Visual target detection and identification system for unmanned aerial vehicle group

Abstract

The invention discloses an unmanned aerial vehicle group visual target detection and identification system, which comprises an unmanned aerial vehicle group, a ground control station and a central server, and is characterized in that the unmanned aerial vehicle group is composed of a plurality of unmanned aerial vehicles and is used for collecting real-time data of a task area; the ground control station is used for receiving real-time data transmitted by the unmanned aerial vehicle, analyzing task requirements, calculating an optimal scheduling scheme of the unmanned aerial vehicle and monitoring the state of the unmanned aerial vehicle; the central server is used for data processing, unmanned aerial vehicle scheduling decision making and reinforcement learning model training. According to the method, a task demand matrix is constructed from a task target, regional characteristics and monitoring requirements; and meanwhile, task requirements are matched by utilizing an unmanned aerial vehicle capability matrix, and task allocation is dynamically adjusted through real-time environment feedback. On this basis, a reinforcement learning algorithm is introduced to continuously optimize a scheduling strategy, and it is ensured that the system can flexibly and efficiently complete tasks in various complex environments.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to a visual target detection and recognition system for a group of unmanned aerial vehicles. Background Art

[0002] Currently, the use of drones for monitoring has shown significant advantages in many fields, especially in the monitoring of wide-area, large-scale targets. The flexibility, cost-effectiveness and real-time response capabilities of drones make them an ideal choice to replace traditional monitoring methods.

[0003] However, the existing technology still has shortcomings. For targets with large volume or range (such as large buildings, transportation facilities, vast farmland or forest fires), a single drone cannot fully cover the target area. In this case, it is necessary for multiple drones to detect and identify targets in real time from different angles and heights to ensure comprehensive observation of the target. However, when multiple drones are used for collaborative detection and identification, the existing drones are often managed according to the predetermined task requirement matrix, which can easily lead to system bloat and waste of resources. Summary of the invention

[0004] The purpose of the present invention is to provide a drone swarm visual target detection and recognition system to solve the problems raised in the above background technology.

[0005] To achieve the above-mentioned purpose of the invention, the present invention provides a drone swarm visual target detection and recognition system, including a drone swarm, a ground control station, and a central server, wherein:

[0006] A drone swarm consists of several drones that collect real-time data in the mission area;

[0007] The ground control station is used to receive real-time data transmitted by the UAV, analyze mission requirements, calculate the best scheduling plan for the UAV, and monitor the status of the UAV;

[0008] The central server is responsible for data processing, drone scheduling decisions, and reinforcement learning model training.

[0009] Furthermore, the drone is configured with a flight controller, a multi-sensor system, a communication module, and a battery management system, wherein the multi-sensor system is equipped with sensors including a visible light camera, an infrared camera, and a lidar for target recognition, real-time monitoring, and environmental data collection.

[0010] Furthermore, the ground control station includes a mission requirement analysis module, a UAV capability analysis module, a mission requirement and capability matching module and an environment feedback module, wherein:

[0011] The mission requirement analysis module is used to process the mission requirement data from the mission command center or the automation system, extract the mission requirements from the mission objectives, regional characteristics and monitoring requirements, and generate a mission requirement matrix;

[0012] The UAV capability analysis module monitors the operating status of the UAV in real time and feeds it back to the dispatch system, generating a UAV capability matrix based on the flight time, load capacity, and sensor accuracy of the UAV;

[0013] The mission requirement and capability matching module is used to match the mission requirements with the UAV capabilities and calculate the number of UAVs required in the mission area;

[0014] The environmental feedback module obtains real-time information about environmental changes through drone sensors and external data sources, and dynamically adjusts mission requirements and drone capabilities.

[0015] Furthermore, the central server includes a dynamic allocation module and a scheduling optimization module, wherein:

[0016] The dynamic allocation module dynamically allocates drones to perform tasks based on updated mission requirements and drone capabilities;

[0017] The scheduling optimization module optimizes the UAV scheduling strategy through reinforcement learning algorithm to improve the scheduling efficiency of the system.

