Agricultural planting decision-making method and system

Abstract

The invention belongs to the technical field of intelligent agricultural planting, and particularly relates to an agricultural planting decision-making method and system based on a large language model.The technological process of the method includes the steps that firstly, historical field data and field data are preprocessed, and then processing such as feature extraction, clustering and sensor data fusion is conducted; and finally, predicting according to the agricultural data set and sending to the user. The concrete structure of the system comprises a data processing module, an intelligent agent module and an agricultural interaction model module. The system can be widely applied to precision agriculture, and helps farmers to implement precise management strategies such as precise fertilization, irrigation and pest control according to specific environment and crop requirements, so that the yield and quality of crops are improved. According to intelligent greenhouse management, environment parameters such as temperature, humidity and illumination are monitored in real time through an environment sensor, and after data are processed, an intelligent agent can provide an optimized management strategy to adjust conditions in the greenhouse, so that an optimal environment is created for growth of plants.

Description

Technical Field

[0001] The present invention belongs to the field of smart agricultural planting technology, and specifically relates to an agricultural planting decision-making method and system based on a large language model, which detects and predicts agriculture through the interaction between an intelligent agent and a large language model, thereby realizing the intelligence and precision of agricultural production. Background Art

[0002] With the rapid development of artificial intelligence technology, more and more artificial intelligence technologies are being applied to the process of automated decision-making, making the decision-making method more efficient and accurate. However, a single model is difficult to meet the accuracy requirements, and complex models require more storage resources and computing power. For example, machine learning models require multiple data sets. If the data set exceeds a certain scale, it will face the problem of low computing efficiency; at the same time, it has poor adaptability when processing new types of tasks, and requires large-scale data training to obtain better fitting results, which requires huge storage space and computing power; in addition, under conditions such as light intensity and posture changes, the robustness of machine learning models is poor. In the prior art, Chinese patent 202410463204.2 discloses a smart home control method based on AI dialogue, comprising the steps of: constructing a model based on a cloud data center to obtain a control voice recognition model, an AI dialogue model, and a voice response generation model; using a smart home control device to connect to the cloud data center, downloading and deploying a reconstructed control voice recognition model, a reconstructed AI dialogue model, and a reconstructed voice response generation model; based on the smart home control device, collecting the user's real-time control voice data, and using the reconstructed control voice recognition model to generate corresponding real-time control text data based on the real-time control voice data; using the reconstructed AI dialogue model to generate corresponding real-time control instructions and real-time response text data based on the real-time control text data; controlling the corresponding smart home device based on the real-time control instruction, and using the reconstructed voice response generation model to generate corresponding real-time response voice data based on the real-time response text data. Chinese patent 202311742169.X discloses a smart agricultural system based on AI technology, including: an AI cloud platform, which is the main control end of the system, used to issue execution commands, and intelligently manage the entire agricultural production process through the Internet of Things; a user end, the user end performs artificial training on the AI ​​cloud platform, and the AI ​​cloud platform feeds back the algorithm solution to the user end; a soil data acquisition module, which is used to monitor the data of the soil in the farmland and feed back the monitored data to the AI ​​cloud platform; an environmental monitoring module, which is used to monitor the environment around the farmland and feed back the monitored data to the AI ​​cloud platform; an irrigation module, which receives an irrigation start / stop instruction sent by the AI ​​cloud platform and then irrigates the farmland; a visual recognition module, which identifies crops in the farmland and then feeds back the identified crop images to the AI ​​cloud platform; an insecticide module, which receives an insecticide start instruction sent by the AI ​​cloud platform and automatically turns off the insecticide module after the insecticide is completed.Chinese patent 202311769495.X discloses a smart agricultural management method based on AI video processing, which includes the following steps: S1. Real-time collection of video information in the target area, and equipping each management object with an intelligent device; S2. Matching the management object with the target object in the video, and calculating the state index of the target object; S3. Analyzing the correlation index and danger index between the target objects, and uploading the analysis results to the management background; S4. Receive the decision results of the management background and execute them, continue to track the target object or light up the light in response. The above patents and prior arts occupy more storage resources and computing power, have inaccurate agricultural ecological monitoring, low data processing efficiency, and cannot scientifically adjust the environment and soil in agriculture according to the actual situation of the plant. Therefore, a method and system for agricultural planting decision-making using a large language model is developed and designed, which is widely used in precision agriculture to help farmers implement precise management strategies based on specific environmental and crop needs. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art, develop and design an agricultural planting decision-making method and system using a large language model, and combine an intelligent agent with the large language model to make agricultural planting decisions.

[0004] In order to achieve the above-mentioned purpose, the specific process of the agricultural planting decision-making method involved in the present invention is:

[0005] First, historical field data are obtained and field data are collected through acquisition equipment, and the historical field data and field data are preprocessed to obtain historical agricultural data sets and agricultural data sets;

[0006] Then, the historical agricultural data set and agricultural data set are processed by feature extraction, clustering and sensor data fusion, and the processed historical agricultural data set and agricultural data set are transmitted to the intelligent agent. At the same time, the historical field data is transmitted to the agricultural development platform;

[0007] Finally, the agent makes predictions based on the agricultural dataset and sends the prediction results and historical field data to users through the agricultural development platform so that users can query solutions.

[0008] In this process, the prediction results and historical field data are analyzed separately, and the prediction result prompt words and historical field data prompt words are obtained respectively. The prediction result prompt words are analyzed with reference to the historical field data prompt words to obtain the final prompt words;

[0009] Output the final prompt words to the agricultural interaction model;

[0010] The agricultural interaction model retrieves the agricultural knowledge base according to the final prompt word, trains the model according to the agricultural knowledge base, obtains planting suggestions, and outputs the planting suggestions in the form of human-computer dialogue or on the corresponding equipment, and performs corresponding operations (including controlling temperature and humidity, spraying water, and fertilizing).

