Agricultural planting system based on artificial intelligence model

Abstract

The embodiment of the invention provides an agricultural planting system based on an artificial intelligence model. The system comprises a data acquisition module used for acquiring multi-modal data of a target planting area, the multi-modal data comprising crop data of a target crop and environmental data of the target planting area; the model deployment module is deployed with an artificial intelligence model and is used for analyzing the multi-modal data by using the artificial intelligence model so as to obtain crop information of the target planting area and generating a decision instruction according to the crop information, the crop information comprises one or more of the following items: crop growth state information, pest and disease prediction information, water and fertilizer demand information and environment adaptation information; and the decision execution module is used for generating a control signal based on the decision instruction and issuing the control signal to the execution equipment to control the execution equipment to execute the corresponding operation task. The system can realize precise, real-time and intelligent management of agricultural planting.

Description

Technical Field

[0001] The present invention relates to the technical field of smart agriculture, and more particularly to an agricultural planting system based on an artificial intelligence model. Background Art

[0002] With the accelerating digital transformation of agriculture, traditional planting models are facing bottlenecks such as inefficient data utilization, lagging decision-making, and extensive management. Precision and intelligence have become the core directions of agricultural development. In current agricultural production, planting decisions mostly rely on empirical judgments or simple statistical analyses, and cannot accurately match the dynamic requirements of crop growth. Especially in key links such as pest and disease control, water and fertilizer regulation, and environmental adaptation, problems such as untimely responses and unreasonable measures are likely to occur.

[0003] To solve the above problems, agricultural big data platforms and decision support systems have emerged in the prior art, but there are still significant limitations: using general-purpose algorithm models, the prediction accuracy is limited, and decision support mostly stays at the level of planting plan recommendations, without realizing the联动 execution with execution devices, and manual intervention is required for operation, resulting in low efficiency and prone to execution deviations. Summary of the Invention

[0004] The present invention is proposed in view of the above problems. The present invention provides an agricultural planting system based on an artificial intelligence model.

[0005] According to one aspect of the present invention, there is provided an agricultural planting system based on an artificial intelligence model, including: a data acquisition module for acquiring multi-modal data of a target planting area, the multi-modal data including crop data of a target crop and environmental data of the target planting area; a model deployment module deployed with an artificial intelligence model, the model deployment module being communicatively connected to the data acquisition module for analyzing the multi-modal data using the artificial intelligence model to obtain crop information of the target planting area, and generating a decision instruction according to the crop information, wherein the crop information includes one or more of the following: crop growth status information, pest and disease prediction information, water and fertilizer requirement information, environmental adaptation information; a decision execution module communicatively connected to the model deployment module for generating a control signal based on the decision instruction and sending the control signal to an execution device to control the execution device to execute a corresponding operation task, the execution device including one or more of the following: an inspection robot, a mobile execution terminal, an irrigation device, a fertilization device, a light supplement device, a ventilation device, a shading device, a pesticide application device, a plant protection robot.

[0006] For example, the model deployment module includes a cloud server and edge nodes that are communicatively connected to each other; the edge nodes are communicatively connected to the data acquisition module and are used to preprocess multimodal data to obtain preprocessed data, and to perform emergency event identification based on the preprocessed data to obtain event identification results, and upload the preprocessed data and abnormal event results to the cloud server, wherein the abnormal event results are event identification results indicating that there are abnormalities in the target planting area; the cloud server is equipped with an artificial intelligence model, which is used to analyze the preprocessed data and / or abnormal event results using the artificial intelligence model to obtain crop information, and to generate decision instructions based on the crop information.

[0007] For example, preprocessing includes: standardizing the multimodal data to obtain standardized multimodal data, establishing data associations between the standardized multimodal data using semantic web technology or federated learning technology to obtain associated multimodal data, and determining preprocessed data based on the associated multimodal data.

[0008] For example, preprocessing further includes: filtering outliers in the multimodal data and filling in missing values ​​in the multimodal data before standardizing the multimodal data to obtain new multimodal data; and / or, determining preprocessed data based on the correlated multimodal data, including: performing data denoising and feature extraction on the correlated multimodal data to obtain preprocessed data.

[0009] For example, the decision execution module is further configured to receive execution feedback data from the execution device and send the execution feedback data to the model deployment module. The model deployment module is further configured to iteratively train the artificial intelligence model based on the execution feedback data and newly collected multimodal data within a preset time period after the execution device performs the task.

[0010] For example, the decision execution module is also used to monitor the operating status of the execution equipment and the execution progress of the task in real time. When it is determined that the execution equipment has malfunctioned based on the operating status, or when it is determined that the task execution is abnormal based on the execution progress, an alarm message is issued and / or the system automatically switches to a backup execution equipment.

[0011] For example, the decision instruction includes a job task; the decision execution module generates control signals based on the decision instruction and sends the control signals to the execution device in the following manner: adding the job task to the task queue, sorting the job tasks in the task queue according to a preset priority rule, generating control signals for the job tasks in the task queue in sequence and sending them to the execution device, wherein, during sorting, if the priority of the first job task currently received is higher than that of the second job task most recently to be executed, the second job task is paused and the first job task is inserted in the task queue before the second job task.

