Intelligent generation method and device of unmanned aerial vehicle countermeasure strategy

Abstract

This invention discloses an intelligent generation method and apparatus for drone countermeasure strategies, relating to the field of drone technology. It addresses the problems of limited applicable scenarios, lack of deep semantic support, and limited methods in existing drone countermeasure technologies, resulting in poor matching effects and low efficiency. The method includes: acquiring multi-source sensor data of a target area to determine environmental information and the situational type information of the suspected drone; extracting features corresponding to the situational type information from the multi-source sensor data and performing similarity matching with features of multiple target matching entity nodes to determine the matching result; based on the matching result and environmental information, acquiring at least one countermeasure from a drone countermeasure knowledge graph and generating an initial countermeasure strategy set; and processing the initial countermeasure strategy set using a target optimization algorithm to obtain the target countermeasure strategy. This invention aims to improve the matching accuracy and efficiency of drone countermeasure technology.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a method and apparatus for intelligently generating UAV countermeasure strategies. Background Technology

[0002] With the rapid development of drone technology, security threats such as illegal intrusion and malicious reconnaissance are becoming increasingly prominent, making drone countermeasure systems a key piece of equipment for ensuring airspace security. Existing methods for generating drone countermeasure strategies mainly rely on preset rule bases or simple numerical matching, which have significant shortcomings in practical applications.

[0003] First, it has poor adaptability to complex scenarios. Existing methods struggle to correlate various factors such as UAV situation and flight environment. When facing unknown UAV models or new tactics, they cannot generate effective countermeasures based on feature correlations and can only passively match preset rules, thus limiting their adaptability.

[0004] Secondly, decision-making lacks deep semantic support. Existing methods determine countermeasures solely through numerical calculations or surface feature matching, failing to uncover the underlying logic of the data. This results in insufficient targeting of countermeasure strategies, making it difficult to address the diverse threats posed by drones.

[0005] Third, the countermeasures are too simplistic. Existing systems typically only provide a single countermeasure, failing to form a comprehensive systemic solution that combines multiple methods. This makes it difficult to balance the effectiveness of countermeasures with the security of the scenario, and can easily lead to secondary risks in complex environments such as airports and cities.

[0006] Therefore, there is an urgent need for a drone countermeasure strategy generation method that can adapt to complex scenarios, uncover deep correlations, and generate diverse solutions to address the aforementioned shortcomings of existing technologies. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent generation method and apparatus for drone countermeasure strategies, which can improve the accuracy and efficiency of drone countermeasures.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] In a first aspect, the present invention provides an intelligent generation method for drone countermeasure strategies, comprising:

[0010] Acquire multi-source sensor data of the target area, and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation;

[0011] The matching features corresponding to the situation type information are extracted from the multi-source sensor data, and the matching features are matched with the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the matching features are either multimodal features corresponding to the single-UAV situation or cluster situation features corresponding to the cluster situation.

[0012] Based on the matching results and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated.

[0013] The initial countermeasure strategy set is processed using a target optimization algorithm to obtain the target countermeasure strategy.

[0014] Optionally, the feature to be matched is compared with the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result, including:

[0015] According to the similarity calculation formula:

[0016] ;

[0017] Calculate the similarity score for each target matching entity node; where, Similarity; The feature vector to be matched; ; Match the feature vectors of entity nodes to the target; ; The first feature vector to be matched Individual feature terms; The first feature vector of the target entity node is matched to the feature vector of the target entity node. Individual feature terms; It is a positive integer greater than 2; For the first The weight coefficients of each sub-feature term; It is a positive integer greater than 2;

[0018] The matching result is determined based on the similarity values ​​of all target matching entity nodes.

[0019] Optionally, the entity nodes in the UAV countermeasure knowledge graph include at least tactical pattern nodes;

[0020] When the situation type information is a cluster situation, the target matching entity node is a tactical mode node, and the matching result is the target tactical mode node;

[0021] The situational characteristics of the cluster to be matched include quantitative and structural characteristics, formation characteristics, cooperative action characteristics, and communication and coordination characteristics;

[0022] The tactical mode nodes include at least cluster-coordinated attack mode nodes, multi-type hybrid flanking penetration mode nodes, and distributed cluster harassment mode nodes;

[0023] Based on the similarity scores of all target matching entity nodes, the matching results are determined, including:

[0024] The tactical mode node corresponding to the maximum similarity value is determined as the target tactical mode node.

[0025] Optionally, the entity nodes in the drone countermeasure knowledge graph also include drone model nodes; when the situation type information is the single-drone situation, the target matching entity node is the drone model node, and the matching result is the target drone model node that matches the suspected drone; the multimodal features to be matched include radio frequency features, flight features, and optical features;

[0026] Based on the similarity scores of all target matching entity nodes, the matching results are determined, including:

[0027] Compare the similarity values ​​corresponding to all drone model nodes with the first preset similarity threshold;

[0028] If there is at least one drone model node whose similarity value is greater than or equal to the first preset similarity threshold, then the drone model node corresponding to the maximum similarity value is determined as the target drone model node.

[0029] If the similarity values ​​corresponding to all drone model nodes are less than the first preset similarity threshold, the model of the suspicious drone is determined to be an unknown model, and matching continues according to a preset priority order to determine the target drone model node; the preset priority order is as follows: communication protocol node, sub-communication protocol node, hardware feature type node, and software feature type node.

[0030] Optionally, the entity nodes in the drone countermeasure knowledge graph also include multiple communication protocol nodes;

[0031] The matching process continues according to a preset priority order to determine the target drone model node, including:

[0032] Multiple communication protocol sub-features are extracted from the multimodal features to be matched, and a communication protocol feature vector is constructed; the multiple communication protocol sub-features include signal frequency band, frequency hopping pattern, and frame structure identifier;

[0033] Using the aforementioned similarity calculation formula, the communication protocol feature vector is matched with the preset feature vectors of multiple communication protocol nodes to obtain the first similarity result for each communication protocol node.

