Methods, devices, equipment, and storage media for agricultural knowledge graph construction and reasoning based on a spatiotemporal intelligent foundation.

Abstract

This application provides a method, apparatus, device, and storage medium for constructing and reasoning agricultural knowledge graphs based on a spatiotemporal intelligent foundation, relating to the fields of geographic information and remote sensing technologies. The method includes: constructing a five-layer partitioned structure of wide-area space, regional space, local space, object space, and signal space; constructing a five-dimensional vector data structure of location, morphology, attributes, structure, and status for each geographic object entity; constructing an agricultural knowledge graph with geographic object entities as nodes and similarity distance as edges; and reasoning about planting patterns by combining information transmission and knowledge constraints. When reasoning about the suitability of target crop planting, environmental features of already planted plots are extracted, typical niche centers are learned through clustering, the similarity score between the plot to be evaluated and the niche center is calculated, and environmental semantic weights are introduced for correction to generate a suitability score and level. This application solves the problems of existing technologies' inability to express spatiotemporal dynamic processes and weak reasoning capabilities.

Description

Technical Field

[0001] This application relates to the fields of geographic information and remote sensing technology, and in particular to a method, apparatus, equipment and storage medium for constructing and reasoning agricultural knowledge graphs based on a spatiotemporal intelligent foundation. Background Technology

[0002] Currently, Geographic Information Systems (GIS) and remote sensing technologies are widely used in various fields such as agricultural monitoring, urban planning, and disaster emergency response. With the continuous improvement of remote sensing data acquisition capabilities, how to extract effective knowledge from massive amounts of spatiotemporal data has become a key problem urgently needing to be solved in this field. Knowledge graphs, as a means of structured knowledge organization, have been introduced into the fields of remote sensing and geographic information in recent years to describe the semantic relationships between ground features.

[0003] However, existing remote sensing knowledge graph technologies have the following main shortcomings: First, existing technologies primarily rely on static maps, lacking the ability to effectively represent dynamic spatiotemporal processes. Most remote sensing knowledge maps remain at the level of describing the static attributes of ground features, recording only static information such as the type and distribution range of ground features at a specific moment or period. For geographical phenomena with significant spatiotemporal dynamic characteristics, such as the evolution of urban expansion, the phenological changes in vegetation, and the movement paths of typhoons, existing maps struggle to express them in a concise and standardized manner. This static representation severs the continuity of geographical entities in the temporal dimension, failing to fully depict the evolutionary process of geographical objects.

[0004] Second, existing technologies have weak spatiotemporal reasoning capabilities. Most current remote sensing knowledge graph reasoning mechanisms are based on symbolic logic and pre-defined rules, such as inferring inclusion and adjacency relationships between features through first-order predicate logic or ontology reasoning. While these methods are adequate for handling simple static spatial relationships, they struggle to effectively address complex spatiotemporal relationships, such as land use change trends in a specific region over a given time period or predicting crop growth status in the next stage based on historical phenological patterns. Furthermore, existing methods lack the ability to autonomously learn and predict spatiotemporal patterns, making it difficult to support high-level spatiotemporal analysis applications that extrapolate future states based on historical patterns.

[0005] The aforementioned shortcomings make it difficult for existing remote sensing knowledge graph technologies to meet the demands of precise and intelligent applications when dealing with complex surface systems and dynamically changing environments. Particularly in agriculture, crop growth is influenced by multiple spatiotemporal factors such as policy, soil, climate, and hydrology, necessitating a novel knowledge graph construction and reasoning method that can effectively integrate multi-source spatiotemporal data, express dynamic processes, and support spatiotemporal reasoning. Summary of the Invention

[0006] This application provides a method, apparatus, device, and storage medium for constructing and reasoning agricultural knowledge graphs based on a spatiotemporal intelligent foundation. It aims to address two major shortcomings of existing remote sensing knowledge graph technologies: static representation makes it difficult to describe dynamic spatiotemporal processes, and symbolic logic reasoning mechanisms struggle to handle complex spatiotemporal relationships and spatiotemporal pattern-based predictions. By constructing a five-layer geospatial partitioning structure, it achieves refined deconstruction of complex terrain; by designing a five-dimensional vector data structure, it endows geographic entities with multi-dimensional expressive capabilities of location, morphology, attributes, structure, and status, enabling them to effectively carry spatiotemporal dynamic information; by constructing a crop knowledge graph, it establishes semantic associations and similarity measurement mechanisms between geographic entities; and by introducing a dual reasoning mechanism of information transmission and knowledge constraints, it achieves intelligent analysis and prediction of crop planting patterns, thereby overcoming the shortcomings of existing technologies in expressing spatiotemporal processes and performing spatiotemporal reasoning.

[0007] In a first aspect, this application provides a method for constructing and reasoning about agricultural knowledge graphs based on a spatiotemporal intelligent foundation, the method comprising: A five-layer zoning structure for geographic entities is constructed. Based on high-resolution remote sensing imagery as the base data, the land surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, in the hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions. Each geographic object entity is defined as a geographic patch container, and a vector is constructed for each geographic object entity to obtain a five-dimensional vector data structure of the geographic object entity. The vector is encapsulated by a set of elements in a linear order, with each element corresponding to its index one-to-one, and supports dynamic adjustment of capacity. The vector includes location vector, morphology vector, attribute vector, structure vector and situation vector. By using IoT sensors, cameras, lidar, and satellite remote sensing technologies, the continuous, simulated, and complex physical phenomena of the physical world are transformed into discrete, digital, and structured data. Using the geographic object entities as nodes, the distance between two nodes represents the similarity between two geographic entity objects, thus constructing a knowledge graph for the planting industry. Agricultural reasoning based on information transmission and knowledge constraints utilizes the aforementioned crop knowledge graph, combined with information transmission relationships and knowledge constraint relationships, to analyze and predict crop planting patterns.