[0012] For example, the model deployment module is also used to: store multimodal data in a distributed file system; and / or synchronize core data to a cloud storage server, the core data including one or more of crop growth status information, decision instructions, and model parameters of the artificial intelligence model.

[0013] For example, the data acquisition module includes one or more of the following: a first environmental sensor set inside the target planting area, a weather station set inside and / or outside the target planting area, a second environmental sensor set on the inspection robot, a first camera set on the inspection robot, a third environmental sensor set on the mobile execution terminal, and a second camera set on the mobile execution terminal; the crop data includes first crop data collected by the first camera and / or second crop data collected by the second camera, and the environmental data includes one or more of the following: first environmental data collected by the first environmental sensor, second environmental data collected by the second environmental sensor, third environmental data collected by the third environmental sensor, and fourth environmental data collected by the weather station.

[0014] For example, the data acquisition module and the model deployment module communicate using a first standardized interface, while the model deployment module and the decision execution module communicate using a second standardized interface.

[0015] According to the embodiment of the present invention, the agricultural planting system based on the artificial intelligence model can construct a full-link architecture of "multimodal data perception - AI intelligent decision-making - equipment linkage execution". The prediction accuracy and environmental adaptability of the artificial intelligence model are better than those of the general algorithm model, and its decision instructions can directly drive the execution equipment to execute automatically without human intervention. It can shorten the response time of the operation task from hours to minutes, reduce labor costs by more than 40%, and realize the precise, real-time and intelligent management of agricultural planting.

[0016] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0017] The above and other objects, features, and advantages of the present invention will become more apparent from the more detailed description of the embodiments of the invention in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same parts or steps.

[0018] Figure 1A schematic block diagram of an agricultural planting system based on an artificial intelligence model according to an embodiment of the present invention is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely a part of the embodiments of the present invention, and not all of the embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention described herein, all other embodiments obtained by those skilled in the art without inventive effort should fall within the protection scope of the present invention.

[0020] To at least partially address the aforementioned problems, embodiments of the present invention provide an agricultural planting system based on an Artificial Intelligence (AI) model. This agricultural planting system can construct a full-link architecture of "multimodal data perception – AI intelligent decision-making – equipment linkage execution," enabling precise, real-time, and intelligent management of agricultural planting.

[0021] Figure 1 A schematic block diagram of an AI-based agricultural planting system 100 according to an embodiment of the present invention is shown. Figure 1 As shown, the agricultural planting system 100 includes a data acquisition module 110, a model deployment module 120, and a decision execution module 130.

[0022] The data acquisition module 110 is used to collect multimodal data of the target planting area, including crop data of the target crop and environmental data of the target planting area.

[0023] The target planting area can be any area, including but not limited to open fields, greenhouses, plastic tunnels, and orchards. The target crop is the crop within the target planting area. The target crop can be any type of crop, including but not limited to grapes, apples, cucumbers, and cabbage. Multimodal data (which can be called multimodal heterogeneous data) can include crop data and environmental data. Crop data is data collected for the target crop. For example, crop data can include one or more of color (RGB) images, depth images, thermal infrared images, and hyperspectral images. Each of the RGB images, depth images, thermal infrared images, and hyperspectral images can be a static image or a continuous dynamic image, i.e., a video stream. Environmental data can include: soil data reflecting the soil conditions within the target planting area, and / or meteorological data reflecting the air conditions within and / or outside the target planting area. For example, environmental data can include one or more of soil temperature, soil moisture, soil electrical conductivity, soil nutrient content, air temperature, air humidity, light intensity, weather conditions (e.g., sunny, cloudy), sunshine duration, precipitation, and air carbon dioxide concentration. Soil electrical conductivity (EC) reflects the total amount of salts and soluble ions in the soil. Soil nutrient content can be measured by the amounts of nitrogen (N), phosphorus (P), and potassium (K).

[0024] The model deployment module 120 deploys an AI model and is connected to the data acquisition module 110. It is used to analyze multimodal data using the AI ​​model to obtain crop information of the target planting area and generate decision instructions based on the crop information. The crop information includes one or more of the following: crop growth status information, pest and disease prediction information, water and fertilizer demand information, and suitable environment information.

[0025] In one embodiment, the model deployment module 120 can be implemented using a collaborative processing architecture of cloud servers and edge nodes, meaning that the processing of multimodal data, model analysis, and decision instruction generation can be distributed across different processing units. In another embodiment, the model deployment module 120 can be implemented using edge nodes alone, meaning the cloud server can be eliminated, and all processing of multimodal data, model analysis, and decision instruction generation can be completed on the edge nodes. Edge nodes can be, for example, local servers. The former embodiment has stronger computing power, supporting the training of complex models and the long-term storage of large-scale data, and is more suitable for scenarios with large-scale planting (e.g., ≥50 greenhouses) and a surge in data processing volume, balancing local real-time performance with the need for cloud computing power expansion. The latter embodiment has limited computing power but can be completely independent of the cloud, with lower data transmission latency and faster response speed, and is more suitable for scenarios with poor network conditions and small-scale planting (e.g., ≤100 acres).

