A zero-residue safe plant protection management method based on intelligent monitoring

Abstract

The application discloses a zero agricultural residue safety plant protection management method based on intelligent monitoring, and comprises the following steps: constructing zero agricultural residue safety plant protection technology (OPPT) formed by utilizing ecological immune isolation technology, pest control technology and bionic stress repair technology and an operation procedure of the OPPT, preparing micron state, nanometer state and quantum dot product combinations, and creating an IRIP technology effect evaluation standard. The intelligent detection database guides the application of the OPPT technology in the prevention and treatment of citrus Huanglongbing, the plant protection technology operation after apple bagging, the implementation of zero agricultural residue plant protection technology for vegetables and fruits, and the setting of zero agricultural residue GAP operation procedures for Chinese herbal medicines, and under the implementation of non-chemical pesticides, a unified pest control standard for multiple crop synchronous diseases and pests is formed, that is, when the IRIP value of crops is greater than or equal to 300 during the implementation of the OPPT technology, the prevention and treatment rate of common diseases and pests is 90%, and the zero chemical pollution of the planting environment, the zero pesticide invasion of the grower, the zero pesticide residue of the agricultural products and the modern agricultural safety production scene are realized.

Description

Technical Field

[0001] This invention relates to the field of plant protection safety management, and in particular to a zero-pesticide residue safe plant protection management method based on intelligent monitoring. Background Technology

[0002] With the increasing awareness of food safety and ecological environmental protection, zero-residue safe plant protection technology (OPPT) has become an important direction for the development of modern agriculture. By developing green, safe and efficient plant protection technologies, precise control of crop diseases and pests and management of healthy plant growth can be achieved, which is of great significance for ensuring the quality and safety of agricultural products and promoting the sustainable development of agriculture.

[0003] Currently, traditional plant protection management relies heavily on experience and extensive operations, lacking real-time monitoring and precise analysis of pest and disease occurrence patterns and plant growth status. This leads to delayed control measures and significant resource waste. Furthermore, existing sensors suffer from isolated data collection and limited analytical dimensions, hindering the collaborative processing and effective utilization of multi-source data. Finally, plant management strategies lack dynamic optimization mechanisms, making it impossible to flexibly adjust them based on actual pest and disease conditions and plant growth needs. The deep integration of intelligent monitoring technology with agricultural plant protection offers an innovative path to solving traditional plant protection challenges, driving the field towards precision, intelligence, and ecological transformation. This invention proposes a zero-pesticide residue safe plant protection management method based on intelligent monitoring. It constructs a reference strategy library through cluster analysis of historical plant protection information, integrates and refines multi-source data using a zero-pesticide residue plant protection monitoring network, and utilizes pest and disease prediction models and plant growth assessment models for accurate prediction. Finally, it optimizes plant management strategies based on an objective function, effectively overcoming the shortcomings of traditional technologies. This method achieves intelligent, refined, and efficient zero-pesticide residue plant protection safe management, significantly improving the scientific nature and effectiveness of plant protection management. It provides a practical solution for achieving zero-pesticide residue safe plant protection and has significant application value in promoting green and high-quality agricultural development. Summary of the Invention

[0004] The purpose of this invention is to provide a method for zero-pesticide residue safe plant protection management based on intelligent monitoring. To achieve the above objective, this invention is implemented according to the following technical solution:

[0005] This invention includes the creation of zero pesticide residue safe plant protection technology (OPPT) and an intelligent detection program; the intelligent detection program is used to guide the implementation of OPPT for the prevention and control of Huanglongbing disease with zero pesticide residue in citrus, the implementation of zero pesticide residue bagging technology in apples, and the implementation of safe plant protection technology with zero pesticide residue in vegetables.

[0006] The Zero Pesticide Residue Safe Plant Protection Technology (OPPT) of this invention includes the innovative OPPT technology concept, OPPT technology product combination, and OPPT technology effect evaluation standards. The innovative OPPT technology concept includes ecological immune isolation and shielding effect, single plant-derived pesticide integrated pest and disease control effect, and biomimetic stress repair effect. The OPPT technology product combination includes micro-nano ecological membrane, nano-carvacrol, and chitin quantum dots. The OPPT technology effect evaluation standards include IRIP≥300, integrated pest and disease control rate≥90%, zero environmental pollution, zero harm to growers, and zero pesticide residue.

[0007] The nano-ecological membrane used for ecological immune isolation comprises one or two of the following materials in a 1:1 ratio: 10-1200 nanometer calcium carbonate particles, nano zinc, nano iodine, nano copper, nano titanium, nano boron, and nano diatomaceous earth. The membrane dispersion material includes soluble chitin, chitosan, resistant starch, alginate, agar, acetate, polylactic acid, and polyglutamic acid. The content of micro- and nanomaterials in the dispersant ranges from 10-50%.

[0008] The plant-derived pesticides used in integrated pest management are carvacrol and thymol;

[0009] Chitosan quantum dots are nano-state chitosan glucosamine;

[0010] The intelligent detection program includes:

[0011] Historical plant protection information is categorized by clustering, and a reference strategy library is constructed based on the clustering results; the clustering results include historical pest and disease data, historical plant data, and historical plant management strategies; this invention specifically refers to OPPT for the integrated prevention and control of crop pests and diseases.

[0012] An OPPT monitoring network is constructed to acquire plant protection sensing data. Multi-source data fusion and data extraction are performed on the plant protection sensing data to obtain pest and disease risk data and plant status data.

[0013] Based on the clustering results, a pest and disease prediction model and a plant growth assessment model are constructed. The pest and disease risk data and the plant status data are respectively input into the pest and disease prediction model and the plant growth assessment model to obtain the predicted pest and disease severity and the predicted plant growth status.

[0014] Based on the predicted severity of pests and diseases and the predicted plant growth status, a reference plant management strategy is obtained by matching the reference strategy library.

[0015] The OPPT management period is determined based on the severity of pests and diseases. The operation frequency of different plant management strategies is determined based on the OPPT management period. The plant management objective function is determined. The reference plant management strategy is optimized based on the plant management objective function to obtain the optimal plant management strategy. The reference strategy library is then updated.

[0016] Furthermore, the method for constructing the reference strategy library includes:

[0017] Historical plant protection information is categorized into historical pest and disease data, historical plant status data, and historical plant management strategies through clustering. The historical pest and disease data includes pest and disease monitoring data, pest and disease indicators, environmental meteorological data, soil biological information, and the corresponding historical pest and disease development levels. The historical plant status data includes comprehensive plant growth characteristics, environmental meteorological data, soil nutrient information, and the corresponding historical plant growth status. The historical plant management strategies are used to select appropriate zero-residue integrated pest management methods based on the historical plant growth status and the severity of historical pests and diseases, including implementation parameters, implementation frequency, and implementation paths for different types of OOPT (Out-of-Pollution-Treating) technologies.

[0018] Based on geographical coordinates and plant categories, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are divided into the same strategy zone. Within the same strategy zone, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are matched and associated according to time periods to form a reference strategy library.

[0019] The time period matching rules for the data in the reference strategy library are as follows: when the overlap between the time period of historical pest and disease development and the time period of historical plant growth status is greater than 80%, the data will be associated; when the time interval between the start time of historical pest and disease development and the start time of historical plant management strategy is less than the remedial time threshold, the data will be associated; when the historical pest and disease development is associated with both historical plant growth status and historical plant management strategy, a set of reference strategies will be formed; otherwise, the data will be defined as invalid.

