Wetland disease and pest monitoring method and device, electronic equipment and storage medium

Abstract

The invention discloses a wetland disease and pest monitoring method and device, electronic equipment and a storage medium, and relates to the technical field of ecological environment monitoring or other related fields, and the monitoring method comprises the steps: collecting environment sample data of a target monitoring region in a first time period; the environment sample data is input into the first model, pest reproduction information is output, and the pest database is used for recording the mapping relation between the reproduction rate of pests and the living environment; inputting the disease and pest reproduction information and the environmental sample data into a second model, and outputting a disease and pest prediction density of the target monitoring area in a second time period which is a future time period of the first time period; and determining a pest treatment scheme based on the pest prediction density. The technical problem that the requirements of early warning and accurate prevention and control of diseases and insect pests cannot be met due to the fact that single model recognition cannot effectively monitor the wetland diseases and insect pests in related technologies is solved.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring technology, and more specifically, to a method and device for monitoring wetland pests and diseases, electronic equipment, and storage medium. Background Technology

[0002] Wetland ecosystems possess unparalleled biodiversity and ecosystem service value, including but not limited to vital functions such as climate regulation, water purification, flood control and disaster reduction, and providing habitats and breeding grounds for wild flora and fauna. However, the invasion of pests and diseases severely impacts the ecological balance and biodiversity of wetlands. Traditional management methods often involve emergency treatment only after symptoms appear, which is not only inefficient but also fails to effectively control the spread of pests and diseases, potentially causing irreversible damage to wetland ecosystems in the long run.

[0003] Current technologies commonly employ methods that directly utilize collected data for model training to identify the presence of pests and diseases. However, this approach lacks the ability to predict pest and disease development trends and relies on the identification results of a single model, leading to instability in identification. Other methods can detect existing pest and disease conditions, but similarly lack the ability to predict future pest and disease dynamics, thus offering insufficient support for prevention and early intervention. Furthermore, existing technologies often fail to comprehensively analyze the reproductive conditions and living environment of pests and diseases, resulting in a lack of targeted and forward-looking pest and disease management measures.

[0004] In summary, existing technologies have significant shortcomings in pest and disease monitoring and prediction. They lack effective prediction models to assess the development trend of pests and diseases, and the identification results of a single model are unstable, which affects the efficiency and effectiveness of pest and disease management.

[0005] There is currently no effective solution to the above problems. Summary of the Invention

[0006] This invention provides a method, device, electronic equipment, and storage medium for monitoring wetland pests and diseases, at least to solve the technical problem in related technologies where single-model identification cannot effectively monitor wetland pests and diseases, thus failing to meet the needs for early warning and precise prevention and control of pests and diseases.

[0007] According to one aspect of the present invention, a method for monitoring wetland pests and diseases is provided, comprising: collecting environmental sample data of a target monitoring area within a first time period, wherein the environmental sample data is used to describe various survival indicators of the target monitoring area as a habitat for pests and diseases; inputting the environmental sample data into a first model and outputting pest and disease reproduction information, wherein the first model is a Naive Bayes model pre-trained based on a pest and disease database, the pest and disease database being used to record the mapping relationship between the reproduction rate of pests and diseases and their habitat; inputting the pest and disease reproduction information and the environmental sample data into a second model and outputting a predicted pest and disease density of the target monitoring area within a second time period, wherein the second model is a time-series prediction model pre-trained based on an improved activation function, the second time period being a future time period of the first time period; and determining a pest and disease treatment plan based on the predicted pest and disease density.

[0008] Further, the step of collecting environmental sample data of the target monitoring area within a first time period includes: receiving and responding to data acquisition signals through a pre-arranged sensor array, wherein the sensor array includes at least: a clock sensor, a temperature sensor, a humidity sensor, a rainfall sensor, a vegetation monitoring sensor group, a biological monitoring sensor group, and an insect lamp; according to the first time period and acquisition frequency indicated by the data acquisition signal, simultaneously controlling the clock sensor, temperature sensor, humidity sensor, rainfall sensor, vegetation monitoring sensor group, biological monitoring sensor group, and insect lamp to record the acquisition time point, temperature, humidity, rainfall, vegetation coverage, specified biological coverage, and actual density of pests and diseases, respectively, to obtain a raw dataset; preprocessing the raw data in the raw dataset to obtain the environmental sample data, wherein the preprocessing includes the following operations: data association, data cleaning, data standardization, and data fusion.

[0009] Further, the first model is trained through the following steps: acquiring historical regional observation data and establishing the pest and disease database based on the historical regional observation data, wherein the historical regional observation data includes pest and disease survival environment data, pest and disease reproduction rate, and pest and disease predicted density development trend within a historical time period; extracting a training set from the pest and disease database, obtaining the Naive Bayes model as the first initial model, and using the pest and disease survival environment data in the training set as input data to iteratively train the first initial model; during the iterative training process, using the pest and disease reproduction rate in the training set as the fitting target, adjusting the parameters of the first initial model; and stopping the iterative training when the expected fitting result is obtained, and determining the corresponding first initial model as the first model.

[0010] Further, the step of inputting the environmental sample data into the first model and outputting pest and disease reproduction information includes: inputting the environmental sample data into the input layer of the first model; calculating conditional probabilities based on the environmental sample data and Bayes' theorem through the computation layer of the first model to obtain the category probabilities corresponding to N pest and disease categories, where N is a positive integer; for each pest and disease category, analyzing the pest and disease reproduction rate under the condition that the target monitoring area is the pest and disease survival environment based on the environmental sample data; generating pest and disease reproduction information containing the pest and disease category, the category probability, and the pest and disease reproduction rate, and outputting the pest and disease reproduction information through the output layer of the first model.

[0011] Further, the second model is obtained through the following steps: extracting a training set from the pest and disease database, and establishing a second initial model based on a time-series prediction model and an improved activation function, wherein the improved activation function is obtained by introducing a positive slope into a specified negative interval of the ReLU activation function; using the pest and disease survival environment data and pest and disease reproduction rate in the training set as input data, iteratively training the second initial model; during the iterative training process, using the development trend of the predicted density of pests and diseases in the training set as the fitting target, adjusting the parameters of the second initial model; and stopping the iterative training when the expected fitting result is obtained, and determining the corresponding second initial model as the second model.

[0012] Further, the step of inputting the pest and disease reproduction information and the environmental sample data into the second model and outputting the predicted pest and disease density in the second time period of the target monitoring area includes: inputting the pest and disease reproduction information obtained from the first model and the collected environmental sample data into the input layer of the second model; mapping and encoding the pest and disease reproduction information and the environmental sample data through the encoding layer of the second model to obtain an encoded information set, wherein the encoded information includes at least the encoded vector corresponding to the pest and disease reproduction information and M time-series encoded segments corresponding to the environmental sample data, where M is a positive integer; performing a pest and disease predicted density development analysis on an extended time scale based on a self-attention mechanism and a residual network through the fitting layer of the second model to obtain a density-time fitting curve, wherein the extended time scale includes at least the union of the first time period and the second time period; extracting the pest and disease predicted density in the second time period based on the density-time fitting curve, and outputting the pest and disease predicted density through the output layer of the second model.

[0013] Further, the step of determining a pest and disease treatment plan based on the predicted pest and disease density includes: obtaining a pre-set level threshold, wherein the level thresholds are arranged in descending order as: normal value > warning value > alert value > risk value; when the predicted pest and disease density matches the normal value, determining the pest and disease treatment plan as maintaining regular frequency monitoring; when the predicted pest and disease density matches the warning value, determining the pest and disease treatment plan as increasing the monitoring frequency according to a first preset granularity and taking preventive biological control measures; when the predicted pest and disease density matches the alert value, determining the pest and disease treatment plan as... The monitoring frequency is increased by a second preset particle size, and environmentally friendly chemical control agents are used for localized control. The second preset particle size is larger than the first preset particle size. The environmentally friendly chemical control agent refers to a chemical substance that meets the following conditions: its toxicity to non-pests and pests is lower than a preset toxicity threshold at a specified dosage, and its degradation cycle is shorter than a preset time threshold. When the predicted density of pests and diseases matches the risk value, the pest and disease treatment plan is determined to be to increase the monitoring frequency by a third preset particle size and use the environmentally friendly chemical control agent for global control. The third preset particle size is larger than the second preset particle size.

