Long-acting prevention and control device and method for solenopsis invicta

Abstract

The invention provides a long-acting solenopsis invicta prevention and control device and method, and belongs to the technical field of agriculture and forestry pest prevention and control. The stainless steel cylinder body is arranged at the top end of the fixed bracket; the bait tray is arranged in the middle of the fixing support and located under the stainless steel cylinder body; the main controller is arranged in the stainless steel cylinder body; the antenna is arranged on the outer side wall of the stainless steel cylinder body and connected with the main controller, and the receiving and transmitting direction is upward; the battery is arranged in the stainless steel cylinder body and is connected with the main controller; the solar panel is obliquely arranged at the top end of the stainless steel cylinder body and is connected with the battery; the bait storing and feeding mechanism is arranged in the stainless steel cylinder body and connected with the main controller, and the feeding direction directly faces the bait tray; and the camera is arranged in the stainless steel cylinder body and is connected with the main controller, and the shooting direction of the camera directly faces the bait tray. The method has the advantages that environmental adaptability and pesticide application flexibility of solenopsis invicta prevention and control are greatly improved, and pesticide residues and environmental pollution risks are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of prevention and control of pests in agriculture and forestry, and particularly relates to a long-acting prevention and control device and method for red imported fire ants. BACKGROUND

[0002] The red imported fire ant (Solenopsis invicta) is listed as one of the world's most harmful invasive species by the International Union for Conservation of Nature (IUCN), and has posed a serious threat to the agroforestry ecosystem, biodiversity, public infrastructure safety and human and animal health in the invaded areas. In response to this ecological risk, a large amount of resources has been invested in systematic prevention and control in various regions. At present, chemical control is still the main means of red imported fire ant control, and bait is the preferred type of pesticide because of its long-lasting effect, complete killing and high efficiency in destroying the overall structure of the ant colony.

[0003] Most of the existing red imported fire ant baits are applied by scattering or fixed-point placement. In order to improve the baiting effect and stability, some simple bait station devices have been developed in some existing technologies to fix the position of the bait and reduce the influence of the external environment. However, the existing control methods still have the following obvious defects: 1. The control effect of the bait is easily affected by the environmental humidity. After application, if it rains or the humidity is high, the physical and chemical properties of the bait are likely to change, resulting in a significant decrease in palatability, thereby inhibiting the red imported fire ant from feeding and seriously affecting the control effect. Therefore, in order to ensure that the bait is effectively ingested, it is usually necessary to select appropriate weather conditions for pesticide application and cooperate with continuous monitoring, which has poor flexibility and high implementation cost.

[0004] 2. The red imported fire ant is widely distributed in southern regions, and its breeding peak is concentrated in spring and autumn, which is the key period for prevention and control. However, in southern regions, it rains frequently and the air humidity is high, and the weather window for pesticide application is short, which significantly increases the difficulty and uncertainty of the prevention and control operation.

[0005] 3. The uningested bait remaining in the environment not only causes waste of pesticides, but also may cause environmental pollution risk due to water and soil migration.

[0006] Therefore, how to provide a long-acting prevention and control device and method for red imported fire ants to improve the environmental adaptability and flexibility of pesticide application for red imported fire ant control and reduce the risk of pesticide residue and environmental pollution has become a technical problem to be solved. SUMMARY

[0007] The technical problem to be solved by the present application is to provide a long-acting prevention and control device and method for red imported fire ants to improve the environmental adaptability and flexibility of pesticide application for red imported fire ant control and reduce the risk of pesticide residue and environmental pollution.

[0008] In a first aspect, the present application provides a long-acting prevention and control device for red imported fire ants, comprising: a fixed support; a stainless steel cylinder body, provided at the top end of the fixed support; a bait tray, provided at the middle of the fixed support, directly below the stainless steel cylinder body; a main controller, provided inside the stainless steel cylinder body; an antenna, provided on the outer wall of the stainless steel cylinder body, connected with the main controller, and the receiving and transmitting direction is upward; a battery, provided inside the stainless steel cylinder body, connected with the main controller; a solar panel, provided at the top end of the stainless steel cylinder body, connected with the battery; a bait storage and dispensing mechanism, provided inside the stainless steel cylinder body, connected with the main controller, and the dispensing direction is directly opposite to the bait tray; a camera, provided inside the stainless steel cylinder body, connected with the main controller, and the shooting direction is directly opposite to the bait tray.

[0009] Further, the main controller is integrated with a wireless communication module, and is connected with the antenna.

[0010] Further, the shape of the solar panel matches the shape of the stainless steel cylinder body.

[0011] Further, the bait storage and dispensing mechanism comprises: a funnel feeding part, provided inside the stainless steel cylinder body; a humidity sensor, installed inside the stainless steel cylinder body through a humidity sensor fixing bracket, located above the funnel feeding part, and connected with the main controller; a heating sheet, installed inside the stainless steel cylinder body through a heating sheet fixing bracket, located above the funnel feeding part, and connected with the main controller; an upper cover plate, installed at the top end of the stainless steel cylinder body through a sealing ring, located below the solar panel; a solenoid valve, with the input end communicated with the output end of the funnel feeding part, and the control end connected with the main controller; a discharging part, with the input end communicated with the output end of the solenoid valve, and the discharging end directly opposite to the bait tray.

[0012] Further, the bait storage and dispensing mechanism further comprises: three PG joints, the humidity sensor, the heating sheet and the solenoid valve are respectively connected with the main controller through a PG joint.