[0008] In one possible design, the wide-area space is an administrative division space, including provincial and municipal division codes, provincial and municipal names, town and township division codes, town and township names, and area size data; The aforementioned regional space is a functional zoning space, including administrative division codes and regional division codes. The regional division codes include codes for cultivated areas, urban areas, and ecological baselines. The local space is a functionally subdivided space, including local feature codes. In agricultural planting areas, the local feature codes include codes for cultivated land, orchards, and forest land. In urban living and production areas, the local feature codes include codes for industrial land, residential land, educational land, sports land, commercial service land, logistics and warehousing land, medical and health land, cultural facilities land, parks and green spaces, transportation land, and water and water conservancy facilities land. The object space is a geographic entity patch space, including agricultural land patches, ecological forest and grassland patches, and building patches; The signal space is a multi-source sensing data space, including spectral signals, spatiotemporal variation signals, and radiometric measurement signal data acquired through satellite remote sensing, image signals, visible light signals, and oblique photography signal data acquired through drones, as well as real-time variation signals, historical survey signals, and social resource signal data acquired through high-definition cameras and on-site sampling.

[0009] In one possible design, the position vector is used to store the orientation information of an object in three-dimensional space; the morphology vector is used to store the morphological feature information of points, lines, surfaces, and volumes; the attribute vector is used to store the parameter system information of the geographic knowledge graph; the structure vector is used to store the geographic relationship information between objects, between objects and the environment, and between objects; and the situation vector is used to store the evolution process and situation change information of the object entity.

[0010] In one possible design, the information transmission relationship includes: the probability of planting the same crop on plots with similar environmental characteristics is higher than a preset threshold; The knowledge constraints include: the guiding constraints of agricultural policies on crop planting, the guiding constraints of historical statistical data on real-time data, the constraints of climatic conditions in the space where agricultural plots are located on local space, the constraints of seasons on crop planting, and the constraints of farmers' planting experience.

[0011] In one possible design, the analysis and prediction of crop planting patterns includes target crop planting suitability reasoning, which comprises the following steps: A knowledge graph RDF is constructed, and an environmental feature vector is constructed for each plot. The environmental feature vector includes at least one of the following characteristics: rock exposure, vegetation cover, slope, annual average soil moisture, soil moisture during the growing season, minimum soil moisture, and soil surface temperature. Environmental features of plots planted with target crops are extracted from RDF, and typical niche centers of target crops are learned through clustering algorithms. For each site to be evaluated, the similarity between its environmental feature vector and the typical niche center is calculated to obtain a niche suitability score. Environmental semantic weights are introduced to correct the niche suitability score, generating a target crop planting suitability score and grade.

[0012] In one possible design, learning the typical niche centers of the target crop through a clustering algorithm includes: Select plots of land known to be planted with the target crop as learning samples. ,in , and These are the first, second, and nth features, respectively; The features are standardized using the following formula:

[0013] in, For the first i The mean of each feature, For the first i The standard deviation of each feature For the first i One characteristic, For the first i One standardized feature; Standardized parameters are permanently saved to ensure consistency between the training and inference phases; In the standardized feature space, the K-means clustering algorithm is used, with the objective function being: 2 in, For the k-th cluster center, k ( i ) as a sample i The cluster number to which the sample belongs, each cluster center represents a typical ecological niche, n is the number of samples, and min is the minimum value function; In the original physical quantity space, the characteristic intervals of each niche are statistically analyzed, including the 10th quantile, median and 90th quantile, to form interpretable niche rules.

[0014] In one possible design, the calculation of the similarity between the environmental feature vector and the typical niche center to obtain the niche suitability score includes: Site under evaluation P Extract its environmental feature vector Then, the saved standardization parameters are used to standardize the vector, resulting in a standardized vector. ; Calculate the Euclidean distances between the plot to be evaluated and all niche centers, and take the minimum distance; The minimum distance is mapped to a niche suitability score using a Gaussian kernel function. eco : eco

[0015] in, Here, d is the ecological niche width control parameter, and d is the minimum distance. It is an exponential function with the natural constant as its base; The environmental semantic weights include slope aspect semantic weights. The introduction of these environmental semantic weights to correct the niche suitability score and generate a target crop planting suitability score and grade includes: Determine the slope aspect correction factor based on slope aspect type. aspect The slope type includes sunny slope, semi-sunny slope, semi-shaded slope and shady slope, and the corresponding slope correction factors are preset values; The final suitability score is calculated using the following formula. final : final eco aspect And apply interval constraints: final

[0016] Where max is the maximum value function; The land parcels are mapped to discrete semantic levels based on the final score to obtain the inference results; The reasoning results are saved as RDF triples and written back to the GIS geographic database, realizing a closed loop of graph reasoning from the geographic database to the knowledge graph and back to the geographic database.

[0017] Secondly, this application provides an agricultural knowledge graph construction and reasoning device based on a spatiotemporal intelligent foundation, the device comprising: The zoning structure construction module is configured to construct a five-layer zoning structure for geographic entities. Based on high-resolution remote sensing imagery as the base data, the surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, in the hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions. The vector data structure construction module is configured to define each geographic object entity as a geographic patch container, construct a vector for each geographic object entity, and obtain a five-dimensional vector data structure of the geographic object entity. The vector is encapsulated by a set of elements in a linear order, with each element corresponding one-to-one with its index, and supports dynamic adjustment of capacity. The vector includes a location vector, a morphological vector, an attribute vector, a structural vector, and a situation vector. The knowledge graph construction module is configured to use IoT sensors, cameras, lidar, and satellite remote sensing technologies to transform continuous, simulated, and complex physical phenomena in the physical world into discrete, digital, and structured data. The geographic object entities are used as nodes, and the distance between two nodes represents the similarity between two geographic entity objects, thereby constructing a knowledge graph for the planting industry. The agricultural reasoning module is configured to perform agricultural reasoning based on information transmission and knowledge constraints. It utilizes the crop knowledge graph and combines information transmission relationships and knowledge constraint relationships to analyze and predict crop planting patterns.

[0018] Thirdly, embodiments of this application provide an electronic device, including: at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, causing the at least one processor to execute the agricultural knowledge graph construction and reasoning method based on spatiotemporal intelligent base as described in the first aspect and various possible designs of the first aspect.

[0019] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions. When a processor executes the computer-executable instructions, it implements the agricultural knowledge graph construction and reasoning method based on a spatiotemporal intelligent base as described in the first aspect and various possible designs of the first aspect.

[0020] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the agricultural knowledge graph construction and reasoning method based on a spatiotemporal intelligent base as described in the first aspect and various possible designs of the first aspect.