[0014] According to another aspect of the present invention, a wetland pest and disease monitoring device is also provided, comprising: a collection unit for collecting environmental sample data of a target monitoring area within a first time period, wherein the environmental sample data is used to describe various survival indicators of the target monitoring area as a pest and disease habitat; a first input unit for inputting the environmental sample data into a first model and outputting pest and disease reproduction information, wherein the first model is a Naive Bayes model pre-trained based on a pest and disease database, the pest and disease database being used to record the mapping relationship between the reproduction rate of pests and diseases and their habitat; a second input unit for inputting the pest and disease reproduction information and the environmental sample data into a second model and outputting a predicted pest and disease density of the target monitoring area within a second time period, wherein the second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period; and a determination unit for determining a pest and disease treatment plan based on the predicted pest and disease density.

[0015] Further, the acquisition unit includes: a receiving module, used to receive and respond to data acquisition signals through a pre-arranged sensor array, wherein the sensor array includes at least: a clock sensor, a temperature sensor, a humidity sensor, a rainfall sensor, a vegetation monitoring sensor group, a biological monitoring sensor group, and an insect lamp; a control module, used to simultaneously control the clock sensor, temperature sensor, humidity sensor, rainfall sensor, vegetation monitoring sensor group, biological monitoring sensor group, and insect lamp to record the acquisition time point, temperature, humidity, rainfall, vegetation coverage, specified biological coverage, and actual density of pests and diseases according to the first time period and acquisition frequency indicated by the data acquisition signal, to obtain a raw dataset; and a preprocessing module, used to preprocess the raw data in the raw dataset to obtain the environmental sample data, wherein the preprocessing includes the following operations: data association, data cleaning, data standardization, and data fusion.

[0016] Furthermore, the wetland pest and disease monitoring device further includes: a first training unit for training the first model, wherein the first training unit includes: a first establishment module for acquiring historical regional observation data and establishing the pest and disease database based on the historical regional observation data, wherein the historical regional observation data includes pest and disease survival environment data, pest and disease reproduction rate, and pest and disease predicted density development trend within a historical time period; a first iterative training module for extracting a training set from the pest and disease database, obtaining the Naive Bayes model as the first initial model, and iteratively training the first initial model using the pest and disease survival environment data in the training set as input data; a first parameter adjustment module for adjusting the parameters of the first initial model during iterative training, using the pest and disease reproduction rate in the training set as the fitting target; and a first determination module for stopping iterative training when the expected fitting result is obtained, and determining the corresponding first initial model as the first model.

[0017] Further, the first input unit includes: a first input module, used to input the environmental sample data into the input layer of the first model; a calculation module, used to perform conditional probability calculation based on the environmental sample data and Bayes' theorem through the calculation layer of the first model to obtain the category probabilities corresponding to N types of pests and diseases, where N is a positive integer; a first analysis module, used to analyze the pest and disease reproduction rate for each type of pest and disease based on the environmental sample data, using the target monitoring area as the pest and disease survival environment; and a first output module, used to generate pest and disease reproduction information containing the pest and disease type, the category probability, and the pest and disease reproduction rate, and output the pest and disease reproduction information through the output layer of the first model.

[0018] Furthermore, the wetland pest and disease monitoring device further includes: a second training unit for training a second model, wherein the second training unit includes: a second establishment module for extracting a training set from the pest and disease database and establishing a second initial model based on a time-series prediction model and an improved activation function, wherein the improved activation function is obtained by introducing a positive slope into a specified negative interval of the ReLU activation function; a second iterative training module for iteratively training the second initial model using pest and disease survival environment data and pest and disease reproduction rate in the training set as input data; a second parameter adjustment module for adjusting the parameters of the second initial model during iterative training, using the development trend of pest and disease prediction density in the training set as the fitting target; and a second determination module for stopping iterative training when the expected fitting result is obtained, and determining the corresponding second initial model as the second model.

[0019] Further, the second input unit includes: a second input module, used to input the pest and disease reproduction information obtained from the first model and the collected environmental sample data into the input layer of the second model; a mapping and encoding module, used to map and encode the pest and disease reproduction information and the environmental sample data through the encoding layer of the second model to obtain an encoded information set, wherein the encoded information includes at least the encoding vector corresponding to the pest and disease reproduction information and M time-series encoded segments corresponding to the environmental sample data, where M is a positive integer; a second analysis module, used to perform pest and disease prediction density development analysis on the encoded information set on an extended time scale through the fitting layer of the second model based on a self-attention mechanism and a residual network to obtain a density-time fitting curve, wherein the extended time scale includes at least the union of the first time period and the second time period; and a second output module, used to extract the pest and disease prediction density in the second time period based on the density-time fitting curve and output the pest and disease prediction density through the output layer of the second model.

[0020] Further, the determining unit includes: an acquisition module, used to acquire a pre-set level threshold, wherein the level thresholds are arranged in descending order as: normal value > warning value > alert value > risk value; a third determining module, used to determine the pest and disease treatment plan as maintaining regular frequency monitoring when the predicted pest and disease density matches the normal value; a fourth determining module, used to determine the pest and disease treatment plan as increasing the monitoring frequency according to a first preset granularity and taking preventive biological control measures when the predicted pest and disease density matches the warning value; and a fifth determining module, used to determine the pest and disease treatment plan when the predicted pest and disease density matches the alert value. To increase the monitoring frequency according to a second preset particle size and to use environmentally friendly chemical control agents for local control, wherein the second preset particle size is larger than the first preset particle size, and the environmentally friendly chemical control agent refers to a chemical substance that meets the following conditions: its toxicity to non-pests and pests is lower than a preset toxicity threshold at a specified dosage, and its degradation cycle is shorter than a preset time threshold; a sixth determining module is used to determine, when the predicted density of pests and diseases matches the risk value, that the pest and disease treatment plan is to increase the monitoring frequency according to a third preset particle size and to use the environmentally friendly chemical control agent for global control, wherein the third preset particle size is larger than the second preset particle size.

[0021] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the wetland pest and disease monitoring method described in any one of the above embodiments.

[0022] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the wetland pest and disease monitoring method described in any of the above embodiments.

[0023] According to another aspect of the present invention, a computer program product is also provided, including computer instructions, wherein when the computer instructions are executed by a processor, they implement the steps of the wetland pest and disease monitoring method described in any one of the above embodiments.

[0024] This invention proposes a method for monitoring wetland pests and diseases. First, environmental sample data is collected from the target monitoring area within a first time period. This environmental sample data describes various survival indicators of the target monitoring area as a habitat for pests and diseases. Then, the environmental sample data is input into a first model, which outputs pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database, which records the mapping relationship between pest and disease reproduction rates and their habitats. Next, the pest and disease reproduction information and the environmental sample data are input into a second model, which outputs a predicted pest and disease density for the target monitoring area within a second time period. This second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period. Finally, a pest and disease treatment plan is determined based on the predicted pest and disease density.

[0025] This invention employs a combination of comprehensive environmental analysis and intelligent prediction. By constructing a database of ecological characteristics of pests and diseases and integrating machine learning algorithms, it achieves the goal of accurately predicting the development trend of wetland pests and diseases, thereby realizing the technical effects of early warning and eco-friendly prevention and control of pests and diseases. Specifically, this invention collects environmental sample data covering key ecological elements affecting pest and disease reproduction in the target monitoring area within a first time period. Then, it identifies pest and disease reproduction information through a first model—a Naive Bayes model trained on a large amount of historical data. Deep analysis of the environmental sample data accurately outputs pest and disease reproduction information, revealing potential risks to the pest and disease's living environment. Next, the reproduction information is combined with real-time environmental data and input into a second model—a time-series prediction model with an improved activation function—to accurately predict pest and disease density in the second time period, effectively capturing the dynamic trend of pest and disease spread. Based on the predicted pest and disease density information, this invention further proposes a tiered response mechanism, automatically generating scientifically sound pest and disease treatment plans according to different risk levels. This ensures rapid response and effective control of wetland ecosystems in the face of pest and disease infestations, greatly enriching wetland protection measures, compensating for the shortcomings of traditional monitoring methods, and solving the technical problem that single-model identification in related technologies cannot effectively monitor wetland pests and diseases, thus failing to meet the needs for early warning and precise prevention and control of pests and diseases. Attached Figure Description

[0026] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0027] Figure 1 This is a flowchart of an optional wetland pest and disease monitoring method according to an embodiment of the present invention;

[0028] Figure 2 This is an optional flowchart for identifying the degree of pest and disease reproduction according to an embodiment of the present invention;

[0029] Figure 3 This is an optional flowchart for identifying pest and disease density according to an embodiment of the present invention;

[0030] Figure 4 This is an optional structural diagram of a pest and disease risk level classification and treatment plan according to an embodiment of the present invention;

[0031] Figure 5 This is an example of an optional treatment method for the East Asian migratory locust according to an embodiment of the present invention;

[0032] Figure 6 This is an example of an optional reed moth disease and pest treatment according to an embodiment of the present invention;

[0033] Figure 7 This is a schematic diagram of an optional wetland pest and disease monitoring device according to an embodiment of the present invention;

[0034] Figure 8 This is a structural block diagram of an electronic device for performing a method for monitoring wetland pests and diseases according to an embodiment of the present invention. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] To facilitate understanding of the present invention by those skilled in the art, some terms or nouns involved in the various embodiments of the present invention are explained below:

[0038] Naive Bayes Model is a classification method based on Bayes' theorem in probability theory. It assumes that features are independent of each other. In this invention, it is used to analyze ecological indicators of the living environment of pests and diseases, predict the reproduction rate of pests and diseases, and thus identify pest and disease reproduction information.

