Crop disease and pest and environment correlation mining method based on knowledge graph

Abstract

This invention discloses a method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs, relating to the fields of crop disease and pest control and multi-source data processing. The specific steps of this method are as follows: collecting and initially organizing multi-source heterogeneous data related to crop diseases and pests and the environment, removing invalid data, and standardizing the format; performing noise reduction, redundancy removal, and standardization on the data, and filtering effective related data using association credibility and dynamic association threshold calculation formulas; achieving data alignment through semantic mapping, similarity calculation, and deviation correction to form a collaborative data matrix; constructing a knowledge graph based on this matrix, mining the inherent relationship between the two, and using incremental processing technology to achieve real-time updates of knowledge and mining results; this invention filters effective data through association verification formulas, achieves automatic data alignment by combining semantic mapping and deviation correction, and uses incremental processing technology to solve the problems of large deviations and cumbersome operations in traditional mining methods.

Description

Technical Field

[0001] This invention relates to the field of crop pest and disease control and multi-source data processing technology, specifically a method for mining the relationship between crop pests and diseases and the environment based on knowledge graphs. Background Technology

[0002] The occurrence and spread of crop diseases and pests are directly influenced by environmental factors such as temperature, humidity, and light. Accurately identifying the correlation between these factors is a crucial prerequisite for precise pest and disease control and reducing production losses in agricultural production. Currently, the agricultural sector can acquire multi-source heterogeneous data, including environmental values, pest and disease images, crop types, and growth cycles, through environmental sensors, industrial cameras, and agricultural production records. Knowledge graphs, with their powerful relational expression and mining capabilities, have been applied to the field of relational mining, integrating multi-source data and uncovering inherent connections.

[0003] Traditional methods for mining the correlation between crop diseases and pests and the environment have two prominent problems. On the one hand, the data preprocessing stage lacks verification of the rationality of the correlation. It can only perform simple noise reduction and redundancy removal on multi-source data, and cannot identify erroneous correlations between irrelevant environmental factors and disease and pest data. This leads to invalid data being mixed into the subsequent processing flow, ultimately causing a large deviation in the correlation mining results and failing to accurately reflect the true correlation between the two. On the other hand, the alignment of multi-source heterogeneous data relies on manual operation, which not only consumes a lot of manpower but is also prone to human error. Moreover, when the mining model incorporates new data, the entire model and algorithm framework need to be reconstructed, which is cumbersome and time-consuming. It cannot adapt to the actual needs of real-time data addition in agricultural production, and its practicality is greatly reduced. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs. This method collects multi-source heterogeneous data related to crop diseases and pests and the environment and performs preliminary processing. After noise reduction, redundancy removal and standardization, effective data is screened by combining association verification formulas. Then, data alignment is achieved through semantic mapping, similarity calculation and deviation correction. Based on the aligned data, a knowledge graph is constructed to mine the inherent relationship between the two. With incremental processing technology, real-time updates are achieved, which effectively solves the problems of large deviation and cumbersome operation of traditional methods, improves the accuracy and continuity of association mining, adapts to the actual needs of agricultural production, and the overall process is coherent and repeatable.

[0005] To solve the above-mentioned technical problems, this invention provides the following technical solution: a method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs, the specific steps of which are as follows:

[0006] S1, Data Acquisition and Preliminary Processing: Collect multi-source heterogeneous data related to crop diseases, pests and the environment, perform preliminary processing on the multi-source heterogeneous data, and then transmit it to the anomaly correlation verification preprocessing stage.

[0007] S2, Anomaly Correlation Verification Preprocessing: The collected multi-source heterogeneous data is preprocessed by noise reduction, redundancy removal and standardization. The correlation credibility between various environmental factors and pest data is calculated by multi-dimensional weighted correlation credibility calculation formula. The dynamic correlation threshold is calculated by dynamic correlation threshold calculation formula. Based on the comparison results of correlation credibility and dynamic correlation threshold, the misalignment of correlation between irrelevant environmental factors and pest data is automatically identified and corrected, and valid correlation data is screened.

