Big data-based agricultural production precision consultation management system

Abstract

This invention belongs to the field of agricultural consulting technology, specifically a big data-based precision agricultural production consulting management system. It includes a plot spatiotemporal grid construction unit, used to divide the target area into multiple rectangular spatiotemporal grids with unique plot IDs based on a preset geographic coordinate system and a minimum management unit area threshold. It outputs specific agricultural operation suggestions, including operation type, recommended time window, material type, and application rate, and pushes these suggestions in text and image format through farmer terminals. By adopting a dynamic adaptation mechanism for plot spatiotemporal grids and a crop growth cycle tag linkage update mechanism, it achieves adaptive reconstruction and synchronous information updates of the spatiotemporal grid under scenarios of plot boundary changes and crop rotation, while retaining historical data relationships. This system enables continuous and precise control of single plots across cycles and a high degree of matching between the grid and the actual production scenario.

Description

Technical Field

[0001] This invention belongs to the field of agricultural consulting technology, specifically a precision consulting and management system for agricultural production based on big data. Background Technology

[0002] Agricultural production management is developing towards digitalization and precision, with big data, remote sensing monitoring, and IoT sensing technologies being increasingly widely applied in modern agriculture. To achieve integrated management of farmland, crop growth, environmental factors, and agricultural activities, building precise consulting and intelligent management systems adapted to agricultural production scenarios has become an industry trend. Through the integration, processing, and intelligent analysis of multi-source agricultural data, data support and technical services can be provided for field production decisions.

[0003] This type of agricultural management system can realize the grid-based division of farmland space, unified access and spatiotemporal matching of multi-dimensional agricultural data, and assist farmers in carrying out precision agricultural operations through comprehensive management of plot information, crop growth status, soil environmental parameters and agricultural operation records, improve the scientificity and timeliness of agricultural production decisions, and realize independent control and data traceability of production information of individual plots, helping agricultural production to transform from traditional extensive management to precise and intelligent control.

[0004] Currently, traditional agricultural production management often employs static plot division methods. Plot grids cannot dynamically and adaptively adjust to actual boundary offsets and crop rotation, leading to mismatches between the grid and actual plots, and the inability to synchronize crop growth cycle labels. This hinders continuous and precise management of single plots across growth cycles. Secondly, agricultural production involves multi-source heterogeneous data such as remote sensing imagery, IoT sensors, and farmer logs. Existing systems generally suffer from inconsistent data access standards, insufficient spatiotemporal coordinate matching accuracy, and a lack of time-granularity aggregation and semantic standardization, making it difficult to effectively integrate and utilize multi-source data and providing reliable data support for precise decision-making. Thirdly, existing agricultural management systems often lack data access isolation mechanisms for individual plots, leading to crosstalk and misuse of data between different farmers and plots. The privacy and security of field production data are compromised, and the lack of a closed-loop decision-making self-optimization mechanism prevents the system's decision-making model from continuously iterating and optimizing based on actual production data, limiting the effectiveness of intelligent consultation and management. Therefore, this paper proposes a big data-based precision agricultural production consultation and management system to address these issues. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies and solve at least one of the technical problems mentioned in the background, this invention proposes a big data-based precision consulting and management system for agricultural production.

[0006] The technical solution adopted by the present invention to solve its technical problem is: the agricultural production precision consultation management system based on big data described in the present invention, 1. the agricultural production precision consultation management system based on big data, characterized in that: it includes a plot spatiotemporal grid construction unit, which is used to divide the target area into multiple rectangular spatiotemporal grids with unique plot IDs according to the preset geographical coordinate system and the minimum management unit area threshold, and bind crop type identifier and current growth stage label to each spatiotemporal grid; The multi-source data access unit is used to receive at least three types of heterogeneous data streams in real time: (a) Multispectral image data from remote sensing satellites, including NDVI vegetation index and land surface temperature fields; (b) Structured sensor data from IoT sensors deployed in the field, including soil moisture, pH, and nitrogen, phosphorus and potassium content; (c) Semi-structured agricultural operation logs uploaded by farmers' terminals, which record operation type, time, amount and location coordinates in JSON format; Spatiotemporal semantic alignment unit, used to perform the following operations: (i) The remote sensing image data is resampled to the boundary range of the spatiotemporal grid according to the shooting time and spatial resolution to generate the rasterized index value corresponding to each plot ID; (ii) Match the sensor data to the nearest spatiotemporal grid based on the sensor's GPS coordinates, and aggregate it into a time series with a time granularity of 5 minutes; (iii) Parse the location coordinates in the agricultural operation log, map them to the corresponding plot ID, and standardize the operation type field into ontology concept nodes based on the predefined agricultural operation ontology library. The agricultural operation ontology library includes four types of operations: fertilization, irrigation, spraying, and sowing, and their subclasses. Each subclass is associated with a standard parameter template. The dynamic data fusion unit is used to extract feature vectors from aligned remote sensing indicators, sensing time series and standardized agricultural logs for each plot ID within the agricultural window period corresponding to the current crop growth stage label, and to perform weighted splicing according to preset weight coefficients to form a fusion feature record of the plot in the window period. The weight coefficients are obtained from the configuration table according to the crop type and growth stage. The decision engine invocation unit is used to input the fused feature records into an agricultural decision model that matches the current crop type and growth stage, and output specific agricultural operation suggestions.