[0039] A sequential forecasting model is used to predict data sequences that change over time, capturing the temporal relationships and potential trends between data. This invention employs a time-series forecasting model based on the Transformer architecture, improving the prediction accuracy of future trends in pest and disease density through self-attention mechanisms and residual network optimization, particularly advantageous when processing long-sequence data.

[0040] Self-attention is a mechanism used to capture the importance relationships between different parts of the input data, even if these parts are far apart in the sequence. Self-attention helps models better understand the long-term dependencies in sequence data, thereby improving prediction accuracy.

[0041] The Modified Activation Function (REF) introduces non-linearity to help the model learn complex functions. This invention optimizes the traditional ReLU function, addressing the instability in predictions in the negative region, improving the model's generalization ability, enabling more stable data processing, reducing the gradient vanishing problem during model training, and ensuring the reliability of prediction results.

[0042] The following embodiments of the present invention can be applied to various systems / applications / equipment that require early warning of wetland pests and diseases and decision support for ecosystem protection. They enable a dynamic monitoring and risk assessment system for pests and diseases based on intelligent analysis of environmental factors and time-series prediction models. This invention uses a Naive Bayes model to identify pest and disease reproduction information, and then combines this with a time-series prediction model to predict future trends in pest and disease density. This allows for better adaptation to changes in the wetland environment and provides data-driven decision-making support for the protection of wetland biodiversity and the sustainable management of ecosystems.

[0043] In its specific implementation, this invention first constructs a database of ecological characteristics of pests and diseases, recording the mapping relationship between pest and disease reproduction rates and their living environment, including key environmental indicators such as temperature, humidity, rainfall, food coverage, and the number of natural enemies. Based on this database, a trained Naive Bayes model is used to analyze environmental sample data from the target monitoring area, outputting pest and disease reproduction information. Subsequently, the pest and disease reproduction information and environmental sample data are input into a second model—an improved time-series prediction model—which uses an optimized self-attention mechanism and activation function to predict the density of pests and diseases in the target area over a specific future time period. Finally, based on the risk level of the predicted density, corresponding pest and disease treatment plans are automatically generated, ranging from routine monitoring to environmentally friendly chemical control, and even to the implementation of drone-based pesticide spraying operations covering the entire area, ensuring rapid response and effective control of wetland ecosystems in the face of pest and disease infestations.

[0044] This invention, by combining comprehensive environmental analysis with intelligent prediction, not only overcomes the limitations of traditional monitoring methods in early warning of pests and diseases, but also solves the instability problem of single-model identification in pest and disease prediction under complex environmental conditions, significantly improving the efficiency and accuracy of wetland pest and disease management. Furthermore, the tiered response mechanism provided by this invention can take appropriate intervention measures for different stages of pest and disease development, minimizing disturbance to the wetland ecosystem and achieving a harmonious unity between pest and disease management and wetland ecological protection, providing strong scientific and technological support for wetland protection and restoration.

[0045] The present invention will now be described in detail with reference to various embodiments.

[0046] Example 1

[0047] According to an embodiment of the present invention, a method for monitoring wetland pests and diseases is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0048] The main body implementing this invention can be an intelligent wetland pest and disease monitoring and early warning system. Combining intelligent analysis of environmental factors and time-series prediction model technology, it is used in wetland ecological protection and management scenarios, especially for the early warning and precise prevention and control of pests and diseases. Through data-driven prediction algorithms, the specific implementation steps include building a pest and disease ecological characteristic database, identifying pest and disease reproduction information, predicting pest and disease density, and generating pest and disease treatment plans. This aims to improve the stability of wetland ecosystems and the efficiency of biodiversity protection. It can effectively utilize real-time and historical environmental data to assist wetland managers in responding quickly to potential pest and disease threats, reducing the risk of ecological damage, and promoting the healthy and sustainable development of the wetland environment.

[0049] The intelligent wetland pest and disease monitoring and early warning system first collects environmental sample data within the target monitoring area, including but not limited to temperature, humidity, rainfall, food coverage, and the number of natural enemies. After analysis using a Naive Bayes model, it outputs pest and disease reproduction information. Subsequently, through an improved time-series prediction model, combined with historical pest and disease density data and reproduction information, it predicts the future pest and disease density in the target area. This process effectively improves the accuracy and stability of the prediction by utilizing an optimized activation function and self-attention mechanism. Finally, based on different risk levels of the predicted density, it generates scientific and reasonable pest and disease treatment plans, including increasing the monitoring frequency during the early warning stage, localized chemical control during the alert stage, and comprehensive regional pesticide spraying during the risk stage, ensuring timely and effective pest and disease management.

[0050] Through this invention, the wetland management system can obtain accurate predictions of pest and disease dynamics, thereby taking targeted prevention and control measures to effectively protect the biodiversity and ecological functions of wetlands, reduce negative environmental impacts, promote the sustainable use of wetland resources and the long-term stability of the ecological environment. It is particularly suitable for wetland ecological environments with frequent climate change and complex and changeable ecology, and can improve the level of intelligent wetland management.

[0051] The embodiments of the present invention will now be described in detail with reference to the specific implementation steps.

[0052] Figure 1 This is a flowchart of an optional wetland pest and disease monitoring method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0053] Step S101: Collect environmental sample data of the target monitoring area within the first time period. The environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases.

[0054] Specifically, the target monitoring area refers to a specific wetland ecosystem segment requiring pest and disease monitoring. This area is pre-designated by the intelligent wetland pest and disease monitoring and early warning system based on the wetland's ecological importance, the historical frequency of pest and disease occurrences, the received monitoring and early warning needs, and the degree of impact on the area's biodiversity and ecological environment. The target monitoring area can include different wetland habitat types (e.g., marshes, lake edges, river deltas, etc.) and pest and disease hotspots (e.g., areas rich in food resources and with few natural predators).

[0055] The first time period refers to the time cycle for collecting environmental sample data determined based on the received monitoring and early warning needs. It can cover a time span of several days, weeks, or months, depending on the life cycle of pests and diseases and the changing patterns of the wetland environment. For example, during the peak breeding season of pests and diseases, the monitoring interval can be shortened to increase the data collection frequency, which can more accurately track the dynamics of pest and disease populations.

[0056] Environmental sample data is a collection of information describing key survival indicators of the target monitoring area as a habitat for pests and diseases. These mainly include, but are not limited to, temperature, humidity, rainfall, food cover, and the number of natural enemies. Temperature primarily affects the physiological activity rate and reproductive cycle of pests and diseases; humidity primarily determines the water balance of pests and diseases and the likelihood of disease transmission; rainfall primarily affects the hydrological conditions of the wetland ecosystem, indirectly determining the food sources and habitat quality of pests and diseases; food cover mainly describes the abundance of food sources for pests and diseases, directly affecting their reproductive capacity and spread rate; and the number of natural enemies refers to the number of predators or parasites naturally present in the wetland environment, playing a crucial role in regulating pest and disease population dynamics.

[0057] The collection of the aforementioned environmental sample data can be achieved through various methods and technologies, including but not limited to: environmental sensor networks, remote sensing satellite image analysis, UAV flight data collection, and real-time data transmission from ground ecological monitoring stations. This can obtain the most authentic, comprehensive, and high-resolution environmental information. After the data collection is completed, it can be used to train and validate the Naive Bayes model to identify the reproduction information of pests and diseases.

