A comprehensive agricultural monitoring system based on big data and its early warning and prediction methods

Abstract

This invention proposes a comprehensive agricultural four-condition monitoring system based on big data and its early warning and prediction method. It belongs to the field of smart agriculture technology. The method includes: performing multi-source sensor spatiotemporal alignment processing on the agricultural four-condition monitoring area to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition, and climate; constructing an agricultural causal knowledge graph based on this dataset; dynamically and adaptively adjusting thresholds according to the agricultural causal knowledge graph and initiating multimodal data acquisition to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; performing joint noise reduction processing on the original agricultural four-condition monitoring data using physical models and data-driven methods to generate high-quality agricultural four-condition feature data; and using this early warning and prediction method of the comprehensive agricultural four-condition monitoring system based on big data, the accuracy of agricultural risk early warning can be greatly improved, allowing farmers to promptly grasp potential risks in agricultural production.

Description

Technical Field

[0001] This invention proposes a comprehensive monitoring system for four agricultural conditions based on big data and its early warning and prediction method, belonging to the field of smart agriculture technology. Background Technology

[0002] With the booming development of smart agriculture, monitoring and early warning of the four agricultural conditions (soil moisture, insect infestation, seedling growth, and climate) are crucial for ensuring agricultural production. However, current agricultural monitoring and early warning technologies have many shortcomings that urgently need to be addressed.

[0003] In terms of data fusion, simple weighted averages or rule-based engines are often used, failing to delve into the spatiotemporal causal relationships between soil moisture, pest infestations, seedling conditions, and climate. This results in low data utilization efficiency and an inability to provide strong support for accurate early warning. Regarding the early warning logic, reliance on fixed thresholds, such as "soil moisture <30% equals drought," makes it difficult to adapt to differences in different crops, growth stages, and regional ecosystems, leading to significant discrepancies between warning results and actual conditions. The models suffer from poor interpretability; while black-box AI models can output risk levels, they cannot explain the causes of the risks, making it difficult for farmers to trust them and take effective action. The systems are closed and non-scalable; adding new sensors often requires retraining the model, lacking transfer learning capabilities and increasing the cost of system upgrades and maintenance.

[0004] In addition, existing patents mostly focus on general architectures such as "four conditions in one", "real-time monitoring" and "APP early warning", failing to deeply integrate early warning and prediction functions with causal reasoning, physical models and dynamic evolution modeling, and to build a cognitive-level agricultural decision-making operating system with "causal explanation - risk deduction - strategy generation", making it difficult to achieve a leap in agricultural early warning paradigm. Summary of the Invention

[0005] This invention provides a comprehensive agricultural four-condition monitoring system based on big data and its early warning and prediction method to solve the problems mentioned in the background art above: The present invention proposes an early warning and prediction method for a comprehensive agricultural four-condition monitoring system based on big data, the method comprising: S1. Perform spatiotemporal alignment processing on multi-source sensors in the agricultural four-condition monitoring area to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition and climate; construct an agricultural causal knowledge graph based on this dataset; S2. Based on the agricultural causal knowledge graph, perform dynamic threshold adaptive adjustment and initiate multimodal data acquisition to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; perform joint noise reduction processing of the original agricultural four-condition monitoring data using physical models and data-driven methods to generate high-quality agricultural four-condition feature data; S3. Based on high-quality agricultural four-condition characteristic data, perform dual-drive reasoning processing to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; resolve conflicts and fuse evidence through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. S4. Dynamically divide the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; use agricultural risk attribution analysis data to simulate risk propagation in the agricultural risk evolution grid model to generate multi-step risk projection data; process the multi-step risk projection data to generate agricultural risk response strategy data. S5. Calculate the weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index. Based on the comprehensive agricultural risk index, initiate a regional transfer learning mechanism to update new knowledge graph nodes online and generate agricultural disaster early warning data with transfer learning capabilities.

