two types of counterexamples: false positive counterexamples (the inference result indicates that a certain situation exists, but the actual observation data shows that it does not exist) and false negative counterexamples (the inference result fails to identify the actual situation). The system analyzes each counterexample in detail, traces the reasoning path that leads to the deviation, and locates possible abstract defects. For example, if the system infers that a farm is at risk of disease based on an abstract model, but actual observations show that the farm is in good health, the system will analyze the path that led to this incorrect judgment and may find that the loss of accuracy is caused by the unified abstraction of farms with different biosafety levels.

[0175] Based on the results of the counterexample analysis, the system enters the abstract refinement phase, performing targeted refinement on identified problem areas. Refinement operations primarily include three types: node decomposition (breaking abstract nodes into more fine-grained, concrete nodes), attribute expansion (adding more attribute dimensions to nodes), and relationship refinement (refining abstract relationships into multiple concrete relationship types). The system employs an information gain-based refinement strategy, prioritizing elements that minimize inference bias. For example, if analysis reveals that errors in disease risk assessment are primarily caused by the abstract representation of "biosafety measures," the system will prioritize refining this attribute from a simple "high / medium / low" grading system to multi-dimensional, specific indicators, such as "disinfection frequency," "strictness of personnel control," and "completeness of quarantine and isolation facilities."

[0176] A key innovation of the CEGAR algorithm is its adaptive abstraction granularity control mechanism. The system dynamically adjusts the abstraction granularity based on the error rate and inference importance of different regions. A fine-grained representation is used for critical regions with high inference error rates, while a coarse-grained representation is maintained for less critical regions with accurate inference. This differentiation strategy significantly improves the overall efficiency and accuracy of the system. For example, in disease risk assessment, the system uses a fine-grained representation for farms in high-risk areas, modeling their internal structure, management measures, and external connections in detail; while a coarse-grained representation is used for farms in low-risk areas, recording only basic features.

[0177] Another innovation is the system's multi-level abstraction library. The system predefines multiple levels of abstraction for each element (node, attribute, and relationship) in the graph model, from most specific to most abstract. During refinement, the system can flexibly switch between these predefined levels as needed without rebuilding the abstract representation. For example, for the "feed" node, the system predefines four levels of abstraction: the most specific level (specific brands and batches of feed products), the medium-specific level (feed types such as piglet feed and fattening feed), the general abstract level (broad feed categories such as compound feed and concentrated feed), and the most abstract level (the simple concept of "feed"). This predefined multi-level abstraction significantly improves the efficiency and consistency of refinement operations.

[0178] The system also implements an automatic termination mechanism for abstract refinement. The algorithm monitors the performance improvement after several rounds of refinement and automatically terminates when the performance improvement falls below a preset threshold or when the maximum number of iterations is reached. Furthermore, the system supports resource budget-based termination control. When computing resources are limited, the system prioritizes refinement of the most critical areas, ensuring optimal inference performance within resource constraints.

[0179] Through multiple rounds of CEGAR iterations, the system ultimately achieves an optimized graphical model that balances accuracy and efficiency, providing a solid foundation for the next step of continuous logical reasoning. This optimization is not a one-time effort but an ongoing process. As new data accumulates and new counterexamples emerge, the system regularly executes CEGAR cycles, continuously improving the graphical model's expressive power and reasoning accuracy.

[0180] Step S8.3: Based on the optimized graphical model, a finite field continuous logic reasoning engine is constructed to map the continuous variables and fuzzy concepts in aquaculture management to the truth space of the interval [0,1], and use fuzzy reasoning rules to generate optimized decision information for aquaculture management.

[0181] In step S8.3, the system constructs a finite domain continuous logic inference engine based on the optimized graphical model. This engine maps continuous variables and fuzzy concepts in livestock management to a truth space in the interval [0, 1] and uses fuzzy inference rules to generate optimized decision-making information for livestock management. Traditional Boolean logic reasoning faces serious limitations when dealing with complex problems in the livestock farming sector. This is because livestock management involves a large number of continuous variables (such as temperature, humidity, and weight growth rate) and fuzzy concepts (such as "suitability," "health," and "high yield"), which are difficult to accurately express using simple "yes / no" binary logic. By introducing a truth space in the interval [0, 1], the finite domain continuous logic inference engine enables precise modeling and reasoning of these continuous variables and fuzzy concepts, significantly improving the accuracy and practicality of the inference results.

