Intelligent agriculture monitoring system based on augmented reality and Internet of Things

Abstract

The invention discloses an intelligent agriculture monitoring system based on augmented reality and Internet of Things, and relates to the technical field of monitoring analysis, and the system comprises a data collection module which is used for collecting environment feature data corresponding to an agricultural production region, and obtaining an agricultural scene dynamic data set; the data processing module is used for performing space-time alignment processing on the environment feature data and the agricultural scene dynamic data set to construct an agricultural scene augmented reality three-dimensional model; the prediction model construction module is used for confirming a hydrothermal interaction feature set and establishing a pest and disease risk prediction model based on the hydrothermal interaction feature set; and the early warning module is used for performing multi-dimensional data mapping on an output result of the pest risk prediction model and the augmented reality three-dimensional model, generating a visual risk labeling layer, performing risk grade division based on the visual risk labeling layer, establishing a real-time monitoring feedback data stream, and performing synchronous early warning on a terminal through an Internet of Things gateway. The agricultural monitoring method has the effect of improving the agricultural monitoring efficiency.

Description

Technical Field

[0001] The present application relates to the field of monitoring and analysis technology, and in particular to a smart agricultural monitoring system based on augmented reality and the Internet of Things. Background Art

[0002] Internet of Things monitoring refers to the real-time collection of any object or process that needs to be monitored, connected, and interacted through various devices and technologies such as various information sensors, radio frequency identification technology, global positioning systems, infrared sensors, laser scanners, etc., and the collection of various required information. Through various possible network accesses, it realizes ubiquitous connection between objects and objects, and objects and people, and realizes intelligent perception, identification, and management of objects and processes.

[0003] Augmented reality technology uses optoelectronic display technology, interactive technology, multiple sensor technologies, computer graphics and multimedia technology to integrate the computer-generated virtual environment with the real environment around the user, so that the user can be convinced from the sensory effect that the virtual environment is an integral part of the real environment around him. Augmented reality has the new characteristics of combining virtual and real, real-time interaction, and three-dimensional registration.

[0004] In related technologies, traditional smart agricultural monitoring usually uses multispectral cameras to obtain vegetation indices, and makes agricultural management decisions by establishing statistical relationships between environmental parameters and crop yields. In terms of pest and disease early warning, existing technologies mostly identify diseases based on single environmental factor threshold judgments, which has significant limitations, reduces agricultural monitoring efficiency, and leaves room for improvement. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, this application provides a smart agricultural monitoring system based on augmented reality and the Internet of Things.

[0006] In a first aspect, the present application provides a smart agricultural monitoring system based on augmented reality and the Internet of Things, comprising: The data acquisition module is used to collect environmental characteristic data corresponding to agricultural production areas through the Internet of Things sensor network. At the same time, it uses drones equipped with augmented reality equipment to collect multispectral images according to adaptive tracks, thereby obtaining a dynamic dataset of agricultural scenes. A data processing module is used to perform spatiotemporal alignment processing on environmental feature data and dynamic agricultural scene datasets, thereby constructing an augmented reality 3D model of the agricultural scene and embedding an updateable annotation layer on the surface of the augmented reality 3D model of the agricultural scene; A prediction model building module is used to analyze the microclimate characteristic gradient based on the augmented reality 3D model of the agricultural scene, thereby confirming the water-heat interaction feature set and establishing a pest and disease risk prediction model based on the water-heat interaction feature set; The early warning module is used to perform multi-dimensional data mapping between the output results of the pest and disease risk prediction model and the augmented reality three-dimensional model, generate a visual risk annotation layer, divide the risk levels based on the visual risk annotation layer, establish a real-time monitoring feedback data stream, and issue synchronous early warnings to the terminal through the Internet of Things gateway.

[0007] Preferably, the step of obtaining a dynamic agricultural scene dataset specifically includes: Deploy an IoT sensor network in agricultural production areas. The IoT sensor network includes soil moisture sensors, leaf wetness sensors, and micro-meteorological stations to collect environmental characteristic data sets corresponding to the agricultural production areas. Using drones equipped with augmented reality devices, edge computing nodes generate adaptive tracks based on farmland terrain features, and dynamically adjust spectral acquisition parameters during flight based on the adaptive tracks to adapt to crop growth conditions. The multispectral image data collected by UAVs are subjected to radiation correction and geometric registration processing, and the processed multispectral image data are associated according to a unified spatiotemporal benchmark to construct a dynamic dataset of agricultural scenes.

[0008] Preferably, the environmental feature data and the agricultural scene dynamic data set are subjected to spatiotemporal alignment processing to construct an augmented reality three-dimensional model of the agricultural scene, which specifically includes the following steps: Interpolate the environmental characteristic data corresponding to the agricultural production area to generate a continuous raster dataset; Reconstruct 3D point clouds from multispectral image data and use LiDAR-assisted point cloud registration to build a digital surface model of farmland terrain. A multi-source data spatiotemporal benchmark alignment matrix is established, and the continuous raster dataset is fused with the digital surface model of farmland terrain in multiple layers to extract the coupling relationship between crop phenotypic characteristics and surface parameters. Based on the coupling relationship between crop phenotypic characteristics and surface parameters, an augmented reality three-dimensional model of the agricultural scene is constructed.

