Knowledge graph dynamic updating method and device based on crop growth process

Abstract

The application provides a kind of knowledge graph dynamic updating method and device based on crop growth process, wherein the above-mentioned method includes: environmental factor data and crop growth stage are input to the trained Transform model, to obtain time series feature;Relationship graph is constructed based on environmental factor data and crop growth index by graph convolution network, to obtain environmental crop growth relationship graph;Feature extraction is carried out based on environmental crop growth relationship graph by graph attention network, to obtain spatial feature matrix;Feature extraction is carried out based on spatial feature matrix by spatial graph convolution network, to obtain high-dimensional spatial feature;Alignment and conversion are carried out based on high-dimensional spatial feature by spatial transformation network, to obtain geographic spatial feature;Feature fusion is carried out based on time series feature and geographic spatial feature by deep fusion model, to obtain comprehensive feature;The preset knowledge graph is incrementally updated based on environmental factor data, crop growth index and comprehensive feature.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and in particular to a method and device for dynamically updating a knowledge graph based on the crop growth process. Background Art

[0002] Knowledge graph is a knowledge management method implemented through semantic network technology, and is widely used in data organization and relationship expression in multiple fields.

[0003] Knowledge graph technology is used to build a crop growth knowledge base to help agricultural practitioners make crop management decisions. Current agricultural knowledge graph systems are usually built based on expert experience, historical data, and fixed rules.

[0004] However, building agricultural knowledge graphs based on historical data or static rules lacks consideration of the time dimension and cannot dynamically adjust and update knowledge content. Summary of the Invention

[0005] The present invention provides a method and device for dynamically updating a knowledge graph based on the crop growth process, which is used to address the defects of the existing technology in that the time dimension is not taken into consideration and the knowledge content cannot be dynamically adjusted and updated, and to realize the dynamic updating of the knowledge graph in a complex and changing environment.

[0006] The present invention provides a method for dynamically updating a knowledge graph based on the crop growth process, comprising the following steps. Acquire environmental factor data and crop growth stages; input the environmental factor data and the crop growth stages into a trained Transformer model to obtain crop growth indicators output by the trained Transformer model that correspond to the crop growth stages respectively; input the crop growth indicators into the trained Transformer model to obtain time series features output by the trained Transformer model; construct a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolution network to obtain an environment-crop growth relationship graph; perform feature extraction based on the environment-crop growth relationship graph through a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between the environmental factor data and the crop growth indicators; perform feature extraction based on the spatial feature matrix through a spatial graph convolution network to obtain high-dimensional spatial features; perform alignment and transformation based on the high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features; perform feature fusion based on the time series features and the geographic spatial features through a deep fusion model to obtain comprehensive features; and incrementally update a preset knowledge graph based on the environmental factor data, the crop growth indicators, and the comprehensive features.

[0007] According to the crop growth process based knowledge graph dynamic updating method provided by the application, the environment factor data and the crop growth stage are input into the trained Transformer model to obtain the crop growth indexes corresponding to the crop growth stages respectively output by the trained Transformer model, the environment factor data includes temperature, humidity and illumination, the crop growth stage includes sowing, germination and seedling stage, the crop growth state is predicted based on the environment factor data and the crop growth stage through the trained Transformer model to obtain the crop growth indexes corresponding to the crop growth stages respectively, and the crop growth indexes include leaf area index and plant height.

[0008] According to the crop growth process based knowledge graph dynamic updating method provided by the application, the crop growth indexes are input into the trained Transformer model to obtain the time sequence features output by the trained Transformer model, the multi-head self-attention mechanism of the trained Transformer model is used to predict the crop growth indexes corresponding to the crop growth stages respectively to obtain the time sequence features, and the time sequence features are used to predict the crop growth trend at a future time.

[0009] According to the crop growth process based knowledge graph dynamic updating method provided by the application, before the preset knowledge graph is incrementally updated based on the environment factor data, the crop growth indexes and the comprehensive features, the method further comprises: constructing a multi-level knowledge graph based on a graph database, wherein the multi-level comprises a basic layer, a rule layer and an inference layer.

[0010] According to the crop growth process based knowledge graph dynamic updating method provided by the application, the multi-level knowledge graph is constructed based on a graph database, the construction of the multi-level knowledge graph comprises: constructing a basic layer of the multi-level knowledge graph based on crop types, environment factors, climate conditions, soil types, crop growth states and farming operations, the basic layer is used to represent each entity and entity relationship based on a resource description framework, a rule layer of the multi-level knowledge graph is constructed based on logical rules and association patterns in the crop growth process, the rule layer is used to represent crop growth rules, environment and crop relationship and farming operation influence, and an inference layer of the multi-level knowledge graph is constructed based on the basic layer and the rule layer, the inference layer is used to generate crop growth decisions and farming operation suggestions.

[0011] According to the knowledge graph dynamic updating method based on crop growth process provided by the application, the preset knowledge graph is incrementally updated based on the environmental factor data, the crop growth indicators and the comprehensive features, and the entity classification and entity relationship in the knowledge graph are adjusted through graph neural network and knowledge graph embedding based on the environmental factor data, the crop growth indicators and the comprehensive features.

[0012] The application further provides a knowledge graph dynamic updating device based on crop growth process, comprising the following modules: an acquisition module, configured to acquire environmental factor data and crop growth stages; a Transformer module, configured to input the environmental factor data and the crop growth stages into a trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages respectively output by the trained Transformer model; the Transformer module is further configured to input the crop growth indicators into the trained Transformer model to obtain time sequence features output by the trained Transformer model; a construction module, configured to construct a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolution network to obtain an environmental crop growth relationship graph; a feature extraction module, configured to extract features based on the environmental crop growth relationship graph through a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is used to represent the correlation between the environmental factor data and the crop growth indicators; the feature extraction module is further configured to extract features based on the spatial feature matrix through a spatial graph convolution network to obtain high-dimensional spatial features; a transformation module, configured to align and convert the high-dimensional spatial features based on a spatial transformation network to obtain geographic spatial features; a fusion module, configured to fuse features based on the time sequence features and the geographic spatial features through a deep fusion model to obtain comprehensive features; and an updating module, configured to incrementally update a preset knowledge graph based on the environmental factor data, the crop growth indicators and the comprehensive features.