[0006] The present invention proposes a comprehensive agricultural four-condition monitoring system based on big data. This monitoring system is used to implement the early warning and prediction method described above. The system includes: The knowledge graph construction module performs spatiotemporal alignment processing on the agricultural four-condition monitoring area using multi-source sensors to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition, and climate; and constructs an agricultural causal knowledge graph based on this dataset. Data acquisition module: Based on the agricultural causal knowledge graph, dynamic threshold adaptive adjustment is performed, and multimodal data acquisition is initiated to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; the original agricultural four-condition monitoring data is subjected to joint noise reduction processing of physical model and data-driven approach to generate high-quality agricultural four-condition feature data; Evidence fusion module: Based on high-quality agricultural four-condition characteristic data, dual-drive reasoning is performed to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; conflict resolution and evidence fusion of the dual-drive reasoning results are performed through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. Strategy generation module: Dynamically divides the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; simulates risk propagation on the agricultural risk evolution grid model using agricultural risk attribution analysis data to generate multi-step risk projection data; and processes the multi-step risk projection data to generate agricultural risk response strategy data. Evolution Update Module: Calculates a weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index; initiates a regional transfer learning mechanism based on the comprehensive agricultural risk index to perform online evolution updates on new knowledge graph nodes, generating agricultural disaster early warning data with transfer learning capabilities.

[0007] The beneficial effects of this invention are as follows: The early warning and prediction method of this big data-based comprehensive agricultural four-condition monitoring system can greatly improve the accuracy of agricultural risk early warning, allowing farmers to promptly grasp potential risks in agricultural production. Simultaneously, it effectively reduces the false alarm rate, minimizing false alarms and avoiding unnecessary resource investment and energy consumption for farmers. This method enhances the interpretability of early warning results, clearly presenting the causes of risks through causal knowledge graphs and dual-drive reasoning, enabling farmers to understand and trust early warning information and take effective countermeasures. It reduces system upgrade and maintenance costs; its regional migration and online evolution capabilities eliminate the need to retrain models for new sensors, saving significant time and money. It can adapt to differences in crops, growth stages, and regional ecology, achieving dynamic threshold adaptive adjustment, and can deeply explore the spatiotemporal causal relationships between soil moisture, insect infestation, seedling conditions, and climate, constructing a cognitive-level agricultural decision-making operating system. This avoids the inaccuracies in early warnings caused by superficial data fusion and static early warning logic in traditional technologies, providing a solid guarantee for the development of smart agriculture. Attached Figure Description

[0008] Figure 1 This is a diagram illustrating the steps of the method described in this invention; Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0009] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0010] One embodiment of the present invention, such as Figure 1 As shown, an early warning and prediction method for a comprehensive agricultural four-condition monitoring system based on big data is described, the method comprising: S1. Perform multi-source sensor spatiotemporal alignment processing on the agricultural four-condition monitoring area to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition and climate; construct an agricultural causal knowledge graph based on the dataset, which is used to reveal the dynamic causal relationship network between soil moisture, insect infestation, seedling condition and climate. S2. Based on the agricultural causal knowledge graph, perform dynamic threshold adaptive adjustment and initiate multimodal data acquisition to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; perform joint noise reduction processing of the original agricultural four-condition monitoring data using physical models and data-driven methods to generate high-quality agricultural four-condition feature data; S3. Based on high-quality agricultural four-condition characteristic data, perform dual-drive reasoning processing to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; resolve conflicts and fuse evidence through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. S4. Dynamically divide the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; use agricultural risk attribution analysis data to simulate risk propagation in the agricultural risk evolution grid model to generate multi-step risk projection data; process the multi-step risk projection data to generate agricultural risk response strategy data. S5. Calculate the weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index. Based on the comprehensive agricultural risk index, initiate a regional transfer learning mechanism to update new knowledge graph nodes online and generate agricultural disaster early warning data with transfer learning capabilities.

