Bird repelling device and method based on AI behavior recognition and dynamic parameters

Abstract

The invention relates to a bird repelling device and method based on AI behavior recognition and dynamic parameters. The device comprises a comprehensive cable, a processor, a camera, a holder, a repelling assembly and a foundation. The processor comprises a data acquisition module, a data preprocessing module, an AI identification module, a behavior decision module, a coordinate calculation module, an equipment control module, a state management module and a communication module; the camera has an image acquisition function; the holder comprises a holder base, a rotating table and a rotating arm; the expelling assembly comprises an electromagnetic expelling assembly and a stroboscopic emitter. The method comprises the steps of data acquisition and preprocessing; carrying out AI behavior recognition and behavior level generation; coordinate calculation and holder steering are carried out; strategy generation and parameter matching; the expelling operation is executed; and state recording and feedback. The YOLOv8 AI deep learning model is adopted to accurately identify bird behaviors, and the electromagnetic and stroboscopic collaborative non-contact repelling technology is combined, so that the repelling efficiency is improved, the accidental injury rate of birds is reduced, and the ecological friendliness is good.

Description

Technical Field

[0001] This invention belongs to the field of bird control technology, specifically relating to a bird control device and method based on AI behavior recognition and dynamic parameters, which is applicable to bird pest control in substations, power transmission lines and other scenarios. Background Technology

[0002] In numerous scenarios, including power transmission lines, airport runways, and agricultural planting areas, the safety hazards and economic losses caused by bird activity have become urgent and pressing issues. Specifically, short-circuit trips caused by bird nesting account for 15%–20% of all power line faults; bird strikes at airports cause over $1.3 billion in losses to the global aviation industry annually; and fruit loss in orchards due to bird pecking generally exceeds 30%. These figures clearly highlight the significant impact of bird activity on human production and life, and reflect the urgency of addressing bird-related hazards. However, traditional bird control methods face the following challenges in practical application: (1) Insufficient recognition accuracy: Most existing bird deterrence systems rely on a single sensor to work, which has obvious limitations and cannot accurately distinguish different bird behaviors, resulting in a high false alarm rate.

[0003] (2) Adaptive defects: Under fixed dispersal patterns such as sound waves and strobe, birds will gradually adapt to these fixed dispersal signals, resulting in reduced dispersal efficiency.

[0004] (3) Poor ecological compatibility: Although bird nets can block birds from entering specific areas to a certain extent, they cause serious harm to birds. More than 5 million birds die every year due to bird nets worldwide. Chemical bird repellents pose a risk of soil pollution. These measures have caused great damage to the survival of birds and the ecological environment.

[0005] Therefore, there is an urgent need for an intelligent bird control method that combines accurate behavior recognition capabilities, dynamic strategy adaptation mechanisms, and eco-friendly characteristics to make up for the shortcomings of traditional bird control methods. Summary of the Invention

[0006] This invention provides a bird deterrence device and method based on AI behavior recognition and dynamic parameters, aiming to effectively solve the problems existing in the current bird deterrence technology and provide a more reliable, efficient and environmentally friendly bird deterrence solution for fields such as power transmission, aviation safety and agricultural planting.

[0007] To achieve the above objectives, this invention provides a bird-repelling device based on AI behavior recognition and dynamic parameters, comprising a comprehensive cable, a processor, a camera, a gimbal, a bird-repelling component, and a foundation; the comprehensive cable includes a main cable, a first branch cable, a second branch cable, a third branch cable, and a fourth branch cable; the processor includes a data acquisition module, a data preprocessing module, an AI recognition module, a behavior decision module, a coordinate calculation module, a device control module, a status management module, and a communication module; the camera has image acquisition capabilities; the gimbal includes a gimbal base, a rotating platform, and a rotating arm; the bird-repelling component includes an electromagnetic bird-repelling component and a strobe emitter, wherein the electromagnetic bird-repelling component includes an electromagnetic array composed of several electromagnetic coils, hexagonal tubes, and a thermally conductive coating, as well as a filling layer, a protective shell, a front cover plate, and a rear cover plate, and the strobe emitter is composed of several LED strobe lights; the foundation is used to support the gimbal.

[0008] Furthermore, the integrated cable has power supply, data transmission and control functions. The main cable connects the processor and the junction box. The first branch cable connects the camera and the junction box. The second branch cable connects the gimbal base and the junction box. One end of the third branch cable is connected to the junction box, and the other end has multiple terminals, which are respectively connected to several electromagnetic coils and the fourth branch cable. The other end of the fourth branch cable is connected to the strobe transmitter.

