Tea leaf insect control system and control method thereof

Abstract

The present application relates to a kind of tea insect control system and its management method, belong to the technical field of agricultural pest control, including the environmental information monitoring unit for real-time collection of tea garden environment data, the pest image monitoring unit for capturing and identifying tea tree pest image, with the central control unit of environmental information monitoring unit and pest image monitoring unit connection, and multi-modal control execution unit, method includes: data acquisition and fusion, pest analysis and prediction, intelligent decision generation and multi-modal precision control execution.In the present application, the real-time, accurate monitoring and early warning of tea garden pest situation is realized, through multi-modal collaborative control strategy, while effectively controlling pests, significantly reducing the amount of chemical pesticides, improving tea quality and safety, in line with the development requirements of green ecological agriculture.

Description

Technical Field

[0001] This invention relates to the field of agricultural pest and disease control technology, and in particular to a pest control system and method for tea. Background Technology

[0002] Tea is an important economic crop in my country, and its safe production is directly related to the industry's benefits and consumers' health. During the tea tree planting process, pests and diseases are one of the main factors affecting tea yield and quality. Traditional tea pest control mainly relies on manual garden inspections and experience-based judgment, and regular, large-scale spraying of chemical pesticides, which leads to problems such as excessive pesticide residues, environmental pollution, increased pesticide resistance in pests, and damage to the ecological balance of tea gardens.

[0003] To improve the level of intelligent pest and disease control, some monitoring and control solutions based on the Internet of Things, sensors and image recognition have emerged in existing technologies, but the following obvious limitations still exist;

[0004] 1. Limited functionality and lack of a closed-loop system: Existing solutions often focus on improving a single aspect. For example, CN113575544B discloses a smart management device for tea orchards with a monitoring system, which integrates monitoring, pest control, and water replenishment into a mobile management device. However, its functionality is limited to triggering fixed operations based on sensor signals, lacking the ability to predict pest trends and make intelligent decisions based on predictions. Another common solution is an organic tea pest monitoring device and control monitoring system disclosed in CN202222986271.1. Although it can monitor pests through video and insect-attracting lamps and send alarms via SMS, control actions still require manual intervention, resulting in a serious disconnect between monitoring and execution. None of these solutions have achieved a fully automated closed-loop process of "monitoring-prediction-decision-execution".

[0005] 2. Emphasis on monitoring over prevention and control, and a passive decision-making model. Many studies have focused on improving the accuracy and real-time nature of monitoring and using IoT sensors to collect environmental data to solve the monitoring delay problem. However, the functions of such systems are limited to data collection and display. They are not linked with prevention and control equipment and cannot form proactive intervention. Their early warning methods are still passive and cannot predict and prevent pest outbreaks.

[0006] 3. The prevention and control methods are extensive and lack green synergy. Existing prevention and control devices mostly adopt regional and full-coverage spraying modes. For example, a tea and fruit garden intelligent management device with a monitoring device published by CN113575544B uses a smoke generator box for chemical pest control, which cannot achieve precise targeted application of pesticides to areas with high pest infestation. At the same time, existing technologies generally lack the design to intelligently coordinate physical, biological and other green prevention and control methods with chemical control, resulting in a high degree of dependence on chemical pesticides.

[0007] 4. Lack of model-driven and adaptive optimization: Although a few solutions introduce predictive models, they are mostly limited to trend display or single disease prediction. They fail to dynamically link the prediction results with multimodal prevention and control strategies. Moreover, the systems usually do not have the ability to continuously optimize the prediction models and decision rules using feedback data on prevention and control effects, and cannot adapt to the dynamic changes in the tea garden environment.

[0008] In summary, existing technologies have not yet provided a comprehensive pest control solution for tea that integrates real-time and accurate monitoring, intelligent prediction and early warning, multimodal green prevention and control, and system self-optimization. Therefore, there is an urgent need for an innovative system and method that can achieve fully closed-loop intelligent management, effectively control pests while minimizing the use of chemical pesticides, and ensure the quality of tea and the ecological safety of tea gardens. Summary of the Invention

[0009] This invention provides a tea pest control system and method, which solves the problems mentioned in the background technology above and realizes intelligent, precise and green integrated management of tea pests.