[0058] Through the above series of steps, the present invention can provide early warning signals before the outbreak of pests and diseases, thereby enabling more proactive and effective management measures to protect the health and biodiversity of wetland ecosystems.

[0059] To address the issues of incomplete data collection and rudimentary processing in traditional pest and disease monitoring, and to further achieve a deeper understanding and intelligent prediction of pest and disease activities in wetland ecosystems, the following steps can be optionally included in the collection of environmental sample data of the target monitoring area within a first time period: receiving and responding to data acquisition signals through a pre-deployed sensor array, wherein the sensor array includes at least: a clock sensor, a temperature sensor, a humidity sensor, a rainfall sensor, a vegetation monitoring sensor group, a biological monitoring sensor group, and an insect monitoring lamp; simultaneously controlling the clock sensor, temperature sensor, humidity sensor, rainfall sensor, vegetation monitoring sensor group, biological monitoring sensor group, and insect monitoring lamp to record the collection time point, temperature, humidity, rainfall, vegetation coverage, specified biological coverage, and actual pest and disease density according to the first time period and collection frequency indicated by the data acquisition signal, thereby obtaining a raw dataset; and preprocessing the raw data in the raw dataset to obtain environmental sample data, wherein the preprocessing includes the following operations: data association, data cleaning, data standardization, and data fusion.

[0060] Specifically, clock sensors provide timestamps to ensure all collected data is recorded correctly in chronological order, facilitating time-series analysis; temperature sensors monitor wetland temperature changes; humidity sensors record air humidity and assess wetland moisture conditions; rainfall sensors measure rainfall, and understanding the hydrological cycle can indirectly reflect potential food sources and habitat quality for pests and diseases; vegetation monitoring sensors collect information on the growth status and species of wetland vegetation, and monitoring food cover can indirectly reflect food resources for pests and diseases; biological monitoring sensors monitor the species and quantity of organisms in wetlands, allowing for the designation of monitoring natural enemies and host plants for pests and diseases, and assessing the health status of the biological community; insect-attracting lamps can record the actual density distribution of pests and diseases by attracting and capturing insects, used for model validation and real-time monitoring.

[0061] In one alternative embodiment, after receiving the data acquisition signal, the sensor array begins to automatically record monitoring data according to a specified first time period (which may be a fixed time every day, every week, etc.) and acquisition frequency (such as once per hour, three times a day, etc.). This automated data acquisition process can reduce human error, ensure the consistency and real-time nature of the data, and provide high-quality raw data for model training and prediction.

[0062] In the optional embodiments, data association refers to integrating raw data from different sensors according to time coordinates to ensure that various environmental data at the same point in time can be correlated to form a complete description of the environmental state. Data cleaning refers to removing outliers and irrelevant data to ensure data accuracy, such as occasional erroneous readings from temperature sensors or missing records due to severe weather. Data standardization refers to converting various data into a unified scale to eliminate the influence of dimensions and make the data more suitable for model training, such as converting temperature and humidity into standard scores. Data fusion refers to comprehensively analyzing data from multiple sensors to generate more comprehensive indicators of the pest and disease survival environment. The preprocessing steps include noise removal, missing value filling, and data format standardization to ultimately form environmental sample data that can be used for model training.

[0063] Step S102: Input environmental sample data into the first model and output pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database. The pest and disease database is used to record the mapping relationship between the reproduction rate of pests and diseases and their living environment.

[0064] Specifically, the first model uses a Naive Bayes classifier and is trained using a machine learning algorithm. It should be noted that the Naive Bayes model is a classic classification method based on probability theory in the field of statistical learning. It assumes that each feature (such as temperature, humidity, rainfall, etc.) independently affects the reproduction status of pests and diseases.

[0065] In this embodiment of the invention, the first model is trained based on a pest and disease database, which stores rich historical data, including information on the reproduction rate of pests and diseases under different environmental conditions and specific index values ​​of the above-mentioned different environmental conditions (e.g., temperature range, humidity percentage, rainfall, etc.). After the above data is preprocessed, feature extracted and model trained, it forms the internal knowledge system of the first model, enabling the first model to predict the reproduction information of pests and diseases based on the input environmental sample data.

[0066] In this embodiment of the invention, pest and disease reproduction information is used to reflect the reproductive potential and activity status of pests and diseases under specific environmental conditions, including the following aspects: reproduction rate, population growth trend, and activity status. Specifically, the reproduction rate represents the speed and efficiency of pest and disease reproduction under current environmental conditions; the population growth trend is the trend of pest and disease numbers changing over time, predicted based on the reproduction rate; and the activity status describes whether pests and diseases are in an active reproductive period and the degree of their activity.

[0067] For example, if the model outputs "the pests and diseases are in an active breeding period and the population is expected to increase at a high growth rate in the future," similar information helps wetland management systems understand the actual threat level of pests and diseases and their possible development trends, thereby enabling them to plan prevention and control measures in advance and avoid large-scale outbreaks of pests and diseases that could cause ecological damage.

[0068] This invention, through the input of environmental sample data into a first model, outputs pest and disease reproduction information, providing crucial intermediate variables for subsequent pest and disease density prediction. This information, along with the environmental sample data, is then fed into a second model (a time-series prediction model) for further analysis to predict future pest and disease densities. This phased prediction method fully utilizes the correlation between environmental factors and pest and disease activity, enhancing the accuracy and reliability of predictions, thus facilitating early warning and ecological management of wetland pests and diseases.

[0069] Figure 2 This is an optional flowchart for identifying the degree of pest and disease reproduction according to an embodiment of the present invention, such as... Figure 2 As shown, using temperature, humidity, rainfall, food cover, and number of natural enemies as input data (X), and reproductive level (Y) as label data, a Naive Bayes classification model is trained to obtain a Naive Bayes pest classifier. Subsequently, inputting local data such as temperature, humidity, rainfall, food cover, and number of natural enemies into this classifier can map and output the local pest reproductive level, including: unable to reproduce, difficult to reproduce, difficult to reproduce, limited reproduction, appropriate reproduction, and easy reproduction.

[0070] To achieve intelligent identification of pest and disease reproduction information, pest and disease reproduction patterns can be extracted from historical regional observation data. This can be combined with iterative training and parameter optimization of a Naive Bayes model to obtain a first model. Optionally, the first model is trained through the following steps: acquiring historical regional observation data and establishing a pest and disease database based on this data. The historical regional observation data includes pest and disease survival environment data, pest and disease reproduction rates, and predicted pest and disease density trends over historical time periods. A training set is extracted from the pest and disease database, and a Naive Bayes model is obtained as the first initial model. The pest and disease survival environment data in the training set is used as input data to iteratively train the first initial model. During iterative training, the pest and disease reproduction rate in the training set is used as the fitting target to adjust the parameters of the first initial model. Once the expected fitting result is obtained, iterative training is stopped, and the corresponding first initial model is determined as the first model.

[0071] In one specific implementation scenario, historical regional observation data is collected from long-term observation records of the target wetland or similar ecosystems, covering multiple key indicators within the historical time period, including but not limited to environmental factors that directly affect the survival and reproduction of pests and diseases such as temperature, humidity, rainfall, food coverage, and the number of natural enemies. Then, a pest and disease database is constructed, and a training set is extracted from the pest and disease database, including pest and disease survival environment data, pest and disease reproduction rate, and predicted density trend, for model learning and optimization.

[0072] In the model learning and optimization process, a Naive Bayes model was first selected as the initial model. Then, iterative training was conducted, using environmental data from the training set as input to the model. The model's goal was to fit the pest and disease reproduction rate as accurately as possible. Through repeated training, the model parameters were gradually adjusted to minimize the difference between the predicted reproduction rate and the actual observed value, thereby improving the model's predictive ability.

[0073] In each iteration of training, the reproduction rate of pests and diseases is used as the fitting target. The internal parameters of the model are fine-tuned according to the changes in the reproduction rate until the model output results highly match the reproduction rate labels in the training set. The above parameter adjustment strategy based on target feedback enables the model to continuously optimize itself and better understand the survival and reproduction patterns of pests and diseases under different environmental conditions.

[0074] Finally, when the model achieves the expected fit—that is, exhibits satisfactory prediction accuracy on the training set—the iteration process stops. The first initial model obtained at this point is confirmed as the first model and used for the identification of pest and disease reproduction information.