[0182] The core of the finite field continuous logic inference engine is the truth-value mapping mechanism. The system defines a mapping function for each continuous variable in the graphical model, mapping its original value range to the standard interval [0, 1]. These mapping functions typically adopt an S-shaped or bell-shaped curve, reflecting the nonlinear relationship between the variable value and the degree of association between the specific concept. For example, for the concept of "suitable growth temperature," the system might define a bell-shaped mapping function that takes a value of 1 at the optimal temperature (e.g., 22°C) and gradually decreases as the temperature deviates from the optimal value. For "weight growth meeting the target," the system might define an S-shaped mapping function that approaches 1 when the growth rate exceeds a certain threshold and approaches 0 when it falls below the threshold. These mapping functions are not fixed; instead, their parameters can be adjusted according to the specific needs of different varieties and growth stages, ensuring the flexibility and adaptability of the model.

[0183] To express fuzzy concepts, the system uses fuzzy set theory. Each fuzzy concept (such as "good health") is defined as a fuzzy set, which is jointly determined by the membership functions of multiple related indicators. For example, "good health" may be composed of multiple indicators such as body temperature, appetite, activity level, growth rate, etc., each of which has its own weight and membership function. The system calculates the overall degree of membership of an entity to a fuzzy concept by weighted aggregation of the membership of these indicators. To enhance the interpretability of the model, the system also implements a natural language generation function that can convert numerical membership into intuitive natural language descriptions, such as "excellent health (0.92)" or "growth rate slightly below standard (0.65)".

[0184] In terms of continuous logic operations, the system implements a complete set of continuous logic operators. Unlike traditional Boolean logic, continuous logic operations such as "AND", "OR", and "NOT" accept inputs in the [0,1] interval and produce outputs in the [0,1] interval. The system supports a variety of continuous logic operator families, such as logic, Logic and Product logic, users can choose the appropriate operator family according to the specific application scenario. For example, in risk assessment, Product logic is usually used because it is more sensitive to the comprehensive evaluation of multiple risk factors; while in suitability assessment, The system also supports the definition of compound operators, allowing users to create specific operation rules based on domain knowledge, further improving the expressiveness of the model.

[0185] Based on these foundational mechanisms, the system has constructed a comprehensive library of continuous logic reasoning rules. These rules employ an "if-then" structure, but both the condition and conclusion are continuous values, such as "If the temperature suitability is x AND the humidity suitability is y, then the environmental comfort is f(x,y)." Through text mining and expert knowledge extraction, the system has collected and formalized a large number of reasoning rules from livestock husbandry literature and expert experience, covering various aspects such as animal husbandry management, disease prevention and control, breeding, and environmental control. These rules are organized into a hierarchical structure, ranging from general principles to specific guidance, supporting multi-level reasoning processes.

[0186] Continuous logical reasoning utilizes a forward-chaining algorithm. Starting from known facts, the system infers new facts by continuously applying applicable rules until no further conclusions can be drawn or the target state is reached. Unlike traditional forward-chaining, each step in continuous logical reasoning calculates and propagates a continuous truth value. The final conclusion includes not only the derivation result but also its confidence. The system also implements a hybrid reasoning mechanism, combining data-driven machine learning methods with rule-driven logical reasoning. In areas where the rule base is insufficiently covered, machine learning models supplement reasoning capabilities, achieving more comprehensive knowledge coverage.