[0021] The agricultural knowledge graph construction and reasoning method, apparatus, equipment, and storage medium based on a spatiotemporal intelligent foundation provided in this application have at least the following beneficial effects: 1) This application constructs a five-layered partitioning structure—wide-area space, regional space, local space, object space, and signal space—to deconstruct the complex land surface into multiple independent sub-regions, providing a spatial framework for expressing spatiotemporal dynamic information. Simultaneously, by constructing a five-dimensional vector for each geographic object entity—location, morphology, attributes, structure, and situation—especially by introducing the situation dimension, geographic entities can record their own evolutionary processes and dynamic changes. This overcomes the limitation of existing maps that can only describe static attributes, effectively expressing geographic phenomena with spatiotemporal dynamic characteristics such as urban expansion, vegetation phenological changes, and crop growth processes.

[0022] 2) This application constructs an agricultural knowledge graph with geographic entities as nodes and similarity distances between entities as edges, quantifying the spatiotemporal relationships between geographic entities into computable distance indicators. Based on this, an information transfer mechanism is introduced, enabling plots with similar environmental characteristics to exchange planting information. Furthermore, a set of reasoning rules based on expert knowledge is introduced to constrain and correct the reasoning results. This dual reasoning mechanism overcomes the limitations of existing technologies that rely on symbolic logic, effectively handling complex spatiotemporal relationships and supporting the prediction of future states based on historical patterns.

[0023] 3) This application uses dragon fruit cultivation as a specific application example to propose a complete method for inferring the suitability of target crop cultivation. This method extracts environmental characteristics of already cultivated plots and uses the K-means clustering algorithm to learn the typical niche centers of the target crop. For plots to be evaluated, the Euclidean distance between the plot and the niche center is calculated and mapped to a niche suitability score using a Gaussian kernel function. Environmental semantic weights such as slope aspect are introduced to correct the score, ultimately generating a cultivation suitability score and level. The entire inference process integrates multi-source spatiotemporal data, machine learning algorithms, and expert knowledge rules, achieving an intelligent closed loop from data to knowledge to decision-making.

[0024] 4) Based on clustering learning, this application further statistically analyzes the characteristic intervals of each niche in the original physical quantity space to form interpretable niche rules, giving the inference results clear physical meaning and facilitating understanding and verification by domain experts. Furthermore, the five-layer spatial structure and five-dimensional vector data structure constructed in this invention have good versatility and can be extended to the construction of digital infrastructure in multiple fields such as natural ecology, urban life, energy security, and emergency disaster relief, demonstrating broad application prospects.

[0025] 5) This application fully leverages the advantages of satellite remote sensing's comprehensive coverage, UAV remote sensing's high precision, and ground sensors and high-definition cameras' real-time monitoring capabilities to construct an integrated space-air-ground signal spatial data acquisition system. In practical application at a certain location, this system has initially achieved comprehensive, full-content, and full-element monitoring of complex surface systems. It features low cost and rapid updates, and can provide high-quality spatiotemporal information support for higher-level spatiotemporal analysis and specialized applications. Attached Figure Description

[0026] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0027] Figure 1 A flowchart illustrating an agricultural knowledge graph construction and reasoning method based on a spatiotemporal intelligent platform, provided for embodiments of this application; Figure 2 A five-layer structure diagram of the knowledge graph ontology provided in the embodiments of this application; Figure 3 Example diagram of knowledge transfer relationship provided for embodiments of this application; Figure 4 A flowchart illustrating the suitability reasoning for dragon fruit cultivation provided in this application embodiment; Figure 5 A flowchart for analyzing and predicting crop planting patterns provided in this application embodiment; Figure 6 This is a grading map generated after assessing the suitability of dragon fruit cultivation in a certain area, as provided in an embodiment of this application. Figure 7 The dragon fruit planting suitability grading map is generated after a plot-level fine-grained assessment of a local area, as provided in this application embodiment. Figure 8 This is a structural diagram of the agricultural knowledge graph construction and reasoning device based on a spatiotemporal intelligent base provided in an embodiment of this application.

[0028] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0029] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0030] The collection, storage, use, processing, transmission, provision, and disclosure of financial data or user data involved in the technical solution of this application all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0031] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0032] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0033] Existing technologies of this kind are mainly static maps: most remote sensing knowledge graphs still focus on describing the static attributes of ground objects, lacking an effective representation of dynamic changes. For example, spatiotemporal processes such as urban expansion, vegetation phenological changes, and typhoon movement paths are difficult to express concisely and systematically using existing maps. Spatiotemporal reasoning capabilities are weak: most existing map reasoning mechanisms are based on symbolic logic, making it difficult to handle complex spatiotemporal relationships and predictions based on spatiotemporal patterns.

[0034] Based on this, this application provides a method for constructing and reasoning agricultural knowledge graphs based on a spatiotemporal intelligent foundation. Based on spatiotemporal uniqueness and spatiotemporal distribution patterns, it proposes a geographic spatiotemporal digital foundation model and design concept centered around three basic elements: skeleton, container, and engine. It focuses on constructing a remote sensing intelligent computing path with system decomposition, knowledge constraint, and information transmission as the main lines, and an embedded knowledge graph within the foundation. This further establishes an intelligent spatiotemporal foundation for the agricultural field. Currently, preliminary application scenario exploration has been carried out in a certain location. It is anticipated that digital foundations for fields such as natural ecology, urban life, energy security, and emergency disaster relief will continue to emerge to provide higher-quality spatiotemporal information, support higher-level spatiotemporal analysis and thematic applications, and empower more industries to achieve high-quality development. Specifically, such as... Figure 1 As shown, the method includes the following steps S10-S40.

[0035] S10: Construct a five-layer zoning structure for geographic entities. Based on high-resolution remote sensing imagery as the base data, the land surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, following the hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions.

[0036] The purpose of step S10 is to construct a five-layer framework. In a specific embodiment, based on high-resolution remote sensing imagery as the base data, and according to elements such as administrative boundaries, elevation, slope, aspect, and land parcels, combined with the concept of geographical zoning, the complex surface is deconstructed. Following a zoning structure of wide-area space (administrative zoning) - regional space (natural zoning) - local space (functional zoning) - object space (geographic entities) - signal space (spatiotemporal data), multiple independent sub-regions of similar scale and function are ultimately formed. Each functional zone has similar terrain, similar geographic entities, similar land parcel textures, and uniform overall image features, ultimately achieving large-scale, high-precision, and high-efficiency extraction of object entities (e.g., cultivated land parcels).