[0010] The solution of the present invention to the above-mentioned technical problems is as follows:

[0011] On the one hand, a method for pest control in tea includes the following steps: S1, multi-source data acquisition and processing, collecting multi-dimensional environmental time-series data of tea garden through a distributed environmental information monitoring unit; at the same time, acquiring images through a pest image monitoring unit, and using a deep learning-based target detection model for real-time pest identification, outputting insect population density data with spatiotemporal labels;

[0012] S2, Data Fusion and Feature Construction: The central control unit receives environmental time-series data and insect population density data, performs spatiotemporal alignment and matching, and constructs a fusion feature vector sequence that includes temperature, humidity, light intensity and historical insect population density.

[0013] S3, Pest Dynamic Prediction, inputs the fused feature vector sequence into the pre-trained pest prediction model, and outputs the predicted probability of pest occurrence and pest density in a specified area within a specified time period in the future; the pest prediction model is a two-layer long short-term memory network model with an attention mechanism. This model is trained using historical tea garden datasets and can capture the non-linear temporal relationship between environmental factors and pests.

[0014] S4, Intelligent Decision Command Generation: The central control unit compares the predicted probability of pest occurrence and pest density with the dynamic prevention and control threshold in the prevention and control strategy knowledge base, and generates prevention and control decision commands that include prevention and control modality, geographical area, operation time and intensity parameters according to the preset green prevention and control priority decision rules; the dynamic prevention and control threshold is periodically adaptively adjusted based on historical prevention and control effect data.

[0015] S5, multimodal precision execution and closed-loop optimization: The multimodal prevention and control execution unit receives and executes prevention and control decision instructions; after execution, the system continues to collect subsequent data to evaluate the prevention and control effect, and feeds back the evaluation results to the central control unit for iterative optimization of the parameters of the pest prediction model and the dynamic prevention and control threshold.

[0016] Based on the above technical solution, the present invention can be further improved as follows.

[0017] Furthermore, in step S4, the green prevention and control priority decision rule is specifically as follows: a. When the predicted insect population density is lower than the first dynamic threshold, only a monitoring record instruction is generated;

[0018] b. When the predicted insect population density is between the first and second dynamic thresholds, the command to activate the physical control module or the biological control module will be generated first.

[0019] c. The instruction to activate the targeted chemical control module is generated only when the predicted insect population density is higher than the second dynamic threshold and historical feedback data indicates that the green control effect has not met expectations.

[0020] A tiered decision-making process of "monitoring and early warning → physical / biological control → precision chemical control" has been established to minimize the use of chemical pesticides, which helps reduce pesticide residues in tea, protect the ecological environment and biodiversity of tea gardens. By setting dynamic thresholds and linking them with historical effect data, decisions are not only based on current predictions but also take into account the actual effectiveness of past control measures, avoiding rigidity and blindness in decision-making. Prioritizing physical and biological control methods can effectively control pests in most cases, thereby saving the cost of purchasing and applying chemical pesticides and delaying the development of pesticide resistance in pests.

[0021] Furthermore, in step S5, when the targeted chemical control module is activated, the control decision instruction includes a variable application prescription map generated based on the predicted spatial distribution map of insect population density. The multimodal control execution unit performs path planning based on the prescription map and controls the nozzles to achieve variable spraying in different areas. This changes the traditional uniform spraying mode and performs "on-demand application" based on the spatial heterogeneity of pest occurrence, which can significantly reduce the total amount of chemical pesticides used and waste. It concentrates pesticide spraying on the areas that truly need control, improves targeting, and helps to better control pests. At the same time, combined with path planning, the operation process is optimized. By reducing pesticide application in non-target areas, the negative impact on soil, water sources, and non-target organisms is effectively reduced. This is a concrete manifestation of smart agriculture and precision agriculture technologies.

[0022] On the other hand, a tea pest control system includes an environmental information monitoring unit, a pest image monitoring unit, a central control unit, and a multimodal control execution unit. The environmental information monitoring unit is distributed and fixedly deployed in the tea garden to collect real-time data on the microenvironment of tea tree growth.

[0023] The pest image monitoring unit is distributed and fixedly deployed in the tea garden to collect image and video data containing tea trees, and to identify and count the pests in the image and video data.