[0007] Preferably, the land parcel spatiotemporal grid construction unit is also linked to a grid dynamic adaptation module, which is used to obtain real-time information on changes in land parcel boundaries and crop rotation data in the target area; For plots whose boundary offset exceeds the preset threshold, the spatiotemporal grid reconstruction process is automatically triggered, and the corresponding plot ID and the bound crop type identifier and growth stage label are updated synchronously, while retaining the historical spatiotemporal data association of the original plot ID; for plots after crop rotation switching, the original growth stage label is cleared with one click, and the corresponding crop full growth cycle node library is rematched to achieve precise cross-cycle management of a single plot.

[0008] Preferably, the spatiotemporal semantic alignment unit further includes an abnormal data removal module, which performs dual verification on the rasterized remote sensing indicators and the aggregated sensing time series respectively; A threshold verification model is constructed based on the crop growth patterns of the plots to remove abnormal sensor data that exceeds the normal value range and remote sensing raster data that is distorted due to cloud interference. At the same time, for missing data, a weighted fitting method is used to complete the missing data by using the historical average of the same plot and the normal data of adjacent spatiotemporal grids. This ensures that there are no gaps or distortions after the various types of data are matched with the spatiotemporal grids, which meets the needs of subsequent fusion feature extraction.

[0009] Preferably, it also includes a decision model self-optimization unit, which establishes a two-way data linkage channel with the dynamic data fusion unit and the decision engine calling unit respectively; The self-optimization unit of the decision model is used to collect the results of agricultural operations and the subsequent crop growth status data of the corresponding plots from the feedback of farmers' terminals in real time. The above data is used as labeled samples and fed back to the agricultural decision model built into the decision engine calling unit to iteratively optimize the internal weight parameters of the model in real time. At the same time, independent decision sub-models are built for different crop types and different crop growth stages to realize differentiated iteration of decision parameters within each agricultural window period, further improving the adaptability of agricultural operation suggestions to the actual production conditions of the plots.

[0010] Preferably, the dynamic data fusion unit has a built-in weight adaptive adjustment module, and the weight adaptive adjustment module communicates bidirectionally with the decision model self-optimization unit and the spatiotemporal semantic alignment unit. The weight adaptive adjustment module is used to read the integrity index of the multi-source data after completion and calibration in real time. Combined with the historical decision error value after the iteration of the agricultural decision model, it dynamically corrects the preset weight coefficients of the feature vectors of each data source. The weight parameters can be adapted in real time without manual intervention, ensuring that the fused feature records fit the real-time growth status of the plot and weakening the impact of the deviation of a single data source on the subsequent decision results.

[0011] Preferably, the decision engine calling unit is linked to an agricultural suggestion hierarchical push module, which pre-stores a table of farmer terminal permission levels and land management permissions. The generated agricultural operation suggestions are divided into multi-level execution instructions according to the urgency and complexity of the operation. Combined with the management permissions of the farmer's terminal, the corresponding level of agricultural operation suggestions and supporting implementation details are pushed to the farmer. At the same time, multiple operation suggestions for the same plot are sorted in time to avoid the problem of overlapping and conflicting agricultural operations.