[0008] S3, Adaptive Alignment of Multi-Source Heterogeneous Data: Performs unified semantic mapping on the filtered effective associated data, calculates the alignment similarity between any two types of heterogeneous data using the semantic alignment similarity calculation formula of multi-source heterogeneous data, calculates the alignment deviation correction amount using the adaptive alignment deviation correction calculation formula of multi-source data, automatically completes the alignment of the time axis and numerical dimension of various types of heterogeneous data, and outputs a collaborative data matrix.

[0009] S4, Knowledge Graph Construction and Association Mining: Based on the collaborative data matrix, a knowledge graph of crop diseases and pests and their relationship with the environment is constructed. The association mining algorithm is used to dynamically mine the relationship between diseases and pests and environmental factors, generate the association mining results and output them.

[0010] Furthermore, the multi-source heterogeneous data in S1 includes environmental numerical data collected by environmental sensors, crop pest and disease image feature data, and agricultural text parameter data. The environmental numerical data includes temperature, humidity, light, and pollutant concentration data, while the agricultural text parameter data includes crop type and growth cycle parameters.

[0011] Furthermore, the formula for calculating the multi-dimensional weighted association credibility in S2 is as follows: The meanings of the letters in the parameters are as follows: For the first Class 1 environmental factors and the first The reliability of the association between pest and disease data For the first Scene adaptation weights based on environmental factors. For the first Sampling data of environmental factors, For the first The average value of all sampled data for environmental factors. For the first Sampling data for pest and disease data, For the first The average value of all sampled data for pest and disease data. This is the abnormal penalty coefficient. For the first Standard deviation of environmental factor sampling data For the first Maximum permissible standard deviation for environmental factors.

[0012] Furthermore, the formula for calculating the dynamic correlation threshold in S2 is as follows: The meanings of the letters in each parameter are as follows: For dynamic association thresholds, Based on the correlation threshold, For threshold adaptive adjustment coefficient, This represents the average value of pest and disease data. This represents the minimum allowable fluctuation value for pest and disease data. This represents the maximum allowable fluctuation value for pest and disease data.

[0013] Furthermore, the judgment logic for misalignment in S2 is as follows: when the absolute value of the association credibility is less than the dynamic association threshold, it is judged as an irrelevant association, and the association relationship and corresponding data are automatically removed; when the absolute value of the association credibility is not less than the dynamic association threshold, it is judged as a valid association, and the corresponding data is retained.

[0014] Furthermore, the formula for calculating the semantic alignment similarity of multi-source heterogeneous data in S3 is as follows: The meanings of the letters in each parameter are as follows: For the first heterogeneous data and the first Semantic alignment similarity of heterogeneous data For the first Adaptive weights for each semantic feature dimension. For semantic mapping functions, For the first heterogeneous data in the first Data in one semantic feature dimension For the first heterogeneous data in the first Data in one semantic feature dimension This is the time synchronization penalty coefficient. For the first heterogeneous data and the first Timestamp discrepancy in heterogeneous data It is the absolute value symbol.

[0015] Furthermore, the multi-source data adaptive alignment deviation correction in S3 is as follows: The meanings of the letters in each parameter are as follows: For the first heterogeneous data and the first Alignment bias correction for heterogeneous data For the first The original encoded values ​​of heterogeneous data For the first The original encoded values ​​of heterogeneous data For the first heterogeneous data and the first Alignment similarity of heterogeneous data To achieve maximum alignment similarity, For the first heterogeneous data and the first Data type compatibility coefficient for heterogeneous data.

[0016] Furthermore, the association mining algorithm in S4 adopts incremental electronic digital data processing technology. Specifically, the newly added crop pests and diseases and environmental data are processed sequentially through S2 abnormal association verification preprocessing and S3 multi-source heterogeneous data adaptive alignment processing, and then integrated into the constructed crop pests and diseases and environmental association knowledge graph. Only the association nodes and association edges related to the newly added data in the knowledge graph are updated. At the same time, the association mining algorithm parameters are adjusted based on the updated knowledge graph. The whole process does not require reconstruction of the entire mining model and algorithm framework.

