Intelligent disease and pest monitoring and prevention system

Abstract

The invention relates to the technical field of intelligent agriculture, in particular to an intelligent disease and pest monitoring and prevention system which comprises an environment self-adaption module, a spectrum collection module, a dynamic threshold value analysis module, a prevention decision module, an execution terminal and a data storage module. A polarization filter is adopted to suppress reflection of light on the blade surface to obtain a spectrum fusion image; the dynamic threshold value analysis module is used for establishing a dynamic proportion threshold value of the insect spot area and the medicinal material biomass by adopting a U-Net network in combination with a growth cycle recognition algorithm so as to obtain graded insect pest evaluation parameters; and the prevention and control decision module is used for combining the graded insect pest evaluation parameters with the environmental parameter set and generating a prevention and control scheme by adopting a pesticide application decision tree model based on northern climate characteristics. According to the invention, through dynamic environment adaptive adjustment and northern climate optimization pesticide application decision, efficient and accurate prevention and control of diseases and insect pests in a complex agricultural environment are realized.

Description

Technical Field

[0001] The present invention relates to the field of smart agricultural technology, and in particular to an intelligent pest monitoring and control system. Background Art

[0002] Pest and disease monitoring and control are core components of ensuring the healthy growth of crops, especially in the cultivation of medicinal plants. Pests and diseases not only lead to yield losses but can also alter the content of active ingredients in medicinal materials, directly impacting drug quality and safety. Traditional manual inspections are inefficient and have limited coverage, making it difficult to detect pests and diseases in their early stages. Over-reliance on chemical control can easily lead to excessive pesticide residues and environmental pollution. By introducing intelligent monitoring systems, real-time dynamic perception of medicinal herb planting areas can be achieved. Combining environmental data with precise analysis of pest and disease characteristics allows for the development of differentiated pesticide application strategies, reducing chemical use by 30% to 50%, while improving pest and disease identification accuracy and the timeliness of control efforts.

[0003] Chinese Patent Publication No. CN102541030B proposes an intelligent crop pest and disease monitoring and control system, comprising a monitoring computer and a pesticide spraying device located at a crop cultivation site. The monitoring computer drives the connected pesticide spraying device. The monitoring computer is equipped with a time controller and a spray selection driver module. The time controller, through the spray selection driver module, activates the corresponding nozzles of the pesticide spraying device. The beneficial effects of this invention include: first, intelligent operation, a high degree of automation, diverse control methods, timely control, and effective operation. Second, the monitoring device utilizes solar energy as a power source, achieving significant energy savings. Third, the monitoring device not only automates water supply and pesticide spraying for cultivated crops but also provides multi-faceted control capabilities for crop pest and disease control in different areas of the same site. Fourth, the monitoring probe monitors crop health in real time, enabling timely early warning and targeted treatment for crop pests and diseases. This not only helps ensure crop growth and a good harvest in agriculture, forestry, and tea production, but also significantly reduces pesticide waste and pesticide pollution.

[0004] Although the invention has a complete process for monitoring and controlling crop pests and diseases, it still has certain limitations. Plants are subject to dynamic interference factors in the actual planting environment, such as light intensity fluctuations, heavy rain or snowfall, especially in the winter-spring transition season in the north. The low light intensity and snowfall will cause the image quality of plants planted during this period to degrade, and the morphological characteristics of new invasive pests exceed the range of the training set, which will cause the model detection accuracy to drop significantly. Secondly, the existing technology relies on the wormhole edge shape matching algorithm to identify the type of pests and diseases, but does not fully consider the impact of variables such as leaf surface reflection and camera resolution differences on the image segmentation effect under low light intensity and snowfall conditions, resulting in feature extraction distortion. These defects highlight the shortcomings of the existing technology in terms of adaptability to dynamic environments and adaptability to medicinal material characteristics. There is an urgent need to propose an intelligent pest and disease monitoring and control system. Summary of the Invention

[0005] To solve the above problems, the present invention provides an intelligent pest and disease monitoring and control system, which realizes efficient and accurate prevention and control of pests and diseases in complex agricultural environments through dynamic environmental adaptive adjustment and northern climate-optimized pesticide application decisions.

[0006] To achieve the above-mentioned objectives, the technical solution of the present invention is as follows: an intelligent pest monitoring and control system, comprising an environment adaptation module, a spectrum acquisition module, a dynamic threshold analysis module, a control decision module, an execution terminal, and a data storage module, which are interconnected by signals. The environment adaptation module is signal-connected to a plurality of sensor units deployed in a farmland environment, and the execution terminal includes a plurality of pesticide application units deployed in the farmland environment for spraying chemical pesticides, wherein:

[0007] The environmental adaptation module is used to collect light intensity, ambient temperature, ambient humidity and meteorological data in real time through the sensing unit, adjust the exposure parameters and white balance of the monitoring probe through a dynamic calibration algorithm, obtain the compensated environmental parameter set, and transmit it to the spectrum acquisition module, prevention and control decision module and data storage module respectively;

[0008] The spectrum acquisition module is used to alternately illuminate the target medicinal material with an annular array of 850nm near-infrared light sources and a visible light source with a wavelength of 380-780nm based on the compensated environmental parameter set. The module receives the leaf surface reflectivity transmitted by the sensing unit, uses a polarization filter to suppress the leaf surface reflection, obtains a spectral fusion image, and transmits it to the dynamic threshold analysis module and the data storage module.

[0009] The dynamic threshold analysis module is used to receive the spectral fusion image and the medicinal material image information transmitted by the sensor unit. It uses the improved U-Net network combined with the growth cycle recognition algorithm to establish a dynamic ratio threshold between the insect spot area and the medicinal material biomass. It obtains the graded pest assessment parameters including the growth stage identification parameters, the graded damage index and the insect spot area data, and transmits them to the prevention and control decision module and the data storage module.

[0010] The pest control decision-making module combines graded pest assessment parameters with environmental parameter sets and uses a pesticide application decision tree model based on northern climate characteristics to generate a pest control plan that includes pesticide types adapted to the characteristics of the medicinal materials, atomized particle sizes suitable for adverse environments, and a pesticide application sequence optimized for temperature differences. The plan is then transmitted to the execution terminal.

[0011] The execution terminal is used to receive the control plan, and use the command signal of the control plan to make each spraying unit spray the chemical medicine on the pest area, and record the spraying data and transmit it to the data storage module;

[0012] The data storage module is used to receive and associate storage of environmental parameter sets, spectral fusion images, graded pest assessment parameters and pesticide application data. It adopts a spatiotemporal coupling incremental learning mechanism to generate feedback data to the environmental adaptation module, dynamic threshold analysis module and prevention and control decision module to update the preset processing parameters and optimize the prevention and control plan. It is also used to dynamically match environmental fluctuation parameters with changes in pest characteristics, and generate a feature comparison library with northern climate adaptation labels for storage.