[0012] Preferably, the abnormal data removal module of the spatiotemporal semantic alignment unit is also linked to a data source tracing and marking module; For the removed abnormal data and the supplemented missing data, each is bound to the corresponding plot ID and timestamp to generate an independent traceability identifier, which is simultaneously stored in the plot's historical data ledger. Furthermore, in the subsequent process of feature extraction and agricultural decision-making, the corresponding data traceability information is retrieved simultaneously, realizing traceable control of multi-source data from collection to decision-making throughout the entire process.

[0013] Preferably, the decision model self-optimization unit is equipped with a sample screening module, which is linked to the data traceability and marking module; For the labeled samples returned to the agricultural decision-making model, invalid samples carrying abnormal data traceability markers or missing value completion markers are first removed. Only valid samples that have passed data collection and calibration and have no data distortion are selected to participate in model iteration, further improving the training accuracy of the decision-making model and the reliability of parameter iteration. Preferably, the land parcel spatiotemporal grid construction unit is linked to a grid permission isolation module, which binds a unique land parcel ID to the farmer's terminal identity information; For multi-source data, fusion feature records, and agricultural decision-making suggestions for a single spatiotemporal grid, access and viewing permissions are only granted to the corresponding bound farmer terminals. This achieves independent isolation and management of data for a single plot, prevents cross-plot data interference, and ensures the privacy and security of field production data.

[0014] Preferably, the weight adaptive adjustment module has a built-in dynamic weight calculation model, which realizes real-time adaptive calculation of the weights of feature vectors from each data source through the following formula: ; in: Indicates the first Class data source in the first The adaptive weighting coefficients for each agricultural window period take values ​​in the range [0,1] and satisfy the following conditions: ; For the first The weighting coefficient of the data source in the previous agricultural window period is used to ensure the continuity of weight changes; This is the weight inertia coefficient, with a value range of [0,1], used to control the degree of influence of historical weights; To complete the integrity index of the i-th type of data source after calibration, the value range is [0,1], and it is calculated by combining the data missing rate and the anomaly removal rate; This represents the average error value of the agricultural decision-making model after making decisions based on the current weighted fusion features, and its value range is [value missing]. The larger the error value, the larger the correction coefficient for the corresponding data source weight; Assign adjustment coefficients to the weights, with values ​​ranging from [0,1], and satisfying the following conditions: + =1, which are used to adjust the contribution of data integrity indicators and decision error values ​​to the weights respectively; the weight adaptive adjustment module calculates the weight coefficients of each data source in real time based on the above formula, and updates them synchronously to the dynamic data fusion unit to ensure that the fused feature records are highly adapted to the actual growth status of the plots, effectively weakening the interference of single data source deviation on agricultural decision results.

[0015] The beneficial effects of this invention are: This invention provides a big data-based precision agricultural production consulting and management system. By adopting a dynamic adaptation mechanism of land plot spatiotemporal grid and a crop growth cycle tag linkage update mechanism, it realizes the adaptive reconstruction and synchronous information update of spatiotemporal grid under scenarios of land plot boundary changes and crop rotation, while retaining historical data correlation. It has the function of continuous and precise control of single plots across cycles and high matching between grid and actual production scenario.

[0016] This invention provides a big data-based precision agricultural production consulting and management system. By unifying the access, spatiotemporal semantic alignment and standardized calibration of multi-source heterogeneous data such as remote sensing, IoT, and agricultural logs, it achieves precise matching and efficient integration of multi-dimensional agricultural data with plot spatiotemporal grids, which can enhance the value of data utilization and provide a reliable data foundation for precise decision-making.

[0017] This invention provides a big data-based precision agricultural production consultation and management system. By binding a unique plot ID with the farmer's terminal identity and implementing access control, it achieves independent access to data for a single plot and privacy protection, avoids cross-plot data interference, and ensures the privacy and security of field production data while improving the standardization of system management. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention.

[0019] In the attached diagram: Figure 1 This is a schematic diagram of the system flow of the present invention; Figure 2 This is a schematic diagram of the spatiotemporal grid construction unit for land parcels in this invention; Figure 3 This is a schematic diagram of the spatiotemporal semantic alignment unit of the present invention; Figure 4 This is a schematic diagram of the self-optimization unit module of the graph decision model in this invention. Detailed Implementation

[0020] 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.

[0021] Specific implementation examples are given below.