[0009] Furthermore, the functional module connection relationship of the processor is as follows: The output of the data acquisition module is connected to the input of the data preprocessing module and is configured to transmit raw image data. The output of the data preprocessing module is connected to the input of the AI ​​recognition module and is configured to optimize image data. The AI ​​recognition module integrates the YOLOv8 AI deep learning model. Its first output is connected to the input of the coordinate calculation module and is configured to transmit the target pixel coordinates (u,v); its second output is connected to the input of the behavior decision module and is configured to transmit behavior level parameters (B0 / B1 / B2 / B3); and its third output is connected to the first input of the state management module and is configured to transmit historical data. The output of the behavior decision module is connected to the first input of the device control module and configured to transmit the drive-off level (L0 / L1 / L2 / L3). The output of the coordinate calculation module is connected to the second input of the device control module and is configured to transmit the real-time turning angle of the pan-tilt unit (horizontal θ / vertical φ). The first output terminal of the device control module is connected to the pan-tilt unit, the electromagnetic array, and the strobe transmitter, and is configured to transmit the real-time turning angle of the pan-tilt unit (horizontal θ / vertical φ), the current / voltage value of the electromagnetic array, and the strobe frequency execution parameters; the second output terminal is connected to the input terminal of the status management module and is configured to report the above execution parameters. The output of the status management module is connected to the input of the communication module and configured for transmission history and real-time recording. The communication module output is connected to an external management platform and configured to transmit history, record in real time, and generate alarms.

[0010] Furthermore, the camera has an image acquisition function, which transmits the acquired image to the processor's data acquisition module via the first branch cable and the main cable; the camera can be arranged independently of the de-energizing component, or it can be mounted on the de-energizing component.

[0011] Furthermore, the gimbal base is mounted on the upper end of the foundation, the rotating platform is mounted on the upper end of the gimbal base and is capable of 360° horizontal rotation, the rotating arms are mounted on both sides of the rotating platform, and the inner sides of the two rotating arms are hinged to the outer wall of the protective shell of the electromagnetic drive assembly, the rotating arms drive the protective shell to perform pitch rotation. Among them, the gimbal with horizontal rotation and pitch rotation is prepared by existing technology, such as the "buffered gimbal mechanism that can carry dual guns" disclosed in CN118856170A.

[0012] Furthermore, the electromagnetic drive-off assembly includes several electromagnetic coils with independent parallel circuits. A single set of electromagnetic coils is tightly attached to the inner wall of a single hexagonal tube in a left-hand spiral shape. The thermally conductive coating is applied to cover the inner side of the electromagnetic coil. Several electromagnetic coils, hexagonal tubes, and thermally conductive coatings are arranged in a hexagonal pattern to form an electromagnetic array. The filling layer wraps around the electromagnetic array and is then installed inside a hexagonal protective shell. The protective shell has a front cover plate and a rear cover plate at the front and rear ends, respectively. The rear cover plate has a small hole for the third branch cable to pass through.

[0013] Furthermore, the hexagonal tube is made of permalloy, the filling layer is made of polyurethane foam, and the protective shell, front cover, and rear cover are made of aluminum alloy.

[0014] Furthermore, the strobe emitter is mounted on the top of the protective shell via four hollow pillars, and the LED strobe light faces the same direction as the electromagnetic array emission direction; a wire hole is opened on the protective shell corresponding to any one of the pillars for the fourth branch cable to pass through.

[0015] To achieve the above objectives, this invention also provides a bird deterrence method based on AI behavior recognition and dynamic parameters. The method uses an AI recognition module configured in the processor to detect and calculate bird behavior characteristics, construct a behavior feature map, and generate B0, B1, B2, and B3 level behavior grades. The behavior decision module maps these grades to L0, L1, L2, and L3 deterrence levels and outputs execution parameters. The device control module dynamically matches the execution parameters to achieve hierarchical dynamic control. The deterrence execution parameters are dynamically adjusted according to the birds' adaptability. A non-contact electromagnetic / stroboscopic collaborative technology is used to replace physical barriers and chemical agents, solving the problems of high false alarm rates, easy failure, and ecological damage associated with traditional technologies. Specifically, the method includes the following steps: Step 1, Data Acquisition and Preprocessing: The data acquisition module acquires image data of the target area captured by the camera in real time and transmits it to the data preprocessing module for noise reduction, enhancement and distortion correction; Step 2, AI Behavior Recognition and Behavior Level Generation: The AI ​​recognition module detects bird targets in the preprocessed image data and outputs their target pixel coordinates (u,v) to the coordinate calculation module. At the same time, the behavior level parameters (B0 / B1 / B2 / B3) are output to the behavior decision module. Step 3, Coordinate Calculation and Gimbal Turning: The coordinate calculation module calculates the real-time turning angle of the gimbal (horizontal θ / vertical φ) based on the target pixel coordinates (u,v). The specific steps are as follows: (1) Normalize the image pixel coordinates to convert them into three-dimensional spatial coordinates. The equation is as follows:

[0016] In the formula: Horizontal and vertical focal lengths (pixel values). Principal point coordinates (pixel values), () represents the normalized coordinates in the camera coordinate system; (2) Calculate the real-time turning angle of the gimbal (horizontal θ / vertical φ), the equation is as follows:

[0017] In the formula: This is the initial pitch angle of the camera (201). It should be noted that the camera (201) is horizontal by default. =0°; Step 4, Strategy Generation and Parameter Matching: The behavior decision module generates the corresponding expulsion level (L0 / L1 / L2 / L3) based on the behavior level mapping and matches the corresponding expulsion strategy; the behavior decision model and coordinate calculation module then output the expulsion strategy and the real-time turning angle of the PTZ to the device control module respectively. Step 5: Perform the drive-away operation: The equipment control module outputs real-time turning angle (horizontal θ / vertical φ), electromagnetic array current / voltage value, and strobe frequency execution parameters to the pan-tilt unit, electromagnetic array, and strobe transmitter (502) to perform the drive-away operation; if the drive-away is successful, the pan-tilt unit is reset and step 6 is executed; if the drive-away is unsuccessful, return to step 1 and repeat the execution. Step 6, Status Recording and Feedback: The status management module records the execution parameters reported by the device control module and uploads the execution parameters to the external management platform through the communication module.

[0018] Furthermore, the mapping relationship between behavior level and behavior in step 2 is as follows: Level B0 (No Risk): Birds do not carry nesting materials, stay for less than 5 minutes, do not fly back and forth, and do not gather in groups, or have any of the above characteristics. Level B1 (Low Risk): There is nesting material and it is >10cm; 10 minutes > stay time ≥5 minutes, 5 times > round trip ≥3 times, 5 birds > number of birds ≥2, and one of the above conditions must be met; Level B2 (Medium Risk): Carrying multiple tree branches, totaling more than 3 branches; stay time greater than 10 minutes in 20 minutes, 7 times greater than 5 round trips, 7 birds greater than 5 birds, having one of the above characteristics; Level B3 (High Risk): Carrying high-risk nesting materials (metal parts), staying for more than 20 minutes within 30 minutes, making more than 5 round trips, or having more than 7 birds; any of the above conditions must be met. The mapping relationship between behavior level and expulsion level in step 4 is as follows: L0 (Security): Maps to B0 level behavior; L1 (Weak Removal): Maps to B1 level behavior level; L2 (Forced Expulsion): Maps to B2 level behavior; L3 (Forced removal ineffective): Maps to B3 behavior level; The mapping relationship between expulsion levels and expulsion strategies in step 4 is as follows: L0 (Safe): Records bird location and duration, then resets the gimbal and does not drive the bird away; L1 (Weak Drive-Off): The gimbal turns to aim at the target area, activates a 50Hz low-frequency magnetic field (intensity ≤2mT) + 5Hz strobe light, and lasts for 30 seconds.

[0019] L2 (Strong Removal): Activate a 10Hz pulsed magnetic field (intensity 8-10mT) + 10Hz strobe light for 20 seconds; L3 (Forced removal ineffective): No further removal, record bird location and duration, reset gimbal, and send an alarm to the management platform.

[0020] The beneficial effects of this invention are as follows: The bird-repelling device of this invention integrates multiple modules, including data acquisition, preprocessing, AI recognition, behavior decision-making, coordinate calculation, device control, status management, and communication, through a processor, achieving full automation from data acquisition to bird removal execution. Simultaneously, the camera, gimbal, electromagnetic repelling component, and strobe transmitter in the device work collaboratively to form a complete bird-repelling system. The complementary functions of each component enhance the bird-repelling effect. Specifically, the YOLOv8 AI deep learning model, which integrates AI recognition, can accurately identify bird behavior, reduce the false positive rate, and improve recognition accuracy. The non-contact repelling technology, which combines electromagnetic and strobe signals, improves repelling efficiency while avoiding physical harm to birds, significantly reducing the rate of accidental bird injury and exhibiting excellent eco-friendly characteristics.

