Agricultural detection laboratory full-process digital information management system and method

Abstract

The invention provides an agricultural detection laboratory full-process digital information management system and method, the system comprises a sample receiving and management module, a data acquisition module, a data processing and analysis module, a report generation module and a data storage and traceability module, firstly, sample basic information is received, and a unique identifier is generated for each sample; then collecting physical and chemical parameters of the sample through a sensor; thirdly, performing preliminary analysis on the data by using an SVR model, further performing fine prediction through a deep neural network model, optimizing model parameters by using reinforcement learning, and finally generating a detection report and storing sample data; by combining the SVR model and the deep neural network model, the data processing efficiency and the analysis precision are improved, the parameters of the deep neural network model are optimized through the reinforcement learning model, the detection process is automatically adjusted, the problems of slow data processing and insufficient analysis precision in a traditional method are solved, and the detection efficiency is improved. And the automation level and the data processing capability of an agricultural detection laboratory are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of agricultural testing technology, and in particular to a full-process digital information management system and method for an agricultural testing laboratory. Background Art

[0002] Agricultural testing laboratories play an increasingly important role in agricultural production, especially in the testing and analysis of agricultural samples. Modern agricultural testing not only requires detailed physical and chemical parameter analysis of soil, crop and environmental samples, but also requires the effective management and storage of large amounts of test data. To meet these needs, agricultural testing laboratories are increasingly adopting digital and information management systems to improve the efficiency and accuracy of sample reception, data collection, analysis and processing, and report generation. However, current agricultural testing laboratories rely on traditional statistical methods or basic algorithms for data analysis in actual operation, lack efficient real-time data processing capabilities, and find it difficult to quickly respond to complex testing tasks, resulting in a slow data analysis process and low accuracy. At the same time, most existing agricultural testing methods use fixed analysis models and methods, lack the ability to automatically adjust analysis algorithms according to sample characteristics, and cannot dynamically optimize test results and provide suggestions for improvement measures. Finally, traditional testing systems fail to achieve automated management of the entire process from sample reception and collection to analysis and report generation. The entry and tracking of sample information mostly rely on manual labor, which is prone to omissions or errors and cannot provide reliable traceability functions. Summary of the Invention

[0003] In order to solve the technical problems mentioned in the current background technology of low sample data processing efficiency, lack of intelligent optimization mechanism and insufficient traceability management of the existing agricultural testing system, the purpose of the present invention is to provide a full-process digital information management system and method for agricultural testing laboratories.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions: A full-process digital information management system for agricultural testing laboratories, including: M1, a sample receiving and management module, for receiving basic information of agricultural samples to be tested, generating a unique identifier for each agricultural sample, and storing the basic information and identifier of the agricultural sample in a database; M2, a data acquisition module, which uses sensor equipment to collect physical and chemical parameters of the agricultural sample to be tested in real time, and stores the collected agricultural sample data corresponding to the identifier in a database; M3, a data processing and analysis module, formats the agricultural sample data and performs analysis operations to generate the agricultural sample analysis results; M4, a report generation module, inputs the formatted sample data into a deep neural network model to predict soil data of the agricultural sample, optimizes the soil data by combining it with real soil data; optimizes the parameters of the deep neural network model through a reinforcement learning model; generates a test report based on the sample analysis results and the predicted soil data and provides improvement measures; M5, data storage and traceability module, establishes an agricultural sample detection database for storing the data to be stored in M1 to M4, and supports the entire process of sample traceability from receipt, collection, analysis, report generation to improvement measures.

[0005] Furthermore, the basic information of the agricultural samples includes but is not limited to the sample source, sampling time, sample type and testing requirements; The sample source indicates the specific collection location of the agricultural sample to be tested; The sampling time indicates the specific date when the agricultural sample to be tested is collected; The sample type indicates the type of the agricultural sample to be tested; The testing requirements represent specific testing items and requirements for the agricultural samples to be tested.

[0006] Furthermore, the unique identifier of the i-th agricultural sample to be tested is represented as:

[0007] in, is the basic information vector of the i-th agricultural sample to be detected, represents the set of all basic information of the i-th agricultural sample to be tested, and n represents the number of basic information.

