A Decision Support Method and System for Agricultural Fertilization and Irrigation Based on Privacy-Preserving Intersection Technology

Abstract

This invention belongs to the field of agricultural information decision-making, and provides a method and system for supporting agricultural fertilization and irrigation decisions based on privacy intersection technology. The method includes: receiving raw agricultural data sent by farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals; encrypting the raw agricultural data and storing the encrypted agricultural data in a distributed database; upon receiving an agricultural fertilization and irrigation decision request, determining key agricultural data matching the request based on privacy intersection technology and the encrypted agricultural data stored in the distributed database; establishing an agricultural fertilization and irrigation decision model based on the key agricultural data; and sending the agricultural fertilization and irrigation decision results output by the model to the farmer terminals. By using encryption and privacy intersection technology to achieve secure data sharing among all parties, and combined with the agricultural fertilization and irrigation decision model, the security of data interaction and the reliability of decision-making are improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural information decision-making technology, and in particular to an agricultural fertilization and irrigation decision support method and system based on privacy intersection technology. Background Technology

[0002] In modern agricultural production, precision fertilization and irrigation are crucial for increasing crop yields and reducing resource waste and environmental pollution. However, achieving precise decision-making faces serious data privacy issues.

[0003] On the one hand, farmers possess detailed information about their land, such as area, location, and historical planting data, which involves personal privacy and business secrets. On the other hand, agricultural research institutions possess a wealth of crop growth models, soil fertility research findings, and data on the effects of different fertilizers and irrigation methods on crops, obtained through experiments. Meanwhile, agricultural input suppliers have access to information on the composition, characteristics, and prices of fertilizers and water resources.

[0004] In the traditional process of agricultural data integration and analysis, data sharing is required to achieve precision fertilization and irrigation decisions, but this often leads to privacy risks. For example, farmers worry that their land information and planting habits may be obtained by competitors or unscrupulous businesses, research results from research institutions may be illegally used, and agricultural input suppliers may lose their commercial advantage by sharing their product information.

[0005] For the reasons mentioned above, traditional agricultural fertilization and irrigation decision-making schemes suffer from low data interaction security and low decision reliability. Summary of the Invention

[0006] This invention provides an agricultural fertilization and irrigation decision support method and system based on privacy-preserving intersection technology, which solves the shortcomings of low data interaction security and low decision reliability in traditional agricultural fertilization and irrigation decision schemes.

[0007] On one hand, this invention provides an agricultural fertilization and irrigation decision support method based on privacy-preserving intersection technology. The method is executed by an agricultural data management server, which is connected to a distributed database, farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals. The method includes:

[0008] Receive raw agricultural data sent by the farmer's terminal, the agricultural research institution's terminal, and the agricultural input supplier's terminal;

[0009] The raw agricultural data is encrypted, and the encrypted agricultural data is stored in the distributed database.

[0010] Upon receiving an agricultural fertilization and irrigation decision request initiated by the farmer's terminal, the system determines key agricultural data that matches the agricultural fertilization and irrigation decision request based on privacy intersection technology and the encrypted agricultural data stored in the distributed database.

[0011] An agricultural fertilization and irrigation decision model is established based on the aforementioned key agricultural data, and the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model are sent to the farmer's terminal.

[0012] According to the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention, the original agricultural data is encrypted, including:

[0013] Assign a corresponding data identifier to each data element in the original agricultural data;

[0014] Each data element is bound to its corresponding data identifier and encrypted to obtain encrypted agricultural data.

[0015] The agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention assigns a corresponding data identifier to each data element in the original agricultural data, including:

[0016] Based on the data source, the raw agricultural data is divided into farmer data, research institution data, and agricultural input supplier data;

[0017] Extract the land number and crop variety of each data element in the farmer data, and assign a corresponding data identifier to each data element in the farmer data based on the land number and crop variety;

[0018] Extract the research project number and experimental conditions of each data element from the research institution's data, and assign a corresponding data identifier to the research institution's data based on the research project number and experimental conditions;

[0019] Extract the product number and scope of application of each data element in the agricultural input supplier data, and assign a corresponding data identifier to the agricultural input supplier data based on the product number and scope of application.

[0020] The agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention determines key agricultural data matching the agricultural fertilization and irrigation decision request based on privacy intersection technology and encrypted agricultural data stored in the distributed database, including:

[0021] Based on the agricultural fertilization and irrigation decision request, retrieve the target encrypted agricultural data from the distributed database;

[0022] Based on the data source, each data element in the target agricultural encrypted data is assigned to different computing nodes;

[0023] Privacy intersection calculations are performed on the data elements on all the computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

[0024] According to the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention, privacy intersection calculation is performed on the data elements on all the computing nodes to obtain key agricultural data matching the agricultural fertilization and irrigation decision request, including:

[0025] Based on the data identifier of each data element, the intersection of data elements between the computing node whose data source is farmers and other computing nodes is calculated to obtain the intersection result dataset;

[0026] The data elements in the intersection result dataset are comprehensively filtered to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

[0027] According to the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention, based on the data identifier of each data element, the intersection of data elements between the computing node corresponding to the farmer's data and other computing nodes is calculated to obtain the intersection result dataset, including:

[0028] Based on the data identifier of each data element, the correlation degree between each data element in the computing node corresponding to the farmer's data and the data elements in other computing nodes is determined.

[0029] Extract the target data elements whose correlation is higher than the preset correlation threshold to obtain the intersection result dataset.