[0009] Preferably, the steps of establishing a pest and disease risk prediction model specifically include: Extract soil-atmosphere interface heat flux parameters, canopy stomatal conductance parameters, and root layer hydraulic conductivity parameters from the augmented reality 3D model of agricultural scenes to construct a water-heat interaction feature set. Acquire microclimate characteristic data based on the water-heat interaction feature set, perform gradient direction analysis on the microclimate characteristics using an attention mechanism, and establish a meteorological element distribution map; Pre-train the pest and disease feature transfer learning model and map the characteristics of historical pest and disease samples to the current growth environment to generate a heat map of disease occurrence probability; The disease occurrence probability heat map and meteorological element distribution map are integrated to establish a disease and insect pest risk prediction model, and then the spatiotemporal distribution data of risk levels within a preset time period in the future are output based on the disease and insect pest risk prediction model.

[0010] Preferably, a pest and disease feature transfer learning model is pre-trained, and historical pest and disease sample features are mapped to the current growth environment to generate a disease occurrence probability heat map, specifically including the following steps: Constructing a regional pest and disease characteristic migration database, wherein the regional pest and disease characteristic migration database includes spectral response characteristics, microenvironmental parameter combination patterns, and phenotypic abnormal change trajectories of typical diseases in different climate zones; Design a dual-channel feature extraction network and use a multi-task learning framework to jointly optimize disease type identification and risk level prediction, and pre-train a pest and disease feature transfer learning model; Based on the output results of the pest and disease feature transfer learning model, a disease occurrence probability heat map is generated, and the disease occurrence probability heat map is divided into low-risk areas, warning areas, and high-risk areas based on preset thresholds.

[0011] Preferably, the output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in a multi-dimensional data manner to generate a visual risk annotation layer, and risk levels are divided based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway, specifically including: The disease occurrence probability heat map is spatially mapped with the augmented reality 3D model of the agricultural scene. The annotation layer corresponding to the augmented reality 3D model of the agricultural scene is color-coded and rendered based on the division of low-risk areas, warning areas, and high-risk areas to generate a visual risk annotation layer. Dynamic pulse warning signs are added to high-risk areas, and a multi-dimensional environmental parameter query interface is embedded at the corresponding geographic coordinates. Build a hierarchical warning rule engine to automatically generate feedback data streams based on low-risk areas, warning areas, and high-risk areas; The feedback data stream is converted into protocol through the IoT gateway, real-time monitoring information is pushed to the terminal, and synchronous early warning is carried out.

[0012] In a second aspect, the present application provides a smart agricultural monitoring method based on augmented reality and the Internet of Things, comprising the following steps: The environmental characteristic data corresponding to the agricultural production area is collected through the Internet of Things sensor network. At the same time, drones equipped with augmented reality equipment are used to collect multispectral images according to adaptive tracks to obtain dynamic data sets of agricultural scenes. The environmental feature data is spatiotemporally aligned with the dynamic agricultural scene dataset to construct an augmented reality 3D model of the agricultural scene. An updateable annotation layer is embedded in the surface of the augmented reality 3D model of the agricultural scene. Microclimate gradient analysis is performed based on an augmented reality 3D model of agricultural scenarios to identify a water-heat interaction feature set, and a pest and disease risk prediction model is established based on this feature set. The output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in multiple dimensions to generate a visual risk annotation layer, and risk levels are divided based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway.

[0013] In a third aspect, the present application provides a computer-readable storage medium storing instructions. When the instructions are executed on a computer, the computer executes any one of the above-mentioned smart agricultural monitoring systems based on augmented reality and the Internet of Things.

[0014] In summary, this application has the following beneficial technical effects: The present application provides a smart agricultural monitoring system based on augmented reality and the Internet of Things. By acquiring and spatiotemporally aligning the environmental feature data and agricultural scene dynamic data sets corresponding to the agricultural production area, an augmented reality three-dimensional model of the agricultural scene is constructed. The microclimate feature gradient is analyzed based on the augmented reality three-dimensional model of the agricultural scene, and the water-heat interaction feature set is confirmed. A pest and disease risk prediction model is established based on the water-heat interaction feature set. The output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in multi-dimensional data to generate a visual risk annotation layer. The risk level is divided based on the visual risk annotation layer, and a real-time monitoring feedback data stream is established. The system sends a synchronous warning to the terminal through the Internet of Things gateway, thereby effectively reducing the occurrence of significant limitations in disease identification based on a single environmental factor threshold judgment, thereby effectively improving the efficiency of agricultural monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0016] Figure 1 This is a system diagram of smart agricultural monitoring based on augmented reality and the Internet of Things in an embodiment of the present application.

[0017] Figure 2This is a flow chart of the method for smart agricultural monitoring based on augmented reality and the Internet of Things in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The following is combined with Figure 1-2 This application is described in further detail.

[0019] Example 1 The embodiment of the present application discloses a smart agricultural monitoring system based on augmented reality and the Internet of Things.

[0020] Reference Figure 1 , a smart agricultural monitoring system based on augmented reality and the Internet of Things, including: The data acquisition module is used to collect environmental characteristic data corresponding to agricultural production areas through the Internet of Things sensor network. At the same time, it uses drones equipped with augmented reality equipment to collect multispectral images according to adaptive tracks, thereby obtaining a dynamic dataset of agricultural scenes. A data processing module is used to perform spatiotemporal alignment processing on environmental feature data and dynamic agricultural scene datasets, thereby constructing an augmented reality 3D model of the agricultural scene and embedding an updateable annotation layer on the surface of the augmented reality 3D model of the agricultural scene; A prediction model building module is used to analyze the microclimate characteristic gradient based on the augmented reality 3D model of the agricultural scene, thereby confirming the water-heat interaction feature set and establishing a pest and disease risk prediction model based on the water-heat interaction feature set; The early warning module is used to perform multi-dimensional data mapping between the output results of the pest and disease risk prediction model and the augmented reality three-dimensional model, generate a visual risk annotation layer, divide the risk levels based on the visual risk annotation layer, establish a real-time monitoring feedback data stream, and issue synchronous early warnings to the terminal through the Internet of Things gateway.