[0034] If there is at least one communication protocol node whose first similarity result is greater than or equal to the second preset similarity threshold, then the communication protocol node corresponding to the maximum value of the first similarity result is determined as the target communication protocol node, and the node of the known drone model corresponding to the target communication protocol node is determined as the target drone model node through the adoption protocol relationship between the communication protocol node and the drone model node.

[0035] If the similarity results of all communication protocol nodes are less than the second preset similarity threshold, then the similarity protocol relationship between communication protocol nodes in the UAV countermeasure knowledge graph is invoked, and multiple sub-communication protocol nodes associated with the target communication protocol node are selected based on the filtering conditions to continue similarity matching, resulting in multiple second similarity results; the filtering conditions are that the protocol similarity attribute value is greater than or equal to the third preset similarity threshold.

[0036] If at least one second similarity result is greater than or equal to the second preset similarity threshold, then the corresponding target UAV model node is obtained;

[0037] If all the second similarity results are less than the second preset similarity threshold, then continue matching according to the preset priority order to determine the target UAV model node.

[0038] Optionally, the relationships in the drone countermeasure knowledge graph include at least the relationships that can be countered, the relationships applicable to the environment, the relationships that are prohibited in the environment, and the relationships that affect the effectiveness of countermeasures in the environment.

[0039] When the situation type information is a single-machine situation, based on the matching result and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated, including:

[0040] Based on the countermeasureable relationship, all countermeasures associated with the matching result are obtained to form a first countermeasure pool;

[0041] Based on the applicable environmental relationship, countermeasures containing the environmental information are selected from the first countermeasures pool to obtain a second countermeasures pool;

[0042] By using the disabled environment relationship, the third countermeasure pool is obtained by excluding countermeasures in the second countermeasure pool whose disabled environment contains the environment information.

[0043] By using the aforementioned environmental impact countermeasure effectiveness relationship, the association attributes between the environmental information in the UAV countermeasure knowledge graph and each countermeasure in the third countermeasure pool are retrieved, and the corrected effectiveness is calculated by combining the original effectiveness of each countermeasure.

[0044] Based on the corrected effectiveness of each means in the third countermeasure pool, all countermeasures are combined and the expected countermeasure effect of each combination is marked to obtain the initial countermeasure strategy set for the single-machine situation.

[0045] Optionally, the relationships in the drone countermeasure knowledge graph also include tactical combination relationships;

[0046] When the situation type information is a cluster situation, based on the matching result and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated, including:

[0047] Invoke the tactical combination relationship to obtain the countermeasure coordination framework associated with the target tactical mode node;

[0048] For each link of the aforementioned countermeasures coordination framework, candidate countermeasures for each link are obtained from the UAV countermeasures knowledge graph;

[0049] Based on the applicable environment relationship, the candidate means for each stage are selected from those whose applicable environment contains the environmental information, thus obtaining the first countermeasure set for each stage;

[0050] Based on the aforementioned disabled environment relationship, the means in the first countermeasure set of each stage whose disabled environment contains the aforementioned environmental information are excluded, thus obtaining the second countermeasure set of each stage;

[0051] Based on the aforementioned environmental impact countermeasure effectiveness relationship, the correlation attributes of each means in the second countermeasure means set at each stage are retrieved, the original effectiveness of each means is corrected, and the corrected effectiveness of each means is obtained.

[0052] Based on the effectiveness of each modified method, methods are selected from the second set of countermeasures in each stage and combined in sequence according to the countermeasures coordination framework. The expected countermeasure effect of each combination is marked to generate an initial set of countermeasure strategies for the cluster situation.

[0053] Optionally, the initial countermeasure strategy set is processed using a target optimization algorithm to obtain a target countermeasure strategy, including:

[0054] A multi-objective optimization set is determined; the multi-objective optimization set includes benefit-oriented optimization objectives and cost-oriented optimization objectives; the benefit-oriented optimization objectives include reaction success rate and collaborative efficiency; the cost-oriented optimization objectives include response time, economic cost, and collateral damage;

[0055] Formula used:

[0056] ;

[0057] The original data of the benefit-oriented optimization objectives are normalized to obtain the normalized values ​​of each benefit-oriented optimization objective; among them, For the first The normalized value of a benefit-oriented optimization objective; For the first The original data for each benefit-oriented optimization objective; For the first The minimum baseline value for each benefit-oriented optimization objective; For the first The maximum baseline value of each benefit-oriented optimization objective;

[0058] Formula used:

[0059] ;

[0060] The original data for cost-type optimization objectives are normalized to obtain normalized values ​​for each cost-type optimization objective; among them, For the first The normalized value of a cost-type optimization objective; For the first The original data for each cost-based optimization objective; For the first The minimum baseline value for a cost-based optimization objective; For the first The maximum baseline value for each cost-based optimization objective;

[0061] The weight coefficients of each optimization objective are determined based on the situation type information.

[0062] Formula used:

[0063] ;

[0064] Calculate the overall score for each initial countermeasure strategy in the initial countermeasure strategy set; where, The overall score; For the first The weight coefficients of the first optimization objective; if the first... If the optimization objective is a benefit-oriented optimization objective, then If the first If the optimization objective is a cost-based optimization objective, then ;

[0065] The initial countermeasure strategy with the highest overall score is selected as the target countermeasure strategy.

[0066] Optionally, the intelligent generation method for the drone countermeasure strategy further includes:

[0067] Based on the target countermeasure strategy, the corresponding countermeasure execution device is controlled to perform countermeasure tasks against the suspicious drone, and the countermeasure process data is recorded in real time during the execution of the countermeasure task.

[0068] If the countermeasure process data shows that the countermeasure effectiveness of the target countermeasure strategy is higher than a preset threshold, then the attribute of the countermeasure effectiveness relationship between the target countermeasure strategy and the corresponding UAV model in the UAV countermeasure knowledge graph is updated.