[0008] Furthermore, the data of the i-th agricultural sample to be tested is expressed as:

[0009] in, is the unique identifier of the i-th sample, represents the physical parameters of the i-th agricultural sample to be tested, represents the chemical parameters of the i-th agricultural sample to be tested.

[0010] Furthermore, the analysis operation is an SVR algorithm, and its objective function is as follows:

[0011] in, is the weight vector; is the bias term; is the tolerance error, and the interval is , used to define the tolerance range of prediction error; The constraints of the objective function are:

[0012]

[0013] in, For the a target value for each of the agricultural samples to be tested; is the weight vector of the SVR model; For the a characteristic vector of the agricultural sample to be tested, including physical parameters and chemical parameters; Represents a vector and eigenvectors The inner product of After the SVR model is trained, a prediction value is provided for each agricultural sample to be tested, representing the analysis result, which is specifically expressed as follows:

[0014] in, is the analysis result of the i-th agricultural sample to be tested.

[0015] Furthermore, the representation of layer I of the deep neural network model is as follows:

[0016] in, is the weighted sum of the agricultural sample data to be detected in the first layer, which is the input of each neuron; is the weight matrix of the agricultural sample data to be tested in layer I, which expresses the weighted degree of the input data; is the output of the agricultural sample data to be tested in the previous layer, The input layer of the deep neural network The formatted agricultural sample data to be tested; is the bias term of the agricultural sample data to be tested in the first layer, which is used to adjust the activation value of each neuron; ReLU is used as the activation function to process the weighted sum data, setting the negative values in the weighted sum data of each layer to 0 and keeping the positive values unchanged:

[0017] in, is the weighted sum data of the Ith layer after being processed by the ReLU activation function; Indicates the processing method of the weighted sum data results of the first layer, The output layer of the deep neural network model outputs the predicted value of soil data, which is expressed as follows:

[0018] in, is the output result, i.e. the predicted value of soil data; is the weight matrix of the output layer; The output of the weighted sum data of the penultimate layer; is the output layer bias term, which is used to adjust the predicted value of soil data; The error between the predicted value of the soil data and the actual soil data is calculated to optimize the parameters of the predicted value of the soil data, and the MSE is used as the loss function for calculation:

[0019] in, is a loss function, which represents the difference between the predicted result and the actual data; m is the number of samples of the agricultural samples to be tested; is the actual soil data of the i-th agricultural sample to be tested; is the predicted value of the soil data of the i-th agricultural sample to be tested.

[0020] Furthermore, the parameters of the deep neural network model are optimized by establishing the reinforcement learning model, and the structure of the reinforcement learning model includes the following parts: The state space of the reinforcement learning model is determined by the deep neural network model parameters and loss, and the state space is defined as:

[0021] in, for The parameter state of the deep neural network model at time , is the weight matrix of the agricultural sample data to be tested in the first layer; is the bias term of the agricultural sample data to be tested in the first layer; L represents the number of layers; represents the loss at time t; The action space represents the adjustment amount of the reinforcement learning model parameters, and the action space is defined as:

[0022] in, is the action state of the reinforcement learning model parameters at time t; Represents the adjustment amount of the data weight matrix of the first layer; Represents the adjustment amount of the bias term in the first layer; Indicates the number of layers; The reward function is designed to evaluate the quality of the results obtained after the deep neural network model parameters perform an action. That is, after the reinforcement learning model parameters select an action, the deep neural network model feeds back the reward to the reinforcement learning model. The reward function is designed as follows:

[0023] in, is the reward value calculated by the action taken by the reinforcement learning model parameters at time t; Represents the reinforcement learning model parameters at time The loss value; represents the loss value of the reinforcement learning model parameters at time t; is the indicator function; is an additional term, expressed as:

[0024] The reinforcement learning model outputs actions for calculating the parameters of the reinforcement learning model through the policy network, that is, actions The generation depends on the current state and the parameters of the policy network , the formula is:

[0025] in, Indicates that the status Next, the deterministic action value output by the policy network; It is represented by a mean of 0 and a variance of Gaussian noise term is used to increase the randomness of the strategy.

