Farmland disease and pest dynamic prediction model construction method driven by multi-source remote sensing data
By using hierarchical preprocessing and ensemble learning algorithms for multi-source remote sensing data, a dynamic prediction model for farmland pests and diseases was constructed, solving the problem of real-time prediction of farmland pests and diseases using remote sensing data, and achieving high-precision and efficient pest and disease prediction and management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-03-10
AI Technical Summary
In existing technologies, remote sensing data lacks the ability to predict dynamic changes in farmland pest and disease monitoring in real time, and the fusion and processing methods of multi-source remote sensing data are difficult to adapt to the changing needs of different industries and different periods, resulting in insufficient timeliness and accuracy of prediction results.
By using a multi-source remote sensing data-driven approach, time-series, optical, and radar remote sensing data are acquired, hierarchical preprocessing and ensemble learning algorithms are used to fuse them, and a multi-dimensional farmland pest and disease monitoring model is constructed. This model combines historical and real-time data to make accurate predictions, and the early warning information is displayed visually and adjusted in real time.
It has improved the accuracy and spatiotemporal resolution of farmland pest and disease monitoring, enhanced agricultural management decision support, enabled precise pest and disease control and optimal resource allocation, and improved the efficiency of farmland management and emergency response capabilities.
Smart Images
Figure CN121640264A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of agricultural remote sensing data application technology, specifically to a method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data. Background Technology
[0002] Farmland pests and diseases are significant factors affecting agricultural production. Timely and accurate monitoring and prediction of their occurrence can effectively improve crop yield and quality. Traditional pest and disease monitoring methods mainly rely on manual inspections and meteorological monitoring, which suffer from information lag and narrow coverage, making it difficult to meet the needs of large-scale, precision agriculture. Currently, there are two main types of methods for predicting farmland pests and diseases: one is prediction models based on meteorological data, and the other is monitoring and analysis methods based on remote sensing data. The former relies on meteorological conditions such as temperature, humidity, and precipitation for prediction, but the timeliness and accuracy of meteorological data still have certain limitations. Remote sensing technology, by obtaining information on the spatial distribution and changes of farmland pests and diseases, can reflect the real-time condition of farmland.
[0003] However, current technologies primarily utilize remote sensing data for static monitoring and localized analysis, lacking the ability to predict dynamic changes in farmland pests and diseases in real time. Regarding the fusion and dynamic prediction of multi-source remote sensing data, existing technologies have failed to effectively address the challenge of flexibly adjusting data processing methods to meet the changing needs of different periods and industries. This means that the timeliness and accuracy of prediction results are often affected by data quality, acquisition frequency, and data processing algorithms, making it difficult to meet the efficiency and accuracy requirements of actual production. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides a method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data, which solves the problems of multi-source remote sensing data fusion processing, improving the accuracy of dynamic prediction of pests and diseases, and automatic adjustment of real-time early warning.
[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data, comprising: S1. acquiring multi-source remote sensing data through a remote sensing platform, wherein the multi-source remote sensing data includes time-series remote sensing data, optical remote sensing data, and radar remote sensing data of the target farmland area; collecting multi-dimensional information through data fusion to improve the monitoring accuracy and spatiotemporal resolution of the target farmland area and provide more comprehensive agricultural management decision support;
[0006] S2. Performing layered preprocessing on the multi-source remote sensing data helps to correct errors and improve data quality;
[0007] S3. The multi-source data of the hierarchical preprocessing are deeply fused and intelligently integrated through an ensemble learning algorithm to construct a multi-dimensional farmland pest and disease monitoring model, improve monitoring accuracy and prediction accuracy, and assist in precision agriculture decision-making;
[0008] S4. Based on historical data of pest and disease occurrence, construct a dynamic prediction model for pests and diseases. The dynamic prediction model for pests and diseases combines real-time data and historical data to make accurate predictions and generate prediction results, so as to achieve precise prevention and control and optimized resource allocation.
[0009] S5. The prediction results are displayed in real time through a visualization module. Based on the real-time data, the prediction results and early warning information are automatically adjusted to improve the efficiency of emergency response and farmland management.
