Low-altitude operation intelligent networking method and system for integrated command and meteorological support

Abstract

The present invention relates to the field of communication networking technology, specifically to a low-altitude operation intelligent networking method and system for integrated command and meteorological support, including obtaining a feature matrix and networking task information of a micro-airport, compressing the feature matrix of the micro-airport to obtain a low-dimensional feature matrix, extracting feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set, obtaining a drone combination sequence through the operation type, operation flight altitude parameter and operation flight speed parameter in the privacy protection coding of the micro-airport, the corresponding task type parameter and the overlapping area of ​​the task coverage rectangle, obtaining a spectrum interference matrix through the interference degree between the communication frequency band parameters in the drone combination sequence, obtaining a communication grouping scheme through the spectrum interference matrix, and obtaining a communication channel allocation sequence through the communication grouping scheme. The present invention performs optimal networking through cross-platform drone feature analysis intelligence, and at the same time solves the information security of the cross-platform drone networking process.

Description

Technical Field

[0001] The present invention relates to the technical field of communication networking, and in particular to a low-altitude operation intelligent networking method and system for integrated command and meteorological support. Background Art

[0002] Drones are widely used in agriculture, environmental monitoring, logistics distribution, public safety and other fields. Some application scenarios require the deployment of a large number of drones and the use of drone clusters for operations. Therefore, for such low-altitude aircraft, it is necessary to build a micro-airport for centralized control. The micro-airport can be understood as a control platform for accommodating and centrally managing drones. The purpose is to monitor and dispatch drone cluster operations and provide comprehensive drone operation, coordination and optimization support. The drone control platform realizes real-time monitoring, task scheduling and data processing of a large number of drones, optimizes the coordination of multiple drones, improves task execution efficiency, and ensures data security. However, with the expansion of the application scale of drone control platforms, the limitations of a single platform have also emerged, such as its "limited operating scope, task complexity and coverage capability, data processing capability, resource optimization and scheduling" and other issues. At present, the connection and interaction between domestic drone control platforms are mostly achieved through data sharing and synchronization or custom communication protocols. Although the synchronization between platforms is guaranteed, data leakage and conflicts may occur. Summary of the invention

[0003] (1) Technical issues to be resolved

[0004] The purpose of the present invention is to provide a low-altitude operation intelligent networking method and system for integrated command and meteorological support, so as to solve the problem of drone networking across multiple drone platforms.

[0005] (2) Technical solution

[0006] To achieve the above objectives, on the one hand, the present invention provides a low-altitude operation intelligent networking method for integrated command and meteorological support, the method comprising:

[0007] Obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain the networking task information, wherein the networking task information includes the task area coordinates and the task type parameter; the operation type parameter and the task type parameter both include the operation type, the operation flight altitude parameter and the operation flight speed parameter;

[0008] The feature matrix of the micro-airport is compressed to obtain a low-dimensional feature matrix, and feature vectors of different granularity levels are extracted from the low-dimensional feature matrix to obtain a multi-scale feature set. The multi-scale feature set is mapped into a binary hash code through a local sensitive hash function to obtain a privacy protection code for the micro-airport. The task area coordinates are processed through the minimum enclosing rectangle algorithm to obtain a task coverage rectangle. The distance matrix is ​​obtained by calculating the Euclidean distance from the geographical coordinates of multiple micro-airports to the center point of the task coverage rectangle. The task matching matrix is ​​obtained by calculating the Hamming distance between the operation type, operation flight altitude parameter and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter. The micro-airport matching vector is obtained by normalizing the distance matrix and the task matching matrix. The candidate micro-airport set is obtained from the micro-airport matching vector through a preset matching threshold. The regional coverage matrix is ​​obtained by the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set. The operation weight coefficient is obtained by mapping the task type parameter to the preset task type dictionary. The task fitness matrix is ​​obtained by multiplying the regional coverage matrix with the operation weight coefficient.

[0009] The task fitness matrix is ​​solved by a linear programming algorithm to obtain the optimal UAV combination sequence. The spectrum interference matrix is ​​obtained by calculating the interference degree between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence. The spectrum interference matrix is ​​processed by a spectral clustering algorithm to obtain a communication grouping scheme. The communication grouping scheme is processed by a graph coloring algorithm to obtain a communication channel allocation sequence.

[0010] Furthermore, the method of compressing the feature matrix of the micro-airport to obtain a low-dimensional feature matrix includes:

[0011] Drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for:

[0012] ;

[0013] in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for:

[0014] ;

[0015] in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for:

[0016] ;

[0017] in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in .

[0018] Furthermore, the method of extracting feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set includes:

[0019] Sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is:

[0020] ;

[0021] in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is:

[0022] ;

[0023] in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The feature vector between .

[0024] Furthermore, the method of mapping the multi-scale feature set into a binary hash code to obtain the micro-airport privacy protection code by using a local sensitive hash function includes:

[0025] The feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules :

[0026] ;

[0027] in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for:

[0028] ;

[0029] in and is the pre-set public key, is a random number, It is a modular operation.

[0030] Furthermore, the method of obtaining the task matching matrix by calculating the Hamming distance between the operation type, the operation flight altitude parameter and the operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter includes:

[0031] Encoding the privacy protection of micro-airports Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes :

[0032] ;

[0033] in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix :

[0034] .