[0187] In the application of reasoning results, the system has designed a decision support generation module to convert the results of continuous logical reasoning into practical and feasible management suggestions. This module first conducts a comprehensive analysis of the reasoning results to identify key issues and room for improvement. Then, based on the built-in policy template library and combined with specific situations, it generates personalized decision-making suggestions. These suggestions are not simple fixed texts, but are dynamically generated based on the specific numerical values ​​of the reasoning results, which can reflect the severity and urgency of the problem. For example, for feed formula optimization, the system may generate specific and feasible suggestions such as "The current protein content is slightly lower than the optimal level (0.78), and it is recommended to increase the proportion of soybean meal by 3-5%" based on the reasoning results; for disease risk warnings, the system may generate targeted guidance such as "The risk of respiratory disease is moderate (0.62), and it is recommended to strengthen ventilation and observe daily. If symptoms worsen, isolate and treat immediately."

[0188] The system pays special attention to expressing uncertainty in decision recommendations. Traditional decision-making systems often oversimplify complex problems, providing deterministic recommendations that ignore the uncertainties and risks of the real world. This system, based on the characteristics of continuous logical reasoning, can naturally express the uncertainty of conclusions, helping users to more comprehensively understand the risks of decision-making. For example, the system will not only provide a decision such as "Recommend changing feed type," but also provide supplementary information such as "This recommendation has a confidence level of 0.85, a success probability of approximately 78%, and the main uncertainty factor comes from individual differences in response." This allows users to make decisions based on a full understanding of the risks.

[0189] To support multi-objective decision-making optimization, the system implements a Pareto-front-based decision-making scheme generation mechanism. In livestock management, it is often necessary to balance multiple objectives, such as production performance, cost control, animal welfare, and environmental impact. The system uses continuous logical reasoning to evaluate the performance of different schemes on each objective, identify the Pareto-optimal set of solutions (i.e., a set of solutions that cannot simultaneously improve all other objectives without sacrificing a certain objective), and provide a recommended ranking based on user preferences. This multi-objective optimization method avoids the excessive focus on a single objective in traditional decision-making systems and provides a more balanced and sustainable decision-making solution.

[0190] The system also features an interactive reasoning explanation feature to help users understand the reasoning process and the reasons for reaching conclusions. Users can view the complete reasoning path, including the rules used at each step, the truth values ​​of the inputs, and the resulting conclusions. The system uses visualization techniques to display the reasoning network, clearly illustrating the relationships between various factors and their contributions to the final conclusion. Furthermore, the system provides a "what-if" analysis function, allowing users to modify input conditions or rule parameters to observe changes in the conclusions, assisting users in hypothesis testing and sensitivity analysis. These features significantly enhance the system's transparency and credibility, enabling users to critically evaluate reasoning results and avoid blindly adopting them.

[0191] By applying a finite domain continuous logic inference engine, the system successfully implements intelligent decision support for livestock farming management, transforming abstract knowledge graphs into concrete and actionable management recommendations. This inference engine not only handles continuous variables and fuzzy concepts in livestock management, but also expresses and handles the uncertainty and diversity of knowledge, providing more accurate, comprehensive, and practical decision support for complex livestock management problems. Compared with traditional expert systems, it significantly improves on handling complexity, adaptability, and transparency, representing the latest developments in intelligent decision-making technology for livestock farming.

[0192] By combining a counterexample-guided abstract refinement algorithm with a finite field continuous logic inference engine, the system constructs a highly efficient and accurate livestock farming knowledge graph reasoning framework. This framework effectively leverages existing knowledge to address complex livestock management issues, providing users with scientific and reliable decision support and significantly improving the intelligence level and production efficiency of livestock farming management.

[0193] In summary, the generalized graph model is the core data structure of the decision support system. It abstracts the various components of the livestock and poultry industry chain into nodes and edges. Nodes represent entities, such as farms, slaughterhouses, feed mills, logistics companies, and sales terminals; edges represent relationships between entities, such as supply relationships, capital flows, and logistics routes. Each node and edge has attributes. Node attributes include information such as entity type, size, and location, while edge attributes include information such as relationship type, strength, and direction.