[0013] Further, the shape of the upper cover plate matches the shape of the stainless steel cylinder body.

[0014] In a second aspect, the present application provides a long-term prevention and control method of Solenopsis invicta, comprising the following steps: Step S1, deploying the long-term prevention and control device of Solenopsis invicta, adding bait to the bait storage and dispensing mechanism, and establishing a connection between the main controller and the server through an antenna; Step S2, the main controller controls the electromagnetic valve to dispense a preset amount of bait onto the bait tray based on the prevention and control instructions sent by the server; Step S3, the main controller collects images of the tray in real time through the camera, inputs the tray images into the pre-deployed ant species identification model, foraging behavior identification model and drug efficacy identification model in real time, and obtains ant species identification results, foraging behavior identification results and drug efficacy identification results; Step S4, the main controller dynamically opens and closes the electromagnetic valve based on the ant species identification results and foraging behavior identification results, dynamically adjusts the sampling rate of the camera, records the opening and closing times of the electromagnetic valve, inputs the ant species identification results, foraging behavior identification results, drug efficacy identification results and opening and closing times into the pre-deployed prevention and control state discrimination model, obtains the prevention and control state discrimination results and uploads them to the server; Step S5, the main controller dynamically controls the working of the heating sheet based on the sensing signal of the humidity sensor.

[0015] Further, in step S3, the ant species identification model is constructed based on an image feature extraction layer, a multi-scale feature fusion layer, an attention feature enhancement layer and a classification prediction layer; The image feature extraction layer is constructed based on a convolution feature extraction module, a Transformer context integration module and an attention gate module; the convolution feature extraction module is used to extract initial multi-scale feature maps from the tray images; the Transformer context integration module is used to integrate global spatial context information in the initial multi-scale feature maps; and the attention gate module is used to filter important features from the initial multi-scale feature maps with integrated global spatial context information, and output multi-scale visual features; The multi-scale feature fusion layer is constructed based on a feature pyramid network; the feature pyramid network is used to fuse visual features of different scales through top-down and horizontal connection, and obtain fusion features; The attention feature enhancement layer is constructed based on a spatial attention module and a channel attention module; the spatial attention module is used to enhance the spatial important region in the fusion features; and the channel attention module is used to weight the important channel features in the fusion features with enhanced spatial important region, and obtain enhanced features; The classification prediction layer is constructed based on a full connection classification module; the full connection classification module is used for mapping the enhanced features to a probability distribution of presence of fire ants, presence of non-fire ants, and absence of ants, and then outputting an ant species identification result of presence of fire ants, presence of non-fire ants, or absence of ants.

[0016] Further, in the step S3, the foraging behavior recognition model is constructed based on an image preprocessing layer, a spatial feature extraction layer, a time modeling layer, a change enhancement layer, and a behavior prediction layer; The image preprocessing layer is used for performing standardization and enhancement operations on the input tray image, and outputs a preprocessed image sequence; The spatial feature extraction layer is used for extracting spatial features of each frame from the preprocessed image sequence, and outputs a feature vector sequence of each frame; The time modeling layer is used for performing time-dependent modeling on the feature vector sequence, capturing the time dynamics of the foraging behavior, and outputting time fusion features; The change enhancement layer is used for enhancing the change information in the time fusion features, highlighting the key patterns of bait reduction, and outputting change enhanced features; The behavior prediction layer is used for predicting a behavior probability distribution based on the change enhanced features, and then outputting a foraging behavior recognition result of bait not reduced, bait reduced, or bait basically disappeared; The efficacy recognition model is constructed based on an image feature abstraction layer, an event detection layer, a relationship fusion layer, a feature enhancement layer, and an efficacy prediction layer; The image abstraction layer is used for extracting tray spatial features from the input tray image through a 2D convolutional neural network, and extracting tray time sequence features from each of the tray spatial features through a ViViT network; The event detection layer is used for detecting bait reduction events from the tray time sequence features through a twin 3D convolutional network and a change point detection network, and detecting fire ant death events from the tray time sequence features through an inflation 3D convolutional network and a graph convolutional network; The relationship fusion layer is used for extracting event order relationships from the bait reduction events and the fire ant death events through a time sequence relationship network, dynamically weighting contributions of the bait reduction events and the fire ant death events through a cross-attention mechanism, and fusing the time sequence relationships through a gated recurrent unit to obtain relationship features; The feature enhancement layer enhances the relationship features through self-supervised contrast learning to obtain an enhanced feature vector; The efficacy prediction layer is used for mapping the enhanced feature vector to hidden features through a full connection layer, and outputting a binary classification probability of bait failure or bait not failed through a Sigmoid activation function, and then obtaining an efficacy recognition result of bait failure or bait not failed.

[0017] Further, in the step S4, the prevention and control state discrimination model is constructed based on a decision tree algorithm, and is used for outputting a prevention and control state discrimination result of continuing prevention and control, supplementing bait, replacing bait or completing prevention and control according to the ant species identification result, the foraging behavior identification result, the drug effect identification result and the switch number.