[0013] The application further provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the knowledge graph dynamic updating method based on crop growth process according to any of the above when executing the program.

[0014] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the knowledge graph dynamic updating method based on crop growth process according to any of the above.

[0015] The application further provides a computer program product comprising a computer program which, when executed by a processor, implements any of the above-mentioned knowledge graph dynamic updating methods based on crop growth processes.

[0016] The knowledge graph dynamic updating method and device based on crop growth processes provided by the application can predict corresponding crop growth indicators according to environmental factor data and crop growth stages through the trained Transformer model, thereby capturing long-term dependencies in the data; the crop growth indicators are input into the Transformer model again, time series features can be extracted to obtain the change rule of crop growth over time; the environmental crop growth relationship graph constructed by the graph convolution network can show the complex relationship between environmental factors and crop growth indicators; the spatial feature matrix can be obtained by feature extraction based on the environmental crop growth relationship graph through the graph attention network, which can represent the correlation between environmental factor data and crop growth indicators; the geographic spatial features can be obtained through the processing of the spatial graph convolution network and the spatial transformation network, which can reflect the differences of crop growth in different geographic spatial positions; the comprehensive features can be obtained by fusing the time series features and the geographic spatial features through the deep fusion model, which can comprehensively reflect the comprehensive influence of time, space and environmental factors in the crop growth process; the preset knowledge graph is incrementally updated based on the environmental factor data, the crop growth indicators and the comprehensive features, the incremental update mechanism adds the newly added environmental factors, crop growth indicators and related relationships (i.e. comprehensive features) into the preset knowledge graph, ensuring the accuracy and timeliness of the graph structure, thereby solving the defects that the existing technology lacks consideration of the time dimension and cannot dynamically adjust and update the knowledge content. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description one by one. Obviously, the drawings in the following description are some embodiments of the application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0018] Figure 1 is a flowchart of the knowledge graph dynamic updating method based on crop growth processes provided by the application.

[0019] Figure 2 is a flowchart of the crop growth process dynamic knowledge expression and graph construction provided by the application.

[0020] Figure 3 is a flowchart of the time and space multi-dimensional feature extraction provided by the application.

[0021] Figure 4 is a structural schematic diagram of the knowledge graph dynamic updating device based on crop growth process provided by the present application.

[0022] Figure 5 is a physical structure schematic diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application.

[0024] With the continuous development of modern agricultural technology, the fine and intelligent demand of crop production management is increasing. The growth process of crops is affected by many factors such as climate, soil conditions, pests and diseases, and the dynamic changes of these factors have a direct impact on different stages of crop growth. These factors not only have different effects on crop growth at different time nodes, but also may show significant dynamic changes due to differences in region and crop variety. Therefore, real-time monitoring and scientific decision-making of crop growth process in agricultural management become the key to improve production efficiency and yield. Knowledge graph is a knowledge management method realized by semantic web technology, which is widely used in data organization and relationship expression in many fields. It expresses data in the form of graph, uses nodes to represent entities and edges to represent the relationship between entities, so as to realize the expression, reasoning and query function of complex knowledge. In the field of agriculture, knowledge graph technology is used to build crop growth knowledge base to help agricultural practitioners make crop management decisions. Current agricultural knowledge graph systems are usually constructed based on expert experience, historical data and fixed rules. These static graphs collect and integrate knowledge from different sources, such as agricultural literature, expert experience and field data, to form a systematic expression of specific crops, pest control, planting techniques, etc., providing reference and support for agricultural production management.

[0025] In view of the significant deficiencies of traditional agricultural knowledge graph systems in the face of complex and variable crop growth environments in actual production, the present application proposes a knowledge graph dynamic updating method and device based on crop growth process. By introducing dynamic modeling, multi-source data fusion and real-time updating mechanism, agricultural managers can make scientific and timely decisions in complex environments, thereby improving the growth efficiency and yield of crops.

[0026] The existing knowledge graph construction method mainly focuses on the static storage and query of knowledge, and at least one of the following defects exists.

[0027] Static: Traditional knowledge graphs are typically built based on historical data or static rules, lacking consideration of the time dimension and unable to dynamically adjust and update knowledge content. This static nature presents significant limitations when faced with the dynamic changes in actual crop growth.

[0028] Single-dimensional knowledge representation models in existing technologies are often single-dimensional, typically covering only a specific aspect of crop growth, such as weather data or soil conditions, and lacking comprehensive consideration of the entire crop growth process. This single-dimensionality fails to meet the needs of multi-dimensional information integration in agricultural management.

[0029] Lag: Due to the lack of a dynamic update mechanism in the knowledge graph, the decision support information it provides often lags behind the actual production situation and cannot respond promptly to sudden changes in the crop growth process, such as sudden pests and diseases, climate anomalies, etc.

[0030] The above technical defects show that the existing knowledge graph construction method has significant shortcomings when facing the complex and changeable crop growth process in agricultural production, especially in actual production, it cannot provide accurate and efficient real-time decision support, which seriously restricts the implementation effect of modern agricultural intelligent management.

