Intelligent monitoring system for rice diseases and insect pests

Abstract

This invention belongs to the field of spectral monitoring technology, specifically an intelligent monitoring system for rice diseases and pests. It includes a ground-based monitoring module, multiple sensing modules such as environmental, wind, sound wave, spore, and soil sensors, an analysis module, and an early warning and control module. The system acquires spectral image data by capturing images of rice using a hyperspectral imager. It utilizes hundreds of bands from invisible light to near-infrared and even short-wave infrared to monitor diseases and pests in real time. Furthermore, it employs a combination of high-altitude monitoring by drones and mobile ground monitoring, enabling real-time monitoring of the status and development of rice diseases and pests in different environments under varying weather conditions. Through ground-based collaboration and threshold judgment, it overcomes the dependence on weather conditions associated with single-drone monitoring, ensuring the continuity of monitoring tasks and the timeliness of data.

Description

Technical Field

[0001] This invention belongs to the field of spectral detection technology, specifically an intelligent monitoring system for rice diseases and pests. Background Technology

[0002] In recent years, remote sensing, Internet of Things (IoT), and artificial intelligence (AI) technologies have been applied in agriculture. For example, drones equipped with multispectral cameras are used to monitor crop growth.

[0003] However, existing technical solutions often suffer from the following shortcomings: monitoring methods are limited, often relying solely on either spectral or image analysis, resulting in limited ability to identify early and concealed characteristics of pests and diseases, and susceptibility to environmental interference; data analysis models are static, unable to adapt to dynamic changes in different regions, growth stages, and environmental conditions, and warning thresholds are often unscientifically set; systems are mostly open-loop monitoring and alarm systems, lacking closed-loop linkage with precision control equipment, failing to achieve full-process intelligent control from perception to execution; they lack comprehensive correlation analysis of field microenvironments (such as soil moisture, nutrients, and spore concentration), making it difficult to provide root-cause control recommendations; and drones are frequently used for large-scale aerial spectral monitoring, which is highly dependent on weather conditions and cannot function properly in windy or rainy weather. This means that if there is prolonged windy or rainy weather, stable collection of spectral information of the rice canopy and stems during that period is impossible, leading to data interruption and an inability to understand the spectral information of pests and diseases under prolonged windy or rainy weather. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention proposes an intelligent monitoring system for rice diseases and pests that can achieve early warning, accurate identification, intelligent prediction, and closed-loop control.

[0005] The technical solution adopted by this invention to solve its technical problem is: the intelligent monitoring system for rice diseases and pests of this invention includes:

[0006] The monitoring module is used to collect multi-channel characteristic spectral band reflectance image data of the rice canopy and high-definition image data of the rice in the monitoring area. The monitoring module includes an unmanned aerial vehicle (UAV) monitoring unit and a ground-based mobile monitoring unit. The UAV monitoring unit is equipped with a hyperspectral imager for large-scale aerial patrol monitoring. The ground-based mobile monitoring unit also includes an unmanned inspection tracked vehicle, a lifting mast, a high-definition camera, and a hyperspectral imager. The lifting mast is mounted on the UAV tracked vehicle. The high-definition camera and hyperspectral imager are fixed to the lifting mast and can rise and fall with it. The high-definition camera can monitor the specific condition of the rice leaves and the types of pests and diseases.

[0007] A wind sensor module, which is used to monitor wind force and direction in real time;

[0008] An environmental sensing module is used to monitor temperature, humidity, light intensity, rainfall, and leaf surface humidity, and to access weather forecast information data.

[0009] The positioning module is used to monitor the position data of the UAV monitoring unit and the ground mobile monitoring unit in real time;

[0010] The data transmission module is used to transmit the data collected by the monitoring module, positioning module, wind sensor module, and environmental sensor module to the analysis module, and at the same time transmit the control commands of the analysis module to the monitoring module.