[0018] The advantages of the present application are: 1. By integrating solar power supply and wireless communication, the red imported fire ant long-acting prevention and control device can be deployed in the wild for a long time and receive remote instructions, breaking through the weather and geographical restrictions and significantly improving the environmental adaptability and flexibility of pesticide application; at the same time, the red imported fire ant long-acting prevention and control device ensures the dry storage of bait through a sealing structure and a heating and moisture-proof system, and uses a camera and an AI model to identify target ant colonies and their foraging behavior in real time, thereby accurately controlling the baiting time and dosage and realizing "on-demand feeding", thereby fundamentally reducing the invalid exposure and residue of bait, and ultimately greatly improving the environmental adaptability and flexibility of red imported fire ant prevention and control, and reducing the risk of pesticide residue and environmental pollution.

[0019] 2. By using a combination of solar panels and batteries for power supply, the solar panels are inclined at the top of the stainless steel cylinder body and match the shape of the cylinder body, which can efficiently collect solar energy and ensure the long-term operation of the device in the wild without relying on external power grids; this design makes the device suitable for remote areas or outdoor environments where red imported fire ants are common, reduces the frequency of manual maintenance, and improves the continuity and reliability of prevention and control; at the same time, the stainless steel cylinder body and the fixed support provide corrosion resistance and stability, enhancing the durability of the device in harsh weather.

[0020] 3. By integrating a camera and a main controller and combining a pre-deployed AI model (such as an ant species identification model, a foraging behavior identification model, and a drug effect identification model), the device can collect images of the feeding tray in real time, automatically identify the presence and behavior dynamics of red imported fire ants, and automatically identify whether the bait is ineffective; based on the identification results, the main controller dynamically adjusts the bait feeding (such as opening and closing the electromagnetic valve) and the camera sampling rate to achieve accurate response; this adaptive mechanism avoids overfeeding of bait, improves prevention and control efficiency, and reduces the need for human intervention through data-driven decision-making, reflecting the technological progress of intelligentization.

[0021] 4. The main controller integrates a wireless communication module and connects with a server through an antenna, allowing remote reception of prevention and control instructions and uploading of data (such as prevention and control state discrimination results), which enables users to monitor the status of multiple devices in real time, conduct centralized scheduling and data analysis, and greatly improve the convenience and scalability of prevention and control management; the wireless communication direction is designed upwards to optimize signal reception and ensure stable communication in complex environments, in line with the trend of the Internet of Things.

[0022] 5、The bait storage and dispensing mechanism includes components such as solenoid valves, humidity sensors, and heating elements, which can dynamically control the operation of the heating elements based on environmental humidity, preventing the bait from getting wet and ensuring its effectiveness. The main controller accurately controls the amount of bait to be dispensed based on control instructions and identification results, and the direction of bait dispensing is directly towards the bait tray, reducing waste. This precise dispensing not only saves bait costs but also reduces environmental impact, highlighting environmental protection and economic benefits.

[0023] 6、The modular structure is adopted, such as the bait storage and dispensing mechanism connected by PG joints, the upper cover plate matched with the barrel body and sealed, facilitating installation and maintenance. This design simplifies bait refilling, component replacement, and other operations, reducing long-term use costs. At the same time, modularization helps standardize production, improves the scalability and adaptability of the device, and meets different scene requirements.

[0024] 7、The control method combines multi-level AI models (such as feature pyramid networks and attention mechanisms) to achieve high-precision ant identification and behavior analysis, and outputs intelligent decisions (such as supplementing bait) through a control state discrimination model (based on decision trees). This method deeply integrates hardware and software to form a closed-loop control system, improving the scientific nature and long-term effectiveness of red ant control, representing a technological innovation in biological control.

[0025] 8、By integrating solar power supply systems, stainless steel barrels, and fixed supports, long-term energy supply and excellent environmental adaptability are achieved, ensuring the stable operation of the device in the wild for a long time. At the same time, with the help of intelligent main controllers, cameras, and pre-deployed AI models (such as ant species identification and foraging behavior identification models), real-time monitoring and adaptive control of red ants are achieved, which can dynamically adjust bait dispensing and sampling rates to improve control accuracy. In addition, the wireless communication module supports remote data interaction and centralized management, while the bait storage and dispensing mechanism combined with humidity sensors and heating elements ensures the effectiveness and resource optimization of the bait. Modular design (such as PG joints and sealed upper covers) simplifies maintenance, and the overall method integrates hardware and advanced algorithms to form a closed-loop control system, significantly improving efficiency, reducing human intervention, and having environmental and economic benefits.

[0026] 9、By integrating humidity sensors and heating elements, a stable microenvironment is created, effectively overcoming the problem of deterioration of the physicochemical properties and decreased palatability of traditional bait in high-humidity environments, significantly improving the stability and persistence of the bait, reducing bait waste and soil and water pollution, thereby reducing the secondary impact on non-target organisms and the ecological environment, and assisting in ecological safety and biodiversity protection.

[0027] 10、Adopting waterproof sealing design and edge computing architecture, it realizes all-weather reliable operation and real-time data processing, combines deep learning and decision tree algorithm, accurately identifies the species and feeding behavior of Solenopsis invicta, realizes dynamic monitoring and intelligent evaluation of prevention and control effect, greatly improves the scientificity and response speed of prevention and control operation. Especially in the climate conditions of high humidity and short prevention and control window period in the south, the prevention and control cycle can be effectively prolonged, and the success rate of prevention and control can be significantly improved. In addition, the device has self-adaptive regulation and control capability, reduces the frequency of manual inspection and labor cost, reduces the complexity and uncertainty of prevention and control management, and promotes the intelligent, precise and sustainable transformation of Solenopsis invicta prevention and control.

