Bird repelling and unmanned aerial vehicle dispatching integrated platform based on big data analysis

Abstract

The invention relates to the technical field of intelligent control and unmanned aerial vehicle application, in particular to a bird repelling and unmanned aerial vehicle dispatching integrated platform based on big data analysis. Comprising a multi-source data acquisition module; a dynamic scheduling decision module; and a federal learning optimization module. According to the invention, the millimeter-wave radar, the infrared camera, the voiceprint recognition unit and the meteorological data interface are integrated through the multi-source data acquisition module, so that the defects of one-sided data and extensive evaluation of a traditional scheme are overcome; a bird threat level, unmanned aerial vehicle endurance time, an electric quantity attenuation model and an airspace control rule are fused through a dynamic scheduling decision module, dynamic balance of priority response of a high-risk area and unmanned aerial vehicle endurance and energy consumption is realized, and the problems of response lag and insufficient endurance caused by unreasonable task allocation in a traditional scheme are avoided; the federal learning optimization module adapts to a specific geographical environment through transfer learning, and the problem of strategy stiffness of a traditional scheme in a complex environment is solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control and drone application technology, and in particular to an integrated bird-repelling and drone scheduling platform based on big data analysis. Background Art

[0002] With the growing demand for bird activity monitoring and repelling in sectors such as aviation, electricity, and agriculture, drone-based intelligent bird repellent technology has become a research hotspot due to its flexibility and efficiency. Traditional bird repellent methods rely on manual patrols or fixed devices, which suffer from delayed response times and limited coverage. The combination of drone swarm technology and big data analytics offers a new path to precise and intelligent bird repellent.

[0003] In the prior art, Chinese patent CN202210852281.8 discloses a "drone-based bird repellent delivery device and its control method", which includes a tube body for placing a bird repellent box, one end of the tube body is connected to the drone, and a delivery assembly is provided at the other end of the tube body, and a guide device is provided on the side of the tube body; the delivery assembly includes relative trigger legs, each trigger leg includes a support rod and an inclined leg connected to each other, the upper end of the inclined leg is provided with an inward protrusion, and the distance from the bottom end of the inclined leg to the bottom end of the tube body is less than the height of a bird repellent box. Although this solution improves the efficiency of bird repellent box delivery, it has the following significant defects: the device only focuses on the drone delivery action itself, and does not involve key data such as bird activity trajectory (such as flight altitude, cluster density), environmental parameters (such as wind speed, precipitation) and drone status (such as power, equipment health), resulting in bird repellent demand assessment completely relying on manually preset scenarios, and unable to dynamically adjust the delivery strategy according to real-time data; at the same time, it lacks environmental adaptability, does not consider the impact of complex meteorological conditions (such as strong winds, heavy rains) on drone flight safety and the effectiveness of bird repellents, and does not design differentiated repellent solutions based on the behavioral characteristics of different birds (such as raptors and resident birds). Another patent, CN202311390280.7, proposes a "UAV agricultural bird-repelling method and system based on a topological sorting reward mechanism", which includes building a bird identifier, collecting three-dimensional target data of birds, and performing three-dimensional target tracking; automatically generating a reward automaton based on the recognition results, and building a connection mechanism between the reward automaton and the underlying algorithm, and using reinforcement learning methods for path planning; the device uses reinforcement learning to achieve path planning to improve task parallelism, but the solution still has technical bottlenecks: its reinforcement learning reward mechanism only focuses on task parallelism and path coverage efficiency, ignoring the drone's endurance (such as the impact of battery degradation rate on task allocation), airspace control rules (such as no-fly zone weights) and equipment health (such as sensor failure history), which may cause the drone to interrupt operations due to power exhaustion or equipment failure during high-load tasks.

[0004] Although the above existing technologies all have their advantages, they still have the following common technical defects in actual use: 1. The data collection dimension is single, and it is impossible to build a multi-source heterogeneous data model that includes bird activities, environmental conditions, and equipment parameters, making it difficult to accurately assess bird repellent needs; 2. The scheduling strategy is extensive and does not integrate threat levels, energy consumption status, and equipment health, resulting in delayed responses in high-risk areas or insufficient drone endurance; 3. The environmental adaptability is weak, lacking ground-air coordination mechanisms under extreme weather conditions and differentiated repellent strategies for multiple bird species. To address the above problems, this application proposes an integrated bird repellent and drone scheduling platform based on big data analysis. Summary of the Invention

[0005] The purpose of the present invention is to provide an integrated bird-repelling and drone scheduling platform based on big data analysis to solve the problems raised in the above background technology.

[0006] To solve the above technical problems, the present invention aims to provide an integrated bird-repelling and drone dispatching platform based on big data analysis, including: a multi-source data acquisition module for integrating meteorological data, geographic information data, historical bird activity data, and real-time monitoring data, the real-time monitoring data including bird activity tracks monitored by radar and population classification information identified by voiceprints;

[0007] The data fusion processing module uses a spatiotemporal feature extraction algorithm to process heterogeneous data sources and generate a spatiotemporal-correlated bird density distribution heat map;

[0008] The dynamic scheduling decision module generates a multi-UAV collaborative path planning solution based on the heat map and UAV status parameters through an improved ant colony algorithm;

[0009] The collaborative control module controls the drone to perform sonic, light, or physical repelling tasks according to the path planning scheme, and adjusts flight parameters in real time to respond to environmental changes;

[0010] Communication module, responsible for data transmission between modules and communication between the UAV and the platform;

[0011] User interaction module, providing an operation interface for users;

[0012] A database for storing various collected data, bird activity model parameters, drone information, and scheduling strategy history records;

[0013] A federated learning optimization module aggregates bird behavior data collected from multiple regions to update prediction model parameters. It adapts general models to specific geographic environments through transfer learning and builds a virtual training environment based on digital twin technology to verify the model's generalization capabilities.

[0014] The energy management module is used to predict the total flight time of the drone swarm, generate a dynamic charging scheduling plan based on the distribution of charging piles, introduce energy consumption weighting factors into route planning, prioritize low-power cruise mode, and optimize task allocation based on battery degradation models to reduce high-energy-consuming routes;

[0015] The emergency response module is used to activate a multi-UAV formation to form a three-dimensional repelling barrier when a sudden bird flock is detected. It generates a dynamic interception path based on reinforcement learning and triggers an airspace emergency avoidance protocol. In extreme weather conditions, it automatically switches to a collaborative mode with ground-based bird repellent equipment.

[0016] The security management module is used to authenticate user logins, set operation permissions for different users, and ensure platform data security and operation security.

[0017] As a further improvement of this technical solution, the multi-source data acquisition module includes:

[0018] A dual-mode monitoring device, integrating millimeter-wave radar and infrared camera, collects real-time data on bird flight altitude, speed, and flock density;

[0019] Voiceprint recognition unit, which identifies high-risk bird species through spectrum analysis and marks the threat level;

[0020] The meteorological data interface obtains the precipitation probability, wind speed gradient and temperature change curve of the target area in real time; and uses a distributed sensor network for data collection to obtain multi-dimensional data information, and collects bird activity data, environmental data and drone status data in the target area in real time.

[0021] Through the above settings, the multi-source data acquisition module integrates multiple monitoring devices and data interfaces, adopts a distributed sensor network, realizes real-time collection of multi-dimensional data, and provides comprehensive and accurate data support for the platform.

[0022] As a further improvement of this technical solution, the working method of the data fusion processing module includes the following steps:

[0023] S1. Use spatiotemporal grid modeling technology to divide the target area into dynamic geographic grids;

[0024] S2, dynamically weighting meteorological data and bird activity characteristics based on the attention mechanism to generate a multi-dimensional fusion feature vector;

[0025] S3, extracting periodic activity patterns from historical data through a sliding time window algorithm;

[0026] S4. Use machine learning algorithms to train historical bird activity data and environmental data to establish a bird activity model and evaluate the priority of bird repellent needs in different regions.