[0026] For example, the AI ​​model can be constructed using one or more algorithms such as Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Random Forest. Preferably, CNN, LSTM, and Random Forest algorithms are integrated to construct a multi-dimensional AI model. For example, the AI ​​model may include one or more of the following: a crop growth status (or growth vigor) monitoring model, a pest and disease prediction model, a water and fertilizer requirement model, and an environmental adaptation model. The crop growth status monitoring model is used to analyze the growth status of a target crop based on at least a portion of the multimodal data to determine crop growth status information. For example, the crop growth status monitoring model can accurately analyze growth indicators such as leaf area index, fruit development progress, and chlorophyll content based on high-resolution images, hyperspectral images, and environmental data of the target crop. These growth indicators can reflect the growth status of the target crop (e.g., growth stage, presence of abnormalities such as yellowing disease, fruit cracking, and powdery mildew). The pest and disease prediction model is used to predict pests and diseases based on at least a portion of the multimodal data to determine pest and disease prediction information. For example, pest and disease prediction models can combine hyperspectral images of the target crop, environmental data (such as air temperature, air humidity, and air carbon dioxide concentration), weather conditions, and historical disease incidence data of the target planting area to construct time-series prediction models, providing early warnings of pest and disease occurrence risks and spread ranges at predetermined times (e.g., 3-7 days). Water and fertilizer demand models are used to analyze water and fertilizer requirements based on at least a portion of multimodal data to determine water and fertilizer demand information. For example, water and fertilizer demand models can calculate precise irrigation amounts, fertilizer types (nitrogen, phosphorus, and potassium ratios), and fertilizer application times based on soil data (such as soil moisture and soil nutrient content), the target crop's growth stage (i.e., growth cycle, such as flowering period), and meteorological data (such as light intensity, weather conditions, and precipitation). Environmental adaptation models are used to perform environmental adaptation analysis based on at least a portion of multimodal data to determine environmental adaptation information. For example, environmental adaptation models can analyze the growth stages of a target crop based on crop data and dynamically match suitable environmental thresholds (i.e., environmental adaptation information) for the target crop at different growth stages, such as suitable air temperature, air humidity, light intensity, and air carbon dioxide concentration at the current growth stage. Environmental adaptation information can provide a basis for decision-making in environmental control in planting scenarios such as greenhouses.

[0027] The model deployment module can generate standardized decision instructions based on the above crop information. For example, the decision instructions may include core information such as the type of the executing device, execution parameters (e.g., irrigation amount, ventilation duration, supplemental lighting intensity, etc.), execution duration, and priority.

[0028] The decision execution module 130 is communicatively connected to the model deployment module 120. It is used to generate control signals based on decision instructions and send the control signals to the execution equipment to control the execution equipment to perform the corresponding work tasks. The execution equipment includes one or more of the following: inspection robot, mobile execution terminal, irrigation equipment, fertilization equipment, supplemental lighting equipment, ventilation equipment, shading equipment, pesticide application equipment, and plant protection robot.

[0029] Those skilled in the art will understand that the main functions of an inspection robot include perception, reconnaissance, and diagnosis (problem detection), and its output is data and information. The main functions of a plant protection robot include execution, intervention, and handling (problem resolution), and its output is physical actions. Inspection robots may have sensors for data collection, including cameras and / or environmental sensors. Inspection robots possess autonomous movement and path planning capabilities, and can trigger close-range data collection on crop growth status and pest and disease conditions according to preset routes or instructions, and can simultaneously acquire environmental data at the collection points. Plant protection robots may have actuators, such as pumps, nozzles, and robotic arms, to perform one or more agricultural operations such as spraying (i.e., applying pesticides), weeding, pruning, and harvesting. Supplemental lighting equipment, ventilation equipment, and shading equipment can be referred to as environmental control equipment. Irrigation equipment, fertilization equipment, and pesticide application equipment can be implemented independently, or any two or more can be integrated together; for example, irrigation equipment and fertilization equipment can be integrated into a single water and fertilizer machine. Irrigation equipment, fertilization equipment, supplemental lighting equipment, ventilation equipment, shading equipment, and pesticide application equipment can be devices that are fixedly installed in the target planting area. Mobile execution terminals can include one or more of the following: drones, underwater robots, and rail robots. They can function as inspection robots and / or agricultural robots, possessing sensors for data collection and / or actuators for performing agricultural operations. Mobile execution terminals can supplement the insufficient coverage of ground-based inspection robots, fixed environmental sensors, and fixed execution equipment, enabling integrated aerial inspection, high-altitude agricultural operations, and large-scale data collection. They are suitable for planting areas such as large-scale open-field cultivation and complex terrain (e.g., mountains and hills).