[0020] The reference strategy library consists of reference strategies.

[0021] Furthermore, the method for obtaining data on potential pest and disease risks and plant status data includes:

[0022] An OPPT monitoring network is constructed to acquire plant protection sensing data; the OPPT monitoring network includes a hyperspectral imaging sensor, a biosensor, a high-definition camera, various environmental parameter sensors, and biochemical analysis experiments;

[0023] A hyperspectral imaging sensor acquires plant spectral reflectance information, and performs data inversion on the spectral reflectance information to obtain chlorophyll concentration and first plant activity. The first plant growth stage is determined based on the chlorophyll concentration; the first plant activity reflects the degree of damage to plant cell structure.

[0024] Biosensors are used to monitor the concentration and types of pest-related microorganisms in the air in real time to obtain primary pest monitoring data.

[0025] A high-definition camera acquires high-definition plant images, and the high-definition plant images are processed to obtain first plant features and second plant features. The second plant growth stage is determined by matching the first plant features with a plant growth domain library. The first plant features represent the plant's growth shape and size. The second plant features represent the plant's external appearance characteristics affected by pests and diseases.

[0026] Plant growth climate environment is obtained through various environmental parameter sensors, and soil nutrient information and primary soil biological information are obtained through biochemical analysis experiments.

[0027] The third plant growth stage is obtained by taking the union of the first and second plant growth stages. The plant activity range is determined based on the third plant growth stage. The second plant activity is obtained by crossing the first plant activity with the second plant characteristics. The third plant activity is determined based on the plant activity range and the second plant activity. The comprehensive plant growth characteristics are composed of the third plant growth stage and the third plant activity. The plant activity range represents the range of various activities corresponding to the growth stage. The third plant activity reflects the overall performance of the plant's physiological and biochemical functions, antioxidant activity, stress resistance, and metabolic activities.

[0028] Based on the second plant characteristics, pest and disease indicators are determined, and the second plant characteristics are used to correct the first soil biological information and the first pest and disease monitoring data to obtain the second soil biological information and the second pest and disease monitoring data.

[0029] The pest and disease indicators, plant growth climate environment, second soil biological information, and second pest and disease monitoring data are combined to form pest and disease risk data, and the comprehensive plant growth characteristics, plant growth climate environment, and soil nutrient information are combined to form plant status data.

[0030] Furthermore, the method for obtaining predictions of pest and disease severity and plant growth status includes:

[0031] Historical pest and disease data were used as a comprehensive pest and disease set. Random forest was used to divide the comprehensive pest and disease set into a pest and disease training set and a pest and disease test set in a ratio of 6:4.

[0032] The pest and disease prediction model is constructed, which specifically includes dynamically adjusting the selection gate, time decay injection layer, multi-head LSTM unit, spatiotemporal attention layer and fully connected layer;

[0033] The dynamic adjustment selection gate dynamically filters and fuses key features based on temporal correlation, and the expression is:

[0034]

[0035]

[0036] in For time step The output of the dynamic feature selection gate, It is the Sigmoid activation function. This is the weight matrix. For time step The hidden state, For time step The input vector, For bias terms, For time step The fused feature vector;

[0037] The time decay injection layer uses a time decay factor. Assign time-time sensitive weights to time-series data. , For learnable parameters, For time intervals, Index for time-aware units. The number of time-aware units;

[0038] The multi-head LSTM unit captures the dependencies of pest and disease data at different time scales in parallel, and the expression is:

[0039]

[0040]

[0041]

[0042] in For time step The hidden state vector after memory enhancement To step the time Global Time Series Knowledge Base Mapping to the hidden state vector Weight matrices of the same dimension For the first Each time-sensing unit at time step The hidden state vector, To update the global time series knowledge base Gated loop unit, For multi-head time sensing units at time steps Hidden state splicing, For the first A long short-term memory network;

[0043] The spatiotemporal attention layer identifies the correlation patterns between key time nodes and spatial dimensions;

[0044] The fully connected layer connects to the spatiotemporal attention layer to generate a predicted degree of pest and disease infestation; the degree of pest and disease infestation includes pest and disease type and level.

[0045] Historical plant state data is used as a comprehensive set of plant states. Random forest is used to divide the comprehensive set of plant states into a plant state training set and a plant state test set in a 6:4 ratio.

[0046] A plant growth assessment model is constructed, which specifically includes a multi-source fusion layer, a deep residual block, a cross-attention layer, a multi-task prediction head, a hybrid loss calculation layer, and an output layer.

[0047] The multi-source fusion layer is used to align the spatiotemporal resolution of different sensor data; the deep residual block is used to extract higher-order nonlinear growth features of plants; the cross-attention layer establishes a dynamic correlation between environmental factors and growth indicators through a cross-attention mechanism; the multi-task prediction head outputs predictions of plant growth status in parallel; the plant growth status includes growth rate, energy conversion, and senescence degree.

[0048] The hybrid loss calculation layer comprehensively optimizes prediction accuracy and biological rationality through a hybrid adaptive loss function, the expression of which is:

[0049]

[0050] in For hybrid adaptive loss function, The weights of the gradient direction penalty term are the coefficients. These are the weighting coefficients for the biological constraint term. This represents the current number of days of growth. The maximum number of days for growth. Mean square error, This is a gradient direction penalty term. For activation function, For the true value Over time rate of change, For predicted values Over time rate of change, This is a biological constraint term, reflecting the energy conservation among photosynthetic accumulation, respiratory consumption, and biomass increment.

[0051] Furthermore, the method for obtaining the reference plant management strategy includes:

[0052] The reference strategy database data is coarsely screened based on environmental meteorology from plant protection sensing data, and then finely screened based on spatiotemporal data from plant protection sensing data.

[0053] The comprehensive similarity of feature vectors between the predicted severity of pests and diseases and the historical severity of pests and diseases in the reference strategy library is calculated. The highest comprehensive similarity is taken as the pest and disease similarity, and the historical plant management strategy corresponding to the highest comprehensive similarity is taken as the first candidate plant management strategy. The calculation steps of the comprehensive similarity are as follows: calculate the average value of time series similarity, spatial propagation similarity and environmental matching degree to obtain spatiotemporal environmental similarity, and calculate the product of spatiotemporal environmental similarity and cosine similarity to obtain the comprehensive similarity.

[0054] Calculate the comprehensive similarity of feature vectors between the predicted plant growth status and the historical plant growth status in the reference strategy library, take the highest comprehensive similarity as the plant growth status similarity, and take the historical plant management strategy corresponding to the highest comprehensive similarity as the second candidate plant management strategy.

[0055] The first candidate weight and the second candidate weight are determined according to the ratio of the similarity between pests and diseases and the similarity between plant growth status. The first candidate plant management strategy and the second candidate plant management strategy are weighted and fused together according to the first candidate weight and the second candidate weight to obtain the reference plant management strategy.

[0056] Furthermore, the method for determining the plant management objective function includes:

[0057] The OOPT treatment period is determined based on the severity of pests and diseases. The frequency of operation of different plant management strategies within the control management period is calculated based on the OOPT management period and the effective pest and disease control time of different plant management strategies. The OOPT management period includes the emergency period, the control period, and the maintenance period.