[0013] Furthermore, the environment adaptation module includes an illumination compensation unit, a weather change response unit, a night imaging unit, and an integration unit, wherein:

[0014] The light compensation unit is used to collect the light intensity in all directions within the farmland planting area through the sensor unit, adjust the camera gain parameters using the adaptive white balance algorithm, obtain a balanced light distribution map, and transmit it to the integration unit;

[0015] The weather change response unit is used to receive meteorological data from the sensing unit, extract precipitation particle distribution data, pre-process the image using a dark channel prior defogging algorithm, obtain the visibility correction coefficient in snowfall or rainstorm scenes, and transmit it to the integration unit;

[0016] The night imaging unit is used to obtain a number of path data transmitted by the sensing unit. When the light intensity is lower than 50 lux, the thermal imaging function of the sensing unit is activated. The edge enhancement algorithm is used to extract the wormhole outline, obtain the wormhole spot morphology data under the light intensity lower than 50 lux, and transmit it to the integration unit.

[0017] The integration unit is used to receive the balanced light distribution map, visibility correction coefficient and insect spot morphology data, bind them with the acquisition timestamp through the spatiotemporal label fusion algorithm, generate the compensated environmental parameter set, and synchronously transmit it to the spectral acquisition module, prevention and control decision module and data storage module.

[0018] Furthermore, the night imaging unit includes a thermal radiation calibration subunit and a multispectral registration subunit, wherein:

[0019] The thermal radiation calibration subunit is used to correct the thermal imaging data using the Stefan-Boltzmann law based on the relationship between the ambient temperature and the blade emissivity transmitted by the sensing unit, obtain an infrared image with a temperature resolution of 0.1°C, and transmit it to the multispectral registration subunit;

[0020] The multispectral registration subunit is used to match the feature points of the infrared image with the visible light with a wavelength of 380 to 780 nm collected by the sensing unit. The SIFT algorithm is used to achieve cross-modal image alignment: the leaf vein feature points enhanced by Gabor filtering are extracted in the visible light image, and the wormhole edge feature points with a temperature gradient of >2°C / cm are located in the infrared image. Finally, fused data with a spatial error of <2 pixels is obtained, and the fused data is transmitted to the integration unit as the worm spot morphology data under light intensity less than 50 lux.

[0021] Furthermore, the spectrum acquisition module includes an active light source control unit, a filter switching unit and an image fusion unit, wherein:

[0022] The active light source control unit is used to receive the environmental parameter set and adjust the wavelength of the near-infrared light source according to the ambient humidity value transmitted by the sensor unit. When the humidity is greater than 80%, it switches to the 950nm band to penetrate the water mist. It uses pulse modulation technology to eliminate motion blur, obtains leaf back image data that matches the environmental parameter set, and transmits it to the image fusion unit.

[0023] A filter switching unit is used to select the optimal filter combination using a rotating filter group according to the blade surface reflectivity transmitted by the spectrometer in the sensing unit, obtain a polarized image that suppresses specular reflection, and transmit it to the image fusion unit;

[0024] The image fusion unit is used to perform wavelet transform fusion on the leaf back image data and the polarization image, and adopts the regional energy weighted algorithm to enhance the wormhole edge contrast, obtain a spectral fusion image with a resolution of not less than 20 million pixels, and transmit it to the dynamic threshold analysis module and data storage module.

[0025] Furthermore, the image fusion unit includes a shadow removal subunit and a multi-scale enhancement subunit, wherein:

[0026] The shadow removal subunit is used to fuse the leaf back image data with the polarization image and detect and mark the occluded areas. The occluded areas include pixel missing areas caused by leaf overlap, external equipment occlusion, or low-light reflection. The occluded areas are defined as continuous pixel blocks with grayscale values ​​lower than 30% of the background mean in the visible light band and a sudden drop in reflectance greater than 15% in the near-infrared band. The leaf surface texture in the occluded area is reconstructed using the Poisson equation to obtain the spectral intermediate image with the occlusion removed, and the image is transmitted to the multi-scale enhancement subunit.

[0027] The multi-scale enhancer unit is used to receive the spectral intermediate image and use the Laplace pyramid fusion algorithm in the scale range of 0.5 to 2 mm to separate the leaf artifacts caused by snow cover in the northern region from the real wormholes in the frequency domain, enhance the gradient response intensity of the wormhole edge to more than twice the background value, generate a spectral fusion image, and transmit it to the dynamic threshold analysis module and data storage module.

[0028] Furthermore, the dynamic threshold analysis module includes a growth cycle identification unit, a hazard coefficient calculation unit, and an insect spot area correction unit, wherein:

[0029] The growth cycle recognition unit is used to receive the medicinal material image information transmitted by the sensing unit, extract the plant height and stem thickness characteristics of the medicinal material through a convolutional neural network, and analyze the reflectance data in the 400-1000nm band based on the information of the spectral fusion image. The improved ResNet50 network is used to distinguish between the seedling stage and the mature stage. The growth cycle recognition unit includes a stress response analysis subunit and a root development monitoring subunit. It is also used to receive the flavonoid substance correlation degree transmitted by the stress response analysis subunit and the underground pest impact degree transmitted by the root development monitoring subunit. The medicinal material growth rate compensation coefficient is combined with the characteristics of the short growth cycle and large day and night temperature difference of the medicinal material in the northern climate to obtain the growth stage identification parameter of the medicinal material and transmit it to the hazard coefficient calculation unit.

[0030] The hazard coefficient calculation unit is used to adjust the insect spot threshold according to the growth stage identification parameters of the medicinal materials. When the medicinal materials are in the seedling stage, a dynamic interval threshold of 5% to 15% of the biomass proportion is adopted. When the medicinal materials are in the mature stage, a progressive threshold of 15% to 30% is adopted to obtain a graded hazard index and transmit it to the insect spot area correction unit;

[0031] The insect spot area correction unit is used to combine the leaf vein distribution characteristics in the spectral fusion image and use morphological opening operation to eliminate the artifact interference caused by leaf surface reflection or low light, obtaining an accuracy of 0.1mm. 2 The insect spot area data is integrated into the graded pest assessment parameters, which are then transmitted to the prevention and control decision module and the data storage module.

[0032] Furthermore, the growth cycle identification unit includes a stress response analysis subunit and a root development monitoring subunit, wherein:

[0033] The stress response analysis subunit is used to receive the 500-600mm band data collected by the sensor unit, combine it with the convolutional neural network to extract the characteristics of the medicinal material plant height and stem diameter, and detect the accumulation characteristics of secondary metabolites in the medicinal material. When the first-order derivative peak at 530nm is detected to drop by more than 15%, it is determined to be a metabolic abnormality caused by pest and disease stress. The correlation degree of flavonoid substances is generated and uploaded to the growth cycle recognition unit;

[0034] The root development monitoring subunit is used to obtain the three-dimensional structure of the root system through the sensing unit and reconstruct the root density distribution. When a root cavity with a diameter greater than 2mm and a soil moisture content greater than 20% is detected, it is determined to be an insect pest infestation event, and the underground insect pest impact degree is generated and uploaded to the growth cycle identification unit.