[0030] The decision execution module 130 can convert standardized decision instructions into control signals recognizable by the execution equipment and send them to the corresponding execution equipment to control the execution equipment to perform corresponding work tasks. Work tasks may include data acquisition and / or agricultural operations. Through control signals, the inspection robot and / or mobile execution terminal can be controlled to collect data from target plots and / or target points, and / or, one or more of the following—mobile execution terminal, irrigation and fertilization equipment, supplemental lighting equipment, ventilation equipment, shading equipment, pesticide application equipment, and plant protection robot—can be controlled to perform corresponding agricultural operations. The "plot" described herein refers to a logical plot, and each plot can be defined by a set of boundary coordinates. Each plot can be further divided into several points, each point can have a unique identifier and be bound to a single crop. For example, the decision execution module 130 can be designed to simultaneously send control signals to multiple types of execution equipment to support the collaborative execution of multiple types of execution equipment. For instance, after controlling the start of the supplemental lighting equipment in a greenhouse, the decision execution module 130 can simultaneously adjust the operating parameters of the greenhouse's ventilation equipment to maintain the temperature and humidity balance within the greenhouse.

[0031] By adopting the above technical solution, a full-link architecture of "multimodal data perception - AI intelligent decision-making - equipment linkage execution" can be constructed. The prediction accuracy and environmental adaptability of the AI ​​model are better than those of general-purpose algorithm models, and its decision commands can directly drive the execution equipment to execute automatically without human intervention. This can shorten the response time of the task from hours to minutes, reduce labor costs by more than 40%, and realize precise, real-time and intelligent management of agricultural planting.

[0032] According to embodiments of the present invention, the data acquisition module includes one or more of the following: a first environmental sensor installed inside the target planting area; a weather station installed inside and / or outside the target planting area; a second environmental sensor installed on an inspection robot; a first camera installed on the inspection robot; a third environmental sensor installed on a mobile execution terminal; and a second camera installed on the mobile execution terminal. Crop data includes first crop data collected by the first camera and / or second crop data collected by the second camera. Environmental data includes one or more of the following: first environmental data collected by the first environmental sensor; second environmental data collected by the second environmental sensor; third environmental data collected by the third environmental sensor; and fourth environmental data collected by the weather station.

[0033] For example, the first camera may include one or more of the following: an RGB camera, a color depth (RGBD) camera, a thermal infrared camera, a hyperspectral camera, etc. Correspondingly, the first crop data may include one or more of the following: RGB images acquired by an RGB camera, RGB images acquired by an RGBD camera, depth images acquired by an RGBD camera, thermal infrared images acquired by a thermal infrared camera, hyperspectral images acquired by a hyperspectral camera, etc. The RGB camera may be, for example, a high-resolution RGB camera with at least 12 million effective pixels. Details of the target crop, such as color, texture, and small lesions, can be clearly identified from the RGB images acquired by the high-resolution RGB camera. The hyperspectral camera of the inspection robot can dynamically adjust imaging parameters through an adaptive spectral selection algorithm to optimize the acquisition quality of crop data. For example, the second camera may include one or more of the following: an RGB camera, a color depth (RGBD) camera, a thermal infrared camera, a hyperspectral camera, etc. Correspondingly, the second crop data may include one or more of the following: RGB images acquired by an RGB camera, RGB images acquired by an RGBD camera, depth images acquired by an RGBD camera, thermal infrared images acquired by a thermal infrared camera, hyperspectral images acquired by a hyperspectral camera, etc. Both the inspection robot and the mobile execution terminal can carry one or more cameras to collect crop data of the target crop.

[0034] For example, the first environmental sensor may include a soil sensor buried in the target planting area and / or a meteorological sensor fixedly installed within the target planting area. For example, in a greenhouse or polytunnel planting scenario, each greenhouse or polytunnel may deploy 6-8 sets of the first environmental sensors according to a "uniform distribution" principle, wherein the installation height of the meteorological sensor can be in the range of 1.5-2m, and the burial depth of the soil sensor can be in the range of 10-15cm. The soil sensor may include one or more of the following: soil temperature sensor, soil moisture sensor, soil conductivity sensor, soil nutrient content sensor, etc. Each of the soil sensors may exist individually or multiple sensors may be integrated together; for example, the soil temperature sensor and the soil moisture sensor may be implemented using an integrated soil temperature and humidity sensor. The meteorological sensor may include one or more of the following: air temperature sensor, air humidity sensor, light sensor, carbon dioxide concentration sensor, etc. Each of the meteorological sensors may exist individually or multiple sensors may be integrated together; for example, the air temperature sensor and the air humidity sensor may be implemented using an integrated air temperature and humidity sensor. For example, the first environmental data may include one or more of the following: soil temperature collected by a soil temperature sensor, soil moisture collected by a soil moisture sensor, soil conductivity collected by a soil conductivity sensor, soil nutrient content collected by a soil nutrient content sensor, air temperature collected by an air temperature sensor, air humidity collected by an air humidity sensor, light intensity collected by a light sensor, and air carbon dioxide concentration collected by a carbon dioxide concentration sensor. The second environmental sensor may include a meteorological sensor mounted on the inspection robot. This meteorological sensor may include one or more of the following: air temperature sensor, air humidity sensor, light intensity sensor, and carbon dioxide concentration sensor. Similarly, each of these meteorological sensors may exist individually or multiple sensors may be integrated together. For example, the second environmental data may include one or more of the following: air temperature collected by an air temperature sensor, air humidity collected by an air humidity sensor, light intensity collected by a light sensor, and air carbon dioxide concentration collected by a carbon dioxide concentration sensor. The third environmental sensor may include a meteorological sensor mounted on a mobile execution terminal. This meteorological sensor may include one or more of the following: air temperature sensor, air humidity sensor, light intensity sensor, and carbon dioxide concentration sensor. Similarly, each of these meteorological sensors may exist individually or multiple sensors may be integrated together. For example, the third environmental data may include one or more of the following: air temperature collected by an air temperature sensor, air humidity collected by an air humidity sensor, light intensity collected by a light sensor, and air carbon dioxide concentration collected by a carbon dioxide concentration sensor. The fourth environmental data is collected by a weather station.Weather stations can be set up in relatively open areas within the target planting area, or around the target planting area. For example, the fourth environmental data may include one or more of the following: air temperature, air humidity, light intensity, air carbon dioxide concentration, weather conditions, sunshine duration, precipitation, etc.