[0028] 11、Through highly automated and intelligent prevention and control process, the precise and efficient Solenopsis invicta management is realized, which has the advantages of using the cooperation of the main controller and the server to automatically release bait and analyze image data in real time based on multiple models (such as ant species identification, feeding behavior identification and drug effect identification model), ensuring the accuracy of identification; the dynamic self-adaptive control mechanism can adjust the electromagnetic valve, camera sampling rate and heating sheet according to the sensor signal and identification result, optimize resource use and improve energy efficiency; at the same time, the prevention and control state discrimination model outputs decisions based on multidimensional data science, supports long-term monitoring and remote management, and the advanced technical architecture (such as Transformer, attention mechanism, etc.) enhances the robustness and scalability of the system, and improves the reliability, sustainability and economy of the prevention and control as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0029] The application will be further described below with reference to the accompanying drawings and embodiments.

[0030] Fig. 1 is a structural schematic view of a Solenopsis invicta long-acting prevention and control device.

[0031] Fig. 2 is a side sectional view of a Solenopsis invicta long-acting prevention and control device.

[0032] Fig. 3 is a bottom view of a Solenopsis invicta long-acting prevention and control device.

[0033] Fig. 4 is a flowchart of a Solenopsis invicta long-acting prevention and control method.

[0034] REFERENCE NUMERALS: 100 - A long-acting red imported fire ant prevention and control device, 1 - A fixed support, 2 - A stainless steel barrel, 3 - A bait tray, 4 - A main controller, 5 - An antenna, 6 - A battery, 7 - A solar panel, 8 - A bait storage and dispensing mechanism, 9 - A camera, 81 - A funnel feeding part, 82 - A humidity sensor, 83 - A heating sheet, 84 - An upper cover plate, 85 - An electromagnetic valve, 86 - An outlet part, 87 - A PG joint, 821 - A humidity sensor fixing bracket, 831 - A heating sheet fixing bracket, 841 - A sealing ring. DETAILED DESCRIPTION

[0035] The technical solution in the embodiments of the present application has the following general idea: by integrating solar power supply and wireless communication, the long-acting red imported fire ant prevention and control device 100 can be deployed in the wild for a long time and receive remote instructions, thereby breaking through the weather and geographical restrictions and significantly improving environmental adaptability and pesticide application flexibility; at the same time, the long-acting red imported fire ant prevention and control device 100 ensures the dry storage of bait through a sealing structure and a heating and moisture-proof system, and uses the camera 9 and an AI model to identify target ant colonies and their foraging behaviors in real time, thereby accurately controlling the bait feeding time and dosage and realizing "on-demand dispensing", which fundamentally reduces the invalid exposure and residue of bait and effectively reduces the risk of pesticide residue and environmental pollution.

[0036] Please refer to Figs. 1 to 4 The preferred embodiment of the long-acting red imported fire ant prevention and control device 100 of the present application comprises: a fixed support 1; a stainless steel barrel 2 arranged at the top end of the fixed support 1; a bait tray 3 arranged at the middle part of the fixed support 1 and located directly below the stainless steel barrel 2; a main controller 4 arranged inside the stainless steel barrel 2; an antenna 5 arranged on the outer side wall of the stainless steel barrel 2 and connected with the main controller 4, with the receiving and transmitting direction upward; a battery 6 arranged inside the stainless steel barrel 2 and connected with the main controller 4; a solar panel 7 arranged obliquely at the top end of the stainless steel barrel 2 and connected with the battery 6; a bait storage and dispensing mechanism 8 arranged inside the stainless steel barrel 2 and connected with the main controller 4, with the feeding direction directly facing the bait tray 3; the bait storage and dispensing mechanism 8 is used for long-acting storage and dispensing of bait; a camera 9 arranged inside the stainless steel barrel 2 and connected with the main controller 4, with the shooting direction directly facing the bait tray 3.

[0037] The main controller 4 integrates a wireless communication module (not shown) and is connected with the antenna 5.

[0038] The shape of the solar panel 7 matches the shape of the stainless steel barrel body 2.

[0039] The bait storage and dispensing mechanism 8 comprises: A funnel feeding piece 81 is arranged in the interior of the stainless steel barrel body 2; A humidity sensor 82 is installed in the interior of the stainless steel barrel body 2 through a humidity sensor fixing bracket 821, is located above the funnel feeding piece 81, and is connected with the main controller 4; the humidity sensor 82 is used for monitoring the environmental humidity in the bait storage and dispensing mechanism 8; A heating sheet 83 is installed in the interior of the stainless steel barrel body 2 through a heating sheet fixing bracket 831, is located above the funnel feeding piece 81, and is connected with the main controller 4; the heating sheet 83 is used for heating the environmental temperature in the bait storage and dispensing mechanism 8; An upper cover plate 84 is installed at the top end of the stainless steel barrel body 2 through a sealing ring 841, and is located below the solar panel 7; An electromagnetic valve 85 is in communication with the discharge end of the funnel feeding piece 81 at the input end, and is connected with the main controller 4 at the control end; A discharge piece 86 is in communication with the output end of the electromagnetic valve 85 at the feeding end, and is opposite to the bait tray 3 at the discharge end.

[0040] The bait storage and dispensing mechanism 8 further comprises: Three PG joints 87 are respectively connected with the main controller 4 through one PG joint 87.

[0041] The shape of the upper cover plate 84 matches the shape of the stainless steel barrel body 2.