[0011] The working principle and effects of the above technical solution are as follows: This solution effectively improves the accuracy of agricultural weather monitoring and the timeliness of early warning and forecasting, enhances the interpretability of risk attribution and the pertinence of response strategies, and reduces the labor intensity and cost of manual monitoring. It reduces errors caused by spatiotemporal misalignment and noise interference from multi-source data, avoids misjudgment and missed judgment of risks, and prevents agricultural losses due to untimely prevention and control. It not only achieves synchronous integration and efficient utilization of the four weather data, but also improves the early warning adaptability to different regions through online evolution and transfer learning of knowledge graphs. Simultaneously, it reduces data redundancy and invalid calculations, enhances the stability and practicality of the monitoring system, and provides reliable support for agricultural disaster prevention and control and field management.

[0012] In one embodiment of the present invention, S1 includes: S11. Collect raw sensor information from multiple sources within the agricultural monitoring area, perform spatiotemporal benchmark unification processing on the raw sensor information, and generate a sensor raw dataset with consistent spatiotemporal benchmark. S12. Perform dimension normalization and format unification processing on the original sensor dataset to generate a standardized soil moisture, insect, seedling, and climate monitoring dataset. S13. Extract the correlation attributes of various monitoring elements in the standardized monitoring dataset, construct the correlation links between elements, and generate basic correlation structure data; S14. Based on the basic association structure data, a multi-level association reasoning framework is built to generate an agricultural causal knowledge graph. The agricultural causal knowledge graph is used to display the dynamic causal relationship network between soil moisture, insects, seedlings, and climate.

[0013] The working principle and effects of the above technical solution are as follows: This solution improves the standardization and consistency of multi-source monitoring data on agricultural soil moisture, pests, and other pests, enhances the rationality of correlation analysis among various monitoring elements, and reduces errors caused by spatiotemporal misalignment and format confusion in the original sensing data. It reduces the difficulty and redundant workload of subsequent data processing, avoids analytical biases caused by inconsistent data benchmarks and ambiguous correlations, and prevents impacts on the accuracy of subsequent early warning and forecasting. It effectively integrates multi-source sensor data, forming a unified and standardized dataset from scattered data on soil moisture, pests, etc., and clearly presents the dynamic causal relationships between the four pest elements, providing solid data and graphical support for subsequent risk inference and early warning forecasting, further improving the data processing efficiency of the entire monitoring system.

[0014] In one embodiment of the present invention, S2 includes: S21. Combining the internal correlation strength of the agricultural causal knowledge graph, adaptively correct the critical range of monitoring indicators to generate a dynamically adapted indicator threshold system. S22. Start the synchronous collection of data from the multimodal sensing terminal across the entire region to generate original agricultural four-condition monitoring data covering the entire region and all time periods; S23. Synchronously collect the time tags and spatial positioning information corresponding to the original agricultural four conditions monitoring data, and generate one-to-one spatiotemporal coordinate data; S24. Perform noise reduction and purification processing on the original agricultural four-condition monitoring data by combining physical model constraints and data-driven methods to generate high-quality agricultural four-condition characteristic data.

[0015] The working principle and effects of the above technical solution are as follows: This solution improves the adaptability of monitoring indicator thresholds and the coverage of raw monitoring data, enhances the purity and usability of the four weather characteristics data, and reduces noise interference and invalid information in the raw data. It reduces monitoring bias caused by fixed thresholds and incomplete data collection, avoids the influence of noisy data on subsequent analysis results, and prevents deviations in early warning and prediction. It allows monitoring thresholds to dynamically adjust with changes in the correlation between the four weather conditions, ensuring that monitoring standards align with actual field conditions. It also enables full-area, all-time data collection and precise noise reduction, while binding spatiotemporal coordinates to make the data traceable, providing high-quality, highly adaptable basic data for subsequent dual-drive inference, further improving the reliability of the entire monitoring process.

[0016] In one embodiment of the present invention, step S21 includes: Read the internal correlation strength of the agricultural causal knowledge graph and generate quantitative data on correlation strength. Perform a full-domain indicator transmission analysis on the quantitative data of correlation strength to generate indicator correlation transmission data; Adaptive correction of critical intervals is performed on the correlation and transmission data of indicators to generate dynamically adjusted threshold data; The dynamically adjusted threshold data is normalized and integrated across the entire scenario to generate a dynamically adapted indicator threshold system.