[0031] The purpose of the present invention is to provide a method and device for dynamically updating a knowledge graph based on the crop growth process, aiming to overcome the defects of staticity, singleness and hysteresis in the prior art, realize dynamic monitoring of the crop growth process and real-time knowledge updating, and thus improve the intelligence and precision level of agricultural production management. Specifically, the present invention realizes the dynamic construction and intelligent reasoning of the knowledge graph by constructing a knowledge expression model that can dynamically adapt to changes in the crop growth environment, and combines multi-dimensional data fusion and real-time update mechanisms. This will enable agricultural production management to make more scientific and timely decisions in the face of complex and changing environments, ensuring the optimal growth state and efficient yield of crops.

[0032] Optionally, the method for dynamically updating the knowledge graph based on the crop growth process of the embodiment of the present application can be executed by a server, or by a terminal device, or jointly by a server and a terminal device, taking the example of the method for dynamically updating the knowledge graph based on the crop growth process in this embodiment being executed by a server.

[0033] Figure 1 This is a flow chart of the method for dynamically updating the knowledge graph based on the crop growth process provided by the present invention. Figure 1 As shown, the method includes the following steps.

[0034] Step 101: Acquire environmental factor data and crop growth stages.

[0035] In this embodiment, IoT sensors (such as soil moisture sensors and weather stations) and remote sensing technologies (such as drones and satellite imagery) are used to collect real-time data on environmental factors. The collected data includes meteorological data (temperature, humidity, precipitation), soil data (moisture content, nutrients), crop growth data (growth rate, leaf area index), and management measures (fertilization, irrigation, etc.).

[0036] In an embodiment of the present invention, the growth process of a crop is divided into several key stages to obtain crop growth stages, wherein the crop growth stages include sowing, germination, seedling stage, tillering stage, heading stage and maturity stage.

[0037] Step 102 : Input the environmental factor data and the crop growth stage into the trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages output by the trained Transformer model.

[0038] In this embodiment of the present invention, a Transformer model is used to predict crop growth status. By inputting environmental factors (such as temperature, humidity, and light), the trained Transformer model predicts crop growth indicators (such as leaf area index and plant height) at various growth stages.

[0039] The core formula of the Transformer model can be referred to the following formula (1):

[0040] (1)

[0041] in, represents the attention mechanism, is the query matrix, is the bond matrix, is the value matrix, is the dimension of the bond matrix, represents the normalization function, Represents the transposed matrix.

[0042] Step 103: Input the crop growth index into the trained Transformer model to obtain the time series features output by the trained Transformer model.

[0043] In an embodiment of the present invention, the Transformer model is used for time series prediction, which can effectively capture long-term dependencies and nonlinear changes. Combined with the multi-head self-attention mechanism, the time series data of the crop growth process is processed to obtain time series features.

[0044] Through the embodiments of the present invention, the trained Transformer model can accurately predict corresponding crop growth indicators based on environmental factor data and crop growth stages, so as to capture long-term dependencies in the data and provide high-precision prediction results. By inputting the crop growth indicators into the Transformer model again, time series features can be extracted, which helps to understand the changes in crop growth over time and provide an important basis for subsequent decision-making and analysis.

[0045] Step 104: construct a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolutional network to obtain an environment-crop growth relationship graph.

[0046] Step 105: extract features based on the environment-crop growth relationship graph through a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between environmental factor data and crop growth indicators.

[0047] In this embodiment of the present invention, crop growth indicators (leaf area index, plant height) at each crop growth stage are input into a graph convolutional network (GCN). This network then analyzes and constructs a relationship graph between environmental factors and crop growth stages. This generated environment-crop growth relationship graph is then input into a graph attention network (GAT), which further focuses on the impact of various environmental factors on different growth stages. Ultimately, a new feature matrix (i.e., a spatial feature matrix) is output. This spatial feature matrix reveals the relevance and importance of various environmental factors on crop growth indicators at different growth stages.

[0048] Graph Convolutional Network (GCN) and Graph Attention Network (GAT) are used to analyze the various dynamic factors affecting crop growth. GCN and GAT can process the complex relationship between environmental data and crop growth, capture the node features and relationships in the graph, and thus provide more accurate dynamic adjustment suggestions. The calculation formula (2) of the GCN model is:

[0049] (2)

[0050] in, Indicates the The nodes of a layer represent matrices, is the activation function, is the degree matrix, is the adjacency matrix and the identity matrix of and, is a learnable weight matrix.

[0051] The calculation formula (3) of the GCN model is:

[0052] (3)

[0053] in, Representation node The updated feature vector, represents the activation function, Representation node The set of neighbor nodes of Representation node To Node The attention weight, represents the weight matrix, Representation node The eigenvector of .

[0054] Through the embodiment of the present invention, a spatial feature matrix with spatial information is obtained for subsequent reasoning tasks.

[0055] Step 106: Extract features based on the spatial feature matrix through a spatial graph convolutional network to obtain high-dimensional spatial features.

[0056] Step 107: align and transform the high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features.

[0057] In this embodiment of the present invention, a spatial graph convolutional network (SGCN) and a spatial transformer network (STN) are used to process geospatial data. These models can map spatial data into a feature space, extract and transform spatial features, and thus achieve accurate modeling and analysis of spatial data.

[0058] In the previous paper, graph convolutional networks (GCN) and graph attention networks (GAT) were used to extract spatial features during crop growth. These features mainly come from the relationship graph between the environment and crop growth indicators, representing local spatial information.

[0059] Next, these extracted spatial feature matrices (such as local spatial relationships) are fed into the Spatial Graph Convolutional Network (SGCN). The SGCN further extracts and analyzes complex spatial features related to geographic location, maps them into the feature space, and captures a wider range of spatial dependencies.

[0060] The high-dimensional spatial features extracted by the spatial graph convolutional network are fed into the spatial transformer network (STN). The STN transforms the high-dimensional spatial features, performs refined spatial feature alignment and conversion, and outputs a geospatial feature matrix for precise modeling and analysis.