[0058] The objective function for plant management is determined based on the management costs, ecological impacts, and strategy effectiveness time during each OPPT management period. The expression is as follows:

[0059]

[0060] in Let the objective function be plant management. for The length of the disinfection management period, for Effective prevention and control time of management strategies for Prevention and control period The operational intensity of management strategies For the corresponding standard operating intensity, for The cost of management strategies for The area of ​​the management strategy for The effective area of ​​a single device in the management strategy This is the ecological cost conversion factor. For use with OPPT Ecological impact index For the number of ecological impact categories, for Prevention and control period The time it takes for management strategies to take effect;

[0061] The constraints of the plant management objective function are determined based on different plant management strategies, and the expression is as follows:

[0062]

[0063] in For strategy Disinfection efficiency for The severity of pests and diseases required during the pest and disease control period. , For equipment The rated maximum power and rated minimum power, For equipment The percentage of time the device is open. For equipment power, natural enemy Ecological impact coefficient, For the quantity released, For ecological carrying capacity threshold, The application concentration of OPPT , These are the minimum and maximum application concentrations of OPPT.

[0064] Furthermore, the method for obtaining the optimal plant management strategy includes:

[0065] Based on the objective function of plant management, the reference plant management strategy is optimized using Nash equilibrium to obtain the optimal plant management strategy. The specific steps are as follows:

[0066] Based on the rational participants in the game defined by management strategies, participant 1 is identified as physical control, participant 2 as biological control, and participant 3 as OPPT control. The strategy variables for each participant are determined. The game payoff function is determined based on the elimination effect of each management strategy and the plant management objective function. The strategy set is determined based on the constraints of the plant management objective function. The strategy variables for participant 1 are equipment density and equipment power; the strategy variables for participant 2 are insect release density and insect species; and the strategy variables for participant 3 are operation type ratio and OPPT concentration.

[0067] The equilibrium point is solved by the asynchronous gradient response method. The existence of Nash equilibrium is proved by Shizuo Kakutani's fixed point theorem. The strategy variables of the participants are repeatedly adjusted until the game payoff function is continuous and quasi-concave in the strategy space and the plant management objective function is maximized. The optimal combination of strategy variables is then output to obtain the optimal plant management strategy.

[0068] The optimal plant management strategy and the corresponding prediction of pest and disease severity and plant growth status are used to update the reference strategy library.

[0069] Secondly, a zero-pesticide residue safe plant protection management method based on intelligent monitoring includes:

[0070] Management strategy module: used to classify historical plant protection information by clustering, construct a reference strategy library based on the clustering results, obtain reference plant management strategies by matching the reference strategy library with the predicted severity of pests and diseases and the predicted plant growth status, and update the reference strategy library based on the optimal plant management strategy;

[0071] Sensing data module: used to build OPPT monitoring network to acquire plant protection sensing data, and to perform multi-source data fusion and data refinement on plant protection sensing data to obtain pest and disease risk data and plant status data;

[0072] Model prediction module: used to construct a pest and disease prediction model and a plant growth assessment model based on the clustering results, and input the pest and disease risk data and the plant status data into the pest and disease prediction model and the plant growth assessment model respectively to obtain the predicted pest and disease severity and the predicted plant growth status;

[0073] Strategy optimization module: used to determine the plant management objective function, and optimize the reference plant management strategy according to the plant management objective function to obtain the optimal plant management strategy;

[0074] Management platform module: used to view, store and manage the predicted severity of pests and diseases, the predicted plant growth status and the optimal plant management strategy, and to perform zero-pesticide residue pest and disease control operations based on the predicted severity of pests and diseases, the predicted plant growth status and the optimal plant management strategy.

[0075] The beneficial effects of this invention are:

[0076] This invention is a zero-pesticide residue safe plant protection management method based on intelligent monitoring. Compared with the prior art, this invention has the following technical advantages:

[0077] This invention, through multi-source data fusion, data refinement, model construction, strategy matching, and equilibrium game theory, can enhance data preprocessing capabilities and improve model adaptability in zero-pesticide residue plant protection safety management. It can improve the efficiency and accuracy of zero-pesticide residue plant protection safety management, optimize the technology, significantly save resources, increase work efficiency, and achieve multi-dimensional, real-time management of plant protection safety. It provides more reliable technical support for zero-pesticide residue plant protection safety, helps improve the scientific nature and effectiveness of plant protection management, and has significant application value in promoting green and high-quality agricultural development. It can adapt to the management needs of different zero-pesticide residue plant protection safety management systems and different users, and has a certain degree of universality. Attached Figure Description

[0078] Figure 1 This is a flowchart illustrating the steps of a zero-pesticide residue safe plant protection management method based on intelligent monitoring according to the present invention. Detailed Implementation

[0079] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0080] The present invention discloses a zero-pesticide residue safe plant protection management method based on intelligent monitoring, comprising the following steps:

[0081] like Figure 1 As shown, this embodiment includes the following steps:

[0082] Historical plant protection information is categorized by clustering, and a reference strategy library is constructed based on the clustering results. The clustering results include historical pest and disease data, historical plant data, and historical plant management strategies. The historical plant management strategies specifically refer to the integrated pest and disease control strategy of OPPT.

[0083] An OPPT monitoring network is constructed to acquire plant protection sensing data. Multi-source data fusion and data extraction are performed on the plant protection sensing data to obtain pest and disease risk data and plant status data.

[0084] Based on the clustering results, a pest and disease prediction model and a plant growth assessment model are constructed. The pest and disease risk data and the plant status data are respectively input into the pest and disease prediction model and the plant growth assessment model to obtain the predicted pest and disease severity and the predicted plant growth status.

[0085] Based on the predicted severity of pests and diseases and the predicted plant growth status, a reference plant management strategy is obtained by matching the reference strategy library.

[0086] The OPPT management period is determined based on the severity of pests and diseases. The operation frequency of different plant management strategies is determined based on the OPPT management period. The plant management objective function is determined. The reference plant management strategy is optimized based on the plant management objective function to obtain the optimal plant management strategy. The reference strategy library is then updated.

[0087] The project establishes management procedures and implementation standards for zero-pesticide residue safe plant protection technology (OPPT), including: constructing an ecological immune shielding technology using nano-sized chitosan ecological membranes to isolate pest eggs from implantation and normal hatching, while inhibiting the germination and reproduction of fungal and bacterial spores and cells; forming a biological pesticide pest and disease control technology through nano-sized plant-derived pesticides; and using chitosan glucosamine as a biomimetic stress factor to create an internal ecological shielding effect and membrane repair effect to resist damage to plants from pests and diseases; the nano-sized chitosan ecological membrane is composed of nano-calcium, nano-zinc, nano-iodine, nano-copper, and chitosan quantum dots; the pests specifically include spider mites, aphids, whiteflies, thrips, and psyllids; the nano-sized plant-derived pesticides include any one or a combination of two of the following: carvacrol, matrine, eugenol, and thymol; the chitosan glucosamine includes glucosamine, amino oligosaccharides, chitosan quantum dots, and chitosan polysaccharides; and the internal ecological shielding effect and membranes include cell membranes, chlorophyll membranes, and mitochondrial membranes.

[0088] The OPPT management procedure reflects the achievement of the effect evaluation standard when the IRIP of crops is ≥300; the effect evaluation standard includes: zero pollution of the planting environment by chemical pesticides, zero harm to growers by chemical pesticides, and zero chemical pesticide residue rate in agricultural products.