[0035] Based on the same inventive concept, on the other hand, the present invention also provides a low-altitude operation intelligent networking system for integrated command and meteorological support, the system comprising:

[0036] A networking node acquisition module is used to obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain networking task information, wherein the networking task information includes the task area coordinates and task type parameters; the operation type parameters and task type parameters both include the type of operation, the operation flight altitude parameter and the operation flight speed parameter;

[0037] A matching encryption module is used to compress the feature matrix of a micro-airport to obtain a low-dimensional feature matrix, extract feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set, and map the multi-scale feature set into a binary hash code through a local sensitive hash function to obtain a privacy protection code for a micro-airport; process the task area coordinates through a minimum enclosing rectangle algorithm to obtain a task coverage rectangle, calculate the Euclidean distance from multiple micro-airport geographic coordinates to the center point of the task coverage rectangle to obtain a distance matrix, calculate the Hamming distance between the operation type, operation flight altitude parameter, and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter to obtain a task matching matrix, obtain a micro-airport matching vector through normalization of the distance matrix and the task matching matrix, and obtain a set of candidate micro-airports from the micro-airport matching vector through a preset matching threshold; obtain an area coverage matrix through the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set, obtain an operation weight coefficient by mapping the task type parameter to a preset task type dictionary, and obtain a task fitness matrix by multiplying the area coverage matrix with the operation weight coefficient;

[0038] The communication grouping module is used to solve the task fitness matrix through a linear programming algorithm to obtain the optimal UAV combination sequence, obtain the spectrum interference matrix by calculating the degree of interference between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence, process the spectrum interference matrix through a spectral clustering algorithm to obtain a communication grouping scheme, and process the communication grouping scheme through a graph coloring algorithm to obtain a communication channel allocation sequence.

[0039] Furthermore, the system comprises:

[0040] Feature extraction and compression module for drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for:

[0041] ;

[0042] in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for:

[0043] ;

[0044] in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for:

[0045] ;

[0046] in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in .

[0047] Furthermore, the system further comprises:

[0048] Feature classification module for sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is:

[0049] ;

[0050] in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is:

[0051] ;

[0052] in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The feature vector between .

[0053] Furthermore, the system further comprises:

[0054] Encryption module, used to convert feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules :

[0055] ;

[0056] in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for:

[0057] ;

[0058] in and is the pre-set public key, is a random number, It is a modular operation.

[0059] Furthermore, the system further comprises:

[0060] Network generation module for encoding micro-airport privacy protection Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes :

[0061] ;

[0062] in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix :

[0063] .

[0064] (3) Beneficial effects

[0065] Compared with the prior art, the beneficial effect of the present invention is to intelligently perform optimal networking through feature analysis of cross-platform drones, while solving the information security problem in the cross-platform drone networking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flowchart of a low-altitude operation intelligent networking method for integrated command and meteorological support according to Embodiment 1 of the present invention;

[0067] Figure 2 This is a module block diagram of a low-altitude operation intelligent networking system for integrated command and meteorological support according to Example 2 of the present invention. DETAILED DESCRIPTION

[0068] 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.

[0069] Before giving examples, it is necessary to explain the application scenarios of the present invention. For example, in a certain urban area, a large number of drones are deployed, some of which are used for meteorological monitoring, some for surveying and mapping, and some for disaster relief and protection, etc.; there are even the same application scenarios, such as several adjacent areas where drones of multiple user units are used for surveying and mapping, but there is no situation where drone clusters are used in coordination with each other. When a large number of drones are required for some large-scale emergency situations, aerial photography of events, etc., the drones deployed on different drone control platforms cannot be centrally managed unless communication protocols are controlled in advance. In an emergency, it is also impossible to form drones from multiple platforms into a communication network hub, and then centrally issue flight plans, etc. Therefore, how to quickly form drones that operate at low altitudes across platforms into a micro-airport network with central control, and also pay attention to data confidentiality between platforms during the networking process is a problem that needs to be solved. Communication confidentiality is divided into several requirements. The first is that the parameter information of the drone configuration must not be leaked during the communication process. The reason is that the operators of major commercial micro-airports need to ensure that the relevant parameter configurations and communication frequency bands of their own drones will not be leaked to each other during the temporary networking process, to prevent subsequent commercial competitors from interfering with the normal drone cruise missions under specific circumstances, such as blocking drone signals through specific frequency bands and specific areas; the second is that the networking communication process needs to be reasonably arranged according to the communication level of the drone. Some drones act as relay transmission signal nodes to convey signals to the drones of the next level, and the signals transmitted between drones can only be received by the set drones. Even if other drones receive the signals, they cannot crack them, to ensure the security of the temporary networking process of the drone.

[0070] Example 1: Figure 1 As shown, this embodiment provides a low-altitude operation intelligent networking method for integrated command and meteorological support, the method comprising:

[0071] Obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain the networking task information, wherein the networking task information includes the task area coordinates and the task type parameter; the operation type parameter and the task type parameter both include the operation type, the operation flight altitude parameter and the operation flight speed parameter;

[0072] The feature matrix of the micro-airport is compressed to obtain a low-dimensional feature matrix, and feature vectors of different granularity levels are extracted from the low-dimensional feature matrix to obtain a multi-scale feature set. The multi-scale feature set is mapped into a binary hash code through a local sensitive hash function to obtain a privacy protection code for the micro-airport. The task area coordinates are processed through the minimum enclosing rectangle algorithm to obtain a task coverage rectangle. The distance matrix is ​​obtained by calculating the Euclidean distance from the geographical coordinates of multiple micro-airports to the center point of the task coverage rectangle. The task matching matrix is ​​obtained by calculating the Hamming distance between the operation type, operation flight altitude parameter and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter. The micro-airport matching vector is obtained by normalizing the distance matrix and the task matching matrix. The candidate micro-airport set is obtained from the micro-airport matching vector through a preset matching threshold. The regional coverage matrix is ​​obtained by the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set. The operation weight coefficient is obtained by mapping the task type parameter to the preset task type dictionary. The task fitness matrix is ​​obtained by multiplying the regional coverage matrix with the operation weight coefficient.