[0030] According to the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention, the data elements in the intersection result dataset are comprehensively screened to obtain key agricultural data matching the agricultural fertilization and irrigation decision request, including:

[0031] Calculate the data completeness of the intersection result dataset, and determine whether the data completeness is higher than a preset completeness threshold to obtain the completeness judgment result;

[0032] If the completeness judgment result is yes, then extract the redundant and invalid elements from the intersection result dataset;

[0033] Redundant and invalid elements are removed from the intersection result dataset to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

[0034] According to the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided by the present invention, an agricultural fertilization and irrigation decision model is established based on the key agricultural data, including:

[0035] Extract key feature data from the aforementioned key agricultural data;

[0036] The key feature data is standardized and normalized to obtain intermediate feature data;

[0037] The intermediate feature data is filtered for validity to obtain sample data;

[0038] The pre-built deep learning network model is trained based on the sample data to obtain an agricultural fertilization and irrigation decision model.

[0039] On the other hand, the present invention also provides an agricultural fertilization and irrigation decision support system based on privacy intersection technology, including: an agricultural data management server, a distributed database, a farmer terminal, an agricultural research institution terminal, and an agricultural input supplier terminal;

[0040] The agricultural data management server is connected to the distributed database, the farmer terminal, the agricultural research institution terminal, and the agricultural input supplier terminal, respectively.

[0041] The agricultural data management server is used to receive raw agricultural data sent by the farmer's terminal, the agricultural research institution's terminal, and the agricultural input supplier's terminal, encrypt the raw agricultural data, and store the encrypted agricultural data in the distributed database.

[0042] The agricultural data management server is also used to receive agricultural fertilization and irrigation decision requests initiated by the farmer's terminal, determine key agricultural data matching the agricultural fertilization and irrigation decision request based on privacy intersection technology and the agricultural encrypted data stored in the distributed database, establish an agricultural fertilization and irrigation decision model based on the key agricultural data, and send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer's terminal.

[0043] The agricultural fertilization and irrigation decision support system based on privacy intersection technology provided by the present invention includes an agricultural data management server comprising:

[0044] The data encryption module is used to receive raw agricultural data sent by the farmer terminal, the agricultural research institution terminal and the agricultural input supplier terminal, assign a corresponding data identifier to each data element in the raw agricultural data, bind and encrypt each data element with the corresponding data identifier to obtain encrypted agricultural data, and store the encrypted agricultural data in the distributed database.

[0045] The privacy intersection calculation module is used to receive agricultural fertilization and irrigation decision requests initiated by the farmer terminal, retrieve target agricultural encrypted data from the distributed database based on the agricultural fertilization and irrigation decision requests, allocate each data element in the target agricultural encrypted data to different computing nodes, and perform privacy intersection calculations on the data elements on all the computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision requests.

[0046] The model building module is used to extract key feature data from the key agricultural data, standardize and normalize the key feature data to obtain intermediate feature data, screen the intermediate feature data for effectiveness to obtain sample data, and train a pre-built deep learning network model based on the sample data to obtain an agricultural fertilization and irrigation decision model.

[0047] The result output module is used to send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer terminal.

[0048] This invention provides an agricultural fertilization and irrigation decision support method and system based on privacy intersection technology. It encrypts raw agricultural data sent from farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals, and stores the encrypted agricultural data in a distributed database. Upon receiving an agricultural fertilization and irrigation decision request from a farmer terminal, the system uses privacy intersection technology and the encrypted agricultural data stored in the distributed database to determine key agricultural data matching the request. Based on this key data, an agricultural fertilization and irrigation decision model is established, and the decision results output by the model are sent to the farmer terminal. Because the decision-making process utilizes encryption and privacy intersection technology to ensure secure data sharing among all parties, and combined with the agricultural fertilization and irrigation decision model, it can efficiently and accurately provide farmers with agricultural fertilization and irrigation decision support services, improving the data interaction security and decision reliability in the agricultural fertilization and irrigation decision-making process. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided in an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of the structure of the agricultural fertilization and irrigation decision support system based on privacy intersection technology provided in an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the structure of the agricultural data management server in an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0054] The following is combined with Figures 1 to 3 This invention describes the detailed scheme of the agricultural fertilization and irrigation decision support method and system based on privacy intersection technology provided in the embodiments of the present invention.

[0055] Figure 1 This is a flowchart illustrating the agricultural fertilization and irrigation decision support method based on privacy intersection technology provided in this embodiment of the invention.

[0056] like Figure 1 As shown in the embodiment of the present invention, the agricultural fertilization and irrigation decision support method based on privacy intersection technology can be implemented by an agricultural data management server, which is connected to a distributed database, a farmer terminal, an agricultural research institution terminal, and an agricultural input supplier terminal. The method mainly includes the following steps:

[0057] Step 110: Receive raw agricultural data sent from farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals.

[0058] In this embodiment, the farmer's terminal, the agricultural research institution's terminal, and the agricultural input supplier's terminal can all be mobile terminals or fixed terminals.

[0059] Understandably, agricultural raw data refers to the collective term for raw data related to agricultural production uploaded by farmers, agricultural research institutions, and agricultural input suppliers.

[0060] Step 120: Encrypt the raw agricultural data and store the encrypted agricultural data in a distributed database.

[0061] In this embodiment, homomorphic encryption algorithms can be used to encrypt the raw agricultural data. Homomorphic encryption allows specific computational operations to be performed on the ciphertext without decryption, thus ensuring the privacy of the raw agricultural data during storage and processing. For example, when calculating the correlation between soil fertility data and fertilizer composition data, the encrypted data can be directly used for calculation to obtain the encrypted result. Decryption is only performed when the result is finally needed. Even if the database is illegally accessed, attackers cannot obtain the plaintext content of the data.

[0062] Step 130: After receiving the agricultural fertilization and irrigation decision request initiated by the farmer's terminal, determine the key agricultural data that matches the agricultural fertilization and irrigation decision request based on privacy intersection technology and the agricultural encrypted data stored in the distributed database.

[0063] Understandably, privacy intersection technology specifically refers to a cryptographic protocol that allows two or more participants to compute the intersection of their sets without revealing other information in their respective sets.