[0018] Furthermore, the task requirement matrix is ​​represented as T(i, j), and its construction method includes the following steps:

[0019] Step S101, obtaining task area characteristics, including area A i , Terrain complexity C i , and environmental impact E;

[0020] Step S102: Obtain monitoring requirements, including accuracy P, real-time performance R, and coverage C r ;

[0021] Step S103, converting the combination of task objectives, regional characteristics and monitoring requirements into specific values ​​in the task requirement matrix. The element calculation formula of the task requirement matrix is ​​as follows:

[0022] T ij =w1·A i +w2·C t +w3·E+w4·P j +w5·R j +w6·C r ;

[0023] w1 to w6 are the area A of the region i , Terrain complexity C i, environmental impact E, accuracy P, real-time R and coverage C r The corresponding weight coefficient indicates the importance of each factor.

[0024] Furthermore, the task requirement and capability matching module calculates the number of drones required for each task area based on the task requirement matrix and the drone capability matrix, and dynamically allocates them in real time;

[0025] The drone capability matrix is ​​represented as C(j, k). The performance parameters of each drone are obtained through the data provided by the drone manufacturer and the drone's own sensors, and then the comprehensive performance coefficient is weighted to determine the drone capability matrix;

[0026] Calculate the number of drones N required for each mission area i The calculation formula is

[0027] In the formula, ∑ j T ij is the total amount of task requirements, summed for each requirement type j, ∑ k C jk The sum of the UAV capabilities.

[0028] Furthermore, the environmental feedback module obtains real-time environmental information through the drone's sensors and dynamically adjusts the mission requirements and drone capabilities; the updated mission requirement matrix and drone capability matrix are used to recalculate the allocation; specifically, the following steps are included:

[0029] Step S201, obtaining real-time environmental feedback information through the drone's sensors and other external data sources, including: target movement status;

[0030] changes in weather conditions, including changes in wind speed, rainfall and other environmental factors;

[0031] Battery consumption;

[0032] Step S202: Update the task requirement matrix T(i, j) and the drone capability matrix C(j, k) according to the feedback information; it can be expressed as the following formula:

[0033] T′ ij =T ij +ΔT ij , C′ jk =C jk +ΔC jk ;

[0034] Where, T ij is the task requirement, ΔT ij The reason for the change in task requirements; C jk is the UAV capability, ΔC jkis the change in the UAV’s capabilities.

[0035] Furthermore, the dynamic allocation module dynamically reallocates drones according to the updated mission requirements and drone capabilities, and updates the drone status information; the number of drones D allocated to each mission area i is determined by adjusting the allocation matrix D(i, j) ij ; The calculation formula is: D ij =min(N i , available_drones(j));

[0036] Where available_drones(j) represents the number of currently available drones.

[0037] Furthermore, the dynamic allocation module updates the status of each drone in the system in real time, including battery status and flight time: records the flight time of the drone, and switches drones or arranges drone rotation when necessary; the calculation formula for the performance status of the current drone is: jk =S jk -ΔS jk ;

[0038] In the formula, S jk is the state of performance parameter k of UAV j during mission execution, ΔS jk Represents the consumption of the drone in the current mission cycle.

[0039] Furthermore, the scheduling optimization module performs the following steps to optimize the scheduling of drones:

[0040] Step S301, design a reward function, and calculate the reward function RR according to the task success rate and resource usage cost; the calculation formula is: RR = α*success_rate-β*cost;

[0041] In the formula, α and β are adjustment factors that control the impact of success rate and cost on rewards. Success_rate is the success rate of task completion, which indicates the percentage of task goal completion. Cost is the resource usage cost, including the energy consumption of the drone and the task execution time.

[0042] Step S302, reinforcement learning process, takes the current task requirement matrix and drone capability matrix as input states, selects actions including dispatching the number of drones, reallocating drones in the task area, changing the flight path of drones, and obtaining rewards RR update strategy according to the task completion situation

[0043] Compared with the prior art, the present system and method have the following advantages:

[0044] The present invention constructs a task requirement matrix based on task objectives, regional characteristics and monitoring requirements; at the same time, it uses the UAV capability matrix to match task requirements and dynamically adjusts task allocation through real-time environmental feedback.

[0045] The present invention introduces a reinforcement learning algorithm to continuously optimize the scheduling strategy, ensuring that the system can complete tasks flexibly and efficiently in various complex environments.

[0046] The present invention ensures that the mission continues by dynamically allocating drones and updating their status in real time, and drones can be adjusted and supplemented in time according to demand, which greatly improves the flexibility and accuracy of mission execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a schematic diagram of the structure of a drone swarm visual target detection and recognition system.