[0061] Among them, the formula (4) of the spatial graph convolutional network model is:

[0062] (4)

[0063] in, is the normalized adjacency matrix, It is The node feature matrix of the layer, is a learnable weight matrix, is the activation function.

[0064] Formula (5) of the spatial transformation network model is:

[0065] (5)

[0066] in, is the spatial transformation parameter, is the input image, represents an affine function, It is the output after spatial transformation.

[0067] Here, first, preliminary spatial features are extracted through GCN / GAT; second, complex spatial dependency features are extracted through SGCN; finally, spatial transformation and alignment are performed through STN to obtain the final geospatial data features.

[0068] Step 108: Perform feature fusion based on time series features and geographic space features through a deep fusion model to obtain comprehensive features.

[0069] Real-time data collection is performed using IoT sensors (such as soil moisture sensors and weather stations) and remote sensing technologies (such as drones and satellite imagery). Key data collected includes: meteorological data (temperature, humidity, precipitation), soil data (moisture content, nutrients), crop growth data (growth rate, leaf area index), and management measures (fertilization, irrigation, etc.).

[0070] In this embodiment of the present invention, a deep fusion model is applied, combining convolutional neural networks and recurrent neural networks to process time series and spatial data from different sensors. The convolutional neural network is used for feature extraction, and the recurrent neural network is used for time series modeling, achieving efficient data fusion and processing.

[0071] Apply a deep fusion model to integrate the temporal data processed by the Transformer model and the geospatial data processed by the graph convolutional network, graph attention network, spatial graph convolutional network, and spatial transformation network.

[0072] Time series data is modeled through the Transformer model, and its multi-head self-attention mechanism is used to effectively capture the long-term dependence and nonlinear changes in the time dimension during crop growth, generating time series features.

[0073] Spatial data is used to extract spatial features through graph convolutional networks and graph attention networks to capture the spatial dependencies between environmental factors and crop growth indicators and generate a spatial feature matrix.

[0074] The spatial feature matrix is ​​input into the spatial graph convolutional network and the spatial transformation network for feature extraction, spatial feature alignment and transformation to obtain geographic spatial features.

[0075] The deep fusion model takes time series features and geographic spatial features as input and performs efficient fusion through the fusion layer; the fusion layer can use a fully connected layer or attention mechanism to achieve collaborative analysis and processing of spatiotemporal features, and ultimately generate comprehensive features to improve the prediction accuracy of crop growth status.

[0076] Step 109: incrementally update the preset knowledge graph based on environmental factor data, crop growth indicators and comprehensive characteristics.

[0077] In an embodiment of the present invention, a time series-based incremental update method is adopted to dynamically input new data; the real-time stream processing framework Apache Flink and incremental learning technology are used to update model parameters and knowledge graphs in real time, supporting rapid processing of data streams and dynamic adjustment of models.

[0078] For example, environmental factors and crop growth-related data are collected in real time through various sensors (such as temperature, humidity, and light). Before entering the system, the data stream is first preprocessed to remove noise and outliers and perform normalization.

[0079] The preprocessed data is stream processed using Apache Flink. Flink handles high-throughput, low-latency real-time data streams and supports windowing, batching real-time data to accommodate analysis at different time granularities. Data within each time window is delivered to the model in a timely manner.

[0080] Using incremental learning technology, model parameters are dynamically updated based on the new input each time new data arrives. This incremental learning module gradually adjusts the model's weights and parameters based on the feature differences between historical and newly added data, without having to retrain the entire model. This significantly improves computational efficiency, ensuring real-time and stable model performance, especially in the case of large-scale data streams.

[0081] As new data and environmental changes occur, the knowledge graph also needs to be updated in real time. The incremental update mechanism dynamically adds new environmental factors, crop growth indicators, and related relationships to the existing knowledge graph, ensuring the accuracy and timeliness of the graph structure. Furthermore, based on changes in the graph structure, new reasoning processes can be triggered to explore potential knowledge connections, further enhancing the level of intelligence.

[0082] According to a method for dynamically updating a knowledge graph based on the crop growth process provided by the present invention, environmental factor data and crop growth stages are input into a trained Transformer model, and crop growth indicators corresponding to the crop growth stages are obtained as outputs of the trained Transformer model, including:

[0083] Input environmental factor data and crop growth stages into the trained Transformer model. Environmental factor data includes temperature, humidity, and light; crop growth stages include sowing, germination, and seedling stages.

[0084] Through the trained Transformer model, crop growth status is predicted based on environmental factor data and crop growth stages, and crop growth indicators corresponding to the crop growth stages are obtained. Among them, crop growth indicators include leaf area index and plant height; crop growth indicators are used to represent crop growth status.

[0085] In this embodiment of the present invention, the input environmental factor data specifically includes temperature, humidity, and light. These environmental factor data, along with the crop growth stage (sowing, germination, and seedling), are used as input to predict crop growth status using a trained Transformer model. The output is growth indicators corresponding to each growth stage, including leaf area index and plant height, which are used to dynamically monitor crop growth.

[0086] According to a method for dynamically updating a knowledge graph based on the crop growth process provided by the present invention, crop growth indicators are input into a trained Transformer model to obtain time series features output by the trained Transformer model, including:

[0087] Through the multi-head self-attention mechanism of the trained Transformer model, predictions are made based on crop growth indicators corresponding to the crop growth stages to obtain time series features, among which the time series features are used to predict crop growth trends in the future.

[0088] The growth indicators of crops at various growth stages (leaf area index, plant height) are predicted using the aforementioned Transformer model, and the growth indicators corresponding to each growth stage of crops are obtained; then, the growth indicators of crops are used as input, and the Transformer model is further used for time series prediction to obtain time series features. Through the multi-head self-attention mechanism of the Transformer model, long-term dependencies and nonlinear changes in the growth process of crops can be captured, and the growth trend at future time points can be predicted.