[0011] The present invention relates to an agricultural planting decision-making system, comprising:

[0012] The data processing module is responsible for acquiring historical field data and collecting field data through acquisition equipment, preprocessing the field data of the historical field data set to obtain the historical agricultural data set and agricultural data set; performing feature extraction, clustering, and sensor data fusion operations on the historical agricultural data set and agricultural data set, and transmitting the operated historical agricultural data set and agricultural data set to the intelligent agent module;

[0013] The intelligent agent module is responsible for making predictions based on the agricultural data set and outputting the prediction results to the agricultural development platform. The agricultural development platform sends the prediction results and historical field data to the user respectively. The user makes a query and sends the query information to the intelligent agent through the agricultural development platform. The intelligent agent extracts prompt words from the query information, historical agricultural data and prediction data, and retrieves the agricultural knowledge base according to the prompt words, and outputs the agricultural knowledge base information to the agricultural interactive model module.

[0014] The agricultural interaction model module is responsible for training the agricultural interaction model according to the information in the agricultural knowledge base; the user asks agricultural knowledge through voice recognition or mobile devices, analyzes the agricultural knowledge asked, and transmits the analysis results to the intelligent body. The intelligent body compares and analyzes the agricultural knowledge with the prompt words, retrieves the agricultural knowledge base, and forms planting suggestions; the planting suggestions are output to the agricultural interaction model, and the agricultural interaction model processes the planting suggestions and outputs them to the user; and the user operates the agricultural device through voice recognition or mobile devices; the agricultural interaction model analyzes the agricultural device operation signal sent by the user, and transmits the analysis results to the intelligent body; the intelligent body transmits the signal to the agricultural device through wireless sensor signals to control the operation of the agricultural device; after the work is completed, feedback information will be provided to the user to inform whether the operation is successful and the operation effect.

[0015] Compared with the existing technology, the present invention relies on advanced information technology and intelligent algorithms, covering the complete chain from data collection to data analysis to decision support, realizing the intelligence and precision of agricultural production; the integrated application not only improves the efficiency and sustainability of agricultural production, but also provides farmers with scientific decision-making support, which helps to improve the yield and quality of crops. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a process flow chart of the agricultural planting decision-making method.

[0017] Figure 2 It is a schematic diagram of the principle of the agricultural planting decision-making method involved in the present invention.

[0018] Figure 3 The present invention provides a process flow chart for feature extraction, clustering and sensor data fusion.

[0019] Figure 4 The present invention is a process flow chart of weighted fusion.

[0020] Figure 5 A process flow chart for making predictions for the intelligent agent involved in the present invention.

[0021] Figure 6 A process flow chart for a user query solution according to the present invention.

[0022] Figure 7 The present invention relates to a process flow chart for answering agricultural knowledge base.

[0023] Figure 8 The present invention provides a process flow chart for extracting final prompt words.

[0024] Fig. 9 This is a process flow chart for training an agricultural interactive model for answering agricultural knowledge bases involved in the present invention.

[0025] Fig.10 A process flow chart for improving feedback and updating feedback of the agricultural knowledge base involved in the present invention.

[0026] Fig.11 This is a process flow chart for adjusting the agricultural interaction model involved in the present invention.

[0027] Fig.12 The present invention is a process flow chart for implementing the solution.

[0028] Fig.13 The present invention is a structural block diagram of the agricultural planting decision-making system involved.

[0029] Fig.14 The present invention is a structural block diagram of an electronic device.

[0030] Fig.15 The present invention is a structural block diagram of a computer-readable storage medium. DETAILED DESCRIPTION

[0031] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0032] Embodiment 1:

[0033] The present embodiment involves an agricultural planting decision-making method, comprising the following steps:

[0034] Step S100: acquiring historical field data and collecting field data through acquisition equipment, preprocessing the historical field data and field data to obtain a historical agricultural data set and an agricultural data set;

[0035] Perform operations such as feature extraction, clustering, and sensor data fusion on the historical agricultural data set and the agricultural data set, and transmit the operated historical agricultural data set and the agricultural data set to the intelligent agent;

[0036] Transmit historical field data to agricultural development platforms;

[0037] Among them, field data include environmental sensing, crop phenotype, and weather forecast data;

[0038] Step S200: the intelligent agent makes predictions based on the agricultural data set and outputs the prediction results to the agricultural development platform;

[0039] The agricultural development platform will send the prediction results and historical field data to users, who will query solutions and execute solutions.

[0040] The prediction results and the historical field data are analyzed separately to obtain the prediction result prompt words and the historical field data prompt words, and the prediction result prompt words are analyzed with reference to the historical field data prompt words to obtain the final prompt words;

[0041] Output the final prompt words to the agricultural interaction model;

[0042] The agricultural interactive model retrieves the agricultural knowledge base according to the prompt words, trains the model according to the agricultural knowledge base, and obtains planting suggestions;

[0043] The agricultural knowledge base includes crop planting codes, agricultural record data, and national agricultural standards and specifications;

[0044] Step S300: The agricultural interaction model outputs the planting suggestions in the form of human-computer dialogue or on the corresponding device, and performs corresponding operations, including controlling temperature and humidity, spraying water, and applying fertilizers.

[0045] The working principle of an agricultural planting decision method involved in this embodiment is as follows: when it is necessary to obtain agricultural planting decisions, the field data collected by the acquisition equipment includes environmental sensing, crop phenotypes, and weather forecast data, etc., and these data are pre-processed to form historical agricultural data sets and agricultural data sets. These data sets are processed by operations such as feature extraction, clustering, and sensor data fusion to improve the accuracy and availability of the data. The intelligent agent uses the processed agricultural data set to make predictions and outputs the prediction results to the agricultural development platform. The agricultural development platform further sends the prediction results and historical field data to the user, and the user can query and execute solutions based on this information. The analysis of the prediction results and historical field data can generate prediction result prompt words and historical field data prompt words, which are parsed to form final prompt words, which are used to guide the agricultural interactive model to retrieve the agricultural knowledge base and train the model, and finally obtain planting suggestions. The agricultural interactive model outputs the planting suggestions in the form of human-computer dialogue or on the corresponding device, and performs corresponding operations, such as controlling temperature and humidity, spraying water, and fertilizing. The entire process shows a complete chain from data collection to data analysis to decision support, and each step relies on advanced information technology and intelligent algorithms to realize the intelligence and precision of agricultural production. This integrated application not only improves the efficiency and sustainability of agricultural production, but also provides farmers with scientific decision-making support, helping to improve crop yield and quality.