[0035] The data acquisition module 110 can support the automatic acquisition of multimodal data from environmental sensors, weather stations, inspection robots, etc., and can convert the acquired data into one or more preset data formats, such as JPG, JSON, CSV, etc.

[0036] The above scheme provides a method for collecting crop and environmental data using data acquisition devices deployed in different ways, making it convenient to use appropriate data acquisition devices for data collection in different planting scenarios.

[0037] According to an embodiment of the present invention, the model deployment module includes a cloud server and edge nodes that are communicatively connected to each other; the edge nodes are communicatively connected to the data acquisition module and are used to preprocess multimodal data to obtain preprocessed data, and to perform emergency event identification based on the preprocessed data to obtain event identification results, and to upload the preprocessed data and abnormal event results to the cloud server, wherein the abnormal event results are event identification results indicating that there are abnormalities in the target planting area; the cloud server is equipped with an AI model, which is used to analyze the preprocessed data and / or abnormal event results using the AI ​​model to obtain crop information, and to generate decision instructions based on the crop information.

[0038] Cloud servers can be deployed with AI model training platforms and big data storage and analysis modules, possessing large-scale data processing and model training capabilities. Edge nodes are the core physical or logical computing units in an edge computing architecture. They can be deployed close to the data source (such as farmland, greenhouses, or factory workshops) for localized data processing, analysis, and response, rather than uploading all data to a distant cloud. Edge nodes can be implemented using any hardware device or software container with local computing, storage, and communication capabilities. For example, an edge node can be part of a control center. The control center can include industrial-grade programmable logic controllers (PLCs) that support MQTT / Modbus protocols, capable of receiving decision instructions from the cloud server in real time and distributing them to the execution devices, while simultaneously monitoring the operating status of the execution devices.

[0039] Edge nodes can preprocess multimodal data, such as denoising and feature extraction, and can identify emergency events based on the preprocessed data. Emergency events can be defined according to preset emergency event rules. For example, emergency events can include pest and disease outbreaks, sudden changes in environmental data (which may correspond to extreme weather conditions), etc., meaning emergency event identification can be the identification of pest and disease outbreaks, sudden changes in environmental data, etc. Edge nodes can upload the preprocessed data to the cloud server. When the event identification result indicates an anomaly in the target planting area, i.e., when the event identification result is an abnormal event result, the edge node can also upload the abnormal event result to the cloud server. For example, edge nodes can perform primary pest and disease detection using crop data to quickly detect suspected pest and disease areas, and the event identification result can include these suspected pest and disease areas. Edge nodes can upload the preprocessed data and suspected pest and disease areas to the cloud server, so that the cloud server can further determine whether pests and diseases have occurred based on the preprocessed data and suspected pest and disease areas, and generate corresponding prevention and control decisions (i.e., decision instructions) when pests and diseases are determined to have occurred.

[0040] In existing technologies, data processing typically relies entirely on cloud architecture, resulting in high data transmission latency and insufficient real-time performance and response speed. This makes it difficult to cope with unexpected scenarios in the field, such as outbreaks of pests and diseases or extreme weather, leading to delayed decision-making and potential production losses. This embodiment employs an architecture that combines edge nodes and cloud servers. Edge nodes are responsible for real-time data preprocessing and rapid response to unexpected scenarios (such as emergency pest and disease warnings), while the cloud server handles the deployment of complex models (including model training and big data analysis). This approach effectively balances real-time performance and computing power, reducing decision-making latency to, for example, less than 2 seconds, thus perfectly meeting the real-time management needs of planting scenarios.

[0041] According to an embodiment of the present invention, preprocessing includes: standardizing the multimodal data to obtain standardized multimodal data, establishing data associations between the standardized multimodal data using semantic web technology or federated learning technology to obtain associated multimodal data, and determining preprocessed data based on the associated multimodal data.