[0194] The generalized graph model employs a multi-level structure, dividing the industry chain into three levels: micro, meso, and macro. The micro level focuses on the management and production activities within a single farm; the meso level examines the connections and collaboration between farms within a specific region; and the macro level examines the industry landscape and development trends for an entire region or industry. This multi-level structure enables analysis of industry chain issues from different perspectives and granularities.

[0195] The counterexample-guided abstract refinement algorithm is a method for dynamically optimizing generalized graph models. It first creates an initial abstract representation of the farm association network, aggregating entities and relationships with similar characteristics to form a highly abstract graph model. It then performs influence propagation analysis on the abstract graph, comparing the results with actual observations to extract key deviations as counterexamples. Based on these counterexamples, a hierarchical decomposition strategy is employed to prioritize graph regions associated with the counterexamples, decomposing them into finer-grained entities and recalculating their relationship attributes. Finally, the algorithm evaluates the information gain of each refinement operation, stopping refinement when the information gain falls below a preset threshold. This results in an optimized graph model that balances analytical accuracy and computational efficiency.

[0196] The Finite Field Continuous Logic Inference Engine is a reasoning system based on fuzzy logic theory that can handle uncertainty and continuous variables in aquaculture management. This engine maps truth values ​​to the interval [0,1] and uses membership functions to express fuzzy concepts such as "high yield" and "good health." The inference engine integrates a probabilistic reasoning module that can quantitatively express uncertainty. Combining forward reasoning (from facts to conclusions) and backward reasoning (from goals to conditions) strategies, it supports analysis and decision-making in complex scenarios.

[0197] Based on a generalized graph model and a continuous logic reasoning engine, the decision support system implements structured continuous optimization capabilities: first, a structured representation of the optimization problem is constructed based on the generalized graph model, mapping variables, objective functions, and constraints to nodes and edges of the graph; then, a continuous optimization algorithm library is constructed, integrating algorithms such as gradient descent, quadratic programming, and interior point method, and automatically selecting a suitable algorithm combination based on the characteristics of the problem; for large-scale optimization problems, a community discovery algorithm is used to decompose the graph, identify weakly connected subgraphs to construct independent optimization subproblems, and coordinate the solutions of the subproblems through boundary conditions to achieve parallel solution; at the same time, adaptive step size control, multi-starting point search, and convergence monitoring are implemented to ensure the stability and reliability of the algorithm; finally, based on the performance data of the optimization algorithm in actual applications, a performance evaluation index system is constructed, and the algorithm configuration parameters are dynamically adjusted to form a closed-loop system for adaptive optimization.

[0198] Example 5

[0199] See also Figure 6The present invention also provides a GIS-based intelligent livestock breeding full-chain visualization system, including:

[0200] The data collection module is used to collect basic farm information, livestock information, production process data, transaction data and regulatory data through a standardized data interface, and clean, convert and standardize the collected data to form a unified data resource pool;

[0201] A data association module is used to associate and map the business data in the data resource pool with the GIS spatial data based on the geographic coordinate system, and to construct a three-dimensional data model containing spatial information, attribute information and relationship information;

[0202] A visualization engine module is used to design a visualization engine with three levels, namely macro, meso, and micro, using the three-dimensional data model and WebGL technology, and to achieve multi-scale visualization from regional overview to individual farm details through a layered design;

[0203] The traceability visualization module is used to build a data model of the entire life cycle of livestock products from breeding and farming to slaughter and sales based on the multi-scale visualization display, and realize the visualization display of the entire chain traceability of livestock products through the combination of dynamic timeline and flow chart;

[0204] The decision support module is used to build a decision support system including industrial layout analysis, breeding scale forecast, disease risk assessment and market supply and demand analysis based on the above-mentioned visual display and full-chain traceability data, and generate a decision support report.

[0205] The modules interact with each other through a data bus to achieve unified management and sharing of data. The system architecture adopts a front-end and back-end separation design. The front-end implements the user interface and interactive functions based on the JavaScript framework, and the back-end implements business logic and data processing functions based on the microservice architecture. The RESTful API interface specification is used to achieve front-end and back-end communication.