[0011] The analysis module receives and analyzes atmospheric environmental data, hyperspectral data, high-definition image data, and monitoring module location data. The analysis module includes a first threshold judgment model and a second threshold judgment model, and deploys a deep learning model for pest and disease identification and prediction. The first threshold judgment model makes judgments based on preset meteorological thresholds for UAV flight safety. These meteorological thresholds include at least the maximum permissible wind speed and rainfall intensity thresholds, which are determined based on UAV model performance parameters and historical flight safety data.

[0012] The second threshold judgment model is based on a dynamic risk threshold for suspected disease and pest spectrum. This threshold is obtained by training a historical hyperspectral database using machine learning methods and is used to identify abnormal spectral features from hyperspectral data.

[0013] The early warning module receives the pest and disease identification results and risk prediction information output by the analysis module, generates early warning information and precise prevention and control suggestions, and pushes them to the user terminal.

[0014] Preferably, the system also includes a sound wave monitoring module, which comprises multiple acoustic sensors deployed in the paddy field to collect acoustic signature data of insects in the field. The analysis module dynamically adjusts the judgment threshold of the sound wave monitoring based on real-time data from the wind and environmental sensing modules, establishes a background noise model with wind force and rainfall intensity as input parameters based on historical data, and predicts the background noise spectrum under the current environment. The real-time collected sound wave signal spectrum is compared with the predicted background noise spectrum, and the signal energy value that exceeds a certain proportion of the background noise intensity and conforms to the specific insect acoustic signature frequency band is set as the effective signal threshold. When an effective signal is detected, the insect species and density are analyzed using an acoustic signature recognition algorithm. When the monitoring module initially determines that there is an abnormality in pests and diseases, the sound wave monitoring module provides auxiliary species identification and activates a preset sound wave generator to emit sound waves of a specific frequency to repel, lure, or interfere with pests.

[0015] Preferably, it also includes a spore-catching module, which includes a spore-catching instrument set in the monitoring area for automatically collecting pathogenic spores in the air and performing microscopic imaging; the analysis module uses image recognition technology to count and identify the types of spores and generate early warning information for diseases.

[0016] Preferably, the system also includes a soil sensing module, which is buried in the paddy field soil to monitor soil temperature, moisture content, pH value, and nitrogen, phosphorus, and potassium nutrient content. The analysis module performs correlation analysis between the soil data monitored by the soil sensing module and the identified pest and disease types and levels, generating an assessment report on the correlation between the soil environment and the occurrence of pests and diseases, as well as soil improvement suggestions. The correlation analysis and level classification are as follows: the analysis module has a built-in pest and disease and soil correlation knowledge base and statistical model, and matches the real-time collected soil data vector with the soil condition threshold ranges recorded in the knowledge base that induce or aggravate specific pests and diseases. Based on the degree of matching and the extent to which soil parameters deviate from the healthy range, the correlation is classified into three levels: high correlation, moderate correlation, and weak correlation, and marked in the assessment report.

[0017] Preferably, it also includes a control module, which is connected to field irrigation, fertilization and pesticide application equipment; the analysis module generates precise irrigation, fertilization or pesticide application control commands based on pest and disease identification results, early warning levels, soil data and environmental data, and automatically executes variable operations through the control module.

[0018] Preferably, the deep learning model for pest and disease identification and prediction in the analysis module performs joint training and inference by fusing historical and real-time hyperspectral feature vectors, high-definition image features, environmental data, soil data, and sound wave or spore capture data, and outputs the type, level, probability of occurrence, and spatial distribution map of pests and diseases.

[0019] Preferably, the method for obtaining the dynamic disease and pest spectral suspected risk threshold in the second threshold judgment model is to collect hyperspectral data samples of historical healthy rice and rice infected with different types and levels of diseases and pests, extract their key spectral feature indices and full-band reflectance curve feature vectors; use support vector machines, random forests or deep autoencoders to perform unsupervised or semi-supervised learning on the samples, and construct a dynamic decision boundary in the feature space that can distinguish between normal and abnormal spectral morphologies; the dynamic decision boundary is the dynamic threshold, which can adaptively adjust with the changes in the learned spectral feature distribution, and is used to determine in real time whether there are suspected disease and pest anomalies in the collected hyperspectral data.