[0010] Optimized, the observation time interval of the time series remote sensing data should not exceed the response time of crop key growth period changes. The optical remote sensing data covers the visible light, near-infrared and short-wave infrared bands. The radar remote sensing data includes synthetic aperture radar and lidar. The optical remote sensing data includes multispectral data with red edge bands of B5-704nm, B6-740nm, and B7-783nm, band data with a resolution of 5×20m, and multispectral imaging data with band centers of 550nm, 660nm, 720nm, and 840nm.
[0011] The optimized, layered preprocessing described in S2 includes:
[0012] S21. Pixel-level preprocessing: Radiometric calibration, atmospheric correction, geometric correction and strip restoration are performed on the original remote sensing images to obtain surface reflectance, brightness temperature and backscattering coefficient;
[0013] S22. Plot-level preprocessing: Based on the farmland plot vector boundary, the pixel-level results are aggregated with intra-plot mean to eliminate the influence of mixed pixels;
[0014] S23. Regional preprocessing: Based on the plot-level results, the sliding window method is used to calculate spatial texture, neighborhood statistics and terrain correction coefficients to generate a regionally consistent multidimensional feature matrix.
[0015] The optimized ensemble learning algorithm includes:
[0016] S31. Integrated Feature Fusion: A weighted decision-level fusion framework is adopted to perform fusion processing on the preprocessed multi-source data;
[0017] S32. Multidimensional Feature Optimization: Based on the occurrence mechanism of pests and diseases, a feature pyramid is constructed. In the spectral dimension, 3 to 5 feature bands with p < 0.01 that are significantly related to the target pests and diseases in the 680-750nm red edge derivative spectrum are selected. In the time dimension, 7 to 15 days are extracted as a sliding window as a time-series statistic. In the spatial dimension, the GLCM texture energy and correlation of the 5×5 neighborhood are calculated.
[0018] The optimized model for the fusion process is as follows:
[0019]
[0020] The final fused feature matrix is defined as follows: α represents the optical feature weights, and β represents the radar feature weights: β = 1 - α. The weight of the i-th optical feature. For the i-th optical characteristic value, The weight of the j-th radar feature is... Let L be the j-th radar feature value. L is obtained by acquiring the three-dimensional structural information of the crop canopy, providing crucial data support for pest and disease prediction. The L-derived canopy structure features are represented by γ, which is a fixed weight of the L-structure features with a value of 0.2. The range of α is 0.4 to 0.7 and is dynamically adjusted based on the band signal-to-noise ratio.
[0021] The optimized approach involves adjusting the real-time data update time by combining the generated historical data on pest and disease occurrence with multi-dimensional data from the target farmland.
[0022] The optimized dynamic prediction model can access remote sensing data streams in real time and dynamically update prediction results, and trigger SMS / email alerts based on user-defined thresholds when the probability of pest occurrence is ≥0.7.
[0023] In an optimized manner, the automatic adjustment prediction results are transmitted and calculated in real time through a cloud platform, ensuring the timely updating and delivery of pest and disease prediction information.
[0024] This invention provides a method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data. It has the following beneficial effects:
[0025] This method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data significantly improves the accuracy and spatiotemporal resolution of farmland pest and disease monitoring through the acquisition and fusion of multi-source remote sensing data. Specifically, it employs multi-dimensional information collection using time-series remote sensing data, optical remote sensing data, and radar remote sensing data, enhancing the monitoring capability of target farmland areas and providing more comprehensive data support for agricultural management decisions.