[0035] Furthermore, the prevention and control decision module includes a time sequence planning unit, a pesticide decision unit, and an abnormal warning unit, among which:

[0036] The timing planning unit receives the environmental parameter set and obtains the day-night temperature difference data based on the temperature marked with the timestamp. When the day-night temperature difference is ≥10°C, a thermodynamic model is used to calculate the dew point condensation time on the leaf surface. This generates a pesticide application schedule that avoids the peak morning dew period of 6:00 to 8:00 a.m. and transmits it to the pesticide decision-making unit.

[0037] The pesticide decision-making unit receives graded pest assessment parameters and extracts growth stage identification parameters, as well as real-time temperature data from the environmental parameter set. It uses a fuzzy logic algorithm to select botanical or chemical pesticides, and generates a control plan for the pesticide type and atomization parameters adapted to the current temperature, as well as the application path, to the execution terminal. When the temperature is less than 5°C, an instruction containing an antifreeze-type atomizing liquid is added to the control plan.

[0038] The abnormal warning unit is used to receive spectral fusion images and extract wormhole image information, detect the morphological characteristics of pests, and use the twin network to compare the pest model with the northern climate adaptation label in the feature comparison library. When the morphological similarity is <70% and the insect spot area is ≥5mm 2 When a new species invades, an early warning plan including the coordinates of the isolation area and emergency agents is generated and transmitted to the user terminal, and a new early warning pest sample is generated and stored in the data storage module.

[0039] Furthermore, the dosage decision unit includes an atomization parameter optimization subunit and an environment adaptation subunit, wherein:

[0040] The atomization parameter optimization subunit is used to extract the polarization image of the spectral fusion image, calculate the contact angle data of the leaf surface based on the polarization image, use the fluid dynamics model to calculate the adhesion efficiency of the droplets on the waxy leaf surface with a contact angle greater than 110°, generate the atomization parameters of the particle size of 30 to 50 μm, and transmit them to the drug decision-making unit;

[0041] The environmental adaptation subunit is used to extract the meteorological data collected from the environmental parameters to obtain the inversion layer height and wind speed data. When the inversion layer height is <100m and the wind speed is >3m / s, the Gaussian diffusion model is used to calculate the horizontal drift distance of the agent, generate the application path 0.5 to 1.2m away from the medicinal material canopy, and transmit it to the agent decision unit.

[0042] Furthermore, the data storage module includes an insect pest signature library update unit and an incremental learning unit, wherein:

[0043] The pest feature library update unit receives new warning pest samples and performs three-dimensional modeling. It uses a generative adversarial network to generate a 200-degree rotational view with the main axis of the insect body as the rotation axis in the visible light and near-infrared dual bands, obtains a pest model with a northern climate label, and transmits it to the incremental learning unit.

[0044] The incremental learning unit is used to receive pest models and pesticide application data. Without restarting the system, it uses an elastic weight solidification algorithm to calculate the Fisher information matrix of the existing pest recognition model weights, retains the parameters ranked in the top 30% in terms of weight importance, dynamically updates the feature extraction layer weights in the U-Net network, generates a dynamically optimized pest and disease recognition model, and feeds back the white balance correction coefficient to the environmental adaptation module in real time. It calibrates the light source wavelength according to the near-infrared reflectivity characteristics of the newly added pest model; pushes the model parameters containing the new species recognition threshold and the incremental update of the growth stage compensation coefficient to the dynamic threshold analysis module; and outputs the optimal pesticide application strategy set for historically similar environments to the prevention and control decision module, screening the pesticide-environment combination scheme with a prevention and control success rate ≥90% in historical data.

[0045] The above scheme has the following beneficial effects:

[0046] This solution addresses image analysis distortion caused by light fluctuations and sudden weather changes. The environmental adaptation module's multispectral fusion and real-time calibration technology significantly improves insect spot edge recognition accuracy in low-light, rain, fog, and snowy conditions. The spectral acquisition module's polarization filtering and shadow elimination mechanisms effectively suppress leaf surface reflections and occlusion interference, ensuring high-fidelity extraction of wormhole morphological features. The dynamic threshold analysis module, combining herb growth stages and biomass changes, establishes differentiated insect spot damage assessment thresholds, addressing misjudgments of insect damage severity during seedling and mature stages and preventing over- or under-application of pesticides.

[0047] 2. This solution addresses the short growth cycles and widely varying environmental tolerances of medicinal herbs. The Dynamic Threshold Analysis module, through secondary metabolite correlation detection and root pest impact assessment, constructs a composite judgment model that considers both above- and below-ground pests. This model accurately quantifies the correlation between active ingredient loss and pest infestation. The Prevention and Control Decision-Making module optimizes application timing and atomization parameters based on northern climate characteristics (such as diurnal temperature differences and the inversion layer effect). It also dynamically adjusts the pesticide attachment strategy based on the waxy leaf characteristics of the medicinal herbs, significantly improving solution utilization and targeted control.

[0048] 3. This solution addresses the lag in sudden pest identification caused by static training models in traditional systems. The data storage module leverages incremental learning and a northern climate-labeled feature library to rapidly model new species morphological characteristics and dynamically update the identification model. This ensures the system maintains high-precision detection capabilities even in complex environments such as snowfall reflections and frozen soil disturbances. The anomaly warning module compares multi-dimensional features with historical strategies to generate emergency isolation and targeted pesticide application plans, blocking the spread of pests.

[0049] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a schematic diagram of the system framework of an embodiment of the intelligent pest monitoring and control system of the present invention;

[0051] Figure 2 This is a schematic diagram of the framework of the environment adaptation module of an embodiment of the intelligent pest monitoring and control system of the present invention;

[0052] Figure 3 This is a schematic diagram of the spectrum acquisition module framework of an embodiment of the intelligent pest monitoring and control system of the present invention;

[0053] Figure 4 This is a schematic diagram of the framework of a dynamic threshold analysis module of an embodiment of the intelligent pest monitoring and control system of the present invention;

[0054] Figure 5 This is a schematic diagram of the prevention and control decision module framework of an embodiment of the intelligent pest monitoring and control system of the present invention;

[0055] Figure 6 This is a schematic diagram of the data storage module framework of an embodiment of the intelligent pest monitoring and control system of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0058] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0059] The following is further described in detail through specific implementation methods:

[0060] Example 1:

[0061] As attached Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 and Figure 6 As shown: An intelligent pest monitoring and control system includes an environment adaptation module, a spectrum acquisition module, a dynamic threshold analysis module, a control decision module, an execution terminal and a data storage module that are mutually signal-connected. The environment adaptation module is signal-connected to a number of sensor units deployed in the farmland environment. The execution terminal includes a number of pesticide application units deployed in the farmland environment for spraying chemical pesticides, wherein:

[0062] The environmental adaptation module is used to collect light intensity, ambient temperature, ambient humidity and meteorological data in real time through the sensing unit, adjust the exposure parameters and white balance of the monitoring probe through a dynamic calibration algorithm, obtain the compensated environmental parameter set, and transmit it to the spectrum acquisition module, prevention and control decision module and data storage module respectively;

[0063] The environmental adaptation module includes an illumination compensation unit, a weather change response unit, a night imaging unit, and an integration unit, among which:

[0064] The light compensation unit is used to collect the light intensity in all directions within the farmland planting area through the sensor unit, adjust the camera gain parameters using the adaptive white balance algorithm, obtain a balanced light distribution map, and transmit it to the integration unit;

[0065] The weather change response unit is used to receive meteorological data from the sensing unit, extract precipitation particle distribution data, pre-process the image using a dark channel prior defogging algorithm, obtain the visibility correction coefficient in snowfall or rainstorm scenes, and transmit it to the integration unit;

[0066] The night imaging unit is used to obtain a number of path data transmitted by the sensing unit. When the light intensity is lower than 50 lux, the thermal imaging function of the sensing unit is activated. The edge enhancement algorithm is used to extract the wormhole outline, obtain the wormhole spot morphology data under the light intensity lower than 50 lux, and transmit it to the integration unit.

[0067] The night imaging unit includes a thermal radiation calibration subunit and a multispectral registration subunit, wherein:

[0068] The thermal radiation calibration subunit is used to correct the thermal imaging data using the Stefan-Boltzmann law based on the relationship between the ambient temperature and the blade emissivity transmitted by the sensing unit, obtain an infrared image with a temperature resolution of 0.1°C, and transmit it to the multispectral registration subunit;

[0069] The multispectral registration subunit is used to match feature points between the infrared image and visible light captured by the sensor unit with a wavelength of 380 to 780 nm. The SIFT algorithm is used to achieve cross-modal image alignment. The subunit extracts Gabor filter-enhanced leaf vein feature points from the visible light image and locates wormhole edge feature points with a temperature gradient greater than 2°C / cm in the infrared image. This fused data is then transmitted to the integration unit as wormhole spot morphology data under illumination intensity less than 50 lux.

[0070] The integration unit is used to receive the balanced light distribution map, visibility correction coefficient and insect spot morphology data, bind them with the acquisition timestamp through the spatiotemporal label fusion algorithm, generate the compensated environmental parameter set, and synchronously transmit it to the spectral acquisition module, prevention and control decision module and data storage module.

[0071] The spectrum acquisition module is used to alternately illuminate the target medicinal material with an annular array of 850nm near-infrared light sources and a visible light source with a wavelength of 380-780nm based on the compensated environmental parameter set. The module receives the leaf surface reflectivity transmitted by the sensing unit, uses a polarization filter to suppress the leaf surface reflection, obtains a spectral fusion image, and transmits it to the dynamic threshold analysis module and the data storage module.

[0072] The spectrum acquisition module includes an active light source control unit, a filter switching unit, and an image fusion unit, wherein:

[0073] The active light source control unit is used to receive the environmental parameter set and adjust the wavelength of the near-infrared light source according to the ambient humidity value transmitted by the sensor unit. When the humidity is greater than 80%, it switches to the 950nm band to penetrate the water mist. It uses pulse modulation technology to eliminate motion blur, obtains leaf back image data that matches the environmental parameter set, and transmits it to the image fusion unit.

[0074] A filter switching unit is used to select the optimal filter combination using a rotating filter group according to the blade surface reflectivity transmitted by the spectrometer in the sensing unit, obtain a polarized image that suppresses specular reflection, and transmit it to the image fusion unit;

[0075] The image fusion unit is used to perform wavelet transformation on the leaf back image data and the polarization image, and use the regional energy weighted algorithm to enhance the wormhole edge contrast to obtain a spectral fusion image with a resolution of not less than 20 million pixels, and transmit it to the dynamic threshold analysis module and data storage module;

[0076] The image fusion unit includes a shadow removal subunit and a multi-scale enhancement subunit, wherein:

[0077] The shadow removal subunit is used to fuse the leaf back image data with the polarization image and detect and mark the occluded areas. The occluded areas include pixel missing areas caused by leaf overlap, external equipment occlusion, or low-light reflection. The occluded areas are defined as continuous pixel blocks with grayscale values ​​lower than 30% of the background mean in the visible light band and a sudden drop in reflectance greater than 15% in the near-infrared band. The leaf surface texture in the occluded area is reconstructed using the Poisson equation to obtain the spectral intermediate image with the occlusion removed, and the image is transmitted to the multi-scale enhancement subunit.

[0078] The multi-scale enhancer unit is used to receive the spectral intermediate image and use the Laplace pyramid fusion algorithm in the scale range of 0.5 to 2 mm to separate the leaf artifacts caused by snow cover in the northern region from the real wormholes in the frequency domain, enhance the gradient response intensity of the wormhole edge to more than twice the background value, generate a spectral fusion image, and transmit it to the dynamic threshold analysis module and data storage module.

[0079] The dynamic threshold analysis module is used to receive the spectral fusion image and the medicinal material image information transmitted by the sensor unit. It uses the improved U-Net network combined with the growth cycle recognition algorithm to establish a dynamic ratio threshold between the insect spot area and the medicinal material biomass. It obtains the graded pest assessment parameters including the growth stage identification parameters, the graded damage index and the insect spot area data, and transmits them to the prevention and control decision module and the data storage module.

[0080] The dynamic threshold analysis module includes a growth cycle identification unit, a hazard coefficient calculation unit, and an insect spot area correction unit, among which:

[0081] The growth cycle recognition unit is used to receive the medicinal material image information transmitted by the sensing unit, extract the plant height and stem thickness characteristics of the medicinal material through a convolutional neural network, and analyze the reflectance data in the 400-1000nm band based on the information of the spectral fusion image. The improved ResNet50 network is used to distinguish between the seedling stage and the mature stage. The growth cycle recognition unit includes a stress response analysis subunit and a root development monitoring subunit. It is also used to receive the flavonoid substance correlation degree transmitted by the stress response analysis subunit and the underground pest impact degree transmitted by the root development monitoring subunit. The medicinal material growth rate compensation coefficient is combined with the characteristics of the short growth cycle and large day and night temperature difference of the medicinal material in the northern climate to obtain the growth stage identification parameter of the medicinal material and transmit it to the hazard coefficient calculation unit.