[0017] The working principle and effects of the above technical solution are as follows: This solution improves the dynamic adaptability and overall consistency of monitoring indicator thresholds, enhances the scientific rigor and rationality of threshold adjustments, and reduces errors caused by inaccurate correlation strength analysis and one-sided threshold correction. It lowers the probability of mismatches between fixed thresholds and actual field conditions, avoiding missed or false alarms due to unreasonable threshold settings, and preventing any impact on the effectiveness of subsequent data collection and analysis. It can accurately quantify the correlation strength of the four weather elements, ensuring that threshold correction aligns with the dynamic correlation between elements, and achieves unified integration across the entire field, ensuring consistent monitoring standards in different regions. This provides a precise and adaptable basis for subsequent high-quality data collection, further enhancing the rigor of the monitoring process.

[0018] In one embodiment of the present invention, S3 includes: S31. Import high-quality agricultural four-condition characteristic data into the physical inference channel to generate deterministic inference results that conform to natural laws; S32. Import high-quality agricultural four-condition characteristic data into the statistical learning channel to generate probabilistic inference results based on data distribution; S33. Input the deterministic reasoning results and probabilistic reasoning results into the agricultural causal knowledge graph to carry out conflict identification and information fusion processing; S34. Perform hierarchical attribution decomposition on the fused results to generate interpretable agricultural risk attribution analysis data.

[0019] The working principle and effects of the above technical solution are as follows: This solution improves the accuracy and comprehensiveness of agricultural risk reasoning, enhances the interpretability and logic of risk attribution, and reduces the bias and errors caused by a single reasoning method. It reduces judgment errors caused by conflicting reasoning results and ambiguity in attribution, avoiding insufficient targeting of subsequent prevention and control strategies due to the inability to trace the causes of risks, thus preventing agricultural losses. It can ensure that the reasoning results conform to natural field laws through physical deduction, while also taking into account the objectivity of data distribution through statistical learning. Simultaneously, it resolves conflicts in dual-path reasoning, deconstructs risk causes, and makes risk analysis more convincing. This provides accurate and reliable support for subsequent risk simulation and response strategy generation, further enhancing the scientific nature of early warning and forecasting.

[0020] In one embodiment of the present invention, S33 includes: S331. Input the deterministic reasoning results and the probabilistic reasoning results into the agricultural causal knowledge graph to generate a dual-path reasoning input dataset; S332. Perform feature similarity comparison on the dual-path reasoning input dataset to generate reasoning result difference feature data; S333. Perform conflict identification operation on the difference feature data of the reasoning results to generate a set of reasoning conflict features; perform conflict resolution processing on the set of reasoning conflict features to generate conflict-free integrated reasoning data. S334. Perform multi-dimensional information fusion on the conflict-free reasoning integrated data to complete conflict identification and information fusion processing.

[0021] The working principle and effects of the above technical solution are as follows: This technical solution improves the integration quality and consistency of dual-path reasoning results, enhances the accuracy of conflict identification and the rationality of information fusion, and reduces biases caused by conflicting reasoning results and insufficient fusion. It reduces risk analysis errors caused by unresolved conflicts and cluttered information, and avoids unreliable fusion results that could affect subsequent risk attribution and prevention strategy formulation. It can comprehensively integrate deterministic and probabilistic reasoning data, accurately capture the conflicts between them, effectively resolve conflicts, and achieve multi-dimensional information fusion, making the reasoning results more complete and credible. This provides unified, high-quality basic data for subsequent hierarchical attribution decomposition, further enhancing the rigor and effectiveness of risk analysis.

[0022] In one embodiment of the present invention, S334 includes: Feature dimensions are extracted from conflict-free reasoning integrated data to generate multi-dimensional reasoning feature data; Assign feature weights to multi-dimensional inference feature data to generate a weighted inference feature set; Cross-dimensional correlation and fusion are performed on the weighted inference feature set to generate fused inference feature data; Consistency verification is performed on the fused inference feature data to generate standardized fusion results; conflict identification and information fusion processing are completed using the standardized fusion results.