[0037] Wide-area spatial data is a data structure with strong geospatial attributes and serves as the foundational data for the geospatial-temporal digital infrastructure. This embodiment utilizes administrative divisions to define wide-area spatial data, firstly delineating the provincial / municipal / district codes and names of the administrative region. The second level delineates spatial data down to the township / town level, including the township / township's administrative division code, name, and area.

[0038] Regional spatial data mainly records the observation range of the target. The administrative division code contained in the regional spatial data points to which wide-area spatial data the regional space belongs to. The observation range of the target is clustered into different types of regions and assigned corresponding regional division codes, such as cultivated areas, urban areas, and ecological background.

[0039] Local spatial data involves further subdividing each spatial region, making the observation scope more specific and giving each region its own local feature code. For example, in agricultural planting areas, land is classified as arable land, orchard land, forest land, etc., based on its planting type. In urban living and production areas, land is classified according to its function type, such as industrial land, residential land, educational land, sports land, commercial and service land, logistics and warehousing land, medical and health land, cultural facilities land, parks and green spaces, transportation land, and water and water conservancy facilities land. This series of classifications plays a crucial role in the fine extraction of the next level of object spatial data.

[0040] Object spatial data represents the granularity of production down to the smallest information model within geographic spatiotemporal data. It carries a wealth of spatial signal data and is an integration of spatiotemporal data, transforming complex landforms into geographic entities such as agricultural plots, ecological forest and grassland plots, and building plots. The currently constructed intelligent spatiotemporal foundation includes eight categories of elements: water systems, roads, buildings, water bodies, arable land, grassland, soil and rock, and others.

[0041] Spatial signal data records information on human activity, real-time geographical changes, historical survey statistics, and ecological information on resource endowments. In the high-altitude domain, this embodiment uses satellite remote sensing (full coverage and backward compatibility) to acquire spectral signals, spatiotemporal variation signals, and radiometric measurement signals of objects. In the low-altitude domain, this embodiment uses unmanned aerial vehicles (UAVs) (high-resolution but regionalized) to observe image signals, visible light signals, and oblique photography signals of objects. In the surface domain (local space and object space), this embodiment uses high-definition cameras and manual on-site sampling to collect and statistically analyze real-time change signals, historical survey signals, and social resource signals of objects.

[0042] S20: Define each geographic object entity as a geographic patch container, construct a vector for each geographic object entity, and obtain a five-dimensional vector data structure for the geographic object entity. The vector is encapsulated by a set of elements in linear order, with each element corresponding to its index one-to-one. It supports dynamic adjustment of capacity. The vector includes location vector, morphology vector, attribute vector, structure vector, and situation vector.

[0043] The purpose of step S20 is to design a base data structure to store massive amounts of data. In some embodiments, an object-oriented approach is used to design the data structure. Each heterogeneous object in the geographic world is a geographic object entity, i.e., a geographic patch container, used as a carrier for the geographic knowledge graph mapped from the physical world. Based on the data definitions of five types of space and the concept of geographic spatiotemporal, this embodiment designs a five-dimensional vector, as shown in Table 1. The five-dimensional vectors are location vector, morphology vector, attribute vector, structure vector, and situation vector. Each vector is encapsulated by a set of elements in a linear order, with each element corresponding one-to-one with its index. It supports dynamic capacity adjustment, allowing storage space to be expanded or reduced as needed. The element type of the vector is not limited; it can be a basic data type or a complex object.

[0044] Table 1 Vector Definition

[0045] S30: Through IoT sensors, cameras, LiDAR, and satellite remote sensing technologies, the continuous, simulated, and complex physical phenomena of the physical world are transformed into discrete, digital, and structured data. Geographic entities are used as nodes, and the distance between two nodes represents the similarity between two geographic entities, thus constructing a knowledge graph for the planting industry.

[0046] The purpose of step S30 is to construct a knowledge graph of the planting industry, thereby analyzing the patterns of agricultural planting. In this embodiment, various sensing technologies (such as IoT sensors, cameras, LiDAR, satellite remote sensing, etc.) are used to transform continuous, simulated, and complex physical phenomena in the real world (such as the position, state, temperature, and movement of objects, as well as human-environment interactions, etc.) into discrete, digital, and structured data. This discrete, digital, and structured data is then modeled and stored to form a geographic knowledge graph that maps the physical world. Geographic object entities serve as containers for the geographic world, i.e., geographic patch containers, used to store the entire parameter system of the geographic knowledge graph mapped from the physical world.

[0047] In some embodiments, an ontology for the knowledge graph is proposed based on a 5-layer skeleton structure, such as... Figure 2 As shown, a geographic object entity is a node, and the distance between two nodes is the similarity between the two geographic entities. The closer the distance, the higher the similarity. After adding a node, the similarity is recalculated and the entire map is updated.

[0048] Specifically, Figure 2 This is a five-layer structure diagram of the knowledge graph ontology. The ontology is built around geographic entities, establishing a five-layer hierarchical framework: wide-area space, regional space, local space, object space, and signal space. The wide-area space layer includes administrative divisions such as cities / autonomous regions, administrative levels, districts / counties, and townships. The regional space layer covers natural and functional zoning elements such as urban areas, cultivated areas, ecological localities, sub-basins, and hydrological response units. The local space layer includes functional zoning and geotechnical unit elements, while also relating to natural endowments such as water, light, heat, and air. The object space layer includes buildings. The knowledge graph ontology includes specific geographic entities such as apartments, water bodies, water systems, rivers, reservoirs, farmland, roads, dry land, ponds, paddy fields, and forest and grassland plots. It is also categorized into the element category of vegetation type. The signal spatial layer is configured with attribute parameters such as soil moisture, soil salinization, soil physicochemical properties, and microbial parameters. In addition, other types of elements are set as supplementary classifications of geographic entities in this five-layer structure knowledge graph ontology. Through the orderly arrangement of the above-mentioned layers and corresponding elements, a five-layer structure knowledge graph ontology adapted to the geographic spatiotemporal digital base is fully constructed.

[0049] In knowledge graphs, land is a relatively important element, as shown in Table 2. Specifically, it can be categorized into five types of parameters: land use parameters (LU), land cover parameters (LCC), soil (LS), land resources (LR), and land type / application (LT / LA).