[0017] Compared with existing technologies, this knowledge graph-based method for mining the relationship between crop diseases, pests, and the environment has the following advantages:

[0018] I. This invention, through a preprocessing stage of anomaly correlation verification, combines a multi-dimensional weighted correlation credibility calculation formula and a dynamic correlation threshold calculation formula to accurately determine the correlation between various environmental factors and pest data. It can automatically identify and eliminate misalignments between irrelevant environmental factors and pest data, filtering out more targeted and effective correlation data. This solves the problem of missing correlation verification in traditional data mining methods, avoids irrelevant data interfering with subsequent processing, and ensures that the mining results truly reflect the intrinsic correlation between pests and environmental factors, eliminating the need for subsequent manual correction of biases and improving the accuracy of correlation mining.

[0019] Second, this invention achieves automatic alignment of three types of heterogeneous data—environmental numerical data, pest and disease images, and agricultural text—by designing a dedicated semantic mapping function, semantic alignment similarity calculation formula, and adaptive deviation correction calculation formula. This allows for synchronized matching of the time axis and numerical dimensions without manual intervention. Furthermore, by combining incremental electronic digital data processing technology, newly added data, after preprocessing and alignment, can be directly integrated into the existing knowledge graph, requiring only updates to relevant nodes and algorithm parameters, without reconstructing the entire mining model. This solves the problems of reliance on manual data alignment and cumbersome model updates in traditional methods, reducing manpower input, lowering model maintenance complexity, and ensuring the continuity of the association mining process.

[0020] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0022] Figure 1 This is an overall flowchart of the knowledge graph-based method for mining the relationship between crop diseases and pests and the environment according to the present invention.

[0023] Figure 2 This is a detailed flowchart of the anomaly correlation verification preprocessing step in this invention;

[0024] Figure 3 This is a detailed flowchart of the adaptive alignment process for multi-source heterogeneous data in this invention. Detailed Implementation

[0025] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0026] This embodiment discloses a knowledge graph-based method for mining the relationship between crop diseases and pests and the environment. The core of the method involves four consecutive and progressive steps: data collection and preliminary processing, abnormal association verification and preprocessing, adaptive alignment of multi-source heterogeneous data, and knowledge graph construction and association mining. Combined with a specially designed algorithm formula, it can achieve accurate mining and real-time updating of the relationship between crop diseases and pests and environmental factors, solve the drawbacks of traditional mining methods, and adapt to the actual application scenarios of agricultural production.

[0027] The overall process is as follows Figure 1 As shown, the specific implementation process is as follows:

[0028] First, the S1 data acquisition and preliminary processing steps are performed. During the data acquisition process, a multi-device collaborative acquisition method is adopted to simultaneously collect three types of core heterogeneous data: environmental numerical data collected by environmental sensors, crop pest and disease image feature data collected by industrial cameras, and agricultural text parameter data formed by agricultural production records. These three types of data cover the core information required for crop pest and disease and environmental correlation analysis, ensuring the comprehensiveness of the data.

[0029] After data collection is completed, preliminary processing is immediately carried out, focusing on the completeness and usability of the data. First, all types of data are screened one by one to remove obviously invalid data, including data missing core information, data that is obviously outside the reasonable range, and data collected repeatedly, to ensure that the retained data can be used for subsequent processing and analysis. Then, the format of the screened data is standardized, and the different formats of data output from different acquisition devices are adjusted to a standard format that is easy to transmit and process in a unified manner, eliminating processing obstacles caused by format differences. Finally, all the processed multi-source heterogeneous data are summarized to form a unified raw data set, which is then transmitted to the anomaly correlation verification preprocessing stage. At this point, step S1 is completed, making full preparations for the subsequent step S2.

[0030] After step S1 is completed, the next step is the S2 anomaly correlation verification preprocessing step. This step is crucial for improving the accuracy of the data mining results. Its core is to deeply optimize and filter the raw, multi-source, heterogeneous data transmitted in S1. Through correlation rationality verification, environmental data unrelated to crop diseases and pests, as well as misaligned data, are removed, while valid correlation data closely related to diseases and pests are retained. The specific implementation process of this step is as follows: Figure 2 As shown, this reduces the interference of invalid data on subsequent processing from the source, improving the large mining deviation caused by the lack of correlation verification in traditional mining methods.