[0082] The growth cycle identification unit includes a stress response analysis subunit and a root development monitoring subunit, wherein:

[0083] The stress response analysis subunit is used to receive the 500-600mm band data collected by the sensor unit, combine it with the convolutional neural network to extract the characteristics of the medicinal material plant height and stem diameter, and detect the accumulation characteristics of secondary metabolites in the medicinal material. When the first-order derivative peak at 530nm is detected to drop by more than 15%, it is determined to be a metabolic abnormality caused by pest and disease stress. The correlation degree of flavonoid substances is generated and uploaded to the growth cycle recognition unit;

[0084] The root development monitoring subunit is used to obtain the three-dimensional structure of the root system through the sensor unit and reconstruct the root density distribution. When a root cavity with a diameter greater than 2mm and a soil moisture content greater than 20% is detected, it is determined to be an insect infestation event, and the underground insect pest impact degree is generated and uploaded to the growth cycle recognition unit;

[0085] The hazard coefficient calculation unit is used to adjust the insect spot threshold according to the growth stage identification parameters of the medicinal materials. When the medicinal materials are in the seedling stage, a dynamic interval threshold of 5% to 15% of the biomass proportion is adopted. When the medicinal materials are in the mature stage, a progressive threshold of 15% to 30% is adopted to obtain a graded hazard index and transmit it to the insect spot area correction unit;

[0086] The insect spot area correction unit is used to combine the leaf vein distribution characteristics in the spectral fusion image and use morphological opening operation to eliminate the artifact interference caused by leaf surface reflection or low light, obtaining an accuracy of 0.1mm. 2 The insect spot area data is integrated into the graded pest assessment parameters, which are then transmitted to the prevention and control decision module and the data storage module.

[0087] The pest control decision-making module combines graded pest assessment parameters with environmental parameter sets and uses a pesticide application decision tree model based on northern climate characteristics to generate a pest control plan that includes pesticide types adapted to the characteristics of the medicinal materials, atomized particle sizes suitable for adverse environments, and a pesticide application sequence optimized for temperature differences. The plan is then transmitted to the execution terminal.

[0088] The prevention and control decision module includes a time sequence planning unit, a pesticide decision unit, and an abnormal warning unit, among which:

[0089] The timing planning unit receives the environmental parameter set and obtains the day-night temperature difference data based on the temperature marked with the timestamp. When the day-night temperature difference is ≥10°C, a thermodynamic model is used to calculate the dew point condensation time on the leaf surface. This generates a pesticide application schedule that avoids the peak morning dew period of 6:00 to 8:00 a.m. and transmits it to the pesticide decision-making unit.

[0090] The pesticide decision-making unit receives graded pest assessment parameters and extracts growth stage identification parameters, as well as real-time temperature data from the environmental parameter set. It uses a fuzzy logic algorithm to select botanical or chemical pesticides, and generates a control plan for the pesticide type and atomization parameters adapted to the current temperature, as well as the application path, to the execution terminal. When the temperature is less than 5°C, an instruction containing an antifreeze-type atomizing liquid is added to the control plan.

[0091] The drug decision unit includes an atomization parameter optimization subunit and an environment adaptation subunit, where:

[0092] The atomization parameter optimization subunit is used to extract the polarization image of the spectral fusion image, calculate the contact angle data of the leaf surface based on the polarization image, use the fluid dynamics model to calculate the adhesion efficiency of the droplets on the waxy leaf surface with a contact angle greater than 110°, generate the atomization parameters of the particle size of 30 to 50 μm, and transmit them to the drug decision-making unit;

[0093] The environmental adaptation subunit is used to extract meteorological data collected from environmental parameters to obtain the inversion layer height and wind speed data. When the inversion layer height is less than 100m and the wind speed is greater than 3m / s, the Gaussian diffusion model is used to calculate the horizontal dispersion distance of the pesticide, generate a pesticide application path 0.5 to 1.2m away from the medicinal material canopy, and transmit it to the pesticide decision unit;

[0094] The abnormal warning unit is used to receive spectral fusion images and extract wormhole image information, detect the morphological characteristics of pests, and use the twin network to compare the pest model with the northern climate adaptation label in the feature comparison library. When the morphological similarity is <70% and the insect spot area is ≥5mm 2 When a new species invades, an early warning plan including the coordinates of the isolation area and emergency agents is generated and transmitted to the user terminal, and a new early warning pest sample is generated and stored in the data storage module.

[0095] The execution terminal is used to receive the control plan, and use the command signal of the control plan to make each spraying unit spray the chemical medicine on the pest area, and record the spraying data and transmit it to the data storage module;

[0096] The data storage module is used to receive and associate storage of environmental parameter sets, spectral fusion images, graded pest assessment parameters and pesticide application data. It adopts a spatiotemporal coupling incremental learning mechanism to generate feedback data to the environmental adaptation module, dynamic threshold analysis module and prevention and control decision module to update the preset processing parameters and optimize the prevention and control plan. It is also used to dynamically match environmental fluctuation parameters with changes in pest characteristics, and generate a feature comparison library with northern climate adaptation labels for storage.

[0097] The specific implementation process is as follows: In northern my country, the pest and disease monitoring and control systems for medicinal material planting often suffer from image distortion during monitoring due to sudden weather changes such as snowfall, sandstorms or heavy rain (especially snowfall). In addition, the low illumination in such weather environments leads to increased errors in the extraction of insect spot morphology on medicinal material leaves. The wormhole shape matching method preset by the traditional system relies on a fixed threshold and cannot adapt to dynamic environments.

[0098] Therefore, the system described in this solution was deployed in a medicinal material base in Chengde, Hebei Province. Taking a 50-mu planting field where Atractylodes lancea and Ziziphus jujuba were intercropped as an example, Atractylodes lancea, as a deep-rooted medicinal material, is susceptible to underground erosion by beetle larvae, while the dense branches and leaves of Ziziphus jujuba easily induce aphid clusters, and aphids will also bite Atractylodes lancea. Traditional prevention and control methods are inaccurate due to the complex microclimate of the intercropping field (such as a 30% difference in canopy humidity) and the interaction of pests. Usually, Atractylodes lancea and Ziziphus jujuba are planted in March, which is the transition period between winter and spring, and there may still be a sharp drop in temperature or snowfall. When encountering snowfall and visibility less than 50m, the traditional system relies on fixed white balance parameters and single-spectrum imaging technology in low light and weather sudden changes, and the error rate of insect spot edges is as high as 30%. This system activates the dark channel defogging algorithm through the weather mutation response unit of the environmental adaptation module, increasing the visibility correction coefficient from 0.3 to 0.8 and reducing the misjudgment rate of insect spot edges from 42% to 8%; in addition, when it is snowing at night, the thermal radiation calibration subunit corrects the leaf emissivity in real time based on the Stefan-Boltzmann law, so that the temperature resolution of the infrared image reaches 0.1°C (traditional systems are only 0.5°C), solving the leaf artifact interference caused by snow reflection at night; at the same time, the multispectral registration subunit aligns the visible light leaf vein features with the infrared wormhole edge (temperature gradient > 2°C / cm) through the SIFT algorithm, making the spatial error <1.5 pixels, thereby improving the insect spot detection accuracy by 60% under light intensity below 50lux.