[0023] The working principle and effects of the above technical solution are as follows: This solution improves the fusion accuracy and standardization of conflict-free inference data, enhances the rationality and reliability of multi-dimensional feature fusion, and reduces errors caused by incomplete feature extraction and messy fusion. It lowers the probability of deviations and inconsistencies in the fusion results, avoids non-standard fused data affecting the overall effect of subsequent conflict identification and information fusion, and prevents delays in risk analysis. It can accurately extract multi-dimensional features from conflict-free data, highlight key features through weight allocation, achieve efficient cross-dimensional fusion, and ensure standardized and unified fusion results through consistency verification. This effectively improves the quality and efficiency of information fusion processing, providing more accurate and reliable support for subsequent risk attribution and decomposition.

[0024] In one embodiment of the present invention, step S4 includes: S41. Utilize spatiotemporal coordinate data to perform refined grid segmentation of the entire agricultural monitoring area, assign independent spatiotemporal attributes to each grid, and generate an agricultural risk evolution grid model. S42. Inject agricultural risk attribution analysis data into the agricultural risk evolution grid model, conduct spatiotemporal risk diffusion simulation, and generate multi-step progressive risk projection data; S43. Based on multi-step risk simulation data, automatically generate response plans to generate agricultural risk response plans by region and time period; S44. Conduct feasibility verification and optimization of agricultural risk response plans, and generate agricultural risk response strategy data.

[0025] The working principle and effects of the above technical solution are as follows: This solution improves the accuracy of agricultural risk evolution simulation and the pertinence of response strategies, enhances the coherence of risk projection and the feasibility of solutions, and reduces prevention and control loopholes caused by one-sided risk simulations and unreasonable response plans. It reduces the workload and error rate of manually developing response plans, avoiding untimely responses and inappropriate measures due to inaccurate risk diffusion simulations, thus preventing greater agricultural losses. It enables refined management and control across the entire monitoring area, clearly presenting the spatiotemporal evolution patterns of risks, and automatically generates regional and time-specific response plans based on projection data. Furthermore, the plans are validated and optimized to ensure their implementation, making risk prevention and control more scientific and efficient, and providing strong support for the prevention and control of agricultural disasters.

[0026] In one embodiment of the present invention, S41 includes: Read spatiotemporal coordinate data, extract spatial boundary and time span information of the entire agricultural monitoring area, and generate spatiotemporal range data of the entire area; Adaptive grid size division is performed on the spatiotemporal data of the entire domain to generate refined grid segmentation parameters; The entire agricultural monitoring area is divided into grids based on refined grid segmentation parameters to generate an initial monitoring grid set; Each grid in the initial monitoring grid set is assigned a unique spatiotemporal identifier and attribute parameters to generate grid cells with attribute identifiers; By integrating grid cells with attribute labels, an agricultural risk evolution grid model is generated.

[0027] The working principle and effects of the above technical solution are as follows: This solution improves the precision and adaptability of the grid segmentation for agricultural monitoring across the entire area, enhances the standardization and practicality of the grid model, and reduces errors caused by uneven grid division and ambiguous attributes. It lowers the difficulty of subsequent risk evolution simulation, avoiding insufficient accuracy in risk simulation due to unreasonable grid size or lack of clear spatiotemporal identification, thus preventing impact on subsequent risk extrapolation and response plan generation. It can accurately capture the spatiotemporal range of the entire monitoring area, achieving adaptive grid size division to meet the monitoring needs of different regions. Furthermore, it assigns a unique identifier and attributes to each grid, making grid units traceable and controllable. The generated risk evolution grid model is more closely aligned with actual monitoring scenarios, providing a solid foundation for subsequent risk diffusion simulation.

[0028] In one embodiment of the present invention, S42 includes: Extract risk source characteristics and transmission parameters from agricultural risk attribution analysis data to generate basic risk evolution data; The basic data of risk evolution is injected into the agricultural risk evolution grid model according to the grid unit to generate the initial grid risk dataset; Spatiotemporal diffusion rule calculations are performed on the initial grid risk dataset to generate dynamic parameters for risk diffusion. Based on the dynamic parameters of risk diffusion, multi-period progressive diffusion simulation is carried out to generate time-period risk evolution data; the risk evolution data of each period are integrated to generate multi-step progressive risk projection data.