[0022] Please see Figures 1-4 This invention provides a big data-based precision agricultural production consulting and management system, including a plot spatiotemporal grid construction unit, a multi-source data access unit, a spatiotemporal semantic alignment unit, a dynamic data fusion unit, a decision engine invocation unit, and a decision model self-optimization unit. The units are closely connected through high-speed data links to jointly construct a closed-loop precision agricultural production consulting and intelligent control architecture.

[0023] First, the spatial foundation layer of this system consists of land parcel spatiotemporal grid construction units. Based on a preset geographic coordinate system and a minimum management unit area threshold, each unit divides the target agricultural production area into multiple rectangular spatiotemporal grids with unique land parcel IDs. Each spatiotemporal grid is then bound to a crop type identifier and a current growth stage label, thus completing the standardized binding management of land parcel space and crop growth cycle. This unit is equipped with a grid dynamic adaptation module, which is used to acquire real-time information on changes in the boundaries of target plots and crop rotation data. For plots whose boundary offset exceeds a preset threshold, a spatiotemporal grid reconstruction process is automatically triggered, synchronously updating the corresponding plot ID and the bound crop type identifier and growth stage label, while retaining the historical spatiotemporal data association of the original plot ID; for plots after crop rotation switching, the original growth stage label is cleared with one click, and the corresponding crop's full growth cycle node library is rematched to achieve precise cross-cycle management of a single plot; Meanwhile, the spatiotemporal grid construction unit of the land parcel is equipped with a grid permission isolation module, which binds the unique land parcel ID with the farmer's terminal identity information. For multi-source data, fused feature records, and agricultural decision-making suggestions of a single spatiotemporal grid, access and viewing permissions are only granted to the corresponding bound farmer terminal, realizing independent isolation and control of data for a single land parcel, preventing cross-plot data interference, and ensuring the privacy and security of field production data; Secondly, the multi-source data access unit, serving as the system's data entry layer, receives at least three types of heterogeneous data streams in real time and performs unified access and format verification. The accessed data includes multispectral image data from remote sensing satellites, containing NDVI vegetation index and land surface temperature fields; structured sensor data from IoT sensors deployed in the fields, including soil moisture, pH value, and nitrogen, phosphorus, and potassium content; and semi-structured agricultural operation logs uploaded by farmers' terminals, recording operation type, time, dosage, and location coordinates in JSON format. After the unit completes data verification, the valid data streams are pushed to the spatiotemporal semantic alignment unit for further processing. Furthermore, the spatiotemporal semantic alignment unit is used to accurately match, standardize, and calibrate heterogeneous data with the spatiotemporal grid of land parcels. Its core operations include: (1) Resample the remote sensing image data to the boundary range of the spatiotemporal grid according to the shooting time and spatial resolution to generate the rasterized index value corresponding to each plot ID; (ii) Match the sensor data to the nearest spatiotemporal grid based on the sensor's GPS coordinates, and aggregate it into a time series with a time granularity of 5 minutes; (iii) Parse the location coordinates in the agricultural operation log, map them to the corresponding plot ID, and standardize the operation type field into ontology concept nodes based on the predefined agricultural operation ontology library. The agricultural operation ontology library includes four types of operations: fertilization, irrigation, spraying, and sowing, and their subclasses. Each subclass is associated with a standard parameter template.

[0024] In some embodiments, such as Figures 1-2 As shown, this unit also includes an anomaly data removal module, which performs dual verification on rasterized remote sensing indicators and aggregated sensing time series. A threshold verification model is constructed based on the crop growth pattern of the plot to remove abnormal sensing data that exceeds the normal value range and remote sensing raster data that is distorted due to cloud interference. At the same time, for missing data, the module completes the missing data by using the historical average of the same plot and the normal data of adjacent spatiotemporal grids in a weighted fitting method, ensuring that there are no gaps or distortions after the various types of data are matched with the spatiotemporal grid, which meets the requirements of subsequent fusion feature extraction. The abnormal data removal module is also linked to a data traceability marking module. For the removed abnormal data and the supplemented missing data, the corresponding plot ID and timestamp are bound to generate an independent traceability mark, which is simultaneously stored in the plot's historical data ledger. In the subsequent process of feature extraction and agricultural decision-making, the corresponding data traceability information is retrieved simultaneously, realizing traceable control of multi-source data from collection to decision-making throughout the entire process. Subsequently, the dynamic data fusion unit extracts feature vectors from aligned remote sensing indicators, sensing time series, and standardized agricultural logs for each plot ID within the agricultural window corresponding to the current crop growth stage label. These feature vectors are then weighted and spliced ​​according to preset weight coefficients to form a fusion feature record for the plot within that window. The weight coefficients are obtained from the configuration table based on the crop type and growth stage.