[0042] Multimodal data can be converted into a unified format through standardization (or normalization), resulting in standardized multimodal data. Standardized multimodal data can then establish data relationships using Semantic Web technology or federated learning, such as "soil moisture data of a greenhouse - corresponding regional crop data - concurrent meteorological data." Semantic Web technology can naturally connect data from different sources and with different structures through unified URIs and RDF links, forming a data network. Federated learning (FL) is a distributed machine learning paradigm where "the data doesn't move, the model moves." It allows multiple participants to exchange only model parameters / gradients, with the central server aggregating a global model, achieving the goal of "sharing knowledge, not sharing data," provided their local data remains within its domain. Federated learning can enable cross-regional joint modeling in scenarios where data from multiple planting areas is not interconnected, ensuring data privacy in each planting area. Federated learning is primarily applicable to large-scale chain farms and cross-regional planting bases. Compared to federated learning, Semantic Web technology has lower modeling complexity and lower requirements for network bandwidth and edge node computing power.

[0043] Existing agricultural planting systems mostly process single-type data (such as sensor data or image data), lacking semantic-level fusion mechanisms for multimodal data. This results in low data correlation, difficulty in fully extracting data value, and an inability to provide comprehensive support for decision-making. This embodiment innovatively employs semantic web technology or federated learning technology to achieve correlation analysis of multimodal data, breaking down data silos and providing AI models with a comprehensive and accurate data source, improving decision accuracy by approximately 25%-30%.

[0044] According to embodiments of the present invention, preprocessing further includes: filtering outliers in the multimodal data and filling in missing values ​​in the multimodal data before standardizing the multimodal data to obtain new multimodal data; and / or determining preprocessed data based on the correlated multimodal data, including: performing data denoising and feature extraction on the correlated multimodal data to obtain preprocessed data.

[0045] In the first embodiment, outliers (e.g., jump points) in the multimodal data can be filtered out and missing values ​​imputed before standardization to obtain new multimodal data. It is understood that subsequent standardization of the multimodal data involves standardizing the new multimodal data. For example, outliers in the multimodal data can be filtered out using statistical methods (e.g., the 3σ principle) and machine learning algorithms. For example, missing values ​​in the multimodal data can be imputed using linear interpolation and the K-nearest neighbor algorithm. By filtering outliers and imputed missing values, the accuracy and comprehensiveness of the multimodal data can be improved, thereby enhancing the accuracy of subsequent analysis and decision-making.

[0046] In the second embodiment, determining preprocessed data based on correlated multimodal data may include: performing data denoising and feature extraction on the correlated multimodal data to obtain preprocessed data. Exemplarily, data denoising may be implemented using one or more algorithms such as moving average or moving median filtering, model-based filtering (e.g., Kalman filtering, particle filtering), and spatial domain filtering. Exemplary methods for feature extraction of various types of data are described below. For light intensity, the rate of change in light intensity can be extracted as a feature, i.e., calculating the difference in light intensity between two consecutive moments ΔL / Δt, reflecting the trend of light intensity change. For temperature data, the corresponding temperature change gradient can be extracted as a feature, i.e., calculating ΔT / Δt, reflecting the potential impact of environmental conditions on light distribution. Similar to temperature data, for humidity data, the corresponding humidity change gradient can be extracted as a feature, i.e., calculating ΔH / Δt, also reflecting the potential impact of environmental conditions on light distribution. For weather conditions, weather trend factors can be extracted as features, i.e., combining the time series of the day's weather conditions to extract the weather change trend for the next few hours, such as the light reduction factor corresponding to "sunny to cloudy".

[0047] The first and second embodiments described above can be combined in the same embodiment as needed.

[0048] According to an embodiment of the present invention, the decision execution module is further configured to receive execution feedback data from the execution device and send the execution feedback data to the model deployment module. The model deployment module is further configured to: iteratively train the AI ​​model based on the execution feedback data and newly collected multimodal data within a preset time period after the execution device performs the task.

[0049] The model deployment module supports continuous learning and transfer learning of the AI ​​model. The decision execution module receives execution feedback data from the execution equipment. The data acquisition module collects multimodal data in real time. After the execution equipment performs its tasks, the multimodal data and execution feedback data are fed back to the model deployment module in real time by the data acquisition module and the decision execution module, respectively, so that the model deployment module can use this data to iteratively train the AI ​​model. Transfer learning can quickly apply existing crop model knowledge to similar crops, reducing the data and time costs required for training new crop models by more than 60%. In addition, customized AI models can be developed for different crop varieties, growth stages, and planting area environments.

[0050] For example, the execution feedback data may include the actual execution parameters of the executing device. For irrigation equipment, actual execution parameters may include: actual operating time, actual water flow (i.e., actual irrigation volume), water pressure, valve opening degree, etc. For fertilization equipment, actual execution parameters may include: EC / pH value, etc. For ventilation equipment, actual execution parameters may include: actual fan speed, actual window opening angle, actual wind speed, actual ventilation time, etc. For supplemental lighting equipment, actual execution parameters may include: actual supplemental lighting intensity, actual operating time, etc. For shading equipment, actual execution parameters may include the actual opening / closing percentage of the shading curtain, etc. Each type of executing device has its own corresponding actual execution parameters, which can be set according to actual needs; these will not be elaborated upon here. The parameter type (first parameter type) of the actual execution parameters of each executing device may be consistent with the parameter type (second parameter type) of the execution parameters of that executing device included in the above decision instruction, or the first parameter type corresponding to each executing device may at least include the second parameter type corresponding to that executing device.