[0027] Through the above settings, the design uses technologies such as spatiotemporal grid modeling and attention mechanism to extract data periodicity and establish a bird activity model, thereby improving the depth of data processing and the ability to assess the priority of bird repellent needs.

[0028] As a further improvement of this technical solution, the dynamic scheduling decision module includes:

[0029] Priority assessment unit, which generates a task priority sequence based on bird threat level, drone endurance, and airspace control rules;

[0030] The path optimization unit uses an improved ant colony algorithm with obstacle avoidance constraints to calculate the energy-optimal path for each UAV;

[0031] The redundant scheduling unit presets backup routes and emergency drone activation strategies; and formulates drone scheduling strategies based on the results of the big data analysis module, comprehensively considering the drone's current location, battery power, remaining bird-repellent materials, and the priority of bird-repellent demand in the target area, to determine the number of drones to perform bird-repellent tasks, flight routes, bird-repellent methods, and task execution time, and give priority to dispatching drones that meet the conditions and are closer to the target area to perform tasks.

[0032] Through the above settings, the dynamic scheduling decision module comprehensively considers multiple factors to generate a task priority sequence, formulates scheduling strategies, presets backup routes and emergency strategies, and realizes efficient and reasonable scheduling of drones.

[0033] As a further improvement of this technical solution, in the path optimization unit, the improved ant colony algorithm with obstacle avoidance constraint is implemented by the heuristic function Optimize the path search process: , Indicates the current location of the drone With target location The distance between and is the weight coefficient, , used to balance the influence of distance factors and the threat level of birds in the target area on path selection; Indicates the target location The bird threat level value at Indicates the maximum bird threat level value among all target areas.

[0034] Through the above settings, by introducing the improved ant colony algorithm into the heuristic function, the distance and threat level factors are balanced, the UAV path selection is optimized, and energy consumption and task execution efficiency are taken into account.

[0035] As a further improvement of this technical solution, the collaborative control module includes:

[0036] Multi-machine communication protocol supports state synchronization and command distribution of heterogeneous drone groups;

[0037] Dynamic parameter adjustment unit, adaptively adjusts the drone's flight altitude and bird-repelling intensity based on real-time wind speed data;

[0038] The task switching unit automatically switches to light deterrence mode when the sonic repelling fails;

[0039] The bird repellent execution module is installed on the UAV, and the bird repellent methods include at least sonic bird repellent, chemical bird repellent and optical bird repellent.

[0040] Through the above settings, the collaborative control module supports heterogeneous drone collaboration, can dynamically adjust parameters, switch bird-repelling modes, and be equipped with multiple bird-repelling methods to improve task execution flexibility and bird-repelling effects.

[0041] As a further improvement of this technical solution, the user interaction module includes:

[0042] The 3D visualization interface uses high-precision 3D modeling and real-time rendering technology to update the drone's position in real time. It also supports user interaction, and clicking the drone icon can check key parameters.

[0043] The manual intervention interface provides a prioritized list of tasks, which users can adjust by dragging and dropping or entering numerical values, with the system recalculating the schedule in real time. Users can also take control of specific drones and use pre-set commands to control flight and bird repelling.

[0044] A multi-tiered warning system uses multi-dimensional bird data and pre-defined algorithms to categorize threat levels, with yellow and red alerts. When an alert is triggered, the interface flashes, accompanied by audible and visual alarms, and a window pops up with warning details and response recommendations.

[0045] Data display and operation, using a variety of charts to dynamically display relevant data.

[0046] Through the above settings, the user interaction module provides a three-dimensional visualization interface, a manual intervention interface and a multi-level early warning system to enhance the user's operating experience and the ability to monitor and intervene in the platform.

[0047] As a further improvement of this technical solution, the federated learning optimization module includes:

[0048] Data aggregation and updating unit: uses secure multi-party computing to encrypt and aggregate bird behavior data from multiple regions, and uses an adaptive weighting strategy to determine the weight of data in each region;

[0049] Transfer learning adaptation unit: uses feature extraction and mapping technology to identify key features of the general model and map them to the feature space of a specific geographical environment;

[0050] Digital twin verification unit: Builds a highly simulated virtual training environment based on multi-source data, simulating various scenarios to generate test sets;

[0051] The evaluation feedback unit regularly evaluates indicators such as model accuracy, feeds the results back to relevant sub-modules, and adjusts strategies, parameters, and structures.

[0052] Through the above settings, data aggregation, transfer learning and digital twin verification are used to optimize the prediction model, improve the model adaptability and generalization ability, and adapt to different geographical environments.

[0053] As a further improvement of this technical solution, the energy management module includes:

[0054] Endurance prediction unit: By analyzing the drone's historical flight data, battery capacity, and real-time load weight, it establishes an endurance prediction model and outputs the remaining endurance of the drone swarm in different mission scenarios;

[0055] Charging scheduling unit: Combines the geographical distribution data of charging piles and the real-time power of drones, uses the shortest path algorithm to generate a dynamic charging scheduling plan, and prioritizes dispatching drones with power levels below 20% to the nearest charging pile for charging;

[0056] Energy consumption optimization unit: Introduces energy consumption weight parameters into the path planning algorithm, presets energy consumption coefficients for different flight modes based on the drone model, and automatically selects low-power cruise mode for non-emergency tasks;

[0057] Task allocation unit: A loss model is established based on historical battery attenuation data. When allocating tasks, low-energy consumption routes are prioritized to drones with a battery attenuation rate exceeding 30%, thereby balancing the overall energy consumption of the fleet.

[0058] Through the above settings, the energy management module can predict the endurance, generate a charging schedule, optimize energy consumption and task allocation, improve the endurance management and energy efficiency of the drone swarm, and balance the use of the swarm.

[0059] As a further improvement of this technical solution, the emergency processing module includes:

[0060] Cluster monitoring unit: By analyzing the bird density and movement speed in radar monitoring data, it can determine whether there is a sudden bird cluster;

[0061] Formation response unit: Launch three or more drones to form a V-shaped formation, and use the sonic repelling device to synchronously emit 15-25kHz pulse sound waves to form a three-dimensional repelling barrier;

[0062] Obstacle avoidance control unit: When the distance between the drone and the bird is detected to be less than 50 meters, the obstacle avoidance program is automatically triggered and the flight altitude and heading are adjusted using the preset safety avoidance strategy;

[0063] Equipment switching unit: When the meteorological sensor detects that the wind speed exceeds level 6 or the rainfall exceeds 20mm / h, it automatically switches to the fixed acoustic bird repellent device deployed on the ground and recalls all drones to a safe area.

[0064] Through the above settings, the design can timely monitor sudden bird flocks, initiate strategies such as formation driving away, obstacle avoidance and equipment switching, effectively respond to emergencies, and ensure safety and bird-repelling effects.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. This invention integrates millimeter-wave radar, infrared cameras, voiceprint recognition units, and meteorological data interfaces through a multi-source data acquisition module, combined with a distributed sensor network, to achieve real-time collection of multi-dimensional data such as bird activity trajectories, population classifications, meteorological parameters, and equipment status. The data fusion processing module utilizes spatiotemporal grid modeling, an attention mechanism, and a sliding time window algorithm to construct a spatiotemporally correlated multi-source heterogeneous data model, accurately extracting the periodic patterns of bird activity and the priority of bird repellent needs, addressing the shortcomings of traditional solutions with incomplete data and extensive evaluation.

[0067] 2. This invention integrates bird threat levels, drone endurance, battery decay models, and airspace control rules through a dynamic scheduling decision module. A priority assessment unit generates a task sequence, while a path optimization unit uses an improved ant colony algorithm to balance distance and threat level. The energy management module introduces an energy consumption weighting factor and optimizes task allocation based on a battery decay model. This achieves a dynamic balance between prioritized response in high-risk areas and drone endurance and energy consumption, avoiding the delayed response and insufficient endurance caused by irrational task allocation in traditional solutions.