[0064] Step 140: Establish an agricultural fertilization and irrigation decision model based on key agricultural data, and send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer's terminal.

[0065] The solution provided in this embodiment can achieve data sharing among all parties without disclosing their privacy data through privacy intersection technology. At the same time, based on the agricultural fertilization and irrigation decision-making model, it can achieve agricultural fertilization and irrigation decision support more efficiently and accurately, thereby improving the data interaction security and decision reliability in the decision-making process.

[0066] In one embodiment, the raw agricultural data is encrypted, specifically including:

[0067] First, assign a corresponding data identifier to each data element in the raw agricultural data.

[0068] In a specific implementation, a corresponding data identifier is assigned to each data element in the raw agricultural data, specifically including:

[0069] The first step is to divide the raw agricultural data into farmer data, research institution data, and agricultural input supplier data based on the data source.

[0070] It is understood that the source of the agricultural raw data in this embodiment can be farmers, research institutions, and agricultural input suppliers. Therefore, the agricultural raw data can be divided into farmer data, research institution data, and agricultural input supplier data according to the data source.

[0071] In this embodiment, farmer data may specifically include: land number, land location (such as the land's latitude and longitude information), land area, soil type, crop variety, and planting history data.

[0072] Soil type can be obtained from the soil type actively selected by the farmer's terminal, or from the soil testing report uploaded by the farmer's terminal. Planting history data specifically includes the yield of previous crops, as well as fertilization and irrigation details.

[0073] In practical applications, the aforementioned farmer data can be initially encrypted on the user terminal before being uploaded to the agricultural data management service terminal for further encryption.

[0074] The agricultural data management server can also perform format verification and integrity checks on the farmer data uploaded by the farmer's terminal to ensure the accuracy and usability of the farmer data. For farmer data that does not meet the requirements, it can promptly notify the farmer's terminal to correct or supplement it.

[0075] The data from research institutions specifically includes: crop growth model data, soil fertility research data, and experimental data on the effects of different fertilizers and irrigation methods on crop growth. The experimental data includes research project numbers and experimental conditions. Before uploading data, research institutions can use encryption algorithms compatible with agricultural data management servers to perform preliminary encryption on the data, ensuring its security.

[0076] After receiving data from research institutions, the agricultural data management server can classify and label the data for later querying and use. Simultaneously, the server can also assess the quality of the research institution data and select high-quality data for labeling.

[0077] Agricultural input supplier data specifically includes: product name, product number, product ingredients, ingredient content, applicable scope, usage method, and product price. Preliminary encryption processing can be performed on the agricultural input supplier data before uploading. Agricultural input supplier data can also include product performance evaluation data and user feedback information to enrich the data resources.

[0078] The agricultural data management server can audit and verify data from agricultural input suppliers to ensure the authenticity and reliability of the data. For false or misleading data, the agricultural data management server has the right to refuse to accept it or require the agricultural input supplier's terminal to correct it.

[0079] The second step is to extract the land number and crop variety of each data element in the farmer's data, and assign a corresponding data identifier to each data element in the farmer's data based on the land number and crop variety.

[0080] The third step is to extract the research project number and experimental conditions for each data element in the research institution's data, and assign corresponding data identifiers to the research institution's data based on the research project number and experimental conditions.

[0081] The fourth step is to extract the product number and scope of application for each data element in the agricultural input supplier data, and assign a corresponding data identifier to the agricultural input supplier data based on the product number and scope of application.

[0082] It is understood that in this embodiment, the data identifier is a string of numbers with a fixed number of digits. During the data identifier allocation process, the data on which it is based can be converted into numbers, and all or part of the numbers can be extracted as the basis for determining the data identifier.

[0083] For example, taking the data identifier of data elements in farmer data as an example, the first three digits of the land number can be extracted, and then different crop varieties can be numbered with three digits to obtain a data identifier composed of six digits.

[0084] Then, each data element is bound to its corresponding data identifier and encrypted to obtain encrypted agricultural data.

[0085] In one embodiment, based on privacy intersection technology and agricultural encrypted data stored in a distributed database, key agricultural data matching agricultural fertilization and irrigation decision requests are determined, specifically including:

[0086] The first step is to retrieve the target agricultural encrypted data from the distributed database based on the agricultural fertilization and irrigation decision request.

[0087] It is understandable that target agricultural encrypted data refers to agricultural encrypted data related to agricultural fertilization and irrigation decision requests. In practical applications, the request content in agricultural fertilization and irrigation decision requests can be extracted. This request content specifically includes key information such as the location of the land to be decided, crop variety, and soil type. Then, based on at least some of the key information in the request content, data indexing can be performed in a distributed database to retrieve target agricultural encrypted data that is related to the aforementioned key information.

[0088] The second step is to allocate each data element in the target agricultural encrypted data to different computing nodes based on the data source.

[0089] In this embodiment, at least three types of computing nodes can be set up based on the data source: computing nodes corresponding to farmers, computing nodes corresponding to agricultural research institutions, and computing nodes corresponding to agricultural input suppliers. Subsequently, the data elements in the encrypted agricultural data can be assigned to the corresponding type of computing nodes based on their data identifiers.

[0090] The third step is to perform privacy-preserving intersection calculations on the data elements on all computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision requests.

[0091] In one specific implementation, privacy-preserving intersection calculations are performed on the data elements across all computing nodes to obtain key agricultural data that matches agricultural fertilization and irrigation decision requests. This data specifically includes:

[0092] First, based on the data identifier of each data element, the intersection of data elements between the computing node whose data source is farmers and other computing nodes is calculated to obtain the intersection result dataset.

[0093] It is understood that the intersection finding process in this embodiment can be a process of matching the correlation between data elements.