[0099] Synchronously, the spectral acquisition module starts an 850nm near-infrared light source to penetrate the waxy leaves of Atractylodes lancea (contact angle of 115°), and at the same time suppresses the reflection of the leaves of Ziziphus jujuba through polarization filtering. The image fusion unit uses wavelet transform to fuse the back of the leaf and the polarization image, and separates the snowfall artifacts within the scale range of 0.5 to 2mm on the leaves of Atractylodes lancea. The gradient response intensity of the wormhole edge is increased to 2.3 times the background value, further improving the accuracy of the system in determining the wormhole type.

[0100] In addition, since medicinal materials are always in a dynamic growth process, the degree of impact of insect pests on medicinal materials varies in different growth stages. For example, 10% insect spot damage in the seedling stage of Atractylodes lancea is equivalent to 25% in the mature stage. However, the traditional system makes inaccurate assessment of insect pests in the seedling stage of Atractylodes lancea, and the static threshold is uniformly determined as 15%, which often leads to excessive or insufficient application of pesticides. Therefore, through the dynamic threshold analysis module combined with the growth cycle recognition unit, it is determined that when Atractylodes lancea is in the seedling stage (plant height <20cm), the stress response analysis subunit detects a 22% decrease in the 530nm band derivative peak (hyperspectral imaging analysis), triggering a flavonoid correlation warning; the root development monitoring subunit simultaneously detects root cavities with a diameter of 2.8mm through ground penetrating radar, which are determined to be infestations by beetle larvae. The insect spot area threshold is dynamically adjusted to 8% (the original threshold in the mature stage is 15%), triggering the correction of the underground pest impact parameter, which reduces the amount of pesticide applied in the seedling stage of Atractylodes lancea by 40% and increases the success rate of underground pest control to 92%.

[0101] Usually, after monitoring and identifying pests, instructions are issued to the field spraying system or drones and other equipment through the execution terminal to perform the pesticide application operation. However, in the transition environment between winter and spring, the low temperature causes the water-based atomized liquid to freeze, resulting in a significant reduction in the attachment rate to the waxy leaves. Therefore, this system uses the environmental adaptation subunit of the prevention and control decision module to calculate the current temperature as 4°C, the inversion layer height of 85m, and the wind speed of 3.5m / s, and generates a drone hovering path 0.7m away from the canopy. By dynamically adjusting the drone's pesticide application path, the pesticide deposition rate is increased to 89%; the atomization parameter optimization subunit switches to 40μm antifreeze droplets containing ethylene glycol at a temperature of 4°C, allowing the pesticide to penetrate the dense wax layer of the branches and leaves of the jujube, increasing the attachment rate from 48% to 78%, and the pesticide utilization rate in the inversion layer scenario to 90%, and the drug efficacy in the low temperature environment is extended by 3 hours.

[0102] When the time planning unit predicted that the morning dew condensation period was 6:20-7:45, the application window was delayed to 8:30 to avoid dilution of the solution. At the same time, based on the temperature difference gradient between the canopies of Atractylodes lancea and Ziziphus jujuba (surface temperature difference of 2-3°C), differentiated pesticide application was carried out in different time periods: the Atractylodes lancea area was sprayed during the low temperature period in the early morning to reduce the volatilization of the solution, and the Ziziphus jujuba area chose the high temperature period in the afternoon to enhance the diffusion of droplets.

[0103] By harvest time, this system reduced the area of ​​Atractylodes macrocephala insect spots by 62% year-on-year (compared to a 28% reduction in traditional plots), and increased the effectiveness of aphid control in wild jujubes by 55%. The average per-acre income from intercropping reached 3,800 yuan, a 40% increase compared to monocropping. Pesticide residues were 30% below the national standard, eliminating the monitoring blind spots of traditional systems in complex planting scenarios.

[0104] Example 2:

[0105] As attached Figure 1 and Figure 6 As shown, the difference from Example 1 is that the data storage module includes a pest feature library update unit and an incremental learning unit, wherein:

[0106] The pest feature library update unit receives new warning pest samples and performs three-dimensional modeling. It uses a generative adversarial network to generate a 200-degree rotational view with the main axis of the insect body as the rotation axis in the visible light and near-infrared dual bands, obtains a pest model with a northern climate label, and transmits it to the incremental learning unit.

[0107] The incremental learning unit is used to receive pest models and pesticide application data. Without restarting the system, it uses an elastic weight solidification algorithm to calculate the Fisher information matrix of the existing pest recognition model weights, retains the parameters ranked in the top 30% in terms of weight importance, dynamically updates the feature extraction layer weights in the U-Net network, generates a dynamically optimized pest and disease recognition model, and feeds back the white balance correction coefficient to the environmental adaptation module in real time. It calibrates the light source wavelength according to the near-infrared reflectivity characteristics of the newly added pest model; pushes the model parameters containing the new species recognition threshold and the incremental update of the growth stage compensation coefficient to the dynamic threshold analysis module; and outputs the optimal pesticide application strategy set for historically similar environments to the prevention and control decision module, screening the pesticide-environment combination scheme with a prevention and control success rate ≥90% in historical data.

[0108] The specific implementation process is as follows: When the system captures new insect pests, taking Mongolian Astragalus as an example, the leaves of Mongolian Astragalus are thin, the wax layer is weak, and the contact angle is only 85°, which is easily interfered by snow reflection. The traditional feature library update requires downtime for maintenance. The new pest identification cycle is as long as 14 days, and the broken veins are mistakenly identified as wormholes against the background of permafrost cracks, with an error rate of 14%. This solution uses the pest feature library update unit to generate a 200° rotated perspective three-dimensional model of Astragalus aphids with a permafrost crack background under the dual channels of visible light (580nm chlorophyll absorption peak) and near-infrared (1300nm permafrost penetration band), simulates the northern climate label of snow reflection +15% and permafrost crack characteristics as interference features, and improves the accuracy of new species identification from 68% to 89%.

[0109] The incremental learning unit calculated the Fisher information matrix of the existing Mongolian Astragalus spot recognition model, locking the first 30% of key parameters and updating only the remaining 70% of weights to accommodate the new insect species. After model iteration, the low-light F1 value increased from 0.72 to 0.88. Based on the new insect species' 22% decrease in reflectivity at 950nm in the near-infrared band, the light source was automatically adjusted to 1050nm for enhanced penetration. This correction coefficient for the permafrost spot area was then sent to the dynamic threshold module. Simultaneously, historical data for temperatures between -5°C and 2°C and humidity >80% was sent to the control decision module, identifying combinations with a control success rate of ≥90%.