[0075] To generate structured pest and disease reproduction information, optionally, the step of inputting environmental sample data into a first model and outputting pest and disease reproduction information includes: inputting environmental sample data into the input layer of the first model; calculating conditional probabilities based on the environmental sample data and Bayes' theorem through the computation layer of the first model to obtain the category probabilities corresponding to N pest and disease categories, where N is a positive integer; for each pest and disease category, analyzing the pest and disease reproduction rate under the condition that the target monitoring area is the pest and disease survival environment based on the environmental sample data; generating pest and disease reproduction information containing pest and disease category, category probability, and pest and disease reproduction rate, and outputting the pest and disease reproduction information through the output layer of the first model.

[0076] In this embodiment of the invention, the input layer of the first model is responsible for receiving external data, converting it into a format that the model can process internally, and serving as the starting point of the model processing flow.

[0077] Next, conditional probability calculations are performed using Bayes' theorem through the computational layer of the first model. Bayes' theorem is used in statistics to update the probability estimate of an event given certain evidence. In this invention, the computational layer analyzes environmental sample data using historical data from a pest and disease database as evidence, estimating the probability of N types of pests and diseases occurring in the target monitoring area. This probability corresponds to the category probability of that pest or disease type, where N represents the number of pest and disease types that the first model can identify. Each pest or disease corresponds to different survival and reproduction conditions, resulting in varying probabilities of occurrence in the target monitoring area. The category probability corresponding to each pest or disease type is used as the reproduction rate of the corresponding pest or disease in the target monitoring area, reflecting the growth trend of the pest or disease population, and thus used to predict the likelihood of a pest or disease outbreak in the short term.

[0078] Finally, the three key pieces of information—pest / disease category, category probability, and pest / disease reproduction rate—are integrated to generate pest / disease reproduction information. This highly structured pest / disease reproduction information not only includes pest / disease classification information but also provides an overview of the occurrence and reproduction of various pests / diseases under current environmental conditions, which is clearly presented through the output layer of the first model.

[0079] Step S103: Input pest and disease reproduction information and environmental sample data into the second model, and output the predicted density of pests and diseases in the second time period of the target monitoring area. The second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is the future time period of the first time period.

[0080] Specifically, unlike static pest and disease reproduction information identification, the second model focuses on the analysis and prediction of time-series data, especially handling dynamic ecological data that changes over time. The second model is trained based on an improved activation function and, combined with a time-series prediction model architecture, can deeply learn and grasp the complex relationship between pest and disease density and environmental factors.

[0081] Time-series forecasting models can not only handle linear changes in time-series data but also capture nonlinear dynamic characteristics, which is particularly crucial for predicting pest and disease density. Pest and disease density is typically influenced by the interaction of multiple environmental factors and changes over time, leading to fluctuations in pest and disease populations. The second model, through a self-attention mechanism and an optimized residual network, can identify and utilize these changes to predict future trends in pest and disease density.

[0082] The predicted density of pests and diseases refers to the quantity level of pests and diseases in the target monitoring area predicted by the second model within a second time period (i.e., a future period) based on current environmental sample data and pest and disease reproduction information. It is presented as a quantitative value, reflecting the potential population density of pests and diseases in the future. By predicting the density of pests and diseases in the future, wetland management systems can anticipate potential risk levels and take appropriate preventative measures to prevent outbreaks and protect the health and stability of the wetland ecosystem.

[0083] The selection of the second time period is usually based on the life cycle of pests and diseases and the ecological response time, generally covering a time range of several weeks to several months. The second model predicts the future change trajectory of pest and disease density in the target area by analyzing historical data and the real-time state of the pest and disease's living environment, including the trend of density rise and fall, the time point of peak occurrence, and the possible range of density fluctuations.

[0084] The predicted density of pests and diseases directly determines subsequent control decisions. When the predicted density is low, only routine monitoring is needed; when the predicted density is close to or higher than the warning value, biological control, chemical control, or even emergency joint prevention and control measures are required to reduce the risk of pests and diseases. The higher the prediction accuracy, the earlier the wetland ecosystem management system can react, keeping pests and diseases within the minimum range of damage.

[0085] Figure 3 This is an optional flowchart for identifying pest and disease density according to an embodiment of the present invention, such as... Figure 3 As shown, using date, temperature, humidity, and reproduction level as input data (X), and using historical pest density (Y) of insect infestation lamps as label data, a time-series prediction model is trained, and then the actual pest density in the local area is predicted based on the local date, temperature, humidity, and reproduction level.

[0086] To improve the accuracy of pest and disease density prediction, a second model can be obtained by extracting a training set from a pest and disease database and iteratively training it using a time-series prediction model architecture and an improved activation function. Optionally, the second model is obtained through the following steps: extracting a training set from a pest and disease database and establishing a second initial model based on a time-series prediction model and an improved activation function, where the improved activation function is obtained by introducing a positive slope into a specified negative interval of the ReLU activation function; using pest and disease survival environment data and pest and disease reproduction rate from the training set as input data, iteratively training the second initial model; during the iterative training process, using the development trend of predicted pest and disease density in the training set as the fitting target, adjusting the parameters of the second initial model; and stopping the iterative training when the expected fitting result is obtained, thus determining the corresponding second initial model as the second model.

[0087] In one specific implementation scenario, a training set selected from a pest and disease database for training a time-series prediction model includes historical data on the survival environment, reproduction rate, and predicted density trends of pests and diseases. Subsequently, a second initial model is built based on the time-series prediction model architecture and combined with an improved activation function (an optimized ReLU, a variant of Leaky ReLU (LeakyRectified Linear Unit)). It should be noted that the improved activation function introduces a positive slope in the specified negative interval of ReLU, solving the neuron "death" problem caused by the original ReLU function having a zero gradient in the negative interval, thus enhancing the model's stability and predictive ability.

[0088] After determining the second initial model, data on the pest and disease survival environment and their reproduction rate were used as input data to iteratively train the model. By continuously optimizing the model parameters, the model was able to more accurately learn the patterns of pest and disease density changes over time. In each iteration, the model attempted to fit the predicted trend of pest and disease density as its objective; that is, the model output should be as close as possible to the trend of pest and disease density changes recorded in the training set. Through repeated training, the model parameters were gradually adjusted to minimize prediction error and improve the model's prediction accuracy.

[0089] During the training process, once the model achieves the expected fitting result—that is, exhibits a satisfactory prediction accuracy on the training set, making the predicted trend of pest and disease density highly consistent with the actual observed values—the iterative training is stopped, and the second initial model is formally confirmed as the second model for future pest and disease density prediction.

[0090] To accurately predict the development trend of pest and disease density, optionally, the step of inputting pest and disease reproduction information and environmental sample data into a second model and outputting the predicted pest and disease density in the target monitoring area within a second time period includes: inputting pest and disease reproduction information obtained from a first model and collected environmental sample data into the input layer of the second model; mapping and encoding the pest and disease reproduction information and environmental sample data through the encoding layer of the second model to obtain a set of encoded information, wherein the encoded information includes at least the encoded vector corresponding to the pest and disease reproduction information and M time-series encoded segments corresponding to the environmental sample data, where M is a positive integer; performing a pest and disease predicted density development analysis on an extended time scale based on a self-attention mechanism and a residual network through the fitting layer of the second model to obtain a density-time fitting curve, wherein the extended time scale includes at least the union of the first time period and the second time period; extracting the predicted pest and disease density within the second time period based on the density-time fitting curve, and outputting the predicted pest and disease density through the output layer of the second model.

[0091] In this embodiment of the invention, the input layer of the second model is the data entry point for receiving external information. The pest and disease reproduction information includes pest and disease categories, category probabilities, and reproduction rates, which are the results of the first model's analysis based on historical data and current environmental conditions. The environmental sample data includes a series of key environmental indicators that are monitored in real time and affect the survival and spread of pests and diseases, such as temperature, humidity, and rainfall.

[0092] Furthermore, the task of the coding layer is to convert the raw pest and disease reproduction information and environmental sample data into a format that the model can understand and process, typically in the form of coding vectors or temporal coding segments. The coding layer maps pest and disease reproduction information into a set of coding vectors, carrying comprehensive information about pest and disease categories and reproduction potential, while decomposing the environmental sample data into M temporal coding segments, where M represents the time step or window size processed by the model. Each temporal coding segment contains environmental data at a specific time point, facilitating analysis by the second model on the time axis.

