Vegetable growth cycle pest control decision system and method based on knowledge graph

Abstract

The application relates to the field of agricultural informatization and intelligent decision-making technology, and discloses a vegetable growth cycle disease and pest control decision-making system and method based on a knowledge graph. The method collects unstructured agricultural data streams containing vegetable varieties, growth stages, environmental parameters, disease and pest records and geographic information from an original database. Through scanning identification and content extraction of the data streams, a time-labeled data tuple set is generated. Subsequently, clustering is performed according to the growth stages to which the tuples belong, and stage feature clusters are formed in the order of the growth cycle. Multidimensional feature correlation analysis is performed on the stage feature clusters, and a dynamic coupling relationship between environmental parameters, disease and pest history and geographic spatial features under the growth cycle dimension is established. Based on the dynamic coupling relationship, disease and pest control decision-making parameters of vegetables in a specific planting area at each growth cycle stage are generated. The application realizes accurate matching of control decision-making, crop physiological timing and regional characteristics.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of agricultural informatization and intelligent decision-making, in particular to a vegetable growth cycle pest control decision-making system and method based on a knowledge graph. BACKGROUND

[0002] In existing agricultural production management, pest control decisions are mostly dependent on the experience of agricultural technicians or rule bases based on static historical data. Common technical means include establishing a pest database to query control solutions through keyword matching, or using environmental sensor data to set fixed threshold values to trigger alarms. These methods treat the entire growth season of vegetables as a homogeneous time period, or store and analyze data from different sources in isolation. Existing solutions deal with static or simply time-sequenced data records, and fail to reorganize the data according to the internal rhythms of vegetable physiological development. There is a disconnect between data records and the actual phenological stages of crops, resulting in poor timeliness of control recommendations.

[0003] The deficiencies of existing technologies lie in the fact that their decision-making models cannot reflect the dynamic interaction between environmental parameters, pest occurrence patterns, and geographical conditions in different growth stages of crops. Environmental data and pest records are mostly associated after the fact, lacking forward-looking coupling analysis based on the growth cycle. Geospatial information is often only used as a background tag and fails to be deeply integrated into decision-making factors. This results in control strategies that are generally applicable but lack specificity, making it difficult to adapt to specific regions, specific varieties, and different growth stages, and failing to achieve true precision control. The present application needs to solve the two core problems of strictly aligning data organization with growth timing and constructing a dynamic association model of multi-source heterogeneous features within this timing framework. SUMMARY

[0004] The present application aims to provide a vegetable growth cycle pest control decision-making system and method based on a knowledge graph to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides a vegetable growth cycle pest control decision-making method based on a knowledge graph, which comprises:

[0006] Collecting unstructured agricultural data streams containing vegetable varieties, growth stages, environmental parameters, pest records, and geographical information from the original database;

[0007] Scanning the unstructured agricultural data streams piece by piece, identifying the segment units associated with the vegetable growth cycle in the data streams, and extracting the content of each segment unit to generate a set of data tuples containing time markers;

[0008] inputting the data tuple set into a feature fusion channel, performing stage clustering operation on the data tuples according to the vegetable growth stage to which the segment unit belongs, and generating stage feature clusters sorted according to the growth cycle;

[0009] performing multi-dimensional feature correlation analysis on the stage feature clusters, and establishing dynamic coupling relationship of environmental parameter features, pest and disease occurrence history features and geographical space features under the growth cycle dimension;

[0010] based on the dynamic coupling relationship, generating pest and disease control decision parameters of the vegetable in each growth cycle stage in a specific planting area.

[0011] Preferably, the unstructured agricultural data stream is scanned piece by piece, the segment units associated with the vegetable growth cycle in the data stream are identified, and each segment unit is content extracted to generate a data tuple set containing time markers, including:

[0012] receiving real-time or historical unstructured agricultural data stream output by the original database through a data interface;

[0013] setting a sliding analysis window on the unstructured agricultural data stream, and performing keyword matching on the data content passing through the sliding analysis window, the keywords including feature descriptors of the vegetable growth cycle stage;

[0014] when the data content in the sliding analysis window matches any one of the feature descriptors successfully, the data content of a preset length before and after the matching position is intercepted as a complete segment unit;

[0015] extracting entities and attributes from each segment unit, identifying the vegetable variety name, specific growth stage, pest and disease type, occurrence time point and geographical location description text therein, and encapsulating the extraction results as a structured data tuple;

[0016] adding a time stamp marker of its original appearance in the unstructured agricultural data stream to each structured data tuple, and finally generating a data tuple set containing time markers.

[0017] Preferably, the data tuple set is input into a feature fusion channel, the data tuples are subjected to stage clustering operation according to the vegetable growth stage to which the segment unit belongs, and stage feature clusters sorted according to the growth cycle are generated, including:

[0018] establishing a feature fusion channel, the feature fusion channel containing a plurality of parallel feature processing branches;

[0019] reading the specific growth stage marked by each data tuple from the data tuple set containing time markers;

[0020] According to the specific growth stage, the data tuples are routed to corresponding feature processing branches respectively;

[0021] In each feature processing branch, the data tuples belonging to the same specific growth stage are rearranged according to the chronological order of their time labels;

[0022] The rearranged data tuples are subjected to feature vectorization conversion according to the pest and disease types and environmental parameter categories represented therein, and the vectorization results with similar features within the same time period are aggregated into a feature subset;

[0023] The feature subsets output by all feature processing branches are assembled in the natural order of their corresponding specific growth stages in the vegetable growth cycle to generate stage feature clusters sorted by the growth cycle.

[0024] Preferably, multi-dimensional feature correlation analysis is performed on the stage feature clusters to establish dynamic coupling relationships of environmental parameter features, pest and disease occurrence history features, and geographical space features under the growth cycle dimension, including:

[0025] The stage feature clusters sorted by the growth cycle are analyzed to separate environmental parameter feature subsequences, pest and disease occurrence history feature subsequences, and geographical space feature subsequences therefrom;

[0026] For each specific growth stage, a dynamic correlation graph model corresponding to the growth stage is created;

[0027] In the dynamic correlation graph model, the feature values of the environmental parameter feature subsequences in the growth stage are taken as environmental nodes, the feature values of the pest and disease occurrence history feature subsequences in the growth stage are taken as pest and disease nodes, and the feature values of the geographical space feature subsequences in the growth stage are taken as space nodes;

[0028] By analyzing the co-occurrence and transition patterns between the environmental nodes, pest and disease nodes, and space nodes in historical data, directed connection edges between the environmental nodes, pest and disease nodes, and space nodes are defined and created;

[0029] Each directed connection edge is accompanied by an association strength weight calculated from historical statistical frequency, thereby forming a dynamic coupling subgraph describing the multi-dimensional feature interaction relationship in the growth stage;

[0030] The dynamic coupling subgraphs corresponding to each growth stage are graph structure fused in the order of the growth cycle to generate a dynamic coupling relationship throughout the entire growth cycle;

[0031] The stage feature clusters sorted by the growth cycle are analyzed to separate environmental parameter feature subsequences, pest and disease occurrence history feature subsequences, and geographical space feature subsequences therefrom, including:

[0032] loading the stage feature clusters sorted by growth cycle, deconstructing the feature subset corresponding to each growth stage in the stage feature cluster;

[0033] From each feature subset, identify and extract data with feature vector labels of temperature, humidity, light, and soil pH, arrange them in chronological order to form the environmental parameter feature segment of the growth stage;

[0034] From the same feature subset, identify and extract data with feature vector labels of pest name, occurrence time, and severity, arrange them in chronological order to form the pest occurrence history feature segment of the growth stage;

[0035] From the same feature subset, identify and extract data with feature vector labels of latitude and longitude coordinates, elevation, and plot number, arrange them in chronological order to form the geospatial feature segment of the growth stage;

[0036] Vertically concatenate the environmental parameter feature segment, pest occurrence history feature segment, and geospatial feature segment corresponding to each growth stage in the order of their growth stages to generate environmental parameter feature subsequence, pest occurrence history feature subsequence, and geospatial feature subsequence.