[0026] By employing ensemble learning algorithms for deep fusion and intelligent integration, an invention has been developed that improves the accuracy of pest and disease prediction and monitoring precision. Furthermore, the dynamic pest and disease prediction model, combining historical and real-time data, can generate precise prediction results and enhance the efficiency of farmland management and emergency response through visualization and real-time adjustment of early warning information. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the structure of the present invention;
[0028] Figure 2 This is a flowchart of the multi-source remote sensing data acquisition and fusion process of the present invention;
[0029] Figure 3 This is a visual representation of the structure of the early warning system of this invention. Detailed Implementation
[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0031] Example 1
[0032] like Figure 1-3 As shown, this embodiment of the invention provides a method for constructing a dynamic prediction model for farmland pests and diseases driven by multi-source remote sensing data, including: S1. Acquiring multi-source remote sensing data through a remote sensing platform. The multi-source remote sensing data includes time-series remote sensing data, optical remote sensing data, and radar remote sensing data of the target farmland area. Multi-dimensional information is collected through data fusion. The observation time interval of the time-series remote sensing data should not exceed the response time of changes in the key growth period of crops. The optical remote sensing data covers the visible light, near-infrared, and short-wave infrared bands. The radar remote sensing data includes synthetic aperture radar and lidar. The optical remote sensing data includes multispectral data with red-edge bands of B5-704nm, B6-740nm, and B7-783nm, band data with a resolution of 5×20m, and multispectral imaging data with band centers of 550nm, 660nm, 720nm, and 840nm.
[0033] S2. Perform layered preprocessing on multi-source remote sensing data. Layered preprocessing includes:
[0034] S21. Pixel-level preprocessing: Radiometric calibration, atmospheric correction, geometric correction and strip restoration are performed on the original remote sensing image to obtain surface reflectance, brightness temperature and backscattering coefficient.
[0035] S22. Plot-level preprocessing: Based on the farmland plot vector boundary, the pixel-level results are aggregated with intra-plot mean to eliminate the influence of mixed pixels.
[0036] S23. Regional preprocessing: Based on the plot-level results, the sliding window method is used to calculate spatial texture, neighborhood statistics and terrain correction coefficients to generate a regionally consistent multidimensional feature matrix.
[0037] S3. The multi-source data, after hierarchical preprocessing, is deeply fused and intelligently integrated using ensemble learning algorithms to construct a multi-dimensional farmland pest and disease monitoring model. The ensemble learning algorithms include:
[0038] S31. Integrated Feature Fusion: A weighted decision-level fusion framework is adopted to perform fusion processing on the preprocessed multi-source data.
[0039] S32. Multidimensional Feature Optimization: Based on the occurrence mechanism of pests and diseases, a feature pyramid is constructed. In the spectral dimension, 3 to 5 feature bands with p < 0.01 that are significantly related to the target pests and diseases in the 680-750nm red edge derivative spectrum are selected. In the time dimension, 7 to 15 days are extracted as a sliding window as a time-series statistic. In the spatial dimension, the GLCM texture energy and correlation of the 5×5 neighborhood are calculated.
[0040] S4. Based on historical data of pest and disease occurrence, a dynamic prediction model for pests and diseases is constructed. This model combines real-time and historical data for accurate prediction and generates prediction results. After the historical data of pest and disease occurrence is generated, the update time of the real-time data is adjusted by combining multi-dimensional data of the target farmland. The dynamic prediction model can access remote sensing data streams in real time and dynamically update prediction results. Furthermore, based on a user-defined threshold, an SMS / email alert is triggered when the probability of pest and disease occurrence is ≥0.7. The fused model is as follows:
[0041]
[0042] The final fused feature matrix is defined as follows: α represents the optical feature weights, and β represents the radar feature weights: β = 1 - α. The weight of the i-th optical feature. For the i-th optical characteristic value, The weight of the j-th radar feature is... Let L be the j-th radar feature value. L is obtained by acquiring the three-dimensional structural information of the crop canopy, providing crucial data support for pest and disease prediction. The L-derived canopy structure features are represented by γ, which is a fixed weight of the L-structure features with a value of 0.2. The range of α is 0.4 to 0.7 and is dynamically adjusted based on the band signal-to-noise ratio.
[0043] S5. The prediction results are displayed in real time through a visualization module. The prediction results are automatically adjusted and early warning information is issued based on the real-time data. The automatic adjustment of the prediction results is achieved through real-time data transmission and calculation via a cloud platform.