[0029] The working principle and effects of the above technical solution are as follows: This solution improves the accuracy and consistency of agricultural risk diffusion simulation, enhances the completeness and reliability of risk projection data, and reduces deviations caused by incomplete risk source extraction and diffusion simulations that do not accurately reflect reality. It lowers the probability of inaccurate judgment of risk evolution patterns leading to a lack of targeted response plans, avoids fragmented simulation data that fails to reflect the spatiotemporal progression of risks, and prevents impacts on the timing of prevention and control. It can accurately extract risk source characteristics and transmission parameters, ensuring that risk injection aligns with the actual grid conditions. Furthermore, through spatiotemporal diffusion calculations and multi-period simulations, it can fully present the risk evolution process. The generated multi-step progressive projection data can clearly predict risk development trends, providing accurate and comprehensive support for the generation of subsequent response plans.

[0030] In one embodiment of the present invention, step S5 includes: S51. Perform multi-dimensional weighted calculations on agricultural risk response strategy data to generate a comprehensive agricultural risk index based on four conditions. S52. Using the comprehensive risk index of agricultural conditions as a trigger, initiate an inter-regional transfer learning mechanism to extract cross-scenario disaster characteristic information; S53. Update the extracted cross-scene feature information to the original knowledge structure to complete the online evolution of knowledge graph nodes; S54. Conduct full-domain extrapolation calculations through the evolved knowledge system to generate agricultural disaster early warning data with transfer learning capabilities.

[0031] The working principle and effects of the above technical solution are as follows: This solution improves the accuracy of comprehensive risk assessment of agricultural conditions (including weather, soil, and water conditions) and the adaptability of early warning data. It enhances the dynamic updating capability of the knowledge graph and the scientific rigor of early warnings, reducing errors caused by biased risk index calculations and a lack of universality in early warnings. It lowers the probability of insufficient adaptation of disaster early warnings across different regions, avoiding the solidification of knowledge graphs and their inability to adapt to new scenarios, which could lead to delayed or misjudged warnings and missed opportunities for optimal prevention and control. It can comprehensively quantify the comprehensive risks of the four conditions through multi-dimensional weighted calculations, and absorb cross-scenario experience through regional transfer learning and knowledge graph evolution, making the generated early warning data more universal and adaptable. It can accurately reflect the current risk situation and adapt to the monitoring needs of different regions, effectively improving the comprehensiveness and effectiveness of agricultural disaster early warnings.

[0032] One embodiment of the present invention, such as Figure 2 As shown, an agricultural four-condition comprehensive monitoring system based on big data is used to implement the early warning and prediction method described above. The system includes: The graph construction module performs spatiotemporal alignment processing on the agricultural four-condition monitoring area using multi-source sensors to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition, and climate; based on this dataset, an agricultural causal knowledge graph is constructed, which is used to reveal the dynamic causal relationship network between soil moisture, insect infestation, seedling condition, and climate. Data acquisition module: Based on the agricultural causal knowledge graph, dynamic threshold adaptive adjustment is performed, and multimodal data acquisition is initiated to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; the original agricultural four-condition monitoring data is subjected to joint noise reduction processing of physical model and data-driven approach to generate high-quality agricultural four-condition feature data; Evidence fusion module: Based on high-quality agricultural four-condition characteristic data, dual-drive reasoning is performed to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; conflict resolution and evidence fusion of the dual-drive reasoning results are performed through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. Strategy generation module: Dynamically divides the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; simulates risk propagation on the agricultural risk evolution grid model using agricultural risk attribution analysis data to generate multi-step risk projection data; and processes the multi-step risk projection data to generate agricultural risk response strategy data. Evolution Update Module: Calculates a weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index; initiates a regional transfer learning mechanism based on the comprehensive agricultural risk index to perform online evolution updates on new knowledge graph nodes, generating agricultural disaster early warning data with transfer learning capabilities.