[0046] The agricultural planting decision-making method involved in this embodiment can be widely used in precision agriculture to help farmers implement precise management strategies such as precise fertilization, irrigation and pest control according to specific environmental and crop needs, thereby improving crop yield and quality. Intelligent greenhouse management, through environmental sensors to monitor environmental parameters such as temperature, humidity, and light in real time. After the data is processed, the intelligent agent can provide optimized management strategies to adjust the conditions in the greenhouse to create the best environment for plant growth; for example, according to real-time data to control temperature and humidity, automatic water spraying and fertilization operations to ensure that plants obtain the best growth conditions. Agricultural production planning, using historical data and prediction results to formulate planting plans. Risk management and post-disaster recovery, through real-time weather forecasts and environmental sensor data, predict potential risks, and provide response strategies based on historical data and knowledge bases, so as to help farmers respond quickly and reduce risk losses. Product traceability and quality control, through the analysis of agricultural record data, can realize the traceability mechanism of agricultural products to ensure the safety and quality of products. Provide decision support for governments and agricultural institutions, including data analysis and decision support; through predictive analysis, historical data trends, etc., policymakers can better understand the current status of agricultural production, formulate practical policies, and promote agricultural development.

[0047] Embodiment 2:

[0048] This embodiment is improved on the basis of embodiment 1. The process of feature extraction, clustering and sensor data fusion in step S100 includes the following steps:

[0049] Step S101: preprocessing the historical field data and the field data, removing noise and outliers in the historical field data and the field data, converting the historical field data and the field data into a data standardized form, and obtaining a historical agricultural data set and an agricultural data set; transmitting the historical agricultural data set and the agricultural data set to the intelligent agent;

[0050] Step S102: Integrate and cluster the historical agricultural data set and the agricultural data set according to the time series respectively; perform weighted fusion of the integrated cluster data on each sensor to form a comprehensive field data set and a historical comprehensive data set, and extract features from the comprehensive field data set and the historical comprehensive field data set;

[0051] Step S103: The intelligent agent analyzes the characteristic data of the comprehensive field data set and the historical comprehensive field data set to identify the crop growth stages or regions under different environmental conditions; compares the characteristic data of the comprehensive field data set and the historical comprehensive field data set, and performs predictive processing on the field data.

[0052] Among them, step S101 improves the quality and consistency of data by preprocessing historical field data and field data, removing noise and outliers, and converting the data into a standardized form; it helps the intelligent agent to better understand and analyze the data and improve the accuracy of prediction and decision-making;

[0053] Step S102 forms a comprehensive field data set and a historical comprehensive data set by integrating clustering and weighted fusion, extracts features, and helps the intelligent agent identify the crop growth stage or area under different environmental conditions, thereby improving the level of refinement of crop management;

[0054] Step S103 analyzes and compares the feature data through the intelligent agent and performs predictive processing on the field data, which helps to improve the efficiency and resource utilization of agricultural production, reduce waste and improve economic benefits.

[0055] Embodiment 3:

[0056] This embodiment is improved on the basis of embodiment 2. In step S102, the cluster data is integrated to perform weighted fusion on each sensor. Through methods such as dynamic time planning, median absolute deviation method, weight allocation and feature-level data fusion, the accuracy and efficiency of multi-sensor data fusion are significantly improved, and the influence of errors and noise is reduced. The specific process includes the following steps:

[0057] Step S1021: The dynamic time planning distance method is used to calculate the distance of the sensors at each time point within a period of time. By calculating the distance of each sensor data, the median absolute deviation method is used to set the correlation threshold. The low correlation sensor data exceeding the threshold will not participate in this fusion.

[0058] Step S1022: Calculate the correlation between the sensors based on the calculated dynamic time planning distance, and assign weights based on the strength of the correlation; the higher the correlation between the terminal perception data, the greater the assigned weight; perform weighted data fusion on the currently collected perception data based on the size of the weight;

[0059] Step S1023: abstract the sensor data, obtain the associated characteristic values ​​between the data through analysis, abstract the sensor data, extract the key characteristic information of each type of perception data, and perform feature-level data fusion by utilizing the association of these characteristic information.

[0060] Among them, step S1021 calculates the distance of the sensors at each time point by the dynamic time planning distance method, and uses the median absolute deviation method to set the correlation threshold, excludes low-correlation sensor data, thereby improving the accuracy and reliability of data fusion, effectively eliminating noise and outliers, and reducing errors in data fusion;

[0061] Step S1022 calculates the correlation between sensors according to the dynamic time planning distance and distributes weights. The higher the correlation between the terminal perception data, the greater the weight allocated. The weight is reasonably allocated to improve the accuracy and efficiency of data fusion.

[0062] Step S1023 abstracts the sensor data, extracts key feature information, and uses the association of these feature information to perform feature-level data fusion, effectively utilizing the redundant and complementary information in the data to improve the effect of data fusion.

[0063] Embodiment 4:

[0064] This embodiment is improved on the basis of embodiment 1. In step S200, the intelligent agent makes predictions based on the agricultural data set. By introducing neural network and intelligent agent technology, the simulation and prediction capabilities of the traditional crop growth model are optimized, and the intelligent level of agricultural production is improved. Not only the accuracy and real-time performance of the simulation are improved, but also the recognition and early warning capabilities of abnormal crop growth are enhanced, providing technical support for the development of precision agriculture and smart agriculture. The specific process includes the following steps:

[0065] Step S201: Initialize the weight and threshold of each neuron in the agent;

[0066] Step S202: Input the features in the agricultural data set into the intelligent agent and calculate the weighted sum of each neuron; each neuron corresponds to each feature, and the neuron will simulate the crop growth process and obtain the final crop growth data results;

[0067] Step S203: Compare the final crop growth data results with the features in the historical agricultural data set to obtain differences and analyze the differences, wherein the differences include slow crop growth cycle, insect pests or nutritional imbalance.