[0110] Agent type: Plant-derived matrine + 40% ethylene glycol antifreeze;

[0111] Atomization parameters: particle size 25 μm, application height 0.3 m (avoiding the permafrost inversion layer);

[0112] The response time for aphid control in the permafrost zone of Mongolian Astragalus was shortened from 14 days to 6 hours, and the adhesion rate of the pesticide solution was increased to 81%, successfully blocking the spread of pests.

[0113] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. An intelligent pest monitoring and control system, characterized in that: The system includes an environment adaptation module, a spectrum acquisition module, a dynamic threshold analysis module, a prevention and control decision module, an execution terminal, and a data storage module, which are mutually signal-connected. The environment adaptation module is signal-connected to a number of sensor units deployed in the farmland environment. The execution terminal includes a number of pesticide application units deployed in the farmland environment for spraying chemical pesticides, wherein: The environmental adaptation module is used to collect light intensity, ambient temperature, ambient humidity and meteorological data in real time through the sensing unit, adjust the exposure parameters and white balance of the monitoring probe through a dynamic calibration algorithm, obtain the compensated environmental parameter set, and transmit it to the spectrum acquisition module, prevention and control decision module and data storage module respectively; The spectrum acquisition module is used to alternately illuminate the target medicinal material through a ring array of 850nm near-infrared light sources and a visible light source with a wavelength of 380-780nm based on the compensated environmental parameter set. By receiving the leaf surface reflectivity transmitted by the sensing unit, a polarization filter is used to suppress the leaf surface reflection to obtain a spectral fusion image, which is then transmitted to the dynamic threshold analysis module and the data storage module; The dynamic threshold analysis module is used to receive the spectral fusion image and the medicinal material image information transmitted by the sensor unit. It uses the improved U-Net network combined with the growth cycle recognition algorithm to establish a dynamic ratio threshold between the insect spot area and the medicinal material biomass. It obtains the graded pest assessment parameters including the growth stage identification parameters, the graded damage index and the insect spot area data, and transmits them to the prevention and control decision module and the data storage module. The pest control decision-making module combines graded pest assessment parameters with environmental parameter sets and uses a pesticide application decision tree model based on northern climate characteristics to generate a pest control plan that includes pesticide types adapted to the characteristics of the medicinal materials, atomized particle sizes suitable for adverse environments, and a pesticide application sequence optimized for temperature differences. The plan is then transmitted to the execution terminal. The execution terminal is used to receive the control plan, and use the command signal of the control plan to make each spraying unit spray the chemical medicine on the pest area, and record the spraying data and transmit it to the data storage module; The data storage module is used to receive and associate storage of environmental parameter sets, spectral fusion images, graded pest assessment parameters and pesticide application data. It adopts a spatiotemporal coupling incremental learning mechanism to generate feedback data to the environmental adaptation module, dynamic threshold analysis module and prevention and control decision module to update the preset processing parameters and optimize the prevention and control plan. It is also used to dynamically match environmental fluctuation parameters with changes in pest characteristics, and generate a feature comparison library with northern climate adaptation labels for storage.

2. The intelligent pest monitoring and control system according to claim 1, characterized in that: The environmental adaptation module includes an illumination compensation unit, a weather change response unit, a night imaging unit, and an integration unit, among which: The light compensation unit is used to collect the light intensity in all directions within the farmland planting area through the sensor unit, adjust the camera gain parameters using the adaptive white balance algorithm, obtain a balanced light distribution map, and transmit it to the integration unit; The weather change response unit is used to receive meteorological data from the sensing unit, extract precipitation particle distribution data, pre-process the image using a dark channel prior defogging algorithm, obtain the visibility correction coefficient in snowfall or rainstorm scenes, and transmit it to the integration unit; The night imaging unit is used to obtain a number of path data transmitted by the sensing unit. When the light intensity is lower than 50 lux, the thermal imaging function of the sensing unit is activated. The edge enhancement algorithm is used to extract the wormhole outline, obtain the wormhole spot morphology data under the light intensity lower than 50 lux, and transmit it to the integration unit. The integration unit is used to receive the balanced light distribution map, visibility correction coefficient and insect spot morphology data, bind them with the acquisition timestamp through the spatiotemporal label fusion algorithm, generate the compensated environmental parameter set, and synchronously transmit it to the spectral acquisition module, prevention and control decision module and data storage module.

3. The intelligent pest monitoring and control system according to claim 2, characterized in that: The night imaging unit includes a thermal radiation calibration subunit and a multispectral registration subunit, wherein: The thermal radiation calibration subunit is used to correct the thermal imaging data using the Stefan-Boltzmann law according to the ambient temperature transmitted by the sensing unit, obtain an infrared image with a temperature resolution of 0.1°C, and transmit it to the multispectral registration subunit; The multispectral registration subunit is used to match the feature points of the infrared image with the visible light with a wavelength of 380 to 780 nm collected by the sensing unit. The SIFT algorithm is used to achieve cross-modal image alignment: the leaf vein feature points enhanced by Gabor filtering are extracted in the visible light image, and the wormhole edge feature points with a temperature gradient of >2°C / cm are located in the infrared image. Finally, fused data with a spatial error of <2 pixels is obtained, and the fused data is transmitted to the integration unit as the worm spot morphology data under light intensity less than 50 lux.

4. The intelligent pest monitoring and control system according to claim 3, characterized in that: The spectrum acquisition module includes an active light source control unit, a filter switching unit, and an image fusion unit, wherein: The active light source control unit is used to receive the environmental parameter set and adjust the wavelength of the near-infrared light source according to the ambient humidity transmitted by the sensor unit. When the humidity is greater than 80%, it switches to the 950nm band to penetrate the water mist. It uses pulse modulation technology to eliminate motion blur, obtains leaf back image data that matches the environmental parameter set, and transmits it to the image fusion unit. A filter switching unit is used to select the optimal filter combination using a rotating filter group according to the blade surface reflectivity transmitted by the spectrometer in the sensing unit, obtain a polarized image that suppresses specular reflection, and transmit it to the image fusion unit; The image fusion unit is used to perform wavelet transform fusion on the leaf back image data and the polarization image, and adopts the regional energy weighted algorithm to enhance the wormhole edge contrast, obtain a spectral fusion image with a resolution of not less than 20 million pixels, and transmit it to the dynamic threshold analysis module and data storage module.

5. The intelligent pest monitoring and control system according to claim 4, characterized in that: The image fusion unit includes a shadow removal subunit and a multi-scale enhancement subunit, wherein: The shadow removal subunit is used to fuse the leaf back image data with the polarization image and detect and mark the occluded areas. The occluded areas include pixel missing areas caused by leaf overlap, external equipment occlusion, or low-light reflection. The occluded areas are defined as continuous pixel blocks with grayscale values ​​lower than 30% of the background mean in the visible light band and a sudden drop in reflectance greater than 15% in the near-infrared band. The leaf surface texture in the occluded area is reconstructed using the Poisson equation to obtain the spectral intermediate image with the occlusion removed, and the image is transmitted to the multi-scale enhancement subunit. The multi-scale enhancer unit is used to receive the spectral intermediate image and use the Laplace pyramid fusion algorithm in the scale range of 0.5 to 2 mm to separate the leaf artifacts caused by snow cover in the northern region from the real wormholes in the frequency domain, enhance the gradient response intensity of the wormhole edge to more than twice the background value, generate a spectral fusion image, and transmit it to the dynamic threshold analysis module and data storage module.