[0069] Compared with existing technologies, the intelligent generation method for UAV countermeasure strategies provided by this invention addresses three major shortcomings of the background technology through a three-layer core design: First, by acquiring multi-source sensor data and determining environmental information and individual and cluster situations, it overcomes the limitation of existing methods in associating multiple factors. Second, relying on a UAV countermeasure knowledge graph, it performs similarity matching between features to be matched and entity nodes, such as matching model nodes with individual multimodal features and tactical mode nodes with cluster features. It utilizes semantic relationships in the graph to mine the underlying logic of data associations, providing deep semantic support for decision-making and solving the problem of insufficient countermeasure targeting. Finally, it first generates an initial strategy set containing combinations of multiple countermeasures based on the matching results and environmental information, and then selects the optimal strategy through a target optimization algorithm, replacing the limitation of single-means output in existing systems. This method can balance countermeasure effectiveness and security in scenarios such as airports and cities, avoiding secondary risks.

[0070] Secondly, the present invention also provides an intelligent generation device for drone countermeasure strategies, comprising:

[0071] The acquisition module is used to acquire multi-source sensor data of the target area, and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation;

[0072] The matching module is used to extract the features to be matched corresponding to the situation type information from the multi-source sensor data, and to perform similarity matching between the features to be matched and the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the features to be matched are multimodal features to be matched corresponding to the single-UAV situation or cluster situation features to be matched corresponding to the cluster situation.

[0073] The generation module is used to obtain at least one countermeasure from the UAV countermeasure knowledge graph and generate an initial set of countermeasure strategies based on the matching results and the environmental information.

[0074] The optimization module is used to process the initial countermeasure strategy set using a target optimization algorithm to obtain the target countermeasure strategy. Attached Figure Description

[0075] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0076] Figure 1 A flowchart illustrating an intelligent generation method for drone countermeasure strategies provided in an embodiment of the present invention;

[0077] Figure 2 A schematic diagram of the structure of an intelligent generation device for drone countermeasure strategies provided in an embodiment of the present invention. Detailed Implementation

[0078] To facilitate a clear description of the technical solutions in the embodiments of the present invention, the terms "first" and "second" are used to distinguish identical or similar items with essentially the same function and effect. For example, the first threshold and the second threshold are merely used to distinguish different thresholds and do not limit their order. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0079] It should be noted that in this invention, the terms "exemplarily" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplarily" or "for example" in this invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplarily" or "for example" is intended to present the relevant concepts in a concrete manner.

[0080] In this invention, "at least one" means one or more, and "more than one" means two or more. "And / or" describes the relationship between the associated objects, indicating that three relationships can exist.

[0081] like Figure 1 As shown, this embodiment of the invention provides an intelligent generation method for drone countermeasure strategies, which may include:

[0082] Step 100: Acquire multi-source sensor data of the target area, and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation;

[0083] First, let's introduce multi-source sensor data. Multi-source sensor data mainly includes the following three core types of data: radar sensor data, photoelectric sensor data, and radio detection sensor data.

[0084] For radar sensor data, the core data are the real-time three-dimensional position coordinates of the target UAV, the number of targets, and the velocity vector. The velocity vector includes the flight direction and flight acceleration. All of these data can directly capture the spatial distribution and motion state of the UAV.

[0085] For photoelectric sensor data, the core data includes the drone's optical imaging (body outline, color, number of propellers, etc.), number of images, and spatial spacing. Continuous frame image analysis can help determine the number of drones and their formation structure, while also identifying visual features of the geographical environment such as airport runways and city buildings.

[0086] For radio detection sensor data, the core data includes the number of independent radio frequency signal sources, signal frequency bands, frequency hopping patterns, frame structure, and communication protocol characteristics.

[0087] Next, we will introduce the process of determining the environmental information of the target area based on multi-source sensor data.

[0088] Environmental information is mainly determined through two dimensions: geographical environment and weather conditions, both of which rely on the comprehensive analysis of multi-source sensor data.

[0089] For example, determining the geographical environment may include the following steps:

[0090] By combining spatial coordinate data from radar sensors, the spatial characteristics of the target area can be determined. Spatial characteristics include the fixed coordinate range of the airport airspace and the location coordinate dispersion caused by the building distribution in densely populated urban areas.

[0091] By using imaging data from photoelectric sensors, we can identify the distinctive features of target areas, such as the visual differences between airport runways, city skyscrapers, and open areas in the suburbs.

[0092] The signal interference characteristics of radio detection data can be used to assist in verification, such as the dense civilian signals in urban environments and the existence of specific civil aviation communication frequency bands in airport environments.

[0093] For example, determining weather conditions may include the following steps:

[0094] Based on the degree of radar signal propagation attenuation, determine whether there is heavy rain or dense fog;

[0095] The image clarity of photoelectric sensors helps determine visibility; for example, images are blurry in foggy weather.

[0096] By combining the drone's flight status data, such as speed fluctuations and hovering stability captured by radar, it can be determined whether there is strong wind.

[0097] The following section introduces a method for distinguishing between single-machine and cluster situations based on multi-source sensor data. This method primarily employs two core logic steps: quantitative feature extraction and collaborative verification. The specific judgment rules are as follows:

[0098] Step 1: Quantitative feature extraction.

[0099] If the radar count, electro-optical imaging count, and radio signal source count all show a target count of 1, it is initially determined to be a single-unit situation.

[0100] If data from multiple sensor sources all show that the number of targets is greater than or equal to a preset threshold (e.g., 3 drones), then proceed to the second step of collaborative verification to determine whether it is a cluster situation rather than multiple scattered drones.

[0101] Step 2: Collaborative verification.

[0102] If the quantity is only 1 and there are no collaborative features, it is ultimately determined to be a single-machine situation; no collaborative features can mean no synchronous actions and no multi-machine communication link.

[0103] If the number is greater than or equal to 3, and any of the following coordination conditions are met, the final determination is a cluster situation:

[0104] ① Trajectory correlation: Radar data shows that the spatial distance between multiple UAVs is stable and their speed direction is consistent. For example, stable spatial distance can be defined as a spatial distance fluctuation of less than or equal to 20% and a speed direction deviation of less than or equal to 10°.

[0105] ② Action coordination: Photoelectric sensors capture multiple drones turning, hovering, or performing sequential actions simultaneously; synchronous sequential actions include, for example, interfering with a drone first, attacking a drone, and then penetrating the defense.

[0106] ③ Communication coordination: Radio detection shows that multiple drones are using synchronized frequency hopping signals or master-slave communication links (one drone sends control commands to the others).