[0042] The preferred embodiment of the long-acting prevention and control method for the red imported fire ants comprises the following steps: Step S1, the long-acting prevention and control device for the red imported fire ants is deployed, bait is filled into the bait storage and dispensing mechanism, and the main controller is connected with the server through an antenna; in specific implementation, the long-acting prevention and control device for the fire ants is arranged in the red imported fire ant occurrence area or the monitoring area, a soft land is selected, and the place is required to be free of vegetation sheltering, so that the solar panel can receive sunlight, the fixing support is inserted into the soil, and the bait tray is tightly attached to the ground; at the same time, the long-acting prevention and control device for the fire ants is initialized, the power supply is connected, and it is ensured that each component operates normally; the network is configured, and it is ensured that the data transmission between the server is smooth; the filled bait does not exceed the humidity sensor; Step S2, the main controller controls the electromagnetic valve to dispense a preset amount of bait onto the bait tray based on the prevention and control instruction sent by the server; In step S3, the main controller collects the bait disc image in real time through the camera, inputs the bait disc image into the pre-deployed ant species identification model, foraging behavior identification model and drug efficacy identification model in real time, and obtains the ant species identification result, foraging behavior identification result and drug efficacy identification result; In step S4, the main controller dynamically opens and closes the electromagnetic valve based on the ant species identification result and the foraging behavior identification result, dynamically adjusts the sampling rate of the camera, records the opening and closing times of the electromagnetic valve, inputs the ant species identification result, foraging behavior identification result, drug efficacy identification result and opening and closing times into the pre-deployed prevention and control state discrimination model, obtains the prevention and control state discrimination result and uploads the server; Based on the locally deployed model, the main controller does not need to return the bait disc image data to the server for processing, which significantly improves the autonomous operation capability and unattended performance of the device in the field environment; In step S5, the main controller dynamically controls the working of the heating sheet based on the sensing signal of the humidity sensor.

[0043] When the environmental humidity exceeds the preset threshold, the main controller automatically activates the heating sheet to actively control the humidity of the internal microenvironment, effectively inhibits the accumulation of water, and ensures that the bait is in a dry and stable state. At the same time, the upper cover plate combines with the sealing ring to form a reliable waterproof barrier, significantly reducing the risk of external precipitation and condensed water intrusion. This synergistic effect significantly improves the environmental stability of the bait storage environment, maximizes the avoidance of bait palatability decline caused by moisture, and ensures the physicochemical properties of the bait during long-term storage and use.

[0044] In specific implementation, the prevention and control of the red imported fire ant is a repeated cycle process. The main controller automatically triggers the prevention and control process based on the prevention and control instructions sent by the server, continuously collects and identifies the bait disc image, dynamically adjusts the bait release speed and image acquisition frequency based on the identification result, dynamically controls the working of the heating sheet based on the humidity to ensure the quality of the bait, and feeds back the prevention and control state discrimination result. The server can dynamically adjust the prevention and control strategy based on the prevention and control state discrimination result, forming a closed loop.

[0045] In step S3, the ant species identification model is constructed based on an image feature extraction layer, a multi-scale feature fusion layer, an attention feature enhancement layer and a classification prediction layer. The image feature extraction layer is constructed based on a convolution feature extraction module, a Transformer context integration module and an attention gate module. The convolution feature extraction module is used to extract initial multi-scale feature maps from the bait disc image. The Transformer context integration module is used to integrate global spatial context information in the initial multi-scale feature maps. The attention gate module is used to filter important features from the initial multi-scale feature maps integrated with global spatial context information, and output multi-scale visual features. The multi-scale feature fusion layer is constructed based on a feature pyramid network; the feature pyramid network is used to fuse visual features of different scales through top-down and lateral connection to obtain fused features; The attention feature enhancement layer is constructed based on a spatial attention module and a channel attention module; the spatial attention module is used to enhance spatial important regions in the fused features; the channel attention module is used to weight important channel features in the fused features of the enhanced spatial important regions to obtain enhanced features; The classification prediction layer is constructed based on a fully connected classification module; the fully connected classification module is used to map the enhanced features to a probability distribution of the presence of fire ants, the presence of non-fire ants, and the absence of ants, and then output an ant species recognition result of the presence of fire ants, the presence of non-fire ants, or the absence of ants.

[0046] In the step S3, the foraging behavior recognition model is constructed based on an image preprocessing layer, a spatial feature extraction layer, a time modeling layer, a change enhancement layer, and a behavior prediction layer; The image preprocessing layer is used to perform standardization and enhancement operations on the input tray image to output a preprocessed image sequence; The spatial feature extraction layer is used to extract spatial features of each frame from the preprocessed image sequence to output a feature vector sequence of each frame; The time modeling layer is used to model the time dependence of the feature vector sequence to capture the time dynamics of the foraging behavior, and output time fusion features; The change enhancement layer is used to enhance the change information in the time fusion features to highlight the key patterns of bait reduction, and output change enhanced features; The behavior prediction layer is used to predict a behavior probability distribution based on the change enhanced features, and then output a foraging behavior recognition result of the absence of bait reduction, the reduction of bait, or the substantial disappearance of bait; The efficacy recognition model is constructed based on an image feature abstraction layer, an event detection layer, a relationship fusion layer, a feature enhancement layer, and an efficacy prediction layer; The image abstraction layer is used to extract tray spatial features from the input tray image through a 2D convolutional neural network, and extract tray temporal features from each of the tray spatial features through a ViViT network; The event detection layer is used to detect bait reduction events from the tray temporal features through a twin 3D convolutional network and a change point detection network, and detect fire ant death events from the tray temporal features through an inflated 3D convolutional network and a graph convolutional network; The relationship fusion layer is used for extracting event order relationships from the bait reduction events and the fire ant death events by a time sequence relationship network, dynamically weighting contributions of the bait reduction events and the fire ant death events by a cross-attention mechanism, fusing time sequence relationships by a gated recurrent unit, and obtaining relationship features; The feature enhancement layer enhances the relationship features by self-supervised contrast learning to obtain an enhanced feature vector; The pharmacodynamic prediction layer is used for mapping the enhanced feature vector to a hidden feature by a full connection layer, and outputting a binary classification probability of bait failure or bait non-failure by a Sigmoid activation function, and then obtaining a pharmacodynamic identification result of bait failure or bait non-failure.