[0021] Using the above technical solution, multi-source data is collected collaboratively through the IoT sensor network and drones. In agricultural production areas, the IoT sensor network deploys temperature and humidity sensors, light sensors, soil moisture sensors, and other sensors to collect real-time environmental characteristic data such as air temperature, humidity, light intensity, and soil moisture content. At the same time, drones equipped with augmented reality devices use adaptive flight paths to collect multispectral images. The adaptive trajectory planning algorithm dynamically adjusts the flight path based on farmland topography, crop distribution, and meteorological conditions to ensure coverage of the entire monitoring area. For example, during the critical period of crop growth, drones increase the frequency of shooting in key areas. Multispectral images contain information in the visible and near-infrared bands and can be used to analyze crop growth, chlorophyll content, etc. Through these two methods, a dynamic dataset of agricultural scenes containing environmental parameters and crop phenotypic characteristics is obtained, providing rich material for subsequent analysis. The data processing module fuses the collected multi-source data and maps the two types of data into a unified spatiotemporal coordinate system through timestamp matching and geographic coordinate calibration. For example, the temperature and humidity data collected by the sensor at a certain moment is associated with the multispectral image of the area taken by a drone at the same time. Based on the aligned data, 3D reconstruction technology is used to construct an augmented reality 3D model of the agricultural scene. The augmented reality 3D model of the agricultural scene intuitively displays the farmland topography, crop distribution and growth status. To facilitate user understanding and operation, an updateable annotation layer is embedded in the surface of the augmented reality 3D model of the agricultural scene. The annotation content includes information such as crop variety, growth cycle, irrigation history, and these annotations can be dynamically updated based on real-time data. The prediction model construction module uses augmented reality 3D models to analyze microclimate characteristics and predict pests and diseases. It uses the rich environmental and crop information in the augmented reality 3D models of agricultural scenarios to analyze microclimate characteristic gradients, such as temperature and humidity differences between different plots. It extracts water-heat interaction feature sets from the microclimate data. The water-heat interaction feature sets are closely related to the occurrence of pests and diseases. For example, high temperature and high humidity environments are prone to breeding fungal diseases. Based on historical data, a pest and disease risk prediction model is established. The pest and disease risk prediction model uses the water-heat interaction feature set as input and outputs the probability and severity of pest and disease occurrence. For example, by analyzing temperature and humidity data for multiple consecutive days, the probability of powdery mildew occurring in a certain plot within the next week can be predicted. The early warning module combines prediction results with augmented reality 3D models to achieve visual early warning. The output of the pest and disease risk prediction model is mapped onto the augmented reality 3D model, generating a visual risk annotation layer. Areas of varying risk are color-coded within this layer, for example, red for high-risk, yellow for medium-risk, and green for low-risk. Warning areas are also classified according to risk grading criteria, and corresponding response measures are developed for each risk level. A real-time monitoring feedback data stream is established to continuously track pest and disease trends. When risk levels change or new risk points emerge, synchronized early warning messages are sent to end devices (such as mobile phones and tablets) via the IoT gateway, prompting farmers to take timely prevention and control measures. For example, if the pest and disease risk level in a particular area rises from low to medium, an automatic early warning message is sent, along with targeted prevention and control recommendations. This collaborative module achieves a closed-loop management process from data collection, processing, analysis, to early warning, providing intelligent decision-making support for agricultural production, effectively reducing pest and disease losses and improving agricultural monitoring efficiency.

[0022] It should be noted that the steps for obtaining a dynamic dataset of agricultural scenarios include: Deploy an IoT sensor network in agricultural production areas. The IoT sensor network includes soil moisture sensors, leaf wetness sensors, and micro-meteorological stations to collect environmental characteristic data sets corresponding to the agricultural production areas. Using drones equipped with augmented reality devices, edge computing nodes generate adaptive tracks based on farmland terrain features, and dynamically adjust spectral acquisition parameters during flight based on the adaptive tracks to adapt to crop growth conditions. The multispectral image data collected by UAVs are subjected to radiation correction and geometric registration processing, and the processed multispectral image data are associated according to a unified spatiotemporal benchmark to construct a dynamic dataset of agricultural scenes.