[0107] Special case: If there are 2 or more drones but no coordinated characteristics, they are judged as multiple scattered drones and handled separately according to the single-drone situation logic.

[0108] Step 200: Extract the features to be matched corresponding to the situation type information from the multi-source sensor data, and perform similarity matching between the features to be matched and the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the features to be matched are multimodal features to be matched corresponding to the single UAV situation or cluster situation features to be matched corresponding to the cluster situation.

[0109] In other words, step 200 means: if the situation type information is a single-machine situation, then extract the multimodal features to be matched from the multi-source sensor data; if the situation type information is a cluster situation, then extract the cluster situation features to be matched from the multi-source sensor data.

[0110] In this embodiment of the invention, the unmatched multimodal features of the single-machine situation are used for the identification of the UAV model, specifically to identify whether the UAV model is a known model or an unknown model.

[0111] For example, the multimodal features to be matched specifically include three categories: radio frequency features, flight features, and optical features.

[0112] Multiple target matching entity nodes are the core entity node set in the UAV countermeasure knowledge graph, which is used to carry out matching feature similarity matching and clarify feature comparison benchmark. Its core function is to establish a mapping relationship of "feature data → entity association" by calculating the vector similarity with the feature to be matched, so as to provide a basis for subsequent extraction of appropriate countermeasures.

[0113] From the perspective of the corresponding logic of the features to be matched, when the feature to be matched is a multimodal feature of a single drone situation, multiple target matching entity nodes focus on nodes related to the individual attributes of the drone, such as drone model nodes, communication protocol nodes, hardware feature nodes, and software feature nodes, and lock the model or underlying attributes of a single drone through feature comparison. When the feature to be matched is a cluster situation feature of a cluster situation, such as the quantity structure, formation, coordinated actions, and communication coordination features mentioned later, multiple target matching entity nodes focus on nodes related to the group behavior of drones, such as tactical mode nodes such as cluster coordinated attack mode nodes, multi-model mixed flanking penetration mode nodes, and distributed cluster harassment mode nodes, and lock the group tactical intentions of multiple drones through feature comparison.

[0114] In short, the essence of multiple target matching entity nodes is a knowledge graph entity set that is adapted to the feature type to be matched.

[0115] Radio frequency characteristics are directly related to UAV communication protocols, and include the following sub-characteristics: signal frequency band, frequency hopping rules, and frame structure.

[0116] For example, the signal frequency bands are common drone control frequency bands or image transmission frequency bands such as 2.4GHz, 5.8GHz, and 1.5GHz.

[0117] For example, the frequency hopping pattern is a fixed frequency hopping interval of 10ms.

[0118] The frame structure is a proprietary frame header format for a specific vendor's protocol, such as the frame identifier encoding unique to the DJI protocol.

[0119] Flight characteristics reflect the dynamic and controllability of a drone, including the following sub-characteristics: maximum speed, turning radius, and hovering stability.

[0120] Hovering stability, for example: hovering position error is less than or equal to 0.5 meters.

[0121] Optical features are visual distinguishing features based on appearance, which may include the following sub-features: fuselage size, exterior color, number of propellers, and special markings; special markings include fuselage logo, exhaust spectral characteristics, etc.

[0122] Regarding the extraction of multimodal features, all features were directionally parsed and extracted from multi-source sensor data. The extraction process can be found in relevant technologies and will not be elaborated here.

[0123] Next, we will introduce the situational characteristics of the cluster to be matched. The situational characteristics of the cluster to be matched are used for tactical pattern identification, and therefore can include four categories: quantity and structure characteristics, formation and spatial distribution characteristics, coordinated action characteristics, and signal coordination characteristics.

[0124] (1) Quantitative structural characteristics include: total quantity and quantity distribution.

[0125] For example, the quantity distribution is such as "10 attack drones and 2 jamming drones".

[0126] (2) The formation and spatial distribution characteristics reflect the multi-aircraft spatial coordination pattern, which includes: formation mode, spatial spacing and spatial coverage.

[0127] For example, the formation can be dense formation, dispersed formation, trapezoidal formation, or diamond formation.

[0128] For example, a dense collaborative spacing with a spatial distance of less than or equal to 50 meters.

[0129] For example, a dispersed collaborative spacing with a spatial distance of 50-100 meters.

[0130] For example, spatial coverage is the three-dimensional spatial volume occupied by the cluster.

[0131] (3) The characteristics of coordinated action reflect the coordination of multi-machine motion, which include: action synchronization and task division action.

[0132] For example, the synchronization of actions can be such that the time difference between synchronous acceleration, synchronous steering, or synchronous hovering is less than 0.5 seconds.

[0133] For example, the task division actions are a sequence of actions such as "the jamming drone suppresses the enemy for 3 seconds first, and then the attacking drone follows up to break through the defense" or a path coordination of "the lead drone explores the way and the follower drone follows".

[0134] (4) Signal coordination characteristics reflect the coordination of multi-machine communication, including: communication signal correlation, signal priority and anti-interference characteristics.

[0135] For example, communication signal correlation can be achieved by multiple machines using the same frequency band for coordinated frequency hopping signals or communication protocols with the same frame structure.

[0136] For example, signal priority can be a master-slave signal link where the lead drone sends control commands to the follower drone, or the signal strength of the jamming drone can be higher than that of the attacking drone.

[0137] For example, anti-interference features can be anti-interference actions for multi-machine synchronous switching of communication frequency bands.

[0138] The following section provides an illustrative example of a method for extracting cluster situational features. The extraction of cluster situational features relies on cross-dimensional fusion analysis of multi-source sensor data.

[0139] First, the extraction of quantitative structural features.

[0140] Data sources: radar sensors and radio detection sensors.

[0141] Extraction process: First, the total number of drones in the cluster is determined by the target counting module of the radar sensor, which is achieved by counting the number of independent position coordinates captured by the radar; then, the "RF signal function analysis" of the radio detection sensor is used to distinguish the signal type, such as jamming signal or control signal, and to correspond to the functional division of the drones. For example, drones that emit jamming signals correspond to jamming drones, and drones that emit control signals correspond to lead drones. Finally, the number distribution is determined. For example, 8 control signal sources plus 2 jamming signal sources correspond to 8 attack drones plus 2 jamming drones.