[0050] Table 2 Land Element Description

[0051] The above parameter system must adhere to two major principles: knowledge constraints and information transmission. 1) Information transmission relationships: Taking agricultural planting as an example, the probability of planting the same crop on plots with similar environmental characteristics is very high. 2) Knowledge constraint relationships: Taking agricultural planting as an example, for instance, agricultural policies guide crop planting, historical statistical data guides real-time data, the climatic conditions of the agricultural plot constrain the local space, the seasons constrain crop planting, and farmers' experience in crop planting are also a type of constraint.

[0052] Examples of knowledge transfer relationships include: Figure 3 As shown, this diagram uses five parameter dimensions—Land Use Parameter (LU), Land Cover Parameter (LCC), Soil (LS), Land Resources (LR), and Land Type / Application (LT / LA)—as a framework to illustrate the transfer and correlation of agricultural knowledge between different plots. At the corresponding positions of each parameter dimension, Plot 1 is configured with specific indicators for morphology, topology, (vegetation) type, crop (vegetation) / variety, biomass / yield, soil type, water, light, heat, and aeration, design planning, and monitoring and supervision; Plot 2 is configured with specific indicators for morphology, topology, (vegetation) type, crop (vegetation) / variety, biomass / yield, soil type, natural endowment, water, light, heat, and aeration, design planning, and monitoring and supervision. Furthermore, the knowledge transfer relationship between the corresponding indicators of Plot 1 and Plot 2 is reflected through transfer markers. Meanwhile, the map treats plots 1, 2, 3, 4, 5, and 6 as geographic entity nodes, and labels adjacent nodes with distance values ​​d representing the similarity between the two geographic entities. Specifically, d=0.1 between plot 1 and plot 5, d=0.06 between plot 1 and plot 2, d=0.16 between plot 5 and plot 2, d=0.12 between plot 5 and plot 6, d=0.08 between plot 2 and plot 6, d=0.2 between plot 5 and plot 4, d=0.5 between plot 6 and plot 3, and d=0.3 between plot 4 and plot 3. In addition, plot 2 is also labeled with d=0.05, d=0.05, d=0.05, d=-0.05, d=-0.05, and d=-0.05, visually presenting the similarity association characteristics between the geographic entities.

[0053] Assuming that plot 1 and plot 2 are most similar in terms of land use parameters, soil, and land resources, and their adjacency distance is d=0.06, it can be inferred that plots 1 and 2 have some similarity in terms of vegetation type / variety or land use planning. If plot 1 is planted with crop A, plot 2 has a high probability of being planted with crop A. This leads to the set of all plots planted with crop A.

[0054] The key to constraint relationships is constructing a set of reasoning rules based on expert knowledge (e.g., the relationship between agricultural policies and crop planting, historical statistics and crop planting, land type and crop planting, farmland climate and crop planting, farmer planting experience and crop planting, etc.). In the aforementioned set of inferred crop results, results that violate the constraints are eliminated. For example, if the plot's slope is shady, the possibility of planting crop A is excluded (assuming crop A is a light-loving crop); if the plot's altitude is higher than 1000 meters, the confidence of planting crop A is reduced, such as to below 10%; if the plot is located within an administrative region, and that region has historically been dominated by crop A, then crop A is preferentially inferred as the crop type planted on the plot.

[0055] S40: Agricultural reasoning based on information transmission and knowledge constraints, utilizing crop knowledge graphs and combining information transmission relationships and knowledge constraint relationships to analyze and predict crop planting patterns.

[0056] In some embodiments, step S40, analyzing and predicting crop planting patterns, includes inferring the suitability of planting the target crop. Taking dragon fruit as an example, for instance... Figure 4 As shown, the implementation method of the target crop planting suitability inference can be as follows: Feature factor screening and feature standardization operations are performed, and the steps of extracting the dragon fruit environmental feature vector matrix are completed simultaneously. Then, typical niche centers of dragon fruit are learned through K-means clustering, and feature factor screening is performed concurrently to obtain typical environmental patterns (clan centers) for dragon fruit cultivation. The selected feature factors include rock exposure rate, vegetation coverage, slope, aspect, annual average soil water, growing season soil water, lowest monthly soil water, and soil surface temperature. After completing the feature factor processing, the calculation of the minimum niche distance is performed. Next, the steps of aspect semantic rule correction and adaptive rule threshold judgment are executed sequentially. Then, the confidence level of the inference results is evaluated. If the confidence level is low, the process returns to re-screening the feature vector dimensions and repeats the aforementioned feature processing and calculation steps. If the confidence level meets the requirements, a dragon fruit planting suitability evaluation result set is directly generated. The planting suitability inference for the target crop, dragon fruit, is completed through the above ordered process steps.

[0057] Specifically, such as Figure 5As shown, the suitability inference for planting the target crop can be carried out through the following steps S401~S404.

[0058] S401: Construct a knowledge graph RDF to build an environmental feature vector for each parcel.

[0059] It should be noted that the knowledge graph RDF can be constructed through steps S10~S30 as described above, and the construction method will not be elaborated here. Each parcel is represented as a semantic entity in the RDF triple, containing environmental attributes as shown in Table 3.

[0060] Table 3 Environmental Attributes

[0061] The selected features collectively constitute an environmental feature vector. In practice, the construction of this environmental feature vector requires selecting characteristic indicators with representational power and discriminative ability from multi-source remote sensing data, ground observation data, and geographic baseline data, based on the physiological and ecological characteristics of the target crop, the regional geographical environment, and data availability. Selection methods include, but are not limited to, the following: Based on domain knowledge and literature review, preliminary geographical environmental factors related to the growth of the target crop were identified, such as topographic factors (slope, aspect, altitude), soil factors (soil type, soil moisture, soil temperature), vegetation factors (vegetation cover, leaf area index), and climatic factors (accumulated temperature, precipitation, sunlight). Taking dragon fruit cultivation as an example, considering its light-loving, drought-tolerant, and waterlogging-intolerant karst crop, rock exposure, vegetation cover, slope, average annual soil moisture, soil moisture during the growing season, minimum soil moisture, and soil surface temperature were initially selected as candidate features. The feasibility of the initially selected features was assessed by considering the availability and spatial resolution of the data sources. For example, vegetation cover and rock exposure were extracted from high-resolution remote sensing imagery, slope and aspect were extracted from digital elevation models (DEMs), and soil moisture and surface temperature were retrieved from long-term remote sensing data to ensure that each feature index met application requirements in terms of spatial coverage and update frequency. Statistical methods such as correlation analysis and principal component analysis were used to eliminate redundant features and reduce feature dimensionality. For example, the Pearson correlation coefficient between candidate features is calculated. If the correlation coefficient between two features is higher than a preset threshold (e.g., 0.85), the feature with the clearer physical meaning or the closer association with crop growth is retained. Through principal component analysis, the original features corresponding to the principal components with a cumulative contribution rate of over 85% are selected, or the principal component scores are used as new feature vectors while retaining the physical meaning of the original features. The ability of the selected features to distinguish crop growth is verified by combining field sampling data and historical planting records. For example, the distribution differences of each feature between planted and unplanted areas are compared, and methods such as random forest or recursive feature elimination are used to assess the importance of each feature. The subset of features that contributes the most to classification or regression is selected, ultimately forming the environmental feature vector for subsequent clustering and inference.

