Intelligent agricultural system based on modular collaboration and AI Agent driving

Abstract

The invention relates to an intelligent agricultural system based on modular collaboration and AI Agent driving. Environmental sensing, video monitoring and equipment operation data are comprehensively acquired through a data acquisition module; the data processing module carries out real-time abnormity diagnosis and mode analysis; the core framework module cooperatively schedules the data storage module, the user management module and the external interface module to realize cross-layer resource and data integration; the business logic module is connected in series with each layer of events to dynamically generate backlogs and drive multi-domain function execution; and the AI service module carries a multi-modal agricultural agent Agent, has the capabilities of natural language understanding, multi-round dialogue memory, agricultural knowledge reasoning, equipment control decision making, abnormity pre-judgment and personalized suggestion generation core AI, coordinates operation of each module, and realizes intelligent jump from passive response to active suggestion. According to the invention, the problems of response lag, operation splitting and data analysis fragmentation of an existing agricultural system are solved, and the real-time performance, the collaboration and the decision accuracy of agricultural management are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart agriculture, in particular to a smart agriculture comprehensive management system based on modular collaboration and AI Agent driving, and more particularly to an agricultural data intelligent perception, closed-loop decision-making and remote control integrated platform. BACKGROUND

[0002] With the transformation of modern agriculture to digitalization and intelligentization, smart agriculture management systems have gradually become the core tool for improving production efficiency. Existing systems usually focus on a single function, such as environmental monitoring or device control, and realize data acquisition, video monitoring or farm management through independent modules to provide basic data support for agricultural production.

[0003] However, such systems have the following significant defects: 1. Data fragmentation: environmental sensing, device control, security monitoring and other modules are isolated from each other, unable to form a data closed-loop collaboration, resulting in delayed abnormal response. For example, when the soil moisture exceeds the standard, the system only records the data and cannot link to the irrigation device, requiring manual intervention; 2. Decision fragmentation: early warning information is scattered in different subsystems: such as environmental alerts, device failures, inventory warnings, lack of a unified to-do hub, resulting in missed or repeated disposal tasks; 3. Operation complexity: remote control relies on professional interface operation and fails to integrate natural language interaction, and poor device compatibility leads to high deployment costs; 4. Lack of intelligent human-computer interaction: existing systems mostly use traditional menu-style operation interfaces, lack of intelligent dialogue assistants, and users need to have professional knowledge to effectively use the system. When facing complex farm decision-making, such as comprehensive consideration of weather, soil, crop growth period irrigation strategy, the system cannot provide intelligent decision support and operation guidance, and cannot achieve convenient operation through natural language interaction, which seriously restricts the popularization and application of smart agriculture technology in ordinary farmers.

[0004] Therefore, it is imperative to develop and deploy a modular collaborative smart agriculture system that integrates AI Agent technology and multi-modal data. SUMMARY

[0005] The present application aims to provide a smart agriculture system based on modular collaboration and AI Agent driving, aiming to solve the problems of data fragmentation, decision fragmentation and operation complexity of existing smart agriculture systems.

[0006] To achieve the above objectives, this invention provides a smart agriculture system based on modular collaboration and AI Agent-driven architecture, comprising a core framework module, a user management module, an external interface module, a business logic module, a data acquisition module, a data processing module, a data storage module, and an AI service module. The core framework module is connected to the user management module, the external interface module, the business logic module, and the AI ​​service module, respectively. The data acquisition module, the data processing module, and the data storage module are interconnected to provide data support for the core framework module and the AI ​​service module. The AI ​​service module is connected to the business logic module. The core framework module is used to implement user interaction, handle HTTP request and response and route management, and implement database access and data caching; The user management module is used for user authentication, user permission management, and user information presentation and management. The external interface module is used to connect to weather API, agricultural technology information API, product information, message push and payment interface access; The business logic module is used to present and manage to-do items, realize remote control and intelligent command issuance, manage warehousing and transportation records, and verify blockchain product traceability. The data acquisition module is used to collect and record environmental sensing time-series data, video surveillance data, and equipment operation data. The data processing module is used to perform anomaly detection and trend analysis on various environmental sensor time-series data collected by the data acquisition module, perform image recognition and pest monitoring on the acquired video surveillance, and asynchronously process and distribute equipment operation data to the message queue. The data storage module is used to store user information, device status, operation records, search record relational data, time-series data transmitted by sensors, and other image and video data; The AI ​​service module integrates a multimodal agricultural intelligent assistant, including a multimodal interaction unit, an intelligent dialogue management unit, an agricultural knowledge reasoning unit, a decision suggestion generation unit, and a personalized learning unit. These units are interconnected and work together to form a complete agricultural intelligent assistant. The multimodal interaction unit is used to process multiple input methods such as voice, text, and images, and supports voice-to-text, text-to-speech, and image recognition to achieve natural and convenient human-computer interaction. The intelligent dialogue management unit is configured to maintain a multi-round dialogue context, support intent recognition, slot filling, dialogue state tracking, and be capable of processing complex agricultural consultation and instructions, and support topic switching and follow-up clarification; the intelligent dialogue management unit adopts a dialogue strategy combining a state machine and deep reinforcement learning, and is capable of processing complex scenes such as multi-intent recognition, slot filling, and dialogue repair.

[0007] The agricultural knowledge reasoning unit is configured to perform logical reasoning based on an embedded agricultural expert knowledge graph, and comprehensively analyze environment sensing time series data obtained by the data acquisition module to output scientific agricultural suggestions; the environment sensing time series data includes current environment data, crop growth stages, and historical data; the agricultural knowledge reasoning unit constructs a knowledge graph covering crop planting, pest control, fertilization and irrigation, and environmental control, and contains more than 100,000 agricultural rules and experiences collected from websites, and supports rule-based logical reasoning and case-based analogical reasoning.

[0008] The decision suggestion generation unit is configured to actively analyze farm data trends according to data processed by the data processing module, generate personalized early warning reminders and optimization suggestions, and support risk prediction, resource allocation optimization, and agricultural plan formulation; the decision suggestion generation unit adopts a multi-objective optimization algorithm to seek a balance among multiple objectives such as crop yield, cost reduction, and environmental friendliness, and generates an optimal agricultural scheme.