[0142] Second, the extraction of formation and spatial distribution characteristics.

[0143] Data sources: radar sensors and photoelectric sensors.

[0144] Extraction process: Based on the three-dimensional position coordinates of multiple drones continuously acquired by radar, the spatial distance between any two drones is calculated. The average distance is used to determine whether the formation is dense or dispersed. At the same time, the geometric shape of the multiple drone positions, such as trapezoid or rhombus, is fitted to determine the formation mode. Combined with wide-angle imaging from photoelectric sensors, the visual shape of the formation is verified. For example, dense formations appear as clusters in the image, while dispersed formations appear as scattered points. The maximum range of cluster position coordinates is calculated to determine the spatial coverage area.

[0145] Third, the extraction of collaborative action features.

[0146] Data sources: radar sensors and photoelectric sensors.

[0147] Extraction process: Using radar velocity vector data, the deviation values ​​of the flight direction and acceleration of multiple drones are calculated. For example, if the deviation of the velocity direction of all drones does not exceed 10°, it is determined to be synchronous turning. At the same time, the time difference of the action trigger is calculated. If the time difference of the simultaneous acceleration of multiple drones does not exceed 0.5 seconds, it is determined to be synchronous acceleration. Through continuous frame images of photoelectric sensors, the temporal relationship of the actions of multiple drones is analyzed. For example, in frame 1, the jamming drone hovers first, and in frame 2, the attacking drone follows up to penetrate. This is determined to be the task division action of "jamming → penetration".

[0148] Fourth, extraction of signal co-features.

[0149] Data source: Radio detection sensor.

[0150] Extraction process: Analyze the signal frequency band, frequency hopping pattern, and frame structure detected by radio to determine the consistency of signals from multiple devices. For example, if multiple devices all use the 2.4GHz frequency band and a fixed 10ms frequency hopping, it is determined to be a communication signal association. Through signal strength analysis, identify the master-slave link. If the strength of a certain signal source is twice that of other signal sources and it sends data to other signal sources, it is determined to be the dominant signal source, corresponding to the leader device. Monitor signal frequency band switching actions. If multiple devices switch frequency bands synchronously, for example, multiple devices simultaneously switch from 2.4GHz to 5.8GHz, it is determined to have anti-interference cooperation characteristics.

[0151] Step 300: Based on the matching results and environmental information, obtain at least one countermeasure from the UAV countermeasure knowledge graph and generate an initial set of countermeasure strategies;

[0152] Understandably, before proceeding to step 100, a drone countermeasure knowledge graph conforming to the embodiments of this invention needs to be constructed in advance. The construction process of the drone countermeasure knowledge graph involves using Natural Language Processing (NLP) and information extraction techniques to continuously extract entities and relationships from drone manufacturer specifications, CVE vulnerability databases, historical case reports, and expert experience manuals to form the entity nodes and relationships in the drone countermeasure knowledge graph. The completed knowledge graph needs to be stored in a graph database (such as Neo4j) to facilitate complex association queries and real-time reasoning.

[0153] The types of entities include drone models, tactical modes, communication protocols, hardware vulnerabilities, software vulnerabilities, countermeasures equipment, geographical environment, and weather conditions, etc.

[0154] Countermeasures include electromagnetic interference devices, navigation deception systems, laser weapons, and net capture systems, among others.

[0155] The types of relationships include protocol adoption, tactical combination, counterability, countermeasure effectiveness, applicable environment, disabled environment, vulnerability, etc. Each relationship can be configured with attribute information. For example, the "countermeasureable" relationship can be set with "effective probability" and "average response time" attributes to improve the semantic expression accuracy and reasoning effectiveness of the knowledge graph.

[0156] The entities mentioned above will be described by example below.

[0157] The drone model refers to the specific drone product model, such as DJI Mini 3, Mavic 3, or Matrice 300RTK. It is the core benchmark for matching countermeasure strategies and is associated with its unique vulnerabilities, communication protocols, and other information.

[0158] Hardware vulnerabilities are defects in the drone's hardware that can be exploited for countermeasures. For example, vulnerabilities related to the frequency of motor control signals are key evidence for deriving countermeasures.

[0159] Software vulnerabilities are defects at the drone software level, such as firmware version-related vulnerabilities, which can be countered through targeted techniques (such as protocol interference).

[0160] Communication protocols are the rules governing communication between the drone and the control unit, such as signal frequency bands, frequency hopping rules, and frame structure identifiers. Different protocols correspond to different anti-jamming schemes.

[0161] Countermeasure equipment refers to hardware or systems used to implement countermeasures, such as electromagnetic interference, navigation deception systems, laser weapons, and net capture systems. Appropriate equipment needs to be selected based on the environment and target characteristics.

[0162] The geographical environment, also known as the environmental information in this embodiment of the invention, refers to the geographical location characteristics of the countermeasure scenario, such as airport airspace, densely populated urban areas, and open suburban areas, which determine the applicable and disabled attributes of the countermeasure equipment.

[0163] Tactical modes refer to the combat or operational methods of drones, such as swarm coordinated attack mode, multi-model mixed flanking penetration mode, and decentralized swarm harassment mode, corresponding to differentiated collaborative countermeasure frameworks.

[0164] The entities mentioned above will be explained next.

[0165] A vulnerability is an edge connecting the "drone model" node and the "hardware vulnerability / software vulnerability" node, indicating a specific vulnerability in a certain drone model. For example, the DJI Mivic 3 has a vulnerability in its motor control signals.

[0166] The adopted protocol is the edge connecting the "UAV Model" node and the "Communication Protocol" node, which represents the exclusive communication rules used by a certain UAV model, such as "Matrice300RTK adopts the 2.4GHz band communication protocol";

[0167] "Can be countered" is an edge connecting the "Drone Model or Tactical Mode" node and the "Countermeasure Equipment" node. It indicates that a certain target, suspicious drone, or tactical mode can be countered by a specific device. The "Can be countered" attribute can be configured with attributes such as effective probability and average response time. For example, DJIMini3 can be countered by a 1.5GHz electromagnetic jammer with an effective probability of 95%.