[0024] The central control unit is connected to the environmental information monitoring unit and the pest image monitoring unit through a communication network, and is used to perform data fusion, dynamic prediction and intelligent decision-making steps.

[0025] The multimodal prevention and control execution unit is connected to the central control unit through a communication network and is used to receive and execute prevention and control decision instructions.

[0026] Furthermore, the central control unit includes a data receiving and storage module for receiving and storing data from the monitoring unit, a data fusion and feature extraction module for constructing a fused feature vector sequence, a pest prediction model module with a built-in two-layer LSTM prediction model incorporating an attention mechanism, and an intelligent decision generation module with a built-in knowledge base of prevention and control strategies and green prevention and control priority decision rules. The modular design makes the data flow and task process clear, with each part performing its own function and working together, ensuring the orderly and stable operation of the system.

[0027] Furthermore, the multimodal control execution unit includes a physical control module, a biological control module, and a targeted chemical control module. The targeted chemical control module is an autonomous mobile application platform equipped with a Beidou / GPS positioning and navigation system, which can perform precise spraying operations according to the variable application prescription map. It integrates physical, biological, and chemical modes, and the system can flexibly call the most suitable single or combined control methods according to decision instructions to cope with complex field conditions.

[0028] The beneficial effects of this invention are as follows: This invention provides a tea pest control system and method, which has the following advantages:

[0029] 1. The prevention and control are highly intelligent and precise, deeply integrating the Internet of Things and precision agricultural equipment, realizing full-process intelligence of pest perception, trend prediction, decision generation, and precise action execution, which greatly reduces reliance on human experience and misjudgment.

[0030] 2. Significantly improves ecological and agricultural product safety benefits. The green-first decision-making rules and the variable spraying technology of applying pesticides on demand can minimize the amount and scope of chemical pesticides used, effectively reduce the risk of pesticide residues in tea, reduce damage to tea garden soil, water sources and beneficial organisms, protect the ecological balance of tea gardens, and ensure the quality and safety of tea.

[0031] 3. Improved prevention and control efficiency and economy: The automated operation of the system saves a lot of manpower costs for manual inspection and decision-making. Precise early warning and targeted prevention and control avoid unnecessary general application and save pesticide costs. Proactive prevention and control can control pests before or in the early stage of outbreaks, reducing yield and quality losses caused by pests, resulting in significant overall economic benefits.

[0032] 4. Strong system adaptability and sustainability: It has a closed-loop optimization function, which makes the system no longer static, but can dynamically adjust its prediction accuracy and decision rationality by continuously learning historical data and prevention and control feedback, so as to better adapt to the environmental characteristics and pest occurrence patterns of specific tea gardens and maintain efficient and scientific prevention and control effects in the long term.

[0033] 5. High degree of technical integration and good scalability: The system adopts a modular design, with clear boundaries between the functional modules of the central control unit, efficient collaboration, and flexible multimodal execution unit structure. This design not only ensures the stability of system operation, but also facilitates the integration of new monitoring sensors, algorithm models, or prevention and control equipment in the future, and has good scalability.

[0034] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0035] 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:

[0036] Figure 1 A flowchart illustrating a tea pest control system and method according to an embodiment of the present invention;

[0037] Figure 2 This is a system architecture diagram of a tea pest control system and method provided in an embodiment of the present invention. Detailed Implementation

[0038] The following is in conjunction with the appendix Figure 1-2The principles and features of the present invention are described below. The examples given are for illustrative purposes only and are not intended to limit the scope of the invention. The invention is described more specifically in the following paragraphs by way of example with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0040] like Figure 1 As shown, the present invention provides a method for pest control in tea leaves, which is implemented according to the following steps;

[0041] Step S1, Multi-source data acquisition and processing. In this embodiment, the system is deployed in a standardized tea garden with an area of ​​about 50 acres. First, the monitoring units are deployed in a network.

[0042] The environmental information monitoring unit consists of 30 IoT sensor nodes, which are uniformly distributed and fixedly deployed in the tea garden in a grid pattern. Each node integrates a high-precision digital temperature and humidity sensor, light intensity sensor and soil moisture sensor, and collects the micro-environment time series data of the tea tree canopy at a frequency of once per minute, and uploads it to the central control unit through the LoRa wireless network.