[0023] Specifically, an IoT sensor network is deployed in agricultural production areas. The IoT sensor network includes soil moisture sensors, leaf wetness sensors, and micro-meteorological stations to collect environmental characteristic data sets corresponding to the agricultural production areas. Soil moisture sensors are buried in the soil at different depths to monitor soil moisture content, nutrient content, and other parameters in real time. For example, data is collected at regular intervals (such as 15 minutes) to obtain changes in soil moisture at different depths. Leaf wetness sensors are installed near crop leaves to accurately measure the humidity on the leaf surface and reflect the moisture status of the crops. Micro-meteorological stations are placed in open areas of agricultural production areas to continuously monitor meteorological factors such as air temperature, humidity, wind speed, wind direction, and light intensity. The above sensors work together to transmit the collected data to the data center via wireless communication technology, forming an environmental characteristic data set for the agricultural production area, providing basic data for subsequent analysis; Using drones equipped with augmented reality devices, edge computing nodes generate adaptive tracks based on farmland terrain features, and dynamically adjust spectral acquisition parameters during flight based on the adaptive track to adapt to crop growth status. The augmented reality devices equipped with drones can perceive farmland terrain, crop distribution, and other information in real time. The edge computing nodes use advanced algorithms based on this information and combined with farmland terrain features to generate adaptive tracks. For example, when the drone detects terrain undulations or obstacles in the farmland, the edge computing nodes adjust the track to avoid the above areas, ensuring flight safety and data acquisition accuracy. During flight, the drone dynamically adjusts spectral acquisition parameters based on crop growth status. When crops are in different growth stages, their reflection and absorption characteristics for different spectra are different. By analyzing the crop growth stage, the drone adjusts spectral acquisition parameters, such as selecting appropriate bands and adjusting shooting resolution, to obtain more accurate crop information. For example, in the early stages of crop growth, the drone may select spectral bands sensitive to chlorophyll for acquisition to monitor crop growth status. The multispectral image data collected by the UAV is subjected to radiation correction and geometric registration processing, and the processed multispectral image data is associated according to a unified time and space benchmark to construct a dynamic data set of agricultural scenes. The multispectral image data collected by the UAV may be affected by factors such as sensor characteristics and lighting conditions, and there are radiation errors and geometric deformations. The multispectral image data is subjected to radiation correction and geometric registration processing. The radiation correction analyzes the image data and corrects its radiation characteristics to make the image data of different bands comparable. The geometric matching principle is to match the multispectral image data with the geographic coordinate system to eliminate the geometric deformation caused by factors such as the UAV's flight posture and terrain changes. After processing, the multispectral image data is Image data is associated according to a unified spatiotemporal benchmark, that is, image data collected at different times and locations are integrated with environmental feature data. For example, the multispectral image data collected by a drone at a certain moment is associated with the environmental feature data collected by the IoT sensor network at the same moment to construct an agricultural scene dynamic dataset. The agricultural scene dynamic dataset contains environmental information and crop growth information of the agricultural production area, providing comprehensive data support for agricultural production decision-making. Through the above steps, from the deployment of the IoT sensor network to the construction of the agricultural scene dynamic dataset, comprehensive monitoring and data collection of the agricultural production area are realized, providing strong data guarantee for the intelligent management and decision-making of agricultural production.

[0024] It should be noted that the spatial and temporal alignment of environmental feature data with the agricultural scene dynamic dataset to construct an augmented reality 3D model of the agricultural scene specifically includes the following steps: Interpolate the environmental characteristic data corresponding to the agricultural production area to generate a continuous raster dataset; Reconstruct 3D point clouds from multispectral image data and use LiDAR-assisted point cloud registration to build a digital surface model of farmland terrain. A multi-source data spatiotemporal benchmark alignment matrix is established, and the continuous raster dataset is fused with the digital surface model of farmland terrain in multiple layers to extract the coupling relationship between crop phenotypic characteristics and surface parameters. Based on the coupling relationship between crop phenotypic characteristics and surface parameters, an augmented reality three-dimensional model of the agricultural scene is constructed.

[0025] Specifically, the environmental characteristic data corresponding to the agricultural production area is interpolated to generate a continuous raster dataset. The environmental characteristic data are usually discrete data points collected by multiple sensors distributed in the agricultural production area, such as soil moisture, temperature, light intensity, etc. Interpolation processing is to expand and fit the above discrete data points through mathematical methods to generate a continuous raster dataset. For example, based on known soil moisture data points, considering the spatial correlation between data points, the soil moisture values of each location in the entire agricultural production area are calculated, thereby obtaining a continuous soil moisture raster dataset. Through the above interpolation processing, the spatial distribution of environmental characteristics in the agricultural production area can be more accurately reflected, providing richer data support for subsequent analysis; The multispectral image data is reconstructed into a three-dimensional point cloud, and a digital surface model of the farmland terrain is established by combining it with lidar-assisted point cloud registration. The multispectral image data contains spectral information of crops and farmland, but lacks three-dimensional spatial information. Three-dimensional point cloud reconstruction is the process of converting multispectral image data into three-dimensional point cloud data. For example, stereo vision technology is used to calculate the three-dimensional coordinates of each pixel based on multispectral images taken at different angles to form three-dimensional point cloud data. The lidar can accurately measure the height information of the farmland terrain. Through lidar-assisted point cloud registration, the three-dimensional point cloud data is matched and fused with the point cloud data obtained by the lidar. For example, the point cloud data generated by the multispectral image and the point cloud data obtained by the lidar scan are adjusted in spatial position and posture so that the two can accurately correspond. In this way, a digital surface model of the farmland terrain is established, which can intuitively display the undulations of the farmland terrain and the spatial distribution of crops. A multi-source data spatiotemporal benchmark alignment matrix is established, and the continuous raster dataset is fused with the farmland terrain digital surface model through multi-layer overlay, thereby extracting the coupling relationship between crop phenotypic characteristics and surface parameters, and then constructing an augmented reality 3D model of the agricultural scene based on the coupling relationship between crop phenotypic characteristics and surface parameters. The multi-source data spatiotemporal benchmark alignment matrix is a tool for unifying the time and space benchmarks of different data sources. For example, the environmental feature data in the continuous raster dataset is aligned with the crop and terrain information in the farmland terrain digital surface model in time and space. Through multi-layer overlay and fusion, the continuous raster dataset is overlaid on the farmland terrain digital surface model. Analyze the relationship between the two. For example, analyze the relationship between environmental characteristics such as soil moisture and crop growth conditions and terrain undulations, extract the coupling relationship between crop phenotypic characteristics (such as crop height, density, etc.) and surface parameters (such as soil fertility, terrain slope, etc.), and construct an augmented reality 3D model of the agricultural scene based on the coupling relationship and combined with augmented reality technology. In the augmented reality 3D model of the agricultural scene, users can use augmented reality devices such as smart glasses or mobile phone applications to intuitively view the environmental characteristics, crop growth conditions and terrain information of the agricultural production area, and make scientific agricultural production decisions based on the coupling relationship between crop phenotypic characteristics and surface parameters.