[0068] Among them, step S201 initializes the weight and threshold of each neuron in the intelligent agent, provides a basic neural network structure for the intelligent agent, ensures that each neuron has an initial state, which is crucial for weighted sum calculation and crop growth simulation, and can start the neural network more quickly, reduce the startup time, and improve the response speed of the system;

[0069] Step S202 inputs the features in the agricultural data set into the intelligent agent, calculates the weighted sum of each neuron, and can dynamically reflect the growth status of crops under different environmental conditions by simulating the crop growth process. It can more accurately simulate the crop growth process and provide more real-time growth data, which helps to timely discover and solve problems in crop growth;

[0070] Step S203 compares the final crop growth data results with the features in the historical agricultural data set, obtains the differences, and analyzes the differences; through comparative analysis, it is possible to identify abnormal conditions in the crop growth process, such as slow growth cycle, insect pests or nutritional imbalance, thereby providing decision support for agricultural production, being able to more comprehensively evaluate the crop growth status, provide more accurate early warning and guidance, and help improve the efficiency and quality of agricultural production.

[0071] Embodiment 5:

[0072] This embodiment is improved on the basis of embodiment 1. In step S200, the user will query the solution. By introducing advanced technologies such as deep learning, natural language processing and machine learning, the intelligence level and response speed of the agricultural question-answering system are significantly improved, human intervention is reduced, and the adaptability and accuracy of the system are improved. The specific process includes the following steps:

[0073] Step S204: Preliminary recognition of the question voice issued by the user is performed to determine the purpose of the user's question, and the intelligent agent extracts the final prompt word and retrieves the relevant answer agricultural knowledge base according to the prompt word;

[0074] Step S205: The agricultural interaction model is trained according to the agricultural knowledge base of the dialogue to obtain a number of dialogue texts; the intelligent agent obtains the dialogue information, performs agricultural prediction on the dialogue information, compares the prediction result with the historical data result, and generates information if it is less than the threshold value and outputs it to the user; if the prediction result is greater than the threshold value, information improvement feedback and agricultural knowledge base update feedback are generated;

[0075] Step S206: adjusting the agricultural interaction model construction method and the agricultural interaction model according to the information improvement feedback and the agricultural knowledge base update feedback.

[0076] Among them, step S204 implements fast and accurate question and answer by parsing user queries, identifying key words, and matching information in the knowledge base. The question and answer system based on deep learning and natural language processing technology can better adapt to the changing Chinese context and improve the robustness and accuracy of the system;

[0077] Step S205 improves the accuracy and efficiency of answer selection, reduces feature loss, and achieves fast and accurate similarity matching of rice question texts through a dynamic attention mechanism and a multi-strategy matching model;

[0078] Step S206 is based on the ontology-based wide-area agricultural information service system framework, which can automatically answer user questions, provide real-time and non-real-time interaction between users and experts, and automatically organize, reorganize and update the knowledge base to achieve information proliferation.

[0079] Embodiment 6:

[0080] This embodiment is improved on the basis of embodiment 5. The process of answering the knowledge base in step S204 includes the following steps:

[0081] Step S2041: obtaining a question voice issued by the user, establishing a voice rhythm of the question voice according to the rhythmic features of the question voice, dividing the question voice into a plurality of rhythm segments according to the voice rhythm, and establishing a recognition frequency of the corresponding rhythm segment according to the rhythm strength corresponding to each rhythm segment;

[0082] Step S2042: Based on the recognition rate, the corresponding rhythm segment is subjected to language recognition using a preset speech recognition technology to obtain semantic information corresponding to the rhythm segment, and the semantic information is reorganized to obtain a user question text;

[0083] Step S2043: semantically analyze the question text to obtain several question purposes of the user, and compare and analyze the question purpose with the final prompt word to obtain the corresponding agricultural knowledge index; search the agricultural knowledge base for the answer whose correlation with the corresponding question purpose is greater than a preset correlation threshold according to the agricultural knowledge index.

[0084] Among them, step S2041 establishes speech rhythm by analyzing the rhythmic features of the user's question speech, and divides the question speech into several rhythm segments, and establishes a recognition frequency according to the rhythm strength of each rhythm segment; the accuracy of speech recognition is improved, and the changes in rhythm and strength can more accurately reflect the user's intention and semantic information;

[0085] In step S2042, based on the recognition frequency, the corresponding rhythm segment is recognized by using the preset speech recognition technology to obtain semantic information corresponding to the rhythm segment, and the semantic information is reorganized to obtain the user's question text; the recognition accuracy and efficiency are improved by the preset speech recognition technology, and the final text is more in line with the user's actual intention by reorganizing the semantic information;

[0086] Step S2043 performs semantic analysis on the question text to obtain several question purposes of the user, and compares and analyzes the question purpose with the final prompt word to obtain the corresponding agricultural knowledge index; searches for the corresponding agricultural knowledge base whose correlation with the corresponding question purpose is greater than a preset correlation threshold according to the agricultural knowledge index; through semantic analysis and comparative analysis, the accuracy and efficiency of agricultural knowledge retrieval are improved, so that users can quickly obtain professional agricultural knowledge related to their questions.

[0087] Embodiment 7:

[0088] This embodiment is improved on the basis of embodiment 5. In step S205, according to the information improvement feedback and the agricultural knowledge base update feedback, through pre-question-answer separation, dynamic prediction, real-time feedback and knowledge base update, the accuracy, adaptability and efficiency of agricultural management are significantly improved, and the application effect of the existing technology in agricultural production is optimized. The specific process includes the following steps:

[0089] Step S2051: obtaining the dialogue information between the user and the agricultural interaction model, performing pre-question-answer separation on the pre-answer information, and obtaining the question information corresponding to the user and the pre-answer information of the agricultural interaction model;

[0090] Step S2052: predicting crop growth based on the question information, comparing the prediction results with historical data, and setting thresholds based on the historical data;

[0091] Step S2053: Output the comparison result to the agricultural interaction model; if it is less than the threshold, generate planting suggestions and output them to the user; if it is greater than the threshold, generate improvement feedback and agricultural knowledge base update feedback; the agricultural interaction model adjusts the construction method according to the information improvement feedback and agricultural knowledge base update feedback.