[0020] Preferably, the precise prevention and control suggestions generated by the early warning module include: recommended drug name, application dosage, optimal application time window, and recommended operation path for drones or ground equipment.

[0021] Preferably, the data transmission module adopts 5G, LoRa or narrowband IoT technology to achieve low-latency and high-reliability transmission of monitoring data and control commands.

[0022] The beneficial effects of this invention are as follows:

[0023] 1. This invention uses a hyperspectral imager to capture spectral image data of rice, and utilizes hundreds of bands from invisible light to near-infrared and even short-wave infrared to monitor pests and diseases in real time. It employs both aerial monitoring by drones and mobile ground monitoring to acquire spectral images under different conditions, enabling real-time monitoring of pest and disease status and development in rice under different weather conditions. Through the combination of air and ground monitoring and threshold judgment, it overcomes the dependence on weather for single drone monitoring, ensuring the continuity of monitoring tasks and the timeliness of data.

[0024] 2. This invention improves the sensitivity of initial spectral judgment by using a dynamic second threshold. Sound waves and spores provide evidence before disease transmission and in the early stages of pest occurrence. The multimodal fusion model significantly improves the accuracy of the final diagnosis. Through linkage with operational equipment, it achieves on-demand, variable, and targeted control, reduces the amount of pesticides and fertilizers used, improves control effectiveness, and reduces labor costs. Attached Figure Description

[0025] The invention will now be further described with reference to the accompanying drawings.

[0026] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0027] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0028] like Figure 1 As shown, the intelligent monitoring system for rice diseases and pests of the present invention includes:

[0029] The monitoring module is used to collect multi-channel characteristic spectral band reflectance image data of the rice canopy and high-definition image data of the rice in the monitoring area. The monitoring module includes an unmanned aerial vehicle (UAV) monitoring unit and a ground-based mobile monitoring unit. The UAV monitoring unit is equipped with a hyperspectral imager for large-scale aerial patrol monitoring. The ground-based mobile monitoring unit also includes an unmanned inspection tracked vehicle, a lifting mast, a high-definition camera, and a hyperspectral imager. The lifting mast is mounted on the UAV tracked vehicle. The high-definition camera and hyperspectral imager are fixed to the lifting mast and can rise and fall with it. The high-definition camera can monitor the specific condition of the rice leaves and the types of pests and diseases.

[0030] A wind sensor module, which is used to monitor wind force and direction in real time;

[0031] An environmental sensing module is used to monitor temperature, humidity, light intensity, rainfall, and leaf surface humidity, and to access weather forecast information data.

[0032] The positioning module is used to monitor the position data of the UAV monitoring unit and the ground mobile monitoring unit in real time;

[0033] The data transmission module is used to transmit the data collected by the monitoring module, positioning module, wind sensor module, and environmental sensor module to the analysis module, and at the same time transmit the control commands of the analysis module to the monitoring module.

[0034] The analysis module receives and analyzes atmospheric environmental data, hyperspectral data, high-definition image data, and monitoring module location data. The analysis module includes a first threshold judgment model and a second threshold judgment model, and deploys a deep learning model for pest and disease identification and prediction. The first threshold judgment model makes judgments based on preset meteorological thresholds for UAV flight safety. These meteorological thresholds include at least the maximum permissible wind speed and rainfall intensity thresholds, which are determined based on UAV model performance parameters and historical flight safety data.

[0035] The second threshold judgment model is based on a dynamic risk threshold for suspected disease and pest spectrum. This threshold is obtained by training a historical hyperspectral database using machine learning methods and is used to identify abnormal spectral features from hyperspectral data.

[0036] The early warning module receives the pest and disease identification results and risk prediction information output by the analysis module, generates early warning information and precise prevention and control suggestions, and pushes them to the user terminal.

