Leaf vegetable plant protection whole-process dynamic prevention and control method based on fusion knowledge graph

Abstract

The invention relates to the field of leaf vegetable plant protection, and discloses a leaf vegetable plant protection whole-process dynamic control method based on a fusion knowledge graph, and the method comprises the following steps: constructing a multi-source fusion knowledge graph for leaf vegetable plant protection, constructing a mathematical model for identifying leaf vegetable insect pests, and carrying out the dynamic control of the insect pests of leaf vegetables and the dynamic early warning of the leaf vegetable insect pests. According to the method, the application of the fusion knowledge graph can be combined, plant protection is turned from passive response to active prevention and control, sustainable development of agricultural production is achieved, the yield is guaranteed, and the nutritional quality of the leaf vegetables can be maintained.

Description

Technical Field

[0001] The present invention relates to the field of leafy vegetable plant protection, and in particular to a full-process dynamic prevention and control method for leafy vegetable plant protection based on a fusion knowledge graph. Background Art

[0002] Fusion knowledge graphs are a knowledge representation technology that integrates multi-source, heterogeneous data and constructs semantic associations. Its core is to extract, align, and fuse data from diverse sources and structures (such as text, images, sensor data, and expert experience) into a unified semantic network. This network, with entities (such as pests and diseases, pesticides, and environmental parameters) as nodes and relationships (such as "cause," "prevent," and "prone to") as edges, is capable of representing complex agricultural knowledge systems.

[0003] Applying a fused knowledge graph to leafy vegetable cultivation can increase diagnostic accuracy to over 90%. The advantages of full-process optimization include selecting appropriate quality crops based on their disease resistance profiles for pre-production prevention, and enabling in-production monitoring through real-time data from IoT devices triggering early warnings. Plant protection is particularly critical for leafy vegetables because they have a short growth cycle and a limited window for pest and disease outbreaks. Furthermore, their large leaf surface area makes them susceptible to pests. Finally, their delicate tissues are more sensitive to pesticide damage. Data shows that scientific plant protection not only ensures yield but also maintains the nutritional quality of leafy vegetables. The application of a fused knowledge graph shifts plant protection from reactive response to proactive prevention and control, achieving sustainable agricultural production. Therefore, a dynamic plant protection method for leafy vegetables based on a fused knowledge graph is proposed for the entire plant protection process. Summary of the Invention

[0004] The present invention overcomes the shortcomings of the existing technology and provides a dynamic prevention and control method for the entire process of leafy vegetable plant protection based on a fusion knowledge graph.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The first aspect of the present invention provides a full-process dynamic control method for leafy vegetable plant protection based on a fusion knowledge graph, comprising the following steps:

[0007] Construct a multi-source fusion knowledge graph on leafy vegetable plant protection, and shape different architectures of the multi-source fusion knowledge graph to obtain the target multi-source fusion knowledge graph;

[0008] Combined with the target multi-source fusion knowledge graph, a mathematical model for identifying leafy vegetable pests was constructed and calibrated as a leafy vegetable pest identification mathematical model;

[0009] Combined with the mathematical model of leafy vegetable pest identification, a dynamic decision-making model for leafy vegetable pest prevention and control is constructed, and combined with the dynamic decision-making model for leafy vegetable pests, dynamic early warning of leafy vegetable pests is carried out.

[0010] Furthermore, in a preferred embodiment of the present invention, the multi-source fusion knowledge graph for leafy vegetable plant protection is constructed, and different architectures of the multi-source fusion knowledge graph are structured to obtain a target multi-source fusion knowledge graph, specifically:

[0011] Collect and obtain the full-process related data of leafy vegetable plant protection, wherein the full-process related data of leafy vegetable plant protection includes structured data and unstructured data;

[0012] Introducing a historical data network, obtaining an agricultural meteorological database in the historical data network, and determining a leafy vegetable plant protection area, retrieving historical environmental parameters of the leafy vegetable plant protection area based on the agricultural meteorological database, and deploying sensors in the leafy vegetable plant protection area;

[0013] Among them, the sensors deployed in the leafy vegetable plant protection area are used to collect the EC value and pH value of the soil in the leafy vegetable plant protection area;

[0014] The historical environmental parameters of the leafy vegetable plant protection area and the EC value and pH value of the soil in the leafy vegetable plant protection area collected by the sensor are organized into institutional data and calibrated as a type of leafy vegetable plant protection data;

[0015] Identify leafy vegetable species, retrieve relevant pest and disease papers based on historical data networks, and perform unstructured data mining and extraction on relevant pest and disease papers.

[0016] The unstructured data in the papers on pests and diseases corresponding to leafy vegetable species include farmers' plant protection question and answer records for leafy vegetable species and text describing the symptoms of leafy vegetable species. The unstructured data are categorized as Class II leafy vegetable plant protection data.