[0089] In this embodiment, the method for constructing the reference policy library includes:

[0090] Historical plant protection information is categorized into historical pest and disease data, historical plant status data, and historical plant management strategies through clustering. The historical pest and disease data includes pest and disease monitoring data, pest and disease indicators, environmental meteorological data, soil biological information, and the corresponding historical pest and disease development levels. The historical plant status data includes comprehensive plant growth characteristics, environmental meteorological data, soil nutrient information, and the corresponding historical plant growth status. The historical plant management strategies are used to select appropriate zero-residue pest and disease control methods based on the historical plant growth status and the severity of historical pests and diseases, including implementation parameters, frequency, and implementation paths for different types of control technologies.

[0091] Based on geographical coordinates and plant categories, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are divided into the same strategy zone. Within the same strategy zone, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are matched and associated according to time periods to form a reference strategy library.

[0092] The time period matching rules for the data in the reference strategy library are as follows: when the overlap between the time period of historical pest and disease development and the time period of historical plant growth status is greater than 80%, the data will be associated; when the time interval between the start time of historical pest and disease development and the start time of historical plant management strategy is less than the remedial time threshold, the data will be associated; when the historical pest and disease development is associated with both historical plant growth status and historical plant management strategy, a set of reference strategies will be formed; otherwise, the data will be defined as invalid.

[0093] The reference strategy library consists of reference strategies;

[0094] In the actual assessment, historical plant protection information from apple orchards in Luochuan County, Shaanxi Province, over the past five years was collected. Strategy zones were divided according to apple variety and planting area. Taking the data correlation of strategy zones corresponding to Shaanxi Province's Apple Variety A as an example, it was found that the period of codling moth infestation was from July 5th to August 29th, 2015, while the period from fruit enlargement to maturity was from June 29th to September 1st, 2015, with a time overlap of 55 / 65 = 85%. Correlation of the data revealed that the first management strategy adopted to address the codling moth infestation was on July 12th, 2015, 7 days after the onset of the infestation, less than the remedial time threshold of 10 days. Correlation of this set of historical pest and disease development levels, along with historical plant growth status and historical plant management strategies, can form a set of reference strategies.

[0095] Multiple sets of reference strategies are combined into a reference strategy library.

[0096] In this embodiment, the method for obtaining pest and disease risk data and plant status data includes:

[0097] An OPPT (Optimal Plant Protection Test) monitoring network is constructed to acquire plant protection sensing data; the zero pesticide residue plant protection monitoring network includes a hyperspectral imaging sensor, a biosensor, a high-definition camera, various environmental parameter sensors, and biochemical analysis experiments;

[0098] A hyperspectral imaging sensor acquires plant spectral reflectance information, and performs data inversion on the spectral reflectance information to obtain chlorophyll concentration and first plant activity. The first plant growth stage is determined based on the chlorophyll concentration; the first plant activity reflects the degree of damage to plant cell structure.

[0099] Biosensors are used to monitor the concentration and types of pest-related microorganisms in the air in real time to obtain primary pest monitoring data.

[0100] A high-definition camera acquires high-definition plant images, and the high-definition plant images are processed to obtain first plant features and second plant features. The second plant growth stage is determined by matching the first plant features with a plant growth domain library. The first plant features represent the plant's growth shape and size. The second plant features represent the plant's external appearance characteristics affected by pests and diseases.

[0101] Plant growth climate environment is obtained through various environmental parameter sensors, and soil nutrient information and primary soil biological information are obtained through biochemical analysis experiments.

[0102] The third plant growth stage is obtained by taking the union of the first and second plant growth stages. The plant activity range is determined based on the third plant growth stage. The second plant activity is obtained by crossing the first plant activity with the second plant characteristics. The third plant activity is determined based on the plant activity range and the second plant activity. The comprehensive plant growth characteristics are composed of the third plant growth stage and the third plant activity. The plant activity range represents the range of various activities corresponding to the growth stage. The third plant activity reflects the overall performance of the plant's physiological and biochemical functions, antioxidant activity, stress resistance, and metabolic activities.

[0103] Based on the second plant characteristics, pest and disease indicators are determined, and the second plant characteristics are used to correct the first soil biological information and the first pest and disease monitoring data to obtain the second soil biological information and the second pest and disease monitoring data.

[0104] The pest and disease indicators, plant growth climate environment, second soil biological information, and second pest and disease monitoring data are combined to form pest and disease risk data, and the comprehensive plant growth characteristics, plant growth climate environment, and soil nutrient information are combined to form plant status data.

[0105] In the actual assessment, taking the plant protection sensing data of the apple orchard on July 15, 2024 as an example, the data inversion of the plant spectral reflectance information collected by the hyperspectral imaging sensor yielded a chlorophyll concentration of 40 mg / g and the first plant activity = (1 - plant cell structure damage degree) = (1 - 0.25) = 0.75, thus determining the first plant growth stage as the first stage of fruit enlargement.

[0106] The biosensor detected an airborne concentration of 80 spider mite-related microorganisms per cubic meter. 3 ，高清摄像头获取的图像显示第一植物特征（苹果树树高 3.5m, crown width 2.5m), second plant characteristics (some leaves have yellowish-white spots, accounting for 0.11), after matching with the plant growth domain library, it was determined that the second plant growth stage is also the first stage of fruit enlargement, and the third plant growth stage is also the first stage of fruit enlargement, with the plant activity range of 0.7-0.82.

[0107] Various environmental parameter sensors measured a temperature of 28℃, humidity of 55%, and light intensity of 7000 lux. Biochemical analysis revealed that the soil contained 1.2% nitrogen and 0.6% phosphorus. The first soil biological information showed that the number of beneficial microorganisms was 800 / g.

[0108] The cross-calculation of the first plant activity and the second plant characteristic yielded a second plant activity of 0.72. The third plant activity = second plant activity * (second plant activity / mean range of plant activity) = 0.72 * (0.72 / 0.76) = 0.68.

[0109] Based on the second plant characteristic, the pest and disease indicators were determined to be a leaf damage rate of 15% and a pest density of 9.1 insects / plant. This was corrected to obtain the second soil biological information (beneficial microorganism count of 800*(1-0.11) = 712 cells / g) and the second pest and disease monitoring data (red spider mite-related microorganism concentration of 80*(1+0.11) = 88.8 cells / m³). 3）；

[0110] The pest and disease indicators, plant growth climate environment, second soil biological information, and second pest and disease monitoring data are combined to form pest and disease risk data. The comprehensive plant growth characteristics (third plant growth stage and third plant activity), plant growth climate environment, and soil nutrient information are combined to form plant status data.

[0111] In this embodiment, the method for obtaining the predicted severity of pests and diseases and the predicted plant growth status includes:

[0112] Historical pest and disease data were used as a comprehensive pest and disease set. Random forest was used to divide the comprehensive pest and disease set into a pest and disease training set and a pest and disease test set in a ratio of 6:4.