[0009] The personalized learning unit is configured to record user operation habits, preference settings, and farm characteristics, and continuously optimize suggestion strategies through machine learning algorithms to realize adaptive evolution of the intelligent assistant; the personalized learning unit realizes continuous optimization and personalized adaptation of suggestion strategies through mechanisms such as user portrait construction, preference learning, and effect feedback. The entire module realizes coordination and operation control of other modules through the core framework module, and assumes the role of a "housekeeper".

[0010] Further, the report generation module is configured to statistically analyze data and its analysis situation within a certain time range, realize multi-dimensional data aggregation and safety factor scoring, and export a report.

[0011] Further, the report generation module includes a data statistical unit and a report exporting unit connected in sequence, The data statistical unit is configured to aggregate environment indicators and multi-dimensional device state data across databases, generate daily trend comparison charts, and support custom report dimension configuration. The report exporting unit is configured to generate a required format document from statistical results according to a template, automatically add a digital signature, and provide a voice reading function.

[0012] Furthermore, the core framework module includes a Web application framework unit, a database access unit, and a caching unit connected in sequence. The Web application framework unit is used to process HTTP request and response in real time, and to implement API routing and load balancing. The database access unit is used to implement object-relational mapping, automatically execute SQL conversion and transaction management, maintain a dynamic database connection pool, and support read-write separation and failover mechanisms; The caching unit is used to implement high-frequency data caching, provide distributed session management capabilities, implement cache eviction policies to ensure data consistency, and reduce database access pressure.

[0013] Furthermore, the user management module includes a user authentication unit, a permission management unit, and a user information management and processing unit connected in sequence. The user authentication unit is used to implement stateless identity authentication, support automatic token refresh and expiration mechanisms, encrypt user passwords, and prevent the risk of credential leakage. The permission management unit is used to control the granularity of role permissions, realize dynamic access authentication of API interfaces, support permission group configuration and real-time effect, and ensure the security boundary of system operation. The user information management unit is used to support users in viewing and managing their personal information.

[0014] Furthermore, the external interface module includes a third-party API integration unit, a payment interface unit, and a message push unit connected in sequence; The third-party API integration unit is used to connect to the meteorological bureau's real-time weather API to provide weather conditions and disaster risk forecasts, aggregate agricultural policy information interfaces, and access voice reading interfaces. The payment interface unit is used to integrate the payment SDK to process commodity transactions and implement dual amount verification to prevent tampering; The message push unit is used to push device alarm notifications in real time and send SMS / email reminders simultaneously. It supports customized subscription rules and do-not-disturb settings.

[0015] Furthermore, the business logic module includes a to-do list management unit, an intelligent control unit, a warehouse management unit, and a blockchain traceability unit connected in sequence. The to-do list management unit is used to aggregate alarm events from various modules, automatically sort them by urgency, track task handling status, and record processing logs. The intelligent control unit is used to translate user operation commands into device control protocols, receive execution feedback in real time, and support control script presets and batch operations. The warehouse management unit generates shelf-life warnings by monitoring storage and transportation environment parameters, associating inventory status with inventory status, and recording equipment operation and maintenance history to form a full life cycle file. The blockchain traceability unit encrypts and uploads agricultural product production and processing data to the blockchain, generates an unalterable traceability QR code, and provides a distributed ledger verification interface for consumers to check authenticity.

[0016] Furthermore, the data acquisition module includes a sensor data receiving unit, a video stream processing unit, and a device communication unit connected in sequence; The sensor data receiving unit is used to collect real-time data streams of temperature and humidity, soil moisture and spore concentration, and filter abnormal fluctuation data of the equipment. The video stream processing unit is used to connect to camera streaming media, complete video encoding and decoding, extract key frames and buffer streaming data, and provide preprocessing support for intelligent analysis. The device communication unit is used to dynamically monitor the online status and operating parameters of the device and implement a command retransmission mechanism to ensure control reliability.

[0017] Furthermore, the data processing module includes an environmental data analysis unit, a video intelligent analysis unit, and a message queue unit connected in sequence. The environmental data analysis unit is used to detect abnormal sensor data and generate dynamic threshold warnings, and to predict environmental change trends by combining time series analysis. The video intelligent analysis unit is used to identify crop diseases and pests and output visual analysis tags and confidence reports; The message queue unit is used to asynchronously process high-concurrency tasks and accurately distribute alarm information.

[0018] Furthermore, the data storage module includes interconnected relational database units, time-series database units, and file storage units. The relational database unit is used to structure and store user account and device metadata and to establish a data dictionary to ensure the consistency of business fields; The time-series database unit is used to write high-frequency monitoring data from sensors and provide time-range aggregated queries; The file storage unit is used for distributed storage of various types of files, including pictures, video recordings, and reports transmitted by users. Beneficial effects of this invention: This invention comprehensively acquires environmental sensing, video monitoring, and equipment operation data through the data acquisition module; the data processing module performs real-time anomaly diagnosis and pattern analysis; the core framework module coordinates and schedules the data storage module, the user management module, and the external interface module to achieve cross-layer resource and data integration; the business logic module connects events at each layer to dynamically generate to-do items and drive the execution of multi-domain functions; the AI ​​service module is equipped with a multimodal agricultural intelligent agent, possessing core AI capabilities such as natural language understanding, multi-turn dialogue memory, agricultural knowledge reasoning, equipment control decision-making, anomaly prediction, and personalized suggestion generation. It can provide users with 24 / 7 smart agriculture assistant services through various interactive methods such as voice, text, and images, and also acts as a "housekeeper" to coordinate the operation of various modules, achieving an intelligent leap from passive response to proactive suggestion. Overall, this invention solves the problems of delayed response, fragmented operation, and fragmented data analysis in existing agricultural systems through a "collection-analysis-decision-response" closed loop, significantly improving the real-time performance, collaboration, and decision-making accuracy of agricultural management.