[0017] Performing structural cleaning and unified analysis on the first-category leafy vegetable plant protection data and the second-category leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection data;

[0018] Combined with the pre-processed leafy vegetable plant protection data, a multi-source fusion knowledge graph about leafy vegetable plant protection is constructed and calibrated as the target multi-source fusion knowledge graph.

[0019] Furthermore, in a preferred embodiment of the present invention, the pre-processed leafy vegetable plant protection data is combined to construct a multi-source fusion knowledge graph about leafy vegetable plant protection, which is labeled as a target multi-source fusion knowledge graph, specifically:

[0020] A data ontology library is preset, and the pre-processed leafy vegetable plant protection data is imported into the data ontology library for storage, and the data ontology library is expanded with spatiotemporal attributes, wherein the spatiotemporal attribute expansion is to combine the pre-processed leafy vegetable plant protection data in the data ontology library with additional disease nodes and prevention and control plan nodes;

[0021] In the data ontology library, free data combination is performed on the pre-processed leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection sub-data combinations, a semantic similarity calculation algorithm is introduced, and based on the semantic similarity calculation algorithm, semantic similarity calculation is performed on all pre-processed leafy vegetable plant protection sub-data combinations, and pre-processed leafy vegetable plant protection sub-data combinations with semantic similarity greater than a preset value are marked as similar data combinations;

[0022] Constructing a loss function for performing knowledge fusion on the pre-processed leafy vegetable plant protection data within the similar data combination to obtain a fused leafy vegetable plant protection sub-data combination, and updating the pre-processed leafy vegetable plant protection data based on the fused leafy vegetable plant protection sub-data combination to obtain updated leafy vegetable plant protection data;

[0023] In the data ontology library, the spatiotemporal constraint reasoning rules are determined. The spatiotemporal constraint reasoning rules are as follows: probability of occurrence of pests and diseases = pathogen survival rate × meteorological suitability × crop sensitivity. Based on the spatiotemporal constraint reasoning rules, a query index for updating leafy vegetable plant protection data is established to realize the construction of a multi-source fusion knowledge graph about leafy vegetable plant protection, which is calibrated as the target multi-source fusion knowledge graph.

[0024] Furthermore, in a preferred embodiment of the present invention, the target multi-source fusion knowledge graph is combined to construct a mathematical model for identifying leafy vegetable pests, which is calibrated as a leafy vegetable pest identification mathematical model, specifically:

[0025] Within the target multi-source fusion knowledge graph, we extract the seasonal occurrence probability, geographical distribution, historical environmental parameters of leafy vegetable plant protection areas, and image feature vectors of symptom characteristics of pests and diseases, and introduce an LSTM algorithm model.

[0026] Import all extracted image feature vectors into the LSTM algorithm model for graph neural modeling, implement updated graph neural boundary of the LSTM algorithm model, and obtain an updated LSTM algorithm model;

[0027] A three-way cross-attention network is designed within the updated LSTM algorithm model to achieve deep interactive fusion of different image feature vectors. At the same time, the updated LSTM algorithm model is combined with the target multi-source fusion knowledge graph and the Bayesian algorithm to construct a causal chain for leafy vegetable protection, thereby obtaining a secondary updated LSTM algorithm model.

[0028] The secondary updated LSTM algorithm model is adaptively trained, wherein the adaptive training is adaptive training combined with an adaptive loss function, and a leafy vegetable pest recognition mathematical model is obtained after the adaptive training.

[0029] Furthermore, in a preferred embodiment of the present invention, the leafy vegetable pest identification mathematical model is combined to construct a leafy vegetable pest dynamic prevention and control decision model, and the leafy vegetable pest dynamic decision model is combined to perform a leafy vegetable pest dynamic early warning, specifically:

[0030] The EC value and pH value of the soil in the leafy vegetable plant protection area are analyzed by using a leafy vegetable pest identification mathematical model. At the same time, leafy vegetable samples are collected in the leafy vegetable plant protection area to obtain sample characteristics of the collected leafy vegetable samples. The sample characteristics of the collected leafy vegetable samples are simultaneously imported into the leafy vegetable pest identification mathematical model for analysis;

[0031] Based on the leafy vegetable pest identification mathematical model, the leafy vegetable pest type in the current leafy vegetable plant protection area is output and calibrated as the real-time pest type;

[0032] Construct a dynamic leafy vegetable pest control decision model, connect it to the leafy vegetable pest identification mathematical model, and retrieve all corresponding leafy vegetable control decisions under different real-time pest and disease types based on the historical data network;

[0033] Import all prevention and control decisions corresponding to different real-time pest and disease types of leafy vegetables into the leafy vegetable pest dynamic prevention and control decision model, ensuring that the leafy vegetable pest dynamic prevention and control decision model directly outputs the corresponding prevention and control decision after determining the real-time pest and disease type through the leafy vegetable pest identification mathematical model;

[0034] Combining the dynamic decision-making model of leafy vegetable pests and the mathematical model of leafy vegetable pest identification, dynamic early warning of leafy vegetable pests is carried out in the leafy vegetable plant protection area.