[0073] The task fitness matrix is ​​solved by a linear programming algorithm to obtain the optimal UAV combination sequence. The spectrum interference matrix is ​​obtained by calculating the interference degree between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence. The spectrum interference matrix is ​​processed by a spectral clustering algorithm to obtain a communication grouping scheme. The communication grouping scheme is processed by a graph coloring algorithm to obtain a communication channel allocation sequence.

[0074] For example, micro-airport A is deployed in the eastern area of ​​city X, and micro-airport B is deployed in the western area, about 15 kilometers apart. Micro-airport A manages 8 drones, mainly used for environmental monitoring. The operation type of each drone is "cruise flight", the flight altitude is 200 meters, the flight speed is 15 meters / second, the operation radius is 5 kilometers, and the communication frequency band is set to A1; micro-airport B manages 5 drones, mainly used for surveying and mapping of construction sites, the operation type is "fixed-point scanning", the flight altitude is 150 meters, the flight speed is 10 meters / second, the operation radius is 3 kilometers, and the communication frequency band is B1. Now a large-scale emergency search and rescue mission is needed. The mission area is in the middle area between the two micro-airports, and the coverage range is required to be 8 kilometers in radius. The mission type is "cruise monitoring", and the drone needs to fly at an altitude of 200 meters and a speed of 15 meters / second. The mission requirements are highly matched with the cruise monitoring drones of micro-airport A, but it is also hoped to use the surveying and mapping drones in micro-airport B as much as possible to expand the coverage. It should be pointed out that the type of operation sometimes determines the external sensors or cameras that the drone needs to mount, such as mounting a camera with a specific shooting frame accuracy, detecting temperature and humidity or specific sensors.

[0075] The specific steps are as follows: The drone feature vectors of micro-airport A are generated based on the identification number, operation type, flight altitude, flight speed, operation radius and communication frequency band of each drone. For example, the feature vector of drone A1 is [1, "cruise", 200, 15, 5, "A1"]. Similarly, the feature vector of drones at micro-airport B, such as drone B1, is [1, "fixed-point scanning", 150, 10, 3, "B1"]. These feature vectors form the feature matrix of micro-airports A and B. In order to improve the computational efficiency, the system compresses the feature matrix to obtain a low-dimensional feature matrix, generates a multi-scale feature set by extracting multi-scale features, and generates a binary hash code for the subsequent use of the local sensitive hash function.

[0076] According to the requirements of the emergency task, the system performs the minimum circumscribed rectangle algorithm on the coordinates of the task area and calculates a coverage rectangle with a radius of 8 kilometers. Then, the distance matrix is ​​generated by calculating the geographical distance from micro-airports A and B to the task center. The results show that micro-airport A is closer and suitable for the main coverage task. At the same time, the task type "cruise monitoring" is highly matched with the drone operation type in micro-airport A, but there is a certain deviation with the "fixed-point scanning" drone in micro-airport B. The generated task matching matrix shows that the matching degree of micro-airport A is higher than that of B. Finally, 5 drones in micro-airport A were selected as the candidate set, and 3 drones that can assist in coverage were selected from micro-airport B.

[0077] In the candidate set, the operating radius of each drone is used to calculate their effective coverage area within the mission coverage area. Assume that the operating radius of the five candidate drones at micro-airport A is 5 kilometers, and the operating radius of the three candidate drones at micro-airport B is 3 kilometers. After calculation, the drones at micro-airport A can effectively cover about 80% of the mission area, while the drones at micro-airport B can cover about 20% of the edge area.

[0078] A fitness matrix was generated based on these coverage data, and the fitness matrix was weighted to prioritize candidate drones based on the importance of the mission and the degree of overlap of the drone coverage locations.

[0079] The linear programming algorithm was used to solve the fitness matrix, and the optimal combination of 6 drones was finally determined (4 of which were from micro-airport A and 2 from micro-airport B). Next, the system analyzed the communication frequency band parameters of the optimal combination and found that there might be some interference between the A1 and B1 bands. The frequency band parameters were clustered using the spectral clustering algorithm, and the communication channels were reallocated using the graph coloring algorithm to ensure that the communication interference between different groups was minimized. Finally, a communication grouping scheme was formed, which enabled drones to perform tasks stably without interfering with each other.

[0080] Among them, the mission area coordinates are processed by the minimum enclosing rectangle algorithm to obtain the mission coverage rectangle. The purpose of obtaining the distance matrix by calculating the Euclidean distance from the geographic coordinates of multiple micro-airports to the center point of the mission coverage rectangle is to determine the coverage of the mission area and calculate the distance from each micro-airport to the mission center. Assume that the center point of the mission area is located at the coordinates (X1, Y1) and needs to cover a radius of 8 kilometers. The rectangular boundary containing the entire mission area is determined by the minimum enclosing rectangle algorithm, that is, the mission coverage rectangle is generated. Then, for each micro-airport (such as A and B), its geographic coordinates are obtained. and , calculate its Euclidean distance to the center point of the mission coverage rectangle. Record the distance from each micro-airport to the mission center point to provide data support for the subsequent screening of micro-airports with appropriate distances.