6. The intelligent pest monitoring and control system according to claim 5, characterized in that: The dynamic threshold analysis module includes a growth cycle identification unit, a hazard coefficient calculation unit, and an insect spot area correction unit, among which: The growth cycle recognition unit is used to receive the medicinal material image information transmitted by the sensing unit, extract the plant height and stem thickness characteristics of the medicinal material through a convolutional neural network, and analyze the reflectance data in the 400-1000nm band based on the information of the spectral fusion image. The improved ResNet50 network is used to distinguish between the seedling stage and the mature stage. The growth cycle recognition unit includes a stress response analysis subunit and a root development monitoring subunit. It is also used to receive the flavonoid substance correlation degree transmitted by the stress response analysis subunit and the underground pest impact degree transmitted by the root development monitoring subunit. The medicinal material growth rate compensation coefficient is calculated based on the characteristics of the short growth cycle and large day and night temperature difference of the medicinal material in the northern climate, and the growth stage identification parameters of the medicinal material are obtained and transmitted to the hazard coefficient calculation unit. The hazard coefficient calculation unit is used to adjust the insect spot threshold according to the growth stage identification parameters of the medicinal materials. When the medicinal materials are in the seedling stage, a dynamic interval threshold of 5% to 15% of the biomass proportion is adopted. When the medicinal materials are in the mature stage, a progressive threshold of 15% to 30% is adopted to obtain a graded hazard index and transmit it to the insect spot area correction unit; The insect spot area correction unit is used to combine the leaf vein distribution characteristics in the spectral fusion image and use morphological opening operation to eliminate the artifact interference caused by leaf surface reflection or low light, obtaining an accuracy of 0.1mm. 2 The insect spot area data is integrated into the graded pest assessment parameters, which are then transmitted to the prevention and control decision module and the data storage module.

7. The intelligent pest monitoring and control system according to claim 6, characterized in that: The growth cycle identification unit includes a stress response analysis subunit and a root development monitoring subunit, wherein: The stress response analysis subunit is used to receive the 500-600mm band data collected by the sensor unit, combine it with the convolutional neural network to extract the characteristics of the medicinal material plant height and stem diameter, and detect the accumulation characteristics of secondary metabolites in the medicinal material. When the first-order derivative peak at 530nm is detected to drop by more than 15%, it is determined to be a metabolic abnormality caused by pest and disease stress. The correlation degree of flavonoid substances is generated and uploaded to the growth cycle recognition unit; The root development monitoring subunit is used to obtain the three-dimensional structure of the root system through the sensing unit and reconstruct the root density distribution. When a root cavity with a diameter greater than 2mm and a soil moisture content greater than 20% is detected, it is determined to be an insect pest infestation event, and the underground insect pest impact degree is generated and uploaded to the growth cycle identification unit.

8. The intelligent pest monitoring and control system according to claim 7, characterized in that: The prevention and control decision module includes a time sequence planning unit, a pesticide decision unit, and an abnormal warning unit, among which: The timing planning unit receives the environmental parameter set and obtains the day-night temperature difference data based on the temperature marked with the timestamp. When the day-night temperature difference is ≥10°C, a thermodynamic model is used to calculate the dew point condensation time on the leaf surface. This generates a pesticide application schedule that avoids the peak morning dew period of 6:00 to 8:00 a.m. and transmits it to the pesticide decision-making unit. The pesticide decision-making unit receives graded pest assessment parameters and extracts growth stage identification parameters, as well as real-time temperature data from the environmental parameter set. It uses a fuzzy logic algorithm to select botanical or chemical pesticides, and generates a control plan for the pesticide type and atomization parameters adapted to the current temperature, as well as the application path, to the execution terminal. When the temperature is less than 5°C, an instruction containing an antifreeze-type atomizing liquid is added to the control plan. The abnormal warning unit is used to receive spectral fusion images and extract wormhole image information, detect the morphological characteristics of pests, and use the twin network to compare the pest model with the northern climate adaptation label in the feature comparison library. When the morphological similarity is <70% and the insect spot area is ≥5mm 2 When a new species invades, an early warning plan including the coordinates of the isolation area and emergency agents is generated and transmitted to the user terminal, and a new early warning pest sample is generated and stored in the data storage module.

9. The intelligent pest monitoring and control system according to claim 8, characterized in that: The drug decision unit includes an atomization parameter optimization subunit and an environment adaptation subunit, where: The atomization parameter optimization subunit is used to extract the polarization image of the spectral fusion image, calculate the contact angle data of the leaf surface based on the polarization image, use the fluid dynamics model to calculate the adhesion efficiency of the droplets on the waxy leaf surface with a contact angle greater than 110°, generate the atomization parameters of the particle size of 30 to 50 μm, and transmit them to the drug decision-making unit; The environmental adaptation subunit is used to extract the meteorological data collected from the environmental parameters to obtain the inversion layer height and wind speed data. When the inversion layer height is <100m and the wind speed is >3m / s, the Gaussian diffusion model is used to calculate the horizontal drift distance of the agent, generate the application path 0.5 to 1.2m away from the medicinal material canopy, and transmit it to the agent decision unit.

10. The intelligent pest monitoring and control system according to claim 9, characterized in that: The data storage module includes a pest signature library update unit and an incremental learning unit, wherein: The pest feature library update unit receives new warning pest samples and performs three-dimensional modeling. It uses a generative adversarial network to generate a 200-degree rotational view with the main axis of the insect body as the rotation axis in the visible light and near-infrared dual bands, obtains a pest model with a northern climate label, and transmits it to the incremental learning unit. The incremental learning unit is used to receive pest models and pesticide application data. Without restarting the system, it uses an elastic weight solidification algorithm to calculate the Fisher information matrix of the existing pest recognition model weights, retains the parameters ranked in the top 30% in terms of weight importance, dynamically updates the feature extraction layer weights in the U-Net network, generates a dynamically optimized pest and disease recognition model, and feeds back the white balance correction coefficient to the environmental adaptation module in real time. It calibrates the light source wavelength according to the near-infrared reflectivity characteristics of the newly added pest model; pushes the model parameters containing the new species recognition threshold and the incremental update of the growth stage compensation coefficient to the dynamic threshold analysis module; and outputs the optimal pesticide application strategy set for historically similar environments to the prevention and control decision module, screening the pesticide-environment combination scheme with a prevention and control success rate ≥90% in historical data.

Images