[0021] The bird deterrence method of this invention constructs a behavioral feature map through an AI recognition module, which can accurately identify various behavioral states of birds, such as whether they carry nesting materials, their dwell time, the number of round trips, and the number of birds gathering. It then maps corresponding deterrence strategies to their behavioral levels, ensuring the targeted and effective deterrence measures and avoiding over-determination and ineffective deterrence. Furthermore, a secondary behavioral level detection is performed after each deterrence, and the deterrence strategy is adjusted promptly based on the detection results to ensure successful deterrence and prevent the deterrence effect from decreasing due to bird adaptation. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall structure of the bird-repelling device of the present invention; Figure 2 This is a schematic diagram showing the disassembled structure of the gimbal and the deflection component of the present invention; Figure 3 This is a schematic diagram of the disassembled structure of the deflection component of the present invention; Figure 4 The images show a cross-sectional view (left) of the electromagnetic drive-away assembly of the present invention and a cross-sectional view (right) of a single electromagnetic array. Figure 5 This is a schematic diagram of the installation of the strobe transmitter of the present invention; Figure 6 This is a schematic diagram of the wiring of the integrated cable of the present invention; Figure 7 This is a data flow diagram between the modules of the processor of this invention; Figure 8 This is a flowchart of the bird-repelling method of the present invention; In the diagram: 1-Main cable; 2-First branch cable; 3-Second branch cable; 4-Third branch cable; 5-Fourth branch cable; 6-Foundation; 101- Junction box; 102- Processor; 201- Camera; 301- Gimbal base; 302- Rotary table; 303- Rotating arm; 401- Electromagnetic coil; 402- Hexagonal tube; 403- Thermally conductive coating; 404- Filler layer; 405- Protective shell; 406- Front cover; 407- Rear cover; 408- Wiring hole; 501- Support column; 502- Strobe emitter; 503- LED strobe light. Detailed Implementation

[0023] To make the technical problems and solutions solved by the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention. Example 1

[0024] like Figure 1 As shown, this embodiment provides a bird deterrent device based on AI behavior recognition and dynamic parameters, including a composite cable, a processor 102, a camera 201, a gimbal, a bird deterrent component, and a base 6. The integrated cable includes a main cable 1, a first branch cable 2, a second branch cable 3, a third branch cable 4, and a fourth branch cable 5; the processor 102 includes a data acquisition module, a data preprocessing module, an AI recognition module, a behavior decision module, a coordinate calculation module, a device control module, a status management module, and a communication module; the camera 201 has image acquisition capabilities; the gimbal includes a gimbal base 301, a rotating platform 302, and a rotating arm 303; the decoupling assembly includes an electromagnetic decoupling assembly and a strobe emitter 502, wherein the electromagnetic decoupling assembly includes an electromagnetic array composed of several electromagnetic coils 401, hexagonal tubes 402, and a thermally conductive coating 403, as well as a filling layer 404, a protective shell 405, a front cover plate 406, and a rear cover plate 407, and the strobe emitter 502 is composed of several LED strobe lights 503; the foundation 6 is used to load the gimbal and can be a specially constructed independent foundation or existing components of an existing building.

[0025] Preferred, combined Figure 6 As shown, the integrated cable has power supply, data transmission and control functions. The main cable 1 connects the processor 102 and the junction box 101. The first branch cable 2 connects the camera 201 and the junction box 101. The second branch cable 3 connects the pan-tilt base 301 and the junction box 101. One end of the third branch cable 4 is connected to the junction box 101, and the other end has multiple terminals, which are respectively connected to several electromagnetic coils 401 and the fourth branch cable 5. The other end of the fourth branch cable 5 is connected to the strobe transmitter 502.

[0026] Preferred, such as Figure 7 As shown, the connection relationship of the functional modules of the processor 102 is as follows: The output of the data acquisition module is connected to the input of the data preprocessing module and configured to transmit raw image data; The output of the data preprocessing module is connected to the input of the AI ​​recognition module and configured to optimize image data; The AI ​​recognition module integrates the YOLOv8 AI deep learning model. Its first output is connected to the input of the coordinate calculation module and is configured to transmit the target pixel coordinates (u,v); its second output is connected to the input of the behavior decision module and is configured to transmit behavior level parameters; and its third output is connected to the first input of the state management module and is configured to transmit historical data. The output of the behavior decision module is connected to the first input of the device control module and configured to transmit the drive-off level. The output of the coordinate calculation module is connected to the second input of the device control module and configured to transmit the real-time turning angle of the pan-tilt unit. The first output of the device control module is connected to the pan-tilt unit, electromagnetic array, and strobe transmitter (502), and is configured to transmit the real-time turning angle of the pan-tilt unit, the current / voltage value of the electromagnetic array, and the strobe frequency execution parameters; the second output is connected to the input of the status management module and is configured to report the above execution parameters. The output of the status management module is connected to the input of the communication module and configured for transmission history and real-time recording. The communication module output is connected to an external management platform and configured to transmit history, record in real time, and generate alarms.