[0044] Multi-source data, including time-series, optical, and radar remote sensing data, is acquired through remote sensing platforms. The observation intervals for time-series data need to be sufficiently short to capture changes during key growth stages in farmland. Optical remote sensing reflects vegetation growth and signs of pests and diseases, while radar data provides high-precision three-dimensional topographic and vegetation structure information. Working together, these technologies ensure the timeliness and accuracy of the data.
[0045] After data acquisition, tiered preprocessing is performed. Pixel-level preprocessing transforms the data into usable data through radiometric calibration and atmospheric correction. Plot-level preprocessing eliminates mixed-pixel effects, while region-level preprocessing calculates texture and neighborhood statistics using the sliding window method to ensure consistent data accuracy and provide a high-quality foundation for subsequent model construction.
[0046] Ensemble learning integrates multi-source data through a weighted decision-level fusion framework. It utilizes red-edge spectral, time window, and neighborhood texture features selected based on the disease and pest occurrence mechanism. These techniques improve model accuracy and help to better reflect the occurrence and development trends of diseases and pests.
[0047] The dynamic pest and disease prediction model combines historical and real-time data for accurate predictions, dynamically adjusts the prediction results, and issues early warnings. When thresholds are set, the system automatically notifies farmland managers via SMS or email, enabling timely measures to be taken and improving control efficiency.
[0048] The forecast results are displayed in real time through charts and heat maps, helping farmland managers quickly identify high-risk areas. Real-time data updates ensure the accuracy of the displayed content and early warning information. The system transmits data through a cloud platform, improving decision support and farmland management efficiency.
[0049] Example 2
[0050] This embodiment describes S2 and S3.
[0051] S2. Layered Preprocessing of Multi-Source Remote Sensing Data
[0052] Objective: To ensure that the acquired remote sensing data undergoes necessary processing to remove errors and improve data quality, thereby ensuring accurate results in subsequent analysis.
[0053] 1. Pixel-level preprocessing:
[0054] Radiometric calibration: Converting remote sensing images into surface reflectance, brightness temperature, and backscattering coefficient to eliminate the influence of changes in instrument response.
[0055] Atmospheric correction: Corrects atmospheric effects and reduces atmospheric interference with remote sensing images, so that the final data can more accurately reflect the surface features.
[0056] Geometric correction: Ensures the geometric accuracy of the image and corrects spatial distortions caused by satellite platform movement.
[0057] Striping repair: If the acquired remote sensing data contains striping phenomena and blank areas between satellite scan lines, repair is required to eliminate these discontinuities.
[0058] 2. Site-level pretreatment:
[0059] Mean aggregation: Based on the vector boundaries of farmland, pixel-level data is aggregated to eliminate the influence of mixed pixels. The pixel data within each plot is averaged to generate a representative value, ensuring greater consistency in remote sensing data for each plot.
[0060] 3. Regional preprocessing:
[0061] Sliding window method: The sliding window method is used to further process plot-level data, calculate spatial texture, neighborhood statistics, and topographic correction coefficients, and generate a regionally consistent multidimensional feature matrix. These feature matrices help enhance the spatiotemporal consistency of farmland monitoring.
[0062] S3. Deep Data Fusion and Intelligent Integration
[0063] Objective: To construct a more accurate monitoring model for farmland pests and diseases by fusing multi-source data after hierarchical preprocessing through an ensemble learning algorithm.
[0064] 1. Integrated Feature Fusion:
[0065] Weighted Decision-Level Fusion Framework: This framework uses a weighted decision-level fusion method to fuse features from different data sources, including optical and radar data. The weight of each data source is dynamically adjusted based on its importance in pest and disease prediction. This helps to obtain more accurate monitoring information from multiple dimensions.
[0066] 2. Optimal selection of multidimensional features:
[0067] Spectral dimension: Feature bands significantly correlated with the target pests and diseases were selected from the red-edge derivative spectrum of 680-750nm. The most relevant bands were screened using statistical methods with p-values <0.01 to ensure high model accuracy.
[0068] Time dimension: 7-15 days are extracted as a sliding window to calculate time-series statistics and capture the dynamic characteristics of pest and disease development during the crop growth cycle.