[0031] In practice, the received raw, multi-source heterogeneous data is first subjected to three processes in sequence: noise reduction, redundancy removal, and standardization. These three processes proceed in an orderly and progressive manner: noise reduction is used to eliminate noise data generated during data acquisition due to equipment errors and environmental interference, ensuring the authenticity and accuracy of the data and preventing noise data from affecting the subsequent correlation calculation results; redundancy removal is used to further remove redundant information hidden in the data, including redundant feature information and duplicate correlation information, reducing the amount of data and improving the efficiency of subsequent data processing; standardization is used to eliminate the differences in dimensions and numerical ranges between different types of heterogeneous data, uniformly mapping all types of data to the same numerical range, allowing different types of data to be compared and calculated, laying the foundation for subsequent correlation credibility calculation and correlation rationality judgment.

[0032] After the above three basic processing steps are completed, the core part of the association rationality verification begins. Since the degree of association between various environmental factors and crop pests and diseases varies, and some environmental factors have no obvious association with pests and diseases, using them directly in subsequent processing without verification would lead to biased data mining results. Therefore, a multi-dimensional weighted association credibility calculation formula is used to quantify the closeness of the association between various environmental factors and crop pest and disease data, providing a quantitative basis for determining the rationality of the association. The calculation formula is as follows: ,in For the i-th type of environmental factor and the i-th type of environmental factor The reliability of the association between pest and disease data As a core quantitative indicator, it is used to accurately quantify the first Class 1 environmental factors and the first The degree of correlation between pest and disease data. The magnitude of the value directly reflects the strength of the relationship between the two, providing a direct basis for determining the rationality of the subsequent relationship; Set as number The scenario adaptation weight for environmental factors is set because different types of environmental factors have varying degrees of impact on the occurrence and spread of crop diseases and pests. It can distinguish the importance of various environmental factors, assign higher weights to environmental factors with high impact and lower weights to those with low impact, ensuring the accuracy of the correlation credibility calculation results and reflecting the differences in the impact of different environmental factors in actual agricultural production. For the first The sampling data of environmental factors is one of the basic data for calculating the correlation confidence, and it comes directly from the environmental numerical data after processing in step S1; For the first The average value of all sampled data for a class of environmental factors is used to eliminate the randomness of a single sampled data, reflect the overall level of that class of environmental factors, and make the correlation calculation results more representative. For the first Sampling data of pests and diseases, and Correspondingly, the pest and disease related data are directly derived from the data compiled in step S1; For the first The average of all sampled data for pest and disease data, and The function is consistent, used to reflect the overall level of this type of pest and disease data and improve the rationality of correlation calculation; Set as the anomaly penalty coefficient. This parameter is set to penalize sampled data that deviates from the normal range, so as to avoid deviations in the correlation confidence calculation results due to individual abnormal sampled data and ensure the stability of the calculation results. For the first The standard deviation of the sampling data for a class of environmental factors is used to reflect the dispersion of the sampling data for that class of environmental factors and to reflect the stability of the data; For the first The maximum permissible standard deviation of environmental factors is used as a benchmark for judging the stability of environmental factor sampling data. and When used in combination, the stability of sampling data for this type of environmental factor can be accurately determined, and the correlation confidence calculation results can be further optimized to ensure that the calculation results can truly reflect the degree of correlation between the two.