[0027] Preferably, the camera 201 has an image acquisition function, which transmits the acquired image to the data acquisition module of the processor 102 via the first branch cable 2 and the main cable 1; the camera 201 can be arranged independently of the de-energizing component, or it can be mounted on the de-energizing component.

[0028] Preferred, such as Figure 2 As shown, the gimbal includes a gimbal base 301, a rotating platform 302, and rotating arms 303. The gimbal base 301 is mounted on the upper end of the base 6, the rotating platform 302 is mounted on the upper end of the gimbal base 301 and can rotate horizontally 360°, and the rotating arms 303 are mounted on both sides of the rotating platform 302. The inner sides of the two rotating arms 303 are hinged to the outer wall of the protective shell 405 of the electromagnetic drive assembly, and the rotating arms 303 drive the protective shell 405 to perform pitch rotation. Among them, the gimbal with horizontal and pitch rotation is prepared by existing technology, such as the "buffered gimbal mechanism that can carry dual guns" disclosed in CN118856170A.

[0029] Preferred, such as Figure 2-4As shown, the decoy assembly includes an electromagnetic decoy assembly and a strobe transmitter 502. The electromagnetic decoy assembly includes several parallel independent circuit electromagnetic coils 401. Each electromagnetic coil 401 is spirally attached to the inner wall of a single hexagonal tube 402. A thermally conductive coating 403 is applied to the inner side of the electromagnetic coils 401. Several electromagnetic coils 401, hexagonal tubes 402, and thermally conductive coating 403 are arranged in a hexagonal pattern to form an electromagnetic array. A filling layer 404 encloses the electromagnetic array and is then installed inside a hexagonal protective shell 405. The protective shell 405 has a front cover plate 406 and a rear cover plate 407 at its front and rear ends, respectively, to protect and fix the electromagnetic array and the filling layer 404. The rear cover plate 407 has a small hole for the third branch cable 4 to pass through.

[0030] More preferably, the hexagonal tube 402 is made of permalloy, which has high magnetic permeability to enhance the magnetic field effect and shield external magnetic field interference; the filling layer 404 is made of polyurethane foam with insulating, shock-absorbing and air-hardening properties; and the protective shell 405, the front cover plate 406 and the rear cover plate 407 are made of aluminum alloy with good electromagnetic shielding and heat dissipation functions.

[0031] Preferred, such as Figure 3 Figure 5 As shown, the strobe emitter 502 is mounted on the top of the protective shell 405 via four hollow pillars 501, and the LED strobe light 503 faces the same direction as the electromagnetic array emission direction; any pillar 501 has a wire hole 408 on the protective shell 405 for passing through the fourth branch cable 5.

[0032] Working principle: The camera 201 acquires image data of the target area in real time and transmits the image data to the data acquisition module of the processor 102 via the first branch cable 2 and the main cable 1. The data acquisition module then transmits the raw image data to the data preprocessing module for noise reduction, enhancement and distortion correction.

[0033] The AI ​​recognition module detects bird targets in the preprocessed image data and outputs their target pixel coordinates (u,v) to the coordinate calculation module. Simultaneously, it outputs behavior level parameters (B0 / B1 / B2 / B3) to the behavior decision module. The behavior decision module then generates the corresponding dispersal level (L0 / L1 / L2 / L3) based on the behavior level mapping, while the coordinate calculation module calculates the real-time turning angle (horizontal θ / vertical φ) of the gimbal based on the target pixel coordinates (u,v).

[0034] Subsequently, the behavioral decision-making model and coordinate calculation module output the decoupling strategy and the real-time turning angle of the PTZ to the equipment control module, respectively. The equipment control module then outputs the real-time turning angle (horizontal θ / vertical φ), electromagnetic array current / voltage value, and strobe frequency execution parameters to the PTZ, electromagnetic array, and strobe transmitter 502. The PTZ's rotary table 302 and rotary arm 303 rotate horizontally and vertically according to the horizontal turning angle θ and vertical turning angle φ, respectively, so that the electromagnetic decoupling component is aligned with the target area. The electromagnetic array and strobe transmitter 502 are then activated to execute the corresponding level of decoupling mode to carry out the decoupling operation.