[0113] The pest and disease prediction model is constructed, which specifically includes dynamically adjusting the selection gate, time decay injection layer, multi-head LSTM unit, spatiotemporal attention layer and fully connected layer;

[0114] The dynamic adjustment selection gate dynamically filters and fuses key features based on temporal correlation, and the expression is:

[0115]

[0116]

[0117] in For time step The output of the dynamic feature selection gate, It is the Sigmoid activation function. This is the weight matrix. For time step The hidden state, For time step The input vector, For bias terms, For time step The fused feature vector;

[0118] The time decay injection layer uses a time decay factor. Assign time-time sensitive weights to time-series data. , For learnable parameters, For time intervals, Index for time-aware units. The number of time-aware units;

[0119] The multi-head LSTM unit captures the dependencies of pest and disease data at different time scales in parallel, and the expression is:

[0120]

[0121]

[0122]

[0123] in For time step The hidden state vector after memory enhancement To step the time Global Time Series Knowledge Base Mapping to the hidden state vector Weight matrices of the same dimension For the first Each time-sensing unit at time step The hidden state vector, To update the global time series knowledge base Gated loop unit, For multi-head time sensing units at time steps Hidden state splicing, For the first A long short-term memory network;

[0124] The spatiotemporal attention layer identifies the correlation patterns between key time nodes and spatial dimensions;

[0125] The fully connected layer connects to the spatiotemporal attention layer to generate a predicted degree of pest and disease infestation; the degree of pest and disease infestation includes pest and disease type and level.

[0126] Historical plant state data is used as a comprehensive set of plant states. Random forest is used to divide the comprehensive set of plant states into a plant state training set and a plant state test set in a 6:4 ratio.

[0127] A plant growth assessment model is constructed, which specifically includes a multi-source fusion layer, a deep residual block, a cross-attention layer, a multi-task prediction head, a hybrid loss calculation layer, and an output layer.

[0128] The multi-source fusion layer is used to align the spatiotemporal resolution of different sensor data; the deep residual block is used to extract higher-order nonlinear growth features of plants; the cross-attention layer establishes a dynamic correlation between environmental factors and growth indicators through a cross-attention mechanism; the multi-task prediction head outputs predictions of plant growth status in parallel; the plant growth status includes growth rate, energy conversion, and senescence degree.

[0129] The hybrid loss calculation layer comprehensively optimizes prediction accuracy and biological rationality through a hybrid adaptive loss function, the expression of which is:

[0130]

[0131] in For hybrid adaptive loss function, The weights of the gradient direction penalty term are the coefficients. These are the weighting coefficients for the biological constraint term. This represents the current number of days of growth. The maximum number of days for growth. Mean square error, This is a gradient direction penalty term. For activation function, For the true value Over time rate of change, For predicted values Over time rate of change, This is a biological constraint term, reflecting the energy conservation among photosynthetic accumulation, respiratory consumption, and biomass increase.

[0132] In the actual assessment, 800 sets of data were selected from historical pest and disease data as a comprehensive pest and disease set. A random forest was used to divide the data into a training set (480 sets) and a test set (320 sets) at a ratio of 6:4. The data was then processed at time steps. Taking dynamic filtering and fusion of key features as an example, the hidden state at time step 19 is: The input vector for time step 20 is The output of the dynamic feature selection gate at time step 20 is calculated. , fused feature vector Taking the calculation of the time decay factor of the third time-sensing unit in time step 20 as an example, the learning parameters of the third time-sensing unit are taken. Time interval ,calculate Calculate the time decay factor of each time sensing unit, based on the fused feature vector after time step 20. The hidden state vector of each time-aware unit is updated at time step 19 with the memory-enhanced hidden state vector at time step 20. ;

[0133] Input the data of potential pests and diseases to be predicted into the pest and disease prediction model, and generate the predicted pest and disease severity in the apple orchard for a period of time in the fully connected layer (red spider mite infestation level is moderate, leaf damage rate is 20%; codling moth infestation level is mild, fruit damage rate is expected to be 10%).

[0134] Eighty hundred data points were selected from historical plant state data to form a comprehensive plant state set, which was then divided into a plant state training set and a plant state test set in a 6:4 ratio. The maximum number of growing days was used in the hybrid loss calculation layer. Gradient direction penalty term weight coefficient Weighting coefficients of biological constraint terms Training and validating plant growth assessment models;

[0135] Plant status data is input into the plant growth assessment model, and the multi-task prediction head outputs the predicted plant growth status in parallel: apple tree height growth rate increases by 0.5 cm / week, trunk diameter growth rate increases by 0.1 cm / week, fruit diameter growth rate increases by 0.3 cm / week, and fruit sugar accumulation rate increases by 0.5 degrees Bx / day.

[0136] In this embodiment, the method for obtaining a reference plant management strategy includes:

[0137] The reference strategy database data is coarsely screened based on environmental meteorology from plant protection sensing data, and then finely screened based on spatiotemporal data from plant protection sensing data.

[0138] The comprehensive similarity of feature vectors between the predicted severity of pests and diseases and the historical severity of pests and diseases in the reference strategy library is calculated. The highest comprehensive similarity is taken as the pest and disease similarity, and the historical plant management strategy corresponding to the highest comprehensive similarity is taken as the first candidate plant management strategy. The calculation steps of the comprehensive similarity are as follows: calculate the average value of time series similarity, spatial propagation similarity and environmental matching degree to obtain spatiotemporal environmental similarity, and calculate the product of spatiotemporal environmental similarity and cosine similarity to obtain the comprehensive similarity.

[0139] Calculate the comprehensive similarity of feature vectors between the predicted plant growth status and the historical plant growth status in the reference strategy library, take the highest comprehensive similarity as the plant growth status similarity, and take the historical plant management strategy corresponding to the highest comprehensive similarity as the second candidate plant management strategy.

[0140] The first candidate weight and the second candidate weight are determined according to the ratio of the similarity between pests and diseases and the similarity between plant growth status. The first candidate plant management strategy and the second candidate plant management strategy are weighted and fused according to the first candidate weight and the second candidate weight to obtain the reference plant management strategy.

[0141] In the actual assessment, in an 80-acre apple orchard, based on the plant protection sensing data to be predicted (environmental meteorological data: temperature 25℃, humidity 65%, light intensity 7500 lux, time: July, latitude and longitude: 118°E / 37°N, altitude: 200 meters), the environmental coarse screening range (20-30℃, humidity range: 60-70%, light intensity range: 6000-8000 lux) and the spatiotemporal fine screening range (month range: June-August, latitude and longitude range: 115°E-120°E / 35°N-38°N, altitude range: 100-300 meters) of the reference strategy library data were determined.

[0142] The highest comprehensive similarity of the feature vectors between the predicted severity of pests and diseases and the historical severity of pests and diseases in the reference strategy database was calculated to be 0.8*0.88=0.704. Among these, the spatiotemporal environmental similarity was (0.8+0.75+0.85) / 3=0.8, and the cosine similarity was 0.88, with temporal series similarity (the chronological order and duration of pest and disease occurrence), spatial propagation similarity (the range and speed of pest and disease spread in the orchard), and environmental matching degree (temperature, humidity, and light intensity) at 0.85. OPPT technology was implemented, using chitosan ecological membranes to isolate pests and diseases, inhibiting egg implantation and hatching, preventing spore germination and reproduction, using carvacrol to kill pests and diseases, and then using chitosan-glucan for biomimetic stress to induce plant resistance to pests and diseases and promote wound repair.

[0143] The highest comprehensive similarity of the feature vectors between the predicted pest and disease severity and the historical pest and disease development severity in the reference strategy library was calculated to be 0.8*0.88=0.6375, of which the spatiotemporal environmental similarity was (0.75+0.7+0.8) / 3=0.75, the cosine similarity was 0.85, the time series similarity was 0.75, the spatial propagation similarity was 0.7, and the environmental matching degree was 0.8. The second candidate plant management strategy was obtained: In terms of physical control, one trap light (power of 50W) was used with 15 sticky insect boards, 2 sets were set up per acre, and one sonic insect repellent device (power of 80W) was placed per acre; In terms of biological control, lacewings were released at a density of 150 lacewings / acre; In terms of OPPT control, 60% of the area was sprayed with a 5% concentration of plant-derived insecticide (carvacrol) by drone, and 40% of the area was injected with a 0.18% concentration of plant-derived insecticide by plant injection.