[0033] Relying solely on association credibility is insufficient to accurately determine the rationality of association relationships. Using a fixed threshold for judgment fails to adapt to varying pest and disease data fluctuations, potentially leading to misjudgments of irrelevant associations or omissions of valid ones. Therefore, a dynamic association threshold calculation formula is employed to calculate a threshold suitable for the current data scenario, serving as the standard for determining the rationality of associations. This formula is as follows: ,in The dynamic correlation threshold is used as the criterion for judgment. Class 1 environmental factors and the first The core standard for the rationality of the correlation between pest and disease data is dynamically adjusted according to the fluctuation of pest and disease data to ensure the adaptability of the judgment standard. The basic association threshold is the benchmark parameter for calculating the dynamic association threshold, providing a basic reference for threshold calculation and ensuring the rationality and stability of the threshold. The threshold adaptive adjustment coefficient is set so that the dynamic correlation threshold can be adaptively adjusted according to the fluctuation of pest and disease data, adapting to different data fluctuation scenarios and avoiding judgment deviation caused by a fixed threshold. The average value of the pest and disease data is used to reflect the overall fluctuation level of all current pest and disease data and is the core reference for threshold adjustment. This represents the minimum allowable fluctuation value for pest and disease data. These two values ​​represent the maximum permissible fluctuation range for pest and disease data. Combined, they define the reasonable fluctuation range for pest and disease data, reflecting the current fluctuation status of the data. , , The three elements work together to ensure that the adjustment of the dynamic association threshold can accurately adapt to the current data scenario, making the determination of the rationality of association more targeted and accurate.

[0034] Relevance and dynamic association threshold After the calculation is completed, the reasonableness of the association is verified according to the established judgment logic. The judgment logic revolves around the numerical comparison between the two. Specifically, when the absolute value of the association confidence is less than the dynamic association threshold, it means that there is no effective association between this type of environmental factor and this type of pest data, and it is judged as an irrelevant association. The association relationship and all corresponding data are automatically removed to avoid invalid data from being mixed into the subsequent processing flow. When the absolute value of the association confidence is not less than the dynamic association threshold, it means that there is a effective association between this type of environmental factor and this type of pest data, and it is judged as a valid association. The corresponding data and association relationship are retained. After all data verification is completed, all valid association data are summarized to form a valid data set, which is then uniformly transmitted to the multi-source heterogeneous data adaptive alignment stage. Step S2 is now complete, providing high-quality valid data for the subsequent step S3.

[0035] After step S2 is completed, the S3 multi-source heterogeneous data adaptive alignment step is executed. The core purpose of this step is to solve the problem of inconsistent formats, semantics, time axes, and numerical dimensions among different types of heterogeneous data in the effective correlated data transmitted in S2, and to achieve synchronous matching and standardization of various types of heterogeneous data. The detailed alignment process of this step is as follows: Figure 3 As shown, this approach enables various types of data to be collaboratively used for subsequent knowledge graph construction and association mining, improving upon the traditional methods where data alignment relies on manual intervention and suffers from significant errors. In practice, the core pain point is first identified: the effective association data transmitted by S2 encompasses three heterogeneous data types: environmental numerical data, crop pest and disease image feature data, and agricultural text parameter data. These data types exhibit significant differences in format, semantic expression, and numerical dimensions, making them unsuitable for direct use in knowledge graph construction and association analysis. Therefore, a unified semantic mapping process is first applied to all effective association data using a semantic mapping function. The core function of this function is to map heterogeneous data of different types and semantics to the same semantic space, convert various types of data into standardized data with a unified semantic dimension, eliminate integration barriers caused by semantic differences, and enable various types of heterogeneous data to be analyzed collaboratively.