[0168] Countermeasure effectiveness is the edge connecting the "countermeasure device" node and the "geographical environment or weather environment" node, representing the degree of influence of the environment on the countermeasure effect of the device. For example, the effectiveness of netting drones in an urban environment is 85%.

[0169] The applicable environment is the edge connecting the "countermeasure device" node and the "geographical environment" node, representing the scenario in which the device can be used normally. For example, the applicable environment for a navigation deception system is the airport airspace.

[0170] The prohibited environment is an edge connecting the "countermeasure device" node and the "geographic environment" node, representing the scenario in which the device is prohibited from use. For example, the prohibited environment for laser weapons is a densely populated urban area.

[0171] Tactical combination is an edge connecting the "tactical mode" node and the "countermeasure device collaboration framework" node, representing the multi-device countermeasure sequence / division of labor corresponding to a certain tactical mode, such as the "cluster collaborative attack mode corresponding to the tactical combination of 'suppression → guidance → cleanup'".

[0172] Step 400: Use the target optimization algorithm to process the initial countermeasure strategy set to obtain the target countermeasure strategy.

[0173] Analysis of the beneficial effects of this embodiment:

[0174] 1) Addressing the issue of poor adaptability of traditional UAV countermeasures technologies to complex scenarios, this embodiment first acquires multi-source sensor data and uses this data to determine environmental and situational information (single-drone / cluster, corresponding to different tactical characteristics), achieving a two-dimensional correlation between scenario and tactics, rather than relying on isolated single data.

[0175] For different situations, corresponding features to be matched are extracted. For example, multimodal features are extracted for single drones and cluster situational features are extracted for clusters. Then, similarity matching is performed with entity nodes in the knowledge graph. In this way, even when facing suspicious drones of unknown models, they can be associated with similar entities in the knowledge graph through feature similarity without relying on preset rules, thus improving adaptability to complex scenarios.

[0176] 2) To address the lack of deep semantic support in traditional UAV countermeasures decision-making, a UAV countermeasures knowledge graph is used as the core carrier. By matching the features to be matched with the similarity of the graph entity nodes, the semantic association between nodes can be activated, rather than relying solely on surface numerical calculations. This gives the decision-making a logical support at the semantic level and improves the targeting of countermeasures.

[0177] 3) Addressing the limitation of traditional UAV countermeasures technologies in their reliance on a single approach. Existing technologies often utilize corresponding knowledge graphs for single-path filtering or single-pattern interference output, meaning they can only output a single countermeasure and cannot form a multi-method collaborative solution. This invention, based on matching results and environmental information, obtains "at least one countermeasure" from a knowledge graph, such as electromagnetic interference, navigation deception, or a net-trapping system, constructing an initial set containing multiple methods instead of outputting a single method. Then, the multiple methods in the initial set are optimized to select suitable combinations for specific scenarios, forming a systematic target countermeasure strategy. This avoids the limitations of a single solution and balances countermeasure effectiveness with scenario security.

[0178] In one optional embodiment, the feature to be matched is compared with the features of the target matching entity node in the UAV countermeasure knowledge graph to determine the matching result, including:

[0179] Step 1: According to the similarity calculation formula:

[0180] (1)

[0181] Calculate the similarity score; where, where, Similarity; The feature vector to be matched; ; Match the feature vectors of entity nodes to the target; ; The first feature vector to be matched Individual feature terms; The first feature vector of the target entity node is matched to the feature vector of the target entity node. Individual feature terms; For the first The weight coefficients of each sub-feature term; It is a positive integer greater than 2;

[0182] Step 2: Determine the matching result based on the similarity values ​​of all target matching entity nodes.

[0183] The entity nodes in the drone countermeasures knowledge graph include multiple types of nodes such as drone model nodes, tactical mode nodes, communication protocol nodes, and sub-communication protocol nodes. Each type of node can include multiple corresponding nodes. For example, tactical mode nodes can include multiple nodes such as cluster coordinated attack mode nodes, multi-model mixed flanking penetration mode nodes, and distributed cluster harassment mode nodes. Other types of nodes will not be elaborated upon or given examples.

[0184] In an optional embodiment, when the situation type information is a single-machine situation, the target matching entity node is the UAV model node, and the matching result is the target UAV model node that matches the suspicious UAV; the multimodal features to be matched include radio frequency features, flight features, and optical features;

[0185] Step 2: Based on the similarity scores of all target matching entity nodes, determine the matching results, which may include:

[0186] Compare the similarity values ​​corresponding to all drone model nodes with the first preset similarity threshold;

[0187] If there is at least one drone model node whose similarity value is greater than or equal to the first preset similarity threshold, then the drone model node corresponding to the maximum similarity value is determined as the target drone model node.

[0188] If the similarity scores of all drone model nodes are less than the first preset similarity threshold, the suspected drone model is determined to be unknown. The model is then matched further using the drone countermeasure knowledge graph's association relationships according to a preset priority order to obtain the target drone model node. The preset priority order is as follows: communication protocol nodes, sub-communication protocol nodes, hardware feature nodes, and software feature nodes.

[0189] The sum of the coefficients for the four features—radio frequency (RF) feature, flight feature, optical feature, and communication protocol node—is 1. For example, the RF feature weight coefficient is 0.4, the flight feature weight coefficient is 0.3, the optical feature weight coefficient is 0.2, and the communication protocol feature weight coefficient is 0.1.

[0190] For example, drone model nodes include known models that are known in the market and included in the drone countermeasure knowledge graph, such as the DJI Mini3 node, Mavic3 node, and Matrice300RTK node. The specific drone models mentioned in this document are for illustrative purposes only and do not constitute any reference to or limitation of actual product functionality.

[0191] The matching process described above is illustrated by example.

[0192] I. Prerequisites.

[0193] The drone model nodes include: DJIMini3 node, Mavic3 node, and Matrice300RTK node.

[0194] The communication protocol node includes:

[0195] 2.4GHz civilian protocol node. Characteristics: signal frequency band 2.4GHz, fixed frequency hopping pattern, frame structure identifier 0x01.