[0037] Preferably, by analyzing the co-occurrence and transition patterns between the environmental nodes, pest nodes, and spatial nodes in the historical data, define and create directed connection edges between the environmental nodes, pest nodes, and spatial nodes, including:

[0038] Collect records of the environmental parameter feature subsequence, pest occurrence history feature subsequence, and geospatial feature subsequence in multiple historical growth cycles;

[0039] Statistically count the number of times that a specific environmental parameter feature node value range, a specific pest occurrence history feature node type, and a specific geospatial feature node location appear simultaneously in the same growth stage as co-occurrence frequency;

[0040] Statistically count the number of times that a specific environmental parameter feature node transitions to a specific pest occurrence history feature node state and the number of times that a specific geospatial feature node transitions to a specific pest occurrence history feature node state between adjacent growth stages as transition frequency;

[0041] Set co-occurrence frequency threshold and transition frequency threshold, when the co-occurrence frequency exceeds the threshold, create undirected connection edges between the environmental nodes, pest nodes, and spatial nodes to represent the existence of association;

[0042] When the transition frequency exceeds its threshold value, a directed connection edge representing a causal or influence relationship is created from the node as the starting state to the node as the arrival state;

[0043] The directed connection edge is merged with the undirected connection edge as the connection edge set between the nodes in the growth stage.

[0044] Preferably, the dynamic coupling subgraph corresponding to each growth stage is fused according to the growth cycle sequence to generate a dynamic coupling relationship throughout the entire growth cycle, including:

[0045] The dynamic coupling subgraph corresponding to the first growth stage is obtained and taken as an initial fusion graph;

[0046] The dynamic coupling subgraph corresponding to the next growth stage is read according to the growth cycle sequence;

[0047] Identify the same nodes in the initial fusion graph and the next growth stage dynamic coupling subgraph, which refer to the same geographic spatial features or the same vegetable varieties;

[0048] Merge the same nodes in the next growth stage dynamic coupling subgraph and the initial fusion graph;

[0049] Add the unique new nodes and new connection edges in the next growth stage dynamic coupling subgraph to the initial fusion graph;

[0050] In the merged graph, between the nodes of adjacent growth stages, according to the disease and pest occurrence history feature subsequence, add a time sequence connection edge representing the evolution of diseases and pests across stages;

[0051] Repeat until the dynamic coupling subgraphs of all growth stages are fused, and finally generate a dynamic coupling relationship graph that integrates all feature nodes and complex connection relationships in the entire growth cycle.

[0052] Preferably, based on the dynamic coupling relationship, generate disease and pest control decision parameters of the vegetable in a specific planting area at each growth cycle stage, including:

[0053] Receive a query request for a target vegetable variety in a specific planting area, the query request containing a geographic location code;

[0054] Traverse the dynamic coupling relationship to filter out the geographic spatial feature nodes matching the geographic location code, and locate the growth stage to which the geographic spatial feature nodes belong;

[0055] Extract the dynamic coupling subgraph corresponding to the growth stage, and obtain all environmental parameter feature nodes and disease and pest occurrence history feature nodes that have directed connection edges with the geographic spatial feature nodes.

[0056] read the current monitoring value of the environmental parameter feature node, compare it with the historical feature value range of the environmental parameter feature node recorded in the dynamic coupling subgraph, and calculate the deviation degree;

[0057] In combination with the correlation strength weight on the directed connection edge, the pest occurrence history feature node associated with the current environmental monitoring value is weighted and evaluated to predict the pest occurrence possibility level;

[0058] According to the pest occurrence possibility level and the associated pest type, a pest control decision parameter including control opportunity, recommended measure and intensity level is generated from a preset strategy library.

[0059] Preferably, reading the current monitoring value of the environmental parameter feature node, comparing it with the historical feature value range of the environmental parameter feature node recorded in the dynamic coupling subgraph, and calculating the deviation degree, comprises:

[0060] Obtain the actual monitoring value of the environmental parameter of the specific planting area at the current growth stage from the real-time data source;

[0061] From the dynamic coupling subgraph, locate the geographical space feature node corresponding to the geographical location code, and find the environmental parameter feature node directly connected thereto;

[0062] Query the feature value recorded in the historical data of the environmental parameter feature node, which includes the maximum value, the minimum value, the average value and the common numerical value distribution interval;

[0063] Compare the actual monitoring value of the environmental parameter with the historical maximum value, the historical minimum value and the historical average value respectively, and calculate the absolute difference value;

[0064] Divide the absolute difference value by the corresponding historical feature value change amplitude to obtain a plurality of relative deviation ratios for the maximum value, the minimum value and the average value;

[0065] Select the maximum value of the plurality of relative deviation ratios as the final deviation degree of the environmental parameter feature node at the current time.

[0066] Preferably, in combination with the correlation strength weight on the directed connection edge, the pest occurrence history feature node associated with the current environmental monitoring value is weighted and evaluated to predict the pest occurrence possibility level, comprising:

[0067] In the dynamic coupling subgraph, find all directed connection edges starting from the environmental parameter feature node and pointing to different pest occurrence history feature nodes;

[0068] obtaining an association strength weight attached to each of the directed connection edges, the association strength weight being calculated based on historical co-occurrence and transition frequency;

[0069] taking the final deviation degree of the environment parameter feature node calculated as an influence factor;

[0070] taking the historical occurrence frequency of the pest and disease occurrence history feature node pointed to by each directed connection edge as another influence factor;

[0071] multiplying the association strength weight, the final deviation degree influence factor and the historical occurrence frequency influence factor to obtain a comprehensive occurrence index for each possible pest and disease occurrence history feature node;

[0072] sorting all pest and disease occurrence history feature nodes according to the comprehensive occurrence index from high to low;

[0073] mapping the comprehensive occurrence index to different level labels according to a preset index threshold range, the level labels including high possibility, medium possibility and low possibility, thereby generating a pest and disease occurrence possibility level.

[0074] Preferably, the processor, when executing the computer program, implements the steps of the vegetable growth cycle pest and disease control decision method based on the knowledge graph according to any one of the above.

[0075] Compared with the prior art, the present application has the following advantages:

[0076] The data tuples are clustered according to the specific growth stages corresponding to their time marks to generate stage feature clusters arranged in the order of the growth cycle. The originally simply listed data stream according to the collection time is reorganized into a data structure strictly corresponding to the physiological development stages of the vegetables. The data of each stage forms an independent and ordered cluster, realizing the precise alignment of the data and the crop life cycle. The analysis deviation caused by the time sequence misalignment is eliminated, a clear boundary and pure content data set is provided for each independent growth stage, and the subsequent analysis can be based on the unique physiological state and vulnerability of each stage, thereby supporting the phased and differentiated decision-making process, rather than providing a general scheme for the entire growth season.