[0036] After semantic mapping, all types of data possess a unified semantic dimension. However, discrepancies still exist in the time axis and numerical dimensions between different types of data. Further calculation of the alignment similarity between different types of data is needed to determine the degree of alignment matching between the two types of data, providing a basis for subsequent deviation correction. Therefore, a multi-source heterogeneous data semantic alignment similarity calculation formula is adopted to quantify the degree of alignment matching between any two types of heterogeneous data. The formula is as follows: ,in For the first heterogeneous data and the first Semantic alignment similarity of heterogeneous data is a core indicator for quantifying the degree of alignment and matching between two types of data. Its value directly reflects the alignment effect between the two types of data and provides a quantitative basis for subsequent deviation correction. For the first The adaptive weights for each semantic feature dimension are set because different semantic feature dimensions have varying impacts on data alignment and matching. Distinguish the importance of each semantic feature dimension, assign higher weights to dimensions with high influence, optimize similarity calculation results, and ensure that the calculation results can truly reflect the degree of alignment and matching between the two types of data. For the first heterogeneous data in the first The data for each semantic feature dimension is standardized data after semantic mapping, which comes directly from the semantic mapping processing results mentioned above, and serves as the basic data for similarity calculation. For the first heterogeneous data in the first Data of semantic feature dimensions, and Correspondingly, the results of semantic mapping are also combined to calculate the degree of matching between the two types of data under the same semantic feature dimension, and then the overall alignment similarity is obtained by summarizing them. The time synchronization penalty coefficient is set to penalize data with large timestamp deviations, ensuring that all types of data can be synchronized and aligned on the timeline, and avoiding deviations in subsequent correlation mining results due to time deviations. For the first heterogeneous data and the first The timestamp deviation of heterogeneous data is used to reflect the differences between the two types of data in the time dimension; The absolute value symbol is used to convert timestamp deviations into non-negative values, ensuring the rationality of penalty calculations and avoiding calculation errors caused by negative deviations.

[0037] Obtain the semantic alignment similarity of the two types of heterogeneous data. Subsequently, the alignment matching degree and deviation direction of the two types of data were clarified. In order to correct the alignment deviation between the two types of data and to ensure that the time axis and numerical dimensions of various heterogeneous data are completely synchronized to meet the standardization requirements, a multi-source data adaptive alignment deviation correction form was adopted to calculate the alignment deviation correction amount of the two types of heterogeneous data and guide the data alignment correction operation. The correction form is as follows: , For the first heterogeneous data and the first The core function of the alignment deviation correction for heterogeneous data is to provide a clear quantitative basis for the alignment correction of two types of data, guide operators or equipment to accurately correct the deviation between the two types of data, and ensure that the aligned data is completely synchronized. For the first The original encoded values ​​of heterogeneous data are one of the basic data for bias correction, and they come from the standardized data after semantic mapping. For the first The original encoded values ​​of heterogeneous data, and Correspondingly, the standardized data, which also comes from semantic mapping, together serve as the basis for bias correction, ensuring the accuracy of the correction calculation. The alignment similarity between the two types of data is used to adjust the deviation correction magnitude based on the degree of matching between the two types of data. The higher the degree of matching, the smaller the correction magnitude, and the lower the degree of matching, the larger the correction magnitude, to ensure the rationality of deviation correction and avoid over-correction or under-correction. The maximum alignment similarity serves as a benchmark parameter for similarity, used to standardize the calculation range of the correction magnitude, ensuring that the correction amount is within a reasonable range and avoiding correction deviation. For the first heterogeneous data and the first This parameter, the data type adaptation coefficient for heterogeneous data, is set because different types of heterogeneous data have different encoding rules and numerical ranges. It adapts to the encoding differences of different types of data, ensures the accuracy of deviation correction, and enables the two types of data to be fully adapted after correction, achieving the requirement of synchronous alignment.

[0038] Following the aforementioned deviation correction process, all heterogeneous data are aligned and corrected one by one. During the correction process, the alignment similarity of various data types is calculated in real time to determine the correction effect, until the alignment similarity between all heterogeneous data reaches the preset standard, completing the alignment operation of the time axis and numerical dimensions of various heterogeneous data. After alignment, all aligned standardized data are summarized, and a collaborative data matrix is ​​constructed according to a preset format. The collaborative data matrix contains all valid, standardized data related to crop diseases and pests, as well as preliminary association information between various data types. Subsequently, the collaborative data matrix is ​​transmitted to the knowledge graph construction and association mining stage. Step S3 is now complete, providing standardized and collaborative data for the subsequent knowledge graph construction and association mining in step S4.

[0039] After step S3 is completed, the final step S4, knowledge graph construction and association mining, is executed. This step is the core of the entire mining method. The core task is to construct a knowledge graph of crop diseases and pests and the environment based on the collaborative data matrix transmitted in S3, mine the inherent relationship between the two through association mining algorithms, and realize the real-time update of association knowledge and mining results to complete the entire association mining process, while adapting to the actual needs of real-time data addition in agricultural production.