[0196] 5.8GHz civilian protocol node. Characteristics: signal frequency band 5.8GHz, fixed frequency hopping pattern, frame structure identifier 0x02.

[0197] High-speed frequency hopping protocol node. Characteristics: signal frequency band 6.0GHz, random frequency hopping pattern, frame structure identifier 0x03.

[0198] Hardware vulnerability nodes include:

[0199] Vulnerable node in motor control signal. Characteristic: Motor control signal frequency 200Hz.

[0200] Vulnerable node in the power system. Characteristic: Power control signal frequency 150Hz.

[0201] The first preset similarity threshold is 85%.

[0202] Feature weighting coefficients: radio frequency feature weighting coefficient 0.4, flight feature weighting coefficient 0.3, optical feature weighting coefficient 0.2, and communication protocol feature weighting coefficient 0.1.

[0203] In specific implementation method 1, the following multimodal features of a suspicious drone to be matched are collected by multi-source sensors: radio frequency characteristics: 2.4GHz communication band, fixed frequency hopping interval of 10ms; flight characteristics: maximum speed 15.5m / s, hovering error 0.4m; optical characteristics: fuselage size 31cm×21cm, white fuselage, 4 propellers. From the above, it can be analyzed that the radio frequency characteristics include 2 sub-features, the flight characteristics include 2 sub-features, and the optical characteristics include 4 sub-features. The 8 sub-features from these three features form the following feature vector to be matched. .

[0204] Feature vector to be matched ;

[0205] Feature vector of DJIMini3 node ;

[0206] Feature vector of Mavic3 node ;

[0207] Feature vectors of Matrice300RTK nodes .

[0208] Substituting into the similarity calculation formula (1), we obtain the following results:

[0209] The similarity score with the DJIMini3 node is 99.71%.

[0210] The similarity value with the Mavic3 node is 98.67%.

[0211] The similarity score with the Matrice300RTK node is 98.72%.

[0212] If the similarity scores of the three drone models are all greater than 85%, then the DJIMini3 node with the highest similarity score among all model nodes is identified as the model of the suspicious drone. In other words, the target drone model node is finally determined to be the DJIMini3 node.

[0213] Analysis of the beneficial effects of this embodiment:

[0214] 1) Addressing the shortcomings of existing countermeasure strategies, which are rigid and lack intelligent association, this embodiment does not rely on fixed rules for judgment when faced with a suspicious drone. Instead, it performs similarity matching between the drone's multimodal features (RF, flight, and optical features) and target matching entity nodes in the drone countermeasure knowledge graph. As in Implementation 1, the similarity calculation formula is used to obtain similarity values ​​with each known drone model. If the similarity is greater than or equal to a preset threshold, the model can be determined; if it is less than the threshold and determined to be an unknown model, it can continue matching according to a preset priority order through the association relationships in the knowledge graph. This method does not rely on pre-set rigid rules, can handle drones of unknown models, and achieves intelligent matching through feature association and graph relationships, effectively solving the problems of rule-based systems being unable to handle drones of unknown models and the isolation and poor adaptability of rules.

[0215] 2) Addressing the deficiency of existing countermeasures in neglecting deep semantic relationships: The knowledge graph in this embodiment contains various entity nodes and relationships. In the process of determining the drone model, when matching unknown drone models according to a preset priority order, communication protocol nodes, etc., are involved. If a suspicious drone has similar relationships with known drone models in terms of communication protocols, the possible model can be inferred using this semantic association.

[0216] 3) Existing technologies include solutions that analyze the dynamic monitoring data stream of target UAVs to identify flight trajectory features and electromagnetic signal spectra, construct a three-dimensional spatial threat situation model to assess the threat level, and generate a dynamic decision factor set based on this to match the basic countermeasure plans in the contingency plan knowledge graph. The key is to generate countermeasure plans based on dynamic monitoring data and threat assessment. However, unlike this embodiment, it does not construct a detailed knowledge graph of entities and relationships for UAV model identification, nor does it determine the UAV model through multimodal feature similarity matching or by utilizing knowledge graph relationships.

[0217] This embodiment employs a deep association matching method using a knowledge graph when dealing with unknown drone models, a feature not found in existing technologies. In accurately identifying unknown drone models, this embodiment leverages the rich entity relationships within the knowledge graph, achieving greater precision and intelligence than existing technologies that rely solely on monitoring data and pre-defined matching. This effectively solves the technical problem of existing technologies' inability to accurately identify drone models by deeply utilizing semantic relationships.

[0218] In one optional embodiment, the preset priority order is as follows: communication protocol node, sub-communication protocol node, hardware feature type node, and software feature type node.

[0219] The matching process continues according to a preset priority order to determine the target drone model node, including:

[0220] Extract multiple communication protocol sub-features from the multimodal features to be matched and construct a communication protocol feature vector; the multiple communication protocol sub-features include signal frequency band, frequency hopping pattern and frame structure identifier (e.g., the dedicated encoding of the protocol frame header).

[0221] Using a similarity calculation formula, the feature vectors of communication protocols are matched with the preset feature vectors of multiple communication protocol nodes to obtain the first similarity result of each communication protocol node.

[0222] If there is at least one communication protocol node whose first similarity result is greater than or equal to the second preset similarity threshold, then the communication protocol node corresponding to the maximum value of the first similarity result is determined as the target communication protocol node, and the node of the known drone model corresponding to the target communication protocol node is determined as the target drone model node by the adoption protocol relationship between the communication protocol node and the drone model node.

[0223] If the similarity results of all communication protocol nodes are less than the second preset similarity threshold, the similarity protocol relationship between communication protocol nodes in the UAV countermeasure knowledge graph is invoked. Based on the filtering conditions, multiple sub-communication protocol nodes associated with the target communication protocol node are selected for further similarity matching to obtain multiple second similarity results. The filtering conditions are that the protocol similarity attribute value is greater than or equal to the third preset similarity threshold.

[0224] If at least one second similarity result is greater than or equal to the second preset similarity threshold, the corresponding target drone model node is obtained;

[0225] If all second similarity results are less than the second preset similarity threshold, then matching continues according to the preset priority order to determine the target UAV model node. Hardware features are extracted from the multimodal features to be matched and a hardware feature vector is constructed for the next similarity matching step. The matching process is similar to that of the communication protocol and will not be elaborated here.