[0068] 3. The collaborative control module in the design of this invention supports dynamic parameter adjustment and multi-mode expulsion strategies for heterogeneous drones. The emergency response module activates a three-dimensional barrier for drone formations and reinforcement learning interception paths in the event of a sudden bird flock, and automatically switches to the collaborative mode of ground equipment under extreme weather conditions. The federated learning optimization module adapts to specific geographical environments through transfer learning, and digital twin technology verifies the model's generalization capabilities. This constructs an adaptive mechanism covering differentiated expulsion of multiple bird species and ground-air collaborative emergency response, solving the problem of strategy rigidity of traditional solutions in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a system framework diagram of the present invention;

[0070] Figure 2 This is a framework diagram of the multi-source data acquisition module in the present invention;

[0071] Figure 3 This is a framework diagram of the collaborative control module in the present invention;

[0072] Figure 4 This is a framework diagram of the federated learning optimization module in the present invention. DETAILED DESCRIPTION

[0073] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0074] like Figures 1-4 As shown, this embodiment provides an integrated platform for bird repelling and drone scheduling based on big data analysis, including:

[0075] Multi-source data acquisition module 100, used to integrate meteorological data, geographic information data, historical bird activity data and real-time monitoring data, including bird activity tracks monitored by radar and population classification information from voiceprint recognition;

[0076] In this embodiment, the multi-source data acquisition module 100 includes:

[0077] The dual-mode monitoring device integrates millimeter-wave radar and infrared camera to collect bird flight altitude, speed and cluster density data in real time. Through the integration of millimeter-wave radar and infrared camera, real-time monitoring of bird three-dimensional trajectories (altitude, speed, density) is achieved, making up for the blind spots of a single sensor in complex environments (such as at night and in fog).

[0078] The voiceprint recognition unit identifies high-risk bird species through spectrum analysis and marks the threat level. The voiceprint recognition unit, combined with spectrum analysis technology, can automatically distinguish high-risk bird species such as magpies and crows and issue graded warnings, providing precise target positioning for subsequent bird-repelling strategies.

[0079] A meteorological data interface provides real-time information on precipitation probability, wind speed gradients, and temperature curves in the target area. A distributed sensor network is used for data collection, acquiring multi-dimensional information, including real-time bird activity data, environmental data, and drone status data within the target area. The design also supports edge computing preprocessing (such as image recognition and data compression at the node), reducing communication bandwidth requirements and improving system response speed.

[0080] As a further explanation of this embodiment, the distributed sensor network mainly includes high-definition cameras, meteorological sensors, ultrasonic sensors, etc., and uses multiple types of sensors to build a heterogeneous network, which can cover the three dimensions of bird activities, environmental conditions, and equipment parameters, forming a full-scene data collection capability.

[0081] As a further illustration of this embodiment, the bird activity data includes: bird species, number, flight trajectory, resting location, etc.;

[0082] Environmental data includes: weather conditions, light intensity, wind speed and direction, etc.

[0083] Drone status data includes: drone battery level, location, and remaining bird repellent.

[0084] It's worth noting that the dual-mode monitoring device integrates a 77GHz millimeter-wave radar (ARS408-21, with a detection range of ≥2km and an angular resolution of ≤1°) and a 640×512 pixel infrared camera (FLIR Boson 640). It utilizes a hardware-synchronized trigger mechanism (with a synchronization error of ≤50μs). A built-in noise reduction process first removes salt-and-pepper noise from the image using a median filter, then dynamically segments the target area using the Otsu algorithm. Finally, a non-local means filter is used to preserve edges, ensuring that ≥90% of pixels are recognized effectively during nighttime detection.

[0085] Furthermore, the data fusion processing module 200 processes the heterogeneous data sources using a spatiotemporal feature extraction algorithm to generate a spatiotemporal-correlated bird density distribution heat map;

[0086] In this embodiment, the working method of the data fusion processing module 200 includes the following steps:

[0087] S1. Use spatiotemporal gridding modeling technology to divide the target area into dynamic geographic grids. This design breaks through the limitations of traditional fixed grid division and uses dynamic geographic grid technology (such as density-based adaptive division) to achieve refined representation of high-frequency activity areas (such as airport runways and farmland core areas) (grid granularity can be as small as 20m×20m), resolving the contradiction between computational redundancy in sparse areas and feature loss in dense areas.

[0088] S2. Dynamically weight meteorological data and bird activity characteristics based on the attention mechanism to generate a multi-dimensional fusion feature vector. This design targets the nonlinear correlation between meteorological data (wind speed, precipitation) and bird activity characteristics (species, density). It dynamically allocates fusion weights through the attention mechanism (such as automatically increasing the weight of the impact of wind speed on bird aggregation before a rainstorm), avoiding the problem of key information being overwhelmed by traditional equal-weight fusion.

[0089] S3. Extract periodic activity patterns from historical data through a sliding time window algorithm. This design uses a sliding time window algorithm (e.g., 15-minute / 30-minute variable windows) to extract diurnal and weekly cycles of bird activity (e.g., morning and evening foraging peaks, and flocking frequency during migration seasons), providing forward-looking input in the time dimension for predictive bird-repelling strategies.

[0090] S4. Use machine learning algorithms to train historical bird activity data and environmental data to build a bird activity model and assess the priority of bird repellent needs in different regions. This design, based on a bird activity model trained with historical data, outputs the priority of bird repellent needs in each region (a quantitative score of 0-10), replacing traditional manual judgment and achieving scientific and automated resource scheduling.

[0091] As a further explanation of this embodiment, this embodiment further deepens the spatiotemporal gridding modeling technology, including:

[0092] The gridding algorithm has been optimized, using a density clustering algorithm (DBSCAN) to dynamically determine grid granularity. A fine 20m×20m grid is automatically generated for hotspots with bird densities ≥10 birds / 100m³; a coarse 100m×100m grid is used for sparse areas with densities <2 birds / 100m³. This design balances computational efficiency and accuracy. The three-dimensional grid has been expanded to include a vertical height dimension (with a tiered accuracy of 10m), such as low altitude (0-50m), mid-altitude (50-200m), and high altitude (>200m). This design, combined with radar altitude data, generates a three-dimensional heat map containing altitude information. This design addresses the problem of traditional two-dimensional models being unable to represent stratified bird activity.

[0093] Grid attribute definition: each grid is associated with geographical features (vegetation coverage, distance to water, terrain slope), airspace features (no-fly zone boundaries, charging station locations), and historical bird-scaring records (number of bird-scaring operations in the past 24 hours, effective radius), forming a multi-dimensional decision-making factor.

[0094] As a further explanation of this embodiment, the specific implementation method of the attention mechanism in this embodiment is as follows:

[0095] The feature weighted calculation method is used, and the calculation method is as follows: define the meteorological-bird feature correlation matrix ,in, ;in: For the meteorological characteristics (such as wind speed, temperature), For the Corr is the Pearson correlation coefficient, which ensures that the weight distribution conforms to the intrinsic correlation of the data. For example, when the wind speed is detected to be greater than 8m / s, the weight of the wind direction feature on the bird flight trajectory is automatically increased by 30%, reflecting the dominant influence of strong wind on the movement direction of the bird flock.

[0096] As a further explanation of this embodiment, this embodiment also improves the sliding time window algorithm, including:

[0097] Adaptive window length selection: Based on Fourier transform analysis of historical data periodicity, the optimal window length is automatically matched: for scenes with obvious daily cycles (such as bird activity at airports), a 60-minute window is used; for high-frequency migration scenes, a 15-minute window is used, and the window overlap rate is set to 50% to avoid boundary effects;

[0098] The dimension of temporal feature extraction has been expanded; semantic features such as lunar phases (new moon, first quarter moon, full moon, last quarter moon) and major event days (such as the start date of the migration season) have been added to enhance the richness of time series analysis.