[0092] Among them, step S2051 can more accurately identify the user's question content and the model's predicted answer content through pre-question-answer separation, thereby improving the accuracy and efficiency of subsequent steps;

[0093] Step S2052 can dynamically adjust the prediction model by comparing crop growth prediction with historical data to improve the accuracy and adaptability of the prediction;

[0094] Step S2053 can adjust agricultural management strategies in real time and improve agricultural production efficiency by generating planting suggestions and improvement feedback.

[0095] Embodiment 8:

[0096] This embodiment is improved on the basis of embodiment 5. The process of adjusting the agricultural interaction model in step S206 includes the following steps:

[0097] Step S2061: Determine the defect features of the agricultural interaction model according to the information improvement feedback, and determine the agricultural knowledge to be supplemented features of the corresponding agricultural knowledge base according to the agricultural knowledge and the new feedback;

[0098] Step S2062: establishing defect positions of the agricultural interaction model according to defect characteristics of the agricultural interaction model, and adjusting the model structure of the agricultural interaction model according to the defect positions;

[0099] Step S2063: retrieve the agricultural knowledge to be supplemented in the agricultural knowledge base update feedback, determine the agricultural knowledge association method between the agricultural knowledge to be supplemented and the existing agricultural knowledge based on the agricultural knowledge to be supplemented, and use the agricultural knowledge to supplement the corresponding agricultural knowledge base.

[0100] Among them, in step S2061, by analyzing user feedback, the defective features in the agricultural interaction model are identified, and the knowledge in the agricultural knowledge base is adjusted according to these features; the content of the knowledge base is optimized to make it more in line with the actual needs of users, and the practicality and accuracy of the knowledge base are improved;

[0101] Step S2062: Based on the identified defect characteristics, the structure of the agricultural interaction model is adjusted so that the model can better adapt to the needs of practical applications; the adjustment helps to improve the overall performance of the system and user experience;

[0102] Step S2063 enhances the information richness and coverage of the knowledge base by updating and supplementing the knowledge in the agricultural knowledge base, thereby improving the intelligence level and decision support capabilities of the system.

[0103] Embodiment 9:

[0104] This embodiment is improved on the basis of embodiment 1. The process of extracting the final prompt word in step S200 includes the following steps:

[0105] Step S207: the intelligent agent compares the final crop growth data results with the features in the agricultural data set to obtain differences (differences include slow crop growth cycle, insect pests or nutritional imbalance, etc.), generates corresponding prediction result prompt words according to the differences; parses the prediction result prompt words according to the historical field data prompt words to generate final prompt words; and inputs the prompt words into the agricultural development platform;

[0106] Step S208: Process and analyze the prompt words, determine the knowledge field of the prompt words, match them with the professional vocabulary in the determined prompt word field, find relevant information, mark the retrieved data and text, and filter them according to the marks; identify the acquired graph, compare it with the graph in the knowledge base, filter the corresponding information again, and extract the filtered information;

[0107] Step S209: retrieve the agricultural knowledge base according to the prompt word, and train the agricultural interaction model based on the retrieved agricultural knowledge base.

[0108] Among them, step S207 generates prediction result prompt words by comparing the final crop growth data results with the features in the agricultural data set, and parses the prediction result prompt words according to the historical field data prompt words to generate the final prompt words; it can help the intelligent agent identify differences such as slow crop growth cycle, insect pests or nutritional imbalance, thereby providing targeted prompt words for the agricultural development platform to optimize the crop growth environment and management strategy;

[0109] Step S208 processes and analyzes the prompt words, determines the knowledge field of the prompt words, matches them with professional vocabulary, finds relevant information, and marks and filters the data and text; this can improve the accuracy and efficiency of information processing, ensure that the extracted information is relevant to the prompt words, and provide a reliable data basis for subsequent decision support;

[0110] Step S209 retrieves the agricultural knowledge base according to the prompt words and trains the agricultural interaction model; it can utilize the rich information in the knowledge base, combined with the prediction results generated by the prompt words, to train a more accurate interaction model, thereby improving the intelligence level of agricultural production and decision-making support capabilities.

[0111] Embodiment 10:

[0112] This embodiment is improved on the basis of embodiment 9. In step S209, the agricultural knowledge base is trained with an agricultural interaction model. By constructing a knowledge network, logical docking nodes and forming an interaction model, the systematic management and efficient retrieval of agricultural knowledge are realized, the intelligent level and response speed of knowledge services are improved, and the accuracy and efficiency in knowledge matching, retrieval and reorganization are higher, and the technical effect and user experience of agricultural knowledge services are optimized. The specific process includes the following steps:

[0113] Step S2091: the intelligent agent obtains several answering agricultural knowledge bases corresponding to the question speech, constructs an agricultural knowledge construction network corresponding to each answering agricultural knowledge base, and analyzes the agricultural knowledge logic between different answering agricultural knowledge bases according to the agricultural knowledge construction network;

[0114] Step S2092: constructing a logical docking node for each pair of agricultural knowledge bases according to the agricultural knowledge logic, and logically arranging the agricultural logic bases according to the docking directions corresponding to the logical docking nodes to form an agricultural interaction model; sorting out the knowledge of the user's questions, and checking them with the key prompt words, and obtaining a number of relevant knowledge points according to the key prompt words; finding each relevant knowledge point in the agricultural interaction model, and building a knowledge structure corresponding to each relevant agricultural knowledge point according to the search results;

[0115] Step S2093: Establishing a dialogue format according to the purpose of the question, and matching each agricultural knowledge structure according to the purpose of the question, and reorganizing the matched knowledge according to the dialogue format to obtain a number of dialogue texts.