[0037] This invention targets the monitoring of pests and diseases in rice. It primarily relies on a hyperspectral imager to capture spectral images of the rice. The hyperspectral imager records a continuous and detailed spectral curve for each pixel in the image, covering hundreds of bands from visible light to near-infrared and even short-wave infrared. The changes caused by pests and diseases are hidden in the subtle differences in these spectral curves. The spectral characteristic data is in the visible light region (400-700nm): pests and diseases can lead to chlorophyll decomposition (chlorosis), cell structure damage (necrosis), and pigment changes (such as anthocyanin). (Accumulation); This manifests as a decrease in reflectivity in the green light band and an abnormal increase in reflectivity in the red light band (due to reduced chlorophyll absorption); Near-infrared region (700-1300nm): a "sensitive zone" for plant cell structure and water content; Healthy leaves have extremely high near-infrared reflectivity due to their complex spongy internal structure; When diseases or pests occur, the cell structure is damaged, and the near-infrared reflectivity will decrease significantly; Short-wave infrared region (1300-2500nm): mainly related to biochemical components such as water, protein, cellulose, and starch in plants. Pest and disease stress affects the content and structure of these components, thereby altering their reflectance spectra. By analyzing these spectral curves, a large number of indices with clear physical meaning can be calculated for precise diagnosis, such as early warning indices for stress: the photochemical reflectance index (PRI), which is extremely sensitive to changes in leaf photosynthetic efficiency and can detect stress before visible symptoms appear; the red-edge parameter, which indicates a sharp increase in reflectance between 680nm and 50nm in the spectrum of healthy plants; pest and disease stress causes a blue shift (movement towards shorter wavelengths) or a change in the slope of this red-edge, making it an extremely sensitive early indicator; and the spectral curves generated from spectral images monitored by a hyperspectral imager also... Physiological and biochemical contents can be estimated: chlorophyll content: the degree of chlorophyll loss can be monitored through inversion of indices such as NDVI (Normalized Difference Vegetation Index) and TCARI / OSAVI; water content: water loss due to pests, diseases, or stress can be monitored through indices such as NDWI (Normalized Difference Water Index); nitrogen content, carotenoid content, etc.: are closely related to the plant's stress resistance and health status; and disease-specific spectral characteristics: certain diseases (such as rust, powdery mildew, and viral diseases) produce unique spores, mycelia, or cause the accumulation of specific metabolites, forming diagnostic absorption or reflectance characteristic peaks / valleys in specific wavelength bands; through spectral matching or machine learning, the type of disease can be directly identified.

[0038] For acquiring spectral images of rice, a first threshold is set for the maximum permissible wind speed and rainfall intensity corresponding to the flight safety of the drone model used. Wind speed is monitored by a wind sensor module, and rainfall intensity is monitored by an environmental sensor module. If neither the wind speed nor the rainfall intensity currently monitored reaches the first threshold, the drone monitoring unit conducts a large-scale aerial monitoring to obtain spectral images of the rice. These images are then fed back to the analysis module for pest and disease detection. Based on the spectral curves, a preliminary analysis of pest and disease types is performed and fed back to the monitoring module. At this time, the ground-based mobile monitoring unit lowers its boom to its lowest position and moves within the rice paddies, using a high-definition camera to capture images of the rice surface condition and pest types. This feedback can be combined with the spectral information acquired by the drone at high altitude for comparison with the high-definition ground images. The first step is to clearly identify the types and distribution of pests and diseases. If either the wind speed or the rainfall intensity in the current monitoring data reaches the first threshold, the UAV monitoring unit cannot collect spectral information over a large area in the air. The analysis module feeds back to the monitoring module. At this time, the ground mobile monitoring unit raises its lifting arm to the highest point and moves in the rice paddies. It uses a hyperspectral imager to acquire spectral images of the rice. Due to the limited height of the lifting arm, only a small area of ​​spectral images can be acquired. However, the ground mobile monitoring unit can move flexibly, and the positioning module can locate the rice paddies, so it can perform zoned monitoring to obtain the surface condition of the rice and the types of pests. Since zoned monitoring is required, the efficiency is lower than that of UAV high-altitude monitoring. Feeding back the spectral images and location information obtained from the zoned monitoring can further clarify the types and distribution of pests and diseases.