[0019] The compatibility and adaptability capabilities are deeply integrated into the system architecture: the data processing module automatically identifies the communication protocols of seven types of agricultural equipment, the intelligent control unit translates natural language commands into device-executable protocols, and the AI ​​service module matches differentiated agricultural technology knowledge based on user profiles, achieving seamless compatibility from traditional agricultural machinery to intelligent terminals, and significantly reducing the cost and learning threshold of upgrading to smart agriculture. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a block diagram of a specific embodiment 1 of the present invention; Figure 2 This is a block diagram of the core framework module of a specific embodiment 1 of the present invention; Figure 3 This is a block diagram of the user management module in a specific embodiment 1 of the present invention; Figure 4 This is a block diagram of the external interface module in a specific embodiment 1 of the present invention; Figure 5 This is a block diagram of the business logic module in a specific embodiment 1 of the present invention; Figure 6 This is a block diagram of the data acquisition module in a specific embodiment 1 of the present invention; Figure 7 This is a block diagram of the data processing module in a specific embodiment 1 of the present invention; Figure 8 This is a block diagram of the data storage module in a specific embodiment 1 of the present invention; Figure 9 This is a block diagram of the report generation module in a specific embodiment 1 of the present invention; Figure 10 This is a block diagram of the AI ​​service module in a specific embodiment of the present invention.

[0022] In the diagram: 1-Core Framework Module, 2-User Management Module, 3-External Interface Module, 4-Business Logic Module, 5-Data Acquisition Module, 6-Data Processing Module, 7-Data Storage Module, 8-Report Generation Module, 9-AI Service Module, 11-Web Application Framework Unit, 12-Database Access Unit, 13-Cache Unit, 21-User Authentication Unit, 22-Permission Management Unit, 23-User Information Management Unit, 31-Third-Party API Integration Unit, 32-Payment Interface Unit, 33-Message Push Unit, 41-To-Do Item Management Unit, 42-Smart Control Sheet Units include: 43-Warehouse Management Unit, 44-Blockchain Traceability Unit, 51-Sensor Data Receiving Unit, 52-Video Stream Processing Unit, 53-Equipment Communication Unit, 61-Environmental Data Analysis Unit, 62-Video Intelligent Analysis Unit, 63-Message Queue Unit, 71-Relational Database Unit, 72-Time Series Database Unit, 73-File Storage Unit, 81-Data Statistics Unit, 82-Report Export Unit, 91-Multimodal Interaction Unit, 92-Intelligent Dialogue Management Unit, 93-Agricultural Knowledge Reasoning Unit, 94-Decision Suggestion Generation Unit, and 95-Personalized Learning Unit. Detailed Implementation

[0023] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention. Specific Implementation Example 1

[0024] like Figures 1 to 10As shown, this invention provides a smart agriculture system based on modular collaboration and AI Agent-driven architecture, comprising a core framework module 1, a user management module 2, an external interface module 3, a business logic module 4, a data acquisition module 5, a data processing module 6, a data storage module 7, a report generation module 8, and an AI service module 9. The core framework module 1 serves as the center and is connected to the user management module 2, the external interface module 3, the business logic module 4, and the AI ​​service module 9 respectively. The data acquisition module 5, the data processing module 6, and the data transmission module 7 are connected in pairs and as a whole to the core framework module 1, the report generation module 8, and the AI ​​service module 9. The AI ​​service module 9 is connected to the business logic module 4. The core framework module 1 is used to implement user interaction, handle HTTP request and response processing and route management, and implement database access and data caching; specifically, it includes: The core framework module 1 includes a Web application framework unit 11, a database access unit 12, and a cache unit 13, which are connected in sequence. The Web application framework unit 11 is used to process HTTP request and response in real time, and to implement API routing and load balancing. The database access unit 12 is used to implement object-relational mapping, automatically execute SQL conversion and transaction management, maintain a dynamic database connection pool, and support read-write separation and failover mechanisms. The cache unit 13 is used to implement high-frequency data caching, provide distributed session management capabilities, implement cache eviction policies to ensure data consistency and reduce database access pressure.

[0025] In this embodiment, the Web application framework unit 11 processes HTTP request responses in real time with default configuration, including API routing and load balancing, and converts the routed data access requirements into database operation instructions. The database access unit 12 then performs object-relational mapping, automated SQL conversion and transaction management, maintains a dynamic database connection pool, and supports read-write separation and failover mechanisms to obtain optimized database result data. Finally, the caching unit 13 implements caching strategies for high-frequency database result data, including providing distributed session management capabilities and implementing cache eviction policies to ensure data consistency and reduce database access pressure.

[0026] The core framework module's database access unit uses a dynamic connection pool management mechanism to achieve read / write separation and failover. The caching unit ensures data consistency through distributed session management and eviction policies. The web application framework unit uses load balancing technology to handle high-concurrency access pressure. This triple linkage ensures that the system maintains millisecond-level response stability in scenarios with 2000+ connected devices.

[0027] The user management module 2 is used for user authentication, user permission management, and user information presentation and management; specifically, it includes: The user management module 2 includes a user authentication unit 21, a permission management unit 22, and a user information management and processing unit 23, which are connected in sequence. The user authentication unit 21 is used to implement stateless identity authentication, support automatic token refresh and expiration mechanism, encrypt user password, and prevent the risk of credential leakage. The permission management unit 22 is used to control the granularity of role permissions, realize dynamic access authentication of API interfaces, support permission group configuration and real-time effect, and ensure the security boundary of system operation. The user information management unit 23 is used to support users in viewing and managing their personal information.

[0028] In this embodiment, the user authentication unit 21 verifies user credentials in real time through a stateless identity authentication mechanism, including encrypting user passwords, supporting automatic token refresh and expiration mechanisms, generating permission verification tokens, and preventing the risk of credential leakage.

[0029] The permission management unit 22 then performs dynamic authentication based on the token, including controlling the granularity of role permissions, implementing API interface access authentication, configuring permission groups and making them effective in real time, and outputting the scope of authorized operations to ensure system security boundaries. Finally, after the permission verification is successful, the user information management unit 23 supports users in performing secure viewing and management operations on their personal information.