[0081] Among them, the micro-airport matching vector is obtained by normalizing the distance matrix and the task matching matrix. The purpose of selecting the candidate micro-airport set from the micro-airport matching vector by the preset matching threshold is to select the micro-airports suitable for the task according to the distance and task matching degree. First, the generated distance matrix is ​​normalized to obtain the standardized distance value of each micro-airport relative to the task center, ranging from 0 to 1 (0 represents the nearest and 1 represents the farthest). At the same time, the task matching matrix is ​​generated according to the task parameters (such as flight altitude, flight speed and operation type), and the matching degree between the operation parameters of the drone in each micro-airport and the task requirements is compared; the higher the matching degree, the more suitable the micro-airport is for participating in the task. The normalized distance matrix and the task matching matrix are weightedly fused to generate the micro-airport matching vector. The micro-airport with a higher value in the matching vector indicates that it is suitable for the task. Finally, the candidate micro-airport set is selected according to the preset matching threshold (such as 0.75) to ensure that the participating micro-airports have both moderate distance and high task matching degree.

[0082] Among them, the regional coverage matrix is ​​obtained by the overlapping area of ​​the operating radius of the UAVs in the candidate micro-airport set and the task coverage rectangle, the operation weight coefficient is obtained by mapping the task type parameter to the preset task type dictionary, and the purpose of obtaining the task adaptability matrix by multiplying the regional coverage matrix with the operation weight coefficient is to determine the coverage capability and adaptability of the candidate micro-airport. In the candidate micro-airport set, each UAV in the micro-airport has its own operating radius. The regional coverage matrix is ​​obtained by calculating the overlapping area with the task coverage rectangle by the operating radius of each UAV. For example, if the mission area radius is 8 kilometers, and the operating radius of the UAV in micro-airport A is 5 kilometers, the proportion of the mission area that can be covered is calculated. Next, the mission type (such as "cruise monitoring") is mapped to the preset mission type dictionary to obtain the corresponding operation weight coefficient (such as the weight coefficient of cruise monitoring may be 0.8, and the weight coefficient of fixed-point scanning may be 0.6). Finally, the regional coverage matrix and the operation weight coefficient are multiplied to obtain the task adaptability matrix. This matrix reflects the coverage degree and task adaptability of each UAV in the mission area. The higher the adaptability, the more suitable the UAV is for performing this mission.

[0083] Among them, the optimal UAV combination sequence is obtained by solving the task fitness matrix through a linear programming algorithm. The purpose of obtaining the spectrum interference matrix by calculating the interference degree between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence is to select the optimal UAV combination from the fitness matrix and analyze its communication interference. Apply a linear programming algorithm (such as the simplex method) to the task fitness matrix, and use the maximum task fitness as the objective function to solve the optimal UAV combination sequence suitable for this task. For example, among multiple UAV combinations, 3 UAVs at micro-airport A and 2 UAVs at micro-airport B may be selected. Once the optimal combination is determined, its communication frequency band parameters are analyzed next. Since the UAV frequency bands of micro-airports A and B are A1 and B1 respectively, a spectrum interference matrix is ​​generated to evaluate the interference degree of these UAVs operating simultaneously in the same area. By calculating the spectrum overlap and the possibility of interference between adjacent frequency bands, the specific interference value in the interference matrix is ​​determined (such as the interference degree between the two frequency bands A1 and B1 is 0.3).

[0084] Among them, the purpose of processing the spectrum interference matrix by the spectral clustering algorithm to obtain the communication grouping scheme and processing the communication grouping scheme by the graph coloring algorithm to obtain the communication channel allocation sequence is to group the drones and allocate communication channels to avoid mutual interference. The drones in the spectrum interference matrix are grouped according to the degree of interference using the spectral clustering algorithm. For example, drones with large frequency band interference are divided into different groups to minimize the communication interference of drones in the same group. After that, the graph coloring algorithm is used to allocate communication channels to each group to ensure that each group of drones transmits data in the designated communication channel without interfering with each other. For example, the drones at micro-airport A are assigned to channel 1, and the drones at micro-airport B are assigned to channel 2. The communication channel allocation sequence finally formed enables drones to work together without interference.

[0085] Furthermore, the method of compressing the feature matrix of the micro-airport to obtain a low-dimensional feature matrix includes:

[0086] Drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for:

[0087] ;

[0088] in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for:

[0089] ;

[0090] in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for:

[0091] ;

[0092] in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in .