[0039] The environmental sensing module can monitor changes in atmospheric temperature and humidity, light conditions, and leaf surface humidity. When acquiring spectral images, it can understand and record changes in pests and diseases under different atmospheric conditions, and obtain information on the development of pests and diseases under different atmospheric conditions. Adjustments can then be made based on this data.

[0040] This invention uses a hyperspectral imager to capture spectral image data of rice, and utilizes hundreds of bands from invisible light to near-infrared and even short-wave infrared to monitor pests and diseases in real time. It employs both aerial monitoring by drones and mobile ground monitoring in combination, enabling real-time monitoring of pest and disease conditions and their development in rice under different weather conditions. Through the synergy between air and ground monitoring and threshold judgment, it overcomes the dependence on weather for single drone monitoring, ensuring the continuity of monitoring tasks and the timeliness of data.

[0041] As one embodiment of the present invention, it also includes a sound wave monitoring module, which includes multiple acoustic sensors deployed in the paddy field to collect field insect voiceprint data; the analysis module dynamically adjusts the judgment threshold of sound wave monitoring based on real-time data from the wind sensing module and the environmental sensing module, establishes a background noise model with wind force level and rainfall intensity as input parameters based on historical data, and predicts the background noise spectrum under the current environment; compares the real-time collected sound wave signal spectrum with the predicted background noise spectrum, and sets the signal energy value that exceeds a certain proportion of the background noise intensity and conforms to the specific insect voiceprint characteristic frequency band as the effective signal threshold; when an effective signal is detected, the insect species and density are analyzed using a voiceprint recognition algorithm; when the monitoring module initially determines that there is an abnormality in pests and diseases, the sound wave monitoring module provides auxiliary species identification and activates a preset sound wave generator to emit sound waves of a specific frequency to repel, lure, or interfere with pests.

[0042] As one embodiment of the present invention, it also includes a spore-catching module, which includes a spore-catching instrument set in the monitoring area for automatically collecting pathogenic spores in the air and performing microscopic imaging; the analysis module uses image recognition technology to count and identify the types of spores and generate early warning information for diseases.

[0043] As one embodiment of the present invention, it also includes a soil sensing module, which is buried in the paddy field soil to monitor soil temperature, moisture content, pH value, and nitrogen, phosphorus, and potassium nutrient content. The analysis module performs correlation analysis on the soil data monitored by the soil sensing module with the identified pest and disease types and levels, generating an assessment report on the correlation between the soil environment and the occurrence of pests and diseases, as well as soil improvement suggestions. The correlation analysis and level classification are as follows: the analysis module has a built-in pest and disease, soil correlation knowledge base and statistical model, and matches the real-time collected soil data vector with the soil condition threshold ranges recorded in the knowledge base that induce or aggravate specific pests and diseases. Based on the degree of matching and the extent to which soil parameters deviate from the healthy range, the correlation is divided into three levels: "highly correlated", "moderately correlated", and "weakly correlated", and marked in the assessment report. This provides data and direction support for subsequent pest and disease control and regulation of rice.

[0044] As one embodiment of the present invention, it also includes a control module, which is connected to field irrigation, fertilization and pesticide application equipment; the analysis module generates precise irrigation, fertilization or pesticide application control commands based on pest and disease identification results, early warning levels, soil data and environmental data, and automatically executes variable operations through the control module.