[0035] During the expulsion process, the status management module synchronously records the execution parameters reported by the device control module and uploads the execution parameters to the external management platform through the communication module for subsequent expulsion strategy optimization. Example 2

[0036] This embodiment takes a 35kV substation transmission line as an example to provide a bird-repelling method based on AI behavior recognition and dynamic parameters, such as... Figure 8 As shown, it includes the following steps: Step 1: Data Acquisition and Preprocessing Camera 201 captures images of the 35kV substation outgoing line structure and surrounding area in real time, and transmits them to the data acquisition module of processor 102 via the first branch cable 2 and the main cable 1. The data acquisition module then transmits the raw image data to the data preprocessing module for noise reduction, enhancement, and distortion correction, eliminating interference from the metal reflection of the transmission tower and improving the contrast between birds and nesting materials.

[0037] Step 2: AI Behavior Recognition and Behavior Level Generation: The AI ​​recognition module, using its built-in YOLOv8 AI deep learning model, identified a magpie in the pre-processed image data. The magpie's beak is carrying a 15cm twig (exceeding the B1 threshold of 10cm). The bounding box pixel coordinates are (…). =800, =600). Based on the magpie's location and behavioral characteristics, its behavior level is mapped to B1. The pixel coordinates and behavior level B1 are then transmitted to the coordinate calculation module and the behavior decision module, respectively.

[0038] Step 3, Coordinate Calculation and Gimbal Rotation: The coordinate calculation module calculates the coordinates based on the pixel coordinates ( =800, =600) and camera 201 calibration parameters (horizontal focal length) =1000 pixels, vertical focal length 800 pixels, principal point coordinates ( =500, =400), obtained through the normalization transformation equation:

[0039] The calculated real-time turning angle of the gimbal is:

[0040] Step 4: Generation and parameter matching of expulsion strategies: The behavior decision module maps the B1 level behavior level to the L1 level drive-away level, with the corresponding drive-away strategy being "the gimbal turns to aim at the target area, activates a 50Hz low-frequency magnetic field (intensity ≤2mT) + a 5Hz strobe light, lasting for 30 seconds"; the behavior decision model and coordinate calculation module outputs execution parameters to the device control module: 50Hz low-frequency magnetic field (intensity 1.8mT) + 5Hz strobe light, lasting for 30 seconds, gimbal turning angle =16.7° =-14.0°.

[0041] Step 5: Perform L1 level removal: The equipment control module controls the pan-tilt unit's movements: the rotary table 302 rotates horizontally by 16.7°, and the rotating arm 303 rotates in pitch by -14.0°, thereby driving the electromagnetic array and strobe transmitter 502 to align with the target area. Activating the electromagnetic array and strobe transmitter 502 involves: the third branch cable 4 supplying 50Hz AC power to several electromagnetic coils 401; the permalloy hexagonal tube 402 enhancing the magnetic field concentration, forming a 1.8mT low-frequency alternating magnetic field in the target area; and the fourth branch cable 5 driving the LED strobe light 503 to flash at a frequency of 5Hz.

[0042] Step 6, Status Recording and Feedback: The status management module records the L1 level execution parameters: magnetic field frequency 50Hz, current 0.5A, flicker frequency 5Hz, and gimbal angle (…). =16.7° =-14.0°), and upload the execution parameters to the external management platform through the communication module.

[0043] Step 7, Secondary Detection and Behavior Level Upgrade: Thirty seconds after the L1 level bird was driven away, camera 201 captured images again. The AI ​​recognition module detected that the bird was still in the target area and that the number of round trips had increased to 5, thus upgrading the behavior level to B2.

[0044] Step 8: Perform Level 2 removal: The behavior decision module maps the B2 level behavior level to the L2 level drive-away level. The device control module switches the drive-away strategy and starts a 10Hz pulsed magnetic field (intensity 9mT) + 10Hz high-frequency flashing for 20 seconds.

[0045] Step 9: Successful removal and gimbal reset: Within 20 seconds, the birds, disturbed by the combined effects of the magnetic field and strobe light, flew away from the target area. The AI ​​recognition module failed to detect the target for 3 consecutive seconds, classifying it as a B0 behavior level. The behavior decision module mapped this to an L0 clearance level. The device control module sent a gimbal reset command: gimbal 302 returned to zero (θ=0°), rotating arm 303 reset to horizontal (φ=0°), electromagnetic coil 401 stopped supplying power, and the strobe light turned off. The status management module recorded the complete clearance process: B1→L1 (ineffective after 30 seconds)→B2→L2 (successful after 20 seconds). This accumulated data was used for subsequent strategy optimization.