[0030] The external interface module 3 is used to connect to interfaces such as weather API, agricultural technology information API, product information, message push, and payment interfaces; specifically, it includes: The external interface module 3 includes a third-party API integration unit 31, a payment interface unit 32, and a message push unit 33, which are connected in sequence. The third-party API integration unit 31 connects to the meteorological bureau's real-time weather API to provide weather conditions display and disaster risk prediction, aggregates agricultural policy information interfaces, and provides voice reading interfaces. The payment interface unit 32 is used to integrate the WeChat / Alipay payment SDK to process commodity transactions and implement dual amount verification to prevent tampering. The message push unit 33 is used to push device alarm notifications in real time and simultaneously send SMS / email reminders. It supports customized subscription rules and do-not-disturb settings.

[0031] The external interface module 3 is also provided with an expansion interface for connecting to other extended information; In this embodiment, the third-party API integration unit 31 aggregates real-time weather data from the meteorological bureau and agricultural policy information, including disaster risk prediction and voice reading interface access, to generate a comprehensive decision-making information flow. The payment interface unit 32 then, based on transaction requests triggered by this information flow, such as agricultural input purchases after a disaster warning, integrates the WeChat / Alipay payment SDK to perform dual amount verification and anti-tampering transaction processing, outputting a payment voucher with anti-tampering markings. Finally, the message push unit 33, based on the payment voucher and device alarm status, triggers multi-channel push notifications in real time, including SMS / email reminders and device alarm notifications, and supports customized subscription rules and do-not-disturb settings for precise targeting.

[0032] The network selection unit of the external interface module automatically switches between LAN and 4G transmission channels according to business priority. The payment interface unit ensures transaction security through dual amount verification and anti-tampering mechanisms. The message push unit achieves accurate delivery of alarm information based on subscription rules, ensuring reliable end-to-end communication from data collection to the business closed loop.

[0033] The business logic module 4 is used to present and manage pending items, realize remote control and intelligent command issuance, manage warehousing and transportation records, and verify blockchain product traceability; specifically, it includes: The business logic module 4 includes a to-do item management unit 41, an intelligent control unit 42, a warehouse management unit 43, and a blockchain traceability unit 44, which are connected in sequence. The to-do list management unit 41 is used to aggregate alarm events from various modules, such as: environmental exceedance / equipment failure, automatic hierarchical sorting by urgency, tracking task handling status and recording processing logs; The intelligent control unit 42 is used to translate user operation commands, such as virtual joystick actions, into device control protocols, receive execution feedback in real time, and support control script presets and batch operations. The warehouse management unit 43 monitors storage and transportation environment parameters, such as temperature, humidity / gas concentration, generates shelf-life warnings based on the associated inventory status, and records equipment operation and maintenance history to form a full life cycle archive. The blockchain traceability unit 44 encrypts and uploads agricultural product production and processing data to the blockchain, generates an unalterable traceability QR code, and provides a distributed ledger verification interface for consumers to check authenticity.

[0034] In this embodiment, the to-do list management unit 41 aggregates multi-source alarm events such as environmental exceedances and equipment malfunctions in real time, sorts them by urgency, tracks their handling status, and records logs to generate a list of instructions to be executed, including equipment control requirements and storage and transportation monitoring tasks. The intelligent control unit 42 then translates user operation instructions in this list, such as virtual joystick actions or preset scripts, into equipment control protocols in real time, receives execution feedback, and outputs closed-loop equipment operation data. The warehouse management unit 43 dynamically correlates inventory status with operation data and external inputs, such as storage and transportation parameters like temperature, humidity, and gas concentration, to generate shelf-life warnings, while simultaneously recording equipment maintenance history to form full lifecycle archived data. Finally, the blockchain traceability unit 44 encrypts and uploads key production and processing information from the archived data to the blockchain, generating an unalterable traceability QR code, and provides authenticity verification services to consumers through a distributed ledger verification interface.

[0035] By encrypting and recording key operations such as pesticide application records and environmental interventions on the blockchain traceability unit, a traceability QR code with tamper-proof attributes is generated. The authenticity can be verified to consumers through a distributed ledger verification interface, eliminating the risk of falsified agricultural production data from the technical level and building a trustworthy agricultural value chain.

[0036] The data acquisition module 5 is used to acquire and record time-series data input from sensors and video stream data presented by cameras, and to perform IoT device protocol adaptation and status monitoring; specifically, it includes: It includes a sensor data receiving unit 51, a video stream processing unit 52, and a device communication unit 53, which are connected in sequence. The sensor data receiving unit 51 is used to collect real-time data streams of temperature and humidity, soil moisture and spore concentration and filter abnormal fluctuation data of the equipment. The video stream processing unit 52 is used to connect to the camera streaming media, complete video encoding and decoding, extract key frames and buffer streaming data, and provide preprocessing support for intelligent analysis. The device communication unit 53 is used to dynamically monitor the online status and operating parameters of the device. A command retransmission mechanism is implemented to ensure control reliability.

[0037] In this embodiment, the sensor data receiving unit 51 collects real-time heterogeneous data streams of temperature, humidity, soil moisture, and spore concentration, such as abnormal fluctuation data from active filtration equipment, and outputs a purification sensor dataset. The video stream processing unit 52 then, based on the monitoring point requirements associated with this dataset (e.g., triggering video verification due to abnormal spore concentration), connects to the camera's streaming media to complete video encoding / decoding and streaming data buffering, extracting keyframes to generate a pre-processed analysis data packet. Finally, the device communication unit 53 monitors the online status and operating parameters of the target device based on the keyframe analysis results and dynamic changes in sensor data, and implements remote control through a command retransmission mechanism to ensure reliability.

[0038] By collecting key agricultural parameters such as temperature, humidity, soil moisture, and spore concentration in real time through a heterogeneous sensor cluster, the sensor data receiving unit dynamically filters abnormal fluctuations in equipment data, the video stream processing unit performs millisecond-level frame extraction and buffering for the target area, and the equipment communication unit implements a command retransmission mechanism to ensure control reliability. This ensures accurate capture of high-frequency environmental data and holographic monitoring of equipment status, providing a high-confidence basis for intelligent decision-making.