[0047] In the step S4, the prevention and control state discrimination model is constructed based on a decision tree algorithm, and is used for outputting a prevention and control state discrimination result of continuing prevention and control, supplementing bait, replacing bait or completing prevention and control according to the ant species identification result, the foraging behavior identification result, the pharmacodynamic identification result and the switch number.

[0048] The core of the present application is to realize the synergistic optimization of long-acting preservation of bait, dynamic evaluation of palatability and real-time monitoring of prevention and control effect. By integrating humidity sensors and heating sheets in the bait storage and release mechanism, a microenvironment control module is constructed to effectively maintain the stability of the internal microenvironment, significantly improve the physicochemical stability of the bait and prolong its effective storage period. At the same time, the overall waterproof sealing structure design ensures the reliability of the bait performance in complex outdoor conditions. For palatability evaluation and prevention and control efficiency monitoring, the image recognition algorithm based on deep learning and the decision tree classification algorithm are fused to realize accurate perception and intelligent discrimination of ant species and bait foraging behavior. Further, the edge computing architecture is adopted, and all data processing and model inference tasks are completed on the local terminal, effectively reducing the dependence on the cloud, significantly improving the real-time performance, autonomy and adaptability of the system, and providing technical support for the precision and intelligent prevention and control of the red imported fire ant.

[0049] In summary, the present application has the following advantages: 1. By integrating solar power supply and wireless communication, the red imported fire ant long-acting prevention and control device can be deployed in the field for a long time and receive remote instructions, breaking through the weather and geographical limitations and significantly improving the environmental adaptability and flexibility of pesticide application; at the same time, the red imported fire ant long-acting prevention and control device ensures the dry storage of bait through the sealing structure, heating and moisture-proof system, and uses cameras and AI models to identify target ant colonies and their foraging behavior in real time, thereby accurately controlling the baiting time and dosage, realizing "on-demand release", and fundamentally reducing the invalid exposure and residue of bait, ultimately greatly improving the environmental adaptability and flexibility of pesticide application for red imported fire ant prevention and control, and reducing the risk of pesticide residue and environmental pollution.

[0050] 2. The device is powered by a combination of solar panels and batteries. The solar panels are inclined and placed at the top of the stainless steel cylinder. They are shaped to match the cylinder and can efficiently collect solar energy, ensuring the device can operate in the wild for a long time without relying on external power grids. This design makes the device suitable for remote areas or outdoor environments where fire ants are common, reducing the frequency of manual maintenance and improving the continuity and reliability of prevention and control. Meanwhile, the stainless steel cylinder and fixed support provide corrosion resistance and stability, enhancing the device's durability in harsh weather.

[0051] 3. By integrating cameras and main controllers, and combining pre-deployed AI models (such as ant species identification models, foraging behavior identification models, and drug efficacy identification models), the device can collect images of the feeding tray in real time, automatically identify the presence and behavior of fire ants, and automatically identify whether the bait is ineffective. Based on the identification results, the main controller dynamically adjusts the bait release (such as opening the electromagnetic valve) and the camera sampling rate, achieving precise response. This adaptive mechanism avoids over-releasing bait, improves control efficiency, and reduces the need for human intervention through data-driven decision-making, reflecting the technological progress of intelligentization.

[0052] 4. The main controller integrates a wireless communication module and connects to a server through an antenna, allowing remote reception of prevention and control instructions and uploading of data (such as prevention and control state judgment results). This allows users to monitor the status of multiple devices in real time, conduct centralized scheduling and data analysis, and greatly improve the convenience and scalability of prevention and control management. The wireless communication direction is designed upwards to optimize signal reception, ensuring stable communication in complex environments and conforming to the trend of the Internet of Things.

[0053] 5. The bait storage and release mechanism includes components such as electromagnetic valves, humidity sensors, and heating elements. It can dynamically control the operation of the heating elements based on environmental humidity to prevent the bait from getting wet and ensure its effectiveness. The main controller accurately controls the release amount based on prevention and control instructions and identification results, and the release direction is directly opposite the bait tray, reducing waste. This precise release not only saves bait costs but also reduces environmental impact, highlighting environmental protection and economic benefits.

[0054] 6. The device uses a modular structure, such as the bait storage and release mechanism connected by PG joints. The upper cover plate matches and seals the cylinder, making it easy to install and maintain. This design simplifies bait refilling, component replacement, and other operations, reducing long-term use costs. Meanwhile, modularity helps standardize production, improves the device's scalability and adaptability, and meets different scene requirements.