[0040] In practice, a knowledge graph relating crop diseases and pests to the environment is first constructed. This construction revolves around a collaborative data matrix, with each type of data in the matrix serving as a node in the knowledge graph. Environment-related data corresponds to environment nodes, crop disease and pest-related data to disease and pest nodes, and agricultural text parameter data to auxiliary nodes. Each type of node contains complete attribute information to ensure the integrity of the node information. Subsequently, the relationships between different types of data in the collaborative data matrix are used as edges in the knowledge graph. The weights of these edges are determined by the association confidence calculated in step S2. It is determined that the higher the credibility of the association, the greater the weight of the association edge, which intuitively reflects the degree of association between various nodes. Through the construction of nodes and association edges, a complete knowledge graph of crop diseases and pests and the environment is formed. This knowledge graph can clearly and intuitively present the relationship between various environmental factors and crop diseases and pests, providing a visualized and structured knowledge reference for subsequent association mining.

[0041] After the knowledge graph is constructed, the association mining algorithm is activated. This algorithm operates around the constructed association knowledge graph. Its core is to analyze the attribute information of various associated nodes and the weights of associated edges in the knowledge graph to mine the intrinsic associations between environmental factors and crop diseases and pests, including direct and indirect associations. During the mining process, the association credibility in step S2 and the alignment similarity in step S3 are combined to further optimize the mining results, ensuring that the mining results can truly and accurately reflect the intrinsic associations between the two. After the mining is completed, the association mining results are generated. These results contain core information such as the association type and association strength between various environmental factors and crop diseases and pests, which can be used for subsequent disease and pest control-related analysis and applications.

[0042] Meanwhile, to adapt to the actual needs of real-time updates of crop pest and disease and environmental data in agricultural production, the association mining algorithm adopts incremental electronic digital data processing technology. The specific application of this technology is as follows: When new crop pest and disease and environmental data are added, the entire mining model and knowledge graph are not directly reconstructed. Instead, the new data is first processed through the S2 anomaly association verification preprocessing step and the S3 multi-source heterogeneous data adaptive alignment step. This process involves noise reduction, redundancy removal, standardization, association verification, semantic mapping, and alignment correction of the new data. Valid association data is then selected from the new data and converted into standardized data. Subsequently, the processed standardized new data is integrated into the already constructed crop pest and disease and environmental association knowledge. The graph update only updates the associated nodes and edges in the knowledge graph related to the newly added data. This includes adding related nodes, updating the attribute information of existing nodes, adding associated edges, or adjusting the weights of existing associated edges. It does not change other nodes and associated edges in the knowledge graph. At the same time, it adjusts the relevant parameters of the association mining algorithm around the updated association knowledge graph to adapt the algorithm to the updated knowledge graph. It re-mines the changes in association relationships brought about by the new data, realizing real-time updates of association mining results. The entire process does not require reconstruction of the entire mining model and algorithm framework, simplifying the update operation, improving the update efficiency, and ensuring that the association mining results can reflect the latest crop pest and disease and environmental correlation in a timely manner, adapting to the actual needs of agricultural production.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs, characterized in that, The specific steps of this method are as follows: S1, Data Acquisition and Preliminary Processing: Collect multi-source heterogeneous data related to crop diseases, pests and the environment, perform preliminary processing on the multi-source heterogeneous data, and then transmit it to the anomaly correlation verification preprocessing stage. S2, Anomaly Correlation Verification Preprocessing: The collected multi-source heterogeneous data undergoes noise reduction, redundancy removal, and standardization preprocessing. The correlation credibility between various environmental factors and pest data is calculated using a multi-dimensional weighted correlation credibility calculation formula. The multi-dimensional weighted correlation credibility calculation formula is as follows: The meanings of the letters in the parameters are as follows: For the first Class 1 environmental factors and the first The reliability of the association between pest and disease data. For the first Scene adaptation weights based on environmental factors. For the first Sampling data of environmental factors, For the first The average value of all sampled data for environmental factors. For the first Sampling data for pest and disease data, For the first The average value of all sampled data for pest and disease data. This is the abnormal penalty coefficient. For the first Standard deviation of environmental factor sampling data For the first Maximum permissible standard deviation for environmental factors; The dynamic association threshold is calculated using the dynamic association threshold calculation formula, which is as follows: The meanings of the letters in each parameter are as follows: For dynamic association thresholds, Based on the correlation threshold, For threshold adaptive adjustment coefficient, This represents the average value of pest and disease data. This represents the minimum allowable fluctuation value for pest and disease data. This represents the maximum allowable fluctuation value for pest and disease data. Based on the comparison between the correlation confidence and the dynamic correlation threshold, the system automatically identifies and corrects misalignments between irrelevant environmental factors and pest and disease data. When the absolute value of the correlation confidence is less than the dynamic correlation threshold, it is determined to be an irrelevant correlation, and the correlation and corresponding data are automatically removed. When the absolute value of the correlation confidence is not less than the dynamic correlation threshold, it is determined to be a valid correlation. S3, Adaptive Alignment of Multi-Source Heterogeneous Data: Performs unified semantic mapping on the filtered effective associated data, calculates the alignment similarity between any two types of heterogeneous data using the semantic alignment similarity calculation formula of multi-source heterogeneous data, calculates the alignment deviation correction amount using the adaptive alignment deviation correction calculation formula of multi-source data, automatically completes the alignment of the time axis and numerical dimension of various types of heterogeneous data, and outputs a collaborative data matrix. S4, Knowledge Graph Construction and Association Mining: Based on the collaborative data matrix, a knowledge graph of crop diseases and pests and their relationship with the environment is constructed. The association mining algorithm is used to dynamically mine the relationship between diseases and pests and environmental factors, generate the association mining results and output them.