[0039] The data processing module 6 is used to perform time-series data anomaly detection and trend analysis on various types of collected environmental information, and to perform frame-by-frame image recognition and pest monitoring on the collected video stream information and asynchronously distribute it to the message queue; specifically, it includes: The data processing module 6 includes an environmental data analysis unit 61, a video intelligent analysis unit 62, and a message queue unit 63, which are connected in sequence. The environmental data analysis unit 61 is used to detect abnormal sensor data and generate dynamic threshold warnings, and to combine time series analysis to predict environmental change trends. The video intelligent analysis unit 62 is used to identify crop diseases and pests and output visual analysis tags and confidence reports. The message queue unit 63 is used to asynchronously process high-concurrency tasks and accurately distribute alarm information.

[0040] In this embodiment, the environmental data analysis unit 61 detects abnormal sensor data in real time, generates dynamic threshold warnings, and combines time series analysis to predict environmental change trends, outputting an environmental risk assessment report including real-time warnings and trend predictions. The video intelligent analysis unit 62 then uses the coordinates of abnormal areas associated with this report, such as crop blocks experiencing sudden temperature and humidity changes, to call video streams for crop pest and disease identification and equipment intrusion detection, automatically triggering tiered alarms and generating visual analysis conclusions with a confidence report. Finally, the message queue unit 63 asynchronously schedules tiered alarms and visual conclusions, accurately distributing them to high-concurrency processing nodes to ensure the real-time availability of alarm information.

[0041] The environmental data analysis unit identifies data anomalies based on a dynamic threshold early warning mechanism and predicts environmental degradation trends by combining time series algorithms. The video intelligent analysis unit outputs disease identification confidence reports through visual models, and the message queue unit implements asynchronous distribution and scheduling of alarm information. This enables a leap in data quality from raw collection to intelligent analysis, establishing a multimodal decision-making foundation for precision agricultural intervention.

[0042] The data storage module 7 is used to store user information, device status, operation records, search record relational data, and time-series data transmitted by sensors, as well as other image and video data; specifically including: The data storage module 7 includes a relational database unit 71, a time-series database unit 72, and a file storage unit 73, which are connected in sequence. The relational database unit 71 is used to structure and store user account and device metadata and to establish a data dictionary to ensure the consistency of business fields; The time-series database unit 72 is used to efficiently write high-frequency monitoring data from sensors and provide time-range aggregated queries. The file storage unit 73 is used for distributed storage of various types of files such as pictures, video recordings, and reports transmitted by users.

[0043] In this embodiment, the relational database unit 71 stores user accounts, device metadata, and a data dictionary in a structured manner, enforces consistency of business fields, and outputs a device topology index containing key metadata such as device ID, sensor type, and spatial location. The time-series database unit 72 then uses this index to locate the target device cluster, performs millisecond-level writing of high-frequency sensor monitoring data, and supports aggregation queries of historical environmental trends by time range, generating timestamp index paths. Finally, the file storage unit 73 associates storage spaces with the timestamp paths in the time-series database, distributively storing user-uploaded images, video recordings, analysis reports, and other files within that time period, while simultaneously establishing a cross-storage system index to ensure traceable association between files and business data.

[0044] The data storage module uses a relational database to solidify the topological relationships of the devices, while the time-series database enables the efficient writing of thousands of sensor data points per second. The file storage unit manages unstructured data such as videos and reports using spatiotemporal indexes. Combined with the cross-database aggregation and analysis capabilities of the report generation module, it outputs a visualized daily agricultural situation report containing trend charts and optimization suggestions, providing multi-dimensional data support for planting decisions.

[0045] The report generation module 8 is used to perform statistical analysis on data and its analysis within a certain time range, achieve multi-dimensional data aggregation and security coefficient scoring, and export reports in PDF or other formats; specifically, it includes: The report generation module 8 includes a data statistics unit 81 and a report export unit 82, which are connected in sequence. The data statistics unit 81 is used to aggregate multi-dimensional data such as environmental indicators and equipment status across databases to generate daily trend comparison charts. Customizable report dimension configuration is supported. The report export unit 82 is used to generate PDF and other format documents from the statistical results according to the template, automatically add digital signatures, and provide voice reading function.

[0046] In this embodiment, the data statistics unit 81 aggregates multi-dimensional data such as environmental indicators and equipment status across databases, supports custom report dimension configuration, dynamically generates statistical intermediate data, including daily trend comparison charts and multi-dimensional analysis results. The report export unit 82 then automatically translates the analysis results into Word / PDF formatted documents based on this data, and provides a distribution interface through the open document download API to achieve efficient delivery of analysis conclusions.

[0047] The AI ​​service module 9 uses a large language model based on the Transformer architecture as its core intelligent engine, fine-tuned and optimized with professional knowledge in the agricultural field, and possesses powerful semantic understanding and generation capabilities. It is used to intelligently identify and execute various instructions issued by the business logic module, configure products, manage automatic language processing and multi-turn dialogue, reason about agricultural knowledge, provide proactive warnings and suggestions, and perform personalized continuous learning. The AI ​​service module 9 includes a multimodal interaction unit 91, an intelligent dialogue unit 92, an agricultural knowledge reasoning unit 93, a decision suggestion generation unit 94, and a personalized learning unit 95, which are connected sequentially. The multimodal interaction unit 91 is used to integrate multiple input methods such as voice, image, and text to realize natural language understanding and speech recognition conversion, support visual question answering and image annotation functions, and provide users with a convenient human-computer interaction experience. The multimodal interaction unit integrates technologies such as automatic speech recognition (ASR), text-to-speech (TTS), and computer vision (CV), and supports functions such as voice wake-up, continuous dialogue, and image understanding.

[0048] It has natural language processing capabilities: it supports speech-to-text, semantic understanding, and intent recognition, and can parse complex agricultural instructions, such as "Based on today's weather conditions, help me formulate tomorrow's irrigation plan."