[0026] It should be noted that the steps for establishing a pest and disease risk prediction model include: Extract soil-atmosphere interface heat flux parameters, canopy stomatal conductance parameters, and root layer hydraulic conductivity parameters from the augmented reality 3D model of agricultural scenes to construct a water-heat interaction feature set. Acquire microclimate characteristic data based on the water-heat interaction feature set, perform gradient direction analysis on the microclimate characteristics using an attention mechanism, and establish a meteorological element distribution map; Pre-train the pest and disease feature transfer learning model and map the characteristics of historical pest and disease samples to the current growth environment to generate a heat map of disease occurrence probability; The disease occurrence probability heat map and meteorological element distribution map are integrated to establish a disease and insect pest risk prediction model, and then the spatiotemporal distribution data of risk levels within a preset time period in the future are output based on the disease and insect pest risk prediction model.

[0027] Specifically, soil-atmosphere interface heat flux parameters, canopy stomatal conductance parameters, and root layer hydraulic conductivity parameters are extracted from the agricultural scene augmented reality 3D model to construct a water-heat interaction feature set. The agricultural scene augmented reality 3D model contains rich agricultural production information. Through professional data analysis tools and algorithms, soil-atmosphere interface heat flux parameters are extracted from the agricultural scene augmented reality 3D model. For example, the energy balance principle is used, combined with temperature, humidity and other data in the model, to calculate the heat exchange rate between the soil and the atmosphere, and obtain the soil-atmosphere interface heat flux parameters. For the canopy stomatal conductance parameters, the influence of the opening and closing degree of the stomata on gas exchange is analyzed according to the spectral characteristics and physiological characteristics of the crop canopy, and then the canopy stomatal conductance parameters are determined. The root layer hydraulic conductivity parameters are obtained by analyzing the water conductivity capacity of the root layer by combining the hydraulic principles with the analysis of soil texture, root distribution and other information, and then the root layer hydraulic conductivity parameters are obtained. The above parameters are integrated to construct a water-heat interaction feature set to reflect the dynamic process of water and heat exchange in the agricultural production area. Based on the water-heat interaction feature set, microclimate characteristic data is obtained, and the attention mechanism is used to perform gradient direction analysis on the microclimate characteristics to establish a meteorological element distribution map. The water-heat interaction feature set provides a basis for obtaining microclimate characteristic data. By analyzing the water-heat interaction feature set and combining it with other environmental data, microclimate characteristic data such as temperature, humidity, and wind speed are obtained. The attention mechanism is used to perform gradient direction analysis on the microclimate characteristics. For example, the temperature gradient and humidity gradient are focused on to analyze the changing trends in different directions. Through the analysis of microclimate characteristics, a meteorological element distribution map is established to intuitively display the distribution of meteorological elements in the agricultural production area. For example, the map can show areas with higher or lower temperatures, and areas with higher or lower humidity, providing meteorological information for subsequent pest and disease risk prediction; Pre-train the pest and disease feature transfer learning model, and map the historical pest and disease sample features to the current growth environment to generate a heat map of the probability of disease occurrence. The pest and disease feature transfer learning model is a model that can use historical data for prediction. By collecting a large amount of historical pest and disease sample data, the pest and disease feature transfer learning model is pre-trained so that the pest and disease feature transfer learning model can learn the characteristics and occurrence patterns of pests and diseases, and map the historical pest and disease sample features to the current growth environment. Considering factors such as meteorological elements, crop varieties, and soil conditions in the current environment, the applicability of historical pest and disease sample features in the current environment is analyzed. For example, according to the current meteorological conditions such as temperature and humidity, as well as the growth stage of the crop, the historical pest and disease sample features are matched with the current environment. Through the above mapping, a heat map of the probability of disease occurrence is generated, and different colors are used to represent the probability of disease occurrence in different areas. For example, red areas indicate a higher probability of disease occurrence, and green areas indicate a lower probability of disease occurrence. By integrating the disease occurrence probability heat map with the meteorological element distribution map, a pest and disease risk prediction model is established. This model then outputs the spatiotemporal distribution data of risk levels within a preset future time period. By integrating the disease occurrence probability heat map with the meteorological element distribution map, the distribution of both the disease occurrence probability and meteorological elements is comprehensively considered. For example, in areas where the disease occurrence probability is high and meteorological conditions are favorable for pest and disease spread, the pest and disease risk level is high; whereas in areas where the disease occurrence probability is low and meteorological conditions are unfavorable for pest and disease spread, the pest and disease risk level is low. Based on this fused data, a pest and disease risk prediction model is established. This model can predict the pest and disease risk level for different regions within a preset future time period based on current environmental data and historical pest and disease data. For example, it can predict the pest and disease risk level for each region within the next week and output the spatiotemporal distribution data of the risk level, providing a decision-making basis for agricultural production managers, allowing them to take timely prevention and control measures and reduce the impact of pests and diseases on agricultural production.

[0028] It should be noted that pre-training the pest and disease feature transfer learning model and mapping the historical pest and disease sample features to the current growth environment to generate the disease occurrence probability heat map specifically includes the following steps: Constructing a regional pest and disease characteristic migration database, wherein the regional pest and disease characteristic migration database includes spectral response characteristics, microenvironmental parameter combination patterns, and phenotypic abnormal change trajectories of typical diseases in different climate zones; Design a dual-channel feature extraction network and use a multi-task learning framework to jointly optimize disease type identification and risk level prediction, and pre-train a pest and disease feature transfer learning model; Based on the output results of the pest and disease feature transfer learning model, a disease occurrence probability heat map is generated, and the disease occurrence probability heat map is divided into low-risk areas, warning areas, and high-risk areas based on preset thresholds.