[0069] Spatial dimension: The gray-level co-occurrence matrix of a 5×5 neighborhood is used to calculate texture energy and correlation, further enhancing the extraction of spatial features and ensuring that the model can make full use of the spatial information of farmland.
[0070] Through these methods, the ensemble learning algorithm in stage S3 can effectively fuse multi-source data to generate a high-precision, high-spatiotemporal-resolution monitoring model for farmland pests and diseases. The implementation of these steps significantly improves the quality of remote sensing data, providing solid data support for subsequent dynamic pest and disease prediction models, thereby enhancing the accuracy and real-time performance of farmland pest and disease prediction.
[0071] Example 3
[0072] This embodiment describes S4.
[0073] S4. Construct a dynamic prediction model for pests and diseases based on historical data on pest and disease occurrence.
[0074] The purpose of this step is to build a dynamic prediction model for pests and diseases, which combines historical data and real-time remote sensing data to make accurate predictions and generate targeted early warning and control plans for farmland pests and diseases.
[0075] 1. Data preparation and analysis:
[0076] Historical pest and disease data: Analysis is conducted based on historical data of past pests and diseases, including the time, location, and intensity of occurrence. This historical data provides training data for building predictive models.
[0077] Real-time remote sensing data: Time series of remote sensing data, optical data, and radar data acquired in real time are used to supplement and correct historical data to ensure the timeliness and accuracy of prediction results.
[0078] 2. Construction of dynamic prediction model for pests and diseases:
[0079] Model Framework: An ensemble learning algorithm is used to deeply fuse historical data and real-time remote sensing data. The model combines historical data on pest and disease occurrence patterns with real-time data to adjust and predict the dynamic changes of pests and diseases in real time.
[0080] Algorithm steps:
[0081] By integrating feature fusion, multi-source remote sensing data is combined with historical pest and disease data to form a comprehensive feature matrix.
[0082] A weighted decision framework is adopted to optimize the model prediction results by weighted synthesis of optical and radar data.
[0083] We selected red-edge spectral, spatial texture, and time-series features that are closely related to the occurrence of pests and diseases. Experimental analysis revealed that these features have a significant correlation with the occurrence of pests and diseases.
[0084] Model optimization:
[0085] The feature weights are dynamically adjusted in the model to adapt to changes in different seasons or crop growth stages.
[0086] Threshold setting: The model sets a threshold for the probability of pest and disease occurrence. When the predicted probability of pest and disease occurrence reaches or exceeds this threshold, the early warning system is triggered.
[0087] 3. Real-time data integration and updating:
[0088] Data Updates: As new remote sensing data becomes available, the dynamic prediction model connects to the remote sensing data stream in real time and automatically updates its predictions. This ensures that the prediction model remains up-to-date and can promptly reflect changes in pests and diseases.
[0089] Combining historical and real-time data: By integrating historical and real-time data, the model can provide more accurate predictions for different disease and pest occurrence cycles. The inclusion of real-time data can effectively correct for potential errors in historical data, ensuring the accuracy and reliability of the model's predictions.
[0090] 4. Early warning mechanism and output:
[0091] Early warning information release: When the prediction results show that the probability of pests and diseases is higher than the preset threshold, the system will automatically trigger an early warning message via SMS or email to notify farmland managers to take emergency measures.
[0092] Prediction Results Display: The prediction results of pests and diseases are displayed in real time through the visualization module's heat map and charts, helping farmland managers to quickly identify potential high-risk areas.
[0093] Visualized content:
[0094] Regional heat map: shows the risk level of pests and diseases; the darker the color, the higher the probability of pests and diseases occurring.
[0095] Time series charts: show the historical trends and future forecasts of pests and diseases.
[0096] This model enables farm managers to obtain real-time forecasts of pest and disease outbreaks and adjust their farm management strategies accordingly to reduce the impact of pests and diseases on crops.
[0097] The core of the above steps is to deeply integrate historical pest and disease data with real-time remote sensing data, and build a dynamic pest and disease prediction model through ensemble learning algorithms. This model can accurately predict the occurrence of pests and diseases and issue early warning information in a timely manner, helping farmland managers to optimize prevention and control measures and achieve precision agricultural management.