[0144] The ratio of pest and disease similarity (0.704) to plant growth status similarity (0.6375) was calculated to obtain the first candidate weight of 0.525 and the second candidate weight of 0.475. Based on the first candidate weight and the second candidate weight, the first candidate plant management strategy and the second candidate plant management strategy were weighted and fused to obtain the reference plant management strategy: For physical control, one trap lamp (power of 50W) was used with 12 sticky insect boards, 2.5 sets were set up per acre, and one sonic insect repellent device (power of 80W) was placed per acre; For biological control, predatory mites were released at a density of 105 per acre and lacewings were released at a density of 71 per acre; For OPPT control, 65% of the area was sprayed with a 5% concentration of plant-derived insecticide (carvacrol) by drone, and 35% of the area was injected with a 0.19% concentration of plant-derived insecticide by plant injection.

[0145] In this embodiment, the method for determining the plant management objective function includes:

[0146] The control and management period is determined based on the severity of pests and diseases. The frequency of operation of different plant management strategies within the control and management period is calculated based on the effective control time of different plant management strategies during the control and management period. The control and management period includes an emergency period, a control period, and a maintenance period.

[0147] The objective function for plant management is determined based on the management costs, ecological impacts, and the time it takes for the strategies to take effect during each prevention and control management period. The expression is as follows:

[0148]

[0149] in Let the objective function be plant management. for The length of the prevention and control management period for Effective prevention and control time of management strategies for Prevention and control period The operational intensity of management strategies For the corresponding standard operating intensity, for The cost of management strategies for The area of ​​the management strategy for The effective area of ​​a single device in the management strategy This is the ecological cost conversion factor. For the production of biological control Ecological impact index For the number of ecological impact categories, for Disinfection Management Period The time it takes for management strategies to take effect;

[0150] The constraints of the plant management objective function are determined based on different plant management strategies, and the expression is as follows:

[0151]

[0152] in For strategy Disinfection efficiency for The severity of pests and diseases required during the pest and disease control period. , For equipment The rated maximum power and rated minimum power, For equipment The percentage of time the device is open. For equipment power, natural enemy Ecological impact coefficient, For the quantity released, For ecological carrying capacity threshold, The application concentration of OPPT , The minimum and maximum application concentrations of OPPT;

[0153] In actual assessment, the management period was determined based on the predicted severity of pests and diseases: 7 days for the emergency period, 15 days for the control period, and 20 days for the maintenance period. The effective period of each management measure was 1 day for trap lamps + sticky boards, 3 days for sonic insect repellent equipment, 15 days for biological control, and 60 days for OPPT. The operation frequency (rounded) of each management measure in different management periods was calculated as 7 / 15 / 20, 3 / 5 / 7, 1 / 1 / 2, and 1 / 3 / 3.

[0154] Disinfection Management Period Operational intensity of management strategies In physical control, the intensity of operation is directly proportional to the power of the equipment; in biological control, the intensity of operation is directly proportional to the insect release density; and in OPPT control, the intensity of operation is related to the proportion and concentration of OPPT used. Cost of management strategy In physical control, the cost of management strategies is related to equipment purchase and electricity costs; in biological control, the cost is related to natural enemy procurement and monitoring costs; and in OPPT control, the cost is related to pesticide costs and manual operation costs. The number of ecological impact categories is also relevant. This includes the impacts on biodiversity and other vegetation;

[0155] Ecological cost conversion factor The required concentration is 0.15. The pest and disease severity requirements are 0.7 for the emergency period, 0.5 for the controller, and 0.3 for the maintenance period. The power range of the trapping lamp is 30-60W, the power range of the sonic insect repellent device is 50-100W, the ecological carrying capacity threshold is 0.175, and the concentration of the pesticide used in OPPT technology is 0.1-0.25%.

[0156] In this embodiment, the method for obtaining the optimal plant management strategy includes:

[0157] Based on the objective function of plant management, the reference plant management strategy is optimized using Nash equilibrium to obtain the optimal plant management strategy. The specific steps are as follows:

[0158] Based on the rational participants in the game defined by management strategies, participant 1 is identified as physical control, participant 2 as biological control, and participant 3 as OPPT control. The strategy variables for each participant are determined. The game payoff function is determined based on the elimination effect of each management strategy and the plant management objective function. The strategy set is determined based on the constraints of the plant management objective function. The strategy variables for participant 1 are equipment density and equipment power; the strategy variables for participant 2 are insect release density and insect species; and the strategy variables for participant 3 are operation type ratio and OPPT concentration.

[0159] The equilibrium point is solved by the asynchronous gradient response method. The existence of Nash equilibrium is proved by Shizuo Kakutani's fixed point theorem. The strategy variables of the participants are repeatedly adjusted until the game payoff function is continuous and quasi-concave in the strategy space and the plant management objective function is maximized. The optimal combination of strategy variables is then output to obtain the optimal plant management strategy.

[0160] The optimal plant management strategy and the corresponding prediction of pest and disease severity and plant growth status are used to update the reference strategy library;

[0161] In the actual assessment, the Nash equilibrium optimization reference plant management strategy was adopted based on the plant management objective function. Participant 1 was identified as physical control (including sound wave repelling, light trapping + insect nets / sticky boards), participant 2 as biological control (release of natural enemy insects, including predatory mites and lacewings), and participant 3 as OPPT control (including drone spraying and plant injection).

[0162] The game payoff function includes physical control payoff, biological control payoff, and OPPT control payoff. Physical control payoff is determined by the physical control effect and the management cost and strategy take-off time in the plant management objective function. Biological control payoff is determined by the biological control effect and the management cost, ecological impact, and strategy take-off time in the plant management objective function. OPPT control payoff is determined by the OPPT control effect and the management cost and strategy take-off time in the plant management objective function.

[0163] The reference plant management strategy is set as the initial strategy. The gradient (payout / operation cost) of each participant is calculated, and the asynchronous (strategy) of each participant is updated based on their gradient. The asynchronous gradient response method is repeated to update the strategy variables until the game payoff function is continuous and quasi-concave in the strategy space, and the plant management objective function is minimized. Based on plant protection sensing data from an 80-acre apple orchard on July 15, 2024, the optimal plant management strategy is output: For physical control, one 50W trap light is paired with 20 sticky insect boards, with 3 traps set per acre. For group control, the operation frequency was 4 / 8 / 10, with one sonic insect repellent device (power of 75W) placed per acre, and the operation frequency was 4 / 8 / 10; for biological control, predatory mites were released at a density of 120 per acre and lacewings were released at a density of 65 per acre, with the operation frequency being 1 / 1 / 2; for OPPT control, 65% of the area was sprayed with a 5% concentration of plant-derived insecticide (carvacrol) by drone, with the operation frequency being 1 / 3 / 3, and 35% of the area was injected into the plants with a concentration of 0.19% plant-derived insecticide, with the operation frequency being 1 / 1 / 2;

[0164] The optimal plant management strategy and the corresponding prediction of pest and disease severity and plant growth status are used to update the reference strategy library.