[0055] 7. The method combines multi-level AI models (such as feature pyramid networks and attention mechanisms) to achieve high-precision ant identification and behavior analysis, and outputs intelligent decisions (such as supplementing bait) through a prevention and control state discrimination model (based on decision trees); this method deeply integrates hardware and software to form a closed-loop prevention and control system, improving the scientificity and long-term effectiveness of the prevention and control of Solenopsis invicta, representing a technological innovation in the field of biological control.

[0056] 8. By integrating a solar power supply system, a stainless steel barrel body, and a fixed support, long-term energy supply and excellent environmental adaptability are achieved, ensuring the stable operation of the device in the wild for a long time; at the same time, with the help of an intelligent main controller, a camera, and a pre-deployed AI model (such as an ant species identification and foraging behavior identification model), real-time monitoring and adaptive control of Solenopsis invicta are achieved, which can dynamically adjust the bait release and sampling rate to improve the precision of prevention and control; in addition, the wireless communication module supports remote data interaction and centralized management, while the bait storage and release mechanism combined with a humidity sensor and a heating sheet ensures the effectiveness and resource optimization of the bait; modular design (such as PG joints and sealed upper covers) simplifies maintenance, and the overall method integrates hardware and advanced algorithms to form a closed-loop prevention and control system, significantly improving efficiency, reducing human intervention, and having environmental and economic benefits.

[0057] 9. By integrating a humidity sensor and a heating sheet, a stable microenvironment is constructed, effectively overcoming the problem of deterioration of the physicochemical properties and decreased palatability of traditional baits in high-humidity environments, significantly improving the stability and persistence of the bait, reducing pesticide waste and soil and water pollution, thereby reducing the secondary impact on non-target organisms and the ecological environment, and assisting in ecological safety and biodiversity protection.

[0058] 10. The use of waterproof sealing design and edge computing architecture enables all-weather reliable operation and real-time data processing, combined with deep learning and decision tree algorithms, to accurately identify Solenopsis invicta species and foraging behavior, enabling dynamic monitoring and intelligent evaluation of prevention and control effectiveness, significantly improving the scientificity and response speed of prevention and control operations. Especially in the southern rainy and humid climate conditions with a short prevention and control window, the prevention and control cycle can be effectively extended, significantly improving the success rate of control. In addition, the device has adaptive control capabilities, reducing the frequency of manual inspections and labor costs, reducing the complexity and uncertainty of prevention and control management, and promoting the intelligent, precise, and sustainable transformation of Solenopsis invicta prevention and control.

[0059] 11、Through highly automated and intelligent prevention and control processes, precise and efficient red imported fire ant management is achieved, including the use of a master controller working in conjunction with a server to automatically release baits and analyze image data in real time based on multiple models (such as ant species identification, foraging behavior identification, and drug efficacy identification models) to ensure identification accuracy; a dynamic adaptive control mechanism can adjust solenoid valves, camera sampling rates, and heating plates based on sensor signals and identification results to optimize resource use and improve energy efficiency; at the same time, a prevention and control state discrimination model outputs decisions based on multidimensional data science, supporting long-term monitoring and remote management, while advanced technical architectures (such as Transformers, attention mechanisms, etc.) enhance the robustness and scalability of the system, overall improving the reliability, sustainability, and economy of prevention and control.

[0060] Although the specific embodiments of the present application are described above, those skilled in the art should understand that the specific examples described are illustrative only and are not intended to limit the scope of the present application, and equivalent modifications and variations made in accordance with the spirit of the present application should be covered within the scope of the claims of the present application.

Claims

1. A long-term control device for red imported fire ants, characterized in that: include: A fixed bracket; A stainless steel cylinder is located at the top of the fixed bracket; A bait tray is located in the middle of the fixed bracket, directly below the stainless steel cylinder. A main controller is located inside the stainless steel cylinder. An antenna is installed on the outer wall of the stainless steel cylinder and connected to the main controller, with the transmitting and receiving direction facing upwards; A battery is located inside the stainless steel cylinder and connected to the main controller; A solar panel is tilted and mounted at the top of the stainless steel cylinder and connected to the battery; A bait storage and dispensing mechanism is located inside the stainless steel cylinder and connected to the main controller, with the dispensing direction facing the bait tray; A camera is located inside the stainless steel cylinder and connected to the main controller, with its shooting direction facing the bait tray.

2. The long-term control device for red imported fire ants as described in claim 1, characterized in that: The main controller integrates a wireless communication module and is connected to the antenna.

3. The long-term control device for red imported fire ants as described in claim 1, characterized in that: The shape of the solar panel matches the shape of the stainless steel cylinder.

4. The long-term control device for red imported fire ants as described in claim 1, characterized in that: The bait storage and dispensing mechanism includes: A funnel-shaped feed component is located inside the stainless steel cylinder. A humidity sensor is mounted inside the stainless steel cylinder via a humidity sensor mounting bracket, located above the funnel feeder, and connected to the main controller; A heating element is installed inside the stainless steel cylinder via a heating element mounting bracket, located above the funnel feeder, and connected to the main controller; A top cover plate is installed at the top of the stainless steel cylinder body via a sealing ring, located below the solar panel; A solenoid valve, with its input end connected to the discharge end of the funnel feeder and its control end connected to the main controller; A discharge component has its inlet end connected to the output end of the solenoid valve and its outlet end facing the feed tray.