[0094] For example, based on the data identifier of each data element, the intersection of data elements between the computing node corresponding to the farmer's data and other computing nodes is calculated to obtain the intersection result dataset, which specifically includes:

[0095] First, based on the data identifier of each data element, the correlation between each data element in the computing node corresponding to the farmer's data and the data elements in other computing nodes is determined.

[0096] In practical applications, a correlation table can be pre-established between the data identifiers of farmers' data, the data identifiers of research institutions' data, and the data identifiers of agricultural input suppliers' data. For example, a correlation table can be established between the crop varieties in the data identifiers of farmers' data, the experimental conditions in the data identifiers of research institutions' data, and the adaptability range in the data identifiers of agricultural input suppliers' data. The correlation table contains the correspondence between each set of data identifiers and the numerical value of the correlation degree.

[0097] Subsequently, we can extract the data identifiers of each data element in the computing node corresponding to the farmer's data and the corresponding data elements in other computing nodes. Then, we can directly query the correlation between each data element in the computing node corresponding to the current farmer's data and the data elements in other computing nodes based on the correlation lookup table.

[0098] Then, target data elements with a correlation degree higher than the preset correlation degree threshold are extracted to obtain the intersection result dataset.

[0099] In this embodiment, target data elements with a correlation degree higher than a preset correlation degree threshold are grouped together to obtain the intersection result dataset.

[0100] Finally, the data elements in the intersection result dataset are comprehensively filtered to obtain key agricultural data that matches the agricultural fertilization and irrigation decision requests.

[0101] In one specific implementation, the data elements in the intersection result dataset are comprehensively filtered to obtain key agricultural data that matches the agricultural fertilization and irrigation decision-making request, specifically including:

[0102] The first step is to calculate the data completeness of the intersection result dataset and determine whether the data completeness is higher than the preset completeness threshold to obtain the completeness judgment result.

[0103] In this embodiment, the actual data volume of the intersection result dataset can be obtained separately, and the total data volume on all computing nodes can be obtained. The total data volume on all nodes is multiplied by a preset allowable deviation coefficient to obtain the expected data volume. The data integrity of the intersection result dataset is obtained by dividing the actual data volume by the expected data volume.

[0104] It is understandable that the preset allowable deviation coefficient and data integrity value are both within the data range of 0 to 1. The preset integrity threshold can be reasonably set according to the actual accuracy requirements.

[0105] This embodiment ensures sufficient data for subsequent filtering processes by first determining the data integrity of the intersection result dataset, thus guaranteeing the effectiveness of the filtering process.

[0106] The second step is to extract redundant and invalid elements from the intersection result dataset if the completeness judgment result is yes.

[0107] In this embodiment, redundant elements refer to duplicate data elements, and invalid elements refer to data elements with obvious deviations.

[0108] The third step is to remove redundant and invalid elements from the intersection result dataset to obtain key agricultural data that matches the agricultural fertilization and irrigation decision-making request.

[0109] This embodiment improves the data quality of subsequent key agricultural data by removing redundant and invalid elements from the intersection result dataset, thereby providing accurate and effective data basis for the establishment of agricultural fertilization and irrigation decision-making models, and further improving the model accuracy of agricultural fertilization and irrigation decision-making models.

[0110] In one embodiment, an agricultural fertilization and irrigation decision-making model is established based on key agricultural data, specifically including:

[0111] The first step is to extract key feature data from key agricultural data.

[0112] In practical applications, during the key feature data extraction stage, we can not rely solely on agricultural key data, but also introduce other auxiliary data, such as meteorological data and geographic information data, so as to capture more comprehensive feature information.

[0113] In this embodiment, the key feature data specifically include soil fertility indicators (such as nitrogen content, phosphorus content, potassium content, and organic matter content), crop growth stage characteristics (such as growth days, plant height, and leaf area index), climate characteristics (such as temperature, humidity, rainfall, and sunshine duration), and agricultural input product characteristics (such as fertilizer composition, irrigation water volume, and irrigation frequency).

[0114] The second step is to standardize and normalize the key feature data to obtain intermediate feature data.

[0115] Understandably, standardizing and normalizing key feature data can make different types of feature data comparable, which facilitates subsequent model training and analysis.

[0116] In practical applications, Z-score standardization can be used to standardize key feature data. Z-score standardization converts the original data into a standardized Z-score by calculating the standard deviation and mean of the original data. Specifically, the Z-score reflects the degree of deviation between the original data and the mean, measured in units of standard deviation.

[0117] In this embodiment, the formula for Z-score standardization is:

[0118] (1)

[0119] in, For any type of key feature data, The mean of any class of key feature data. denoted as the standard deviation of any class of key feature data.

[0120] The above formula can be used to convert correlation feature data of different magnitudes into Z-score scores of a unified metric for comparison, thereby improving the comparability of key feature data of different classes.

[0121] The third step is to filter the intermediate feature data for validity to obtain sample data.

[0122] After obtaining the intermediate feature data, a filtering range can be set. Each intermediate feature data is compared with the filtering range. Data within the filtering range can be determined as valid, while data outside the filtering range can be determined as invalid, thereby achieving the validity filtering of the intermediate feature data.

[0123] The fourth step involves training a pre-built deep learning network model based on sample data to obtain an agricultural fertilization and irrigation decision model.

[0124] In the model architecture selection stage, machine learning and deep learning network models suitable for the characteristics of agricultural data can be chosen, such as deep learning network models with neural network architectures, to build agricultural fertilization and irrigation decision models. Subsequently, the model parameters and structure can be reasonably adjusted according to the scale and complexity of the data to improve the model's performance and generalization ability.

[0125] The preprocessed sample data is divided into training and test sets. The deep learning network model is trained using the training set, and the model's parameters and weights are continuously adjusted to make the model's predictions as close as possible to the actual situation. Then, the trained agricultural fertilization and irrigation decision model is evaluated using the test set, and the model's accuracy, recall, F1 score, and other indicators are calculated to determine whether the performance of the agricultural fertilization and irrigation decision model meets the requirements.