[0025] In some embodiments, such as Figures 1-4 As shown, this unit incorporates a weight adaptive adjustment module that communicates bidirectionally with the decision model self-optimization unit and the spatiotemporal semantic alignment unit. The weight adaptive adjustment module reads the completeness index of the multi-source data after completion and calibration in real time, and dynamically adjusts the preset weight coefficients of the feature vectors of each data source by combining the historical decision error values ​​after the agricultural decision model iteration. This allows for real-time adaptation of weight parameters without manual intervention, ensuring that the fused feature records closely match the real-time growth status of the plots and mitigating the impact of single data source bias on subsequent decision-making results.

[0026] The weight adaptive adjustment module has a built-in dynamic weight calculation model, which realizes real-time adaptive calculation of the weights of feature vectors from various data sources through the following formula: ; in: Indicates the first Class data source in the first The adaptive weighting coefficients for each agricultural window period take values ​​in the range [0,1] and satisfy the following conditions: ; For the first The weighting coefficient of the data source in the previous agricultural window period is used to ensure the continuity of weight changes; This is the weight inertia coefficient, with a value range of [0,1], used to control the degree of influence of historical weights; To complete the integrity index of the i-th type of data source after calibration, the value range is [0,1], and it is calculated by combining the data missing rate and the anomaly removal rate; This represents the average error value of the agricultural decision-making model after making decisions based on the current weighted fusion features, and its value range is [value missing]. The larger the error value, the larger the correction coefficient for the corresponding data source weight; Assign adjustment coefficients to the weights, with values ​​ranging from [0,1], and satisfying the following conditions: + =1, which are used to adjust the contribution of data integrity indicators and decision error values ​​to the weights respectively; the weight adaptive adjustment module calculates the weight coefficients of each data source in real time based on the above formula, and updates them synchronously to the dynamic data fusion unit to ensure that the fused feature records are highly adapted to the actual growth status of the plots, effectively weakening the interference of single data source deviation on agricultural decision results.

[0027] In some embodiments, such as Figures 1-4 As shown, the decision engine calling unit is used to input the fused feature records into the agricultural decision model that matches the current crop type and growth stage, and output specific agricultural operation suggestions, including operation type, recommended time window, material type and application amount, and push them in the form of pictures and text through the farmer terminal. This unit is equipped with a tiered agricultural suggestion push module, which pre-stores a table mapping farmer terminal permission levels to plot management permissions. For generated agricultural operation suggestions, they are categorized into multi-level execution instructions based on urgency and complexity. Combined with the farmer's terminal management permissions, the module pushes the corresponding level of agricultural operation suggestions and supporting execution details. Furthermore, multiple operation suggestions for the same plot are sorted chronologically to avoid conflicts arising from overlapping agricultural operations. In addition, the system also includes a decision model self-optimization unit, which establishes a two-way data linkage channel with the dynamic data fusion unit and the decision engine invocation unit. The decision model self-optimization unit is used to collect the results of agricultural operations and the subsequent crop growth status data of the corresponding plots from the farmer's terminal in real time. The above data is used as labeled samples and fed back to the agricultural decision model built into the decision engine invocation unit to iteratively optimize the internal weight parameters of the model in real time. At the same time, independent decision sub-models are built for different crop types and different crop growth stages to achieve differentiated iteration of decision parameters within each agricultural window period, thereby improving the adaptability of agricultural operation suggestions to the actual production conditions of the plots.

[0028] In some embodiments, such as Figures 1-4 As shown, the self-optimization unit of the decision-making model is equipped with a sample screening module, which is linked to the data traceability and labeling module. For the labeled samples returned to the agricultural decision-making model, invalid samples carrying abnormal data traceability tags and missing value completion tags are first removed. Only valid samples that have passed the collection and calibration and have no data distortion are selected to participate in the model iteration, thereby improving the training accuracy of the decision-making model and the reliability of parameter iteration.