[0035] Furthermore, in a preferred embodiment of the present invention, the leafy vegetable pest dynamic decision-making model and the leafy vegetable pest identification mathematical model are combined to perform a dynamic early warning of leafy vegetable pests in the leafy vegetable plant protection area, specifically:

[0036] Introducing a buzzer alarm into the leafy vegetable protection area and modularly connecting the buzzer alarm with a leafy vegetable pest identification mathematical model;

[0037] The leafy vegetable pest identification mathematical model is used to determine in real time whether leafy vegetables in the leafy vegetable protection area are affected by pests and diseases. In combination with the historical data network, the impact range of the pests and diseases can be determined based on the real-time pest and disease type.

[0038] A standard pest and disease impact range is preset. If pests and diseases exist in leafy vegetables in the leafy vegetable protection area, and the pest and disease impact range of the real-time pest and disease type is greater than the standard pest and disease impact range, the buzzer alarm will be controlled to sound to ensure that the staff in the leafy vegetable protection area are aware of it. At the same time, the corresponding prevention and control decision will be determined through the dynamic decision-making model of leafy vegetable pests.

[0039] A second aspect of the present invention further provides a leafy vegetable plant protection full-process dynamic prevention and control system based on a fusion knowledge graph, the leafy vegetable plant protection full-process dynamic prevention and control system comprising a memory and a processor, the memory storing a leafy vegetable plant protection full-process dynamic prevention and control method, and the leafy vegetable plant protection full-process dynamic prevention and control method, when executed by the processor, implementing the following steps:

[0040] Construct a multi-source fusion knowledge graph on leafy vegetable plant protection, and shape different architectures of the multi-source fusion knowledge graph to obtain the target multi-source fusion knowledge graph;

[0041] Combined with the target multi-source fusion knowledge graph, a mathematical model for identifying leafy vegetable pests was constructed and calibrated as a leafy vegetable pest identification mathematical model;

[0042] Combined with the mathematical model of leafy vegetable pest identification, a dynamic decision-making model for leafy vegetable pest prevention and control is constructed, and combined with the dynamic decision-making model for leafy vegetable pests, dynamic early warning of leafy vegetable pests is carried out.

[0043] This invention addresses the technical deficiencies in the background art and has the following beneficial effects: constructing a multi-source fusion knowledge graph for leafy vegetable plant protection and a mathematical model for identifying leafy vegetable pests, enabling dynamic pest control and early warning for leafy vegetable pests. By combining the application of the fusion knowledge graph, this invention can shift plant protection from passive response to active prevention and control, achieving sustainable agricultural production, ensuring both yield and nutritional quality of leafy vegetables. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying any creative work, they can also obtain drawings of other embodiments based on these drawings.

[0045] Figure 1 A flow chart of a dynamic control method for the entire leafy vegetable plant protection process based on the fusion knowledge graph is shown;

[0046] Figure 2 A flow chart of a method for constructing a fusion knowledge graph is shown;

[0047] Figure 3 The program view of the full-process dynamic control system for leafy vegetable plant protection based on the fusion knowledge graph is shown. DETAILED DESCRIPTION

[0048] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0049] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0050] Figure 1 The flowchart of the dynamic control method for leafy vegetable plant protection based on the fusion knowledge graph is shown, which includes the following steps:

[0051] S102: constructing a multi-source fusion knowledge graph on leafy vegetable plant protection, and shaping different architectures of the multi-source fusion knowledge graph to obtain a target multi-source fusion knowledge graph;

[0052] S104: combining the target multi-source fusion knowledge graph, constructing a mathematical model for identifying leafy vegetable pests, and calibrating it as a leafy vegetable pest identification mathematical model;

[0053] S106: Combined with the leafy vegetable pest identification mathematical model, a leafy vegetable pest dynamic prevention and control decision model is constructed, and combined with the leafy vegetable pest dynamic decision model, a leafy vegetable pest dynamic early warning is performed.

[0054] Furthermore, in a preferred embodiment of the present invention, the target multi-source fusion knowledge graph is combined to construct a mathematical model for identifying leafy vegetable pests, which is calibrated as a leafy vegetable pest identification mathematical model, specifically:

[0055] Within the target multi-source fusion knowledge graph, we extract the seasonal occurrence probability, geographical distribution, historical environmental parameters of leafy vegetable plant protection areas, and image feature vectors of symptom characteristics of pests and diseases, and introduce an LSTM algorithm model.