[0206] The data acquisition module includes a data access submodule, a data processing submodule and a data storage submodule: the data access submodule is responsible for connecting to various data sources and supports multiple data acquisition methods such as REST API, direct database connection, IoT devices and file import; the data processing submodule is responsible for data cleaning, conversion and normalization processing to achieve data integrity, consistency and timeliness checks; the data storage submodule is responsible for data storage and management, adopts a distributed database architecture, integrates relational databases, time series databases and spatial databases, and realizes unified management of structured, semi-structured and unstructured data.

[0207] The data association module includes a spatial basic data processing submodule, a geocoding submodule and a spatial association submodule: the spatial basic data processing submodule is responsible for processing basic geographic information data such as administrative boundaries, road networks, water systems and terrain, and building a standardized spatial reference system; the geocoding submodule is responsible for locating the address information and GPS coordinates of the farm to the spatial reference system and establishing the spatial index of the farm; the spatial association submodule is responsible for associating the farm spatial index with business data, constructing a spatial attribute association table, and using multi-level level of orientation (LOD) technology to process data expressions at different scales.

[0208] The visualization engine module includes a macro-visualization submodule, a meso-visualization submodule and a micro-visualization submodule: the macro-visualization submodule is responsible for generating heat maps and bar charts of regional livestock breeding resource distribution, displaying macro indicators such as the total number of farms, total livestock and breed distribution; the meso-visualization submodule is responsible for generating three-dimensional scenes of farm groups in a specific area, displaying the spatial distribution, scale structure and production capacity of the farms; the micro-visualization submodule is responsible for generating three-dimensional visualization inside a single farm, displaying the building layout, livestock distribution and equipment configuration, and supporting interactive query of equipment status, environmental parameters and livestock information.

[0209] The traceability visualization module includes a traceability data model submodule, a timeline display submodule, a spatial trajectory submodule and a relationship network submodule: the traceability data model submodule is responsible for designing a traceability data model covering the reproduction, breeding, slaughtering, processing and sales links, and defining the key information points and data association relationships of each link; the timeline display submodule is responsible for realizing the temporal browsing of traceability data, and displaying the time nodes and processes of products from breeding to sales; the spatial trajectory submodule is responsible for generating spatial flow visualization of products, and displaying the spatial movement path of products from farms to slaughterhouses, processing plants, and sales terminals; the relationship network submodule is responsible for displaying the association relationship between products and related entities through a force-directed graph, and displaying the responsible entities and handover relationships in the product flow process.

[0210] The decision support module includes an industrial layout analysis submodule, a breeding scale prediction submodule, a disease risk assessment submodule, a market supply and demand analysis submodule and a visualization report submodule: the industrial layout analysis submodule is responsible for analyzing industrial agglomeration areas and development potential areas based on farm distribution and scale data, combined with land use, environmental capacity and transportation conditions; the breeding scale prediction submodule is responsible for predicting future breeding volume, slaughter volume and market supply based on historical data and growth trends using a time series analysis algorithm; the disease risk assessment submodule is responsible for calculating the probability of disease transmission through network analysis methods based on farm distribution, livestock density and historical disease records, and identifying high-risk areas; the market supply and demand analysis submodule is responsible for integrating breeding production data, slaughtering and processing data and market sales data, and analyzing market supply and demand relationships and price change trends; the visualization report submodule is responsible for integrating the analysis results into a decision-making report, which includes data analysis charts, evaluation conclusions and recommended measures.

[0211] In addition, the system also includes a privacy protection module and a system management module: the privacy protection module is responsible for implementing data differential privacy protection, including data sensitivity classification, privacy parameter optimization and privacy protection mechanism implementation; the system management module is responsible for user authority management, system configuration management and log audit management to ensure the safe and stable operation of the system.