[0049] The intelligent dialogue unit 92 is used to implement multi-turn dialogue management based on a Large Language Model (LLM), maintain context understanding capabilities, and parse user commands into executable operation commands, such as "turn off the irrigation system" and "check soil moisture," ensuring the accuracy of command execution. It employs a dialogue strategy combining state machines and deep reinforcement learning, enabling it to handle complex scenarios such as multi-intent recognition, slot filling, and dialogue repair. It achieves multi-turn dialogue management: maintaining the dialogue context and supporting follow-up questions and clarifications; for example, when a user asks about "the pests and diseases mentioned earlier," it can accurately link the information to previous dialogue content. The agricultural knowledge reasoning unit 93 is used to perform logical reasoning, pest and disease diagnosis and analysis, growth cycle prediction and agricultural timing judgment based on the agricultural expert knowledge base, and provides professional agricultural technical support; it constructs an agricultural expert knowledge graph covering crop planting, pest and disease control, fertilization and irrigation, environmental control and other fields, containing more than 100,000 agricultural rules and experiences collected from the website, supporting logical reasoning based on current environmental data, crop growth stage and historical experience rules and case-based analogical reasoning, and providing decision suggestions; The decision suggestion generation unit 94 is used to combine real-time environmental data and historical experience to generate accurate planting decision suggestions and resource allocation optimization schemes, realizing data-driven intelligent decision-making. It employs a multi-objective optimization algorithm to seek a balance among multiple objectives such as ensuring crop yield, reducing costs, and environmental friendliness, generating the optimal agricultural plan. It can also proactively issue warnings and suggestions, actively pushing risk warnings and optimization suggestions to users based on data trend analysis, without requiring users to actively inquire.

[0050] The personalized learning unit 95 is used to apply collaborative filtering algorithms to match agricultural input demand, recommend agricultural technology knowledge based on user behavior profiles, and continuously optimize the recommendation strategy. In other words, it dynamically adjusts the suggestion strategy according to user operating habits and farm characteristics to achieve personalized services tailored to each individual. Through user profile construction, preference learning, and effect feedback mechanisms, it achieves continuous optimization and personalized adaptation of the suggestion strategy. The entire module coordinates and controls the operation of other modules through the core framework module, acting as a "housekeeper."

[0051] In this embodiment, the multimodal interaction unit 91 integrates speech recognition, image parsing, and text understanding technologies to support farmer dialect recognition and intelligent annotation of disease images, generating a unified multimodal semantic representation vector. The intelligent dialogue unit 92 then implements multi-turn dialogue management based on this vector within a large language model framework, maintaining planting history context, parsing complex agricultural instructions such as "adjust today's irrigation amount based on yesterday's humidity data," and outputting structured operation instructions and demand intentions. The agricultural knowledge reasoning unit 93 invokes an expert knowledge base to perform logical reasoning based on this intention, including matching pest and disease symptoms, predicting growth cycles, and determining the optimal agricultural timing, generating professional diagnostic conclusions and suggested solutions. The decision suggestion generation unit 94 deeply integrates professional solutions with real-time environmental data, such as combining current soil temperature and humidity, weather forecasts, and crop growth stages, outputting precise planting decisions and resource allocation optimization suggestions. Finally, the personalized learning unit 95, based on user historical behavior profiles and decision feedback, continuously optimizes agricultural input product recommendations and agricultural technology knowledge matching strategies through collaborative filtering algorithms, achieving autonomous evolution of AI service capabilities.

[0052] The multimodal intelligent assistant Agent, utilizing the aforementioned AI service module, represents a leap from a traditional "tool-type" system to an "assistant-type" system. Natural language processing technology reduces user learning costs, allowing agricultural novices to quickly master complex operations through voice dialogue. Intelligent dialogue management supports multi-round interactions and follow-up clarification, providing personalized advice as if conversing with an expert. The agricultural knowledge reasoning engine integrates expert experience and scientific data, offering interpretable intelligent support for decision-making. Proactive early warning and suggestion functions transform passive response into proactive service, automatically detecting data or situational anomalies and mobilizing various modules to nip potential risks in the bud. A personalized learning mechanism makes the system increasingly intelligent with use, with suggestion accuracy continuously improving over time, ultimately achieving a personalized agricultural intelligent assistant service, significantly enhancing the user experience and practical value of the smart agriculture system.

[0053] It possesses the following intelligent service capabilities: (1) Proactive intelligent monitoring: Based on data trend analysis, it proactively discovers anomalies and pushes warnings to users without requiring users to actively query; (2) Conversational operation guidance: It guides users to complete complex agricultural operations through natural language interaction, supporting step-by-step teaching and real-time Q&A; (3) Intelligent decision recommendation: It automatically generates the optimal agricultural decision-making scheme by integrating multi-dimensional information such as environmental data, crop status, and weather forecast; (4) Personalized service customization: It provides differentiated intelligent services based on the user's farm size, crop type, management experience, and other characteristics; (5) Continuous learning and evolution: It continuously optimizes service quality through user feedback and effect evaluation, realizing the self-evolution of the intelligent assistant.

[0054] In this embodiment of the invention, the data acquisition module 6 comprehensively acquires environmental sensing, video monitoring, and equipment operation data; the data processing module 5 performs real-time anomaly diagnosis and pattern analysis; the core framework module 1 coordinates and schedules the data storage module 7, user management module 2, and external interface module 3 to achieve cross-layer resource and data integration; the business logic module 4 connects events at each layer to dynamically generate to-do items and drive the execution of multi-domain functions; the report generation module 8 can generate decision reports within a certain time range based on data; in this specific embodiment, it generates a decision report for the day based on data; the AI ​​service module 9 is equipped with a multimodal agricultural intelligent agent, possessing core AI capabilities such as natural language understanding, multi-turn dialogue memory, agricultural knowledge reasoning, equipment control decision-making, anomaly prediction, and personalized suggestion generation, and also undertakes the "housekeeper" function to coordinate the operation of each module. Overall, this invention solves the problems of delayed response, fragmented operation, and fragmented data analysis in existing agricultural systems through a "collection-analysis-decision-response" closed loop, significantly improving the real-time performance, collaboration, and decision-making accuracy of agricultural management.