[0089] The environmental factor data and the growth stage of crops are input into the trained Transformer model, and the Transformer model predicts the growth indicators of crops at different growth stages based on the environmental factor data. The Transformer model is mainly used to capture time series features of the growth state here, helping to predict the changes in indicators at different growth stages under specific environmental conditions.

[0090] According to the knowledge graph dynamic updating method based on the growth process of crops provided by the application, before incrementally updating the preset knowledge graph based on the environmental factor data, the growth indicators of crops and the comprehensive features, the method further comprises:

[0091] A multi-level knowledge graph is constructed based on a graph database, wherein the multi-level includes a basic layer, a rule layer and an inference layer.

[0092] In the embodiments of the application, a multi-level knowledge graph is designed, including a basic layer (crop species, environmental factors), a rule layer (crop growth rules) and an inference layer (inference logic and decision support). The basic layer uses Resource Description Framework (RDF) to represent data, and the rule layer uses SPARQL query language to define rules.

[0093] The knowledge graph is stored using a graph database Neo4j, and the nodes and relationships in the knowledge graph are dynamically adjusted. In combination with a graph neural network GNN and a graph embedding technology, the knowledge graph is dynamically updated to reflect the knowledge in new data in real time.

[0094] According to the knowledge graph dynamic updating method based on the growth process of crops provided by the application, a multi-level knowledge graph is constructed based on a graph database, including:

[0095] A basic layer of the multi-level knowledge graph is constructed based on crop species, environmental factors, climate conditions, soil types, crop growth states and farming operations; the basic layer is used to represent each entity and entity relationship based on a resource description framework;

[0096] A rule layer of a multi-level knowledge graph is constructed based on the logical rules and association patterns in the crop growth process. The rule layer is used to represent crop growth patterns, the relationship between the environment and crops, and the impact of farming operations.

[0097] A multi-level knowledge graph reasoning layer is constructed based on the basic layer and the rule layer, where the reasoning layer is used to generate crop growth decisions and agricultural operation recommendations.

[0098] In an embodiment of the present invention, the base layer (Data Layer) is the bottom layer of the knowledge graph, which contains basic information related to crop growth, including crop types, environmental factors (such as temperature, humidity, light, etc.), climatic conditions, soil types, crop growth status (such as leaf area index, plant height, etc.) and agricultural operations (seeding, irrigation, fertilization, etc.).

[0099] Representation: The base layer uses RDF (Resource Description Framework) to represent entities and their relationships. The RDF triple structure (subject-predicate-object) is used to describe crops and their related environmental factors. For example:

[0100] Example of a triple:

[0101] Corn - Suitable temperature -20°C

[0102] Wheat - Suitable light - 10 hours

[0103] Soil moisture-impact-crop growth rate

[0104] The basic layer data uses a unified namespace and standardized description method to ensure that data from different sources can be uniformly mapped to the knowledge graph, supporting upper-level rules and reasoning operations.

[0105] The Rule Layer defines the logical rules and association patterns of crop growth, ensuring the system can make reasonable judgments and inferences based on the dynamic changes in crop growth. The Rule Layer encompasses crop growth patterns, the relationship between the environment and crops, and the impact of agricultural operations.

[0106] The rule layer uses the SPARQL (SPARQL Protocol and RDF Query Language) query language to define various rules. SPARQL enables querying, manipulating, and reasoning based on the RDF data in the base layer. Rules can set specific actions and outcomes based on conditions (such as temperature ranges and duration of sunlight).

[0107] Example SPARQL rules:

[0108] sparql

[0109] SELECT ?crop WHERE {

[0110] ?croprdf:type:crop.

[0111] ?Crops: Suitable temperature?Temperature.

[0112] FILTER (?temperature>=15&&?temperature<=25)

[0113] }

[0114] This rule can be used to filter crop types suitable for current climate conditions based on temperature range.

[0115] The rule layer is closely connected to the base layer, extracting data from the base layer through SPARQL and making conditional judgments based on the set rules. It also provides data support for the reasoning layer as the basis for reasoning.

[0116] The inference layer is the decision-support layer of the knowledge graph. It derives new knowledge and decision recommendations based on the data in the base layer and the logic in the rule layer. The inference layer is primarily responsible for intelligent reasoning from existing data to generate crop growth decisions, agricultural operation recommendations, and more.

[0117] The reasoning layer uses OWL (Web Ontology Language) to represent knowledge and define reasoning logic. OWL-based reasoning can determine new knowledge relationships based on semantic features such as classes, attributes, and relationships.

[0118] At the same time, the reasoning layer integrates a rule engine (such as Drools), uses inference rules to perform reasoning analysis on the input data, and combines environmental changes, crop growth status and historical data to provide support for decision-making.

[0119] An example of the reasoning process:

[0120] Input: Current temperature is 22°C, humidity is 60%, and sunlight duration is 12 hours.

[0121] Reasoning rule: If the temperature is between 15-25°C, the humidity is between 50-70%, and the daylight hours exceed 10 hours, then it is recommended to increase irrigation to promote crop growth.

[0122] Output: The system outputs a recommendation based on the rule: "The current environment is suitable for irrigation. It is recommended to water 50 liters per mu of land."

[0123] The reasoning layer uses SPARQL queries to obtain the conditions provided by the rule layer, and then uses the rule engine to reason about the data to generate new decision recommendations and knowledge. The reasoning results are then updated to the basic layer and the rule layer, forming dynamic feedback.

[0124] The connection between the basic layer and the rule layer is that the basic layer provides data such as crops, environment, and agricultural operations, and the rule layer uses SPARQL to query this data and apply the set logical rules.

[0125] The connection between the rule layer and the reasoning layer is that the rules defined in the rule layer provide a basis for the reasoning layer to make judgments, and the reasoning layer makes intelligent decisions based on the information in the rule layer.