[0046] The present invention has been described in detail above through specific and preferred embodiments. However, those skilled in the art should understand that the present invention is not limited to the embodiments described above. Any modifications, equivalent substitutions, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A bird-repelling device based on AI behavior recognition and dynamic parameters, characterized in that, The system includes a composite cable, a processor (102), a camera (201), a gimbal, a decoy assembly, and a base (6); the composite cable includes a main cable (1), a first branch cable (2), a second branch cable (3), a third branch cable (4), and a fourth branch cable (5); the processor (102) includes a data acquisition module, a data preprocessing module, an AI recognition module, a behavior decision module, a coordinate calculation module, a device control module, a status management module, and a communication module; the camera (201) has image acquisition capabilities; the gimbal includes... The gimbal base (301), the rotating platform (302), and the rotating arm (303) are included. The decoupling assembly includes an electromagnetic decoupling assembly and a strobe emitter (502). The electromagnetic decoupling assembly includes an electromagnetic array consisting of several electromagnetic coils (401), hexagonal tubes (402), and a thermally conductive coating (403), as well as a filling layer (404), a protective shell (405), a front cover plate (406), and a rear cover plate (407). The strobe emitter (502) consists of several LED strobe lights (503). The base (6) is used to load the gimbal.

2. The bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 1, characterized in that: The integrated cable has power supply, data transmission and control functions. The main cable (1) connects the processor (102) and the junction box (101). The first branch cable (2) connects the camera (201) and the junction box (101). The second branch cable (3) connects the pan-tilt base (301) and the junction box (101). One end of the third branch cable (4) is connected to the junction box (101), and the other end has multiple terminals, which are connected to several electromagnetic coils (401) and the fourth branch cable (5) respectively. The other end of the fourth branch cable (5) is connected to the strobe transmitter (502).

3. The bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 1, characterized in that: The functional module connection relationship of the processor (102) is as follows: The output of the data acquisition module is connected to the input of the data preprocessing module and is configured to transmit raw image data. The output of the data preprocessing module is connected to the input of the AI ​​recognition module and is configured to optimize image data. The AI ​​recognition module integrates the YOLOv8 AI deep learning model. Its first output is connected to the input of the coordinate calculation module and is configured to transmit the target pixel coordinates (u,v); its second output is connected to the input of the behavior decision module and is configured to transmit behavior level parameters; and its third output is connected to the first input of the state management module and is configured to transmit historical data. The output of the behavior decision module is connected to the first input of the device control module and configured to transmit the drive-away level. The output of the coordinate calculation module is connected to the second input of the device control module and is configured to transmit the real-time turning angle of the pan-tilt unit. The first output terminal of the device control module is connected to the pan-tilt unit, the electromagnetic array, and the strobe transmitter (502), and is configured to transmit the real-time turning angle of the pan-tilt unit, the current / voltage value of the electromagnetic array, and the strobe frequency execution parameters; the second output terminal is connected to the input terminal of the status management module and is configured to report the above execution parameters. The output of the status management module is connected to the input of the communication module and configured for transmission history and real-time recording. The communication module output is connected to an external management platform and configured to transmit history, record in real time, and generate alarms.

4. The bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 1, characterized in that: The camera (201) has an image acquisition function, and transmits the acquired image to the data acquisition module of the processor (102) through the first branch cable (2) and the main cable (1); the camera (201) can be arranged independently of the de-ejection component, or it can be mounted on the de-ejection component.

5. A bird-repelling device based on AI behavior recognition and dynamic parameters according to claim 1, characterized in that: The gimbal base (301) is installed on the upper end of the base (6), the rotating platform (302) is installed on the upper end of the gimbal base (301) and can rotate horizontally 360°, the rotating arms (303) are installed on both sides of the rotating platform (302), the inner sides of the two rotating arms (303) are hinged to the outer wall of the protective shell (405) of the electromagnetic drive assembly, and the rotating arms (303) drive the protective shell (405) to pitch and rotate.

6. The bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 1, characterized in that: The electromagnetic drive-away assembly includes several electromagnetic coils (401) connected in parallel with independent circuits. A single set of electromagnetic coils (401) is tightly attached to the inner wall of a single hexagonal tube (402) in a left-hand spiral shape. The thermally conductive coating (403) is applied to cover the inner side of the electromagnetic coils (401). Several electromagnetic coils (401), hexagonal tubes (402) and thermally conductive coating (403) are arranged in a hexagonal shape to form an electromagnetic array. The filling layer (404) wraps the electromagnetic array and is installed in a hexagonal protective shell (405). The protective shell (405) has a front cover plate (406) and a rear cover plate (407) at the front and rear ends, respectively. The rear cover plate (407) has a small hole for the third branch cable (4) to pass through.