[0029] This system relies on the collaborative operation of cloud servers and edge computing nodes, adopts a highly reliable data transmission protocol and encrypted storage mechanism, supports 24 / 7 uninterrupted operation, and can stably adapt to various production scenarios such as large-scale farmland and facility agriculture, meeting the needs of precise consultation and management of agricultural production around the clock, with high real-time performance and high security.

[0030] After the system starts running, the plot spatiotemporal grid construction unit completes the grid division, label binding, and permission isolation of the target area; the multi-source data access unit accesses three types of heterogeneous data in real time: satellite remote sensing, field IoT, and farmer uploads; the spatiotemporal semantic alignment unit performs spatiotemporal matching, standardization, anomaly removal, missing data completion, and source tracing marking on the data; the dynamic data fusion unit extracts multi-dimensional features and completes data weighted fusion through a weighted adaptive algorithm to generate fusion feature records for the corresponding agricultural window period; the decision engine calling unit generates hierarchical and orderly agricultural operation suggestions based on the fusion features and matching decision model and pushes them to the corresponding farmer terminals; the decision model self-optimization unit collects feedback data and filters effective samples for iterative return, synchronously driving weight adaptive adjustment, forming a closed-loop management system of data collection, spatiotemporal alignment, data fusion, intelligent decision-making, and feedback optimization, realizing data traceability, precise decision-making, and intelligent management throughout the entire agricultural production process.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A precision consultation management system for agricultural production based on big data, characterized in that: It includes a land parcel spatiotemporal grid construction unit, which is used to divide the target area into multiple rectangular spatiotemporal grids with unique land parcel IDs according to a preset geographic coordinate system and a minimum management unit area threshold, and bind crop type identifier and current growth stage label to each spatiotemporal grid; The multi-source data access unit is used to receive at least three types of heterogeneous data streams in real time: (a) Multispectral image data from remote sensing satellites, including NDVI vegetation index and land surface temperature fields; (b) Structured sensor data from IoT sensors deployed in the field, including soil moisture, pH, and nitrogen, phosphorus and potassium content; (c) Semi-structured agricultural operation logs uploaded by farmers' terminals, which record operation type, time, amount and location coordinates in JSON format; Spatiotemporal semantic alignment unit, used to perform the following operations: (i) The remote sensing image data is resampled to the boundary range of the spatiotemporal grid according to the shooting time and spatial resolution to generate the rasterized index value corresponding to each plot ID; (ii) Match the sensor data to the nearest spatiotemporal grid based on the sensor's GPS coordinates, and aggregate it into a time series with a time granularity of 5 minutes; (iii) Parse the location coordinates in the agricultural operation log, map them to the corresponding plot ID, and standardize the operation type field into ontology concept nodes based on the predefined agricultural operation ontology library. The agricultural operation ontology library includes four types of operations: fertilization, irrigation, spraying, and sowing, and their subclasses. Each subclass is associated with a standard parameter template. The dynamic data fusion unit is used to extract feature vectors from aligned remote sensing indicators, sensing time series and standardized agricultural logs for each plot ID within the agricultural window period corresponding to the current crop growth stage label, and to perform weighted splicing according to preset weight coefficients to form a fusion feature record of the plot in the window period. The weight coefficients are obtained from the configuration table according to the crop type and growth stage. The decision engine invocation unit is used to input the fused feature records into an agricultural decision model that matches the current crop type and growth stage, and output operation suggestions.

2. The big data-based precision consultation management system for agricultural production of claim 1, wherein: The spatiotemporal grid construction unit of the land parcel is also linked to a grid dynamic adaptation module, which is used to obtain information on changes in the boundaries of land parcels in the target area and data on crop rotation in real time. For plots whose boundary offset exceeds the preset threshold, the spatiotemporal grid reconstruction process is automatically triggered, and the corresponding plot ID and the bound crop type identifier and growth stage label are updated synchronously, while retaining the historical spatiotemporal data association of the original plot ID; for plots after crop rotation switching, the original growth stage label is cleared with one click, and the corresponding crop full growth cycle node library is rematched to achieve precise cross-cycle management of a single plot. 3.The big data based agricultural production precision consultation management system according to claim 1, characterized in that: The spatiotemporal semantic alignment unit also includes an abnormal data removal module, which performs dual verification on rasterized remote sensing indicators and aggregated sensing time series respectively. A threshold verification model is constructed based on the crop growth patterns of the plots to remove abnormal sensor data that exceeds the normal value range and remote sensing raster data that is distorted due to cloud interference. At the same time, for missing data, a weighted fitting method is used to complete the missing data by using the historical average of the same plot and the normal data of adjacent spatiotemporal grids. This ensures that there are no gaps or distortions after the various types of data are matched with the spatiotemporal grids, which meets the needs of subsequent fusion feature extraction.