[0126] The connection between the inference layer and the base layer is that the output of the inference layer (such as agricultural operation suggestions and decision results) will be fed back to the base layer to form new triple data and further improve the knowledge graph.

[0127] The connection between the reasoning layer and the rule layer is that the output of the reasoning layer (such as agricultural operation suggestions and decision results) will be fed back to the basic layer to form new triple data and further improve the knowledge graph.

[0128] In this embodiment of the present invention, graph database technology (such as Neo4j) is used to create and update the knowledge graph in real time. Nodes include crop growth status, environmental factors, and management measures, and relationship edges represent the impact between different nodes.

[0129] Leveraging graph database technology, this data is converted into a graph structure, creating nodes (crop types, environmental factors, and management measures) and edges (representing the impact between different nodes). Based on this, new data inputs are monitored in real time, and through an incremental update mechanism, the nodes and relationships in the knowledge graph are dynamically adjusted to ensure that the knowledge graph always reflects the latest crop growth status and environmental changes.

[0130] According to the present invention, a method for dynamically updating a knowledge graph based on the crop growth process is provided, which incrementally updates a preset knowledge graph based on environmental factor data, crop growth indicators, and comprehensive characteristics, including:

[0131] Through graph neural network and knowledge graph embedding, the preset knowledge graph is incrementally updated based on environmental factor data, crop growth indicators and comprehensive characteristics to adjust the entity classification and entity relationship in the knowledge graph.

[0132] For example, we continuously collect new data on environmental factors, crop growth indicators, and comprehensive characteristic data. We preprocess this new data to ensure it aligns with the data structure in the knowledge graph. We use graph neural network models to analyze this new data, identifying new entity classifications and relationships and generating analysis results. Based on these analysis results, we dynamically adjust the entity classifications and relationships in the knowledge graph, performing incremental updates to incorporate new entities, attributes, and relationships into the knowledge graph. This ensures the efficiency and accuracy of the update process and avoids damage to the existing knowledge graph.

[0133] In this embodiment of the present invention, a graph neural network (GNN) and knowledge graph embedding technology (TransE) are used for node classification and relationship prediction. Combined with the rule-based reasoning engine Drools, knowledge reasoning and application are achieved.

[0134] Combining historical data with real-time information, Deep Q-Learning and decision tree algorithms are used to generate targeted agricultural management recommendations, such as optimal fertilization times and pest and disease control measures. Decision strategies are optimized based on feedback mechanisms.

[0135] By integrating historical data (such as crop growth records, fertilization, and pest control measures) with real-time information (such as current weather conditions and soil moisture), a multidimensional dataset is constructed. Next, the Deep Q-Learning algorithm, a reinforcement learning algorithm, is used to build an intelligent agent model. This model takes actions (such as fertilization, irrigation, or pest control) based on the environmental state (such as the current crop growth stage and environmental factors) and evaluates the effectiveness of these actions through a reward mechanism (such as increased crop yield or growth rate).

[0136] At the same time, the decision tree algorithm is combined with historical data analysis to generate targeted agricultural management recommendations. This process includes feature selection, data segmentation, and decision tree construction, ultimately forming clear decision rules to guide agricultural management strategies.

[0137] During implementation, a feedback mechanism is used to monitor the effectiveness of management measures and adjust decision-making strategies based on the results. Through continuous learning and optimization, the system can more accurately recommend optimal fertilization times, pest and disease control measures, and other measures to improve the efficiency and sustainability of agricultural production.

[0138] In some embodiments, the present invention provides a knowledge graph dynamic update system based on the crop growth process, which specifically includes the following modules.

[0139] Data acquisition module: includes sensor data acquisition, remote sensing data processing, edge computing processing, data preprocessing and storage sub-modules.

[0140] Dynamic knowledge expression module: includes dynamic model construction, deep fusion, and time series prediction sub-modules.

[0141] Knowledge graph construction and update module: includes graph construction, graph update, and knowledge management sub-modules.

[0142] Intelligent reasoning and decision support module: including reasoning engine, decision support system, and user interface module.

[0143] User interface module: provides visual display and user interaction functions, including data visualization, report generation, and real-time feedback sub-modules.

[0144] The above-mentioned embodiments of the present invention utilize a Transformer model and a multi-head self-attention mechanism to accurately capture changes in time series data during crop growth, thereby improving the accuracy of predicting crop states at each growth stage. The combination of GCN and GAT further enhances the ability to process the complex relationship between environmental factors and crop growth, optimizing adaptability to dynamic factors.

[0145] The deep fusion model, combined with CNN and RNN, efficiently integrates data from different sensors, ensuring data integrity and accuracy. Dynamic update technology, Apache Flink, and incremental learning enable real-time updates of models and knowledge graphs, ensuring the system can quickly adapt to dynamic changes in actual production.

[0146] By building and updating a dynamic knowledge graph using RDF and a graph database (Neo4j), the system accurately reflects the latest information on crop growth and environmental factors. Combining GNN with TransE, a knowledge graph embedding technology, and reinforcement learning, the system provides precise agricultural management recommendations and optimization strategies, enhancing the intelligence of decision support.

[0147] This invention achieves precise prediction and dynamic response capabilities for crop growth processes through a dynamic knowledge representation model and efficient data fusion and real-time update mechanisms. It accurately captures crop status at different growth stages, adapts promptly to environmental changes, and provides optimized agricultural management recommendations through intelligent reasoning. This not only improves the efficiency and accuracy of crop production, but also enhances the intelligence level of agricultural management, significantly improving the scientific nature and practicality of production decisions.

[0148] refer to Figure 2 , Figure 2 This is a flow chart of the dynamic knowledge expression and graph construction of the crop growth process provided by the present invention, which includes data collection, knowledge expression model, dynamic knowledge graph construction and intelligent reasoning.