[0045] In this invention, the environmental sensing module can monitor atmospheric temperature, humidity, light, and rainfall in real time, and the soil sensing module can monitor soil data in real time. When the monitoring module detects the status and distribution of pests and diseases in rice, it connects to field irrigation, fertilization, and pesticide application equipment through the control module. The analysis module generates precise irrigation, fertilization, or pesticide application control commands based on pest and disease identification results, early warning levels, soil data, and environmental data. The control module automatically executes variable operations, and the monitoring module ultimately detects that the pests and diseases have been controlled. During the process, atmospheric and soil data are recorded and analyzed by the analysis module. This allows for the determination of the optimal atmospheric and soil data for treating specific types of pests and diseases under certain atmospheric or soil conditions, ensuring better treatment efficiency and effectiveness. This analysis and integration ensures that subsequent problem-solving can effectively regulate treatment time and adjust soil data in advance to achieve better pesticide application. Furthermore, because the monitoring module can obtain the distribution and specific location of pests and diseases, the control unit can effectively carry out localized control or zoned treatment using different methods, greatly improving treatment effectiveness and efficiency, and realizing a truly autonomous learning and self-evolving intelligent system.

[0046] As one embodiment of the present invention, the deep learning model for pest and disease identification and prediction in the analysis module performs joint training and inference by fusing historical and real-time hyperspectral feature vectors, high-definition image features, environmental data, soil data, and sound wave or spore capture data, and outputs the type, level, probability of occurrence and spatial distribution map of pests and diseases.

[0047] As one embodiment of the present invention, the method for obtaining the dynamic disease and pest spectral suspected risk threshold in the second threshold judgment model is to collect hyperspectral data samples of historical healthy rice and rice infected with different types and levels of diseases and pests, extract their key spectral feature indices and full-band reflectance curve feature vectors; use support vector machines, random forests or deep autoencoders to perform unsupervised or semi-supervised learning on the samples, and construct a dynamic decision boundary in the feature space that can distinguish between normal and abnormal spectral morphologies; the dynamic decision boundary is the dynamic threshold, which can adaptively adjust with the changes in the learned spectral feature distribution, and is used to determine in real time whether there are suspected disease and pest abnormalities in the collected hyperspectral data.

[0048] As one embodiment of the present invention, the precise prevention and control suggestions generated by the early warning module include: recommended drug name, application dosage, optimal application time window, and recommended operation path for drones or ground equipment.

[0049] As one embodiment of the present invention, the data transmission module adopts 5G, LoRa or narrowband IoT technology to achieve low-latency and high-reliability transmission of monitoring data and control commands.

[0050] The monitoring modules of this invention include: a multi-rotor drone monitoring unit equipped with a push-broom hyperspectral imager (spectral range 400-1000nm); a ground mobile monitoring unit using a wide-tracked unmanned vehicle to adapt to muddy field ridges, with a lifting pole that can be raised to a maximum height of 3 meters, and a multispectral imager (balancing cost and practicality) and a 20-megapixel high-definition zoom camera fixed at the top; a multi-source sensing network: four weather stations (integrating wind, temperature, humidity, rainfall, and light sensors) deployed around the paddy field; six acoustic sensor nodes evenly distributed within the field; an intelligent spore trapping device installed at the upwind end of the field; and soil three-parameter (temperature, humidity, salinity) and NPK sensors buried at representative locations; and an analysis and early warning module: deployed on a cloud server or local edge computing server, responsible for receiving, storing, calculating, and issuing commands for all data. The user terminal is a mobile APP or a computer web interface; the second threshold (dynamic spectral risk threshold) is obtained by collecting hyperspectral data of rice from the past 3-5 growing seasons in this region or similar ecological zones. The data must include healthy samples and samples with different degrees of infection by major pests and diseases such as rice blast, sheath blight, and rice planthopper; all samples have field validation labels; for each sample data, a series of vegetation indices sensitive to pests and diseases (such as NDVI, PRI, red edge parameters, etc.) are calculated, and reflectance values ​​of characteristic band ranges are selected to form a feature vector; a support vector machine or deep autoencoder algorithm is used for training. Healthy samples are used as the "positive class" to allow the model to learn the normal distribution range of spectral characteristics of healthy rice in multivariate space; the training goal is to find a boundary that can surround most of the healthy sample data (i.e., decision hyperplane or reconstruction error threshold); after training, this boundary is the dynamic "suspected risk threshold"; real-time flight-collected hyperspectral data, after the same feature extraction, is input into the model; if the sample point falls within the boundary, it is judged as "normal"; if it falls outside the boundary, it is judged as a "spectral anomaly suspected area". This threshold will be periodically retrained as new labeled health and disease data are continuously added to the database, achieving adaptive updates.