[0077] On the basis of the feature clusters organized by growth stages, the dynamic coupling relationship between environmental parameters, pest history and geographical spatial features is established in the growth cycle dimension. Instead of simple feature splicing or static association, the growth stage is taken as the analysis unit to dynamically model the interaction mode of the three types of heterogeneous features in this unit. Geographical spatial features are integrated into the association network as core variables rather than background information. A feature association graph containing spatial context that evolves with the growth stage is generated. The coupling relationship can explain why the pest occurrence risk is different in different regions or different stages under the same environmental conditions, thereby directly driving the generation of highly contextualized prevention and control parameters. The prevention and control parameters are essentially the output results of the dynamic model under the current stage and current regional conditions, and have stronger explainability and regional adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0078] Figure 1 The working principle diagram of the vegetable growth cycle pest control decision method based on the knowledge graph is described.

[0079] Figure 2 The flowchart for generating stage feature clusters is described.

[0080] Figure 3 The flowchart for defining and creating directed connection edges is described.

[0081] Figure 4 The column chart for environmental parameters and pest association strength is described.

[0082] Figure 5 The comparative column chart for pest occurrence frequency and comprehensive occurrence index is described. DETAILED DESCRIPTION

[0083] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only 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 are within the scope of protection of the present application.

[0084] Please refer to Figure 1The application provides a vegetable growth cycle pest control decision method based on a knowledge graph, which comprises the following steps: collecting unstructured agricultural data streams containing vegetable varieties, growth stages, environmental parameters, pest records and geographical information from an original database; performing a piece-by-piece scan on the unstructured agricultural data streams and identifying fragment units associated with the vegetable growth cycle in the data streams; performing content extraction on each fragment unit to generate a time-labeled data tuple set; inputting the data tuple set into a feature fusion channel and performing stage clustering operation on the data tuples according to the growth stages of the fragment units, to generate stage feature clusters sorted according to the growth cycle; performing multi-dimensional feature association analysis on the stage feature clusters, establishing a dynamic coupling relationship between the environmental parameter features, the pest occurrence history features and the geographical space features in the growth cycle dimension, and generating pest control decision parameters of the vegetable in each growth cycle stage in a specific planting area based on the dynamic coupling relationship.

[0085] In one embodiment of the application, referring to Figure 2 , real-time or historical unstructured agricultural data streams output by the original database through a data interface are received, a sliding analysis window is set on the unstructured agricultural data streams, and keyword matching is performed on the data content passing through the sliding analysis window, the keywords including feature descriptors of the vegetable growth cycle stages; when the data content in the sliding analysis window matches any one of the feature descriptors successfully, the data content with a preset length before and after the matching position is intercepted as a complete fragment unit; entity and attribute extraction is performed on each fragment unit, and the vegetable variety name, the specific growth stage, the pest type, the occurrence time point and the geographical position description text are identified; the extraction results are packaged into a structured data tuple, a time stamp of the original appearance of each structured data tuple in the unstructured agricultural data stream is added, and a time-labeled data tuple set is finally generated. A feature fusion channel is established, and the feature fusion channel comprises a plurality of parallel feature processing branches; the specific growth stage of each data tuple is read from the time-labeled data tuple set; the data tuples belonging to the same specific growth stage are routed to the corresponding feature processing branch according to the specific growth stage; the data tuples belonging to the same specific growth stage are rearranged according to their time labels in each feature processing branch; the rearranged data tuples are converted into feature vectors according to the pest types and environmental parameter categories represented by the data tuples, and the vectorized results with similar features in the same time period are aggregated into a feature subset; the feature subsets output by all feature processing branches are assembled according to the natural order of the specific growth stages in the whole vegetable growth cycle, to generate stage feature clusters sorted according to the growth cycle.

[0086] In a specific implementation, a real-time unstructured agricultural data stream is outputted from a raw database of a vegetable planting management system through an application programming interface, the unstructured agricultural data stream containing temperature values recorded by sensors, pest observation log texts inputted by agricultural personnel, geographic position coordinate information obtained by satellite remote sensing, and growth stage descriptions in planting archives. In a specific implementation, a sliding analysis window with a length of 500 characters is set on the unstructured agricultural data stream, the sliding analysis window moving forward by steps of 100 characters, and the data content passing through the sliding analysis window each time being matched with a preset keyword list containing characteristic descriptors of vegetable growth cycle stages such as "sowing period", "seedling period", "flowering period", and "fruiting period". When the data content in the sliding analysis window is successfully matched with the characteristic descriptor "seedling period", the system intercepts data content with a length of 300 characters before and after the matching position, forming a complete segment unit about seedling growth information.

[0087] An entity and attribute extraction operation based on a named entity recognition model is performed on each segment unit, and the vegetable variety name "tomato", the specific growth stage "seedling period", the pest type "aphid", the time point "2023-05-10", and the geographic position description text "north latitude 40 degrees, east longitude 116 degrees" are recognized and extracted from the segment unit, and the extraction result is packaged as a structured data tuple with uniform fields. A time stamp "1683724800" corresponding to the time when each structured data tuple appears in the unstructured agricultural data stream is added to each structured data tuple, and a data tuple set containing time marks is finally generated. In some embodiments, the raw database also outputs a historical unstructured agricultural data stream, and the historical unstructured agricultural data stream is processed by the same sliding analysis window and feature descriptor matching mechanism, and the segment units in the historical unstructured agricultural data stream are also extracted and added with time stamp marks corresponding to their recording times, to jointly constitute the data tuple set.

[0088] A feature fusion channel is established, and the feature fusion channel contains multiple parallel feature processing branches corresponding to the sowing period, the seedling period, the flowering period, and the fruiting period. From the data tuple set containing time marks, the specific growth stage field content labeled by each data tuple is read. According to the specific growth stage field content, the data tuples are routed to the "sowing period" feature processing branch, the "seedling period" feature processing branch, the "flowering period" feature processing branch, or the "fruiting period" feature processing branch. In each feature processing branch, the data tuples belonging to the same specific growth stage are arranged in ascending order according to the numerical value of their time marks to achieve rearrangement. Optionally, the format of the time mark is Unix timestamp.

[0089] For the rearranged data tuples, feature vectorization conversion is performed according to the internal representation of the pest and disease types and the environmental parameter categories. In specific implementation, the pest and disease types such as "aphid" and "powdery mildew" and the environmental parameter categories such as "temperature" and "humidity" are converted into fixed-dimension sparse vectors by using one-hot encoding. In specific implementation, the vectors with cosine similarity higher than a set threshold in the feature vectorization conversion results within the same natural day are aggregated into a feature subset. The feature subsets output by all feature processing branches are assembled in the natural order of the specific growth stages in the whole growth cycle of the tomato, i.e., the order of the sowing period, the seedling period, the flowering period, and the fruiting period, to generate the stage feature cluster sorted by the growth cycle. In some embodiments, the number of feature processing branches is strictly consistent with the number of standard growth stages defined for the target vegetable variety.

[0090] Optionally, the length and step of the sliding analysis window can be preset according to the average length of the sentences in the unstructured agricultural data stream. It can be understood that the list of feature descriptors for the growth cycle stages of the vegetable can be extended and maintained according to the agronomic knowledge of different vegetable varieties. It can be understood that in the feature vectorization conversion process, for the specific numerical value of the environmental parameter, such as the temperature value of 25 degrees Celsius, normalization processing can be performed to convert it into a scalar value before being used as a dimension of the vector. The judgment criterion for feature similarity can be realized by calculating the Euclidean distance or cosine similarity between vectors and comparing it with a preset threshold. The feature subsets output by all feature processing branches are assembled in the natural order of the specific growth stages in the whole growth cycle of the vegetable, and the assembly process can be represented as a sequential splicing operation to generate the stage feature cluster sorted by the growth cycle.