[0149] Data collection includes: sensing devices, Ethernet, mobile Internet, Internet of Things and cloud storage.

[0150] The knowledge representation model includes: dynamic modeling of growth stages and the combination of time and space dimensions; input data includes varieties, physiological growth and environment; specific models include: spatial graph convolutional network (SGCN), spatial transformer network (STN), Transformer model, graph convolutional network (GCN) and graph attention network (GAT).

[0151] The construction of dynamic knowledge graphs includes: research on weakly supervised learning, dynamic update strategies, and internal update mechanisms; for example, crops grown in the field, crop diseases, disease characteristics, and corresponding (solution) methods; collaborative completion and association rule mining.

[0152] Intelligent reasoning includes: obtaining environmental factor data, such as weather, soil, diseases, and varieties, and making decisions based on reasoning algorithms through GNN and TransE to perform growth forecasts, pest and disease diagnosis, water and fertilizer irrigation, and yield estimation.

[0153] refer to Figure 3 , Figure 3 It is a schematic diagram of the time and space multi-dimensional feature extraction process provided by the present invention.

[0154] like Figure 3 As shown in the figure, environmental factor data is obtained, where the environmental factor data includes weather, soil, disease, and variety. Spatial features are extracted through the spatial graph convolutional network (SGCN), and then temporal features are extracted through the Transformer, and edge learning is performed to obtain multiple time series features and geographic spatial features to facilitate subsequent knowledge fusion.

[0155] The following describes the knowledge graph dynamic updating device based on the crop growth process provided by the present invention. The knowledge graph dynamic updating device based on the crop growth process described below and the knowledge graph dynamic updating method based on the crop growth process described above can be referenced to each other.

[0156] refer to Figure 4 , Figure 4 It is a structural diagram of the knowledge graph dynamic update device based on the crop growth process provided by the present invention, which includes an acquisition module 401, a Transformer module 402, a construction module 403, a feature extraction module 404, a transformation module 405, a fusion module 406 and an update module 407.

[0157] Acquisition module 401, for acquiring environmental factor data and crop growth stage;

[0158] Transformer module 402, for inputting environmental factor data and crop growth stages into the trained Transformer model, and obtaining crop growth indicators corresponding to the crop growth stages output by the trained Transformer model;

[0159] The Transformer module 402 is further used to input the crop growth index into the trained Transformer model to obtain the time series features output by the trained Transformer model;

[0160] A construction module 403 is configured to construct a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolutional network to obtain an environment-crop growth relationship graph;

[0161] A feature extraction module 404 is configured to extract features based on the environment-crop growth relationship graph using a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between environmental factor data and crop growth indicators;

[0162] The feature extraction module 404 is further configured to extract features based on the spatial feature matrix using a spatial graph convolutional network to obtain high-dimensional spatial features;

[0163] Transformation module 405, configured to align and transform high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features;

[0164] A fusion module 406 is configured to fuse features based on time series features and geographic spatial features using a deep fusion model to obtain comprehensive features;

[0165] The updating module 407 is used to incrementally update the preset knowledge graph based on environmental factor data, crop growth indicators and comprehensive characteristics.

[0166] Specifically, the above-mentioned knowledge graph dynamic update device based on the crop growth process provided by the present invention can implement all the method steps implemented by the above-mentioned knowledge graph dynamic update method embodiment based on the crop growth process, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as the method embodiment will not be described in detail here.

[0167] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 5As shown, the electronic device may include: a processor (processor) 510, a communication interface (Communications Interface) 520, a memory (memory) 530 and a communication bus 540, wherein the processor 510, the communication interface 520, and the memory 530 communicate with each other through the communication bus 540. The processor 510 can call the logic instructions in the memory 530 to execute a dynamic updating method for a knowledge graph based on the crop growth process, the method including: obtaining environmental factor data and crop growth stages; inputting the environmental factor data and crop growth stages into the trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages output by the trained Transformer model; inputting the crop growth indicators into the trained Transformer model to obtain time series features output by the trained Transformer model; constructing a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolution network to obtain an environment-crop growth relationship graph; extracting features based on the environment-crop growth relationship graph through a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between the environmental factor data and the crop growth indicators; extracting features based on the spatial feature matrix through a spatial graph convolution network to obtain high-dimensional spatial features; aligning and transforming the high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features; fusing features based on time series features and geographic spatial features through a deep fusion model to obtain comprehensive features; and incrementally updating a preset knowledge graph based on the environmental factor data, crop growth indicators, and comprehensive features.

[0168] Furthermore, the logic instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0169] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the knowledge graph dynamic update method based on the crop growth process provided by the above methods, which includes: obtaining environmental factor data and crop growth stage; inputting the environmental factor data and crop growth stage into the trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages output by the trained Transformer model; inputting the crop growth indicators into the trained Transformer model to obtain the trained Transformer model. Output time series features; construct a relationship graph based on environmental factor data and crop growth indicators through a graph convolutional network to obtain an environment-crop growth relationship graph; extract features based on the environment-crop growth relationship graph through a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between environmental factor data and crop growth indicators; extract features based on the spatial feature matrix through a spatial graph convolutional network to obtain high-dimensional spatial features; align and transform based on high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features; fuse features based on time series features and geographic spatial features through a deep fusion model to obtain comprehensive features; incrementally update the preset knowledge graph based on environmental factor data, crop growth indicators and comprehensive features.