[0093] Furthermore, the second model performs in-depth analysis of the encoded information in the fitting layer based on a self-attention mechanism and a residual network. The self-attention mechanism focuses on key information points in the sequence, while the residual network maintains a stable gradient flow during deep learning through a series of residual blocks, avoiding potential information attenuation or amplification issues in deep networks. The core task of the fitting layer is to perform evolutionary analysis of pest and disease prediction density on an extended time scale, i.e., analyzing how pest and disease density evolves over time. Here, the extended time scale covers the union of the first time period (the current and past monitoring periods) and the second time period (the future prediction period), ensuring the continuity and completeness of the model's predictions.

[0094] The fitting layer produces one or a set of density-time fitting curves to depict the trend of predicted pest and disease density changes over time. The model can extract the predicted pest and disease density for a second time period based on these curves, serving as an estimate of future pest and disease density based on historical data, current environmental conditions, and pest and disease reproduction patterns.

[0095] Finally, the output layer of the second model converts the results of the deep learning process into a readable and intuitive form, outputting the predicted density of pests and diseases in the target monitoring area during the second time period.

[0096] Step S104: Determine the pest and disease treatment plan based on the predicted pest and disease density.

[0097] Specifically, the pest and disease management plan follows the principles of risk level classification and targeted intervention, aiming to achieve effective pest and disease management and the protection of wetland ecosystems. The core of the plan lies in translating predicted pest and disease densities into actionable management measures, ranging from mild monitoring to strict chemical control based on different levels of pest and disease density, forming a comprehensive response system from early warning to emergency response.

[0098] Specifically, based on the predicted pest and disease density, the risk level is divided into four levels: normal value, warning value, alert value, and risk value. Each level corresponds to different treatment plans and intervention intensities.

[0099] Optionally, the step of determining a pest and disease treatment plan based on the predicted pest and disease density includes: obtaining a pre-set level threshold, wherein the level thresholds are arranged in descending order as: normal value > warning value > alert value > risk value; if the predicted pest and disease density matches the normal value, determining the pest and disease treatment plan as maintaining regular frequency monitoring; if the predicted pest and disease density matches the warning value, determining the pest and disease treatment plan as increasing the monitoring frequency according to a first preset granularity and taking preventive biological control measures; if the predicted pest and disease density matches the alert value, determining the pest and disease treatment plan as... The second preset particle size increases the monitoring frequency, and environmentally friendly chemical control agents are used for local control. The second preset particle size is larger than the first preset particle size. Environmentally friendly chemical control agents are chemical substances that meet the following conditions: their toxicity to non-pests and pests is lower than the preset toxicity threshold at the specified dosage, and their degradation cycle is shorter than the preset time threshold. When the predicted density of pests and diseases matches the risk value, the pest and disease treatment plan is determined to be to increase the monitoring frequency according to the third preset particle size and use environmentally friendly chemical control agents for global control. The third preset particle size is larger than the second preset particle size.

[0100] Specifically, when the predicted density is below the normal value, it indicates that the threat of pests and diseases to the wetland ecosystem is relatively small. At this time, only routine ecological monitoring needs to be maintained, environmental data needs to be collected regularly, and the dynamics of pests and diseases need to be tracked. When the predicted density exceeds the normal value but is still below the warning value, it means that it is at a controllable level. The monitoring frequency should be increased immediately, and monitoring of high-incidence areas of pests and diseases should be strengthened. At the same time, eco-friendly control measures can be considered, such as increasing the number of natural enemies and using biological pesticides, to prevent the further spread of pests and diseases. If the predicted density reaches the warning value, it means that pests and diseases may have a significant impact on local biodiversity. More proactive intervention measures need to be taken, including implementing precise chemical control in specific areas, selecting low-toxicity and high-efficiency chemical agents for spraying, and controlling the number of pests and diseases. When the predicted density is higher than the risk value, it reaches the highest risk level, which means that pests and diseases are about to or have already posed a major threat to the wetland ecosystem. At this time, the emergency response mechanism should be activated, a comprehensive pesticide spraying campaign should be organized, and agricultural aviation technology and drones should be used to carry out full-area pesticide delivery to quickly suppress the spread of pests and diseases and prevent ecological disasters.

[0101] Figure 4 This is an optional structural diagram of a pest and disease risk level classification and treatment plan according to an embodiment of the present invention, such as... Figure 4 As shown, the threshold judgment is performed based on the density of pests and diseases, including normal value, warning value, alert value and risk value. When it matches the normal value, routine monitoring can be maintained. When it matches the warning value, preventive chemical control is required. When it matches the alert value, local chemical control is implemented. Finally, when it matches the risk value, regional joint prevention and control is required.

[0102] For example, if the pest is specifically the East Asian migratory locust, then the normal value is no more than 1 locust per square meter, the warning value is more than 1 but no more than 5 locusts per square meter, the alert value is more than 5 but no more than 10 locusts per square meter, and the risk value is more than 10 locusts per square meter. Therefore, if... Figure 5 As shown, Figure 5 This is an example diagram of an optional treatment method for the Oriental migratory locust according to an embodiment of the present invention. The risk level is determined by the density of the Oriental migratory locust: normal value corresponds to routine monitoring; warning value corresponds to enhanced monitoring and biological control, with Bt Bacillus thuringiensis and abamectin recommended as pesticides; alert value corresponds to emergency monitoring and assessment and localized chemical control, with fipronil and imidacloprid recommended as pesticides; risk value corresponds to comprehensive monitoring and assessment and large-scale chemical control, with acetamiprid and thiamethoxam recommended as pesticides.

[0103] Another example: if the pest is specifically the reed shoot moth, then the normal value is no more than 2 moths per square meter, the warning value is more than 2 but no more than 6 moths per square meter, the alert value is more than 6 but no more than 13 moths per square meter, and the risk value is ≥13 moths per square meter. Therefore, if... Figure 6 As shown, Figure 6 This is an example diagram of an optional reed tip moth pest control method according to an embodiment of the present invention. The risk level is determined by the density of reed tip moths: normal value corresponds to routine monitoring; warning value corresponds to enhanced monitoring and biological control, with chrysanthemum larvae and Trichogramma wasps recommended; alert value corresponds to emergency monitoring and assessment and localized chemical control, with permethrin and abamectin recommended; risk value corresponds to comprehensive monitoring and assessment and large-scale chemical control, with cypermethrin and fipronil recommended.

[0104] Through steps S101 to S104, environmental sample data of the target monitoring area within the first time period can be collected first. The environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases. Then, the environmental sample data is input into the first model, which outputs pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database. The pest and disease database is used to record the mapping relationship between the reproduction rate of pests and diseases and their living environment. Then, the pest and disease reproduction information and the environmental sample data are input into the second model, which outputs the predicted density of pests and diseases in the target monitoring area within the second time period. The second model is a time-series prediction model pre-trained based on an improved activation function. The second time period is the future time period of the first time period. Finally, the pest and disease treatment plan is determined based on the predicted density of pests and diseases.

[0105] In this embodiment of the invention, a combination of comprehensive environmental analysis and intelligent prediction is adopted. By constructing a database of ecological characteristics of pests and diseases and integrating machine learning algorithms, the goal of accurately predicting the development trend of wetland pests and diseases is achieved, thereby realizing the technical effects of early warning of pests and diseases and eco-friendly prevention and control. Specifically, this invention collects environmental sample data covering key ecological elements affecting pest and disease reproduction in the target monitoring area within a first time period. Then, it identifies pest and disease reproduction information through a first model—a Naive Bayes model trained on a large amount of historical data. Deep analysis of the environmental sample data accurately outputs pest and disease reproduction information, revealing potential risks to the pest and disease's living environment. Next, the reproduction information is combined with real-time environmental data and input into a second model—a time-series prediction model with an improved activation function—to accurately predict pest and disease density in the second time period, effectively capturing the dynamic trend of pest and disease spread. Furthermore, this invention proposes a tiered response mechanism based on the predicted pest and disease density information. It automatically generates scientifically sound pest and disease treatment plans according to different risk levels, ensuring rapid response and effective control of wetland ecosystems in the face of pest and disease invasions. This greatly enriches wetland protection measures, compensates for the shortcomings of traditional monitoring methods, and solves the technical problem in related technologies where single-model identification cannot effectively monitor wetland pests and diseases, thus failing to meet the needs for early warning and precise prevention and control of pests and diseases.

[0106] The invention will now be described in conjunction with another alternative embodiment.

[0107] Example 2

[0108] The wetland pest and disease monitoring device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in Embodiment 1 above.