[0093] For example, micro-airport A and micro-airport B are deployed in two different areas of the city. Micro-airport A is located in the east and manages 10 drones mainly used for environmental monitoring; micro-airport B is located in the west and is responsible for 8 drones mainly used for terrain mapping. The information of each drone includes the identity number, operation type parameters, operation radius and communication frequency band parameters. In addition, it also contains the geographic coordinate information of each drone. First, the information of these drones is converted into a unified numerical feature. The feature vector generated by the drone information of micro-airport A is such as A1 = [1, "cruise monitoring", 5, "A1"], A2 = [2, "cruise monitoring", 5, "A1"], etc. Similarly, the drone feature vector of micro-airport B is such as B1 = [1, "terrain mapping", 3, "B1"], B2 = [2, "terrain mapping", 3, "B1"]. The drone feature vectors of each micro-airport form a feature matrix to describe the overall operation characteristics of the micro-airport. Next, the feature matrix of the micro-airport is compressed, and the information such as the operation type, operation radius, and communication frequency band of the drone is normalized so that the data can be processed at the same scale. Specifically, the feature vector of the drone is mapped to a high-dimensional space using the Gaussian kernel function, and the kernel matrix is ​​applied to the high-dimensional feature matrix to compress the information into a low-dimensional feature matrix, thereby obtaining the low-dimensional feature matrix of micro-airports A and B, ensuring the representativeness and computational convenience of the micro-airport data. In order to improve privacy and security, the low-dimensional feature matrix is ​​further generated into a multi-scale feature set. Features at different granularity levels help to represent different information of the drone in more detail. Subsequently, the multi-scale feature set is mapped into a binary hash code through the local sensitive hashing (LSH) algorithm to generate the privacy protection code of each micro-airport. For example, the privacy protection code of micro-airport A may generate a binary code similar to 1101010101, while the privacy protection code of micro-airport B is 1110101100. This privacy code is used for identity authentication and information encryption in the subsequent networking process, avoiding the leakage of drone configuration parameters in temporary networking. At the same time, the encrypted code ensures that each micro-airport can cooperate with each other in the execution of the task without exposing its own communication and configuration characteristics. Finally, the privacy protection code also enables the micro-airport to quickly find the task information that best matches itself when it needs to be networked. For example, the mission center can match the privacy protection code of the micro-airport and select suitable micro-airports for temporary networking, thereby realizing flexible task scheduling and coordination with drones.

[0094] Furthermore, the method of extracting feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set includes:

[0095] Sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is:

[0096] ;

[0097] in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is:

[0098] ;

[0099] in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The feature vector between .

[0100] For example, micro-airport A and micro-airport B are set up in the eastern and western areas of the city, respectively, each managing a group of drones with different tasks. The drones at micro-airport A are mainly used for environmental monitoring, while the drones at micro-airport B are used for terrain mapping. In order to ensure the security of cross-platform tasks and the effectiveness of information transmission, multi-granularity feature sets are extracted according to the encryption authority level based on the low-dimensional feature matrix of the micro-airport, so that the safe relay and directional transmission of drone signals can be achieved through different authority levels when networking. At the first-level granularity (the lowest authority level), only basic operation information such as flight altitude, speed, and operation radius are extracted from micro-airport A to form a first-level feature vector set. For example, the feature vector of drone A1 contains [“cruise monitoring”, 200, 15, 5]. This type of drone can only receive signal transmission from high-level nodes, but cannot decode the signal content. The basic drone only executes in the mission area without the authority to transmit or modify signals. When the task requires higher security, switch to the second-level granularity (medium authority level) to extract drone information with higher authority from micro-airport B. These drones not only have basic flight parameters, but also have communication frequency band information and can decode specific signals from neighboring drones. For example, the feature vector of drone B1 is [“terrain surveying”, 150, 10, “B1”]. This level of drone can directional transmit and decode information from low-authority drones, but can only send signals to high-authority drones and cannot decode higher-level signal content. The second-level granularity ensures that medium-authority drones can relay signals in the mission and realize the transmission of mission information between low-authority nodes and high-authority nodes. At the third-level granularity (the highest authority level), only some drones are selected. These drones not only have all the operating parameters and communication frequency band information, but also can perform more complex directional signal transmission and encrypted communication functions. For example, A3 in micro-airport A and B2 in micro-airport B are high-authority nodes that can receive, decode and selectively transmit signals from low-authority and medium-authority drones. The feature vector of A3 contains more detailed information, such as [“environmental monitoring”, 200, 15, “A1”, 5, “decoding authority”]. Such high-authority drones can serve as communication hub nodes, securely transmitting mission information to the next-level nodes through directional transmission and encryption protection, while low-authority drones cannot decipher the signal content.

[0101] Ultimately, the multi-granular feature set based on authority levels enables the drones at micro-airports A and B to form a networking structure with clear encryption levels. High-authority drones act as relay points to transmit signals in a directed manner during the mission, and authority control is used to prevent low-authority drones from deciphering mission information.

[0102] Furthermore, the method of mapping the multi-scale feature set into a binary hash code to obtain the micro-airport privacy protection code by using a local sensitive hash function includes:

[0103] The feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules :

[0104] ;

[0105] in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for:

[0106] ;

[0107] in and is the pre-set public key, is a random number, It is a modular operation.