[0055] For example, in a large-scale smart agriculture base, 2,000 environmental sensors, 100 sets of intelligent irrigation equipment, and 50 high-definition cameras are deployed, achieving intelligent management of the entire process through the collaboration of nine modules. Data acquisition module 5 captures key parameters such as temperature, humidity, soil moisture, and spore concentration in real time, filtering out abnormal fluctuations in the equipment. The video stream processing unit extracts high-definition disease feature frames from pest-risk areas, and the equipment communication unit simultaneously monitors the irrigation valve status and implements remote start / stop control. When data processing module 6 detects that the spore concentration exceeds the threshold, a dynamic early warning signal drives the video intelligent analysis unit to identify anthrax lesions (92% confidence level), and the message queue immediately pushes a Level I alarm asynchronously to the administrator terminal.

[0056] At this point, business logic module 4 initiates a closed-loop response: the to-do list unit classifies the alarm as an emergency work order "B5 Zone Disease Treatment," the intelligent control unit automatically executes a preset script—shutting down the current spraying system and activating the pesticide sprayer, the spraying amount is fed back to the warehouse management unit in real time, triggering a pesticide inventory update and expiration date warning, and simultaneously the blockchain traceability unit encrypts and uploads key data such as application time and concentration to the blockchain, generating a green traceability QR code that consumers can scan to verify. In the face of a sudden rainstorm warning (obtained by external interface module 3 calling the meteorological bureau API), the payment interface unit completes the encrypted transaction of the rainproof film, and the message push unit synchronizes the delivery information to the responsible party.

[0057] The system's stability is ensured throughout operation by the core framework module 1: the Web application framework unit efficiently distributes concurrent requests from 500+ terminals, the database access unit manages the read / write separation of device control commands through a dynamic connection pool, and the caching unit improves the cache hit rate of frequently accessed crop models to 98%. All environmental data is managed hierarchically by the data storage module 7: the relational database records device topology relationships, the time-series database writes sensor dynamics at a rate of thousands of records per second, and the file database stores disease treatment videos according to spatiotemporal indexes.

[0058] In daily management, the report generation module 8 aggregates data such as monitoring trends and operation records, outputting a PDF version of the Smart Agriculture Daily Report with optimization suggestions. The entire system is protected by the user management module 2: agricultural technicians with dual-factor authentication obtain "emergency equipment operation" permissions with dynamic tokens, and all operation logs are linked to personal information files.

[0059] In actual farm management scenarios, the AI ​​service module 9 demonstrates powerful intelligent service capabilities. When a farmer takes a picture of suspected diseased crop leaves with their mobile phone, the multimodal interaction unit 91 first preprocesses and extracts features from the image, while simultaneously recognizing the farmer's voice description, "Why are these leaves turning yellow? Are they spoiled?", and fusing visual and voice information into a unified semantic vector. After receiving this vector, the intelligent dialogue unit 92, while maintaining the historical context of crop planting dialogue, understands that the user is concerned about the yellowing of leaves and automatically associates it with the environmental monitoring data of the plot.

[0060] The agricultural knowledge reasoning unit 93 then initiates expert system reasoning: based on leaf yellowing characteristics, distribution of diseased parts, and current environmental conditions such as high humidity (85%) and temperature (28℃), combined with the built-in plant pathology knowledge base, it quickly diagnoses "bacterial leaf blight" with a confidence level of 89%, and predicts that if not treated promptly, the disease will spread to adjacent plants within 3-5 days. After receiving the diagnostic conclusion, the decision suggestion generation unit 94 comprehensively considers the types of pesticides currently in stock and weather forecasts, such as no rain in the next two days and the crop's early flowering stage, and generates a precise treatment plan: "It is recommended to immediately spray with a copper-based fungicide at a concentration of 800 times, avoiding the midday high-temperature period, with an estimated cost of 45 yuan per mu."

[0061] The personalized learning unit 95 continuously learns throughout the interaction: recording the farmer's preference for organic pest control methods, cost sensitivity, and habit of operating at night, adjusting the recommendation strategy accordingly. When the system encounters similar disease inquiries again, it will prioritize recommending biological agents instead of chemical pesticides and proactively remind users of the optimal application window. Through the accumulation of data from multiple rounds of interactions, the system gradually builds personalized service profiles tailored to each individual, focusing on improving recommendation accuracy and user satisfaction.

[0062] The entire AI service process has transformed from "passive response" to "proactive suggestion": the system can not only accurately diagnose current problems, but also predict potential risks, optimize resource allocation, and personalize service experiences, giving modern agriculture the wings of a "smart brain".

[0063] Through the above process, the nine modules of this invention form a closed loop of "monitoring-analysis-decision-execution-certification", providing modern farms with highly reliable, highly compatible, and fully controllable technical support.

[0064] The above-disclosed embodiments are merely preferred embodiments of a smart agriculture system based on modular collaboration and AI Agent driven by the present invention. They should not be construed as limiting the scope of the present invention. Those skilled in the art can understand that implementing all or part of the above embodiments and making equivalent changes in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A smart agriculture system based on modular collaboration and AI agent-driven operation, characterized in that, It includes a core framework module, a user management module, an external interface module, a business logic module, a data acquisition module, a data processing module, a data storage module, and an AI service module. The core framework module is connected to the user management module, the external interface module, the business logic module, and the AI ​​service module respectively. The data acquisition module, the data processing module, and the data storage module are interconnected and used to provide data support for the core framework module and the AI ​​service module. The AI ​​service module is connected to the business logic module. The core framework module is used to implement user interaction, handle request and response and route management, and implement database access and data caching; The user management module is used for user authentication and permission management; The external interface module is used to connect to weather, agricultural technology information, product information, message push, and payment interface access; The business logic module is used to present and manage to-do items, realize remote control and intelligent command issuance, manage warehousing and transportation records, and verify blockchain product traceability. The data acquisition module is used to collect and record environmental sensing time-series data, video surveillance data, and equipment operation data. The data processing module is used to perform anomaly detection and trend analysis on various environmental sensor time-series data collected by the data acquisition module, perform image recognition and pest monitoring on the acquired video surveillance, and asynchronously process and distribute equipment operation data to the message queue. The data storage module is used to store user information, device status, operation records, search record relational data, time-series data transmitted by sensors, and other image and video data; The AI ​​service module integrates a multimodal agricultural intelligent assistant (Agent) for intelligently identifying and executing various instructions issued by the business logic module, configuring products, recommending system algorithms based on content, automatic language processing, multi-turn dialogue management, agricultural knowledge reasoning, proactive early warning and suggestions, and personalized continuous learning.

2. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, It also includes a report generation module, which is used to perform statistics on data and its analysis within a certain time range, realize multi-dimensional data aggregation and security coefficient scoring, and export reports.

3. The smart agriculture system based on intelligent modular collaboration as described in claim 2, characterized in that, The report generation module includes a data statistics unit and a report export unit connected in sequence. The data statistics unit is used to aggregate environmental indicators and multi-dimensional data on equipment status across databases, generate daily trend comparison charts, and supports custom report dimension configuration. The report export unit is used to generate a document in the required format from the statistical results according to the template, automatically add a digital signature, and provide a voice reading function.

4. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The core framework module includes a Web application framework unit, a database access unit, and a caching unit connected in sequence. The Web application framework unit is used to process HTTP request and response in real time, and to implement API routing and load balancing. The database access unit is used to implement object-relational mapping, automatically execute SQL conversion and transaction management, maintain a dynamic database connection pool, and support read-write separation and failover mechanisms; The caching unit is used to implement high-frequency data caching, provide distributed session management capabilities, implement cache eviction policies to ensure data consistency, and reduce database access pressure.

5. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The external interface module includes a third-party API integration unit, a payment interface unit, and a message push unit connected in sequence. The third-party API integration unit is used to connect to the meteorological bureau's real-time weather API to provide weather conditions and disaster risk forecasts, aggregate agricultural policy information interfaces, and access voice reading interfaces. The payment interface unit is used to integrate the payment SDK to process commodity transactions and implement dual amount verification to prevent tampering; The message push unit is used to push device alarm notifications in real time and send SMS / email reminders simultaneously. It supports customized subscription rules and do-not-disturb settings.

6. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The business logic module includes a to-do item management unit, an intelligent control unit, a warehouse management unit, and a blockchain traceability unit, which are connected in sequence. The to-do list management unit is used to aggregate alarm events from various modules, automatically sort them by urgency, track task handling status, and record processing logs. The intelligent control unit is used to translate user operation commands into device control protocols, receive execution feedback in real time, and support control script presets and batch operations. The warehouse management unit generates shelf-life warnings by monitoring storage and transportation environment parameters, associating inventory status with inventory status, and recording equipment operation and maintenance history to form a full life cycle file. The blockchain traceability unit encrypts and uploads agricultural product production and processing data to the blockchain, generates an unalterable traceability QR code, and provides a distributed ledger verification interface for consumers to check authenticity.

7. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The data acquisition module includes a sensor data receiving unit, a video stream processing unit, and a device communication unit connected in sequence. The sensor data receiving unit is used to collect real-time data streams of temperature and humidity, soil moisture and spore concentration, and filter abnormal fluctuation data of the equipment. The video stream processing unit is used to connect to camera streaming media, complete video encoding and decoding, extract key frames and buffer streaming data, and provide preprocessing support for intelligent analysis. The device communication unit is used to dynamically monitor the online status and operating parameters of the device and implement a command retransmission mechanism to ensure control reliability.

8. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The data processing module includes an environmental data analysis unit, a video intelligent analysis unit, and a message queue unit connected in sequence. The environmental data analysis unit is used to detect abnormal sensor data and generate dynamic threshold warnings, and to predict environmental change trends by combining time series analysis. The video intelligent analysis unit is used to identify crop diseases and pests and output visual analysis tags and confidence reports; The message queue unit is used to asynchronously process high-concurrency tasks and accurately distribute alarm information.

9. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The data storage module includes a relational database unit, a time-series database unit, and a file storage unit. The relational database unit is used to structure and store user account and device metadata and to establish a data dictionary to ensure the consistency of business fields; The time-series database unit is used to write high-frequency monitoring data from sensors and provide time-range aggregated queries; The file storage unit is used for distributed storage of various types of files, including pictures, video recordings, and reports transmitted by users.

10. The smart agriculture system based on intelligent modular collaboration as described in claim 1, characterized in that, The AI ​​service module includes a multimodal interaction unit, an intelligent dialogue management unit, an agricultural knowledge reasoning unit, a decision suggestion generation unit, and a personalized learning unit, which are connected in sequence. The multimodal interaction unit is used to process multiple input methods such as voice, text, and images, and supports voice-to-text, text-to-speech, and image recognition to achieve natural and convenient human-computer interaction. The intelligent dialogue management unit is used to maintain multi-turn dialogue context, supports intent recognition, slot filling, and dialogue status tracking, and can handle complex agricultural consultations and instructions, and supports topic switching and follow-up questions for clarification. The agricultural knowledge reasoning unit is used to perform logical reasoning based on the built-in agricultural expert knowledge graph, and to perform comprehensive analysis in combination with the environmental sensing time-series data obtained by the data acquisition module, and output scientific agricultural suggestions. The environmental sensing time-series data includes current environmental data, crop growth stage, and historical data. The decision suggestion generation unit is used to proactively analyze farm data trends based on the data processed by the data processing module, and generate personalized early warning reminders and optimization suggestions to support risk prediction, resource allocation optimization, and agricultural planning. The personalized learning unit is used to record user operating habits, preference settings, and farm characteristics, and continuously optimizes the suggestion strategy through machine learning algorithms to achieve adaptive evolution of the intelligent assistant.

Images