[0056] Import all extracted image feature vectors into the LSTM algorithm model for graph neural modeling, implement updated graph neural boundary of the LSTM algorithm model, and obtain an updated LSTM algorithm model;

[0057] A three-way cross-attention network is designed within the updated LSTM algorithm model to achieve deep interactive fusion of different image feature vectors. At the same time, the updated LSTM algorithm model is combined with the target multi-source fusion knowledge graph and the Bayesian algorithm to construct a causal chain for leafy vegetable protection, thereby obtaining a secondary updated LSTM algorithm model.

[0058] The secondary updated LSTM algorithm model is adaptively trained, wherein the adaptive training is adaptive training combined with an adaptive loss function, and a leafy vegetable pest recognition mathematical model is obtained after the adaptive training.

[0059] It should be noted that the LSTM algorithm model is a predictive algorithm. First, a spatiotemporal feature expression system is established through the deep coupling of knowledge graphs and LSTMs. This system extracts structured attributes such as seasonal patterns and regional characteristics of pests and diseases from a multi-source fusion knowledge graph. Combined with image feature vectors, the LSTM gating structure is improved using a graph attention mechanism, enabling the model to dynamically perceive the impact of environmental evolution on pests and diseases. Second, a three-way cross-attention fusion module is designed to achieve deep interaction between visual features, environmental temporal features, and knowledge semantic features. This structure effectively addresses the semantic gap problem caused by the simple splicing of multimodal features in traditional methods. Finally, an interpretable reasoning framework is constructed through a Bayesian causal chain, and dynamic optimization of model parameters is achieved by combining an adaptive loss function.

[0060] Furthermore, in a preferred embodiment of the present invention, the leafy vegetable pest identification mathematical model is combined to construct a leafy vegetable pest dynamic prevention and control decision model, and the leafy vegetable pest dynamic decision model is combined to perform a leafy vegetable pest dynamic early warning, specifically:

[0061] The EC value and pH value of the soil in the leafy vegetable plant protection area are analyzed by using a leafy vegetable pest identification mathematical model. At the same time, leafy vegetable samples are collected in the leafy vegetable plant protection area to obtain sample characteristics of the collected leafy vegetable samples. The sample characteristics of the collected leafy vegetable samples are simultaneously imported into the leafy vegetable pest identification mathematical model for analysis;

[0062] Based on the leafy vegetable pest identification mathematical model, the leafy vegetable pest type in the current leafy vegetable plant protection area is output and calibrated as the real-time pest type;

[0063] Construct a dynamic leafy vegetable pest control decision model, connect it to the leafy vegetable pest identification mathematical model, and retrieve all corresponding leafy vegetable control decisions under different real-time pest and disease types based on the historical data network;

[0064] Import all prevention and control decisions corresponding to different real-time pest and disease types of leafy vegetables into the leafy vegetable pest dynamic prevention and control decision model, ensuring that the leafy vegetable pest dynamic prevention and control decision model directly outputs the corresponding prevention and control decision after determining the real-time pest and disease type through the leafy vegetable pest identification mathematical model;

[0065] Combining the dynamic decision-making model of leafy vegetable pests and the mathematical model of leafy vegetable pest identification, dynamic early warning of leafy vegetable pests is carried out in the leafy vegetable plant protection area.

[0066] It should be noted that the claim constructs a closed-loop intelligent prevention and control system for leafy vegetable pests and diseases, which is used to analyze whether leafy vegetables have pests and diseases, and to provide pest and disease prevention and control plans based on the types of pests and diseases. Leafy vegetable samples are collected within the leafy vegetable plant protection area and the sample characteristics are imported into the leafy vegetable pest identification mathematical model for analysis, with the aim of determining the type of pests and diseases. Different types of pests and diseases correspond to different prevention and control plans, and all prevention and control decisions corresponding to leafy vegetables under different real-time pest and disease types can be retrieved in the historical data network. A dynamic leafy vegetable pest prevention and control decision model is constructed, which can directly output corresponding prevention and control decisions based on the real-time pest and disease types of leafy vegetables. Compared with traditional methods, the prevention and control response speed is improved. Its innovation lies in the real-time coupled analysis of multi-dimensional data of "soil-crop-environment-pathogens", which enables plant protection decisions to shift from passive response to active prevention and control.