[0043] The pest image monitoring unit has 15 high-definition smart cameras with waterproof shells fixedly deployed in key locations in the tea garden (such as different variety areas and areas with high incidence of pests in previous years). These cameras automatically collect high-definition images of tea leaves and tender shoots every day in the early morning and evening when pests are most active. The collected images are transmitted to the edge server deployed in the central control unit via 4G / 5G network. The server has a built-in pest target detection model trained and optimized based on the YOLOv5 framework. This model is specifically trained for major tea tree pests such as tea green leafhopper and tea geometrid moth. It can automatically identify and count the uploaded images in real time and automatically add spatiotemporal tags such as collection time and camera number (corresponding geographical coordinates) to the identification results. Finally, it outputs structured "pest population density data".

[0044] Step S2, data fusion and feature construction: The data receiving and storage module of the central control unit continuously receives data streams from the two monitoring units, and the data fusion and feature extraction module initiates a fusion task every 2 hours.

[0045] Now, spatiotemporal alignment is performed. Based on the geographic coordinates of each environmental sensor node and camera, the environmental time series data (temperature, humidity, light) within the time window (such as the previous 24 hours) is matched and aligned with the insect population density data of the nearest spatial location in the same time period.

[0046] Next, feature construction is performed. For a specific "virtual grid area" within the tea garden, defined by coordinates, the average environmental data for every 2 hours over the past 24 hours is extracted and combined with the insect population density value at the corresponding time point. Simultaneously, the historical insect population density value for the area over the past 3 days is used as an additional temporal feature. Finally, a fused feature vector sequence is constructed for each area, for example: In this sequence, T, H, and L represent temperature, humidity, and light intensity, P represents insect population density, and the subscript represents the historical time step. This sequence reflects the coupling relationship between environmental dynamics and insect population evolution.

[0047] Step S3: Pest dynamic prediction. The constructed fusion feature vector sequence is fed into the pre-trained pest prediction model. In this embodiment, the model is a two-layer long short-term memory network (Attention-based Bi-LSTM) that introduces an attention mechanism.

[0048] The first LSTM layer of the model structure is used to learn the deep temporal dependencies of the input feature sequence; the attention mechanism layer can automatically "focus" on the key historical moments (such as specific temperature change periods) that have the greatest impact on predicting future insect infestations; the second LSTM layer further integrates attention-weighted information.

[0049] For training and application, the model is trained and validated offline using historical environmental and pest monitoring datasets from the tea garden and neighboring similar tea gardens over the past three years. When applied online, the model receives the current feature sequence and outputs the "predicted population density of tea green leafhopper" and its "probability level of pest occurrence (high, medium, low)" for the specific area in the next 24-72 hours. The model can effectively capture nonlinear and lagging temporal relationships such as "the population density may surge after continuous high temperature and humidity".

[0050] Step S4: Intelligent decision-making instruction generation. The intelligent decision-making generation module compares the prediction results obtained in step S3 with the rules in the prevention and control strategy knowledge base.

[0051] The knowledge base has a first dynamic threshold (prevention threshold) and a second dynamic threshold (intervention threshold). The initial values ​​are set by agronomic experts. The system will fine-tune these two thresholds once a week based on the actual effect feedback of control measures taken for similar prediction levels in the past month (such as the insect population reduction rate 3 days after pesticide application) to achieve self-adaptation.

[0052] The green prevention and control priority decision-making rules are applied as follows:

[0053] Scenario A: If the insect population density in a certain area is predicted to be 3 insects / 100 leaves in the next 48 hours (lower than the first dynamic threshold of 5 insects / 100 leaves), the decision module will only generate one "strengthen monitoring record" instruction and mark the area, and will no longer execute any control actions triggered by the unit.

[0054] Scenario B: Predicting an insect population density of 8 insects per 100 leaves in another area (between the first threshold of 5 and the second threshold of 15), the decision module first queries the knowledge base. If the current season and temperature are suitable, it generates the instruction: "At 6:00 AM tomorrow, activate the physical control module (such as a cluster of solar-powered insecticidal lamps) in area G-07 for 2 hours."