[0116] Among them, step S2091 can effectively organize and manage scattered agricultural knowledge by building an agricultural knowledge construction network, improve the accessibility and utilization efficiency of knowledge; help solve the problem of scattered and disordered agricultural knowledge, and improve the coverage and utilization efficiency of knowledge;

[0117] Step S2092 can achieve systematic management and efficient retrieval of agricultural knowledge by constructing logical docking nodes and forming an agricultural interaction model, which helps to improve the intelligent level of knowledge services;

[0118] Step S2093 can achieve accurate matching and reorganization of agricultural knowledge by establishing a dialogue format and performing knowledge matching, thereby improving the practicality and pertinence of knowledge services.

[0119] Embodiment 11:

[0120] This embodiment is improved on the basis of embodiment 1. The operation of executing the solution in step S300 has significant optimization effects in terms of automation, intelligence, remote control and real-time feedback, especially in terms of improving operation accuracy, reducing human errors, enhancing user operation experience and improving agricultural management efficiency. The specific process includes the following steps:

[0121] Step S301: Send a target instruction through the mobile device, the agricultural interaction model receives the target instruction, and parses the target instruction to determine the specific operation to be performed (specific operations include watering, fertilizing, and insect repelling operations); transmit the parsed target instruction to the intelligent agent;

[0122] Step S302: The agent identifies the operation to be performed and the target area to be performed according to the target instruction content, and sends a working signal to the corresponding device through wireless communication technology; after the agent completes the execution, the operation result (the operation result includes the amount of water for sprinkling, the amount of fertilizer applied, etc.) is fed back to the agricultural interaction model through wireless communication technology;

[0123] Step S303: The agricultural interaction model processes the operation result data and generates agricultural operation feedback information, and informs the user of whether the operation is successful and the operation effect through the mobile device.

[0124] Among them, step S301 realizes automation and intelligence of agricultural operations, reduces manual intervention, and improves the accuracy and efficiency of operations;

[0125] Step S302 implements remote control and real-time monitoring, ensuring the timeliness and accuracy of the operation;

[0126] In step S303, a user-friendly feedback interface is provided, so that the user can intuitively understand the operation result.

[0127] Embodiment 12:

[0128] This embodiment is improved on the basis of Embodiments 1-11. The agricultural planting decision-making system includes:

[0129] Data processing module 1 is responsible for acquiring historical field data and collecting field data through acquisition equipment, preprocessing the field data of the historical field data set to obtain the historical agricultural data set and agricultural data set; performing feature extraction, clustering and sensor data fusion operations on the historical agricultural data set and agricultural data set, and transmitting the operated historical agricultural data set and agricultural data set to intelligent agent module 2;

[0130] Agent module 2 is responsible for making predictions based on agricultural data sets and outputting the prediction results to the agricultural development platform. The agricultural development platform sends the prediction results and historical field data to the user respectively. The user makes a query and sends the query information to the agent through the agricultural development platform. The agent extracts prompt words from the query information, historical agricultural data and prediction data, and retrieves the agricultural knowledge base according to the prompt words, and outputs the agricultural knowledge base information to the agricultural interaction model module 3.

[0131] The agricultural interaction model module 3 is responsible for training the agricultural interaction model according to the information of the agricultural knowledge base; the user asks agricultural knowledge through voice recognition or mobile devices, analyzes the agricultural knowledge asked, and transmits the analysis results to the intelligent body. The intelligent body compares and analyzes the agricultural knowledge with the prompt words, retrieves the agricultural knowledge base, and forms planting suggestions; the planting suggestions are output to the agricultural interaction model, and the agricultural interaction model processes the planting suggestions and outputs them to the user; and the user operates the agricultural device through voice recognition or mobile devices; the agricultural interaction model analyzes the agricultural device operation signal sent by the user, and transmits the analysis results to the intelligent body; the intelligent body transmits the signal to the agricultural device through wireless sensor signals to control the operation of the agricultural device; after the work is completed, the feedback information is informed to the user whether the operation is successful and the operation effect.

[0132] Among them, the data processing module 1 obtains the environmental data of the farmland, crop growth status and other information in real time through the collection equipment; performs pre-processing operations such as cleaning and standardization on the acquired data to ensure data quality; extracts useful features from the pre-processed data, such as temperature, humidity, soil moisture, etc.; uses clustering algorithms to classify the data and identify the agricultural production patterns in different regions; and fuses the data from different sensors to improve the accuracy and completeness of the data;

[0133] Agent module 2 uses machine learning or deep learning models to predict agricultural data sets, such as crop yields, the probability of pests and diseases, etc. The prediction results and historical data are integrated and output to the agricultural development platform for users to view and analyze. Users can query the prediction results and historical data for a specific time period or region through the agricultural development platform. The agent extracts relevant prompt words based on the user's query information and calls the agricultural knowledge base for further analysis.

[0134] The agricultural interaction model module 3 trains the agricultural interaction model based on the information in the agricultural knowledge base, so that it can understand and answer the user's questions; the intelligent agent compares the user's questions with the prompt words, and calls the agricultural knowledge base for detailed analysis; based on the analysis results, the intelligent agent generates specific planting suggestions and outputs them to the agricultural interaction model; the agricultural interaction model processes the planting suggestions and finally outputs them to the user for reference and implementation; the user operates the agricultural equipment through voice recognition or mobile devices, and the intelligent agent receives and analyzes these operation signals to further optimize the planting suggestions.

[0135] In this embodiment, the data processing module 1 acquires and preprocesses data, the intelligent agent module 2 performs prediction and data analysis, and the agricultural interaction model module 3 provides interaction and suggestion generation. The modules work closely together to achieve full process automation and intelligence from data collection to final planting suggestions; this not only improves agricultural production efficiency, but also provides scientific decision-making support for farmers.