[0067] Furthermore, in a preferred embodiment of the present invention, the leafy vegetable pest dynamic decision-making model and the leafy vegetable pest identification mathematical model are combined to perform a dynamic early warning of leafy vegetable pests in the leafy vegetable plant protection area, specifically:

[0068] Introducing a buzzer alarm into the leafy vegetable protection area and modularly connecting the buzzer alarm with a leafy vegetable pest identification mathematical model;

[0069] The leafy vegetable pest identification mathematical model is used to determine in real time whether leafy vegetables in the leafy vegetable protection area are affected by pests and diseases. In combination with the historical data network, the impact range of the pests and diseases can be determined based on the real-time pest and disease type.

[0070] A standard pest and disease impact range is preset. If pests and diseases exist in leafy vegetables in the leafy vegetable protection area, and the pest and disease impact range of the real-time pest and disease type is greater than the standard pest and disease impact range, the buzzer alarm will be controlled to sound to ensure that the staff in the leafy vegetable protection area are aware of it. At the same time, the corresponding prevention and control decision will be determined through the dynamic decision-making model of leafy vegetable pests.

[0071] It's important to note that pests and diseases can threaten the health of leafy vegetables in the protected area. Therefore, if a large-scale pest and disease is detected within the protected area, staff must be alerted to take timely action. However, the presence of pests and diseases on leafy vegetables is normal, and if the impact is small, they can be ignored for now. The buzzer alarm provides a timely and clear alert to staff in the protected area.

[0072] Figure 2 The flowchart of the method for constructing a fusion knowledge graph is shown, which includes the following steps:

[0073] S202: constructing a multi-source fusion knowledge graph on leafy vegetable plant protection, and shaping different architectures of the multi-source fusion knowledge graph to obtain a target multi-source fusion knowledge graph;

[0074] S204: Combine the pre-processed leafy vegetable plant protection data to construct a multi-source fusion knowledge graph about leafy vegetable plant protection, and mark it as a target multi-source fusion knowledge graph.

[0075] Furthermore, in a preferred embodiment of the present invention, the multi-source fusion knowledge graph for leafy vegetable plant protection is constructed, and different architectures of the multi-source fusion knowledge graph are structured to obtain a target multi-source fusion knowledge graph, specifically:

[0076] Collect and obtain the full-process related data of leafy vegetable plant protection, wherein the full-process related data of leafy vegetable plant protection includes structured data and unstructured data;

[0077] Introducing a historical data network, obtaining an agricultural meteorological database in the historical data network, and determining a leafy vegetable plant protection area, retrieving historical environmental parameters of the leafy vegetable plant protection area based on the agricultural meteorological database, and deploying sensors in the leafy vegetable plant protection area;

[0078] Among them, the sensors deployed in the leafy vegetable plant protection area are used to collect the EC value and pH value of the soil in the leafy vegetable plant protection area;

[0079] The historical environmental parameters of the leafy vegetable plant protection area and the EC value and pH value of the soil in the leafy vegetable plant protection area collected by the sensor are organized into institutional data and calibrated as a type of leafy vegetable plant protection data;

[0080] Identify leafy vegetable species, retrieve relevant pest and disease papers based on historical data networks, and perform unstructured data mining and extraction on relevant pest and disease papers.

[0081] The unstructured data in the papers on pests and diseases corresponding to leafy vegetable species include farmers' plant protection question and answer records for leafy vegetable species and text describing the symptoms of leafy vegetable species. The unstructured data are categorized as Class II leafy vegetable plant protection data.

[0082] Performing structural cleaning and unified analysis on the first-category leafy vegetable plant protection data and the second-category leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection data;

[0083] Combined with the pre-processed leafy vegetable plant protection data, a multi-source fusion knowledge graph about leafy vegetable plant protection is constructed and calibrated as the target multi-source fusion knowledge graph.

[0084] It should be noted that this article constructs a complete technical framework for data collection and processing throughout the entire leafy vegetable plant protection process. The core of this framework is to provide a high-quality data base for subsequent intelligent decision-making through the systematic integration and governance of multi-source heterogeneous data. Specifically: First, through the two-way linkage of the historical meteorological database and the real-time sensor network, a spatiotemporal continuous environmental monitoring system is formed, namely, historical environmental parameters, which are used to reveal the climate patterns of the region, while high-precision sensors deployed in the field capture key parameters such as soil conductivity and pH in real time. The two together constitute a structured type of data, laying the foundation for establishing a quantitative relationship model between the occurrence of pests and diseases and environmental factors, that is, collecting the EC value and pH value of the soil in the leafy vegetable plant protection area. Unstructured data mining and extraction are carried out on papers on pests and diseases corresponding to leafy vegetable species, with the aim of filling the cognitive gap between laboratory research and field practice; finally, standardized conversion and semantic alignment of multi-source data are achieved through a data cleaning engine, and cross-modal correlation features are constructed with the help of feature fusion algorithms (such as deep tensor networks), and the output is a preprocessed data set that can be directly used for knowledge graph construction and machine learning model training, that is, preprocessed leafy vegetable plant protection data, which is used to construct a fused knowledge graph.