[0026] A full-process digital information management method for an agricultural testing laboratory, characterized by comprising: S1. Receive basic information of agricultural samples to be tested, generate a unique identifier for each agricultural sample to be tested, and store the basic information and identifier of the agricultural samples to be tested in a database; S2. Using a sensor device to collect physical and chemical parameters of the agricultural sample to be tested in real time, and associating the collected data with a unique identifier of the sample, and storing the data in a database; S3. Cleaning and formatting the collected data of the agricultural samples to be tested, and generating analysis results of the agricultural samples to be tested through an analysis method; S4. Input the formatted agricultural sample data to be tested into a deep neural network model for prediction, and optimize the parameters of the deep neural network model through a reinforcement learning model. The optimized deep neural network model outputs the final soil data prediction results of the agricultural sample to be tested, and generates a test report in combination with the analysis results in S3, and provides improvement measures; S5. Establish an agricultural sample detection database, and store all relevant data of the agricultural samples to be detected in the database, supporting data traceability based on unique identifiers.

[0027] Compared with the prior art, the advantages of the present invention are: 1. This invention significantly improves data processing efficiency by introducing the SVR model and deep neural network model. The SVR algorithm first quickly analyzes the collected sample data, extracts key features, and provides optimized input for subsequent deep neural network analysis. The deep neural network further accurately predicts and analyzes the data, showing higher accuracy and faster speed when processing complex data. 2. This invention achieves intelligent optimization of the sample analysis process by introducing a reinforcement learning model to optimize the training process and parameter adjustment of the deep neural network. Reinforcement learning can dynamically adjust the parameters of the analysis model based on real-time data feedback to adapt to the characteristics of different samples. This mechanism solves the problem of the lack of intelligent optimization in the existing technology to flexibly respond to complex samples and testing requirements. The testing process of each sample can be automatically optimized according to the actual situation, ensuring that the test reports and improvement measures are more accurate, reliable, and personalized. 3. The present invention assigns a unique identifier to each agricultural sample to be tested and records the basic information of the sample, test data and the entire experimental process in a digital management system, thereby realizing the full automated management of agricultural samples from reception and collection to analysis and report generation. It effectively solves the problem of incomplete sample information recording and non-traceability in traditional technologies, ensures the traceability of all data of each sample, and greatly improves the traceability of data and the accuracy of system management. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0029] Figure 1 Schematic diagram of the system workflow of the present invention; Figure 2 Schematic diagram of the working process of the method of the present invention; Figure 3 This is a schematic diagram of the agricultural sample data flow of the present invention. DETAILED DESCRIPTION

[0030] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] To achieve the above objectives, the present invention is implemented through the following technical solutions: the present invention provides a digital information management system for the entire process of agricultural testing laboratories, such as Figure 1 As shown, the system includes: M1, a sample receiving and management module, for receiving basic information of agricultural samples to be tested, generating a unique identifier for each agricultural sample, and storing the basic information and identifier of the agricultural sample in a database; The basic information of the agricultural samples includes but is not limited to sample source, sampling time, sample type and testing requirements; The sample source indicates the specific collection location of the agricultural sample to be tested, which is used to describe the environmental background of the agricultural sample to be tested; the sampling time indicates the specific date and time when the agricultural sample to be tested was collected, which is used to describe the timeliness of the agricultural sample and the impact of its environmental changes on the sample characteristics; the sample type indicates the type of agricultural sample to be tested, which is used to describe the specific category of the agricultural sample and help determine the corresponding detection method and experimental plan; the detection requirements indicate the specific detection items and requirements for the agricultural sample to be tested, which are used to clarify the experimental direction and technical route, thereby ensuring that the detection results meet expectations and standards; The unique identifier is generated based on the basic information of the agricultural sample to be tested, and the expression formula is:

[0032] in, Indicates the a unique identifier of the agricultural sample to be tested, is the basic information vector of the i-th agricultural sample to be detected, represents a set of all basic information of the i-th agricultural sample to be tested; M2, a data acquisition module, which uses sensor equipment to collect physical and chemical parameters of the agricultural sample to be tested in real time, and stores the collected agricultural sample data corresponding to the identifier in a database; Select appropriate sensor equipment based on the basic information of the agricultural sample to be detected in M1, and use the corresponding sensor equipment to collect data information; The physical parameters refer to the numerical values of the physical properties of agricultural samples, which are used to reflect the physical state and environmental conditions of the samples, including but not limited to temperature parameters, humidity parameters, conductivity parameters and light intensity parameters; the temperature affects the growth cycle of soil and plants; humidity is directly related to water management and the health of crop growth; conductivity measures the concentration of dissolved ions in soil or water, reflecting the salt or mineral content, and is used to assess soil fertility and water quality; light intensity affects plant photosynthesis and growth efficiency; Chemical parameters refer to the numerical values of chemical components in agricultural samples, which are used to reflect the chemical properties and nutrient content of the agricultural samples to be tested, including but not limited to pH parameters, nutrient content parameters, heavy metal concentration parameters, and dissolved oxygen concentration parameters. The pH value affects the acidity and alkalinity of the soil, which in turn affects the nutrient absorption of crops. The nutrient content, such as nitrogen, phosphorus, and potassium, is directly related to the growth and yield of crops. Heavy metal concentration is used to detect the content of harmful metals in soil or water to ensure environmental safety. Dissolved oxygen concentration is an important indicator of water health and directly affects the survival of aquatic organisms. After data collection is completed, the obtained physical and chemical parameter data of the agricultural sample to be tested are associated with the unique identifier previously assigned to each sample, and the expression formula is:

[0033] in, represents the data of the i-th agricultural sample to be tested, is the unique identifier of the i-th sample, represents the physical parameters of the i-th agricultural sample to be tested, represents the chemical parameters of the i-th agricultural sample to be tested.

[0034] M3, a data processing and analysis module, formats the agricultural sample data and performs analysis operations to generate the agricultural sample analysis results; The cleaning ensures the data reliability of the agricultural sample to be tested by removing outliers, filling missing values and performing consistency checks; the formatting unifies the cleaned agricultural sample data to be tested into the same unit and format; the analysis operation is used to extract valuable information from the cleaned and formatted agricultural sample data to be tested; In this embodiment, the analysis operation is a support vector regression (SVR) algorithm, which is used to analyze the data of the agricultural samples to be tested and generate analysis results corresponding to the agricultural samples. The SVR algorithm finds the optimal regression hyperplane from the training data through learning, and the SVR model is the final model trained in this process and is used to make predictions based on new input data. (1) The training goal of the SVR model is to minimize the weight of the regression hyperplane, and the training process is optimized by the objective function:

[0035] in, is the weight vector, which is used to define the direction and slope of the regression hyperplane; is a bias term used to adjust the position of the regression hyperplane; is the tolerance error, and the interval is , used to define the tolerance range of prediction error; is the part of the objective function that minimizes the sum of the squares of the weight vector; The constraints of the objective function are:

[0036]

[0037] in, For the a target value for each of the agricultural samples to be tested; is the weight vector of the SVR model; For the a feature vector of the agricultural sample to be detected; Represents a vector and eigenvectors The inner product of is the bias term; Tolerance error; (2) After the SVR model training is completed, a prediction value is provided for each agricultural sample to be tested, representing the analysis result, and the specific expression formula is:

[0038] in, The analysis result of the i-th agricultural sample to be tested; M4, a report generation module, inputs the formatted sample data into a deep neural network model to predict soil data of the agricultural sample, optimizes the soil data by combining it with real soil data; optimizes the parameters of the deep neural network model through a reinforcement learning model; generates a test report based on the sample analysis results and the predicted soil data and provides improvement measures; The deep neural network model is used to extract key features from the formatted agricultural sample data to be tested and generate prediction results for soil data through a multi-layer neural network architecture; the deep neural network model processes the agricultural sample data to be tested in M3 layer by layer; (1) The deep neural network model performs weighted summation on the formatted agricultural sample data to be tested through a weight matrix and adds a bias term. The calculation formula for the weighted sum of the first layer is:

[0039] in, is the weighted sum of the agricultural sample data to be detected in the first layer, which is the input of each neuron; is the weight matrix of the agricultural sample data to be tested in layer I, which expresses the weighted degree of the input data; is the output of the agricultural sample data to be tested in the previous layer. For the input layer, That is, the formatted agricultural sample data to be tested; is the bias term of the agricultural sample data to be tested in the first layer, which is used to adjust the activation value of each neuron; (2) The activation function plays a role in introducing nonlinearity in the deep neural network model. In this embodiment, ReLU is used as the activation function to process the weighted sum data, and the negative values in the weighted sum data of each layer are set to 0, while the positive values remain unchanged:

[0040] in, is the weighted sum data of the Ith layer after being processed by the ReLU activation function; ReLU activation function is used to process the weighted sum data z of layer I; The processing method of the weighted sum data result of the first layer is specifically expressed as follows:

[0041] (3) The weighted sum data output of the last layer of the deep neural network model is the predicted value of the soil data:

[0042] in, The output result of the weighted sum data of the last layer, that is, the predicted value of the soil data; The weight matrix of the output layer is used to connect the weighted sum data output of the second to last layer and the weighted sum data output of the last layer; The output of the weighted sum data of the second-to-last layer serves as the input of the weighted sum data of the last layer; The weighted and data-dependent bias terms for all levels are used to adjust the predicted values of soil data; (4) After the soil data prediction value is generated, the deep neural network model optimizes the parameters of the soil data prediction value by calculating the error between the soil data prediction value and the actual soil data. In this embodiment, MSE is used as the loss function for calculation:

[0043] in, is a loss function, which represents the difference between the predicted result and the actual data; n is the number of samples of the agricultural samples to be tested; is the actual soil data of the i-th agricultural sample to be tested; is the predicted value of the soil data of the i-th agricultural sample to be tested; The parameters of the deep neural network model are optimized by establishing the reinforcement learning model. The structure of the reinforcement learning model includes the following parts: The state space of the reinforcement learning model is determined by the deep neural network model parameters and loss, and the state space is defined as:

[0044] in, for The parameter state of the deep neural network model at time , is the weight matrix of the agricultural sample data to be tested in the first layer; is the bias term of the agricultural sample data to be tested in the first layer; L represents the number of layers; represents the loss at time t; The action space represents the adjustment amount of the reinforcement learning model parameters, and the action space is defined as:

[0045] in, is the action state of the reinforcement learning model parameters at time t; Represents the adjustment amount of the data weight matrix of the first layer; Represents the adjustment amount of the bias term in the first layer; Indicates the number of layers; The reward function is designed to evaluate the quality of the result obtained after the reinforcement learning model parameters perform a certain action. That is, after the reinforcement learning model parameters select a certain action, the environment feeds back the reward to the reinforcement learning model. The reward function is designed as follows:

[0046] in, is the reward value calculated by the action taken by the reinforcement learning model parameters at time t; Represents the reinforcement learning model parameters at time The loss value; represents the loss value of the reinforcement learning model parameters at time t; is the indicator function; is an additional item, which is expressed as follows in this embodiment:

[0047] The policy network output is used to calculate the action that the reinforcement learning model parameters should take at a certain moment, where Represents the action selection of the reinforcement learning model parameters at time t. The generation of this action depends on the current state and the parameters of the policy network , the formula is:

[0048] in, Indicates that the status Next, the deterministic action value output by the policy network; represents the Gaussian noise term, which is used to increase the randomness of the strategy; Output the final soil data prediction results of the agricultural sample to be tested according to the optimized deep neural network model, and generate a test report in combination with the sample analysis results in M3; This embodiment provides some examples of the improvement measures: After optimization, the deep neural network model outputs the final soil data prediction result, which shows that the soil moisture of the agricultural sample to be tested is low, and it is recommended to increase the irrigation amount; after optimization, the deep neural network model outputs the final soil data prediction result, which shows that the soil nutrients of the agricultural sample to be tested are insufficient, and it is recommended to increase the amount of fertilizer used; after optimization, the deep neural network model outputs the final soil data prediction result, which shows that the soil pH value of the agricultural sample to be tested is too high, and it is recommended to use lime to adjust the soil pH value.

[0049] M5, data storage and traceability module, establishes an agricultural sample detection database for storing the data to be stored in M1 to M4, and supports the entire process of sample traceability from receipt, collection, analysis, report generation to improvement measures.