[0091] In an embodiment of the present application, the stage feature clusters sorted by the growth cycle are parsed and the environmental parameter feature subsequences, the pest and disease occurrence history feature subsequences, and the geographical space feature subsequences are separated therefrom, and a dynamic correlation graph model corresponding to each specific growth stage is created, in which the feature values of the environmental parameter feature subsequences in the growth stage are taken as environmental nodes, the feature values of the pest and disease occurrence history feature subsequences in the growth stage are taken as pest and disease nodes, and the feature values of the geographical space feature subsequences in the growth stage are taken as space nodes. The directed connection edges between the environmental nodes, the pest and disease nodes, and the space nodes are defined and created by analyzing the co-occurrence and transfer patterns between the environmental nodes, the pest and disease nodes, and the space nodes in the historical data, and each directed connection edge is attached with an association strength weight calculated from the historical statistical frequency, thereby forming a dynamic coupling subgraph describing the multi-dimensional feature interaction relationship in the growth stage. The stage feature clusters sorted by the growth cycle are loaded and the feature subsets corresponding to each growth stage in the stage feature clusters are deconstructed, the data with feature vector labels of temperature, humidity, light, and soil pH are identified and extracted from each feature subset and arranged in time sequence to form the environmental parameter feature segments of the growth stage, the data with feature vector labels of pest and disease name, occurrence time, and severity are identified and extracted from the same feature subset and arranged in time sequence to form the pest and disease occurrence history feature segments of the growth stage, and the data with feature vector labels of latitude and longitude coordinates, elevation, and plot number are identified and extracted from the same feature subset and arranged in time sequence to form the geographical space feature segments of the growth stage. The environmental parameter feature segments, the pest and disease occurrence history feature segments, and the geographical space feature segments corresponding to each growth stage are respectively longitudinally concatenated in the order of their growth stages to generate the environmental parameter feature subsequences, the pest and disease occurrence history feature subsequences, and the geographical space feature subsequences.

[0092] In a specific implementation, the stage feature cluster sorted by growth period generated by the feature fusion channel is loaded, and the stage feature cluster contains a sowing period feature subset, a seedling period feature subset, a flowering period feature subset, and a fruiting period feature subset. The feature subset corresponding to each growth period in the stage feature cluster is deconstructed. From the sowing period feature subset, data points with feature vector labels of "22°C", "24°C", and "20°C" for temperature, data points with feature vector labels of "65% RH", "70% RH", and "68% RH" for humidity, data points with feature vector labels of "12000 lux" and "11500 lux" for light, and data points with feature vector labels of "pH 6.5" and "pH 6.8" for soil pH are identified and extracted. These data points are arranged in the order of their recorded time markers "t1", "t2", and "t3" to form an environmental parameter feature segment of the sowing period growth stage. In a specific implementation, from the same sowing period feature subset, data points with feature vector labels of "damping-off disease" for disease and pest name, data points with feature vector labels of "t2" for occurrence time, and data points with feature vector labels of "mild" for severity are identified and extracted. These data points are arranged in time order to form a disease and pest occurrence history feature segment of the sowing period growth stage. From the sowing period feature subset, data points with feature vector labels of "40.1° N, 116.2° E" for latitude and longitude coordinates, data points with feature vector labels of "50 meters" for altitude, and data points with feature vector labels of "B-05" for plot number are identified and extracted. These data points are arranged in time order to form a geographic space feature segment of the sowing period growth stage.

[0093] For each specific growth period, such as for the seedling period growth stage, a dynamic association graph model corresponding to the seedling period growth stage is created. In the dynamic association graph model corresponding to the seedling period growth stage, the feature value "daily average temperature 25°C" of the environmental parameter feature subsequence in the seedling period growth stage is taken as an environmental node, the feature value "aphid outbreak" of the disease and pest occurrence history feature subsequence in the seedling period growth stage is taken as a disease and pest node, and the feature value "plot B-05" of the geographic space feature subsequence in the seedling period growth stage is taken as a space node. By analyzing the co-occurrence and transition patterns between the environmental nodes, disease and pest nodes, and space nodes in the historical data, directed connection edges between the environmental nodes, disease and pest nodes, and space nodes are defined and created. Each directed connection edge is accompanied by an association strength weight calculated from historical statistical frequency, thereby forming a dynamic coupling subgraph that describes the multi-dimensional feature interaction relationship within the seedling period growth stage.

[0094] The environment parameter feature sub-sequence, the pest and disease occurrence history feature sub-sequence, and the geographical space feature sub-sequence are respectively generated by longitudinally concatenating the environment parameter feature segments, the pest and disease occurrence history feature segments, and the geographical space feature segments corresponding to each growth stage in the order of the growth stages. In a specific implementation, the environment parameter feature segments of the sowing period, the seedling period, the flowering period, and the fruiting period are longitudinally connected in the order of the sowing period, the seedling period, the flowering period, and the fruiting period to generate the environment parameter feature sub-sequence covering the whole cycle. The dynamic coupling sub-graphs corresponding to the respective growth stages are fused in the graph structure in the order of the growth cycle to generate the dynamic coupling relationship throughout the whole growth cycle. In some embodiments, the generated geographical space feature sub-sequence is a list of geographical position descriptions arranged in the order of time stages. In some embodiments, the pest and disease occurrence history feature sub-sequence includes an ordered set of pest and disease types and their occurrence time points recorded in different growth stages.

[0095] Optionally, the identification of the feature vector label is based on an exact match of the metadata label field attached to each data point in the feature sub-sequence. It can be understood that the environment parameter feature sub-sequence is generated by connecting the environment parameter feature segments of each stage end to end. It can be understood that the node attribute values in the dynamic association graph model are directly derived from the feature values of the corresponding feature sub-sequence in the corresponding growth stage. By analyzing the co-occurrence and transition patterns between the environment nodes, the pest and disease nodes, and the space nodes in the historical data, the co-occurrence pattern refers to the pattern in which multiple node feature values appear simultaneously in the same time or the same growth stage in the historical record. The creation logic of the directed connection edge is based on the sequence of feature state changes in the historical events, for example, when the historical data shows that the environment node "temperature continuously higher than 28℃" appears multiple times, followed by the pest and disease node "red spider disease", a directed connection edge from the former to the latter is created. The calculation of the association strength weight can be based on the proportion of the co-occurrence or transition event in the total events in the history, and the association strength weight value is between 0 and 1.

[0096] In an embodiment of the present application, refer to Figure 3, collect records of the environmental parameter feature subsequence, the pest and disease occurrence history feature subsequence and the geographical space feature subsequence in multiple historical growth cycles, count the number of simultaneous occurrence of the value interval of the environmental parameter feature node, the type of the pest and disease occurrence history feature node and the position of the geographical space feature node in the same growth stage as the co-occurrence frequency, count the number of state transition of the environmental parameter feature node to the pest and disease occurrence history feature node and the number of state transition of the geographical space feature node to the pest and disease occurrence history feature node between adjacent growth stages as the transition frequency, set the co-occurrence frequency threshold and the transition frequency threshold, when the co-occurrence frequency exceeds the threshold, create a non-directional connection edge between the environmental node, the pest and disease node and the space node to represent the associated existence, when the transition frequency exceeds the threshold, create a directional connection edge from the node as the starting state to the node as the arrival state to represent the causal or influence relationship, combine the directional connection edge and the non-directional connection edge as the connection edge set between the nodes in the growth stage. Obtain the dynamic coupling subgraph corresponding to the first growth stage and take it as the initial fusion graph, read the dynamic coupling subgraph corresponding to the next growth stage in the order of the growth cycle, identify the same nodes in the initial fusion graph and the next growth stage dynamic coupling subgraph, the same nodes refer to the same geographical space features or the same vegetable varieties, merge the same nodes in the next growth stage dynamic coupling subgraph and the initial fusion graph, add the new nodes and new connection edges in the next growth stage dynamic coupling subgraph to the initial fusion graph, add the time sequence connection edge representing the evolution of the pest and disease across stages between the nodes of adjacent growth stages in the merged graph according to the pest and disease occurrence history feature subsequence, repeat the execution until the dynamic coupling subgraphs of all growth stages are fused, and finally generate a dynamic coupling relationship graph atlas integrating all feature nodes and complex connection relationships in the whole growth cycle.