[0055] Scenario C: The insect population density in a certain area is predicted to reach 20 insects / 100 leaves (higher than the second threshold of 15), and the system log shows that biological control (releasing natural enemies) was used in this area last week, but the effect evaluation did not meet the standard. At this time, the decision module generates the instruction: "Start the targeted chemical control module and carry out precise spraying after 4 pm according to the variable application prescription map of area R-12".

[0056] Step S5, Multimodal Precise Execution and Closed-Loop Optimization: The multimodal prevention and control execution unit receives and parses the instructions.

[0057] Physical / biological control implementation: After receiving the instruction from scenario B, the insecticidal lamp group located in area G-07 is remotely controlled to turn on at a set time via the IoT gateway;

[0058] Targeted chemical control implementation (combined) Figure 2 For scenario C, the system first generates a "variable pesticide application prescription map" based on the spatial interpolation of the predicted insect population density on the tea garden map. The map divides the area R-12 into three pesticide-required zones: high, medium, and low. The command is sent to the unmanned spraying vehicle equipped with the Beidou RTK positioning and navigation system. After the prescription map is loaded, the vehicle automatically plans a full-coverage route and dynamically controls the opening and closing of each nozzle and the flow rate based on real-time location information during the journey, so as to achieve "on-demand pesticide application" by focusing on spraying high-density areas and reducing or closing spraying in low-density areas.

[0059] Closed-loop optimization: After the control action is implemented, the system continues to collect actual environmental and pest data for the next 3-7 days through steps S1 and S2 to evaluate the "effectiveness coefficient" (such as the pest population reduction rate) of this control. This coefficient, together with the predicted data and decision instructions on which this decision was based, forms a feedback data package, which is used for:

[0060] Model optimization: Regularly (e.g., monthly) add new feedback data to the training set to incrementally train or fine-tune the pest prediction model and improve its prediction accuracy.

[0061] Threshold optimization: The dynamic control threshold is automatically adjusted based on the effectiveness coefficient. For example, if multiple green control measures have shown good results, the second threshold can be appropriately increased to further delay the initiation of chemical control.

[0062] like Figure 2 As shown, the hardware and software architecture of the tea pest control system used to implement the above method is as follows:

[0063] 1. Sensing layer: includes environmental information monitoring unit and pest image monitoring unit;

[0064] The environmental information monitoring unit consists of distributed, fixed-deployment Internet of Things (IoT) sensor nodes and wireless transmission networks (such as LoRa gateways);

[0065] The pest image monitoring unit consists of a high-definition intelligent camera and an image preprocessing server;

[0066] The transmission layer adopts a hybrid network of 4G / 5G and LoRa to ensure reliable data backhaul with low power consumption;

[0067] 2. Platform Layer (Central Control Unit): Includes data receiving and storage module, data fusion and feature extraction module, pest prediction model module, intelligent decision generation module, and model and strategy optimization module;

[0068] The data receiving and storage module uses time-series and relational databases to classify and store all raw and processed data; the data fusion and feature extraction module executes spatiotemporal alignment, data cleaning, and feature vector construction algorithms; the pest prediction model module deploys and runs a trained Attention-based Bi-LSTM model and provides a prediction API interface; the intelligent decision generation module has a built-in knowledge base of prevention and control strategies containing dynamic thresholds and green priority rules, as well as a decision reasoning engine; the model and strategy optimization module is responsible for processing feedback data and executing model retraining and threshold adjustment algorithms.

[0069] 3. Execution Layer (Multimodal Prevention and Control Execution Unit): Includes physical control module, biological control module, and targeted chemical control module;

[0070] The physical control module uses remotely controllable insecticidal lamps, insect-attracting boards, etc.; the biological control module uses automatic release devices for natural enemy insects, pheromone traps, etc.; the targeted chemical control module uses unmanned spraying vehicles that integrate high-precision navigation, variable spraying controllers, and pesticide management systems.

[0071] 4. The application layer provides web and mobile visualization interfaces for data dashboards, display of early warning information, manual review of instructions, and configuration of system parameters;

[0072] The various units of the system work together through a communication network to form a closed loop of "sensing-transmission-analysis-decision-execution-feedback", realizing intelligent, precise and green management of tea pests.

[0073] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Content not described in detail in this specification is prior art known to those skilled in the art.