5. The long-term control device for red imported fire ants as described in claim 4, characterized in that: The bait storage and dispensing mechanism also includes: The system has three PG connectors, with the humidity sensor, heating element, and solenoid valve each connected to the main controller via a PG connector.

6. The long-term control device for red imported fire ants as described in claim 4, characterized in that: The shape of the upper cover plate matches the shape of the stainless steel cylinder body.

7. A long-term control method for red imported fire ants, characterized in that: The method requires the use of the long-term red imported fire ant control device as described in any one of claims 1 to 6, and includes the following steps: Step S1: Deploy the long-term control device for red imported fire ants, add bait to the bait storage and delivery mechanism, and establish a connection between the main controller and the server via the antenna; Step S2: Based on the prevention and control instructions sent by the server, the main controller controls the solenoid valve to release a preset amount of bait onto the bait tray; Step S3: The main controller acquires images of the feeding tray in real time through the camera, and inputs the images of the feeding tray into the pre-deployed ant species identification model, feeding behavior identification model and drug efficacy identification model in real time to obtain ant species identification results, feeding behavior identification results and drug efficacy identification results; Step S4: The main controller dynamically switches the solenoid valve based on the ant species identification results and feeding behavior identification results, and dynamically adjusts the camera sampling rate. It records the number of times the solenoid valve is switched on and off. The ant species identification results, feeding behavior identification results, drug efficacy identification results, and the number of switching times are input into the pre-deployed prevention and control status discrimination model to obtain the prevention and control status discrimination results and upload them to the server. Step S5: The main controller dynamically controls the operation of the heating element based on the sensing signal from the humidity sensor.

8. The method for long-term control of red imported fire ants as described in claim 7, characterized in that: In step S3, the ant species identification model is constructed based on an image feature extraction layer, a multi-scale feature fusion layer, an attention feature enhancement layer, and a classification prediction layer. The image feature extraction layer is constructed based on a convolutional feature extraction module, a Transformer context integration module, and an attention gating module. The convolutional feature extraction module is used to extract an initial multi-scale feature map from the material tray image. The Transformer context integration module is used to integrate global spatial context information in the initial multi-scale feature map. The attention gating module is used to filter important features from the initial multi-scale feature map that integrates global spatial context information and output multi-scale visual features. The multi-scale feature fusion layer is constructed based on a feature pyramid network; The feature pyramid network is used to fuse visual features of different scales through top-down and lateral connections to obtain fused features; The attention feature enhancement layer is constructed based on a spatial attention module and a channel attention module; the spatial attention module is used to enhance spatially important regions in the fused features. The channel attention module is used to weight the important channel features in the fusion features of important regions in the enhanced space to obtain enhanced features; The classification prediction layer is built based on a fully connected classification module; The fully connected classification module is used to map the enhanced features to the probability distribution of the presence of red ants, the presence of non-red ants, and the absence of ants, and then outputs the ant species identification results of the presence of red ants, the presence of non-red ants, or the absence of ants.

9. A long-term control method for red imported fire ants as described in claim 7, characterized in that: In step S3, the feeding behavior recognition model is constructed based on an image preprocessing layer, a spatial feature extraction layer, a temporal modeling layer, a change enhancement layer, and a behavior prediction layer. The image preprocessing layer is used to standardize and enhance the input tray image and output a preprocessed image sequence. The spatial feature extraction layer is used to extract spatial features of each frame from the preprocessed image sequence and output a feature vector sequence of each frame. The time modeling layer is used to model the time dependency of the feature vector sequence, capture the temporal dynamics of feeding behavior, and output time fusion features. The variation enhancement layer is used to enhance the variation information in the temporal fusion features, highlight the key pattern of bait reduction, and output variation enhancement features; The behavior prediction layer is used to predict the probability distribution of behavior based on the change enhancement features, and then outputs the feeding behavior identification results as the bait not reduced, the bait reduced, or the bait basically disappeared. The drug efficacy recognition model is constructed based on an image feature abstraction layer, an event detection layer, a relationship fusion layer, a feature enhancement layer, and a drug efficacy prediction layer. The image abstraction layer is used to extract the spatial features of the material tray from the input material tray image through a 2D convolutional neural network, and to extract the temporal features of the material tray from each of the spatial features of the material tray through a ViViT network; The event detection layer is used to detect bait reduction events from the temporal features of the feed tray using a twin 3D convolutional network and a change point detection network, and to detect red imported fire ant death events from the temporal features of the feed tray using a dilated 3D convolutional network and a graph convolutional network. The relation fusion layer is used to extract the event sequence relationship from bait reduction events and red imported fire ant death events through a temporal relation network, dynamically weight the contribution of bait reduction events and red imported fire ant death events through a cross-attention mechanism, and fuse the temporal relationship through a gated recurrent unit to obtain relation features; The feature enhancement layer enhances relational features through self-supervised contrastive learning to obtain enhanced feature vectors; The efficacy prediction layer is used to map the enhanced feature vector to the hidden features through a fully connected layer, and outputs the binary classification probability of bait failure or bait non-failure through the Sigmoid activation function, thereby obtaining the efficacy identification result of bait failure or bait non-failure.

10. A long-term control method for red imported fire ants as described in claim 7, characterized in that: In step S4, the prevention and control status discrimination model is constructed based on the decision tree algorithm. It is used to output the prevention and control status discrimination result, which is to continue prevention and control, replenish bait, change bait, or complete prevention and control, based on the ant species identification result, feeding behavior identification result, drug efficacy identification result, and number of switching.

Images