[0051] Existing general-purpose algorithm models lack a continuous learning mechanism and cannot dynamically update model parameters based on new data, resulting in performance degradation after long-term use. However, the AI ​​model according to embodiments of the present invention can combine transfer learning and continuous learning techniques to dynamically update model parameters, further improving prediction accuracy and environmental adaptability compared to general-purpose algorithm models.

[0052] According to an embodiment of the present invention, the decision execution module is also used to monitor the operating status of the execution equipment and the execution progress of the task in real time. When it is determined that the execution equipment has malfunctioned based on the operating status, or when it is determined that the task execution is abnormal based on the execution progress, an alarm message is issued and / or the system automatically switches to a backup execution equipment.

[0053] The decision execution module may include a status monitoring unit for real-time monitoring of the operating status and task progress of the execution equipment. Operating status includes, for example, the flow rate of irrigation equipment, the battery level of the plant protection robot, and the working status of environmental control equipment. When the operating status indicates a malfunction in the execution equipment, or when the execution progress indicates an abnormal task execution, the status monitoring unit can immediately trigger an alarm, such as by pushing a notification to the manager's mobile terminal. This promptly reminds the manager to inspect and maintain the execution equipment, ensuring the normal execution of the task. When the operating status indicates a malfunction in the execution equipment, or when the execution progress indicates an abnormal task execution, the status monitoring unit can also automatically switch to a backup execution equipment. The backup execution equipment then takes over from the malfunctioning or abnormal execution equipment to perform subsequent tasks, ensuring the continuity of the task. For example, issuing alarm information and automatically switching to the backup execution equipment can be performed simultaneously.

[0054] According to an embodiment of the present invention, the decision instruction includes a job task; the decision execution module generates a control signal based on the decision instruction and sends the control signal to the execution device in the following manner: adding the job task to the task queue, sorting the job tasks in the task queue according to a preset priority rule, generating control signals for the job tasks in the task queue in sequence and sending them to the execution device, wherein, during sorting, if the priority of the first job task currently received is higher than that of the second job task most recently to be executed, the second job task is paused and the first job task is inserted in the task queue before the second job task.

[0055] The decision execution module may include an instruction scheduling unit. This unit adds tasks to a task queue, sorts them according to a preset priority rule, and generates control signals for the tasks in the queue in that order, sending them to the execution equipment. For example, the preset priority rule could be something like "emergency control (weight 0.6) > critical growth period (weight 0.3) > routine monitoring (weight 0.1)". The instruction scheduling unit supports the execution of high-priority tasks (hereinafter referred to as high-priority tasks) in the queue. When a high-priority task is generated, it can pause the execution of lower-priority tasks (hereinafter referred to as low-priority tasks). High-priority tasks could be, for example, emergency pest and disease control or sudden changes in environmental data. This scheduling scheme allows for timely responses to events requiring urgent handling by executing corresponding tasks in the queue, thus improving the management of the planting area.

[0056] According to an embodiment of the present invention, the model deployment module is further configured to: store multimodal data in a distributed file system; and / or synchronize core data to a cloud storage server, wherein the core data includes one or more of crop growth status information, decision instructions, and model parameters of the AI ​​model.

[0057] For example, massive amounts of multimodal data can be stored using a distributed file system such as HDFS or other similar distributed file systems, and an indexing engine (such as Elasticsearch) can be optionally configured to support fast data retrieval, historical backtracking, and multi-dimensional relational queries.

[0058] For example, core data (such as crop growth status information, decision-making instructions, and AI model parameters) can be synchronized to a cloud storage server to enable rapid data recovery in the event of local server failure, ensuring data security and business continuity. The above-described solutions of storing multimodal data in a distributed file system and synchronizing core data to a cloud storage server can be implemented in the same embodiment.

[0059] Existing agricultural planting systems typically employ a monolithic architecture. Adding new data types and crop varieties necessitates large-scale system reconstruction, resulting in high adaptation costs and a lack of data backup and fault redundancy mechanisms, leading to weak system stability and resilience. The agricultural planting system described in this embodiment boasts high scalability and reliability, making it suitable for large-scale planting. By employing distributed storage, it supports rapid integration of new data types and crop varieties. Combining cloud backup and fault redundancy mechanisms improves system stability by approximately 35%, adapting to various planting scenarios from small-scale individual farmers to large-scale farms.

[0060] According to an embodiment of the present invention, the data acquisition module and the model deployment module communicate using a first standardized interface, and the model deployment module and the decision execution module communicate using a second standardized interface.

[0061] The standardized interfaces described in this paper (including the first and second standardized interfaces) can be RESTful API interfaces and / or WebSocket interfaces. These standardized interfaces support the transmission of video streams (RTSP) and data in standard formats such as JSON, Protobuf, and XML, thus facilitating third-party integration. Adopting a "modular + standardized interface" design philosophy allows for the independent decoupling of various functional modules (sensing, analysis and decision-making, execution and control) within the agricultural planting system, facilitating module replacement and upgrades.

[0062] Although exemplary embodiments have been described herein with reference to the accompanying drawings, it should be understood that the above exemplary embodiments are merely illustrative and are not intended to limit the scope of the invention. Various changes and modifications can be made therein by those skilled in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the invention as claimed in the appended claims.