[0074] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Those skilled in the art can readily implement the present invention based on the accompanying drawings and the above description. However, any modifications, alterations, or variations made by those skilled in the art without departing from the scope of the present invention, utilizing the disclosed technical content, are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, or variations made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the present invention.

Claims

1. A method for controlling insect pests in tea leaves, characterized in that, Includes the following steps: S1, multi-source data acquisition and processing, collects multi-dimensional environmental time-series data of tea garden through distributed environmental information monitoring units; at the same time, it collects images through pest image monitoring units and uses a deep learning-based target detection model to identify pests in real time, outputting pest population density data with spatiotemporal labels. S2, Data Fusion and Feature Construction: The central control unit receives environmental time-series data and insect population density data, performs spatiotemporal alignment and matching, and constructs a fusion feature vector sequence that includes temperature, humidity, light intensity and historical insect population density. S3, Pest Dynamic Prediction, inputs the fused feature vector sequence into the pre-trained pest prediction model, and outputs the predicted probability of pest occurrence and pest density in a specified area within a specified time period in the future; the pest prediction model is a two-layer long short-term memory network model with an attention mechanism. This model is trained using historical tea garden datasets and can capture the non-linear temporal relationship between environmental factors and pests. S4, Intelligent Decision Command Generation: The central control unit compares the predicted probability of pest occurrence and pest density with the dynamic prevention and control threshold in the prevention and control strategy knowledge base, and generates prevention and control decision commands that include prevention and control modality, geographical area, operation time and intensity parameters according to the preset green prevention and control priority decision rules; the dynamic prevention and control threshold is periodically adaptively adjusted based on historical prevention and control effect data. S5, multimodal precision execution and closed-loop optimization: The multimodal prevention and control execution unit receives and executes prevention and control decision instructions; after execution, the system continues to collect subsequent data to evaluate the prevention and control effect, and feeds back the evaluation results to the central control unit for iterative optimization of the parameters of the pest prediction model and the dynamic prevention and control threshold.

2. The method for controlling insect pests in tea leaves according to claim 1, characterized in that, In step S4, the green prevention and control priority decision rule is specifically as follows: a. When the predicted insect population density is lower than the first dynamic threshold, only a monitoring record instruction is generated; b. When the predicted insect population density is between the first and second dynamic thresholds, the command to activate the physical control module or the biological control module will be generated first. c. The instruction to activate the targeted chemical control module is generated only when the predicted insect population density is higher than the second dynamic threshold and historical feedback data indicates that the green control effect has not met expectations.

3. The method for controlling insect pests in tea leaves according to claim 1, characterized in that, In step S5, when the targeted chemical control module is activated, the control decision instruction includes a variable application prescription map generated based on the predicted insect population density spatial distribution map. The multimodal control execution unit performs path planning based on the prescription map and controls the nozzles to achieve variable spraying in different areas.

4. A tea pest control system for implementing any one of the tea pest control methods of claims 1 to 3, comprising an environmental information monitoring unit, a pest image monitoring unit, a central control unit, and a multimodal control execution unit, characterized in that, The environmental information monitoring units are distributed and fixedly deployed in the tea garden to collect real-time data on the microenvironment for tea tree growth. The pest image monitoring unit is distributed and fixedly deployed in the tea garden to collect image and video data containing tea trees, and to identify and count the pests in the image and video data. The central control unit is connected to the environmental information monitoring unit and the pest image monitoring unit through a communication network, and is used to perform data fusion, dynamic prediction and intelligent decision-making steps. The multimodal prevention and control execution unit is connected to the central control unit through a communication network and is used to receive and execute prevention and control decision instructions.

5. The tea pest control system according to claim 4, characterized in that, The central control unit includes a data receiving and storage module for receiving and storing data from the monitoring unit, a data fusion and feature extraction module for constructing a fused feature vector sequence, a pest prediction model module with a built-in two-layer LSTM prediction model incorporating an attention mechanism, and an intelligent decision generation module with a built-in knowledge base of prevention and control strategies and green prevention and control priority decision rules.

6. The tea pest control system according to claim 4, characterized in that, The multimodal control execution unit includes a physical control module, a biological control module, and a targeted chemical control module. The targeted chemical control module is an autonomous mobile application platform equipped with a Beidou / GPS positioning and navigation system, which can perform precise spraying operations according to the variable application prescription map.

Images