7. A bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 6, characterized in that: The hexagonal tube (402) is made of permalloy, the filling layer (404) is made of polyurethane foam, and the protective shell (405), front cover plate (406) and rear cover plate (407) are made of aluminum alloy.

8. A bird deterrent device based on AI behavior recognition and dynamic parameters according to claim 6, characterized in that: The strobe emitter (502) is mounted on the top of the protective shell (405) via four hollow pillars (501), and the LED strobe light (503) faces the same direction as the electromagnetic array emission direction; a wire hole (408) is opened on the protective shell (405) corresponding to any pillar (501) for passing through the fourth branch cable (5).

9. A bird-repelling method based on AI behavior recognition and dynamic parameters, characterized in that, Includes the following steps: Step 1, Data Acquisition and Preprocessing: The data acquisition module acquires the target area image data captured by the camera (201) in real time and transmits it to the data preprocessing module for noise reduction, enhancement and distortion correction; Step 2, AI Behavior Recognition and Behavior Level Generation: The AI ​​recognition module detects bird targets in the preprocessed image data and outputs their target pixel coordinates (u,v) to the coordinate calculation module. At the same time, the behavior level parameters are output to the behavior decision module. Step 3, Coordinate Calculation and Gimbal Turning: The coordinate calculation module calculates the real-time turning angle of the gimbal based on the target pixel coordinates (u,v). The specific steps are as follows: (1) Normalize the image pixel coordinates to convert them into three-dimensional spatial coordinates. The equation is as follows: ； In the formula: For horizontal and vertical focal lengths, Principal point coordinates, ( () represents the normalized coordinates in the camera coordinate system; (2) Calculate the real-time turning angle of the gimbal (horizontal θ / vertical φ), the equation is as follows: ； In the formula: Let the initial pitch angle of the camera (201) be ; Step 4, Strategy Generation and Parameter Matching: The behavior decision module generates the corresponding expulsion level based on the behavior level mapping and matches the corresponding expulsion strategy; the behavior decision model and coordinate calculation module then output the expulsion strategy and the real-time turning angle of the PTZ to the device control module respectively. Step 5: Perform the driving-away operation: The equipment control module outputs real-time turning angle, electromagnetic array current / voltage value, and strobe frequency execution parameters to the pan-tilt unit, electromagnetic array, and strobe transmitter (502) to perform the driving-away operation. If the expulsion is successful, the PTZ will reset and proceed to step 6; if the expulsion is unsuccessful, return to step 1 and repeat the process. Step 6, Status Recording and Feedback: The status management module records the execution parameters reported by the device control module and uploads the execution parameters to the external management platform through the communication module.

10. A bird-repelling method based on AI behavior recognition and dynamic parameters according to claim 9, characterized in that, The mapping relationship between behavior level and behavior in step 2 is as follows: Level B0 (No Risk): Birds do not carry nesting materials, stay for less than 5 minutes, do not fly back and forth, and do not gather in groups, or have any of the above characteristics. Level B1 (Low Risk): There is nesting material and it is >10cm; 10 minutes > stay time ≥5 minutes, 5 times > round trip ≥3 times, 5 birds > number of birds ≥2, and one of the above conditions must be met; Level B2 (Medium Risk): Carrying multiple tree branches, totaling more than 3 branches; stay time greater than 10 minutes in 20 minutes, 7 times greater than 5 round trips, 7 birds greater than 5 birds, having one of the above characteristics; Level B3 (High Risk): Carrying high-risk nesting materials, staying for more than 20 minutes within 30 minutes, making more than 5 round trips, or having more than 7 birds; any of the above conditions must be met. The mapping relationship between behavior level and expulsion level in step 4 is as follows: L0 (Security): Maps to B0 level behavior; L1 (Weak Removal): Maps to B1 level behavior level; L2 (Forced Expulsion): Maps to B2 level behavior; L3 (Forced removal ineffective): Maps to B3 level behavior; The mapping relationship between expulsion levels and expulsion strategies in step 4 is as follows: L0 (Safe): Records bird location and duration, then resets the gimbal and does not drive the bird away; L1 (weak drive-away): The gimbal turns to aim at the target area and activates a 50Hz low-frequency magnetic field + 5Hz strobe light for 30 seconds. L2 (Strong Disengagement): Activates a 10Hz pulsed magnetic field and a 10Hz strobe light for 20 seconds; L3 (Forced removal ineffective): No further removal, record bird location and duration, reset gimbal, and send an alarm to the management platform.

Images