[0062] Following the above screening process, the environmental feature vectors used for dragon fruit cultivation suitability inference include at least one of the following: rock exposure, vegetation cover, slope, average annual soil moisture, soil moisture during the growing season, minimum soil moisture, and soil surface temperature. It should be noted that the above features are merely exemplary selections for dragon fruit as a specific crop. In practical applications, the same screening approach can be used to determine the corresponding feature vector composition based on the characteristics of different target crops and different geographical regions.

[0063] S402: Extract the environmental characteristics of plots where the target crop has been planted from the RDF, and learn the typical niche centers of the target crop through a clustering algorithm.

[0064] In some embodiments, step S402 specifically includes the following steps S4021 to S4024.

[0065] S4021: Select only known plots of land where dragon fruit is grown as learning samples:

[0066] in , and These are the first, second, and nth features, respectively.

[0067] S4022: Due to the different dimensions of each feature, the features are standardized before clustering.

[0068] in, For the first i The mean of each feature, For the first i The standard deviation of each feature For the first i One characteristic, For the first i A standardized feature.

[0069] Standardized vector for:

[0070] in, , and These represent the first, second, and sixth standardized features, respectively. It should be noted that this embodiment exemplifies a total of six standardized features; this is merely an exemplary implementation and does not constitute a limitation on the number of standardized features.

[0071] Standardized parameters are permanently saved to ensure consistency between the training and inference phases.

[0072] S4023: In the standardized feature space, the K-means clustering algorithm is used, and its objective function is: 2 in, For the first k Cluster centers, k ( i ) as a sample i The cluster number to which the sample belongs, each cluster center represents a typical ecological niche, n is the number of samples, and min is the minimum value function.

[0073] Each cluster center can be understood as a typical dragon fruit ecological niche; ultimately, a standardized mean can be obtained. Standardization Scale Niche center set }

[0074] S4024: Generate interpretable niche rules.

[0075] To enhance the interpretability of the model, after completing K-means clustering and obtaining the centers of each niche, this method further performs statistical analysis on the plot samples included in each niche within the original physical quantity space. Specifically, for each environmental feature participating in the clustering (such as soil moisture, slope, rock exposure, etc.), its 10th quantile (Q10), median, and 90th quantile (Q90) are calculated, thereby quantitatively describing the typical value range of the niche in each feature dimension using interval estimation. For example, for the target crop dragon fruit, the distribution of soil moisture characteristics in a certain niche can be described as: "In this niche, the soil water content is usually between 15% and 25%, with a typical value of about 20%." In this way, the abstract cluster centers are restored to feature intervals with clear physical meaning, enabling domain experts to intuitively understand the actual geographical environmental conditions corresponding to each niche and verify their consistency with the physiological and ecological characteristics of crops.

[0076] S403: For the site to be evaluated, calculate the similarity between its environmental feature vector and the typical niche center to obtain the niche suitability score.

[0077] In some embodiments, step S403 specifically includes the following steps S4031 to S4033.

[0078] S4031: Land parcels to be evaluated P Extract its environmental feature vector Then, the saved standardization parameters are used to standardize the vector, resulting in a standardized vector. .

[0079] In this embodiment, the standardized vector is calculated using the following formula. :

[0080] S4032: Calculate the Euclidean distances between the plot to be evaluated and all niche centers, and take the minimum distance.

[0081] In this embodiment, the plot to be evaluated and all niche centers are calculated. Euclidean distance The formula is: The formula for finding the minimum distance is: .

[0082] S4033: Map the minimum distance to a niche suitability score using a Gaussian kernel function. eco : eco

[0083] in, The niche width control parameter determines how far from the "ideal growing environment for dragon fruit" is considered "approximately suitable" (using an empirical value of 2.0), where d is the minimum distance. It is an exponential function with the natural constant as its base. Niche suitability score. eco The score range is The closer to the center of the ecological niche, the higher the score.

[0084] S404: Introduce environmental semantic weights to correct the niche suitability score and generate a target crop planting suitability score and grade.

[0085] In some embodiments, the environmental semantic weights include slope semantic weights, and step S404 specifically includes the following steps S4041 to S4044.

[0086] S4041: Determine the slope aspect correction factor based on slope aspect type aspect The slope type includes sunny slope, semi-sunny slope, semi-shaded slope and shady slope, and the corresponding slope correction factors are preset values.

[0087] Considering that dragon fruit is a light-loving crop, slope aspect is introduced as a semantic proxy variable for light conditions, and the weight values ​​of each slope aspect are shown in Table 4.

[0088] Table 4 Aspect Weighting Parameter Table

[0089] It should be noted that the values ​​of the weights corresponding to each slope aspect in Table 4 are the slope aspect correction factors. aspect The values ​​mentioned above are exemplary values ​​when dragon fruit is selected as the target crop. The values ​​may differ when other crop varieties are chosen as the target crop. The range of this correction factor is [value missing]. aspect .

[0090] S4042: Calculate the final suitability score final And apply interval constraints.

[0091] Among them, the final suitability score fina The calculation formula is as follows: final eco aspect The formula for calculating interval constraints is: final

[0092] Where max is the maximum value function.

[0093] S4043: Map the land parcels to discrete semantic levels based on the final score to obtain the inference results.

[0094] In this embodiment, based on the final score, the land parcel is mapped to a discrete semantic level using the following formula. :

[0095] in, For the final score, = fina .

[0096] S4044: Save the reasoning results in the form of RDF triples and write them back to the GIS geographic database to realize the closed loop of graph reasoning from the geographic database to the knowledge graph and back to the geographic database.