[0085] Furthermore, in a preferred embodiment of the present invention, the pre-processed leafy vegetable plant protection data is combined to construct a multi-source fusion knowledge graph about leafy vegetable plant protection, which is labeled as a target multi-source fusion knowledge graph, specifically:

[0086] A data ontology library is preset, and the pre-processed leafy vegetable plant protection data is imported into the data ontology library for storage, and the data ontology library is expanded with spatiotemporal attributes, wherein the spatiotemporal attribute expansion is to combine the pre-processed leafy vegetable plant protection data in the data ontology library with additional disease nodes and prevention and control plan nodes;

[0087] In the data ontology library, free data combination is performed on the pre-processed leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection sub-data combinations, a semantic similarity calculation algorithm is introduced, and based on the semantic similarity calculation algorithm, semantic similarity calculation is performed on all pre-processed leafy vegetable plant protection sub-data combinations, and pre-processed leafy vegetable plant protection sub-data combinations with semantic similarity greater than a preset value are marked as similar data combinations;

[0088] Constructing a loss function for performing knowledge fusion on the pre-processed leafy vegetable plant protection data within the similar data combination to obtain a fused leafy vegetable plant protection sub-data combination, and updating the pre-processed leafy vegetable plant protection data based on the fused leafy vegetable plant protection sub-data combination to obtain updated leafy vegetable plant protection data;

[0089] In the data ontology library, the spatiotemporal constraint reasoning rules are determined. The spatiotemporal constraint reasoning rules are as follows: probability of occurrence of pests and diseases = pathogen survival rate × meteorological suitability × crop sensitivity. Based on the spatiotemporal constraint reasoning rules, a query index for updating leafy vegetable plant protection data is established to realize the construction of a multi-source fusion knowledge graph about leafy vegetable plant protection, which is calibrated as the target multi-source fusion knowledge graph.

[0090] It should be noted that the data ontology library is the prototype of the fusion knowledge graph. In the data ontology library, it is necessary to combine the leafy vegetable plant protection data to construct the database parameters and update the architecture to achieve the purpose of building a multi-source fusion knowledge graph. The purpose of adding disease nodes and prevention and control plan nodes is that, because the main purpose of the fusion knowledge graph is to dynamically prevent and control the entire process of leafy vegetable plant protection, it is necessary to know the types of diseases existing in leafy vegetables and the types of prevention and control plans. Therefore, it is necessary to determine the disease nodes and prevention and control plan nodes in the fusion knowledge graph to achieve the purpose of focused analysis. There may be semantic similarities between different data in the pre-processed leafy vegetable plant protection sub-data combination. For example, the meanings of words such as "yellowing" and "turning yellow" are similar. It is necessary to fuse words with similar meanings to improve the efficiency of the fusion knowledge graph during the analysis process and ensure accuracy. Therefore, in combination with the semantic similarity calculation algorithm, semantic similarity calculation is performed on all pre-processed leafy vegetable plant protection sub-data combinations to determine similar data combinations. After data fusion, a loss function is constructed. The loss function is to calculate the mean square error between the data to obtain updated leafy vegetable plant protection data. The spatiotemporal constraint reasoning rule is: probability of pest and disease occurrence = pathogen survival rate × meteorological suitability × crop sensitivity. The purpose of proposing this spatiotemporal constraint reasoning rule is to calculate the probability of pest and disease occurrence based on the data input into the fused knowledge graph, thereby achieving the goal of formulating analysis of leafy vegetable plant protection using the fused knowledge graph. A query index for leafy vegetable plant protection data is established and updated, enabling the construction of a multi-source fused knowledge graph for leafy vegetable plant protection, which is then labeled as the target multi-source fused knowledge graph. This means that issues encountered during leafy vegetable quality assurance can be analyzed using the spatiotemporal constraint reasoning rule.

[0091] like Figure 3 As shown, the second aspect of the present invention further provides a leafy vegetable plant protection full-process dynamic prevention and control system based on a fusion knowledge graph, the leafy vegetable plant protection full-process dynamic prevention and control system includes a memory 31 and a processor 32, the memory 31 stores a leafy vegetable plant protection full-process dynamic prevention and control method, and when the leafy vegetable plant protection full-process dynamic prevention and control method is executed by the processor 32, the following steps are implemented:

[0092] Construct a multi-source fusion knowledge graph on leafy vegetable plant protection, and shape different architectures of the multi-source fusion knowledge graph to obtain the target multi-source fusion knowledge graph;

[0093] Combined with the target multi-source fusion knowledge graph, a mathematical model for identifying leafy vegetable pests was constructed and calibrated as a leafy vegetable pest identification mathematical model;

[0094] Combined with the mathematical model of leafy vegetable pest identification, a dynamic decision-making model for leafy vegetable pest prevention and control is constructed, and combined with the dynamic decision-making model for leafy vegetable pests, dynamic early warning of leafy vegetable pests is carried out.