2. The method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs according to claim 1, characterized in that, The multi-source heterogeneous data in S1 includes environmental numerical data collected by environmental sensors, crop disease and pest image feature data, and agricultural text parameter data. The environmental numerical data includes temperature, humidity, light intensity, and pollutant concentration data, while the agricultural text parameter data includes crop type and growth cycle parameters.

3. The method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs according to claim 1, characterized in that, The formula for calculating the semantic alignment similarity of multi-source heterogeneous data in S3 is as follows: The meanings of the letters in each parameter are as follows: For the first heterogeneous data and the first Semantic alignment similarity of heterogeneous data For the first Adaptive weights for each semantic feature dimension For semantic mapping functions, For the first heterogeneous data in the first Data in one semantic feature dimension For the first Heterogeneous data is data in the first semantic feature dimension. This is the time synchronization penalty coefficient. For the first heterogeneous data and the first Timestamp discrepancy in heterogeneous data It is the absolute value symbol.

4. The method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs according to claim 1, characterized in that, The multi-source data adaptive alignment deviation correction in S3 is as follows: The meanings of the letters in each parameter are as follows: For the first heterogeneous data and the first Alignment bias correction for heterogeneous data For the first The original encoded values ​​of heterogeneous data For the first The original encoded values ​​of heterogeneous data For the first heterogeneous data and the first Alignment similarity of heterogeneous data To achieve maximum alignment similarity, For the first heterogeneous data and the first Data type compatibility coefficient for heterogeneous data.

5. The method for mining the relationship between crop diseases and pests and the environment based on knowledge graphs according to claim 1, characterized in that, The association mining algorithm in S4 adopts incremental electronic digital data processing technology. Specifically, the newly added crop pests and diseases and environmental data are processed sequentially through S2 abnormal association verification preprocessing and S3 multi-source heterogeneous data adaptive alignment processing, and then integrated into the constructed crop pests and diseases and environmental association knowledge graph. Only the association nodes and association edges related to the newly added data in the knowledge graph are updated. At the same time, the association mining algorithm parameters are adjusted based on the updated knowledge graph. The whole process does not require reconstruction of the entire mining model and algorithm framework.

Images