[0097] For example, the inference results are written back to the GDB layer field names and their meanings as shown in Table 5.

[0098] Table 5 Field Names and Meanings

[0099] like Figure 6 The image shows a grading map generated after assessing the suitability of dragon fruit cultivation in a specific area using the method described in this application. This area is located in a typical karst landform zone, characterized by fragmented terrain, high rock exposure, and uneven soil distribution, making precise suitability assessment difficult using traditional methods. As can be seen from the map, this method divides the study area into multiple continuously distributed grid units, mapping each grid unit to three levels—high suitability, medium suitability, and low suitability—based on the final suitability score, represented by different gray levels or legends. High suitability areas are mainly concentrated in areas with gentle terrain, thick soil layers, and good moisture conditions on both sides of river valleys, exhibiting a continuous, strip-like distribution. Medium suitability areas are distributed around the high suitability areas, in regions with moderate terrain undulation and increased rock exposure. Low suitability areas are widely distributed on steep slopes, mountaintops, and karst desertification areas with high rock exposure. Figure 6The assessment results are highly consistent with field surveys and the experience of agricultural experts, verifying the feasibility and accuracy of this method in complex land systems, and providing a scientific basis for agricultural planting planning on a large regional scale.

[0100] like Figure 7 As shown, in Figure 6 Based on the shown area, a dragon fruit planting suitability grading map was generated after further refined assessment of local areas at the plot level. This method, based on the object space in a five-layer spatial structure, deconstructs the complex surface into independent agricultural plots, with each plot participating as an independent geographic object entity in knowledge graph construction and reasoning. Figure 7 It is clear that different plots exhibit varying suitability levels: some plots are assessed as highly suitable, typically characterized by gentle slopes, stable soil moisture, low rock exposure, and sunny or semi-sunny slopes; some plots are assessed as moderately suitable, with environmental characteristics close to the ecological niche boundary for dragon fruit growth; and some plots are assessed as poorly suitable, either located on shady slopes, steep slopes, or with poor soil moisture conditions, resulting in significantly lower scores after adjustment by the aspect correction factor. Figure 7 This visually demonstrates the method's ability to perform refined assessments at the plot scale, distinguishing suitability changes between adjacent plots due to differences in micro-topography and microclimate, thus overcoming the limitation of traditional methods that can only perform regional-level coarse-grained assessments. Figure 6 and Figure 7 As can be seen, this method achieves seamless multi-scale integration from regional macro-planning to plot-level micro-decision-making, fully demonstrating the technological advancement and practical value of agricultural knowledge graph construction and reasoning methods based on spatiotemporal intelligence in complex geographical environments.

[0101] This application also provides an agricultural knowledge graph construction and reasoning device based on a spatiotemporal intelligent foundation, used to implement the methods described in any of the above embodiments, such as... Figure 8 As shown, the agricultural knowledge graph construction and reasoning device based on a spatiotemporal intelligence platform includes: The zoning structure construction module 801 is configured to construct a five-layer zoning structure for geographic entities. Based on high-resolution remote sensing imagery as the base data, the surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, in the hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions. The vector data structure construction module 802 is configured to define each geographic object entity as a geographic patch container, construct a vector for each geographic object entity, and obtain a five-dimensional vector data structure of the geographic object entity. The vector is encapsulated by a set of elements in a linear order, with each element corresponding one-to-one with its index, and supports dynamic adjustment of capacity. The vector includes a location vector, a morphological vector, an attribute vector, a structural vector, and a situation vector. The knowledge graph construction module 803 is configured to use IoT sensors, cameras, lidar, and satellite remote sensing technologies to transform continuous, simulated, and complex physical phenomena in the physical world into discrete, digital, and structured data. It uses the geographic object entities as nodes and the distance between two nodes to represent the similarity between two geographic entity objects, thereby constructing a knowledge graph for the planting industry. The agricultural reasoning module 804 is configured to perform agricultural reasoning based on information transmission and knowledge constraints. It utilizes the crop knowledge graph and combines information transmission relationships and knowledge constraint relationships to analyze and predict crop planting patterns.

[0102] This application provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0103] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0104] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0105] The electronic device provided in this application embodiment can be the terminal device described in the above embodiments.

[0106] This application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solution of the agricultural knowledge graph construction and reasoning method based on the spatiotemporal intelligent base described in the above embodiments.

[0107] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solution of the agricultural knowledge graph construction and reasoning method based on the spatiotemporal intelligent base described in the above embodiments.

[0108] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0109] The modules described as separate components may or may not be physically separate. The components shown as modules 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 the modules can be selected to implement the solution of this embodiment according to actual needs.

[0110] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0111] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0112] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0113] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0114] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0115] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0116] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0117] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A spatio-temporal intelligent base-based agricultural knowledge graph construction and reasoning method, characterized in that, The method includes: A five-layer zoning structure for geographic entities is constructed. Based on high-resolution remote sensing imagery, the land surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, following a hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions. The wide-area space is the administrative division space, the regional space is the functional zoning space, the local space is the functional subdivision space, the object space is the geographic entity patch space, and the signal space is the multi-source sensing data space. Each geographic object entity is defined as a geographic patch container, and a vector is constructed for each geographic object entity to obtain a five-dimensional vector data structure of the geographic object entity. The vector is encapsulated by a set of elements in a linear order, with each element corresponding to its index one-to-one, and supports dynamic adjustment of capacity. The vector includes location vector, morphology vector, attribute vector, structure vector and situation vector. By using IoT sensors, cameras, lidar, and satellite remote sensing technologies, the continuous, simulated, and complex physical phenomena of the physical world are transformed into discrete, digital, and structured data. Using the geographic object entities as nodes, the distance between two nodes represents the similarity between two geographic entity objects, thus constructing a knowledge graph for the planting industry. Agricultural reasoning based on information transmission and knowledge constraints utilizes the aforementioned crop knowledge graph, combined with information transmission relationships and knowledge constraint relationships, to analyze and predict crop planting patterns. The analysis and prediction of crop planting patterns includes inference on the suitability of target crop planting, which includes the following steps: A knowledge graph RDF is constructed, and an environmental feature vector is constructed for each plot. The environmental feature vector includes at least one of the following characteristics: rock exposure, vegetation cover, slope, annual average soil moisture, soil moisture during the growing season, minimum soil moisture, and soil surface temperature. Environmental features of plots planted with target crops are extracted from RDF, and typical niche centers of target crops are learned through clustering algorithms. For each site to be evaluated, the similarity between its environmental feature vector and the typical niche center is calculated to obtain a niche suitability score. Environmental semantic weights are introduced to correct the niche suitability score, generating a target crop planting suitability score and grade.