[0212] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program executes the steps of the GIS-based intelligent livestock farming full-chain visualization method described in the above-mentioned method embodiment. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0213] In addition, an embodiment of the present disclosure also provides a computer program product, which stores a computer program. When the computer program is run by a processor, it executes the steps of a GIS-based smart livestock breeding full-chain visualization method provided by any of the above embodiments of the present disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0214] The computer program product may be implemented in hardware, software, or a combination thereof. In one embodiment, the computer program product is implemented as a computer storage medium, which may be a volatile or non-volatile computer-readable storage medium. In another embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0215] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment and devices can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0216] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0217] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0218] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0219] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The scope of protection of the present disclosure is not limited thereto. Although the present disclosure has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-mentioned embodiments within the technical scope disclosed in the present disclosure, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims.

Claims

1. A GIS-based visualization method for the entire chain of intelligent animal husbandry, characterized in that: include: Collect basic farm information, livestock information, production process data, transaction data, and regulatory data through standardized data interfaces, and clean, convert, and normalize the collected data to form a unified data resource pool; Based on the geographic coordinate system, the business data in the data resource pool is associated and mapped with the GIS spatial data to construct a three-dimensional data model containing spatial information, attribute information and relationship information; Using the three-dimensional data model and WebGL technology, a visualization engine with three levels, macro, meso, and micro, was designed. Through layered design, multi-scale visualization from regional overview to individual farm details was achieved. Based on the multi-scale visualization, a data model covering the entire life cycle of livestock products, from breeding and farming to slaughter and sales, is constructed. By combining a dynamic timeline with a flow chart, a visualization of the entire chain of livestock product traceability is achieved. Based on the above-mentioned visual display and full-chain traceability data, a decision support system is constructed that includes industrial layout analysis, breeding scale forecast, disease risk assessment and market supply and demand analysis to generate a decision support report.

2. The method according to claim 1, characterized in that Collect basic farm information, livestock information, production process data, transaction data, and regulatory data through standardized data interfaces, cleanse, convert, and normalize the collected data to form a unified data resource pool, including: Check the integrity, consistency and timeliness of the collected farm basic information, livestock information, production process data, transaction data and regulatory data, identify and mark abnormal data according to preset data verification rules, perform data error correction and compensation, and output data stream; A distributed database is used to store the data stream, and a hybrid storage system including a relational database, a time series database and a spatial database is established to achieve unified management of structured, semi-structured and unstructured data in the data stream.

3. The method according to claim 1, characterized in that Based on the geographic coordinate system, the business data is associated and mapped with the GIS spatial data to construct a three-dimensional data model containing spatial information, attribute information, and relationship information, including: Acquire and process basic geographic information data including administrative boundaries, road networks, water systems, and topography, and build a standardized spatial reference system; Based on the address information and GPS coordinates of the farm, the farm location is located in the spatial reference system through geocoding to establish a spatial index of the farm; The spatial index of the farm is associated with business data including livestock quantity, breed structure and production status, a spatial attribute association table is constructed, and multi-level LOD technology is used to process data expressions at different scales.

4. The method according to claim 1, wherein Using the three-dimensional data model and WebGL technology, a visualization engine with three levels of macro, meso, and micro is designed, including: Generate heat maps and bar charts of regional livestock breeding resource distribution based on regional administrative boundaries and statistical data, displaying macro indicators including the total number of farms, total livestock population, and breed distribution; For farm groups in a specific area, custom icons, color coding, and shape transformations are used to generate a 3D scene showing the spatial distribution, scale structure, and production capacity of the farms. For a single farm, a three-dimensional visualization interface is generated for building layout, livestock distribution, and equipment configuration, which can interactively display equipment operating status, environmental parameters, and individual livestock information through clicks.

5. The method according to claim 1, wherein Build a data model for the entire life cycle of livestock products, from breeding and farming to slaughter and sales, including: Design a traceability data model covering the breeding, farming, slaughtering, processing and sales links, and define the data structure of key information points and data association relationships in each link; Based on the data structure and GIS map, a dynamic path map of the spatial movement trajectory of the product from the farm to the slaughterhouse, processing plant, and sales terminal is generated; A force-directed graph algorithm is used to calculate and display the relationship between livestock products and related entities, and a visual view of the responsible entities and handover relationships in the product flow process is generated through nodes and lines.