[0050] (1) Establishing an agricultural sample testing database Naming each of the agricultural samples to be tested with the unique identifier initially assigned to the agricultural samples to be tested in M1, building an agricultural sample testing database, and dividing the agricultural sample testing database into a plurality of storage units; (2) Store all data of agricultural samples to be tested Naming each storage unit with a unique identifier of each agricultural sample to be tested, and storing the data to be stored from M1 to M4 into the storage unit accordingly; Specifically, the stored data includes: basic information of the agricultural sample to be tested in M1; physical and chemical parameters of the agricultural sample to be tested in M2; formatted data of the agricultural sample to be tested and sample analysis results of the agricultural sample data to be tested in M3; the final soil data prediction results output by the reinforcement learning model in M4, the test report, and corresponding improvement measures; (3) Data traceability The user can directly retrieve the corresponding single agricultural sample data in the agricultural sample detection database by searching the unique identifier.

[0051] The present invention provides a digital information management method for the entire process of agricultural testing laboratories. Figure 2 As shown, the method includes: S1. Receive basic information of agricultural samples to be tested, generate a unique identifier for each agricultural sample to be tested, and store the basic information and identifier of the agricultural samples to be tested in a database; S2. Using a sensor device to collect physical and chemical parameters of the agricultural sample to be tested in real time, and associating the collected data with a unique identifier of the sample, and storing the data in a database; S3. Cleaning and formatting the collected data of the agricultural samples to be tested, and generating analysis results of the agricultural samples to be tested through an analysis method; S4. Inputting the formatted agricultural sample data to be tested into a deep neural network model for prediction, and optimizing the parameters of the deep neural network model through a reinforcement learning model. The deep neural network model after optimizing the parameters outputs a test report of the agricultural sample to be tested and provides improvement measures; S5. Establish an agricultural sample detection database, and store all relevant data of the agricultural samples to be detected in the database, supporting data traceability based on unique identifiers.

[0052] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0053] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A digital information management system for the entire process of agricultural testing laboratories, characterized by: include: M1, a sample receiving and management module, for receiving basic information of agricultural samples to be tested, generating a unique identifier for each agricultural sample, and storing the basic information and identifier of the agricultural sample in a database; M2, a data acquisition module, which uses sensor equipment to collect physical and chemical parameters of the agricultural sample to be tested in real time, and stores the collected agricultural sample data corresponding to the identifier in a database; M3, a data processing and analysis module, formats the agricultural sample data and performs analysis operations to generate the agricultural sample analysis results; M4, a report generation module, inputs the formatted sample data into a deep neural network model to predict the soil data of the agricultural sample, and optimizes the soil data by combining it with real soil data; Optimizing the parameters of the deep neural network model through a reinforcement learning model; generating a test report based on the sample analysis results and the predicted soil data and providing improvement measures; M5, data storage and traceability module, establishes an agricultural sample detection database for storing the data to be stored in M1 to M4, and supports the entire process of sample traceability from receipt, collection, analysis, report generation to improvement measures.

2. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The basic information of the agricultural samples includes but is not limited to sample source, sampling time, sample type and testing requirements; The sample source indicates the specific collection location of the agricultural sample to be tested; The sampling time indicates the specific date when the agricultural sample to be tested is collected; The sample type indicates the type of the agricultural sample to be tested; The testing requirements represent specific testing items and requirements for the agricultural samples to be tested.

3. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The unique identifier of the i-th agricultural sample to be tested is represented by: ； in, is the basic information vector of the i-th agricultural sample to be detected, represents the set of all basic information of the i-th agricultural sample to be tested, and n represents the number of basic information.

4. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The data of the i-th agricultural sample to be tested is expressed as: ； in, is the unique identifier of the i-th sample, represents the physical parameters of the i-th agricultural sample to be tested, represents the chemical parameters of the i-th agricultural sample to be tested.

5. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The analysis operation is the SVR algorithm, and its objective function is as follows: ； in, is the weight vector; is the bias term; is the tolerance error, and the interval is , used to define the tolerance range of prediction error; The constraints of the objective function are: ； ； in, For the a target value for each of the agricultural samples to be tested; is the weight vector of the SVR model; For the a characteristic vector of the agricultural sample to be tested, including physical parameters and chemical parameters; Represents a vector and eigenvectors The inner product of After the SVR model training is completed, a prediction value is provided for each agricultural sample to be tested, representing the analysis result, which is specifically expressed as follows: ； in, is the analysis result of the i-th agricultural sample to be tested.

6. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The representation of layer I of the deep neural network model is as follows: ； in, is the weighted sum of the agricultural sample data to be detected in the first layer, which is the input of each neuron; is the weight matrix of the agricultural sample data to be tested in layer I, which expresses the weighted degree of the input data; is the output of the agricultural sample data to be tested in layer I-1, The input layer of the deep neural network The formatted agricultural sample data to be tested; is the bias term of the agricultural sample data to be tested in the first layer, which is used to adjust the activation value of each neuron; ReLU is used as the activation function to process the weighted sum data, setting the negative values in the weighted sum data of each layer to 0 and keeping the positive values unchanged: ； in, is the weighted sum data of the Ith layer after being processed by the ReLU activation function; Indicates the processing method of the weighted sum data results of the first layer, The output layer of the deep neural network model outputs the predicted value of soil data, which is expressed as follows: ； in, is the output result, i.e. the predicted value of soil data; is the weight matrix of the output layer; The output of the weighted sum data of the penultimate layer; is the bias term of the output layer, which is used to adjust the predicted value of soil data; The error between the predicted value of the soil data and the actual soil data is calculated to optimize the parameters of the predicted value of the soil data, and the MSE is used as the loss function for calculation: ； in, is a loss function, which represents the difference between the predicted result and the actual data; m is the number of samples of the agricultural samples to be tested; is the actual soil data of the i-th agricultural sample to be tested; is the predicted value of the soil data of the i-th agricultural sample to be tested.

7. The agricultural testing laboratory full-process digital information management system and method according to claim 1 is characterized in that: The parameters of the deep neural network model are optimized by establishing the reinforcement learning model. The structure of the reinforcement learning model includes the following parts: The state space of the reinforcement learning model is determined by the deep neural network model parameters and loss, and the state space is defined as: ； in, for The parameter state of the deep neural network model at time , is the weight matrix of the agricultural sample data to be tested in the first layer; is the bias term of the agricultural sample data to be tested in the first layer; L represents the number of layers; represents the loss at time t; The action space represents the adjustment amount of the reinforcement learning model parameters, and the action space is defined as: ； in, is the action state of the reinforcement learning model parameters at time t; Represents the adjustment amount of the data weight matrix of the first layer; Represents the adjustment amount of the bias term in the first layer; Indicates the number of layers; The reward function is designed to evaluate the quality of the results obtained after the deep neural network model parameters perform an action. That is, after the reinforcement learning model parameters select an action, the deep neural network model feeds back the reward to the reinforcement learning model. The reward function is designed as follows: ； in, is the reward value calculated by the action taken by the reinforcement learning model parameters at time t; Represents the reinforcement learning model parameters at time The loss value; represents the loss value of the reinforcement learning model parameters at time t; is the indicator function; is an additional term, expressed as: ； The reinforcement learning model outputs actions for calculating the parameters of the reinforcement learning model through the policy network, that is, actions The generation depends on the current state and the parameters of the policy network , the formula is: ； in, Indicates that the status Next, the deterministic action value output by the policy network; It is represented by a mean of 0 and a variance of Gaussian noise term is used to increase the randomness of the strategy.

8. A method for digital information management of the entire process of an agricultural testing laboratory, characterized in that: include: S1. Receive basic information of agricultural samples to be tested, generate a unique identifier for each agricultural sample to be tested, and store the basic information and identifier of the agricultural samples to be tested in a database; S2. Using a sensor device to collect physical and chemical parameters of the agricultural sample to be tested in real time, and associating the collected data with a unique identifier of the sample, and storing the data in a database; S3. Cleaning and formatting the collected data of the agricultural samples to be tested, and generating analysis results of the agricultural samples to be tested through an analysis method; S4. Input the formatted agricultural sample data to be tested into a deep neural network model for prediction, and optimize the parameters of the deep neural network model through a reinforcement learning model. The optimized deep neural network model outputs the final soil data prediction results of the agricultural sample to be tested, and generates a test report in combination with the analysis results in S3, and provides improvement measures; S5. Establish an agricultural sample detection database, and store all relevant data of the agricultural samples to be detected in the database, supporting data traceability based on unique identifiers.

Images