[0126] In the model architecture construction phase, the appropriate number of neural network layers and the number of nodes per layer can be selected based on the complexity of the data and the difficulty of the problem. For relatively simple agricultural decision-making problems, a neural network with 3 to 5 layers can be used, including an input layer, 1 to 3 hidden layers, and an output layer. The number of nodes in the input layer equals the number of features. For example, if there are 10 features after feature engineering, then the number of nodes in the input layer is 10. The number of nodes in the hidden layer can be determined through empirical formulas or experiments. In some embodiments, empirical formulas can be expressed as follows:

[0127] (2)

[0128] in, It is the number of hidden layer nodes. It is the number of nodes in the input layer. This is the number of output layer nodes.

[0129] In practical applications, you can first set a general range, such as 10-50 nodes, and then determine the optimal number of nodes by experimenting with different combinations of node numbers and observing the model's performance on the validation set.

[0130] In the hidden layer, the ReLU (Rectified Linear Unit) activation function is usually chosen. The ReLU function is simple to compute and can effectively alleviate the gradient vanishing problem, thus speeding up the training of neural networks.

[0131] For the output layer, an appropriate activation function can be selected based on the type of output data. If the output data is numerical, such as predicting continuous fertilizer application or irrigation water volume, no activation function or a linear activation function may be used. If it involves classification decisions (such as whether fertilization is needed or which fertilizer to choose), the Sigmoid or Softmax activation function can be selected. For example, in a binary classification problem (such as whether irrigation is needed), the Sigmoid activation function can be used; in a multi-class classification problem (such as choosing different types of fertilizer), the Softmax activation function can be used, which transforms the output data into probability distributions for each class.

[0132] During model training, the weights and biases of the deep learning network model first need to be initialized. Initialization methods can include random initialization and Xavier initialization. Random initialization generates weight and bias values ​​randomly within a certain range (such as uniform or normal distribution). Xavier initialization determines an appropriate initialization range based on the number of input and output nodes, ensuring that the weights have suitable variance during initialization, which helps to speed up training and improve model stability. For example, for weight W, if Xavier initialization is used, the initialization formula is:

[0133] (3)

[0134] In the formula, n in and n out These are the number of input nodes and the number of output nodes in the current layer, respectively.

[0135] Then, an appropriate loss function can be selected based on the nature of the problem. For regression problems (such as predicting fertilizer application or irrigation water volume), the mean squared error (MSE) loss function is usually used. For classification problems (such as determining whether fertilizer needs to be applied or selecting fertilizer type), the cross-entropy loss function can be used.

[0136] During the training loop, preprocessed sample data can be input into the deep learning network model in batches. For example, setting the batch size to 32 means that 32 samples are treated as a group for one forward and back propagation calculation.

[0137] During the training process of each batch, forward propagation is performed first, which passes the input data through the layers of the neural network to obtain the output data. Then, the loss value is calculated according to the defined loss function.

[0138] Next, backpropagation is performed. By calculating the gradient of the loss function with respect to the weights and biases, optimization algorithms (such as stochastic gradient descent SGD, Adagrad, Adadelta, Adam, etc.) are used to update the weights and biases.

[0139] The choice of learning rate is also crucial during weight and bias updates. If the learning rate is too large, the model may fail to converge or even diverge; if the learning rate is too small, the training speed will be very slow. A learning rate decay strategy can usually be adopted, gradually reducing the learning rate as training progresses. For example, after a certain number of training steps or when the validation set loss no longer decreases, the learning rate is multiplied by a decay factor less than 1 (such as a value of 0.9).

[0140] Repeat the above process until the preset number of training rounds is reached or the loss value on the validation set no longer decreases, indicating that the model has converged or is close to the optimal solution.

[0141] The trained agricultural fertilization and irrigation decision-making model is then evaluated on a validation set. During training, validation set data can be periodically input into the model to calculate the loss and evaluation metrics (such as precision, recall, F1 score, etc.; for regression problems, root mean square error (RMSE) can be calculated). By observing changes in the validation set metrics, it can be determined whether the model is overfitting or underfitting. If the validation set loss continues to rise while the training set loss continues to fall, it indicates that the model may be overfitting, requiring adjustments such as increasing the amount of data, adding regularization terms, or adjusting the model complexity.

[0142] In the actual verification stage, the hyperparameters of the deep learning network model (such as learning rate, number of hidden layer nodes, number of layers, regularization parameters, etc.) can be adjusted and optimized. For example, grid search, random search or more advanced hyperparameter optimization algorithms (such as genetic algorithm, simulated annealing algorithm, etc.) can be used.

[0143] For example, in grid search, the number of hidden layer nodes can be searched in combination with several values ​​such as 10, 20, and 30. The model is trained under different hyperparameter combinations, and the optimal hyperparameter combination is selected based on the performance on the validation set.

[0144] In some embodiments, the agricultural fertilization and irrigation decision-making model can be optimized and updated based on test results and actual feedback from farmers. If the performance of the agricultural fertilization and irrigation decision-making model is not ideal, the reasons can be analyzed and corresponding measures can be taken, such as increasing the amount of training data, adjusting the feature selection strategy, or improving the model algorithm.

[0145] The model can be updated periodically to adapt to changes in the agricultural production environment and the accumulation of new data. For example, when new crop varieties, fertilizer products, or abnormal climate changes occur, the model's parameters and structure can be adjusted in a timely manner to ensure the effectiveness and reliability of the agricultural fertilization and irrigation decision-making model.

[0146] In practical applications, personalized fertilization and irrigation plans can be generated based on the agricultural fertilization and irrigation decision-making results output by the agricultural fertilization and irrigation decision-making model, combined with the actual needs and production conditions of farmers. These plans specifically include information such as fertilization time, fertilization amount, fertilizer type selection, irrigation time, irrigation water volume, and irrigation frequency.