6. The method according to claim 1, characterized in that The method further comprises: Based on farm distribution and livestock density data, spatial statistical methods are used to calculate the number of susceptible animals and contact risk coefficients per unit area, and a disease risk assessment model is constructed; Based on the risk assessment model and combined with historical disease records, network analysis methods are used to calculate the probability of disease transmission and identify high-risk areas; Perform spatial cluster analysis on the high-risk areas to generate a disease risk distribution map, and generate early warning information for areas where the risk index exceeds a preset threshold.

7. The method according to claim 6, characterized in that The network analysis method is used to calculate the probability of disease spread and identify high-risk areas, including: Collect records of personnel flow, vehicle transportation, material exchange, and livestock transactions between farms to build a farm connection network; The farm association network is converted into an adjacency matrix, and the fourth power of the adjacency matrix is ​​calculated using the Strassen algorithm to calculate the number of four-loops involved by each farm. The four-loop density is defined as the ratio of the number of four-loops involved by a farm to the number of its directly connected neighboring farms. Based on the density of the four ring roads, the DBSCAN algorithm was used to spatially cluster the farms, and areas with a density greater than a preset threshold in the clustering results were marked as transmission risk hotspots.

8. The method according to claim 1, characterized in that After collecting basic farm information, livestock information, production process data, transaction data, and regulatory data through a standardized data interface, the method further includes: Perform differential privacy protection on the collected data, including: Categorize livestock breeding data into four levels based on their sensitivity: public data, internal data, sensitive data, and highly sensitive data, and establish access rights and protection strategies for data at different levels; Construct a linear Markov decision process model and design an adaptive differential privacy parameter optimization algorithm. The decision regret is defined as the difference between the expected loss of the actual decision and the ideal decision. Through iterative optimization, the decision regret is controlled within a preset range. According to the data sensitivity level and query type, a Laplace noise addition mechanism is implemented to add random noise that meets the differential privacy requirements to the query results, ensuring that the privacy of individual data is not leaked while providing analysis results.

9. The method according to claim 1, characterized in that When constructing the decision support system, the method further includes: The entities and relationships in the aquaculture industry chain are abstracted into nodes and edges of a generalized graph model, where nodes represent entities including farms, slaughterhouses, and feed mills, and edges represent relationships including supply relationships, capital flows, and logistics paths. The generalized graph model is optimized using a counterexample-guided abstraction refinement algorithm. By calculating the deviation between the inference results and the actual observation data, the graph regions that need to be refined are determined, and the abstraction level of nodes and edges is dynamically adjusted. Based on the optimized graphical model, a finite domain continuous logic reasoning engine is constructed to map continuous variables and fuzzy concepts in aquaculture management to a truth space in the interval [0,1], and fuzzy reasoning rules are used to generate optimized decision information for aquaculture management.

10. A GIS-based intelligent livestock farming full-chain visualization system, characterized by: include: The data collection module is used to collect basic farm information, livestock information, production process data, transaction data and regulatory data through a standardized data interface, and clean, convert and standardize the collected data to form a unified data resource pool; A data association module is used to associate and map the business data in the data resource pool with the GIS spatial data based on the geographic coordinate system, and to construct a three-dimensional data model containing spatial information, attribute information and relationship information; A visualization engine module is used to design a visualization engine with three levels, namely macro, meso, and micro, using the three-dimensional data model and WebGL technology, and to achieve multi-scale visualization from regional overview to individual farm details through a layered design; The traceability visualization module is used to build a data model of the entire life cycle of livestock products from breeding and farming to slaughter and sales based on the multi-scale visualization display, and realize the visualization display of the entire chain traceability of livestock products through the combination of dynamic timeline and flow chart; The decision support module is used to build a decision support system including industrial layout analysis, breeding scale forecast, disease risk assessment and market supply and demand analysis based on the above-mentioned visual display and full-chain traceability data, and generate a decision support report.

Images