[0097] In a specific implementation, records of the environmental parameter feature subsequence, the pest and disease occurrence history feature subsequence, and the geographical space feature subsequence in multiple historical growth periods are collected, and these historical records cover the complete planting process of the same vegetable variety in the same geographical area in the past three years. The number of times that the environmental parameter feature node “daily average temperature between 25°C and 28°C”, the pest and disease occurrence history feature node “aphid”, and the geographical space feature node “plot B-05” appear at the same time in the same seedling growth stage is counted, and the counted number of times is recorded as the co-occurrence frequency. The number of times that the environmental parameter feature node “air humidity continuously higher than 85%” state transitions to the pest and disease occurrence history feature node “downy mildew” between adjacent seedling growth stages and flowering growth stages is counted, and the number of times that the geographical space feature node “low-lying land” state transitions to the pest and disease occurrence history feature node “root rot” is counted, and the counted number of times is recorded as the transition frequency. A co-occurrence frequency threshold and a transition frequency threshold are set, the co-occurrence frequency threshold is 5 times, and the transition frequency threshold is 3 times. When the co-occurrence frequency of “daily average temperature between 25°C and 28°C”, “aphid”, and “plot B-05” is more than 5 times, a non-directional connection edge representing the existence of association is created between the “daily average temperature between 25°C and 28°C” environmental node, the “aphid” pest and disease node, and the “plot B-05” space node. When the transition frequency of the “air humidity continuously higher than 85%” environmental node to the “downy mildew” pest and disease node is more than 3 times, a directional connection edge representing a cause-and-effect or influence relationship is created from the “air humidity continuously higher than 85%” environmental node to the “downy mildew” pest and disease node. The created directional connection edge and non-directional connection edge are combined as a connection edge set between nodes in the seedling growth stage.

[0098] The first growth stage, i.e., the sowing stage, is obtained, and a dynamic coupling subgraph corresponding to the sowing stage is obtained. The dynamic coupling subgraph corresponding to the sowing stage is taken as an initial fusion graph. In order, the next growth stage, i.e., the seedling stage, is read, and a dynamic coupling subgraph corresponding to the seedling stage is obtained. The same nodes in the initial fusion graph and the seedling stage dynamic coupling subgraph are identified, and the same nodes refer to the same geographical space feature "plot B-05" or the same vegetable variety "tomato". The geographical space feature node "plot B-05" in the seedling stage dynamic coupling subgraph that is the same as the initial fusion graph is merged, and the merging operation means that only one node representing "plot B-05" is retained in the fused graph. New nodes unique to the seedling stage, such as the environmental parameter feature node "daylight duration more than 12 hours" specific to the seedling stage, and new connection edges connecting the new nodes and other nodes are added to the initial fusion graph. In the merged graph, between the disease and pest occurrence history feature node "stand disease" in the sowing stage and the disease and pest occurrence history feature node "sudden collapse disease" in the seedling stage, a time sequence connection edge representing the evolution of the disease and pest across stages is added according to the disease evolution path recorded in the disease and pest occurrence history feature subsequence. The operation of reading the next growth stage dynamic coupling subgraph in order, identifying and merging the same nodes, adding new nodes and new connection edges, and adding time sequence connection edges is repeatedly performed until the dynamic coupling subgraphs corresponding to the flowering stage and the fruiting stage are fused, and a dynamic coupling relationship graph atlas integrating all environmental parameter feature nodes, disease and pest occurrence history feature nodes, geographical space feature nodes, and complex connection relationships in the whole growth cycle of the tomato is finally generated. In some embodiments, the specific values of the co-occurrence frequency threshold and the transition frequency threshold are dynamically adjusted according to the total amount of historical data. In some embodiments, the direction of the time sequence connection edge is always directed to the later growth stage in time.

[0099] Optionally, the calculation of the association strength weight can be based on the historical statistical frequency, and the calculation formula is:

[0100]

[0101] wherein: represents the association strength weight, represents the co-occurrence frequency or the transition frequency obtained by statistical analysis of the historical data, represents the total number of historical observations. It can be understood that the node merging operation includes the integration and aggregation of attribute information carried by each node before merging. It can be understood that the dynamic coupling relationship graph atlas is a knowledge representation stored in the form of a graph structure, wherein the nodes in the graph represent entities or states, and the edges represent relationships or influences. In specific implementations, the total number of historical observations is referred to as the total number of times of interpreting the target growth stage or state transition within the time window of analyzing the historical data. The correlation strength weight The value range of is between 0 and 1, and the closer the value is to 1, the stronger the historical correlation. In a specific implementation, the time sequence connection edge is not limited to the pest node, and the significant change relationship between the environmental parameter feature node in adjacent growth stages can also be expressed through the time sequence connection edge.

[0102] In an embodiment of the present application, a query request for a target vegetable variety in a specific planting area is received, and the query request includes a geographic location code. The dynamic coupling relationship is traversed, and the geographic space feature node matching the geographic location code is filtered out and located to the growth stage to which the geographic space feature node belongs. The dynamic coupling subgraph corresponding to the growth stage is extracted, and all environmental parameter feature nodes and pest occurrence history feature nodes having a directed connection edge with the geographic space feature node are obtained. The current monitoring value of the environmental parameter feature node is read and compared with the historical feature value range of the environmental parameter feature node recorded in the dynamic coupling subgraph to calculate the deviation. The pest occurrence history feature nodes associated with the current environmental monitoring value are weighted and evaluated in combination with the correlation strength weight on the directed connection edge to predict the pest occurrence possibility level. According to the pest occurrence possibility level and the associated pest type, a pest control decision parameter including the control opportunity, the recommended measures, and the intensity level is generated from the preset strategy library.

[0103] In a specific implementation, a query request for a target vegetable variety "tomato" in a specific planting area is received, and the query request includes a geographic location code "B-05". By traversing the dynamic coupling graph, the geographic spatial feature node matching the geographic location code "B-05" is filtered out, and the growth stage "seedling stage" to which the geographic spatial feature node "B-05" belongs is located. The dynamic coupling subgraph corresponding to the seedling stage is extracted, and all environment parameter feature nodes and pest and disease occurrence history feature nodes having a directed connection edge with the geographic spatial feature node "B-05" are obtained, for example, the environment parameter feature node "daily average temperature" and the pest and disease occurrence history feature node "aphid". The current monitoring value "25°C" of the environment parameter feature node "daily average temperature" is read, which is compared with the historical feature value range "20°C to 30°C" of the environment parameter feature node "daily average temperature" recorded in the dynamic coupling subgraph, and the deviation degree of the current monitoring value relative to the historical range is calculated. In combination with the correlation strength weight on the directed connection edge, the pest and disease occurrence history feature node "aphid" associated with the current environment monitoring value is weighted and evaluated, and the pest and disease occurrence possibility level is predicted. According to the predicted pest and disease occurrence possibility level "medium possibility" and the associated pest and disease type "aphid", the pest and disease control decision parameters including the control opportunity "before present budding", the recommended measure "spraying imidacloprid", and the intensity level "regular dose" are generated from the preset strategy library. In some embodiments, the query request can be input through a graphical user interface, and the query request can include target growth stage information in addition to the geographic location code. In some embodiments, the nodes having a directed connection edge with the geographic spatial feature node in the dynamic coupling subgraph can include multiple environment parameter feature nodes and multiple pest and disease occurrence history feature nodes.