4. The agricultural production precision consulting management system based on big data as described in claim 1, characterized in that: It also includes a decision model self-optimization unit, which establishes a two-way data linkage channel with the dynamic data fusion unit and the decision engine invocation unit, respectively. The self-optimization unit of the decision model is used to collect the results of agricultural operations and the subsequent crop growth status data of the corresponding plots from the feedback of farmers' terminals in real time. The above data is used as labeled samples and fed back to the agricultural decision model built into the decision engine calling unit to iteratively optimize the internal weight parameters of the model in real time. Independent decision sub-models are built for different crop types and different crop growth stages to realize differentiated iteration of decision parameters within each agricultural window period.

5. The agricultural production precision consulting management system based on big data as described in claim 1, characterized in that: The dynamic data fusion unit has a built-in weight adaptive adjustment module, which communicates bidirectionally with the decision model self-optimization unit and the spatiotemporal semantic alignment unit. The weight adaptive adjustment module is used to read the integrity index of the multi-source data after completion and calibration in real time, and dynamically correct the preset weight coefficients of the feature vectors of each data source by combining the historical decision error values ​​after the iteration of the agricultural decision model.

6. The agricultural production precision consulting management system based on big data as described in claim 1, characterized in that: The decision engine calling unit is linked to a hierarchical agricultural suggestion push module, which pre-stores a table comparing the farmer's terminal permission level with the land management permission.

7. The agricultural production precision consulting management system based on big data as described in claim 3, characterized in that: The abnormal data removal module of the spatiotemporal semantic alignment unit is also linked to a data tracing and marking module; For the removed abnormal data and the supplemented missing data, each is bound to the corresponding plot ID and timestamp to generate an independent traceability identifier, which is simultaneously stored in the plot's historical data ledger. Furthermore, in the subsequent process of feature extraction and agricultural decision-making, the corresponding data traceability information is retrieved simultaneously, realizing traceable control of multi-source data from collection to decision-making throughout the entire process.

8. The agricultural production precision consulting management system based on big data as described in claim 4, characterized in that: The decision model self-optimization unit is equipped with a sample screening module, which is linked to the data traceability and marking module. For the labeled samples returned to the agricultural decision-making model, invalid samples carrying abnormal data traceability markers and missing value completion markers are first removed, and valid samples that have passed the collection and calibration and have no data distortion are selected to participate in the model iteration.

9. The agricultural production precision consulting management system based on big data as described in claim 2, characterized in that: The land parcel spatiotemporal grid construction unit is linked to a grid permission isolation module, which binds a unique land parcel ID to the farmer's terminal identity information; For multi-source data, fusion feature records, and agricultural decision-making suggestions for a single spatiotemporal grid, access and viewing permissions are only granted to the corresponding bound farmer terminals, thereby achieving independent isolation and control of data for a single plot and preventing cross-plot data interference.

10. The agricultural production precision consulting management system based on big data as described in claim 2, characterized in that: The weight adaptive adjustment module has a built-in dynamic weight calculation model, which realizes real-time adaptive calculation of the weights of feature vectors from various data sources through the following formula: ； in: Indicates the first Class data source in the first The adaptive weighting coefficients for each agricultural window period take values ​​in the range [0,1] and satisfy the following conditions: ; For the first The weighting coefficient of the data source in the previous agricultural window period is used to ensure the continuity of weight changes; This is the weight inertia coefficient, with a value range of [0,1], used to control the degree of influence of historical weights; To complete the integrity index of the i-th type of data source after calibration, the value range is [0,1], and it is calculated by combining the data missing rate and the anomaly removal rate; This represents the average error value of the agricultural decision-making model after making decisions based on the current weighted fusion features, and its value range is [value missing]. The larger the error value, the larger the correction coefficient for the corresponding data source weight; Assign adjustment coefficients to the weights, with values ​​ranging from [0,1], and satisfying the following conditions: + =1, which is used to adjust the contribution of data integrity indicators and decision error values ​​to the weights, respectively.

Images