[0099] Furthermore, the dynamic scheduling decision module 300 generates a multi-UAV collaborative path planning solution through an improved ant colony algorithm based on the heat map and UAV state parameters;

[0100] In this embodiment, the dynamic scheduling decision module 300 includes:

[0101] The priority assessment unit generates a task priority sequence based on the bird threat level, drone flight time, and airspace control regulations. This design generates a task priority sequence based on the bird threat level (0-10, combining species, density, and flight altitude), drone flight capacity (battery charge, cargo volume, and payload), and airspace control regulations (weighting of no-fly and restricted-fly zones). For example, a low-altitude swarm in an airport clear area (threat value of 9 points + no-fly zone weight of 3 points) is marked as the highest priority (10 points), triggering a response within 5 minutes. A single sparrow in a farmland area (threat value of 3 points) is marked as low priority (3 points), giving priority to low-battery drones.

[0102] The path optimization unit uses an improved ant colony algorithm with obstacle avoidance constraints to calculate the energy-optimal path for each UAV;

[0103] The redundant dispatch unit pre-sets backup routes and emergency drone activation strategies. Based on the results of the big data analysis module, it formulates a drone dispatch strategy. This strategy comprehensively considers the drone's current location, battery level, remaining bird repellent supplies, and the priority of bird repellent needs in the target area. It determines the number of drones to be assigned to the bird repellent mission, the flight route, the bird repellent method, and the mission execution time, prioritizing drones that meet the requirements and are closer to the target area. By pre-setting backup routes (generating two to three alternative routes based on historical data and risk assessment of the primary route, with a distance deviation of ≤15%) and establishing an emergency drone activation strategy, a backup drone (pre-deployed with a battery level of ≥80%) is activated within 10 seconds if the drone's battery level drops below 20%, its supply level drops below 15%, or a malfunction occurs. The backup drone automatically takes over the mission and resumes the interrupted mission parameters (such as sonic frequency and flight altitude). For example, if a drone detects a sudden drop in battery level to 15% during a bird repellent mission, the system immediately dispatches a backup drone within 3 km, ensuring uninterrupted repellent operations and a mission interruption rate of less than 5%.

[0104] In this embodiment, in the path optimization unit, the improved ant colony algorithm with obstacle avoidance constraint is implemented by the heuristic function Optimize the path search process: , Indicates the current location of the drone With target location The distance between and is the weight coefficient, , used to balance the influence of distance factors and the threat level of birds in the target area on path selection; Indicates the target location The bird threat level value at Indicates the maximum bird threat level value among all target areas.

[0105] As a further explanation of this embodiment, this embodiment uses the weight coefficient and For further explanation: Airport scene :Airport clear areas are highly sensitive to bird threats, and birds approaching them can easily cause safety accidents. (Threat level weight) allows the algorithm to prioritize bird-repelling routes in high-threat areas, such as during aircraft takeoff and landing. Even if the target is far away, high-threat areas are prioritized to ensure aviation safety. :Farmland bird repellent has relatively low requirements for timeliness and is more concerned with energy consumption costs. (Distance weight) makes the algorithm tend to choose short-distance paths, reducing the energy consumption of drone flights, such as driving away scattered flocks of sparrows, giving priority to dispatching short-range drones, and reducing overall operating costs.

[0106] As a further illustration of this embodiment, this heuristic function enables the drone to consider not only the distance between itself and the target location when planning its path, but also the degree of bird threat in the target area. This allows the drone to more efficiently complete the bird-scaring mission while ensuring optimal energy consumption. In the redundant scheduling unit, the backup route is planned based on the flight data of the main route and an assessment of risk factors such as obstacles and weather changes that may appear in the flight area. This ensures that if a problem occurs on the main route, the drone can quickly switch to the backup route to continue the mission. The emergency drone activation strategy automatically triggers the emergency drone to be put into operation based on the drone's real-time status parameters, such as when the battery level drops below a set threshold, the remaining amount of bird-scaring supplies drops below a set value, or a fault occurs, thereby ensuring the continuity of the bird-scaring operation.

[0107] As a further explanation of this embodiment, this embodiment further deepens the priority evaluation unit, including:

[0108] A three-level threat assessment system (low / medium / high) was established, combining bird species (e.g., birds of prey are considered high threat), flock density (>50 birds / grid are considered high threat), and flight altitude (<50 meters approaching the target area is considered high threat), outputting a threat score of 0-10 (e.g., a flock of kestrels is considered a 9-point threat, a single sparrow is considered a 3-point threat). Airspace control rules were integrated: no-fly zones (weighted +3 points), restricted flight zones (weighted +2 points), and regular areas (weighted 0 points) to ensure that dispatching complies with aviation safety regulations.

[0109] Endurance Assessment: A flight endurance prediction model (with an error of ≤10%) is developed based on the remaining battery charge (%), remaining bird repellent material (such as acoustic device operating hours and chemical capacity), and payload weight (kg). Equipment Health: This includes the battery degradation rate (>30% indicates an inefficient drone, prioritizing low-energy tasks) and sensor failure history (≥2 failures in the past 7 days triggers priority dispatch for backup drones).

[0110] As a further illustration of this embodiment, the heuristic function This is the algorithm's core guidance mechanism, designed to balance two key factors in drone path planning: distance cost and target threat severity. It provides an "intelligent guide" for the algorithm, allowing the drone to consider both its proximity to the target and the severity of the bird threat in the target area when selecting a path, thereby achieving a more efficient and reasonable mission execution strategy.

[0111] in, : Indicates the current location of the drone With target location The Euclidean distance between (the unit can be adjusted according to the actual scenario, such as kilometers). The closer the distance, The larger the value of , the stronger the inspiration for path selection, which reflects the algorithm's pursuit of "optimal energy consumption";

[0112] : Target location The bird threat level value of the area (e.g., derived through comprehensive assessment of voiceprint recognition, flock density, etc., ranging from 0 to 10 points). The higher the value, the greater the threat posed by bird activity in the area to the target (e.g., airport, farmland);

[0113] : The maximum bird threat level value in all target areas. Divide by , to achieve the normalization of threat level (the result is between 0-1), to ensure the relative balance with the distance term in terms of dimension, so that the two can be reasonably weighted and integrated;

[0114] and : Weight coefficient ( , ,and ), which is used to dynamically adjust the relative importance of the distance factor and the threat level factor.

[0115] like Larger values ​​(such as ), the algorithm will tend to choose targets that are closer and give priority to reducing energy consumption;

[0116] like Larger values ​​(such as ), the algorithm will prioritize responding to areas with high threat levels to ensure timely processing of high-risk areas.

[0117] It is understandable that traditional ant colony algorithms usually rely only on pheromone concentration and heuristic information (such as distance) for path search, which is difficult to adapt to the needs of multi-factor decision-making in complex scenarios. However, this improved algorithm introduces the bird threat level factor to make path planning more suitable for actual application scenarios (such as airport bird control). For example, when high-threat areas (such as densely populated birds near the runway) and low-threat areas coexist, the algorithm can adjust the and , prioritizes planning routes to high-threat areas, avoiding neglecting key threats due to simply pursuing proximity, significantly improving the safety and effectiveness of mission execution;

[0118] For example, taking the example of bird control at an airport, assume there are two target areas:

[0119] Area A: 2km from the drone, threat level 8 ;

[0120] Area B: 1.5km from the drone, threat level 4 ;

[0121] If you take , ,but:

[0122] ;

[0123] ,

[0124] at this time ,The algorithm will prioritize guiding the drone to area A, which,although it is farther away, but has a higher threat level, which meets the need of prioritizing high-risk areas in the ,airport scenario;

[0125] Through the above design, the improved ant colony algorithm with obstacle avoidance constraints can comprehensively consider the distance and threat level in complex environments, realize intelligent optimization of the drone path, and effectively improve the execution efficiency and safety of bird-scaring tasks.