[0104] In a specific implementation, the calculation of the deviation degree involves comparing the current monitoring value with the historical feature value range. The current monitoring value of the environment parameter feature node is read, which is compared with the historical feature value range of the environment parameter feature node recorded in the dynamic coupling subgraph, and the process of calculating the deviation degree can be represented as the formula:

[0105]

[0106] Wherein: represents the final deviation degree calculated, represents the actual monitoring value of the environment parameter obtained from the real-time data source, represents the historical maximum value of the environment parameter feature node queried in the dynamic coupling subgraph, represents the historical minimum value of the environment parameter feature node queried in the dynamic coupling subgraph, represents the historical average value of the environment parameter feature node queried in the dynamic coupling subgraph, represents the historical feature value change amplitude, which is calculated in the following manner In a specific implementation, the weighted evaluation process utilizes the weight of the directed connection edge and the calculated deviation degree. The preset strategy library is a database that stores the mapping relationship between disease and pest types, occurrence probability levels, and recommended prevention and control parameters. Optionally, the disease and pest occurrence probability level can be further refined into multiple discrete levels. It can be understood that the generation of disease and pest prevention and control decision parameters is the final output result of the query request. The mapping process looks up the preset strategy library according to the disease and pest type and the probability level as the joint primary key. The content of the preset strategy library is pre-defined based on agricultural expert knowledge, refer to Table 1, part of the mapping relationship.

[0107] Table 1: Disease and pest control strategy mapping table

[0108] Disease / pest type Occurrence likelihood Control timing Suggested measure Intensity level Aphid High likelihood Immediately implement Spray flonicamid Intensified dose Aphid Medium likelihood Before present bud Spray dinotefuran Regular dose Downy mildew Medium likelihood Initial stage of occurrence Spray oxycarboxin Regular dose Root rot Low likelihood After planting Drench anilazine Preventive dose

[0109] Referring to Figure 4 This is a column chart of the correlation strength between environmental parameters and diseases and pests. The relative humidity has the highest correlation weight with downy mildew (about 0.9), which is the core influencing environmental factor of the disease; the daily average temperature has a relatively high correlation weight with aphids (about 0.85), which is the main environmental driving factor of aphid occurrence; the soil pH value has a higher correlation weight with root rot (0.8) than other diseases, which reflects the significant influence of soil acidity and alkalinity on root rot; the light duration has a low correlation degree with the three types of diseases and pests, which is a secondary influencing factor. This kind of chart is usually used for decision analysis of agricultural disease and pest control, which can help growers to monitor the high correlation environmental parameters; it is a visual presentation of the dynamic coupling relationship between "environment-disease and pest", which provides data support for subsequent prediction of disease and pest occurrence probability.

[0110] In an embodiment of the present application, the actual monitoring value of the environmental parameter of the specific planting area at the current growth stage is obtained from the real-time data source, the geospatial feature node corresponding to the geographic location code is located from the dynamically coupled subgraph, and the environmental parameter feature node directly connected thereto is found, the characteristic values recorded in the historical data of the environmental parameter feature node are queried, and the characteristic values include the maximum value, the minimum value, the average value, and the common numerical value distribution interval. The environmental parameter actual monitoring value is compared with the historical maximum value, the historical minimum value, and the historical average value respectively to calculate the absolute difference value, and the absolute difference value is divided by the corresponding historical characteristic value change amplitude to obtain a plurality of relative deviation ratios for the maximum value, the minimum value, and the average value. The maximum value in the plurality of relative deviation ratios is selected as the final deviation degree of the environmental parameter feature node at the current time. In the dynamically coupled subgraph, all directed connection edges starting from the environmental parameter feature node and pointing to different pest and disease occurrence history feature nodes are found, the associated strength weight attached to each directed connection edge is obtained, and the associated strength weight is calculated based on the historical co-occurrence and transition frequency. The final deviation degree of the environmental parameter feature node calculated is taken as an influence factor, the historical occurrence frequency of the pest and disease occurrence history feature node pointed to by each directed connection edge is taken as another influence factor, and the associated strength weight, the final deviation degree influence factor, and the historical occurrence frequency influence factor are multiplied to obtain a comprehensive occurrence index for each possible pest and disease occurrence history feature node. All pest and disease occurrence history feature nodes are sorted in descending order of the comprehensive occurrence index, the comprehensive occurrence index is mapped to different level labels according to a preset index threshold range, and the level labels include high possibility, medium possibility, and low possibility, thereby generating a pest and disease occurrence possibility level.

[0111] In a specific content. Implementation, the actual monitoring value of the environmental parameter of the specific planting area "plot B-05" in the current seedling stage growth stage is obtained from the real-time data source, and the actual monitoring value of the environmental parameter includes "daily average temperature 26.5℃" and "air relative humidity 82%". From the dynamic coupling subgraph corresponding to the seedling stage growth stage, the geographical space feature node "B-05" corresponding to the geographical position code "B-05" is located, and the environmental parameter feature node "daily average temperature" and the environmental parameter feature node "air relative humidity" directly connected with it are found. Query the feature value recorded in the historical data of the environmental parameter feature node "daily average temperature", and the feature value includes the historical maximum value "30℃", the historical minimum value "20℃", the historical average value "25℃" and the common numerical value distribution interval "23℃ to 27℃". The absolute difference value is obtained by comparing and calculating the actual monitoring value of the environmental parameter "26.5℃" with the historical maximum value "30℃", the historical minimum value "20℃" and the historical average value "25℃" respectively, which is the absolute difference value "3.5℃" of the environmental parameter feature node historical maximum value, the absolute difference value "6.5℃" of the environmental parameter feature node historical minimum value and the absolute difference value "1.5℃" of the environmental parameter feature node historical average value. Divide the absolute difference value "3.5℃" by the corresponding historical feature value change amplitude "10℃" to obtain the relative deviation ratio "0.35" of the environmental parameter feature node historical maximum value, divide the absolute difference value "6.5℃" by the historical feature value change amplitude "10℃" to obtain the relative deviation ratio "0.65" of the environmental parameter feature node historical minimum value, and divide the absolute difference value "1.5℃" by the historical feature value change amplitude "10℃" to obtain the relative deviation ratio "0.15" of the environmental parameter feature node historical average value. Select the maximum "0.65" among the relative deviation ratios "0.35", "0.65" and "0.15" as the final deviation degree of the environmental parameter feature node "daily average temperature" at the current time.