[0029] Specifically, a regional pest and disease characteristic migration database is constructed. The regional pest and disease characteristic migration database includes the spectral response characteristics of typical diseases in different climate zones, the combination pattern of microenvironmental parameters and the trajectory of abnormal phenotypic changes. Monitoring points are set up in different climate zones to conduct long-term monitoring of a variety of typical diseases. Wheat rust is monitored in temperate climate zones, and rice blast is monitored in tropical climate zones. The spectral response characteristics of different diseases at different growth stages are obtained through hyperspectral remote sensing technology. For example, in the early stage of wheat rust, the leaf spectrum will have unique reflection and absorption characteristics in specific bands. At the same time, microenvironmental parameters are collected. The combination patterns of parameters such as temperature, humidity, and soil fertility are crucial for understanding the environmental conditions under which diseases occur. For example, rice blast is more likely to occur under conditions of high humidity, suitable temperature, and specific soil fertility. These parameter combination patterns are crucial for understanding the environmental conditions under which diseases occur. Furthermore, through field observations and image recognition technology, the trajectories of abnormal crop phenotypic changes, such as leaf color changes and plant growth morphology changes, are recorded. By integrating these spectral response characteristics, microenvironmental parameter combination patterns, and phenotypic abnormality change trajectories, a regional pest and disease migration database is constructed, providing rich data support for subsequent analysis and prediction. A dual-channel feature extraction network is designed, and a multi-task learning framework is used to jointly optimize disease type identification and risk level prediction. A pre-trained pest and disease feature transfer learning model is used. The dual-channel feature extraction network is designed to more comprehensively extract disease-related features. One channel focuses on spectral feature extraction, and the other channel focuses on phenotypic feature extraction. For example, in the spectral feature extraction channel, a convolutional neural network is used to process hyperspectral images to extract the spectral features of the disease; in the phenotypic feature extraction channel, the phenotypic features of the crop are extracted through image segmentation and feature extraction technology. The multi-task learning framework combines the two tasks of disease type identification and risk level prediction. Combined, during the training process, the objective function is optimized simultaneously. For example, when identifying disease types, the disease type is determined by comparing the spectral and phenotypic characteristics of typical diseases in the database. When predicting risk levels, the possibility of disease occurrence and the scope of impact are predicted by combining microenvironmental parameters and disease transmission models. Through the multi-task learning framework, spectral and phenotypic characteristics can be better utilized to improve the accuracy of disease type identification and risk level prediction. When pre-training the pest and disease feature transfer learning model, data from the regional pest and disease feature transfer database is used. Through a large number of sample training, the model learns the characteristics and occurrence patterns of typical diseases in different climate zones. Based on the output of the pest and disease feature transfer learning model, a heat map of disease occurrence probability is generated, and based on a preset threshold, the heat map of disease occurrence probability is divided into low-risk areas, warning areas, and high-risk areas. The output of the pest and disease feature transfer learning model includes information such as disease type and risk level. Using this information, a heat map of disease occurrence probability is generated. For example, based on the disease occurrence probability predicted by the model, different colors are used on the map to represent the disease occurrence probability in different areas. The darker the color, the higher the probability of disease occurrence. The heat map of disease occurrence probability is divided based on the preset threshold. For example, the disease occurrence probability of 0 to 0.3 is set as a low-risk area, 0.3 to 0.7 as a warning area, and 0.7 to 1 as a high-risk area. Through the above division method, agricultural production managers can intuitively understand the disease risk situation in different areas and take corresponding prevention and control measures. For example, in high-risk areas, monitoring and prevention are strengthened, and routine monitoring is carried out in low-risk areas to improve the efficiency and benefits of agricultural production.

[0030] It should be noted that the output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in a multi-dimensional data manner to generate a visual risk annotation layer, and risk levels are divided based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway. Specifically, the following steps are involved: The disease occurrence probability heat map is spatially mapped with the augmented reality 3D model of the agricultural scene. The annotation layer corresponding to the augmented reality 3D model of the agricultural scene is color-coded and rendered based on the division of low-risk areas, warning areas, and high-risk areas to generate a visual risk annotation layer. Dynamic pulse warning signs are added to high-risk areas, and a multi-dimensional environmental parameter query interface is embedded at the corresponding geographic coordinates. Build a hierarchical warning rule engine to automatically generate feedback data streams based on low-risk areas, warning areas, and high-risk areas; The feedback data stream is converted into protocol through the IoT gateway, real-time monitoring information is pushed to the terminal, and synchronous early warning is carried out.