[0063] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0064] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0065] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0066] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, this approach should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the corresponding claims, its inventive point lies in solving the corresponding technical problem with fewer features than all of those in a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0067] Those skilled in the art will understand that, apart from the mutual exclusion of features, all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or apparatus so disclosed can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0068] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments can be used in any combination.

[0069] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some modules in the AI ​​model-based agricultural planting system according to embodiments of the present invention. The present invention can also be implemented as an apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0070] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0071] The above description is merely a specific embodiment of the present invention or an explanation of that embodiment. The scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An agricultural planting system based on an artificial intelligence model, characterized in that, include: The data acquisition module is used to collect multimodal data of the target planting area, including crop data of the target crop and environmental data of the target planting area. The model deployment module deploys an artificial intelligence model and is communicatively connected to the data acquisition module. It is used to analyze the multimodal data using the artificial intelligence model to obtain crop information of the target planting area and generate decision instructions based on the crop information. The crop information includes one or more of the following: crop growth status information, pest and disease prediction information, water and fertilizer demand information, and environmental adaptation information. The decision execution module is communicatively connected to the model deployment module. It is used to generate control signals based on the decision instructions and send the control signals to the execution devices to control the execution devices to perform corresponding work tasks. The execution devices include one or more of the following: inspection robots, mobile execution terminals, irrigation equipment, fertilization equipment, supplemental lighting equipment, ventilation equipment, shading equipment, pesticide application equipment, and plant protection robots.

2. The agricultural planting system according to claim 1, characterized in that, The model deployment module includes cloud servers and edge nodes that are interconnected. The edge node is communicatively connected to the data acquisition module and is used to preprocess the multimodal data to obtain preprocessed data, and to perform emergency event identification based on the preprocessed data to obtain event identification results, and to upload the preprocessed data and abnormal event results to the cloud server. The abnormal event results are the event identification results indicating that there is an anomaly in the target planting area. The cloud server is equipped with the artificial intelligence model, which is used to analyze the preprocessed data and / or the results of the abnormal events to obtain the crop information and generate the decision instructions based on the crop information.

3. The agricultural planting system according to claim 2, characterized in that, The preprocessing includes: The multimodal data is standardized to obtain standardized multimodal data, and data associations are established between the standardized multimodal data using semantic web technology or federated learning technology to obtain associated multimodal data. The preprocessed data is then determined based on the associated multimodal data.

4. The agricultural planting system according to claim 3, characterized in that, The preprocessing further includes: filtering outliers in the multimodal data and filling in missing values ​​in the multimodal data before standardizing the multimodal data to obtain new multimodal data; And / or, The step of determining the preprocessed data based on the correlated multimodal data includes: performing data denoising and feature extraction on the correlated multimodal data to obtain the preprocessed data.

5. The agricultural planting system according to any one of claims 1-4, characterized in that, The decision execution module is further configured to receive execution feedback data from the execution device and send the execution feedback data to the model deployment module. The model deployment module is also used to iteratively train the artificial intelligence model based on the execution feedback data and the multimodal data newly collected within a preset time period after the execution device performs the task.

6. The agricultural planting system according to any one of claims 1-4, characterized in that, The decision execution module is also used to monitor the operating status of the execution equipment and the execution progress of the task in real time. When it is determined that the execution equipment has malfunctioned based on the operating status, or when it is determined that the task is abnormal based on the execution progress, an alarm message is issued and / or the system automatically switches to a backup execution equipment.

7. The agricultural planting system according to any one of claims 1-4, characterized in that, The decision instruction includes the work task; the decision execution module generates a control signal based on the decision instruction and sends the control signal to the execution device in the following manner: The task is added to the task queue, and the tasks in the task queue are sorted according to a preset priority rule. The task in the task queue generates the control signal in sequence and sends it to the execution device. During sorting, if the priority of the first task received is higher than that of the second task to be executed most recently, the second task is paused and the first task is inserted in the task queue before the second task.

8. The agricultural planting system according to any one of claims 1-4, characterized in that, The model deployment module is also used for: The multimodal data is stored in a distributed file system; and / or, The core data is synchronized to the cloud storage server. The core data includes one or more of the following: crop growth status information, decision instructions, and model parameters of the artificial intelligence model.

9. The agricultural planting system according to any one of claims 1-4, characterized in that, The data acquisition module includes one or more of the following: a first environmental sensor installed inside the target planting area, a weather station installed inside and / or outside the target planting area, a second environmental sensor installed on the inspection robot, a first camera installed on the inspection robot, a third environmental sensor installed on the mobile execution terminal, and a second camera installed on the mobile execution terminal. The crop data includes first crop data collected by the first camera and / or second crop data collected by the second camera, and the environmental data includes one or more of the following: first environmental data collected by the first environmental sensor, second environmental data collected by the second environmental sensor, third environmental data collected by the third environmental sensor, and fourth environmental data collected by the weather station.

10. The agricultural planting system according to any one of claims 1-4, characterized in that, The data acquisition module and the model deployment module communicate using a first standardized interface, and the model deployment module and the decision execution module communicate using a second standardized interface.

Images