[0109] Figure 7 This is a schematic diagram of an optional wetland pest and disease monitoring device according to an embodiment of the present invention, such as... Figure 7 As shown, the device may include: a data acquisition unit 701, a first input unit 702, a second input unit 703, and a determination unit 704.

[0110] The acquisition unit 701 is used to acquire environmental sample data of the target monitoring area within the first time period. The environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases.

[0111] The first input unit 702 is used to input environmental sample data into the first model and output pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database, which is used to record the mapping relationship between the reproduction rate of pests and diseases and their living environment.

[0112] The second input unit 703 is used to input pest and disease reproduction information and environmental sample data into the second model and output the predicted density of pests and diseases in the target monitoring area during the second time period. The second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is the future time period of the first time period.

[0113] The determination unit 704 is used to determine the pest and disease treatment plan based on the predicted density of pests and diseases.

[0114] The aforementioned wetland pest and disease monitoring device can first collect environmental sample data of the target monitoring area within a first time period through the acquisition unit 701. The environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases. Then, the environmental sample data is input into a first model through the first input unit 702, which outputs pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database, which records the mapping relationship between the reproduction rate of pests and diseases and their living environment. Then, the pest and disease reproduction information and environmental sample data are input into a second model through the second input unit 703, which outputs the predicted density of pests and diseases in the target monitoring area within a second time period. The second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period. Finally, the pest and disease treatment plan is determined by the determination unit 704 based on the predicted pest and disease density.

[0115] In this embodiment of the invention, a combination of comprehensive environmental analysis and intelligent prediction is adopted. By constructing a database of ecological characteristics of pests and diseases and integrating machine learning algorithms, the goal of accurately predicting the development trend of wetland pests and diseases is achieved, thereby realizing the technical effects of early warning of pests and diseases and eco-friendly prevention and control. Specifically, this invention collects environmental sample data covering key ecological elements affecting pest and disease reproduction in the target monitoring area within a first time period. Then, it identifies pest and disease reproduction information through a first model—a Naive Bayes model trained on a large amount of historical data. Deep analysis of the environmental sample data accurately outputs pest and disease reproduction information, revealing potential risks to the pest and disease's living environment. Next, the reproduction information is combined with real-time environmental data and input into a second model—a time-series prediction model with an improved activation function—to accurately predict pest and disease density in the second time period, effectively capturing the dynamic trend of pest and disease spread. Furthermore, this invention proposes a tiered response mechanism based on the predicted pest and disease density information. It automatically generates scientifically sound pest and disease treatment plans according to different risk levels, ensuring rapid response and effective control of wetland ecosystems in the face of pest and disease invasions. This greatly enriches wetland protection measures, compensates for the shortcomings of traditional monitoring methods, and solves the technical problem in related technologies where single-model identification cannot effectively monitor wetland pests and diseases, thus failing to meet the needs for early warning and precise prevention and control of pests and diseases.

[0116] Optionally, the acquisition unit includes: a receiving module for receiving and responding to data acquisition signals through a pre-arranged sensor array, wherein the sensor array includes at least: a clock sensor, a temperature sensor, a humidity sensor, a rainfall sensor, a vegetation monitoring sensor group, a biological monitoring sensor group, and an insect lamp; a control module for simultaneously controlling the clock sensor, temperature sensor, humidity sensor, rainfall sensor, vegetation monitoring sensor group, biological monitoring sensor group, and insect lamp to record the acquisition time point, temperature, humidity, rainfall, vegetation coverage, specified biological coverage, and actual density of pests and diseases according to the first time period and acquisition frequency indicated by the data acquisition signal, thereby obtaining a raw dataset; and a preprocessing module for preprocessing the raw data in the raw dataset to obtain environmental sample data, wherein the preprocessing includes the following operations: data association, data cleaning, data standardization, and data fusion.

[0117] Optionally, the wetland pest and disease monitoring device further includes: a first training unit for training a first model, wherein the first training unit includes: a first establishment module for acquiring historical regional observation data and establishing a pest and disease database based on the historical regional observation data, wherein the historical regional observation data includes pest and disease survival environment data, pest and disease reproduction rate, and pest and disease predicted density development trend within a historical time period; a first iterative training module for extracting a training set from the pest and disease database, obtaining a Naive Bayes model as the first initial model, and iteratively training the first initial model using pest and disease survival environment data in the training set as input data; a first parameter adjustment module for adjusting the parameters of the first initial model during iterative training, using the pest and disease reproduction rate in the training set as the fitting target; and a first determination module for stopping iterative training and determining the corresponding first initial model as the first model when the expected fitting result is obtained.

[0118] Optionally, the first input unit includes: a first input module for inputting environmental sample data into the input layer of the first model; a calculation module for performing conditional probability calculations based on the environmental sample data and Bayes' theorem through the calculation layer of the first model to obtain the category probabilities corresponding to N types of pests and diseases, where N is a positive integer; a first analysis module for analyzing the pest and disease reproduction rate for each type of pest and disease based on the environmental sample data, using the target monitoring area as the pest and disease survival environment; and a first output module for generating pest and disease reproduction information containing pest and disease categories, category probabilities, and pest and disease reproduction rates, and outputting the pest and disease reproduction information through the output layer of the first model.

[0119] Optionally, the wetland pest and disease monitoring device further includes: a second training unit for training a second model, wherein the second training unit includes: a second establishment module for extracting a training set from a pest and disease database and establishing a second initial model based on a time-series prediction model and an improved activation function, wherein the improved activation function is obtained by introducing a positive slope into a specified negative interval of the ReLU activation function; a second iterative training module for iteratively training the second initial model using pest and disease survival environment data and pest and disease reproduction rate in the training set as input data; a second parameter adjustment module for adjusting the parameters of the second initial model during iterative training, using the development trend of pest and disease prediction density in the training set as the fitting target; and a second determination module for stopping iterative training and determining the corresponding second initial model as the second model when the expected fitting result is obtained.

[0120] Optionally, the second input unit includes: a second input module, used to input the pest and disease reproduction information obtained from the first model and the collected environmental sample data into the input layer of the second model; a mapping and encoding module, used to map and encode the pest and disease reproduction information and the environmental sample data through the encoding layer of the second model to obtain a set of encoded information, wherein the encoded information includes at least the encoding vector corresponding to the pest and disease reproduction information and M time-series encoded segments corresponding to the environmental sample data, where M is a positive integer; a second analysis module, used to perform pest and disease prediction density development analysis on the set of encoded information on an extended time scale through the fitting layer of the second model based on a self-attention mechanism and a residual network to obtain a density-time fitting curve, wherein the extended time scale includes at least the union of the first time period and the second time period; and a second output module, used to extract the pest and disease prediction density in the second time period based on the density-time fitting curve and output the pest and disease prediction density through the output layer of the second model.

[0121] Optionally, the determining unit includes: an acquisition module for acquiring pre-set level thresholds, wherein the level thresholds, from largest to smallest, are: normal value > warning value > alert value > risk value; a third determining module for determining the pest and disease treatment plan as maintaining regular frequency monitoring when the predicted pest and disease density matches the normal value; a fourth determining module for determining the pest and disease treatment plan as increasing the monitoring frequency according to a first preset granularity and taking preventive biological control measures when the predicted pest and disease density matches the warning value; and a fifth determining module for determining the pest and disease treatment plan as... The monitoring frequency is increased according to the second preset particle size, and environmentally friendly chemical control agents are used for local control. The second preset particle size is larger than the first preset particle size. Environmentally friendly chemical control agents refer to chemical substances that meet the following conditions: their toxicity to non-pests and pests is lower than a preset toxicity threshold at a specified dosage, and their degradation cycle is shorter than a preset time threshold. The sixth determination module is used to determine the pest and disease treatment plan as follows when the predicted pest and disease density matches the risk value: the monitoring frequency is increased according to the third preset particle size, and environmentally friendly chemical control agents are used for global control. The third preset particle size is larger than the second preset particle size.

[0122] The aforementioned wetland pest and disease monitoring device may also include a processor and a memory. The aforementioned acquisition unit 701, first input unit 702, second input unit 703, determination unit 704, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0123] The aforementioned processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, environmental sample data of the target monitoring area within a first time period is collected. This environmental sample data is input into a first model, which outputs pest and disease reproduction information. The pest and disease reproduction information and environmental sample data are then input into a second model, which outputs the predicted pest and disease density of the target monitoring area within a second time period. Based on this predicted density, a pest and disease treatment plan is determined.