[0126] Furthermore, the collaborative control module 400 controls the UAV to perform sonic, light, or physical repelling tasks according to the path planning scheme, and adjusts flight parameters in real time to respond to environmental changes;

[0127] In this embodiment, the collaborative control module 400 includes:

[0128] Multi-machine communication protocol supports state synchronization and command distribution of heterogeneous drone groups; through customized multi-machine communication protocols (such as extended protocols based on MQTT), it can support state synchronization and command distribution of different types of drones (fixed-wing, multi-rotor), solving the problem of poor compatibility of traditional protocols.

[0129] The dynamic parameter adjustment unit adaptively adjusts the drone's flight altitude and bird-repelling intensity based on real-time wind speed data. The dynamic parameter adjustment unit can automatically optimize the flight altitude (adjustment step size 1m) and bird-repelling intensity (sound wave power adjustment range 80-120dB) according to the real-time wind speed (error ±0.5m / s), maintaining a high repelling success rate even in a level 6 strong wind environment.

[0130] The task switching unit automatically switches to light deterrence mode when the sonic repellent fails. The task switching unit automatically switches the bird repellent mode (response time < 3 seconds) based on dual-condition judgment (distance change < threshold and device is normal). For example, it can seamlessly switch from sonic bird repellent (frequency range 1-25kHz) to light deterrence (light intensity range 1000-5000lux), improving the repellent effect in complex scenarios. The conditions for determining sonic repellent failure are: The distance change between the bird and the target area is less than the set distance change threshold. , and the sound wave emitting device works normally, it is considered that the sound wave expulsion has failed. At this time, the task switching unit triggers the light deterrence mode to start.

[0131] The bird repellent execution module is installed on the UAV, and the bird repellent methods include at least sonic bird repellent, chemical bird repellent and optical bird repellent.

[0132] As a further illustration of this embodiment, the bird repellent execution module can automatically switch or combine bird repellent methods and adjust repellent parameters based on the bird species and activity, as well as the bird's response to different repellent methods. For example, for birds sensitive to sound, acoustic repellent is prioritized, with the frequency and intensity of the sound waves adjusted based on the bird's response. For birds that are attracted to light, optical repellent is combined with light color and flashing frequency to adjust parameters. For some difficult-to-repel birds, a strategy of synergizing chemical repellent with other repellent methods is adopted, dynamically adjusting the amount and range of the chemical repellent sprayed based on the bird's behavior.

[0133] As a further illustration of this embodiment, the decision logic of the task switching unit includes:

[0134] The failure determination algorithm first defines a dual threshold mechanism, including a spatial threshold: the distance between the bird and the target area changes by less than 10 meters within 5 consecutive minutes; and a time threshold: continuous acoustic emission for ≥3 minutes and normal device operation (e.g., amplifier temperature <70°C). A confidence assessment is then introduced: when the dual thresholds are met, three rapid verifications are triggered (with 10-second intervals). If all conditions are met, the device is deemed to have failed.

[0135] The mode switching process includes sending a switching command to the bird-repelling execution module after determining that the sound wave has failed; starting the light deterrent (initial parameters: white light, frequency 2Hz, intensity 3000lux); and synchronously recording bird behavior data before and after the switch (such as gathering density and flight direction) for subsequent strategy optimization.

[0136] It should be noted that the dynamic parameter adjustment unit receives real-time wind speed data (from a meteorological sensor with an accuracy of ±0.5m / s) and adjusts the flight altitude according to the following rules:

[0137] When the wind speed is ≤ Level 6 (≤13.8m / s), the flight altitude = preset altitude + wind speed × 2m / s (adjustment step size 1m);

[0138] When the wind speed is greater than level 6, the altitude lock mode is triggered, maintaining the current altitude within a range of ±5m, and at the same time increasing the bird-repelling sound wave power to 110dB.

[0139] Furthermore, the communication module 500 is responsible for data transmission between modules and between the drone and the platform. This module ensures data transmission between various modules within the system (such as the data fusion processing module 200, the dynamic scheduling decision module 300, and the collaborative control module 400), enabling effective collaboration between the different functional modules to form an integrated whole, ensuring the proper operation of the entire bird-repelling system. This design also enables two-way communication between the drone and the platform. The platform can obtain real-time drone status information (such as location, battery level, and the operating status of the bird-repelling device) and send instructions to the drone (such as flight path, bird-repelling mission, etc.), thereby enabling remote control and management of the drone. Furthermore, through stable and reliable data transmission, the communication module 500 helps improve the system's response speed and decision-making accuracy, thereby enhancing the efficiency and reliability of the entire bird-repelling system.

[0140] As a further illustration of this embodiment, the communication module 500 should support multiple communication protocols to adapt to different application scenarios and equipment requirements. For example, for short-range communication (such as communication between modules), the CAN bus protocol can be used, which has the advantages of high reliability, strong real-time performance, and good anti-interference ability. The data transmission rate can reach 1Mbps, which can meet the requirements of high-speed data transmission between modules. For long-distance communication (such as communication between drones and platforms), 4G / 5G network communication protocols can be used to achieve large-scale coverage and high-speed data transmission, ensuring that drones can maintain a stable connection with the platform in different areas; at the same time, in order to achieve seamless communication between different modules and devices, the communication module 500 should have a protocol adaptation function. When different modules use different communication protocols, the communication module 500 can perform protocol conversion so that data can be correctly transmitted between different protocols. For example, convert the data of the CAN bus protocol into data of the 4G / 5G network protocol so that the drone can upload the data to the platform;

[0141] It should be noted that in complex environments, communication signals may be interfered with or attenuated, affecting the stability of data transmission. The communication module 500 can use signal enhancement technology (such as using high-gain antennas) and relay technology (such as setting up signal relay stations) to enhance signal strength and coverage. For example, in mountainous areas or areas with dense buildings, setting up signal relay stations can effectively solve the problem of signal obstruction and ensure smooth communication between the drone and the platform. At the same time, to improve the reliability of communication, the communication module 500 can adopt a redundant communication link design. For example, the drone is equipped with both 4G / 5G network communication and satellite communication. When the 4G / 5G network signal is poor, it automatically switches to satellite communication to ensure the continuity of data transmission.

[0142] Furthermore, the user interaction module 600 provides an operation interface for the user;

[0143] In this embodiment, the user interaction module 600 includes:

[0144] The 3D visualization interface uses high-precision 3D modeling and real-time rendering technology to update the drone's position in real time. It also supports user interaction, and clicking the drone icon can view key parameters. This design can use oblique photogrammetry (accuracy ±5cm) and laser point cloud fusion technology to construct high-precision 3D scenes that include terrain, buildings, and obstacles, and supports custom landmark annotation (such as airport runways and no-fly zones). Key parameters include: battery level, altitude, cargo volume, flight trajectory, etc.

[0145] The manual intervention interface provides a task priority list, which users can adjust by dragging and dropping or entering numerical values, and the system recalculates the schedule in real time. Users can take over a specific drone and control its flight and bird-scaring capabilities using preset commands. When users adjust the priority, the system automatically recommends the optimal scheduling plan (e.g., "Recommended plan: Setting this task priority to 7 will reduce response time by 20%"). The takeover mode also supports importing preset routes (KML / GPX format) and editing waypoints (dragging to adjust coordinates), and provides a "follow mode" in which the drone automatically follows a user-specified moving target.

[0146] A multi-tiered warning system uses multi-dimensional bird data and pre-defined algorithms to categorize threat levels, with yellow and red alerts. When an alert is triggered, the interface flashes, accompanied by audible and visual alarms, and a window pops up with warning details and response recommendations.

[0147] Data display and manipulation uses a variety of charts to dynamically display relevant data. This design allows for dynamic data display using a variety of charts, including line charts (showing bird population trends), heat maps (with a resolution of 10m×10m, displaying bird density distribution), and bar charts (comparing bird repellent effectiveness). This system supports regional filtering, conditional filtering (by species, time, and level), and trend prediction (based on the ARIMA model, with an accuracy rate of ≥70%). Data can be exported to CSV / Excel formats.