[0147] Ideally, fertilization and irrigation plans should be presented to farmers in a visual manner (such as charts, reports, etc.) to facilitate their understanding and implementation. Detailed explanations and instructions should also be provided to help farmers understand the basis and principles of the plans.

[0148] In a specific implementation, effective feature data, such as concept names and numbers, can be extracted from fertilization and irrigation schemes. Then, visualization charts and / or visualization reports can be generated based on the effective feature data.

[0149] Subsequently, farmers carry out agricultural production operations according to the fertilization and irrigation plan. During the implementation process, data such as soil fertility, crop growth status and weather conditions can be monitored in real time through sensors and other equipment. The measured data can be compared with the expected values ​​in the fertilization and irrigation plan. Then, the decision deviation information can be determined based on the comparison results, and the agricultural fertilization and irrigation decision model can be adjusted based on the decision deviation information.

[0150] If the decision deviation exceeds the preset deviation threshold, a deviation alert can be sent to the farmer's terminal in a timely manner to notify the farmer and make a new decision based on the new agricultural fertilization and irrigation decision results output by the adjusted agricultural fertilization and irrigation decision model, so as to ensure the smooth progress of agricultural production.

[0151] It should be noted that this embodiment extensively utilizes encryption technology throughout the entire implementation process, specifically including data transmission encryption and data storage encryption. This embodiment employs encryption protocols such as SSL / TLS to ensure the security of data during network transmission, preventing data theft or tampering. Regarding data storage, encryption algorithms can be used to encrypt the data, ensuring that even if the data is illegally obtained, its content cannot be accessed without decryption.

[0152] For example, in the data transmission encryption stage, the SSL / TLS (Secure Sockets Layer / Transport Layer Security) protocol can be used to ensure the security of data transmitted over the network. The SSL / TLS protocol prevents data from being stolen or tampered with by establishing a secure encrypted channel between each user terminal and the agricultural data management server.

[0153] Before data transmission, each user terminal and the agricultural data management server undergo a handshake process to negotiate the encryption algorithm (such as AES, RSA, etc.) and key exchange method. For example, the RSA algorithm is used for key exchange to generate a symmetric encryption key, which is then used to encrypt the data during transmission using the AES algorithm. This hybrid encryption method combines the key exchange security of asymmetric encryption (such as RSA) with the efficiency of symmetric encryption (such as AES), ensuring the confidentiality and integrity of the data during transmission.

[0154] In the data storage encryption stage, homomorphic encryption algorithms can be used to encrypt data stored in distributed databases. Homomorphic encryption allows specific computational operations to be performed on ciphertext without decrypting the data, thus ensuring data privacy during storage and processing. For example, when calculating the correlation between soil fertility data and fertilizer composition data, the encrypted data can be directly used for calculations to obtain the encrypted result, which is only decrypted when the final result is needed. In this way, even if the database is illegally accessed, attackers cannot obtain the plaintext content of the data.

[0155] In some embodiments, a strict access control mechanism can be established to authenticate and authorize users on each user terminal. Different users (farmers, agricultural research institutions, agricultural input suppliers, etc.) have different access permissions and operational privileges. For example, farmers can only access and modify their own land and production data, agricultural research institutions can access relevant research data and experimental results, but cannot modify farmers' data, and agricultural input suppliers can only view and update their own product information.

[0156] In practical applications, user permissions can be reviewed periodically and adjusted according to changes in user roles and business needs to ensure the effectiveness and security of permission management.

[0157] In some embodiments, a multi-factor authentication mechanism can be employed, combining username and password, SMS verification code, and biometric technologies (such as fingerprint recognition and facial recognition) to verify user identity. For example, when a farmer logs into the system, they first enter their username and password. After successful system verification, an SMS verification code is sent to the farmer's linked mobile phone. After entering the correct verification code, the farmer can choose to undergo fingerprint or facial recognition to further confirm their identity. Only users who pass all identity verification steps can access system resources.

[0158] In this embodiment, a role-based access control (RBAC) model is used to manage user permissions, specifically defining different roles such as farmers, agricultural research personnel, agricultural input suppliers, and administrators. Each role is assigned a different set of permissions. For example, farmers can only access and modify their own land and production data, including viewing soil test reports, historical planting records, and receiving fertilization and irrigation plans for their land; agricultural research personnel can access relevant research data and experimental results, but cannot modify farmers' data, and their access to research data is also restricted by project and data level; agricultural input suppliers can only view and update their own product information, such as detailed parameters and prices of fertilizers and irrigation equipment; administrators are responsible for the overall management and maintenance of the system, including the allocation and adjustment of user permissions, data backup and recovery, etc. The permission management system implements these permission restrictions through access control lists (ACLs) or attribute-based access control (ABAC) policies to ensure that users can only operate on data within their authorized scope.

[0159] In some embodiments, data operations throughout the entire data interaction and decision-making process can be audited and recorded, including operations such as uploading, downloading, modifying, and deleting data. Audit logs record information such as the time, user, and content of the operation, enabling tracing and investigation in the event of a data breach or other security incident.

[0160] In addition, audit logs can be analyzed regularly to identify potential security risks and abnormal behaviors, and timely measures can be taken to prevent and deal with them.

[0161] In some embodiments, the logging capabilities of a distributed database and specialized auditing software can be used to comprehensively audit and record data operations. Audit logs include the timestamp of the operation, the user's identity, the specific content of the operation (such as detailed information about data upload, download, modification, and deletion), and the source and destination IP addresses of the operation.

[0162] For example, when farmers upload new soil testing data, the audit log records information such as the upload time, the farmer's username, the file name and size of the uploaded data, and the IP address of the farmer's device. These audit logs are stored in a separate storage area with strict access control to prevent tampering or deletion.