[0048] Figure 2 This is a working diagram of a drone swarm visual target detection and recognition system. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The present invention calculates the number of drones required by initializing the drone mission requirement matrix and capability matrix, and dynamically adjusts the task allocation in combination with real-time environmental feedback. During the drone mission execution process, the system continuously optimizes resource allocation based on feedback, and improves the scheduling strategy through reinforcement learning to achieve flexible regulation of the number of drones and efficient execution of tasks. The system dynamically adjusts the number of drones to meet the needs of different mission areas, optimizes drone resource allocation, and improves the efficiency and accuracy of task completion.

[0051] like Figure 1 The structure diagram of the present invention is shown. The present invention includes the following modules:

[0052] Drone swarm: Carrying a variety of sensors (such as visible light cameras, infrared cameras, and lidar), the drones are able to collect real-time data in the mission area. The drones are equipped with flight controllers, communication modules (such as 4G / 5G or dedicated communication networks), navigation systems (GPS / Beidou), and battery monitoring systems.

[0053] Ground control station: used to receive real-time data transmitted by drones, analyze mission requirements, and calculate the best scheduling plan for drones. The ground station is also used to monitor the status of drones (such as battery power, location, flight time, etc.).

[0054] Central server: responsible for data processing, drone scheduling decisions and reinforcement learning model training. The server is connected to the ground control station and drones through a communication network.

[0055] Among them, the drone hardware configuration includes:

[0056] Flight controller: responsible for the flight attitude and position control of the drone, supporting path planning, obstacle avoidance and precise positioning.

[0057] Multi-sensor system: The drone is equipped with multiple sensors such as visible light cameras, infrared cameras, and lidar to meet different mission requirements. These sensors can be used for target recognition, real-time monitoring, and environmental data collection.

[0058] Communication module: Through 4G, 5G or dedicated communication networks, drones can maintain real-time communication with ground control stations and other drones to ensure real-time transmission and sharing of mission data.

[0059] Battery Management System: monitors the battery level of the drone in real time to ensure that the drone can return to the charging station or be replaced safely, ensuring that the mission is not interrupted.

[0060] The ground control station configuration includes:

[0061] Mission Requirements Analysis Module: Processes mission requirements data from the mission command center or automated system.

[0062] Drone capability analysis module: monitors the drone’s flight status, battery power, sensor operating status, etc. in real time, and feeds back to the dispatch system.

[0063] Mission requirement and capability matching module: matches mission requirements with drone capabilities and calculates the number of drones required in the mission area.

[0064] Environmental feedback module: Receives environmental change information from drone sensors, ground sensors, and external weather data sources (such as weather stations).

[0065] The central server configuration includes:

[0066] Dynamic allocation module: Assign tasks to each drone and re-plan the path according to real-time environmental changes to ensure that the drone completes the task efficiently.

[0067] Scheduling optimization module: Train the scheduling optimization model through historical data and real-time feedback to gradually improve the utilization efficiency of drone resources.

[0068] like Figure 2 The system working diagram of the present invention is shown. The core functional modules of the system have the following specific functions:

[0069] Mission requirement analysis module: extracts mission requirements from mission objectives, regional characteristics and monitoring requirements, and generates a mission requirement matrix T(i, j);

[0070] UAV capability analysis module: Generates the UAV capability matrix C(j, k) based on the UAV’s flight time, load capacity, sensor accuracy and other information;

[0071] Mission requirements and capability matching module: matches mission requirements with drone capabilities and calculates the number of drones required in the mission area;

[0072] Environmental feedback module: obtains real-time information about environmental changes through drone sensors and external data sources, and dynamically adjusts mission requirements and drone capabilities;

[0073] Dynamic allocation module: dynamically allocates drones to perform tasks based on updated mission requirements and drone capabilities;

[0074] Scheduling optimization module: Optimize the drone scheduling strategy through reinforcement learning algorithm to improve the scheduling efficiency of the system.