[0033] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An early warning and prediction method for a comprehensive agricultural four-condition monitoring system based on big data, characterized in that, The method includes: S1. Perform spatiotemporal alignment processing on multi-source sensors in the agricultural four-condition monitoring area to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition and climate; construct an agricultural causal knowledge graph based on this dataset; S2. Based on the agricultural causal knowledge graph, perform dynamic threshold adaptive adjustment and initiate multimodal data acquisition to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; perform joint noise reduction processing of the original agricultural four-condition monitoring data using physical models and data-driven methods to generate high-quality agricultural four-condition feature data; S3. Based on high-quality agricultural four-condition characteristic data, perform dual-drive reasoning processing to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; resolve conflicts and fuse evidence through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. S4. Dynamically divide the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; use agricultural risk attribution analysis data to simulate risk propagation in the agricultural risk evolution grid model to generate multi-step risk projection data; process the multi-step risk projection data to generate agricultural risk response strategy data. S5. Calculate the weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index. Based on the comprehensive agricultural risk index, initiate a regional transfer learning mechanism to update new knowledge graph nodes online and generate agricultural disaster early warning data with transfer learning capabilities.

2. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 1, characterized in that, S1 includes: S11. Collect raw sensor information from multiple sources within the agricultural monitoring area, perform spatiotemporal benchmark unification processing on the raw sensor information, and generate a sensor raw dataset with consistent spatiotemporal benchmark. S12. Perform dimension normalization and format unification processing on the original sensor dataset to generate a standardized soil moisture, insect, seedling, and climate monitoring dataset. S13. Extract the correlation attributes of various monitoring elements in the standardized monitoring dataset, construct the correlation links between elements, and generate basic correlation structure data; S14. Based on the basic association structure data, a multi-level association reasoning framework is built to generate an agricultural causal knowledge graph. The agricultural causal knowledge graph is used to display the dynamic causal relationship network between soil moisture, insects, seedlings, and climate.

3. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 1, characterized in that, S2 includes: S21. Combining the internal correlation strength of the agricultural causal knowledge graph, adaptively correct the critical range of monitoring indicators to generate a dynamically adapted indicator threshold system. S22. Start the synchronous collection of data from the multimodal sensing terminal across the entire region to generate original agricultural four-condition monitoring data covering the entire region and all time periods; S23. Synchronously collect the time tags and spatial positioning information corresponding to the original agricultural four conditions monitoring data, and generate one-to-one spatiotemporal coordinate data; S24. Perform noise reduction and purification processing on the original agricultural four-condition monitoring data by combining physical model constraints and data-driven methods to generate high-quality agricultural four-condition characteristic data.

4. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 1, characterized in that, The S3 includes: S31. Import high-quality agricultural four-condition characteristic data into the physical inference channel to generate deterministic inference results that conform to natural laws; S32. Import high-quality agricultural four-condition characteristic data into the statistical learning channel to generate probabilistic inference results based on data distribution; S33. Input the deterministic reasoning results and probabilistic reasoning results into the agricultural causal knowledge graph to carry out conflict identification and information fusion processing; S34. Perform hierarchical attribution decomposition on the fused results to generate interpretable agricultural risk attribution analysis data.

5. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 4, characterized in that, S33 includes: S331. Input the deterministic reasoning results and the probabilistic reasoning results into the agricultural causal knowledge graph to generate a dual-path reasoning input dataset; S332. Perform feature similarity comparison on the dual-path reasoning input dataset to generate reasoning result difference feature data; S333. Perform conflict identification operation on the difference feature data of the reasoning results to generate a set of reasoning conflict features; perform conflict resolution processing on the set of reasoning conflict features to generate conflict-free integrated reasoning data. S334. Perform multi-dimensional information fusion on the conflict-free reasoning integrated data to complete conflict identification and information fusion processing.

6. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 5, characterized in that, S334 includes: Feature dimensions are extracted from conflict-free reasoning integrated data to generate multi-dimensional reasoning feature data; Assign feature weights to multi-dimensional inference feature data to generate a weighted inference feature set; Cross-dimensional correlation and fusion are performed on the weighted inference feature set to generate fused inference feature data; Consistency verification is performed on the fused inference feature data to generate standardized fusion results; conflict identification and information fusion processing are completed using the standardized fusion results.

7. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 1, characterized in that, The S4 includes: S41. Utilize spatiotemporal coordinate data to perform refined grid segmentation of the entire agricultural monitoring area, assign independent spatiotemporal attributes to each grid, and generate an agricultural risk evolution grid model. S42. Inject agricultural risk attribution analysis data into the agricultural risk evolution grid model, conduct spatiotemporal risk diffusion simulation, and generate multi-step progressive risk projection data; S43. Based on multi-step risk simulation data, automatically generate response plans to generate agricultural risk response plans by region and time period; S44. Conduct feasibility verification and optimization of agricultural risk response plans, and generate agricultural risk response strategy data.

8. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data according to claim 7, characterized in that, S41 includes: Read spatiotemporal coordinate data, extract spatial boundary and time span information of the entire agricultural monitoring area, and generate spatiotemporal range data of the entire area; Adaptive grid size division is performed on the spatiotemporal data of the entire domain to generate refined grid segmentation parameters; The entire agricultural monitoring area is divided into grids based on refined grid segmentation parameters to generate an initial monitoring grid set; Each grid in the initial monitoring grid set is assigned a unique spatiotemporal identifier and attribute parameters to generate grid cells with attribute identifiers; By integrating grid cells with attribute labels, an agricultural risk evolution grid model is generated.

9. The early warning and prediction method of the agricultural four-condition comprehensive monitoring system based on big data as described in claim 1, characterized in that, The S5 includes: S51. Perform multi-dimensional weighted calculations on agricultural risk response strategy data to generate a comprehensive agricultural risk index based on four conditions. S52. Using the comprehensive risk index of agricultural conditions as a trigger, initiate an inter-regional transfer learning mechanism to extract cross-scenario disaster characteristic information; S53. Update the extracted cross-scene feature information to the original knowledge structure to complete the online evolution of knowledge graph nodes; S54. Conduct full-domain extrapolation calculations through the evolved knowledge system to generate agricultural disaster early warning data with transfer learning capabilities.

10. A comprehensive agricultural four-condition monitoring system based on big data, wherein the monitoring system is used to implement the early warning and prediction method as described in claim 1, characterized in that, The system includes: The knowledge graph construction module performs spatiotemporal alignment processing on the agricultural four-condition monitoring area using multi-source sensors to generate a spatiotemporally synchronized multi-dimensional monitoring dataset of soil moisture, insect infestation, seedling condition, and climate; and constructs an agricultural causal knowledge graph based on this dataset. Data acquisition module: Based on the agricultural causal knowledge graph, dynamic threshold adaptive adjustment is performed, and multimodal data acquisition is initiated to obtain the original agricultural four-condition monitoring data and corresponding spatiotemporal coordinate data; the original agricultural four-condition monitoring data is subjected to joint noise reduction processing of physical model and data-driven approach to generate high-quality agricultural four-condition feature data; Evidence fusion module: Based on high-quality agricultural four-condition characteristic data, dual-drive reasoning is performed to obtain deterministic reasoning results based on physical models and probabilistic reasoning results based on data-driven approaches; conflict resolution and evidence fusion of the dual-drive reasoning results are performed through causal knowledge graphs to generate interpretable agricultural risk attribution analysis data. Strategy generation module: Dynamically divides the agricultural monitoring area into grids using spatiotemporal coordinate data to generate an agricultural risk evolution grid model; simulates risk propagation on the agricultural risk evolution grid model using agricultural risk attribution analysis data to generate multi-step risk projection data; and processes the multi-step risk projection data to generate agricultural risk response strategy data. Evolution Update Module: Calculates a weighted risk index based on agricultural risk response strategy data to generate a comprehensive agricultural risk index; initiates a regional transfer learning mechanism based on the comprehensive agricultural risk index to perform online evolution updates on new knowledge graph nodes, generating agricultural disaster early warning data with transfer learning capabilities.

Images