[0136] Fig.14 The structure of the electronic device is shown, including a central processing unit / microprocessor / main control chip etc. 4; a storage medium 5, coupled to the central processing unit / microprocessor / main control chip etc. 4, and storing computer executable instructions therein, for performing the steps of each method of the embodiment of the present invention when executed by the processor;

[0137] Wherein, the central processing unit / microprocessor / main control chip etc. 4 may include but is not limited to, for example, one or more processors or microprocessors etc.;

[0138] The storage medium 5 may include, but is not limited to, for example, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (such as hard disk, floppy disk, solid state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0139] In addition, the electronic device may also include (but not limited to) a data bus 6, an input / output bus / external bus / device bus 7, a display 8, and input / output devices (9) (for example, a keyboard, a mouse, a speaker, etc.).

[0140] The central processing unit / microprocessor / main control chip etc. 4 can communicate with external devices ( 8 , 9 etc.) through the I / O bus 7 via a wired or wireless network (not shown).

[0141] The storage medium 5 may also store at least one computer executable instruction for executing the various functions and / or method steps in the embodiments described in the present technology when executed by the central processing unit / microprocessor / main control chip 4.

[0142] In one embodiment, the at least one computer executable instruction may also be compiled into or constitute a software product, wherein one or more computer executable instructions are executed by a processor to perform the various functions and / or method steps in the embodiments described in the present technology.

[0143] Fig.15A computer-readable storage medium is shown, and instructions are stored on a non-transitory computer-readable storage medium 11, and the instructions are, for example, computer-readable instructions 10. When the computer-readable instructions 10 are executed by the processor, the various methods described above can be executed. The non-transitory computer-readable storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. Non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 11 can be connected to a computing device such as a computer, and then, when the computing device runs the computer-readable instructions 10 stored on the computer-readable storage medium 11, the various methods described above can be performed.

Claims

1. An agricultural planting decision-making method, characterized in that: The process is: First, historical field data is obtained, field data is collected, and preprocessed to obtain historical agricultural data sets and agricultural data sets; Then, feature extraction, clustering and sensor data fusion are performed and transmitted to the intelligent agent and agricultural development platform; Finally, the agent makes predictions and sends the prediction results and historical field data to the user respectively; Based on the implementation of the agricultural planting decision-making system, the main structure of the agricultural planting decision-making system includes a data processing module, an intelligent agent module and an agricultural interaction model module.

2. The agricultural planting decision-making method according to claim 1, characterized in that: The prediction results and the historical field data are analyzed separately, and the prediction result prompt words and the historical field data prompt words are obtained respectively. The prediction result prompt words are analyzed with reference to the historical field data prompt words to obtain the final prompt words; Output the final prompt words to the agricultural interaction model; The agricultural interaction model retrieves the agricultural knowledge base according to the final prompt word, trains the model according to the agricultural knowledge base, obtains planting suggestions, outputs the planting suggestions in the form of human-computer dialogue or on the corresponding device, and performs corresponding operations.

3. The agricultural planting decision-making method according to claim 1 or 2, characterized in that: The specific process is: First, historical field data are obtained and field data are collected through acquisition equipment, and the historical field data and field data are preprocessed to obtain historical agricultural data sets and agricultural data sets; Perform operations such as feature extraction, clustering, and sensor data fusion on the historical agricultural data set and the agricultural data set, and transmit the operated historical agricultural data set and the agricultural data set to the intelligent agent; Transmit historical field data to agricultural development platforms; Among them, field data include environmental sensing, crop phenotyping, and weather forecast data; Then, the agent makes predictions based on the agricultural dataset and outputs the prediction results to the agricultural development platform; The agricultural development platform will send the prediction results and historical field data to users, who will query the solutions and execute the solutions; Finally, the agricultural interaction model outputs the planting suggestions in the form of human-computer dialogue or on the corresponding equipment, and performs corresponding operations, including controlling temperature and humidity, spraying water, and applying fertilizers.

4. The agricultural planting decision-making method according to claim 3, characterized in that: The process of feature extraction, clustering and sensor data fusion includes the following steps: Step S101: preprocessing the historical field data and the field data, removing noise and outliers in the historical field data and the field data, converting the historical field data and the field data into a data standardized form, and obtaining a historical agricultural data set and an agricultural data set; transmitting the historical agricultural data set and the agricultural data set to the intelligent agent; Step S102: Integrate and cluster the historical agricultural data set and the agricultural data set according to the time series respectively; perform weighted fusion of the integrated cluster data on each sensor to form a comprehensive field data set and a historical comprehensive data set, and extract features from the comprehensive field data set and the historical comprehensive field data set; Step S103: the intelligent agent analyzes the characteristic data of the comprehensive field data set and the historical comprehensive field data set to identify the crop growth stages or regions under different environmental conditions; compares the characteristic data of the comprehensive field data set and the historical comprehensive field data set, and performs prediction processing on the field data; The specific process of integrating cluster data to perform weighted fusion on each sensor includes the following steps: Step S1021: The dynamic time planning distance method is used to calculate the distance of the sensors at each time point within a period of time. By calculating the distance of each sensor data, the median absolute deviation method is used to set the correlation threshold. The low correlation sensor data exceeding the threshold will not participate in this fusion. Step S1022: Calculate the correlation between the sensors based on the calculated dynamic time planning distance, and assign weights according to the strength of the correlation; the higher the correlation between the terminal perception data, the greater the assigned weight; perform weighted data fusion on the currently collected perception data according to the size of the weight; Step S1023: abstract the sensor data, obtain the associated characteristic values ​​between the data through analysis, abstract the sensor data, extract the key characteristic information of each type of perception data, and perform feature-level data fusion by utilizing the association of these characteristic information.

5. The agricultural planting decision-making method according to claim 3, characterized in that ,The specific process of the agent making predictions based on the agricultural dataset includes the following steps: Step S201: Initialize the weight and threshold of each neuron in the agent; Step S202: Input the features in the agricultural data set into the intelligent agent and calculate the weighted sum of each neuron; each neuron corresponds to each feature, and the neuron will simulate the crop growth process and obtain the final crop growth data results; Step S203: Compare the final crop growth data results with the features in the historical agricultural data set to obtain differences and analyze the differences, wherein the differences include slow crop growth cycle, insect pests or nutritional imbalance.