[0148] It should be noted that the warning trigger algorithm in this embodiment is to combine the three dimensions of bird aggregation density (threshold ≥ 50 birds / hectare), flight altitude (threshold ≤ 50m), and approach speed (threshold ≥ 5m / s) to generate a threat index through weighted calculation (weights are configurable). The formula is: Threat Index = 0.5 × Density Factor + 0.3 × Altitude Factor + 0.2 × Speed ​​Factor. Among them, yellow warning: threat index ≥ 0.4 and < 0.7; red warning: threat index ≥ 0.7. When the red warning is triggered, the system automatically performs the following operations:

[0149] Circle the warning area (500m radius highlighted);

[0150] Dispatch the three nearest drones to the destination (flight speed increased to 20m / s);

[0151] Send SMS / email to notify the designated person in charge (supports polling of multiple contacts);

[0152] Start the recording function (save the monitoring images 5 minutes before and after the warning).

[0153] Furthermore, database 700 is used to store various types of collected data, bird activity model parameters, drone information, and scheduling strategy history records; this design centrally stores collected bird activity data (species, number, trajectory, etc.), meteorological data (wind speed, precipitation, etc.), drone information (location, power, model, etc.) and scheduling strategy history records, and adopts a unified data structure and storage specifications to facilitate rapid retrieval and call, with a retrieval response time of ≤0.5 seconds.

[0154] As a further illustration of this embodiment, the database 700 can use a relational database 700 (such as PostgreSQL) to store structured data (such as drone numbers and mission timestamps), and a non-relational database (such as MongoDB) to store semi-structured / unstructured data (such as JSON logs of bird activity trajectories and surveillance video clips). This design takes into account both data consistency and storage flexibility.

[0155] Furthermore, the federated learning optimization module 800 is used to aggregate bird behavior data collected from multiple regions to update the prediction model parameters, adapt the general model to a specific geographical environment through transfer learning, and build a virtual training environment based on digital twin technology to verify the model's generalization ability;

[0156] In this embodiment, the federated learning optimization module 800 includes:

[0157] Data Aggregation and Update Unit: Secure multi-party computation (SMC) is used to encrypt and aggregate bird behavior data from multiple regions, and an adaptive weighting strategy is used to determine the weight of each region's data. Secure multi-party computation (such as Paillier homomorphic encryption) is used to aggregate multi-region data and update model parameters without leaking the original data, ensuring data privacy. Adaptive weighting strategies (such as dynamically adjusting weights based on data quality and sample diversity) improve model training efficiency.

[0158] Transfer learning adaptation unit: Using feature extraction and mapping technology, it identifies the key features of the general model and maps them to the feature space of a specific geographical environment. Transfer learning technology (such as the deep domain adaptation network (DAN)) enables rapid adaptation of the general model to a specific geographical environment, while identifying key features (such as migratory bird migration paths and habitat characteristics) improves mapping accuracy.

[0159] Digital Twin Verification Unit: Builds a highly simulated virtual training environment based on multi-source data, simulating various scenarios to generate test sets. Builds a digital twin environment of the mountain airport, including real terrain and typical meteorological conditions (such as valley winds). Simulates and generates 1,000 test scenarios (including extreme weather and high-density clusters).

[0160] The evaluation feedback unit regularly evaluates indicators such as model accuracy, feeds the results back to relevant sub-modules, and adjusts strategies, parameters, and structures.

[0161] It should be noted that the federated learning optimization module 800 achieves continuous optimization of the bird behavior prediction model through privacy-preserving data aggregation, environment-adaptive transfer learning, and high-precision digital twin verification, specifically including:

[0162] 1. Data aggregation and update unit: This unit uses a hybrid encryption scheme (AES + Paillier homomorphic encryption) to protect the privacy of original data and aggregates multi-region gradient parameters through an adaptive weighting strategy.

[0163] The weight calculation formula is: Data quality is assessed through completeness (missing values ​​<5%) and accuracy (error <10%), and sample diversity is calculated through species richness. This solution ensures that the risk of information leakage is ≤ 0.01% and increases convergence speed by 30%.

[0164] 2. Transfer Learning Adaptation Unit: A Deep Domain Adaptation Network (DAN) is used to enable cross-geographic model migration. Image features are extracted using ResNet-50 and behavioral sequence features are extracted using BiLSTM. The maximum mean difference (MMD) between the source and target domains is calculated (with an MMD threshold of 0.15). Adversarial training is used to align feature distributions. The parameters of the first three layers of the network are frozen, and only the last two layers are fine-tuned (with a learning rate of 0.001). This achieves adaptation speeds exceeding those of traditional methods.

[0165] 3. Digital Twin Verification Unit: A high-precision virtual environment is built using the UnityPhysX engine, integrating terrain data (1m resolution), meteorological data (wind speed accuracy ±0.5m / s), and bird biometrics. This simulates scenarios such as extreme weather (wind speeds of 10-20m / s) and complex terrain (slopes of 15-45°), generating over 100,000 test samples per day. Data consistent with real-world equipment is generated by deploying simulated sensors (120° field of view, 1920×1080 resolution).

[0166] 4. Evaluation and Feedback Unit: Establish a multi-indicator evaluation system. When performance is found to be substandard, it automatically triggers transfer learning parameter adjustments, digital twin scenario optimization, and data weight updates. This forms a continuously iterative closed-loop system, and the model performance improves monthly.

[0167] Through the above design, the federated learning optimization module 800 achieves rapid environmental adaptation and continuous optimization of the model while protecting data privacy, which can improve the prediction accuracy of the cross-regional bird-repellent system and enhance the adaptability and reliability of the system in different geographical environments.

[0168] Furthermore, the energy management module 900 is used to predict the total flight time of the drone swarm, generate a dynamic charging scheduling plan based on the distribution of charging piles, introduce energy consumption weighting factors into route planning, give priority to low-power cruise mode, and optimize task allocation based on the battery degradation model to reduce high-energy-consuming routes;

[0169] In this embodiment, the energy management module 900 includes:

[0170] The Endurance Prediction Unit analyzes historical drone flight data, battery capacity, and real-time payload weight to build a flight time prediction model, outputting the remaining flight time of the drone swarm in different mission scenarios. This unit uses an LSTM neural network to construct the endurance model, taking historical flight data (speed, altitude), battery parameters (charge level, health), and real-time payload as inputs. Dropout regularization is used to prevent overfitting, achieving a validation set root mean square error of ≤12 minutes. During flight, an extended Kalman filter is used to correct the prediction every 30 seconds to account for sudden weather changes (e.g., gust response time ≤2 seconds), resulting in improved prediction accuracy compared to traditional methods.

[0171] Charging scheduling unit: Combines the geographical distribution data of charging piles and the real-time power of drones, uses the shortest path algorithm to generate a dynamic charging scheduling plan, and prioritizes dispatching drones with power levels below 20% to the nearest charging pile for charging;

[0172] Energy consumption optimization unit: Introduces energy consumption weight parameters into the path planning algorithm, presets energy consumption coefficients for different flight modes based on the drone model, and automatically selects low-power cruise mode for non-emergency tasks;

[0173] Task allocation unit: A loss model is established based on historical battery attenuation data. When allocating tasks, low-energy consumption routes are given priority to drones with a battery attenuation rate exceeding 30%, thereby balancing the overall energy consumption of the fleet.

[0174] As a further illustration of this embodiment, the charging scheduling unit generates an efficient dynamic charging plan based on the geographical distribution of charging piles and the real-time status of drones. The core mechanism is as follows:

[0175] Data fusion decision-making: Combining the location information of charging piles (supporting fixed charging piles and mobile charging facilities), the real-time power level of the drone and the mission priority, the optimal charging route is generated through path optimization algorithms (such as the shortest path algorithm).