[0124] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0125] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program with the following method steps: collecting environmental sample data of a target monitoring area within a first time period, wherein the environmental sample data is used to describe various survival indicators of the target monitoring area as a habitat for pests and diseases; inputting the environmental sample data into a first model and outputting pest and disease reproduction information, wherein the first model is a Naive Bayes model pre-trained based on a pest and disease database, and the pest and disease database is used to record the mapping relationship between the reproduction rate of pests and diseases and their habitat; inputting the pest and disease reproduction information and the environmental sample data into a second model and outputting a predicted pest and disease density of the target monitoring area within a second time period, wherein the second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period; and determining a pest and disease treatment plan based on the predicted pest and disease density.

[0126] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the wetland pest and disease monitoring method of any one of the above embodiments.

[0127] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the wetland pest and disease monitoring method of any one of the above embodiments.

[0128] Figure 8 This is a structural block diagram of an electronic device for performing a method for monitoring wetland pests and diseases according to an embodiment of the present invention, such as... Figure 8 As shown, the electronic device may include: one or more ( Figure 8 (Only one is shown) processor 802, memory 804, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0129] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the wetland pest and disease monitoring method and device in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned wetland pest and disease monitoring method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0130] Those skilled in the art will understand that Figure 8 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 8 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 8 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 8 The different configurations shown.

[0131] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0132] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0133] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0135] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0136] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0137] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0138] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method of monitoring for pests in a wetland, the method comprising: include: Environmental sample data of the target monitoring area are collected within a first time period, wherein the environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases. The environmental sample data is input into the first model, and the pest and disease reproduction information is output. The first model is a Naive Bayes model pre-trained based on a pest and disease database, which is used to record the mapping relationship between the reproduction rate of pests and diseases and their living environment. The pest and disease reproduction information and the environmental sample data are input into the second model, and the predicted density of pests and diseases in the target monitoring area during the second time period is output. The second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period. The pest and disease treatment plan is determined based on the predicted pest and disease density.

2. The method of claim 1, wherein, The steps for collecting environmental sample data of the target monitoring area within the first time period include: The system receives and responds to data acquisition signals through a pre-arranged sensor array, wherein the sensor array includes at least: a clock sensor, a temperature sensor, a humidity sensor, a rainfall sensor, a vegetation monitoring sensor group, a biological monitoring sensor group, and an insect monitoring lamp. According to the first time period and acquisition frequency indicated by the data acquisition signal, the clock sensor, temperature sensor, humidity sensor, rainfall sensor, vegetation monitoring sensor group, biological monitoring sensor group and insect lamp are controlled to record the acquisition time point, temperature, humidity, rainfall, vegetation coverage, specified biological coverage and actual density of pests and diseases, respectively, to obtain the original dataset. The raw data in the original dataset is preprocessed to obtain the environmental sample data. The preprocessing includes the following operations: data association, data cleaning, data standardization, and data fusion.

3. The method of claim 1, wherein the method is performed in a wetland. The first model was trained through the following steps: Historical regional observation data is acquired, and the pest and disease database is established based on the historical regional observation data. The historical regional observation data includes pest and disease survival environment data, pest and disease reproduction rate, and pest and disease predicted density development trend within a historical time period. A training set is extracted from the pest and disease database, the Naive Bayes model is obtained as the first initial model, and the pest and disease survival environment data in the training set is used as input data to iteratively train the first initial model. During the iterative training process, the pest and disease reproduction rate in the training set is used as the fitting target to adjust the parameters of the first initial model. If the expected fitting result is obtained, stop iterative training and determine the corresponding first initial model as the first model.

4. The method of claim 3, wherein the method further comprises, The steps of inputting the environmental sample data into the first model and outputting pest and disease reproduction information include: The environmental sample data is input into the input layer of the first model; Through the computational layer of the first model, conditional probability calculations are performed based on the environmental sample data and Bayes' theorem to obtain the category probabilities corresponding to N types of pests and diseases, where N is a positive integer; For each of the aforementioned pest and disease categories, the pest and disease reproduction rate is analyzed based on the environmental sample data, taking the target monitoring area as the pest and disease survival environment. Generate pest and disease reproduction information that includes the pest and disease category, the category probability, and the pest and disease reproduction rate, and output the pest and disease reproduction information through the output layer of the first model.

5. The method of claim 1, wherein the method is used for monitoring of wetland pests. The second model was obtained through the following steps: A training set is extracted from a pest and disease database, and a second initial model is established based on a time-series prediction model and an improved activation function. The improved activation function is obtained by introducing a positive slope into a specified negative interval of the ReLU activation function. The second initial model is iteratively trained using the pest and disease survival environment data and pest and disease reproduction rate in the training set as input data. During the iterative training process, the development trend of the predicted density of pests and diseases in the training set is used as the fitting target to adjust the parameters of the second initial model. If the expected fitting result is obtained, stop the iterative training and determine the corresponding second initial model as the second model.

6. The method of claim 5, wherein the method further comprises, The steps of inputting the pest and disease reproduction information and the environmental sample data into the second model, and outputting the predicted pest and disease density in the target monitoring area during the second time period, include: The pest and disease reproduction information obtained from the first model and the collected environmental sample data are input into the input layer of the second model; The pest and disease reproduction information and the environmental sample data are mapped and encoded through the encoding layer of the second model to obtain a set of encoded information. The encoded information includes at least the encoding vector corresponding to the pest and disease reproduction information and M time-series encoded segments corresponding to the environmental sample data, where M is a positive integer. Through the fitting layer of the second model, based on the self-attention mechanism and residual network, the pest and disease prediction density development analysis of the encoded information set is performed on an extended time scale to obtain a density-time fitting curve, wherein the extended time scale includes at least the union of the first time period and the second time period; Based on the density-time fitting curve, the predicted density of pests and diseases within the second time period is extracted, and the predicted density of pests and diseases is output through the output layer of the second model.

7. The method of claim 1, wherein the method is used for monitoring of wetland pests. The steps for determining pest and disease treatment plans based on the predicted pest and disease density include: Obtain a pre-set level threshold, wherein the level thresholds are ordered from largest to smallest as follows: normal value > warning value > alert value > risk value; If the predicted density of pests and diseases matches the normal value, the pest and disease treatment plan is determined to be to maintain regular frequency monitoring. If the predicted density of pests and diseases matches the warning value, the pest and disease treatment plan is determined to be to increase the monitoring frequency according to the first preset granularity and to take preventive biological control measures. When the predicted density of pests and diseases matches the warning value, the pest and disease treatment plan is determined to be to increase the monitoring frequency according to the second preset particle size and to use an environmentally friendly chemical control agent for local control. The second preset particle size is larger than the first preset particle size. The environmentally friendly chemical control agent refers to a chemical substance that meets the following conditions: its toxicity to non-pest and disease organisms is lower than a preset toxicity threshold at the specified dosage, and its degradation cycle is shorter than a preset time threshold. If the predicted density of pests and diseases matches the risk value, the pest and disease treatment plan is determined to be to increase the monitoring frequency according to the third preset granularity and to use the environmentally friendly chemical control agent for global control, wherein the third preset granularity is greater than the second preset granularity.

8. A device for monitoring wetland pests, characterized by include: The data acquisition unit is used to acquire environmental sample data of the target monitoring area within a first time period, wherein the environmental sample data is used to describe various survival indicators of the target monitoring area as a living environment for pests and diseases. The first input unit is used to input the environmental sample data into the first model and output pest and disease reproduction information. The first model is a Naive Bayes model pre-trained based on a pest and disease database, which is used to record the mapping relationship between the reproduction rate of pests and diseases and their living environment. The second input unit is used to input the pest and disease reproduction information and the environmental sample data into the second model and output the predicted density of pests and diseases in the target monitoring area during the second time period. The second model is a time-series prediction model pre-trained based on an improved activation function, and the second time period is a future time period of the first time period. The determining unit is used to determine the pest and disease treatment plan based on the predicted pest and disease density.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the wetland pest and disease monitoring method according to any one of claims 1 to 7.

10. An electronic device, comprising: It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the wetland pest and disease monitoring method according to any one of claims 1 to 7.

11. A computer program product, characterised in that, The method includes computer instructions, wherein when executed by a processor, the computer instructions implement the steps of the wetland pest and disease monitoring method according to any one of claims 1 to 7.

Images