[0075] The construction of the task requirement matrix T(i, j) of the present invention is based on the characteristics of the task area, the task type and the monitoring requirements. The task area i can be a building, a traffic section, a farmland, a forest fire area, etc. The task type j includes monitoring, identification, search, data collection, etc. The method for constructing the task requirement matrix T(i, j) includes the following steps:

[0076] Step S101, obtaining the characteristics of the mission area. When constructing the mission requirement matrix, the characteristics of the mission area determine the demand for drone resources in different areas. To this end, the system collects the following detailed data through drones or other sensor devices:

[0077] The area Ai affects the range that the drone needs to fly. If the monitoring area is large, more drones are needed to cover it.

[0078] Terrain complexity Ci: Complex terrain (such as mountains and densely populated urban areas) will create obstacles for drone flight and monitoring, requiring more drones to cover different angles.

[0079] Environmental influences E include weather (such as wind speed and rainfall), obstacles (such as trees and buildings), etc. These factors may affect the monitoring effect and path planning of the drone.

[0080] It should be noted that the area can be extracted from geographic information system (GIS) data. The terrain complexity and environmental impact can be obtained through historical meteorological data, satellite images or lidar scanning data.

[0081] Step S201, obtain monitoring requirements. Monitoring requirements define the responsiveness requirements of the UAV to the task. Different tasks have different monitoring requirements; including:

[0082] Precision P, for example, millimeter-level precision is required in building inspection, while farmland monitoring may only require lower precision;

[0083] Real-time performance R, time-sensitive tasks (such as traffic accident monitoring or natural disaster response) require high real-time performance, while routine inspections can accept longer delays.

[0084] Coverage rate Cr, some tasks require full coverage, such as disaster emergency monitoring; while other tasks (such as low-risk area monitoring) can accept partial coverage.

[0085] It should be noted that the monitoring requirements are determined based on actual conditions, usually by the task definer (such as project managers, operators) based on the task objectives, and will not be elaborated here.

[0086] Step S103, converting the combination of mission objectives, regional characteristics and monitoring requirements into specific values ​​in the mission requirement matrix. Since the mission requirement matrix T (i, j) represents the requirements of mission area i for drones in mission type j, the present invention is constructed by converting the combination of mission objectives, regional characteristics and monitoring requirements into specific values ​​in the mission requirement matrix. Assume that there are m mission areas, each with different demand types (such as monitoring, identification, etc.), and the mission requirement matrix T (i, j) is an m*n matrix, where n is the number of mission demand types. Each mission area i corresponds to a row in the matrix. These mission areas can be different geographical locations or mission partitions, such as multiple buildings, farmlands, traffic sections or disaster monitoring areas, etc.; each mission demand j corresponds to a column in the matrix. Mission requirement type refers to the different requirements that need to be met in each mission area, such as monitoring accuracy, coverage, real-time performance, etc. Each column j corresponds to a specific mission requirement type. The specific matrix element calculation formula is as follows:

[0087] T ij =w1·A i +w2·C t +w3·E+w4·P j +w5·R j +w6·C r ;

[0088] w1 to w6 are the corresponding weight coefficients of regional area Ai, terrain complexity Ci, environmental impact E, accuracy P, real-time R, and coverage Cr, respectively, indicating the importance of each factor. These weights are adjusted through experience or historical data. For example, for disaster relief tasks, the weight w5 of real-time R is higher than that of other tasks.

[0089] Here are some examples:

[0090] Suppose we have a building inspection task, area A1 is 10 square kilometers, the terrain complexity is medium (value 5), the environment is harsh (high wind speed, many obstacles, value 8), high accuracy (millimeter level, value 10) and real-time response (value 9) are required, and 100% coverage (value 10) is required. Then:

[0091] T ij =0.5·10+0.2·5+0.1·8+0.3·10+0.4·9+0.6·10=21.2;

[0092] That is, the value of the task requirement matrix is ​​21.2 at this time.

[0093] The process of generating the drone capability matrix by the drone capability analysis module is as follows:

[0094] First, obtain the performance parameters of the drone. The performance parameters of each drone are obtained through the data provided by the drone manufacturer and the drone's own sensors, including flight time, load capacity, sensor resolution, communication capability, and flight altitude.

[0095] Then, the comprehensive performance coefficient is weighted and calculated to determine the UAV capability matrix value. The calculation method is similar to the above-mentioned task requirement matrix and will not be repeated here.

[0096] The mission requirement and capability matching module calculates the number of drones required for each mission area based on the mission requirement matrix and the drone capability matrix, and dynamically allocates them in real time.