[0176] Tiered charging strategy: Emergency charging: For low-battery drones (e.g., battery level below a preset threshold), priority is given to the nearest available charging station to ensure rapid energy replenishment to maintain mission continuity;

[0177] Conventional charging: For non-emergency drones, charging is carried out along the way based on mission path planning, reducing detour losses and improving the efficiency of charging piles.

[0178] Load balancing: Real-time monitoring of charging pile load status, dynamic adjustment of charging distribution strategy, and avoidance of waiting delays caused by local overload.

[0179] Cross-device collaboration: Supports communication and interaction between charging piles and drones (such as charging status feedback and device compatibility verification), realizing the automation and intelligence of the charging process.

[0180] As a further illustration of this embodiment, the energy consumption optimization unit reduces the overall energy consumption of the drone swarm by intelligently matching missions with flight modes, specifically including:

[0181] Multi-mode energy consumption management: Based on the drone model, mission type (such as patrol, expulsion) and environmental conditions, differentiated flight modes (such as low-power cruise mode and high-speed response mode) are preset, and the execution strategy with optimal energy consumption is automatically selected.

[0182] Path Planning Integration: Introducing an energy consumption weighting factor into the drone's path planning algorithm to balance mission urgency with energy consumption. For example, for non-urgent missions, low-power cruise mode is prioritized, reducing unnecessary energy loss by adjusting flight parameters such as speed and altitude.

[0183] Dynamic parameter adjustment: Combining real-time meteorological data (such as wind speed and direction) with load changes, flight parameters are adaptively adjusted to further optimize energy efficiency and extend the drone's single mission endurance.

[0184] It should be added that the task allocation unit optimizes task scheduling based on the battery health status to balance the overall energy consumption of the fleet. The technical key points are as follows:

[0185] Battery degradation modeling: By analyzing historical battery usage data (such as the number of charge and discharge cycles and operating temperature), a battery degradation model is established to evaluate the health status of each drone (such as high, medium, and low degradation levels).

[0186] Differentiated task assignment: For high-attenuation UAVs, low-energy consumption routes (such as plain areas and low-load tasks) are prioritized to avoid high-energy consumption tasks (such as complex mountain terrain and strong wind environments); for newly invested or low-attenuation UAVs, high-priority or high-energy consumption tasks are reasonably assigned to balance the overall loss of the fleet.

[0187] Multi-objective optimization: Intelligent algorithms (such as multi-objective evolutionary algorithms) are used to comprehensively consider mission urgency, drone health status, and path energy consumption to generate the optimal task allocation plan and extend the overall service life of the fleet.

[0188] Furthermore, the emergency response module 1000 is used to activate a multi-UAV formation to form a three-dimensional repelling barrier when a sudden bird flock is detected, generate a dynamic interception path based on reinforcement learning, trigger an airspace emergency avoidance protocol, and automatically switch to a ground-based bird repelling equipment collaborative mode under extreme weather conditions;

[0189] In this embodiment, the emergency processing module 1000 includes:

[0190] Cluster monitoring unit: By analyzing the bird density and movement speed in the radar monitoring data, it determines whether it is a sudden bird cluster. For example, when the bird density exceeds 8 / 100m³ and the movement speed exceeds 4m / s for 5 consecutive minutes, it is determined to be a sudden bird cluster. The cluster monitoring unit realizes automatic identification of sudden bird clusters through quantitative analysis of radar data. Compared with manual observation, the response speed is much faster and the missed detection rate is reduced.

[0191] Formation response unit: Launch 3 or more (preferably an odd number) drones to form a V-shaped formation, and synchronously emit 15-25kHz pulse sound waves through the sonic repelling device to form a three-dimensional repelling barrier; launch ≥3 drones to form diversified formations (V-shaped, diamond or circular), and dynamically select the optimal formation through reinforcement learning; at the same time, synchronously emit 15-25kHz pulse sound waves, support adaptive parameter adjustment for bird species (such as 20kHz and 100dB for sparrows), forming a three-dimensional repelling barrier, which is more efficient than a single machine.

[0192] Obstacle Avoidance Control Unit: When the distance between the drone and a bird is detected to be less than 50 meters, the obstacle avoidance program is automatically triggered, and the flight altitude and heading are adjusted using a preset safety avoidance strategy. This design integrates millimeter-wave radar (accuracy ±0.5m), vision, and ultrasonic sensor data to build a 360-degree detection model. When the distance is detected to be less than 50m, the improved artificial potential field method is triggered for obstacle avoidance, dynamically adjusting the flight altitude (±50m) and heading (±30°), improving the success rate of obstacle avoidance in complex scenarios.

[0193] Equipment Switching Unit: When weather sensors detect wind speeds exceeding level 6 or rainfall exceeding 20 mm / h, the system automatically switches to a fixed ground-based acoustic bird repellent device and recalls all drones to a safe area. When wind speeds exceed level 6 or rainfall exceeds 20 mm / h, the fixed ground-based devices (acoustic / laser / strobe light) are automatically activated, synchronously transmitting cluster information to adjust parameters (e.g., an eagle triggers a 20kHz + laser combination). Drones are recalled along pre-set safe routes, prioritizing charging stations (battery level <30%) and automatically reporting to the air traffic control system (response time <10 seconds), ensuring continuous repelling in extreme weather conditions.

[0194] As a further illustration of this embodiment, the obstacle avoidance control unit in this embodiment can adopt an improved artificial potential field method (incorporating a bird motion prediction model) to calculate the obstacle avoidance path in real time, avoiding the mechanical nature of traditional preset strategies, thereby improving the obstacle avoidance success rate in complex scenarios.

[0195] Furthermore, the security management module 1100 is used to authenticate user logins, set operation permissions for different users, and ensure platform data security and operation security.

[0196] It should be noted that the security management module 1100 ensures the operational and data security of the platform through layered authentication, fine-grained permission control, and full-link security technology, specifically including:

[0197] 1. Multi-factor authentication: Utilizes a three-tiered authentication system: password + dynamic token + biometrics. Passwords must meet complexity requirements of 8-20 characters. Dynamic tokens support Google Authenticator, SMS verification codes, and biometrics. First login requires a bound terminal device; unbound devices require secondary confirmation to minimize the risk of identity theft.

[0198] 2. Role-based permission management:

[0199] Divide the roles into three levels: administrator, operator, and observer:

[0200] Administrator: Has full data read and write, device management, and policy configuration permissions. Key operations require double review.

[0201] Operator: Can perform task scheduling and drone control, and can only access data in the jurisdiction area;

[0202] Observer: Read-only visual interface, no control permissions.

[0203] Supports temporary permission authorization (valid for ≤ 4 hours), minimizes permissions through a two-level approval process, and improves the interception rate of abnormal permission applications.

[0204] 3. Data security technology solutions:

[0205] Transmission encryption: Using TLS1.3 protocol (AES-256-GCM encryption, HMAC-SHA384 integrity check) to ensure data transmission security;

[0206] Storage security: Sensitive data is processed using SHA-512 hashing and salting (64-bit salt value), data repository field-level encryption (e.g., longitude and latitude fields are encrypted separately), key operation logs are stored on read-only media (supporting blockchain evidence storage), and query response time is ≤ 2 seconds.

[0207] 4. Operational safety assurance:

[0208] A session timeout mechanism (30 minutes for operators, 60 minutes for observers) automatically logs out, and remote logins trigger risk alerts (SMS notification + verification code confirmation). Machine learning-based detection of abnormal behavior (such as high-frequency command operations) blocks suspicious operations in real time and logs them, ensuring a response time of less than 10 seconds.

[0209] Through the above design, the security management module 1100 realizes full-process security protection from user authentication to data operation, meets the high security level requirements of aviation, agriculture and other fields, improves the compliance rate of key data access, and raises the system security level to the industry-leading level.