[0112] In the dynamic coupling subgraph, find all directed connection edges starting from the environmental parameter feature node "daily average temperature" and pointing to different pest and disease occurrence history feature nodes, for example, find the directed connection edge pointing to the pest and disease occurrence history feature node "aphid" and the directed connection edge pointing to the pest and disease occurrence history feature node "powdery mildew". Obtain the association strength weight attached to each directed connection edge, which is calculated based on historical co-occurrence and transition frequency. The association strength weight of the directed connection edge pointing to the pest and disease occurrence history feature node "aphid" is "0.8", and the association strength weight of the directed connection edge pointing to the pest and disease occurrence history feature node "powdery mildew" is "0.3". Take the final deviation "0.65" of the environmental parameter feature node "daily average temperature" calculated as an influence factor. Take the historical occurrence frequency of the pest and disease occurrence history feature node to which each directed connection edge points as another influence factor. The historical occurrence frequency of the pest and disease occurrence history feature node "aphid" is "0.4", and the historical occurrence frequency of the pest and disease occurrence history feature node "powdery mildew" is "0.1". In specific implementation, multiply the association strength weight, the final deviation influence factor and the historical occurrence frequency influence factor to obtain the comprehensive occurrence index for each possible pest and disease occurrence history feature node. The calculation formula is:

[0113]

[0114] wherein: represents the comprehensive occurrence index for a certain pest and disease occurrence history feature node, represents the association strength weight attached to the directed connection edge pointing to the node from the dynamic coupling subgraph, represents the final deviation of the environmental parameter feature node calculated, represents the occurrence frequency of the pest and disease occurrence history feature node in the historical data. For the pest and disease occurrence history feature node "aphid", its comprehensive occurrence index is , and for the pest and disease occurrence history feature node "powdery mildew", its comprehensive occurrence index is The history characteristic nodes of all diseases and pests are sorted in descending order according to the comprehensive occurrence indexes, and a sorting result of "aphid (0.208)" and "powdery mildew (0.0195)" is obtained. According to a preset index threshold range, for example, the comprehensive occurrence index greater than 0.2 is mapped to the level label "high possibility", the comprehensive occurrence index between 0.1 and 0.2 is mapped to the level label "medium possibility", and the comprehensive occurrence index less than 0.1 is mapped to the level label "low possibility", the comprehensive occurrence index "0.208" is mapped to the level label "high possibility", and the comprehensive occurrence index "0.0195" is mapped to the level label "low possibility", so as to generate the disease and pest occurrence possibility level. In some embodiments, the historical occurrence frequency is calculated based on the proportion of the number of times that the disease and pest node appears in all relevant growth stages in the past to the total observation times.

[0115] Referring to Figure 5 This is a comparison column chart of disease and pest occurrence frequency and comprehensive occurrence index. The aphid occurrence frequency (0.4) and the comprehensive occurrence index (0.21) are both the highest, and are the most prominent diseases and pests in the current monitoring; the occurrence frequency (0.25) of cabbage worm is relatively high, but the comprehensive occurrence index (0.13) is relatively low, indicating that its harm degree is limited; the occurrence frequency (0.1) of powdery mildew and the comprehensive occurrence index (0.02) are both low, which belongs to low-risk diseases and pests; the occurrence frequency (0.1) of root rot is higher than the comprehensive occurrence index (0.05), which reflects its occurrence but light harm characteristics. Such charts are mainly used for agricultural disease and pest risk assessment, and can quickly identify high-risk diseases and pests (such as aphids) to provide a basis for priority prevention.

[0116] It should be noted that, in the present text, relational terms such as first and second and the like can only be used to distinguish one entity or action from another entity or action, without necessarily requiring or implying that there is any such actual relationship or order between these entities or actions. Moreover, the terms "comprises", "comprising", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article, or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or inherent to such a process, method, article, or apparatus.