[0097] The calculation formula for the number of drones Ni required for each mission area is expressed as Obviously, the ratio of mission requirements to drone capabilities determines the number of drones.

[0098] In the formula, ∑ j T ij is the total amount of task requirements, summed for each requirement type j, ∑ k C jk The sum of the UAV capabilities.

[0099] When mission requirements increase (such as larger areas, higher precision requirements), the number of drones required also increases. When drone capabilities improve (such as longer endurance, higher sensor accuracy), the number of drones required decreases.

[0100] For example: the total demand for mission area i=1 is 500, and the sum of the drone capability matrix is ​​100, so the number of drones required is 5.

[0101] The real-time environmental information is obtained through the drone's sensors, and the mission requirements and drone capabilities are dynamically adjusted. The updated mission requirement matrix T(i, j) and drone capability matrix C(j, k) are used to recalculate the allocation. The following steps are included:

[0102] Step S201, through the drone's sensors and other external data sources (such as weather stations, ground sensors), we obtain real-time environmental feedback information, including:

[0103] Target movement: If the target being monitored is moving, this will change the requirements of the mission area;

[0104] Changing weather conditions: Changes in wind speed, rainfall, or other environmental factors can affect the drone’s flight capabilities or sensor performance;

[0105] Battery consumption: As the mission progresses, the drone’s battery power will decrease, affecting its mission execution time;

[0106] Step S202: Based on the feedback information, we update the task requirement matrix T(i, j) and the drone capability matrix C(j, k); the specific calculation expression can be: T′ ij =T ij +ΔT ij , C′ jk =C jk +ΔC jk ;

[0107] Where, ΔT ij The reasons for the change in task requirements may be changes in the target location (such as the vehicle leaving a certain area), the extension of the monitoring time, etc. jk The amount of change in the drone's capabilities (such as battery consumption resulting in a reduction in flight time, or environmental impact causing sensor performance to degrade).

[0108] It should be noted that ΔT ij and ΔC jk They are all affected by multiple real-time factors, with ΔT ij For example, the calculation expression can be: ΔT ij=w7·ΔE+w8·ΔP+w9·ΔM, w7-w9 represents the influence of environmental changes, task progress and target movement on task requirements, ΔE represents the impact of environmental changes on task requirements. For example, an increase in wind speed or rainfall will reduce the flight efficiency of the drone, resulting in an increase in demand; ΔP represents task progress. If a task has been partially completed, the demand will decrease accordingly; ΔM represents the impact of target movement. For example, if the target leaves the monitoring area, the flight path of the drone needs to be readjusted, resulting in a change in demand.

[0109] For example, suppose that the mission area i is affected by the environmental change ΔE, and the wind speed increases from the initial 10km / h to 20km / h, making the mission execution more difficult. At this time:

[0110] Increased wind speed may affect the stability and monitoring accuracy of drones, requiring the dispatch of more drones.

[0111] If every 10 km / h increase in wind speed results in a 10% increase in demand, and w7 is 0.2, then: ΔT ij =0.2×(20-10)=2; that is, T(i, j) in the task requirement matrix is ​​updated to: T′ ij =T ij +2.

[0112] Dynamic Allocation Module Based on the updated mission requirements and drone capabilities, we need to dynamically reallocate drones to ensure that mission requirements are met and update drone status information (such as battery power, flight time, etc.) at the same time. The process is as follows:

[0113] The number of drones D assigned to each mission area i is determined by adjusting the allocation matrix D(i, j) ij ; The calculation expression can be: D ij =min(N i , available_drones(j));

[0114] Where available_drones(j) represents the number of currently available drones.

[0115] During the execution of each mission, the status of the drone will change (such as battery power decreases, mission progress advances). Therefore, we need to update the status of each drone in the system in real time, including:

[0116] Battery Status: Track battery charge in real time to ensure the drone can complete its mission or return to a charging station before it runs out of power.

[0117] Flight time: Record the flight time of the drone and switch drones or arrange drone rotation when necessary. The calculation expression can be: S′jk =S jk -ΔS jk ;

[0118] In the formula, S jk is the state of performance parameter k of drone j during mission execution (such as battery power, flight time), ΔS jk Represents the consumption of the drone in the current mission cycle.