[0051] A dynamic adjustment method for sound wave thresholds is proposed. First, a noise baseline database is established: During the experimental phase, typical environmental noise spectra under different wind speeds (levels 1-5) and rainfall intensities (no rain, light rain, moderate rain) are recorded, and environmental parameters, noise spectrum lookup tables, or regression models are established. Second, real-time prediction and threshold setting: During system operation, the predicted current background noise spectrum B(f) is matched or calculated from the baseline database based on real-time data from the wind and environmental sensing modules. Then, signal extraction is performed: The raw spectrum acquired in real-time by the acoustic sensors is R(f); the system calculates the difference spectrum D(f) = R(f) - α * B(f), where α is a safety factor (e.g., 1.2) to ensure sufficient noise deduction. Typical acoustic signature frequency bands F of target pests (e.g., rice stem borer, rice planthopper) are pre-stored in the database. pest Final determination: Calculate frequency band F pest The D(f) energy and E within pest Set a minimum effective signal energy threshold E. min If E pest >E min If the signal is detected, it is considered a valid insect sound signal, and then species identification is performed. This method effectively suppresses environmental noise on windy and rainy days, reducing the false alarm rate.

[0052] The correlation analysis and classification of soil data with pests and diseases are based on agricultural expert knowledge and historical data statistics to construct a rule base. For example, long-term deep irrigation (high soil moisture saturation) is highly positively correlated with the occurrence of sheath blight; excessive nitrogen fertilizer (abnormally high soil nitrogen content) easily leads to the occurrence of rice blast and the increase of rice planthopper population. When the analysis module determines that a specific pest or disease (such as rice blast) has occurred in a certain area, it immediately retrieves the recent data of soil sensors in that area (such as the average nitrogen content, pH value, and moisture in the past week). The classification is divided into three categories, which are highly correlated: the soil parameter values ​​continuously and significantly exceed the threshold range for the occurrence of the pest or disease (such as nitrogen content exceeding the upper limit of the healthy range by 30%), and completely match the dominant inducing factors in the knowledge base.

[0053] Moderate correlation: Soil parameter values ​​fluctuate around the susceptibility threshold range, or only some parameters match the triggers in the knowledge base; Weak correlation: Soil parameter values ​​are basically within the healthy range and have no clear statistical relationship with the current occurrence of pests and diseases; The generated pest and disease diagnosis report not only includes pest and disease information, but also adds soil correlation analysis, clearly pointing out the soil factors that may aggravate the disease and their correlation level, and giving specific agronomic measures such as controlling nitrogen fertilizer and timely field drying.