[0095] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A dynamic control method for the entire process of leafy vegetable plant protection based on the fusion knowledge graph, characterized in that: The following steps are involved: Construct a multi-source fusion knowledge graph on leafy vegetable plant protection, and shape different architectures of the multi-source fusion knowledge graph to obtain the target multi-source fusion knowledge graph; Combined with the target multi-source fusion knowledge graph, a mathematical model for identifying leafy vegetable pests was constructed and calibrated as a leafy vegetable pest identification mathematical model; Combined with the mathematical model of leafy vegetable pest identification, a dynamic decision-making model for leafy vegetable pest prevention and control is constructed, and combined with the dynamic decision-making model for leafy vegetable pests, dynamic early warning of leafy vegetable pests is carried out.

2. The full-process dynamic control method for leafy vegetable plant protection based on the fusion knowledge graph according to claim 1 is characterized in that: The multi-source fusion knowledge graph for leafy vegetable plant protection is constructed, and different architectures of the multi-source fusion knowledge graph are structured to obtain a target multi-source fusion knowledge graph, specifically: Collect and obtain the full-process related data of leafy vegetable plant protection, wherein the full-process related data of leafy vegetable plant protection includes structured data and unstructured data; Introducing a historical data network, obtaining an agricultural meteorological database in the historical data network, and determining a leafy vegetable plant protection area, retrieving historical environmental parameters of the leafy vegetable plant protection area based on the agricultural meteorological database, and deploying sensors in the leafy vegetable plant protection area; Among them, the sensors deployed in the leafy vegetable plant protection area are used to collect the EC value and pH value of the soil in the leafy vegetable plant protection area; The historical environmental parameters of the leafy vegetable plant protection area and the EC value and pH value of the soil in the leafy vegetable plant protection area collected by the sensor are organized into institutional data and calibrated as a type of leafy vegetable plant protection data; Identify leafy vegetable species, retrieve relevant pest and disease papers based on historical data networks, and perform unstructured data mining and extraction on relevant pest and disease papers. The unstructured data in the papers on pests and diseases corresponding to leafy vegetable species include farmers' plant protection question and answer records for leafy vegetable species and text describing the symptoms of leafy vegetable species. The unstructured data are categorized as Class II leafy vegetable plant protection data. Performing structural cleaning and unified analysis on the first-category leafy vegetable plant protection data and the second-category leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection data; Combined with the pre-processed leafy vegetable plant protection data, a multi-source fusion knowledge graph about leafy vegetable plant protection is constructed and calibrated as the target multi-source fusion knowledge graph.

3. The full-process dynamic control method for leafy vegetable plant protection based on fusion knowledge graph according to claim 2 is characterized in that: The pre-processed leafy vegetable plant protection data is combined to construct a multi-source fusion knowledge graph on leafy vegetable plant protection, which is calibrated as the target multi-source fusion knowledge graph, specifically: A data ontology library is preset, and the pre-processed leafy vegetable plant protection data is imported into the data ontology library for storage, and the data ontology library is expanded with spatiotemporal attributes, wherein the spatiotemporal attribute expansion is to combine the pre-processed leafy vegetable plant protection data in the data ontology library with additional disease nodes and prevention and control plan nodes; In the data ontology library, free data combination is performed on the pre-processed leafy vegetable plant protection data to obtain pre-processed leafy vegetable plant protection sub-data combinations, a semantic similarity calculation algorithm is introduced, and based on the semantic similarity calculation algorithm, semantic similarity calculation is performed on all pre-processed leafy vegetable plant protection sub-data combinations, and pre-processed leafy vegetable plant protection sub-data combinations with semantic similarity greater than a preset value are marked as similar data combinations; Constructing a loss function for performing knowledge fusion on the pre-processed leafy vegetable plant protection data within the similar data combination to obtain a fused leafy vegetable plant protection sub-data combination, and updating the pre-processed leafy vegetable plant protection data based on the fused leafy vegetable plant protection sub-data combination to obtain updated leafy vegetable plant protection data; In the data ontology library, the spatiotemporal constraint reasoning rules are determined. The spatiotemporal constraint reasoning rules are as follows: probability of occurrence of pests and diseases = pathogen survival rate × meteorological suitability × crop sensitivity. Based on the spatiotemporal constraint reasoning rules, a query index for updating leafy vegetable plant protection data is established to realize the construction of a multi-source fusion knowledge graph about leafy vegetable plant protection, which is calibrated as the target multi-source fusion knowledge graph.