[0165] Secondly, a zero-pesticide residue safe plant protection management system based on intelligent monitoring includes:

[0166] Management strategy module: used to classify historical plant protection information by clustering, construct a reference strategy library based on the clustering results, obtain reference plant management strategies by matching the reference strategy library with the predicted severity of pests and diseases and the predicted plant growth status, and update the reference strategy library based on the optimal plant management strategy;

[0167] Sensing data module: used to build a zero-pesticide residue plant protection monitoring network to acquire plant protection sensing data, and to perform multi-source data fusion and data refinement on plant protection sensing data to obtain pest and disease risk data and plant status data;

[0168] Model prediction module: used to construct a pest and disease prediction model and a plant growth assessment model based on the clustering results, and input the pest and disease risk data and the plant status data into the pest and disease prediction model and the plant growth assessment model respectively to obtain the predicted pest and disease severity and the predicted plant growth status;

[0169] Strategy optimization module: used to determine the plant management objective function, and optimize the reference plant management strategy according to the plant management objective function to obtain the optimal plant management strategy;

[0170] Management platform module: used to view, store and manage the predicted severity of pests and diseases, the predicted plant growth status and the optimal plant management strategy, and to perform zero-pesticide residue pest and disease control operations based on the predicted severity of pests and diseases, the predicted plant growth status and the optimal plant management strategy.

[0171] The above description is only a preferred embodiment of the present invention, and is particularly suitable for the prevention and control of citrus Huanglongbing, strawberry zero-pesticide residue plant protection management procedures, watermelon zero-pesticide residue plant protection management procedures, fragrant pear zero-pesticide residue plant protection management procedures, tomato zero-pesticide residue plant protection management procedures, etc. It is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for zero-pesticide residue safe plant protection management based on intelligent monitoring, characterized in that, Includes the following steps: S1. Classify historical plant protection information by clustering, and construct a reference strategy library based on the clustering results; The clustering results include historical pest and disease data, historical plant data, and historical plant management strategies. S2. Construct a monitoring network for zero pesticide residue safe plant protection technology (OPPT) to obtain plant protection sensing data, and perform multi-source data fusion and data extraction on the plant protection sensing data to obtain pest and disease risk data and plant status data; S3. Construct a pest and disease prediction model and a plant growth assessment model based on the clustering results. Input the pest and disease risk data and the plant status data into the pest and disease prediction model and the plant growth assessment model respectively to obtain the predicted pest and disease severity and the predicted plant growth status. S4. Based on the predicted severity of pests and diseases and the predicted plant growth status, match the reference strategy library to obtain a reference plant management strategy; S5. Determine the OPPT management period based on the severity of pests and diseases, determine the operation frequency of different types of plant management strategies based on the OPPT management period, determine the plant management objective function, optimize the reference plant management strategy based on the plant management objective function to obtain the optimal plant management strategy, and update the reference strategy library. The method for constructing the reference strategy library includes: Historical plant protection information is categorized into historical pest and disease data, historical plant status data, and historical plant management strategies through clustering. The historical pest and disease data includes pest and disease monitoring data, pest and disease indicators, environmental meteorological data, soil biological information, and the corresponding historical pest and disease development levels. The historical plant status data includes comprehensive plant growth characteristics, environmental meteorological data, soil nutrient information, and the corresponding historical plant growth status. The historical plant management strategies are used to select appropriate OPPT (Optimized Plant Protection Test) procedures based on the historical plant growth status and the severity of historical pests and diseases, including implementation parameters, implementation frequency, and implementation paths for different types of OPPT. Based on geographical coordinates and plant categories, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are divided into the same strategy area. Within the same strategy area, historical pest and disease development levels, historical plant growth status, and historical plant management strategies are matched and associated according to time periods to form a reference strategy library. The time period matching rules for the data in the reference strategy library are as follows: when the overlap between the time period of historical pest and disease development and the time period of historical plant growth status is greater than 80%, the data will be associated; when the time interval between the start time of historical pest and disease development and the start time of historical plant management strategy is less than the remedial time threshold, the data will be associated; when the historical pest and disease development is associated with both historical plant growth status and historical plant management strategy, a set of reference strategies will be formed; otherwise, the data will be defined as invalid. The reference strategy library consists of reference strategies.

2. The method for zero-pesticide residue safe plant protection management based on intelligent monitoring according to claim 1, characterized in that, The method for obtaining data on potential pest and disease risks and plant status data includes: An OPPT monitoring network is constructed to acquire plant protection sensing data; the OPPT monitoring network includes a hyperspectral imaging sensor, a biosensor, a high-definition camera, various environmental parameter sensors, and biochemical analysis experiments; A hyperspectral imaging sensor acquires plant spectral reflectance information, and performs data inversion on the spectral reflectance information to obtain chlorophyll concentration and first plant activity. The first plant growth stage is determined based on the chlorophyll concentration; the first plant activity reflects the degree of damage to plant cell structure. Biosensors are used to monitor the concentration and types of pest-related microorganisms in the air in real time to obtain primary pest monitoring data. A high-definition camera acquires high-definition plant images, and the high-definition plant images are processed to obtain first plant features and second plant features. The second plant growth stage is determined by matching the first plant features with a plant growth domain library. The first plant features represent the plant's growth shape and size. The second plant features represent the plant's external appearance characteristics affected by pests and diseases. Plant growth climate environment is obtained through various environmental parameter sensors, and soil nutrient information and primary soil biological information are obtained through biochemical analysis experiments. The third plant growth stage is obtained by taking the union of the first and second plant growth stages. The plant activity range is determined based on the third plant growth stage. The second plant activity is obtained by crossing the first plant activity with the second plant characteristics. The third plant activity is determined based on the plant activity range and the second plant activity. The comprehensive plant growth characteristics are composed of the third plant growth stage and the third plant activity. The plant activity range represents the range of various activities corresponding to the growth stage. The third plant activity reflects the overall performance of the plant's physiological and biochemical functions, antioxidant activity, stress resistance, and metabolic activities. Based on the second plant characteristics, pest and disease indicators are determined, and the second plant characteristics are used to correct the first soil biological information and the first pest and disease monitoring data to obtain the second soil biological information and the second pest and disease monitoring data. The pest and disease indicators, plant growth climate environment, second soil biological information, and second pest and disease monitoring data are combined to form pest and disease risk data, and the comprehensive plant growth characteristics, plant growth climate environment, and soil nutrient information are combined to form plant status data.