[0108] For example, in an emergency mission in a certain urban area, it is necessary to form a temporary network of micro-airports A and B to support cross-platform collaboration of drone swarms. The drone swarm of micro-airport A is mainly responsible for environmental monitoring tasks, while the drone swarm of micro-airport B is mainly used for terrain mapping tasks. In order to ensure the secure transmission of information between drones during the mission, the multi-granularity feature set of the micro-airport is encrypted according to the authority level. First, the feature vector of each drone is extracted hierarchically according to the authority level to form a multi-granularity feature set. For example, A1 and A2 in micro-airport A are given the first-level authority to perform basic monitoring tasks, A3 is given the second-level authority with relay function, and A4 is given the third-level authority to perform encrypted directional transmission and task coordination. The feature vector of each granularity level is hashed using the local sensitive hashing (LSH) algorithm, and the information of each drone is encoded into a binary hash code as a privacy protection code. For example, the first-level authority of A1 is encoded as "110010", the second-level authority of A3 is encoded as "101110", and the third-level authority of A4 is encoded as "111101". After the privacy protection codes are generated, they are used for encrypted networking of cross-platform tasks. The low-privilege level 1 code only allows the drone to receive instructions from the level 2 and level 3 drones, but cannot decode the signal content, thereby ensuring the privacy protection and data security of instruction transmission. In a specific area, when the task requirement switches to level 3 authority, the privacy protection code of A4 is encrypted by the public key. For example, the encrypted code "011010111" is obtained by encrypting with the public key P1. This encrypted code ensures that the characteristic information of the A4 drone cannot be decoded by low-privilege drones during transmission. Furthermore, during task scheduling, the encrypted codes of each drone are transmitted to the central control node, and the drone combination that meets the task requirements is screened out by matching the semantic relevance of the codes.

[0109] Furthermore, the method of obtaining the task matching matrix by calculating the Hamming distance between the operation type, the operation flight altitude parameter and the operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter includes:

[0110] Encoding the privacy protection of micro-airports Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes :

[0111] ;

[0112] in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix :

[0113] .

[0114] For example, in a large-scale emergency mission in an urban area, drones at micro-airports A and B need to work together to complete complex multiple tasks. The drones at micro-airport A are mainly used for environmental monitoring, while drones at B are responsible for terrain mapping. This mission requires both high-altitude drones to conduct large-scale aerial photography and low-altitude drones to conduct precise detection in specific areas. The mission requires drones with different permissions and different mission types to cooperate highly. Therefore, the privacy protection code of the micro-airport is decrypted to obtain the original binary hash code of each drone, and a task matching matrix is ​​generated to ensure the reasonable selection of drones. First, the Hamming distance between the binary task hash code (such as "101001") generated by the task type parameter and the binary hash code of each drone is calculated. Assuming that the task requirements include two types of cruise monitoring and terrain mapping, the Hamming distance between the drone A1 code "101000" of micro-airport A in the matching matrix and the task hash code is 1, indicating that A1 has a high matching degree with the cruise monitoring requirement; the drone B1 code "101010" of micro-airport B has a Hamming distance of 1 with the task hash code, which meets the terrain mapping requirement. Therefore, A1 and B1 were selected as preliminary candidate drones for this mission. Further analysis of the mission requirements revealed that the mission area coverage radius was 10 kilometers, requiring high-altitude drones to perform wide-area aerial photography, while low-altitude drones were required to perform intensive collection in the detection area. The regional coverage matrix was calculated by the overlapping area of ​​each drone's operating radius and the mission coverage area. For example, the operating radius of the high-altitude drone A1 is 8 kilometers, and the overlapping area with the mission area is 80%; the operating radius of the low-altitude drone B1 is 3 kilometers, which can cover 60% of the specific detection area. The weight coefficients of the coverage area and the mission type parameters are weighted to generate the mission fitness matrix. At the same time, the fitness matrix uses a linear programming algorithm to generate a drone combination, including A1 and A3 of micro-airport A and B1 and B3 of micro-airport B. According to the mission requirements, A1 and A3 are responsible for wide-area cruising and data collection at a speed of 15 meters per second at high altitudes to cover the entire mission area. At the same time, B1 and B3, as low-altitude detection drones, perform fixed-point scanning and real-time image analysis at a speed of 10 meters per second. High-altitude drones A1 and A3 take large-scale photos from an altitude of 200 meters to capture aerial image data of the entire area and provide a panoramic background for low-altitude drones; while B1 and B3 cooperate to conduct precise detection at a low altitude of 50 meters, and upload detailed data of key areas to the central node synchronously to ensure dual coverage of terrain mapping and environmental monitoring data. In order to ensure collaborative communication between drones, the degree of interference between the communication frequency bands of each drone is calculated to generate a spectrum interference matrix. For example, the communication frequency bands of A1 and B1 are A1 band and B1 band, respectively, with an interference degree of 0.15, while A1 and A3 are in the same frequency band with a higher interference degree of 0.3. The interference matrix is ​​grouped by the spectral clustering algorithm, and high-interference drones are assigned to different groups to reduce communication interference.Subsequently, the communication channels of each group were allocated through the graph coloring algorithm. Finally, A1 and A3 were allocated to channel 1, and B1 and B3 were allocated to channel 2, ensuring that drones at different operating altitudes and operating tasks could communicate independently and not interfere with each other. In actual operations, the high-altitude aerial photography data of A1 and A3 provided the mission center with a large range of real-time images, while B1 and B3 simultaneously detected environmental changes in specific areas at low altitudes. When A1 found that a specific area needed further investigation, it immediately sent instructions to B1 at low altitude through an encrypted channel. After receiving the instructions, B1 went directly to the target area for detailed data collection. At the same time, A3, as a high-authority drone, can send encrypted signals to B3 with low authority to command it to conduct additional surveying and mapping in the designated area, while B3 cannot decode all the data content of the high-authority drone and only receives directional instructions to perform detection. Through the precise division of labor and encrypted communication between high- and low-altitude drones, efficient collaboration under different task types is achieved.