4. The full-process dynamic control method for leafy vegetable plant protection based on fusion knowledge graph according to claim 1 is characterized in that: The target multi-source fusion knowledge graph is combined to construct a mathematical model for identifying leafy vegetable pests, which is calibrated as a leafy vegetable pest identification mathematical model. Specifically, Within the target multi-source fusion knowledge graph, we extract the seasonal occurrence probability, geographical distribution, historical environmental parameters of leafy vegetable plant protection areas, and image feature vectors of symptom characteristics of pests and diseases, and introduce an LSTM algorithm model. Import all extracted image feature vectors into the LSTM algorithm model for graph neural modeling, implement updated graph neural boundary of the LSTM algorithm model, and obtain an updated LSTM algorithm model; A three-way cross-attention network is designed within the updated LSTM algorithm model to achieve deep interactive fusion of different image feature vectors. At the same time, the updated LSTM algorithm model is combined with the target multi-source fusion knowledge graph and the Bayesian algorithm to construct a causal chain for leafy vegetable protection, thereby obtaining a secondary updated LSTM algorithm model. The secondary updated LSTM algorithm model is adaptively trained, wherein the adaptive training is adaptive training combined with an adaptive loss function, and a leafy vegetable pest recognition mathematical model is obtained after the adaptive training.

5. The full-process dynamic control method for leafy vegetable plant protection based on fusion knowledge graph according to claim 1 is characterized in that: The leafy vegetable pest identification mathematical model is combined to construct a leafy vegetable pest dynamic prevention and control decision model, and the leafy vegetable pest dynamic decision model is combined to perform a leafy vegetable pest dynamic early warning, specifically: The EC value and pH value of the soil in the leafy vegetable plant protection area are analyzed by using a leafy vegetable pest identification mathematical model. At the same time, leafy vegetable samples are collected in the leafy vegetable plant protection area to obtain sample characteristics of the collected leafy vegetable samples. The sample characteristics of the collected leafy vegetable samples are simultaneously imported into the leafy vegetable pest identification mathematical model for analysis; Based on the leafy vegetable pest identification mathematical model, the leafy vegetable pest type in the current leafy vegetable plant protection area is output and calibrated as the real-time pest type; Construct a dynamic leafy vegetable pest control decision model, connect it to the leafy vegetable pest identification mathematical model, and retrieve all corresponding leafy vegetable control decisions under different real-time pest and disease types based on the historical data network; Import all prevention and control decisions corresponding to different real-time pest and disease types of leafy vegetables into the leafy vegetable pest dynamic prevention and control decision model, ensuring that the leafy vegetable pest dynamic prevention and control decision model directly outputs the corresponding prevention and control decision after determining the real-time pest and disease type through the leafy vegetable pest identification mathematical model; Combining the dynamic decision-making model of leafy vegetable pests and the mathematical model of leafy vegetable pest identification, dynamic early warning of leafy vegetable pests is carried out in the leafy vegetable plant protection area.

6. The full-process dynamic control method for leafy vegetable plant protection based on fusion knowledge graph according to claim 5 is characterized in that: The leafy vegetable pest dynamic decision-making model and the leafy vegetable pest identification mathematical model are combined to provide a dynamic early warning of leafy vegetable pests in the leafy vegetable plant protection area, specifically: Introducing a buzzer alarm into the leafy vegetable protection area and modularly connecting the buzzer alarm to a leafy vegetable pest identification mathematical model; The leafy vegetable pest identification mathematical model is used to determine in real time whether leafy vegetables in the leafy vegetable protection area are affected by pests and diseases. In combination with the historical data network, the impact range of the pests and diseases can be determined based on the real-time pest and disease type. A standard pest and disease impact range is preset. If pests and diseases exist in leafy vegetables in the leafy vegetable protection area, and the pest and disease impact range of the real-time pest and disease type is greater than the standard pest and disease impact range, the buzzer alarm will be controlled to sound to ensure that the staff in the leafy vegetable protection area are aware of it. At the same time, the corresponding prevention and control decision will be determined through the dynamic decision-making model of leafy vegetable pests.

7. A dynamic prevention and control system for leafy vegetable plant protection based on the fusion knowledge graph, characterized by: The full-process dynamic prevention and control system for leafy vegetable plant protection includes a memory and a processor. The memory stores a full-process dynamic prevention and control method program for leafy vegetable plant protection. When the full-process dynamic prevention and control method program for leafy vegetable plant protection is executed by the processor, the full-process dynamic prevention and control method steps for leafy vegetable plant protection as described in any one of claims 1-6 are implemented.

Images