[0119] In order to further optimize the scheduling and allocation of drones, this paper introduces reinforcement learning such as Q-learning or PPO (Proximal Policy Optimization) to dynamically learn the optimal scheduling strategy. Through system feedback and mission history data, the model can continuously improve the efficiency of drone resource allocation.

[0120] Specifically, the scheduling optimization module UAV scheduling strategy includes the following steps:

[0121] Step S301, design a reward function. The core of reinforcement learning is to define a reasonable reward function to measure the success rate and efficiency of task execution. The reward function RR can be defined based on the task success rate and resource usage cost. The specific calculation expression is: RR = α*success_rate-β*cost; where α and β are adjustment factors that control the impact of success rate and cost on rewards; success_rate is the success rate of task completion, indicating the percentage of task goal completion, and cost is the resource usage cost, including the energy consumption of the drone, task execution time, etc.

[0122] Step S302, reinforcement learning process. The current task requirement matrix and drone capability matrix are used as input states, and the selected actions include scheduling the number of drones, reallocating drones in the task area, changing the flight path of drones, etc., and the reward RR is obtained according to the completion of the task to update the strategy.

[0123] It should be noted that the success rate of the present invention is the task success rate, which is the positive feedback of the system and reflects the completion of the task goal. It can be evaluated according to the degree of achievement of the task requirements. The specific calculation method can be as follows:

[0124] In the formula, completed_objectives is the number of sub-objectives that have been completed in the task. Sub-objectives can be the coverage of the monitoring area, the number of cruises completed, the number of targets identified, etc.; total_objectives is the total number of all targets in the task.

[0125] For example, suppose the drone needs to monitor 5 key locations in an area and has already covered 3 locations. Then:

[0126] At the same time, in a multi-task scenario, the success rate of the task can be weighted averaged according to the importance of the task, which will not be elaborated here.

[0127] The cost of the present invention mainly measures the resources consumed by the system to complete the task. These resource consumptions may include:

[0128] Energy consumption: The usage of the drone’s battery.

[0129] Time consumption: The length of time it takes to execute a task, especially in scenarios with high real-time requirements.

[0130] Number of drones: The more drones involved in the mission, the higher the cost.

[0131] Communication resources: data transmission volume between drones and ground stations, bandwidth usage, etc.

[0132] The total cost can usually be expressed as the weighted sum of the consumption of each resource, which will not be elaborated here.

[0133] Through the reinforcement learning algorithm, the UAV dispatching system can continuously optimize task allocation and resource utilization strategies as the task execution progresses. This learning mechanism can gradually improve the dispatching efficiency in different tasks and environments, and achieve the optimal number of UAVs for dynamic control and efficient task execution.

[0134] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0135] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0136] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0137] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0138] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A drone swarm visual target detection and recognition system, characterized in that: It includes drone swarm, ground control station, and central server, where: A drone swarm consists of several drones that collect real-time data in the mission area; The ground control station is used to receive real-time data transmitted by the UAV, analyze mission requirements, calculate the best scheduling plan for the UAV, and monitor the status of the UAV; The central server is responsible for data processing, drone scheduling decisions, and reinforcement learning model training.

2. The drone swarm visual target detection and recognition system according to claim 1, characterized in that: The drone is equipped with a flight controller, a multi-sensor system, a communication module, and a battery management system. The multi-sensor system is equipped with sensors such as a visible light camera, an infrared camera, and a lidar for target recognition, real-time monitoring, and environmental data collection.

3. The drone swarm visual target detection and recognition system according to claim 1, characterized in that: The ground control station includes a mission requirement analysis module, a UAV capability analysis module, a mission requirement and capability matching module and an environment feedback module, wherein: The mission requirement analysis module is used to process the mission requirement data from the mission command center or the automation system, extract the mission requirements from the mission objectives, regional characteristics and monitoring requirements, and generate a mission requirement matrix; The UAV capability analysis module monitors the operating status of the UAV in real time and feeds it back to the dispatch system, generating a UAV capability matrix based on the flight time, load capacity, and sensor accuracy of the UAV; The mission requirement and capability matching module is used to match the mission requirements with the UAV capabilities and calculate the number of UAVs required in the mission area; The environmental feedback module obtains real-time information about environmental changes through drone sensors and external data sources, and dynamically adjusts mission requirements and drone capabilities.