[0170] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the knowledge graph dynamic updating method based on the crop growth process provided by the above methods, the method comprising: obtaining environmental factor data and crop growth stage; inputting the environmental factor data and crop growth stage into the trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages respectively output by the trained Transformer model; inputting the crop growth indicators into the trained Transformer model to obtain time series features output by the trained Transformer model; and The convolution network constructs a relationship graph based on environmental factor data and crop growth indicators to obtain an environment-crop growth relationship graph; the graph attention network is used to extract features based on the environment-crop growth relationship graph to obtain a spatial feature matrix, where the spatial feature matrix is ​​used to represent the correlation between environmental factor data and crop growth indicators; the spatial graph convolution network is used to extract features based on the spatial feature matrix to obtain high-dimensional spatial features; the spatial transformation network is used to align and transform the high-dimensional spatial features to obtain geographic spatial features; the deep fusion model is used to fuse features based on time series features and geographic spatial features to obtain comprehensive features; the preset knowledge graph is incrementally updated based on environmental factor data, crop growth indicators and comprehensive features.

[0171] The device embodiments described above are merely illustrative. 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, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0172] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for dynamically updating a knowledge graph based on the crop growth process, characterized in that: include: Obtain data on environmental factors and crop growth stages; Inputting the environmental factor data and the crop growth stage into a trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages, which are output by the trained Transformer model; Inputting the crop growth indicator into the trained Transformer model to obtain a time series feature output by the trained Transformer model; Constructing a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolutional network to obtain an environment-crop growth relationship graph; Extracting features based on the environment-crop growth relationship graph using a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between the environmental factor data and the crop growth indicator; Performing feature extraction based on the spatial feature matrix through a spatial graph convolutional network to obtain high-dimensional spatial features; Aligning and transforming the high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features; Performing feature fusion based on the time series features and the geographic space features through a deep fusion model to obtain comprehensive features; The preset knowledge graph is incrementally updated based on the environmental factor data, the crop growth indicators and the comprehensive characteristics.

2. The method for dynamically updating a knowledge graph based on the crop growth process according to claim 1, characterized in that: Inputting the environmental factor data and the crop growth stage into the trained Transformer model to obtain crop growth indicators corresponding to the crop growth stages output by the trained Transformer model, including: Inputting the environmental factor data and the crop growth stage into the trained Transformer model, wherein the environmental factor data includes temperature, humidity, and light; and the crop growth stage includes sowing, germination, and seedling stage; The trained Transformer model is used to predict crop growth status based on the environmental factor data and the crop growth stage, thereby obtaining crop growth indicators corresponding to the crop growth stages, wherein the crop growth indicators include leaf area index and plant height; and the crop growth indicators are used to represent the crop growth status.

3. The method for dynamically updating a knowledge graph based on the crop growth process according to claim 2, characterized in that: Inputting the crop growth indicator into the trained Transformer model to obtain the time series features output by the trained Transformer model includes: Through the multi-head self-attention mechanism of the trained Transformer model, predictions are made based on the crop growth indicators corresponding to the crop growth stages to obtain time series features, wherein the time series features are used to predict crop growth trends in the future.

4. The method for dynamically updating a knowledge graph based on the crop growth process according to claim 1, characterized in that: Before incrementally updating the preset knowledge graph based on the environmental factor data, the crop growth index, and the comprehensive features, the method further includes: A multi-level knowledge graph is constructed based on a graph database, wherein the multi-levels include a basic layer, a rule layer, and an inference layer.

5. The method for dynamically updating a knowledge graph based on the crop growth process according to claim 4, characterized in that: The multi-level knowledge graph constructed based on the graph database includes: A foundational layer of a multi-level knowledge graph is constructed based on crop types, environmental factors, climate conditions, soil types, crop growth status, and farming operations; the foundational layer is used to represent entities and entity relationships based on a resource description framework; Constructing a rule layer of the multi-level knowledge graph based on logical rules and association patterns in the crop growth process, wherein the rule layer is used to represent crop growth patterns, the relationship between the environment and crops, and the impact of farming operations; A reasoning layer of the multi-level knowledge graph is constructed based on the basic layer and the rule layer, wherein the reasoning layer is used to generate crop growth decisions and agricultural operation suggestions.

6. The method for dynamically updating a knowledge graph based on the crop growth process according to claim 1, characterized in that: The incremental updating of the preset knowledge graph based on the environmental factor data, the crop growth index, and the comprehensive characteristics includes: By embedding graph neural networks and knowledge graphs, the preset knowledge graph is incrementally updated based on the environmental factor data, the crop growth indicators and the comprehensive characteristics to adjust the entity classification and entity relationship in the knowledge graph.

7. A knowledge graph dynamic updating device based on the crop growth process, characterized in that: include: Acquisition module, used to obtain environmental factor data and crop growth stages; A Transformer module, configured to input the environmental factor data and the crop growth stage into a trained Transformer model, and obtain crop growth indicators corresponding to the crop growth stages, which are output by the trained Transformer model; The Transformer module is further configured to input the crop growth indicator into the trained Transformer model to obtain a time series feature output by the trained Transformer model; A construction module, configured to construct a relationship graph based on the environmental factor data and the crop growth indicators through a graph convolutional network to obtain an environment-crop growth relationship graph; a feature extraction module, configured to extract features based on the environment-crop growth relationship graph using a graph attention network to obtain a spatial feature matrix, wherein the spatial feature matrix is ​​used to represent the correlation between the environmental factor data and the crop growth index; The feature extraction module is further configured to extract features based on the spatial feature matrix through a spatial graph convolutional network to obtain high-dimensional spatial features; A transformation module, configured to align and transform the high-dimensional spatial features through a spatial transformation network to obtain geographic spatial features; A fusion module, configured to perform feature fusion based on the time series features and the geographic space features through a deep fusion model to obtain comprehensive features; An updating module is used to incrementally update a preset knowledge graph based on the environmental factor data, the crop growth indicators and the comprehensive features.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for dynamically updating the knowledge graph based on the crop growth process as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for dynamically updating a knowledge graph based on the crop growth process as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for dynamically updating a knowledge graph based on the crop growth process as claimed in any one of claims 1 to 6 is implemented.

Images