[0098] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for constructing a multi-source remote sensing data-driven dynamic prediction model of crop diseases and insect pests, characterized in that, The application relates to a multi-dimensional crop disease and pest monitoring method based on multi-source remote sensing data. S1. Obtain multi-source remote sensing data through a remote sensing platform, wherein the multi-source remote sensing data comprises time series remote sensing data, optical remote sensing data and radar remote sensing data of a target farmland area, and multi-dimensional information is collected through data fusion; S2. Layered preprocessing is performed on the multi-source remote sensing data; S3. The layered preprocessed multi-source data is subjected to deep fusion and intelligent integration through an integrated learning algorithm to construct a multi-dimensional farmland disease and pest monitoring model; S4. A dynamic prediction model of diseases and pests is constructed based on historical data of diseases and pests, the dynamic prediction model of diseases and pests combines real-time data and historical data to perform accurate prediction, and a prediction result is generated; S5. The prediction result is displayed in real time through a visual display module, and the prediction result is automatically adjusted based on the real-time data and early warning information is issued.
2. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 1, characterized in that: The observation time interval of the time series remote sensing data should not exceed the response time of crop key growth period change, the optical remote sensing data covers visible light, near-infrared and short-wave infrared bands, the radar remote sensing data comprises synthetic aperture radar and laser radar, the optical remote sensing data comprises multispectral data with red edge bands of B5-704nm, B6-740nm and B7-783nm, band data with a resolution of 5*20m, and multispectral imaging data with band centers of 550nm, 660nm, 720nm and 840nm.
3. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 1, characterized in that: The layered preprocessing in S2 comprises: S21. Pixel-level preprocessing: radiation calibration, atmospheric correction, geometric correction and strip repair are performed on the original remote sensing image to obtain ground reflectivity, brightness temperature and backscattering coefficient; S22. Plot-level preprocessing: based on the vector boundary of the farmland plot, the pixel-level result is subjected to plot-in-mean aggregation to eliminate the influence of mixed pixels; S23. Area-level preprocessing: on the plot-level result, a sliding window method is used to calculate spatial texture, neighborhood statistics and terrain correction coefficient to generate a region-consistent multi-dimensional feature matrix.
4. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 1, characterized in that: The integrated learning algorithm comprises: S31. Integrated feature fusion: a weighted decision-level fusion framework is used to perform fusion processing on the preprocessed multi-source data; S32. Multi-dimensional feature optimization: a feature pyramid is constructed based on the occurrence mechanism of diseases and pests, 3-5 feature bands with p<0.01 that are significantly related to target diseases and pests are selected in the spectral dimension of 680-750nm red edge derivative spectrum, 7-15 days are extracted as a sliding window as time series statistics, and GLCM texture energy and correlation of a 5*5 neighborhood are calculated in the spatial dimension.
5. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 4, characterized in that: The model of the fusion processing is: is the final fused feature matrix, a is the optical feature weight, and β is the radar feature weight. β = 1 - a, is the weight of the i-th optical feature, is the i-th optical feature value, is the weight of the j-th radar feature, is the j-th radar feature value, L is the three-dimensional structure information of the crop canopy obtained by acquisition, which provides key data support for pest prediction, is the L-derived canopy structure feature, and γ is the fixed weight of the L structure feature, with a value of 0.2, and a range of 0.4-0.7 and dynamic adjustment based on the signal-to-noise ratio of the wave band.
6. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 1, characterized in that: The historical data of diseases and pests are combined with multi-dimensional data of target farmland to update real-time data in S4.
7. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 1, characterized in that: The dynamic prediction model can access remote sensing data flow and dynamically update the prediction result in real time, and triggers an SMS / email warning when the probability of disease and pest occurrence is greater than or equal to 0.7 based on a user-defined threshold.
8. The multi-source remote sensing data-driven dynamic prediction model construction method for crop diseases and insect pests according to claim 6, characterized in that: The automatic adjustment of the prediction result in S5 is realized through real-time data transmission and calculation of a cloud platform.