[0115] Embodiment 2: Based on the same inventive concept, Figure 2 As shown, this embodiment also provides a low-altitude operation intelligent networking system for integrated command and meteorological support, the system comprising:

[0116] A networking node acquisition module is used to obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain networking task information, wherein the networking task information includes the task area coordinates and task type parameters; the operation type parameters and task type parameters both include the type of operation, the operation flight altitude parameter and the operation flight speed parameter;

[0117] A matching encryption module is used to compress the feature matrix of a micro-airport to obtain a low-dimensional feature matrix, extract feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set, and map the multi-scale feature set into a binary hash code through a local sensitive hash function to obtain a privacy protection code for a micro-airport; process the task area coordinates through a minimum enclosing rectangle algorithm to obtain a task coverage rectangle, calculate the Euclidean distance from multiple micro-airport geographic coordinates to the center point of the task coverage rectangle to obtain a distance matrix, calculate the Hamming distance between the operation type, operation flight altitude parameter, and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter to obtain a task matching matrix, obtain a micro-airport matching vector through normalization of the distance matrix and the task matching matrix, and obtain a set of candidate micro-airports from the micro-airport matching vector through a preset matching threshold; obtain an area coverage matrix through the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set, obtain an operation weight coefficient by mapping the task type parameter to a preset task type dictionary, and obtain a task fitness matrix by multiplying the area coverage matrix with the operation weight coefficient;

[0118] The communication grouping module is used to solve the task fitness matrix through a linear programming algorithm to obtain the optimal UAV combination sequence, obtain the spectrum interference matrix by calculating the degree of interference between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence, process the spectrum interference matrix through a spectral clustering algorithm to obtain a communication grouping scheme, and process the communication grouping scheme through a graph coloring algorithm to obtain a communication channel allocation sequence.

[0119] Furthermore, the system comprises:

[0120] Feature extraction and compression module for drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for:

[0121] ;

[0122] in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for:

[0123] ;

[0124] in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for:

[0125] ;

[0126] in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in .

[0127] Furthermore, the system further comprises:

[0128] Feature classification module for sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is:

[0129] ;

[0130] in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is:

[0131] ;

[0132] in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The feature vector between .

[0133] Furthermore, the system further comprises:

[0134] Encryption module, used to convert feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules :

[0135] ;

[0136] in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for:

[0137] ;

[0138] in and is the pre-set public key, is a random number, It is a modular operation.

[0139] Furthermore, the system further comprises:

[0140] Network generation module for encoding micro-airport privacy protection Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes :

[0141] ;

[0142] in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix :

[0143] .