[0117] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for vegetable growth cycle pest control decision-making based on a knowledge graph, characterized in that, The method comprises: Collecting unstructured agricultural data streams containing vegetable varieties, growth stages, environmental parameters, pest and disease records, and geographic information from the original database; Performing a piece-by-piece scan on the unstructured agricultural data streams, identifying segment units associated with the vegetable growth cycle in the data streams, and performing content extraction on each segment unit to generate a set of data tuples containing time markers; Input the data tuple set into the feature fusion channel, perform stage clustering operation on the data tuples according to the growth stage of the segment unit, and generate stage feature clusters sorted by growth cycle; Performing multi-dimensional feature correlation analysis on the stage feature clusters to establish the dynamic coupling relationship between environmental parameter features, pest and disease occurrence history features, and geographic space features in the growth cycle dimension; Based on the dynamic coupling relationship, generate pest and disease control decision parameters for vegetables in a specific planting area at each growth cycle stage; Performing multi-dimensional feature correlation analysis on the stage feature clusters to establish the dynamic coupling relationship between environmental parameter features, pest and disease occurrence history features, and geographic space features in the growth cycle dimension, including: Analyzing the stage feature clusters sorted by growth cycle to separate environmental parameter feature subsequences, pest and disease occurrence history feature subsequences, and geographic space feature subsequences; For each specific growth stage, create a dynamic association graph model corresponding to the growth stage; In the dynamic association graph model, the feature values of the environmental parameter feature subsequences in the growth stage are used as environmental nodes, the feature values of the pest and disease occurrence history feature subsequences in the growth stage are used as pest and disease nodes, and the feature values of the geographic space feature subsequences in the growth stage are used as space nodes; By analyzing the co-occurrence and transfer patterns between the environmental nodes, pest and disease nodes, and space nodes in the historical data, define and create directed connection edges between the environmental nodes, pest and disease nodes, and space nodes; Each directed connection edge is accompanied by an association strength weight calculated from historical statistical frequency, forming a dynamic coupling subgraph that describes the multi-dimensional feature interaction relationship in the growth stage; Fuse the dynamic coupling subgraphs corresponding to each growth stage according to the growth cycle order to generate a dynamic coupling relationship throughout the entire growth cycle; The analysis of the stage feature clusters sorted by growth cycle to separate environmental parameter feature subsequences, pest and disease occurrence history feature subsequences, and geographic space feature subsequences includes: Load the stage feature clusters sorted by growth cycle, and deconstruct the feature subsets corresponding to each growth stage in the stage feature clusters; From each feature subset, identify and extract data with feature vector labels such as temperature, humidity, light, and soil pH, arrange them in chronological order, and construct the environmental parameter feature segment of the growth stage; From the same feature subset, identify and extract data with feature vector labels such as pest and disease name, occurrence time, and severity, arrange them in chronological order, and construct the pest and disease occurrence history feature segment of the growth stage; From the same feature subset, identify and extract feature vector labels as latitude and longitude coordinates, elevation, plot number data, arrange them in chronological order, and construct the geographical spatial feature segment of the growth stage; Vertically concatenate the environmental parameter feature segment, pest and disease occurrence history feature segment, and geographical spatial feature segment corresponding to each growth stage in the order of their growth stages to generate the environmental parameter feature subsequence, pest and disease occurrence history feature subsequence, and geographical spatial feature subsequence; By analyzing the co-occurrence and transition patterns between the environmental nodes, pest and disease nodes, and spatial nodes in the historical data, define and create directed connection edges between the environmental nodes, pest and disease nodes, and spatial nodes, including: Collect records of the environmental parameter feature subsequence, pest and disease occurrence history feature subsequence, and geographical spatial feature subsequence in multiple historical growth cycles; Statistically analyze the number of times that a specific environmental parameter feature node value range, a specific pest and disease occurrence history feature node type, and a specific geographical spatial feature node location appear simultaneously within the same growth stage as the co-occurrence frequency; Statistically analyze the number of times that a specific environmental parameter feature node transitions to a specific pest and disease occurrence history feature node state and the number of times that a specific geographical spatial feature node transitions to a specific pest and disease occurrence history feature node state between adjacent growth stages as the transition frequency; Set a co-occurrence frequency threshold and a transition frequency threshold. When the co-occurrence frequency exceeds the threshold, create undirected connection edges between the environmental nodes, pest and disease nodes, and spatial nodes to represent the existence of associations. When the transition frequency exceeds the threshold, create directed connection edges from the starting state node to the arrival state node to represent causal or influence relationships. Combine the directed connection edges and undirected connection edges as the connection edge set between nodes within the growth stage. 2.The knowledge graph-based vegetable growth period pest control decision method according to claim 1, characterized in that, Scan the non-structured agricultural data stream piece by piece, identify the segment units associated with the vegetable growth cycle in the data stream, and extract the content of each segment unit to generate a set of data tuples containing time markers, including: Receive real-time or historical non-structured agricultural data streams output by the original database through the data interface; Set a sliding analysis window on the non-structured agricultural data stream, and perform keyword matching on the data content passing through the sliding analysis window. The keywords include feature descriptors of the vegetable growth cycle stages. When the data content within the sliding analysis window matches any one of the feature descriptors successfully, extract the data content of a predetermined length before and after the matching position as a complete segment unit. Extract entities and attributes from each segment unit, identify the vegetable variety name, specific growth stage, pest and disease type, occurrence time point, and geographical location description text, and encapsulate the extraction results as a structured data tuple. Add a timestamp marker to each structured data tuple indicating its original appearance in the non-structured agricultural data stream, and finally generate a set of data tuples containing time markers. 3.The knowledge graph-based vegetable growth period pest control decision method according to claim 2, characterized in that, Inputting the data tuple set into a feature fusion channel, performing stage clustering operation on the data tuples according to the vegetable growth stages to which the segment units belong, and generating stage feature clusters sorted by growth periods, including: Establishing a feature fusion channel, wherein the feature fusion channel comprises a plurality of parallel feature processing branches; Reading the specific growth stage marked by each data tuple from the data tuple set comprising time marks; Routing the data tuples to corresponding feature processing branches according to the specific growth stages; In each feature processing branch, rearranging the data tuples belonging to the same specific growth stage according to their time marks in chronological order; Performing feature vectorization conversion on the rearranged data tuples according to the pest and disease types and environmental parameter categories represented by the data tuples, and aggregating the vectorization results with similar features in the same time period into a feature subset; Assembling the feature subsets output by all feature processing branches in the natural order of the specific growth stages in the whole growth period of the vegetables, to generate stage feature clusters sorted by growth periods. 4.The knowledge graph-based vegetable growth period pest control decision method according to claim 1, characterized in that, Fusing the dynamic coupling subgraphs corresponding to each growth stage in the order of growth periods to generate dynamic coupling relationships throughout the whole growth period, including: Obtaining the dynamic coupling subgraph corresponding to the first growth stage as an initial fusion graph; Reading the dynamic coupling subgraph corresponding to the next growth stage in the order of growth periods; Identifying the same nodes in the initial fusion graph and the dynamic coupling subgraph of the next growth stage, wherein the same nodes refer to the same geographical space features or the same vegetable varieties; Merging the same nodes in the dynamic coupling subgraph of the next growth stage and the initial fusion graph; Adding the unique new nodes and new connection edges in the dynamic coupling subgraph of the next growth stage to the initial fusion graph; In the merged graph, adding time sequence connection edges representing the evolution of pests and diseases across stages between the nodes of adjacent growth stages according to the pest and disease occurrence history feature subsequence; Repeating the execution until the dynamic coupling subgraphs of all growth stages are fused, and finally generating a dynamic coupling relationship graph that integrates all feature nodes and complex connection relationships in the whole growth period. 5.The knowledge graph-based vegetable growth period pest control decision method according to claim 4, characterized in that, Based on the dynamic coupling relationship, generating pest and disease control decision parameters for the vegetables in each growth period stage in a specific planting area, including: Receiving a query request for a target vegetable variety in a specific planting area, wherein the query request comprises a geographical location code; Traversing the dynamic coupling relationship to filter out geographical space feature nodes matching the geographical location code and locate the growth stages to which the geographical space feature nodes belong; Extracting the dynamic coupling subgraph corresponding to the growth stage to obtain all environmental parameter feature nodes and pest and disease occurrence history feature nodes having directed connection edges with the geographical space feature nodes; Reading the current monitoring values of the environmental parameter feature nodes and comparing them with the historical feature value ranges of the environmental parameter feature nodes recorded in the dynamic coupling subgraph to calculate the deviation degree; In combination with the correlation strength weight on the directed connection edge, the pest and disease occurrence history characteristic node associated with the current environment monitoring value is weighted and evaluated to predict a pest and disease occurrence possibility level; According to the pest and disease occurrence possibility level and the associated pest and disease type, a pest and disease control decision parameter including a control time, a recommended measure and an intensity level is generated from a preset strategy library. 6.The knowledge graph-based vegetable growth period pest control decision method according to claim 5, characterized in that, The current monitoring value of the environment parameter characteristic node is read, and is compared with the historical characteristic value range of the environment parameter characteristic node recorded in the dynamic coupling subgraph to calculate a deviation degree, including: The environment parameter actual monitoring value of the specific planting area at the current growth stage is obtained from a real-time data source; From the dynamic coupling subgraph, a geographical space characteristic node corresponding to the geographical location code is located, and an environment parameter characteristic node directly connected thereto is found; The characteristic value recorded in the historical data of the environment parameter characteristic node is queried, and the characteristic value includes a maximum value, a minimum value, an average value and a common numerical value distribution interval; The environment parameter actual monitoring value is compared with the historical maximum value, the historical minimum value and the historical average value respectively to calculate absolute difference values; The absolute difference values are divided by the corresponding historical characteristic value variation amplitudes respectively to obtain a plurality of relative deviation ratios for the maximum value, the minimum value and the average value; The maximum value in the plurality of relative deviation ratios is selected as the final deviation degree of the environment parameter characteristic node at the current time. 7.The knowledge graph-based vegetable growth period pest control decision method according to claim 6, characterized in that, In combination with the correlation strength weight on the directed connection edge, the pest and disease occurrence history characteristic node associated with the current environment monitoring value is weighted and evaluated to predict a pest and disease occurrence possibility level, including: In the dynamic coupling subgraph, all directed connection edges starting from the environment parameter characteristic node and pointing to different pest and disease occurrence history characteristic nodes are found; The correlation strength weight attached to each directed connection edge is obtained, and the correlation strength weight is calculated based on historical co-occurrence and transfer frequency; The final deviation degree of the environment parameter characteristic node calculated is taken as an influence factor; The historical occurrence frequency of the pest and disease occurrence history characteristic node pointed to by each directed connection edge is taken as another influence factor; The correlation strength weight, the final deviation degree influence factor and the historical occurrence frequency influence factor are multiplied to obtain a comprehensive occurrence index for each possible pest and disease occurrence history characteristic node; All pest and disease occurrence history characteristic nodes are sorted in descending order of the comprehensive occurrence index; According to a preset index threshold range, the comprehensive occurrence index is mapped to different level labels, and the level labels include high possibility, medium possibility and low possibility, thereby generating a pest and disease occurrence possibility level.

8. A vegetable growth cycle pest and disease prevention decision system based on a knowledge graph, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor, when executing the computer program, implements the steps of the pest and disease control decision method based on the knowledge graph for a vegetable growth cycle according to any one of claims 1 to 7.

Images