6. The agricultural planting decision-making method according to claim 3, characterized in that: The specific process of user query solution includes the following steps: Step S204: Preliminary recognition of the question voice issued by the user is performed to determine the purpose of the user's question, and the intelligent agent extracts the final prompt word and retrieves the relevant answer agricultural knowledge base according to the prompt word; Step S205: The agricultural interaction model is trained according to the agricultural knowledge base of the dialogue to obtain a number of dialogue texts; the intelligent agent obtains the dialogue information, performs agricultural prediction on the dialogue information, compares the prediction result with the historical data result, and generates information if it is less than the threshold value and outputs it to the user; if the prediction result is greater than the threshold value, information improvement feedback and agricultural knowledge base update feedback are generated; Step S206: adjusting the agricultural interaction model construction method according to the information improvement feedback and the agricultural knowledge base update feedback, and adjusting the agricultural interaction model; The process of answering the knowledge base includes the following steps: Step S2041: obtaining a question voice issued by the user, establishing a voice rhythm of the question voice according to the rhythmic features of the question voice, dividing the question voice into a plurality of rhythm segments according to the voice rhythm, and establishing a recognition frequency of the corresponding rhythm segment according to the rhythm strength corresponding to each rhythm segment; Step S2042: Based on the recognition rate, the corresponding rhythm segment is subjected to language recognition using a preset speech recognition technology to obtain semantic information corresponding to the rhythm segment, and the semantic information is reorganized to obtain a user question text; Step S2043: semantically analyze the question text to obtain several question purposes of the user, and compare and analyze the question purpose with the final prompt word to obtain the corresponding agricultural knowledge index; search the agricultural knowledge base for the answer whose correlation with the corresponding question purpose is greater than a preset correlation threshold according to the agricultural knowledge index.

7. The agricultural planting decision-making method according to claim 3, characterized in that: The specific process of improving feedback based on information and updating feedback based on the agricultural knowledge base includes the following steps: Step S2051: obtaining the dialogue information between the user and the agricultural interaction model, performing pre-question-answer separation on the pre-answer information, and obtaining the question information corresponding to the user and the pre-answer information of the agricultural interaction model; Step S2052: predicting crop growth based on the question information, comparing the prediction results with historical data, and setting thresholds based on the historical data; Step S2053: Output the comparison result to the agricultural interaction model; if it is less than the threshold, generate planting suggestions and output them to the user; If it is greater than the threshold, improvement feedback and agricultural knowledge base update feedback are generated; The agricultural interaction model is constructed based on feedback from information improvement and agricultural knowledge base update.

8. The agricultural planting decision-making method according to claim 3, characterized in that: The process of adjusting the agricultural interaction model includes the following steps: Step S2061: Determine the defect features of the agricultural interaction model according to the information improvement feedback, and determine the agricultural knowledge to be supplemented features of the corresponding agricultural knowledge base according to the agricultural knowledge and the new feedback; Step S2062: establishing defect positions of the agricultural interaction model according to defect characteristics of the agricultural interaction model, and adjusting the model structure of the agricultural interaction model according to the defect positions; Step S2063: retrieve the agricultural knowledge to be supplemented in the agricultural knowledge base update feedback, determine the agricultural knowledge association method between the agricultural knowledge to be supplemented and the existing agricultural knowledge based on the agricultural knowledge to be supplemented, and use the agricultural knowledge to supplement the corresponding agricultural knowledge base.

9. The agricultural planting decision-making method according to claim 3, characterized in that: The process of extracting the final prompt word includes the following steps: Step S207: the intelligent agent compares the final crop growth data results with the features in the agricultural data set to obtain differences, and generates corresponding prediction result prompt words according to the differences; Parse the prediction result prompt words according to the historical field data prompt words to generate the final prompt words; And input the prompt words into the agricultural development platform; Step S208: Process and analyze the prompt words, determine the knowledge field of the prompt words, match them with the professional vocabulary in the determined prompt word field, find relevant information, mark the retrieved data and text, and filter them according to the marks; identify the acquired graph, compare it with the graph in the knowledge base, filter the corresponding information again, and extract the filtered information; Step S209: Retrieving an agricultural knowledge base according to the prompt word, and training an agricultural interaction model according to the retrieved agricultural knowledge base; The specific process of training the agricultural interactive model for the agricultural knowledge base includes the following steps: Step S2091: the intelligent agent obtains several answering agricultural knowledge bases corresponding to the question speech, constructs an agricultural knowledge construction network corresponding to each answering agricultural knowledge base, and analyzes the agricultural knowledge logic between different answering agricultural knowledge bases according to the agricultural knowledge construction network; Step S2092: constructing a logical docking node for each pair of agricultural knowledge bases according to the agricultural knowledge logic, and logically arranging the agricultural logic bases according to the docking directions corresponding to the logical docking nodes to form an agricultural interaction model; sorting out the knowledge of the user's questions, and checking them with the key prompt words, and obtaining a number of relevant knowledge points according to the key prompt words; finding each relevant knowledge point in the agricultural interaction model, and building a knowledge structure corresponding to each relevant agricultural knowledge point according to the search results; Step S2093: Establishing a dialogue format according to the purpose of the question, and matching each agricultural knowledge structure according to the purpose of the question, and reorganizing the matched knowledge according to the dialogue format to obtain a number of dialogue texts.

10. The agricultural planting decision-making method according to claim 3, characterized in that: The operational process of implementing the solution includes the following steps: Step S301: a target instruction is issued through a mobile device, and the agricultural interaction model receives the target instruction and parses the target instruction to determine the specific operation to be performed; the parsed target instruction is transmitted to the intelligent agent; Step S302: The agent identifies the operation to be performed and the target area to be performed according to the target instruction content, and sends a working signal to the corresponding device through wireless communication technology; after the agent completes the execution, the operation result is fed back to the agricultural interaction model through wireless communication technology; Step S303: The agricultural interaction model processes the operation result data and generates agricultural operation feedback information, and informs the user of whether the operation is successful and the operation effect through the mobile device.

Images