[0163] When data breaches or other security incidents occur, data analysis and mining of audit logs can trace the source and process of the event. For example, by analyzing the time series of operations and access patterns to related data, it can be determined whether there were abnormally large amounts of data downloaded or unauthorized modifications. Data mining algorithms (such as association rule mining and anomaly detection algorithms) can uncover potential security risks and malicious behavior patterns. For instance, association rule mining can identify users frequently accessing specific types of data within a specific timeframe, which may be an anomaly signal requiring further investigation. Anomaly detection algorithms can build normal behavior models based on users' historical operational behavior; when operations deviate from the normal model, timely alerts are issued and retrospective analysis is performed to determine if a security threat exists.

[0164] In summary, the core objective of this invention is to break down data silos using privacy-preserving intersection technology and establish a secure and reliable agricultural data management server. Through this server, all parties can achieve effective data sharing and integration without exposing the original content of their sensitive data. Specifically, during data interaction, encryption algorithms and security protocols are used to process the data from all parties, performing intersection calculations only in an encrypted state to identify key data intersections related to specific land and crops, thereby providing accurate data support for subsequent decision analysis.

[0165] Based on this precise data, a personalized agricultural fertilization and irrigation decision-making model can be further constructed. This model can generate customized agricultural fertilization and irrigation decisions for farmers based on differences in soil fertility, changes in crop nutrient requirements at different growth stages, and local climate conditions (such as rainfall, temperature, and sunshine duration), combined with the characteristics of agricultural resources. For example, in areas with low soil fertility, the model will suggest increasing the application of specific fertilizers and rationally scheduling irrigation time and water volume to ensure that crops can fully absorb nutrients and water to promote growth; while in seasons or areas with more rainfall, the model will correspondingly reduce irrigation frequency and water volume to avoid water waste and disease problems caused by excessive soil moisture.

[0166] Ultimately, the solution provided by this invention enables refined management of agricultural production, improves the efficiency of agricultural resource utilization, reduces waste of fertilizer and water resources, lowers agricultural production costs, and simultaneously increases the yield and quality of agricultural products, promoting sustainable agricultural development and making a positive contribution to ensuring food security and protecting the ecological environment.

[0167] Based on the same general inventive concept, this invention also protects an agricultural fertilization and irrigation decision support system based on privacy-based intersection technology. The agricultural fertilization and irrigation decision support system based on privacy-based intersection technology provided by this invention will be described below. The agricultural fertilization and irrigation decision support system based on privacy-based intersection technology described below can be referred to in correspondence with the agricultural fertilization and irrigation decision support method based on privacy-based intersection technology described above.

[0168] Figure 2 This is a schematic diagram of the structure of an agricultural fertilization and irrigation decision support system based on privacy intersection technology provided in an embodiment of the present invention.

[0169] like Figure 2 As shown, the agricultural fertilization and irrigation decision support system based on privacy intersection technology provided in this embodiment of the invention specifically includes: an agricultural data management server 210, a distributed database 220, a farmer terminal 230, an agricultural research institution terminal 240, and an agricultural input supplier terminal 250.

[0170] The agricultural data management server 210 is connected to the distributed database 220, the farmer terminal 230, the agricultural research institution terminal 240, and the agricultural input supplier terminal 250.

[0171] The agricultural data management server 210 is used to receive raw agricultural data sent by farmer terminals 230, agricultural research institution terminals 240 and agricultural input supplier terminals 250, encrypt the raw agricultural data, and store the encrypted agricultural data in the distributed database 220.

[0172] The agricultural data management server 210 is also used to receive agricultural fertilization and irrigation decision requests initiated by the farmer terminal 230, determine the key agricultural data that matches the agricultural fertilization and irrigation decision request based on privacy intersection technology and the agricultural encrypted data stored in the distributed database 220, establish an agricultural fertilization and irrigation decision model based on the key agricultural data, and send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer terminal 230.

[0173] See Figure 3 In one embodiment, the agricultural data management server specifically includes:

[0174] The data encryption module 310 is used to receive raw agricultural data sent by farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals, assign a corresponding data identifier to each data element in the raw agricultural data, bind and encrypt each data element with the corresponding data identifier to obtain encrypted agricultural data, and store the encrypted agricultural data in a distributed database.

[0175] The privacy intersection calculation module 320 is used to receive agricultural fertilization and irrigation decision requests initiated by farmers' terminals, retrieve target agricultural encrypted data from the distributed database based on the agricultural fertilization and irrigation decision requests, distribute each data element in the target agricultural encrypted data to different computing nodes, and perform privacy intersection calculations on the data elements on all computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision requests.

[0176] The model building module 330 is used to extract key feature data from key agricultural data, standardize and normalize the key feature data to obtain intermediate feature data, screen the effectiveness of the intermediate feature data to obtain sample data, and train the pre-built deep learning network model based on the sample data to obtain an agricultural fertilization and irrigation decision model.

[0177] The result output module 340 is used to send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer's terminal.

[0178] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated further here.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A decision support method for agricultural fertilization and irrigation based on privacy-preserving intersection technology, characterized in that, The method is executed by an agricultural data management server, which is connected to a distributed database, farmer terminals, agricultural research institution terminals, and agricultural input supplier terminals; the method includes: Receive raw agricultural data sent by the farmer's terminal, the agricultural research institution's terminal, and the agricultural input supplier's terminal; The raw agricultural data is encrypted, and the encrypted agricultural data is stored in the distributed database. Upon receiving an agricultural fertilization and irrigation decision request initiated by the farmer's terminal, the system determines key agricultural data that matches the agricultural fertilization and irrigation decision request based on privacy intersection technology and the encrypted agricultural data stored in the distributed database. An agricultural fertilization and irrigation decision model is established based on the aforementioned key agricultural data, and the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model are sent to the farmer's terminal. Based on the aforementioned key agricultural data, an agricultural fertilization and irrigation decision-making model is established, including: Extract key feature data from the aforementioned key agricultural data; The key feature data is standardized and normalized to obtain intermediate feature data; The intermediate feature data is filtered for validity to obtain sample data; The pre-built deep learning network model is trained based on the sample data to obtain an agricultural fertilization and irrigation decision model.

2. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 1, characterized in that, Encrypting the raw agricultural data includes: Assign a corresponding data identifier to each data element in the original agricultural data; Each data element is bound to its corresponding data identifier and encrypted to obtain encrypted agricultural data.

3. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 2, characterized in that, Assigning a corresponding data identifier to each data element in the raw agricultural data, including: Based on the data source, the raw agricultural data is divided into farmer data, research institution data, and agricultural input supplier data; Extract the land number and crop variety of each data element in the farmer data, and assign a corresponding data identifier to each data element in the farmer data based on the land number and crop variety; Extract the research project number and experimental conditions of each data element from the research institution's data, and assign a corresponding data identifier to the research institution's data based on the research project number and experimental conditions; Extract the product number and scope of application of each data element in the agricultural input supplier data, and assign a corresponding data identifier to the agricultural input supplier data based on the product number and scope of application.

4. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 2, characterized in that, Based on privacy-preserving intersection techniques and the encrypted agricultural data stored in the distributed database, key agricultural data matching the agricultural fertilization and irrigation decision request is determined, including: Based on the agricultural fertilization and irrigation decision request, retrieve the target encrypted agricultural data from the distributed database; Based on the data source, each data element in the target agricultural encrypted data is assigned to different computing nodes; Privacy intersection calculations are performed on the data elements on all the computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

5. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 4, characterized in that, Privacy-preserving intersection calculations are performed on the data elements on all the computing nodes to obtain key agricultural data matching the agricultural fertilization and irrigation decision request, including: Based on the data identifier of each data element, the intersection of data elements between the computing node whose data source is farmers and other computing nodes is calculated to obtain the intersection result dataset; The data elements in the intersection result dataset are comprehensively filtered to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

6. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 5, characterized in that, Based on the data identifier of each data element, the intersection of data elements between the computing node corresponding to the farmer's data and other computing nodes is calculated to obtain the intersection result dataset, including: Based on the data identifier of each data element, the correlation degree between each data element in the computing node corresponding to the farmer's data and the data elements in other computing nodes is determined. Extract the target data elements whose correlation is higher than the preset correlation threshold to obtain the intersection result dataset.

7. The agricultural fertilization and irrigation decision support method based on privacy intersection technology according to claim 5, characterized in that, The data elements in the intersection result dataset are comprehensively filtered to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request, including: Calculate the data completeness of the intersection result dataset, and determine whether the data completeness is higher than a preset completeness threshold to obtain the completeness judgment result; If the completeness judgment result is yes, then extract the redundant and invalid elements from the intersection result dataset; Redundant and invalid elements are removed from the intersection result dataset to obtain key agricultural data that matches the agricultural fertilization and irrigation decision request.

8. An agricultural fertilization and irrigation decision support system based on privacy-preserving intersection technology, characterized in that, Based on the method according to any one of claims 1 to 7, the system includes: an agricultural data management server, a distributed database, a farmer terminal, an agricultural research institution terminal, and an agricultural input supplier terminal; The agricultural data management server is connected to the distributed database, the farmer terminal, the agricultural research institution terminal, and the agricultural input supplier terminal, respectively. The agricultural data management server is used to receive raw agricultural data sent by the farmer's terminal, the agricultural research institution's terminal, and the agricultural input supplier's terminal, encrypt the raw agricultural data, and store the encrypted agricultural data in the distributed database. The agricultural data management server is also used to receive agricultural fertilization and irrigation decision requests initiated by the farmer's terminal, determine key agricultural data matching the agricultural fertilization and irrigation decision request based on privacy intersection technology and the encrypted agricultural data stored in the distributed database, establish an agricultural fertilization and irrigation decision model based on the key agricultural data, and send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer's terminal; establishing the agricultural fertilization and irrigation decision model based on the key agricultural data includes: extracting key feature data from the key agricultural data; standardizing and normalizing the key feature data to obtain intermediate feature data; filtering the intermediate feature data for validity to obtain sample data; and training a pre-constructed deep learning network model based on the sample data to obtain the agricultural fertilization and irrigation decision model.

9. The agricultural fertilization and irrigation decision support system based on privacy intersection technology according to claim 8, characterized in that, The agricultural data management server includes: The data encryption module is used to receive raw agricultural data sent by the farmer terminal, the agricultural research institution terminal and the agricultural input supplier terminal, assign a corresponding data identifier to each data element in the raw agricultural data, bind and encrypt each data element with the corresponding data identifier to obtain encrypted agricultural data, and store the encrypted agricultural data in the distributed database. The privacy intersection calculation module is used to receive agricultural fertilization and irrigation decision requests initiated by the farmer terminal, retrieve target agricultural encrypted data from the distributed database based on the agricultural fertilization and irrigation decision requests, allocate each data element in the target agricultural encrypted data to different computing nodes, and perform privacy intersection calculations on the data elements on all the computing nodes to obtain key agricultural data that matches the agricultural fertilization and irrigation decision requests. The model building module is used to extract key feature data from the key agricultural data, standardize and normalize the key feature data to obtain intermediate feature data, screen the intermediate feature data for effectiveness to obtain sample data, and train a pre-built deep learning network model based on the sample data to obtain an agricultural fertilization and irrigation decision model. The result output module is used to send the agricultural fertilization and irrigation decision results output by the agricultural fertilization and irrigation decision model to the farmer terminal.

Images