[0144] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0145] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A low-altitude operation intelligent networking method for integrated command and meteorological support, characterized in that: The method comprises: Obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain the networking task information, wherein the networking task information includes the task area coordinates and the task type parameter; the operation type parameter and the task type parameter both include the operation type, the operation flight altitude parameter and the operation flight speed parameter; The feature matrix of the micro-airport is compressed to obtain a low-dimensional feature matrix, and feature vectors of different granularity levels are extracted from the low-dimensional feature matrix to obtain a multi-scale feature set. The multi-scale feature set is mapped into a binary hash code through a local sensitive hash function to obtain a privacy protection code for the micro-airport. The task area coordinates are processed through the minimum enclosing rectangle algorithm to obtain a task coverage rectangle. The distance matrix is ​​obtained by calculating the Euclidean distance from the geographical coordinates of multiple micro-airports to the center point of the task coverage rectangle. The task matching matrix is ​​obtained by calculating the Hamming distance between the operation type, operation flight altitude parameter and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter. The micro-airport matching vector is obtained by normalizing the distance matrix and the task matching matrix. The candidate micro-airport set is obtained from the micro-airport matching vector through a preset matching threshold. The regional coverage matrix is ​​obtained by the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set. The operation weight coefficient is obtained by mapping the task type parameter to the preset task type dictionary. The task fitness matrix is ​​obtained by multiplying the regional coverage matrix with the operation weight coefficient. The optimal UAV combination sequence is obtained by solving the task fitness matrix through a linear programming algorithm. The spectrum interference matrix is ​​obtained by calculating the interference degree between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence. The spectrum interference matrix is ​​processed by a spectral clustering algorithm to obtain a communication grouping scheme. The communication grouping scheme is processed by a graph coloring algorithm to obtain a communication channel allocation sequence. The method of compressing the feature matrix of the micro-airport to obtain a low-dimensional feature matrix comprises: Drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for: ； in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for: ； in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for: ； in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, Is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in ; The method of extracting feature vectors of different granularity levels from a low-dimensional feature matrix to obtain a multi-scale feature set includes: Sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is: ； in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is: ； in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The eigenvector between ; The method for mapping a multi-scale feature set into a binary hash code to obtain a micro-airport privacy protection code by using a local sensitive hash function includes: The feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules : ； in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for: ； in and is the pre-set public key, is a random number, is a modular operation; The method for obtaining a task matching matrix by calculating the Hamming distance between the operation type, the operation flight altitude parameter, and the operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter includes: Encoding the privacy protection of micro-airports Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes : ； in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix : 。 2. Low-altitude operation intelligent networking system for comprehensive command and meteorological support, characterized by: The system comprises: A networking node acquisition module is used to obtain the geographic coordinates of at least two micro-airports and the corresponding deployed drone information, wherein the drone information includes the drone's identity number, operation type parameter, operation radius and communication frequency band parameter, obtain the feature vector of a single drone through the drone information and the corresponding geographic coordinates, and obtain the feature matrix of the micro-airport through the feature vectors of all drones in the micro-airport; obtain networking task information, wherein the networking task information includes the task area coordinates and task type parameters; the operation type parameters and task type parameters both include the type of operation, the operation flight altitude parameter and the operation flight speed parameter; A matching encryption module is used to compress the feature matrix of a micro-airport to obtain a low-dimensional feature matrix, extract feature vectors of different granularity levels from the low-dimensional feature matrix to obtain a multi-scale feature set, and map the multi-scale feature set into a binary hash code through a local sensitive hash function to obtain a privacy protection code for a micro-airport; process the task area coordinates through a minimum enclosing rectangle algorithm to obtain a task coverage rectangle, calculate the Euclidean distance from multiple micro-airport geographic coordinates to the center point of the task coverage rectangle to obtain a distance matrix, calculate the Hamming distance between the operation type, operation flight altitude parameter, and operation flight speed parameter in the privacy protection code of the micro-airport and the corresponding task type parameter to obtain a task matching matrix, obtain a micro-airport matching vector through normalization of the distance matrix and the task matching matrix, and obtain a set of candidate micro-airports from the micro-airport matching vector through a preset matching threshold; obtain an area coverage matrix through the overlapping area of ​​the UAV operation radius and the task coverage rectangle in the candidate micro-airport set, obtain an operation weight coefficient by mapping the task type parameter to a preset task type dictionary, and obtain a task fitness matrix by multiplying the area coverage matrix with the operation weight coefficient; The communication grouping module is used to solve the task fitness matrix through a linear programming algorithm to obtain the optimal UAV combination sequence, calculate the interference degree between the communication frequency band parameters of the UAVs in the optimal UAV combination sequence to obtain the spectrum interference matrix, process the spectrum interference matrix through a spectral clustering algorithm to obtain a communication grouping scheme, and process the communication grouping scheme through a graph coloring algorithm to obtain a communication channel allocation sequence; The system further comprises: Feature extraction and compression module for drones deployed at micro-airports The corresponding drone information is constructed as ,in Representative drone identity number, Representative drone The job type parameter, Representative drone The operating radius, Representative drone Communication frequency band parameters; suppose the UAV The geographical coordinates are ,in Representative drone The longitude coordinates of Representative drone The latitude coordinates of the drone and geographic coordinates After normalization, we get and , through the Gaussian kernel function and Mapping to high-dimensional feature space Get high-dimensional feature vector for: ； in is the number of Gaussian kernel functions and is a high-dimensional feature space The dimension of For the preset The bandwidth parameter of the Gaussian kernel function, is the base of the natural logarithm; the high-dimensional feature vectors corresponding to two random drones at the micro-airport are and Calculate the kernel matrix for: ； in is the global bandwidth parameter of the preset Gaussian kernel function; through the kernel matrix Embed high-dimensional feature vectors into low-dimensional manifold space to obtain high-dimensional dynamic manifold structure , the objective function for: ； in For high-dimensional dynamic manifold structures Sparse representation matrix, high-dimensional dynamic manifold structure The original manifold features Drones for micro-airports The high-dimensional feature vector corresponding to the information The matrix composed of is to ask of norm, is to find the sparse representation matrix of norm, is the preset sparse regularization parameter; the minimum value of the first objective formula is obtained by the gradient descent method, and the sparse representation matrix is ​​obtained according to the minimum value of the first objective formula ; The sparse representation matrix Reconstructing the original manifold features in high-dimensional dynamic manifold structures Get a low-dimensional feature matrix ,in ; The system further comprises: Feature classification module for sampling drone nodes from micro-airports , drone node The corresponding low-dimensional feature matrix satisfy ; Initialize the granularity level according to the preset security level , a low-dimensional feature matrix The corresponding drone information embedding vector The formula is: ； in is the first activation function, and Granularity level The weight matrix and bias vector of ,in To obtain the drone collection, To obtain the set of communication links, The value ranges from 0 to the drone set An integer variable between the number of The value ranges from 0 to the communication link set Integer variable between the number of drone nodes , communication link node , drone node and communication link nodes The formed edge is recorded as an edge node And edge nodes Composition edge set Satisfaction ; Initialize the drone node Feature vector at granularity level 0 for , No. Feature vector at granular level The calculation formula is: ； in is the second activation function, The value ranges from 0 to the granularity level integer variable, For drone nodes and communication link nodes The set of neighbor nodes of The value is a set of neighbor nodes The drone nodes within For drone nodes and communication link nodes The attention coefficient between For the The weight matrix of the layer, The granularity level is Corresponding drone node and communication link nodes The eigenvector between ; The system further comprises: Encryption module, used to convert feature vectors at each granularity level Map it into semantically relevant binary hash codes at the corresponding level through cascaded hash modules : ； in is a locality sensitive hash function, For the A micro-airport. The value ranges from 1 to the number of micro-airports participating in the smart network. An integer between Denoted as the micro-airport privacy protection code, the encrypted micro-airport privacy protection code for: ； in and is the pre-set public key, is a random number, is a modular operation; The system further comprises: Network generation module for encoding micro-airport privacy protection Decrypt to get the original binary hash code , the task type parameter is mapped through the same local sensitive hash function as the micro-airport privacy protection encoding to obtain the task hash code , calculate the Hamming distance of binary hash codes : ； in is the hash code length, is the task number, and The first Bit binary number, The value range is 1 to An integer; The Hamming distances between micro-airports and tasks form the task matching matrix : 。

Images