2. The spatio-temporal intelligent base pad-based agricultural knowledge graph construction and reasoning method according to claim 1, characterized in that, The wide area space includes provincial and municipal division codes, provincial and municipal names, township and town division codes, township and town names, and area size data; The regional space includes administrative division codes and regional division codes, and the regional division codes include codes for cultivated areas, urban areas and ecological baselines; The local space includes local feature codes, which include codes for cultivated land, orchards, and forest land in agricultural planting areas, and codes for industrial land, residential land, educational land, sports land, commercial service land, logistics and warehousing land, medical and health land, cultural facilities land, parks and green spaces, transportation land, and water and water conservancy facilities land in urban living and production areas. The object space includes agricultural land parcels, ecological forest and grassland parcels, and building parcels; The signal space includes spectral signals, spatiotemporal variation signals, and radiometric measurement signal data acquired through satellite remote sensing, image signals, visible light signals, and oblique photography signal data acquired through drones, as well as real-time variation signals, historical survey signals, and social resource signal data acquired through high-definition cameras and on-site sampling.

3. The spatio-temporal intelligent base pad-based agricultural knowledge graph construction and reasoning method according to claim 1, characterized in that, The position vector is used to store the orientation information of the object in three-dimensional space; the shape vector is used to store the shape feature information of points, lines, surfaces, and volumes. The attribute vector is used to store the parameter system information of the geographic knowledge graph; the structure vector is used to store the geographic relationship information between objects, between objects and the environment, and between objects; the situation vector is used to store the evolution process and situation change information of object entities.

4. The spatio-temporal intelligent base pad-based agricultural knowledge graph construction and reasoning method according to claim 1, characterized in that, The information transmission relationship includes: the probability of planting the same crop on plots with the same environmental characteristics is higher than a preset threshold; The knowledge constraints include: the guiding constraints of agricultural policies on crop planting, the guiding constraints of historical statistical data on real-time data, the constraints of climatic conditions in the space where agricultural plots are located on local space, the constraints of seasons on crop planting, and the constraints of farmers' planting experience.

5. The method for constructing and reasoning agricultural knowledge graphs based on a spatiotemporal intelligent foundation according to claim 1, characterized in that, The method of learning the typical niche centers of the target crop through clustering algorithms includes: Select plots of land known to be planted with the target crop as learning samples. ,in , and These are the first, second, and nth features, respectively; The features are standardized using the following formula: in, For the first i The mean of each feature, For the first i The standard deviation of each feature For the first i One characteristic, For the first i One standardized feature; Standardized parameters are permanently saved to ensure consistency between the training and inference phases; In the standardized feature space, the K-means clustering algorithm is used, with the objective function being: in, For the k-th cluster center, k ( i ) as a sample i The cluster number to which the sample belongs, each cluster center represents a typical ecological niche, n is the number of samples, and min is the minimum value function; In the original physical quantity space, the characteristic intervals of each niche are statistically analyzed, including the 10th quantile, median and 90th quantile, to form interpretable niche rules.

6. The spatio-temporal intelligent base pad-based agricultural knowledge graph construction and reasoning method according to claim 5, characterized in that, The similarity between environmental feature vectors and typical niche centers is calculated to obtain a niche suitability score, including: Site under evaluation P Extract its environmental feature vector Then, the saved standardization parameters are used to standardize the vector, resulting in a standardized vector. ; Calculate the Euclidean distances between the plot to be evaluated and all niche centers, and take the minimum distance; The minimum distance is mapped to a niche suitability score using a Gaussian kernel function. eco : in, Here, d is the ecological niche width control parameter, and d is the minimum distance. It is an exponential function with the natural constant as its base; The environmental semantic weights include slope aspect semantic weights. The introduction of these environmental semantic weights to correct the niche suitability score and generate a target crop planting suitability score and grade includes: Determine the slope aspect correction factor based on slope aspect type. aspect The slope type includes sunny slope, semi-sunny slope, semi-shaded slope and shady slope, and the corresponding slope correction factors are preset values; The final suitability score is calculated by the following equation final : And apply interval constraints: Where max is the maximum value function; The land parcels are mapped to discrete semantic levels based on the final score to obtain the inference results; The reasoning results are saved as RDF triples and written back to the GIS geographic database, realizing a closed loop of graph reasoning from the geographic database to the knowledge graph and back to the geographic database.

7. A spatio-temporal intelligent base station-based agricultural knowledge graph construction and reasoning device for implementing the method of any one of claims 1-6, characterized in that, The device includes: The zoning structure construction module is configured to construct a five-layer zoning structure for geographic entities. Based on high-resolution remote sensing imagery as the base data, the surface is deconstructed according to administrative boundaries, elevation, slope, aspect, and land parcel elements, in the hierarchical order of wide-area space, regional space, local space, object space, and signal space, forming multiple independent sub-regions. The vector data structure construction module is configured to define each geographic object entity as a geographic patch container, construct a vector for each geographic object entity, and obtain a five-dimensional vector data structure of the geographic object entity. The vector is encapsulated by a set of elements in a linear order, with each element corresponding one-to-one with its index, and supports dynamic adjustment of capacity. The vector includes a location vector, a morphological vector, an attribute vector, a structural vector, and a situation vector. The knowledge graph construction module is configured to use IoT sensors, cameras, lidar, and satellite remote sensing technologies to transform continuous, simulated, and complex physical phenomena in the physical world into discrete, digital, and structured data. The geographic object entities are used as nodes, and the distance between two nodes represents the similarity between two geographic entity objects, thereby constructing a knowledge graph for the planting industry. The agricultural reasoning module is configured to perform agricultural reasoning based on information transmission and knowledge constraints. It utilizes the crop knowledge graph and combines information transmission relationships and knowledge constraint relationships to analyze and predict crop planting patterns.

8. An electronic device, comprising: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes the computer execution instructions stored in the memory to implement the agricultural knowledge graph construction and reasoning method based on the spatiotemporal intelligent base as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the agricultural knowledge graph construction and reasoning method based on a spatiotemporal intelligent base as described in any one of claims 1-6.

Images