[0054] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent monitoring system for rice diseases and pests, characterized in that, include: The monitoring module is used to acquire multispectral reflectance images and high-resolution images of the rice canopy in the monitoring area; It includes an unmanned aerial vehicle (UAV) monitoring unit and a ground mobile monitoring unit; the UAV monitoring unit is equipped with a hyperspectral imager; the ground mobile monitoring unit includes an unmanned inspection tracked vehicle, a lifting rod installed on the unmanned inspection tracked vehicle, and a high-definition camera and a hyperspectral imager fixed on the lifting rod. Wind sensor module for real-time monitoring of wind speed and direction; The environmental sensing module is used to monitor temperature, humidity, light, rainfall, and leaf surface humidity data, and to connect to weather forecast information; A positioning module is used to acquire the real-time location data of the monitoring module; The data transmission module is used to transmit monitoring data and control commands; The analysis module is used to receive and process data from the monitoring module, wind sensor module, environmental sensor module, and positioning module; it has a first threshold judgment model and a second threshold judgment model preset within it, and deploys a deep learning model for pest and disease identification and prediction; the first threshold judgment model is based on preset wind speed and rainfall intensity thresholds for safe drone flight; the second threshold judgment model is based on dynamic pest and disease spectral suspected risk thresholds obtained through machine learning of historical hyperspectral data, and is used to identify spectral anomalies. If the current monitoring data for wind speed and rainfall intensity does not reach the first threshold, the UAV monitoring unit conducts large-scale monitoring in the air, obtaining spectral images of the rice. This data is then fed back to the analysis module for pest and disease analysis. Based on the spectral curves, the module performs preliminary analysis of pest and disease types and feeds this information back to the monitoring module. At this time, the ground-based mobile monitoring unit lowers its boom to its lowest position and moves within the rice paddies, using a high-definition camera to capture images of the rice's surface condition and pest types. This feedback can be combined with the spectral information collected by the UAV at high altitude and the high-definition ground images to further clarify the types and distribution of pests and diseases. If either the current monitoring data for wind speed or rainfall intensity reaches the first threshold, the UAV monitoring unit cannot collect spectral information over a large area in the air. The analysis module feeds this information back to the monitoring module, and the ground-based mobile monitoring unit raises its boom to its highest position and moves within the rice paddies. The early warning module is used to receive the pest and disease identification and prediction results output by the analysis module, generate and push early warning information and prevention and control suggestions.

2. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, It also includes a sound wave monitoring module, which includes multiple acoustic sensors deployed in the paddy field to collect acoustic data of insects in the field; the analysis module dynamically adjusts the judgment threshold of sound wave monitoring based on the real-time data of the wind sensing module and the environmental sensing module.

3. The intelligent monitoring system for rice diseases and pests according to claim 1 or 2, characterized in that, It also includes a spore-catching module, which includes a spore-catching instrument set in the monitoring area for automatically collecting pathogenic spores in the air and performing microscopic imaging; the analysis module uses image recognition technology to count and identify the types of spores and generate early warning information for diseases.

4. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, It also includes a soil sensing module, which is buried in the paddy field soil to monitor soil temperature, moisture content, pH value, and nitrogen, phosphorus, and potassium nutrient content; the analysis module performs correlation analysis between the soil data monitored by the soil sensing module and the identified pest and disease types and levels, and generates an assessment report on the correlation between soil environment and pest and disease occurrence and soil improvement suggestions.

5. The intelligent monitoring system for rice diseases and pests according to claim 1 or 4, characterized in that, It also includes a control module, which is connected to field irrigation, fertilization and pesticide application equipment; the analysis module generates precise irrigation, fertilization or pesticide application control commands based on pest and disease identification results, early warning levels, soil data and environmental data, and automatically executes variable operations through the control module.

6. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, The deep learning model for pest and disease identification and prediction in the analysis module integrates historical and real-time hyperspectral feature vectors, high-definition image features, environmental data, soil data, and sound wave or spore capture data for joint training and inference, and outputs the type, level, probability of occurrence, and spatial distribution map of pests and diseases.

7. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, The method for obtaining the dynamic disease and pest spectral suspected risk threshold in the second threshold judgment model is to collect hyperspectral data samples of historical healthy rice and rice infected with different types and levels of diseases and pests, extract their key spectral feature indices and full-band reflectance curve feature vectors; use support vector machines, random forests or deep autoencoders to perform unsupervised or semi-supervised learning on the samples, and construct a dynamic decision boundary in the feature space that can distinguish between normal and abnormal spectral morphologies; the dynamic decision boundary is the dynamic threshold, which can adaptively adjust with the changes in the learned spectral feature distribution, and is used to determine in real time whether there are suspected disease and pest anomalies in the collected hyperspectral data.

8. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, The precise prevention and control recommendations generated by the early warning module include: recommended drug names, application dosages, optimal application time windows, and recommended operating paths for drones or ground equipment.

9. The intelligent monitoring system for rice diseases and pests according to claim 1, characterized in that, The data transmission module uses 5G, LoRa, or narrowband IoT technology to achieve low-latency and high-reliability transmission of monitoring data and control commands.

Images