[0031] Specifically, the disease occurrence probability heat map is spatially mapped with the agricultural scene augmented reality three-dimensional model, and the annotation layer corresponding to the agricultural scene augmented reality three-dimensional model is color-coded and rendered based on the divided low-risk areas, warning areas and high-risk areas to generate a visual risk annotation layer, and a dynamic pulse warning mark is added to the high-risk area, and a multi-dimensional environmental parameter query interface is embedded at the corresponding geographic coordinates. The disease occurrence probability heat map reflects the possibility of disease occurrence in different areas, and the agricultural scene augmented reality three-dimensional model contains information such as farmland topography and crop distribution. Through spatial mapping technology, the disease occurrence probability heat map is integrated with the agricultural scene augmented reality three-dimensional model. For example, the risk areas in the disease occurrence probability heat map are corresponded to the farmland areas in the three-dimensional model, so that users can intuitively see the distribution of disease risks in the three-dimensional model. Based on the divided low, medium and high risk areas, the annotation layer of the agricultural scene augmented reality three-dimensional model is color-coded and rendered. For example, low-risk areas are marked green, early warning areas are marked yellow, and high-risk areas are marked red. For high-risk areas, dynamic pulse warning signs are added to attract users' attention. For example, a flashing pulse sign will appear in the red area, reminding users that the probability of disease occurrence in this area is high. A multi-dimensional environmental parameter query interface is embedded at the corresponding geographic coordinates, through which environmental parameters such as temperature, humidity, and soil fertility of the area can be queried. For example, when a user clicks on the query interface of a high-risk area, detailed environmental parameter information of the area can be obtained, providing a basis for disease and pest control. A hierarchical early warning rule engine is constructed to automatically generate feedback data streams based on low-risk areas, early warning areas, and high-risk areas. The hierarchical early warning rule engine is constructed to take different early warning measures according to different risk levels. When designing early warning rules, factors such as the probability of disease occurrence, disease type, and environmental parameters are considered. For example, in low-risk areas, if the probability of disease occurrence remains stable for a certain period of time and the environmental parameters are conducive to crop growth, a low-risk feedback data stream will be generated to prompt users to continue monitoring. In early warning areas, when the probability of disease occurrence increases or environmental parameters change unfavorably to crop growth, an early warning feedback data stream will be generated to advise users to take preventive measures, such as strengthening field management and spraying preventive pesticides. For high-risk areas, when the probability of disease occurrence exceeds a certain threshold and the environmental parameters are unfavorable for crop growth, a high-risk feedback data stream will be generated to prompt users to take immediate preventive measures, such as emergency spraying of pesticides, adjustment of irrigation and fertilization strategies, etc. The hierarchical early warning rule engine can automatically generate corresponding feedback data streams according to different risk levels, providing users with targeted decision support; The feedback data stream is encapsulated and converted through the IoT gateway to push real-time monitoring information to the terminal and synchronized early warning. The IoT gateway is a key component that connects sensors in agricultural production areas, augmented reality three-dimensional models of agricultural scenes and terminal devices. The feedback data stream contains information such as disease risk level, environmental parameters, and prevention and control recommendations. The feedback data stream is encapsulated and converted through the IoT gateway to convert the data format into a format suitable for terminal devices to receive. For example, the feedback data stream is encapsulated in JSON format so that it can be displayed on mobile devices and real-time monitoring information is pushed to the terminal. Users can obtain disease risk information in agricultural production areas in real time through mobile phones, tablets and other terminal devices. When the system detects a high-risk area, it will issue a synchronized early warning and send early warning information to users through terminal devices, such as push notifications and alarm sounds. Users can take corresponding prevention and control measures in a timely manner based on the early warning information to improve the safety and efficiency of agricultural production.

[0032] Example 2 The embodiment of the present application also discloses a smart agriculture monitoring method based on augmented reality and the Internet of Things.

[0033] Reference Figure 2 , a smart agricultural monitoring method based on augmented reality and the Internet of Things, comprising the following steps: The environmental characteristic data corresponding to the agricultural production area is collected through the Internet of Things sensor network. At the same time, drones equipped with augmented reality equipment are used to collect multispectral images according to adaptive tracks to obtain dynamic data sets of agricultural scenes. The environmental feature data is spatiotemporally aligned with the dynamic agricultural scene dataset to construct an augmented reality 3D model of the agricultural scene. An updateable annotation layer is embedded in the surface of the augmented reality 3D model of the agricultural scene. Microclimate gradient analysis is performed based on an augmented reality 3D model of agricultural scenarios to identify a water-heat interaction feature set, and a pest and disease risk prediction model is established based on this feature set. The output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in multiple dimensions to generate a visual risk annotation layer, and risk levels are divided based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway.

[0034] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in similar ways. As long as they do not deviate from the concept of the invention, they should all fall within the scope of protection of the present invention.

[0035] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0036] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention.

Claims

1. A smart agricultural monitoring system based on augmented reality and the Internet of Things, characterized by: include: The data acquisition module is used to collect environmental characteristic data corresponding to agricultural production areas through the Internet of Things sensor network. At the same time, it uses drones equipped with augmented reality equipment to collect multispectral images according to adaptive tracks, thereby obtaining a dynamic dataset of agricultural scenes. A data processing module is used to perform spatiotemporal alignment processing on environmental feature data and dynamic agricultural scene datasets, thereby constructing an augmented reality 3D model of the agricultural scene and embedding an updateable annotation layer on the surface of the augmented reality 3D model of the agricultural scene; A prediction model building module is used to analyze the microclimate characteristic gradient based on the augmented reality 3D model of the agricultural scene, thereby confirming the water-heat interaction feature set and establishing a pest and disease risk prediction model based on the water-heat interaction feature set; The early warning module is used to perform multi-dimensional data mapping between the output results of the pest and disease risk prediction model and the augmented reality three-dimensional model, generate a visual risk annotation layer, divide the risk levels based on the visual risk annotation layer, establish a real-time monitoring feedback data stream, and issue synchronous early warnings to the terminal through the Internet of Things gateway.