3. The method for zero-pesticide residue safe plant protection management based on intelligent monitoring according to claim 1, characterized in that, The methods for obtaining predictions of pest and disease severity and plant growth status include: Historical pest and disease data were used as a comprehensive pest and disease set. Random forest was used to divide the comprehensive pest and disease set into a pest and disease training set and a pest and disease test set in a ratio of 6:

4. The pest and disease prediction model is constructed, which specifically includes dynamically adjusting the selection gate, time decay injection layer, multi-head LSTM unit, spatiotemporal attention layer and fully connected layer; The dynamic adjustment selection gate dynamically filters and fuses key features based on temporal correlation, and the expression is: in For time step The output of the dynamic feature selection gate, It is the Sigmoid activation function. This is the weight matrix. For time step The hidden state, For time step The input vector, For bias terms, For time step The fused feature vector; The time decay injection layer uses a time decay factor. Assign time-time sensitive weights to time-series data. , For learnable parameters, For time intervals, Index for time-aware units. The number of time-aware units; The multi-head LSTM unit captures the dependencies of pest and disease data at different time scales in parallel, and the expression is: in For time step The hidden state vector after memory enhancement To step the time Global Time Series Knowledge Base Mapping to the hidden state vector Weight matrices of the same dimension For the first Each time-sensing unit at time step The hidden state vector, To update the global time series knowledge base Gated loop unit, For multi-head time sensing units at time steps Hidden state splicing, For the first A long short-term memory network; The spatiotemporal attention layer identifies the correlation patterns between key time nodes and spatial dimensions; The fully connected layer connects to the spatiotemporal attention layer to generate a predicted degree of pest and disease infestation; the degree of pest and disease infestation includes pest and disease type and level. Historical plant state data is used as a comprehensive set of plant states. Random forest is used to divide the comprehensive set of plant states into a plant state training set and a plant state test set in a 6:4 ratio. A plant growth assessment model is constructed, which specifically includes a multi-source fusion layer, a deep residual block, a cross-attention layer, a multi-task prediction head, a hybrid loss calculation layer, and an output layer. The multi-source fusion layer is used to align the spatiotemporal resolution of different sensor data; the deep residual block is used to extract higher-order nonlinear growth features of plants; the cross-attention layer establishes a dynamic correlation between environmental factors and growth indicators through a cross-attention mechanism; the multi-task prediction head outputs predictions of plant growth status in parallel; the plant growth status includes growth rate, energy conversion, and senescence degree. The hybrid loss calculation layer comprehensively optimizes prediction accuracy and biological rationality through a hybrid adaptive loss function, the expression of which is: in For a hybrid adaptive loss function, The weights of the gradient direction penalty term are the coefficients. These are the weighting coefficients for the biological constraint term. This represents the current number of days of growth. The maximum number of days for growth. Mean square error, This is a gradient direction penalty term. For activation function, For the true value Over time rate of change, For predicted values Over time rate of change, This is a biological constraint term, reflecting the energy conservation among photosynthetic accumulation, respiratory consumption, and biomass increment.

4. The method for zero-pesticide residue safe plant protection management based on intelligent monitoring according to claim 1, characterized in that, The method for obtaining reference plant management strategies includes: The reference strategy database data is coarsely screened based on environmental meteorology from plant protection sensing data, and then finely screened based on spatiotemporal data from plant protection sensing data. The comprehensive similarity of feature vectors between the predicted severity of pests and diseases and the historical severity of pests and diseases in the reference strategy library is calculated. The highest comprehensive similarity is taken as the pest and disease similarity, and the historical plant management strategy corresponding to the highest comprehensive similarity is taken as the first candidate plant management strategy. The calculation steps of the comprehensive similarity are as follows: calculate the average value of time series similarity, spatial propagation similarity and environmental matching degree to obtain spatiotemporal environmental similarity, and calculate the product of spatiotemporal environmental similarity and cosine similarity to obtain the comprehensive similarity. Calculate the comprehensive similarity of feature vectors between the predicted plant growth status and the historical plant growth status in the reference strategy library, take the highest comprehensive similarity as the plant growth status similarity, and take the historical plant management strategy corresponding to the highest comprehensive similarity as the second candidate plant management strategy. The first candidate weight and the second candidate weight are determined according to the ratio of the similarity between pests and diseases and the similarity between plant growth status. The first candidate plant management strategy and the second candidate plant management strategy are weighted and fused together according to the first candidate weight and the second candidate weight to obtain the reference plant management strategy.

5. The method for zero-pesticide residue safe plant protection management based on intelligent monitoring according to claim 1, characterized in that, The method for determining the objective function of plant management includes: The OPPT management period is determined based on the severity of pests and diseases. The frequency of operation of different plant management strategies within the OPPT management period is calculated based on the OPPT management period and the effective OPPT implementation time of different plant management strategies. The OPPT management period includes the emergency period, the control period, and the maintenance period. The objective function for plant management is determined based on the management costs, ecological impacts, and strategy effectiveness time during each OPPT management period. The expression is as follows: in Let the objective function be plant management. for The length of the disinfection management period, for Effective disinfection time of management strategies for Disinfection Management Period The operational intensity of management strategies For the corresponding standard operating intensity, for The cost of management strategies for The area of ​​the management strategy for The effective area of ​​a single device in the management strategy This is the ecological cost conversion factor. To use OPPT to control the generation Ecological impact index For the number of ecological impact categories, for OPPT Management Period The time it takes for management strategies to take effect; The constraints of the management objective function are determined based on the OPPT management strategy for different crops, and the expression is as follows: in For strategy OPPT efficiency for The required level of pests and diseases during the OPPT management period. , For equipment The rated maximum power and rated minimum power, For equipment The percentage of time the device is open. For equipment power, natural enemy Ecological impact coefficient, For the quantity released, For ecological carrying capacity threshold, The application concentration of OPPT , These are the minimum and maximum application concentrations of OPPT.

6. The method for zero-pesticide residue safe plant protection management based on intelligent monitoring according to claim 1, characterized in that, The method for obtaining the optimal plant management strategy includes: Based on the OPPT management objective function, the optimal plant management strategy is obtained by optimizing the reference plant management strategy using Nash equilibrium. The specific steps are as follows: Based on the rational participants in the game defined by management strategies, participant 1 is identified as physical control, participant 2 as biological control, and participant 3 as OPPT control. The strategy variables for each participant are determined. The game payoff function is determined based on the pest and disease control effects of each management strategy and the plant management objective function. The strategy set is determined based on the constraints of the plant management objective function. The strategy variables for participant 1 are equipment density and equipment power; the strategy variables for participant 2 are insect release density and insect species; and the strategy variables for participant 3 are the proportion of operation types and the proportion and concentration of OPPT product used. The equilibrium point is solved by the asynchronous gradient response method. The existence of Nash equilibrium is proved by Shizuo Kakutani's fixed point theorem. The strategy variables of the participants are repeatedly adjusted until the game payoff function is continuous and quasi-concave in the strategy space and the plant management objective function is maximized. The optimal combination of strategy variables is then output to obtain the optimal plant management strategy. The optimal plant management strategy and the corresponding prediction of pest and disease severity and plant growth status are used to update the reference strategy library.

7. A zero-pesticide residue safe plant protection management system based on intelligent monitoring, used to execute the method according to any one of claims 1-6, characterized in that, include: Management strategy module: used to classify historical plant protection information by clustering, construct a reference strategy library based on the clustering results, obtain reference management strategies by matching the reference strategy library with the predicted severity of pests and diseases and the predicted plant growth status, and update the reference strategy library based on the optimal plant management strategy; Sensing data module: used to build OPPT monitoring network to acquire plant protection sensing data, and to perform multi-source data fusion and data refinement on plant protection sensing data to obtain pest and disease risk data and plant status data; Model prediction module: used to construct a pest and disease prediction model and a plant growth assessment model based on the clustering results, and input the pest and disease risk data and the plant status data into the pest and disease prediction model and the plant growth assessment model respectively to obtain the predicted pest and disease severity and the predicted plant growth status; Strategy optimization module: used to determine the plant management objective function, and optimize the reference plant management strategy according to the plant management objective function to obtain the optimal plant management strategy; Management platform module: used to view, store and manage the predicted pest and disease severity, the predicted plant growth status and the optimal plant management strategy, and to perform OPPT operations based on the predicted pest and disease severity, the predicted plant growth status and the optimal plant management strategy.

Images