4. The drone swarm visual target detection and recognition system according to claim 1, characterized in that: The central server includes a dynamic allocation module and a scheduling optimization module, wherein: The dynamic allocation module dynamically allocates drones to perform tasks based on updated mission requirements and drone capabilities; The scheduling optimization module optimizes the UAV scheduling strategy through reinforcement learning algorithm to improve the scheduling efficiency of the system.

5. The drone swarm visual target detection and recognition system according to claim 3, characterized in that: The task requirement matrix is ​​represented as T(i, j), and its construction method includes the following steps: Step S101, obtaining task area characteristics, including area A i , Terrain complexity C i , and environmental impact E; Step S102: Obtain monitoring requirements, including accuracy P, real-time performance R, and coverage C r ; Step S103, converting the combination of task objectives, regional characteristics and monitoring requirements into specific values ​​in the task requirement matrix. The element calculation formula of the task requirement matrix is ​​as follows: T ij =w1·A i +w2·C t +w3·E+w4·P j +w5·R j +w6·C r ; w1 to w6 are the area A of the region i , Terrain complexity C i , environmental impact E, accuracy P, real-time R and coverage C r The corresponding weight coefficient indicates the importance of each factor.

6. The drone swarm visual target detection and recognition system according to claim 5, characterized in that: The task demand and capability matching module calculates the number of drones required for each task area based on the task demand matrix and drone capability matrix, and dynamically allocates them in real time; The drone capability matrix is ​​represented as C(j, k). The performance parameters of each drone are obtained through the data provided by the drone manufacturer and the drone's own sensors, and then the comprehensive performance coefficient is weighted to determine the drone capability matrix; The calculation formula for the number of drones Ni required for each mission area is: In the formula, ∑ j T ij is the total amount of task requirements, summed for each requirement type j, ∑ k C jk The sum of the UAV capabilities.

7. The drone swarm visual target detection and recognition system according to claim 3 is characterized in that: The environmental feedback module obtains real-time environmental information through the drone's sensors and dynamically adjusts the mission requirements and drone capabilities. The updated mission requirement matrix and drone capability matrix are used to recalculate the allocation. The following steps are included: Step S201, obtaining real-time environmental feedback information through the drone's sensors and other external data sources, including: target movement status; changes in weather conditions, including changes in wind speed, rainfall and other environmental factors; Battery consumption; Step S202: Update the task requirement matrix T(i, j) and the drone capability matrix C(j, k) according to the feedback information; it can be expressed as the following formula: T′ ij =T ij +ΔT ij ,C′ jk =C jk +ΔC jk ; Where, T ij is the task requirement, ΔT ij The reason for the change in task requirements; C jk is the UAV capability, ΔC jk is the change in the UAV’s capabilities.

8. The drone swarm visual target detection and recognition system according to claim 4, characterized in that: The dynamic allocation module dynamically reallocates drones according to the updated mission requirements and drone capabilities, and updates the drone status information at the same time; the number of drones D allocated to each mission area i is determined by adjusting the allocation matrix D(i, j) ij ; The calculation formula is: D ij =min(N i , available_drones(j)); Where available_drones(j) represents the number of currently available drones.

9. The drone swarm visual target detection and recognition system according to claim 8, characterized in that: The dynamic allocation module updates the status of each drone in the system in real time, including battery status and flight time: records the flight time of the drone, and switches drones or arranges drone rotation when necessary; the calculation formula for the current drone performance status is: S′ jk =S jk -ΔS jk ; In the formula, S jk is the state of performance parameter k of UAV j during mission execution, ΔS jk Represents the consumption of the drone in the current mission cycle.

10. The drone swarm visual target detection and recognition system according to claim 1, characterized in that: The scheduling optimization module's strategy for optimizing drone scheduling includes the following steps: Step S301, design a reward function, and calculate the reward function RR according to the task success rate and resource usage cost, and the calculation formula is: RR = α*success_rate-β*cost; In the formula, α and β are adjustment factors that control the impact of success rate and cost on rewards. Success_rate is the success rate of task completion, which indicates the percentage of task goal completion. Cost is the resource usage cost, including the energy consumption of the drone and the task execution time. Step S302, reinforcement learning process, takes the current task requirement matrix and drone capability matrix as input states, and the selected actions include scheduling the number of drones, reallocating drones in the task area, changing the flight path of drones, and obtaining rewards RR update strategies based on the task completion status.

Images