[0210] Those skilled in the art will appreciate that the process of implementing all or part of the steps of the above embodiments may be accomplished by hardware, or by instructing related hardware through a program, which may be stored in a computer-readable storage medium.

[0211] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An integrated platform for bird repelling and drone dispatching based on big data analysis, characterized by: include: A multi-source data acquisition module (100) is used to integrate meteorological data, geographic information data, historical bird activity data, and real-time monitoring data, wherein the real-time monitoring data includes bird activity tracks monitored by radar and population classification information identified by voiceprint; A data fusion processing module (200) processes heterogeneous data sources using a spatiotemporal feature extraction algorithm to generate a spatiotemporal-correlated bird density distribution heat map; A dynamic scheduling decision module (300) generates a multi-UAV collaborative path planning scheme based on the heat map and UAV state parameters using an improved ant colony algorithm; A collaborative control module (400) controls the UAV to perform acoustic, light, or physical driving tasks according to a path planning scheme, and adjusts flight parameters in real time to respond to environmental changes; Communication module (500), responsible for data transmission between modules and communication between the UAV and the platform; A user interaction module (600) provides an operation interface for the user; A database (700) for storing various collected data, bird activity model parameters, drone information, and scheduling strategy history records; A federated learning optimization module (800) is used to aggregate bird behavior data collected from multiple regions to update prediction model parameters, adapt the general model to a specific geographical environment through transfer learning, and build a virtual training environment based on digital twin technology to verify the model generalization ability; Energy management module (900), used to predict the total flight time of the drone group, generate a dynamic charging scheduling plan based on the distribution of charging piles, introduce energy consumption weight factors into path planning, give priority to low-power cruise mode, and optimize task allocation according to the battery degradation model to reduce high-energy consumption routes; The emergency processing module (1000) is used to activate a multi-UAV formation to form a three-dimensional repelling barrier when a sudden bird flock is detected, generate a dynamic interception path based on reinforcement learning, trigger an airspace emergency avoidance protocol, and automatically switch to a ground bird repelling equipment coordination mode under extreme weather conditions; The security management module (1100) is used to authenticate user logins, set operational permissions for different users, and ensure platform data security and operational security.

2. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The multi-source data acquisition module (100) comprises: A dual-mode monitoring device, integrating millimeter-wave radar and infrared camera, collects real-time data on bird flight altitude, speed, and flock density; Voiceprint recognition unit, which identifies high-risk bird species through spectrum analysis and marks the threat level; The meteorological data interface can obtain the precipitation probability, wind speed gradient and temperature change curve of the target area in real time; and a distributed sensor network is used for data collection to obtain multi-dimensional data information, and collect bird activity data, environmental data and drone status data in the target area in real time.

3. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The working method of the data fusion processing module (200) comprises the following steps: S1. Use spatiotemporal grid modeling technology to divide the target area into dynamic geographic grids; S2, dynamically weighting meteorological data and bird activity characteristics based on the attention mechanism to generate a multi-dimensional fusion feature vector; S3, extracting periodic activity patterns from historical data through a sliding time window algorithm; S4. Use machine learning algorithms to train historical bird activity data and environmental data to establish a bird activity model and evaluate the priority of bird repellent needs in different regions.

4. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The dynamic scheduling decision module (300) includes: Priority assessment unit, which generates a task priority sequence based on bird threat level, drone endurance, and airspace control rules; The path optimization unit uses an improved ant colony algorithm with obstacle avoidance constraints to calculate the energy-optimal path for each UAV; The redundant scheduling unit presets backup routes and emergency drone activation strategies; and formulates drone scheduling strategies based on the results of the big data analysis module, comprehensively considering the drone's current location, battery power, remaining bird-repelling material volume, and the priority of bird-repelling demand in the target area, to determine the number of drones to perform bird-repelling tasks, flight routes, bird-repelling methods, and task execution time, and give priority to dispatching drones that meet the conditions and are closer to the target area to perform tasks.

5. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 4 is characterized by: In the path optimization unit, the improved ant colony algorithm with obstacle avoidance constraint is implemented by the heuristic function Optimize the path search process: , Indicates the current location of the drone With target location The distance between and is the weight coefficient, , used to balance the influence of distance factors and the threat level of birds in the target area on path selection; Indicates the target location The bird threat level value at Indicates the maximum bird threat level value among all target areas.

6. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The collaborative control module (400) includes: Multi-machine communication protocol supports state synchronization and command distribution of heterogeneous drone groups; Dynamic parameter adjustment unit, adaptively adjusts the drone's flight altitude and bird-repelling intensity based on real-time wind speed data; The task switching unit automatically switches to light deterrence mode when the sonic repelling fails; The bird repellent execution module is installed on the UAV, and the bird repellent methods include at least sonic bird repellent, chemical bird repellent and optical bird repellent.

7. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The user interaction module (600) includes: The 3D visualization interface uses high-precision 3D modeling and real-time rendering technology to update the drone's position in real time. It also supports user interaction. Clicking the drone icon can check key parameters. Key parameters include: The manual intervention interface provides a task priority list that users can adjust by dragging or entering values, and the system will recalculate the schedule in real time. Users can take over a specific drone and use preset commands to control flight and bird repelling. A multi-level early warning system divides threat levels into yellow and red warnings based on multi-dimensional bird data and preset algorithms. When an alert is triggered, the interface flashes and an audible and visual alarm is accompanied, and a window pops up with alert details and response suggestions. Data display and operation, using a variety of charts to dynamically display relevant data.

8. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The federated learning optimization module (800) includes: Data aggregation and updating unit: uses secure multi-party computing to encrypt and aggregate bird behavior data from multiple regions, and uses an adaptive weighting strategy to determine the weight of data in each region; Transfer learning adaptation unit: uses feature extraction and mapping technology to identify key features of the general model and map them to the feature space of a specific geographical environment; Digital twin verification unit: Builds a highly simulated virtual training environment based on multi-source data, simulating various scenarios to generate test sets; The evaluation feedback unit regularly evaluates indicators such as model accuracy, feeds the results back to relevant sub-modules, and adjusts strategies, parameters, and structures.

9. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The energy management module (900) includes: Endurance prediction unit: By analyzing the drone's historical flight data, battery capacity, and real-time load weight, it establishes an endurance prediction model and outputs the remaining endurance of the drone swarm in different mission scenarios; Charging scheduling unit: Combines the geographical distribution data of charging piles and the real-time power of drones, uses the shortest path algorithm to generate a dynamic charging scheduling plan, and prioritizes dispatching drones with power levels below 20% to the nearest charging pile for charging; Energy consumption optimization unit: Introduces energy consumption weight parameters into the path planning algorithm, presets energy consumption coefficients for different flight modes based on the drone model, and automatically selects low-power cruise mode for non-emergency tasks; Task allocation unit: A loss model is established based on historical battery attenuation data. When allocating tasks, low-energy consumption routes are prioritized to drones with a battery attenuation rate exceeding 30%, thereby balancing the overall energy consumption of the fleet.

10. The integrated bird-repelling and drone dispatching platform based on big data analysis according to claim 1 is characterized in that: The emergency processing module (1000) comprises: Cluster monitoring unit: By analyzing the bird density and movement speed in radar monitoring data, it can determine whether there is a sudden bird cluster; Formation response unit: Launch three or more drones to form a V-shaped formation, and use the sonic repelling device to synchronously emit 15-25kHz pulse sound waves to form a three-dimensional repelling barrier; Obstacle avoidance control unit: When the distance between the drone and the bird is detected to be less than 50 meters, the obstacle avoidance program is automatically triggered and the flight altitude and heading are adjusted using the preset safety avoidance strategy; Equipment switching unit: When the meteorological sensor detects that the wind speed exceeds level 6 or the rainfall exceeds 20mm / h, it automatically switches to the fixed acoustic bird repellent device deployed on the ground and recalls all drones to a safe area.

Images