2. The smart agricultural monitoring system based on augmented reality and the Internet of Things according to claim 1, characterized in that: The steps to obtain a dynamic dataset of agricultural scenarios include: Deploy an IoT sensor network in agricultural production areas. The IoT sensor network includes soil moisture sensors, leaf wetness sensors, and micro-meteorological stations to collect environmental characteristic data sets corresponding to the agricultural production areas. Using drones equipped with augmented reality devices, edge computing nodes generate adaptive tracks based on farmland terrain features, and dynamically adjust spectral acquisition parameters during flight based on the adaptive tracks to adapt to crop growth conditions. The multispectral image data collected by UAVs are subjected to radiation correction and geometric registration processing, and the processed multispectral image data are associated according to a unified spatiotemporal benchmark to construct a dynamic dataset of agricultural scenes.

3. The smart agricultural monitoring system based on augmented reality and the Internet of Things according to claim 1, characterized in that: The environmental feature data is spatiotemporally aligned with the agricultural scene dynamic dataset to construct an augmented reality 3D model of the agricultural scene. The specific steps include: Interpolate the environmental characteristic data corresponding to the agricultural production area to generate a continuous raster dataset; Reconstruct 3D point clouds from multispectral image data and use LiDAR-assisted point cloud registration to build a digital surface model of farmland terrain. A multi-source data spatiotemporal benchmark alignment matrix is established, and the continuous raster dataset is fused with the digital surface model of farmland terrain in multiple layers to extract the coupling relationship between crop phenotypic characteristics and surface parameters. Based on the coupling relationship between crop phenotypic characteristics and surface parameters, an augmented reality three-dimensional model of the agricultural scene is constructed.

4. The smart agricultural monitoring system based on augmented reality and the Internet of Things according to claim 1, characterized in that: The steps to establish a pest and disease risk prediction model include: Extract soil-atmosphere interface heat flux parameters, canopy stomatal conductance parameters, and root layer hydraulic conductivity parameters from the augmented reality 3D model of agricultural scenes to construct a water-heat interaction feature set. Acquire microclimate characteristic data based on the water-heat interaction feature set, perform gradient direction analysis on the microclimate characteristics using an attention mechanism, and establish a meteorological element distribution map; Pre-train the pest and disease feature transfer learning model and map the characteristics of historical pest and disease samples to the current growth environment to generate a heat map of disease occurrence probability; The disease occurrence probability heat map and meteorological element distribution map are integrated to establish a disease and insect pest risk prediction model, and then the spatiotemporal distribution data of risk levels within a preset time period in the future are output based on the disease and insect pest risk prediction model.

5. The smart agricultural monitoring system based on augmented reality and the Internet of Things according to claim 4, characterized in that: Pre-train the pest and disease feature transfer learning model, map the historical pest and disease sample features to the current growth environment, and generate a disease occurrence probability heat map. The specific steps include: Constructing a regional pest and disease characteristic migration database, wherein the regional pest and disease characteristic migration database includes spectral response characteristics, microenvironmental parameter combination patterns, and phenotypic abnormal change trajectories of typical diseases in different climate zones; Design a dual-channel feature extraction network and use a multi-task learning framework to jointly optimize disease type identification and risk level prediction, and pre-train a pest and disease feature transfer learning model; Based on the output results of the pest and disease feature transfer learning model, a disease occurrence probability heat map is generated, and the disease occurrence probability heat map is divided into low-risk areas, warning areas, and high-risk areas based on preset thresholds.

6. The smart agricultural monitoring system based on augmented reality and the Internet of Things according to claim 5, characterized in that: The output of the pest and disease risk prediction model is mapped to the augmented reality three-dimensional model in multiple dimensions to generate a visual risk annotation layer. Risk levels are classified based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway. Specifically, the following steps are performed: The disease occurrence probability heat map is spatially mapped with the augmented reality 3D model of the agricultural scene. The annotation layer corresponding to the augmented reality 3D model of the agricultural scene is color-coded and rendered based on the division of low-risk areas, warning areas, and high-risk areas to generate a visual risk annotation layer. Dynamic pulse warning signs are added to high-risk areas, and a multi-dimensional environmental parameter query interface is embedded at the corresponding geographic coordinates. Build a hierarchical warning rule engine to automatically generate feedback data streams based on low-risk areas, warning areas, and high-risk areas; The feedback data stream is converted into protocol through the IoT gateway, real-time monitoring information is pushed to the terminal, and synchronous early warning is carried out.

7. A smart agriculture monitoring method based on augmented reality and the Internet of Things, applied to a smart agriculture monitoring system based on augmented reality and the Internet of Things as described in any one of claims 1 to 6, characterized in that: The following steps are involved: The environmental characteristic data corresponding to the agricultural production area is collected through the Internet of Things sensor network. At the same time, drones equipped with augmented reality equipment are used to collect multispectral images according to adaptive tracks to obtain dynamic data sets of agricultural scenes. The environmental feature data is spatiotemporally aligned with the dynamic agricultural scene dataset to construct an augmented reality 3D model of the agricultural scene. An updateable annotation layer is embedded in the surface of the augmented reality 3D model of the agricultural scene. Microclimate gradient analysis is performed based on an augmented reality 3D model of agricultural scenarios to identify a water-heat interaction feature set, and a pest and disease risk prediction model is established based on this feature set. The output results of the pest and disease risk prediction model are mapped to the augmented reality three-dimensional model in multiple dimensions to generate a visual risk annotation layer, and risk levels are divided based on the visual risk annotation layer. A real-time monitoring feedback data stream is established, and a synchronous early warning is sent to the terminal through the Internet of Things gateway.

8. A computer-readable storage medium, characterized in that: Instructions are stored, and when the instructions are run on a computer, the computer is caused to execute a smart agricultural monitoring system based on augmented reality and the Internet of Things as described in any one of claims 1 to 6.

Images