[0226] If the hardware features cannot be matched, the matching process continues using software features, including firmware version identifiers and other features related to the drone software, until a match is found.

[0227] For example, the second preset similarity threshold is 80%, and the third preset similarity threshold is 75%. The communication protocol feature weight is 0.1, wherein: the signal frequency band sub-feature weight is 0.04; the frequency hopping rule sub-feature weight is 0.03; and the frame structure identifier sub-feature weight is 0.03.

[0228] Extracting the communication protocol sub-features from the multimodal features to be matched: signal frequency band 6.0GHz, random frequency hopping pattern, and frame structure identifier 0x03; then the communication protocol feature vector can be constructed as follows: .

[0229] The pre-defined feature vectors of high-speed frequency hopping protocol nodes in the knowledge graph are: The preset feature vector of the 2.4GHz civilian protocol node is: .

[0230] Using the similarity calculation formula (1), the similarity was matched with the preset feature vector of the high-speed frequency hopping protocol node in the knowledge graph. The similarity result was 83%, which is greater than the second preset similarity threshold of 80%. Through the association relationship between "communication protocol and UAV model", the target UAV model node was determined to be "a cluster of known civilian UAV models associated with the high-speed frequency hopping protocol".

[0231] It should be noted that matching other communication protocol nodes in the drone countermeasure knowledge graph with this feature vector is not applicable. For example, the similarity value of the 2.4GHz civilian stable protocol node after feature vector matching is far below 80%, mainly because the frequency band 2.4GHz ≠ 6.0GHz, the frequency hopping pattern (fixed ≠ random), and the frame structure identifier 0x01 ≠ 0x03, which are significantly different. Similarly, the similarity of the Bluetooth classic protocol node after matching is also below 80%, because the frequency band 2.4GHz ≠ 6.0GHz and the frame structure identifier does not match 0x03. Therefore, neither of these can be used to determine the target drone model node.

[0232] Analysis of the beneficial effects of this embodiment:

[0233] I. Existing rule-based systems, due to their "isolated rules and reliance on presets," are unable to handle novel threat scenarios involving unknown drone models or multiple interconnected factors. This embodiment overcomes this limitation through the following design:

[0234] 1) Break through the fixed rules of model and countermeasure, and establish multi-dimensional correlation and matching.

[0235] When the suspected drone model is unknown (the similarity of all known models is below a first threshold), the implementation does not rely on the fixed rule of "IF - model XTHEN countermeasure Y", but instead prioritizes matching based on lower-level features such as communication protocols, hardware, and software. For example, in an exemplary scenario, an unknown drone model cannot be matched with known model nodes due to its radio frequency characteristics. However, by extracting communication protocol sub-features such as signal frequency bands, frequency hopping patterns, and frame structure identifiers, a feature vector is constructed and matched with "high-speed frequency hopping protocol nodes" in the knowledge graph. Ultimately, through the association between communication protocols and drone models, a cluster of known civilian drone models associated with high-speed frequency hopping protocols is identified, achieving indirect matching of unknown models and solving the problem of not being able to handle unknown models.

[0236] 2) To address complex scenarios involving multiple factors and improve adaptability.

[0237] The matching logic in this embodiment does not rely on a single feature in isolation (such as only looking at the model number), but rather connects model number, communication protocol, hardware, and software in a multi-dimensional way. If the communication protocol matching fails, the "similar protocol relationship" can be used to filter sub-communication protocol nodes, and then hardware and software features can be matched. This forms a dynamic logic that automatically triggers the next layer of association if one layer of matching fails. This design can cope with new threat scenarios where unknown models are associated with new communication protocols or special hardware vulnerabilities, avoiding the problems of isolated and poorly adaptable existing technical rules.

[0238] Second, address the shortcomings of neglecting deeper semantic relationships.

[0239] Existing drone countermeasures often rely solely on numerical calculations, failing to grasp semantic relationships such as "drones A and B use the same communication protocol," thus hindering the derivation of associated countermeasure strategies. This embodiment precisely addresses this deficiency through the semantic association design of a knowledge graph:

[0240] 1) Explore the semantic associations between "entity-relationships" to replace simple numerical matching.

[0241] The core of this implementation is to leverage predefined semantic relationships in the drone countermeasure knowledge graph to achieve cross-entity reasoning. For example, through the "adopted protocol" relationship, the binding logic between the "communication protocol node" and the "drone model node" is clearly defined. When a target communication protocol is matched, it can be directly associated with the corresponding known model, rather than simply calculating the similarity of feature values.

[0242] If the target communication protocol fails to match, the similarity protocol relationship can be invoked to filter out sub-communication protocol nodes with a protocol similarity ≥ the third threshold (such as derived protocols from the same source as the target protocol), and then the model can be associated based on the sub-protocol. This design can understand the semantic connotation of the same-source protocol, and thus deduce that effective countermeasures against drones with the same-source protocol can be reused, making up for the shortcomings of existing technologies that only look at numerical values ​​and do not understand semantics.

[0243] 2) The transferability of countermeasure logic is supported by semantic relationships.

[0244] In an exemplary scenario, although an unknown model of drone cannot be directly matched with a known model, it can reuse the countermeasure strategy corresponding to the protocol by adopting the protocol relationship because it matches the high-speed frequency hopping protocol. This is based on the semantic reasoning of "same source protocol → reusable countermeasure strategy". Existing technologies cannot achieve this cross-model countermeasure strategy migration due to the lack of semantic relationship recognition.

[0245] In one alternative embodiment, the relationships in the drone countermeasure knowledge graph include at least countermeasureable relationships, applicable environment relationships, prohibited environment relationships, and environmental impact on the effectiveness of countermeasures.

[0246] When the situation type information is a single-machine situation, step 300: Based on the matching results and environmental information, obtain at least one countermeasure from the UAV countermeasure knowledge graph and generate an initial countermeasure strategy set, including:

[0247] By leveraging counter-relationships, all countermeasures associated with the matching results are obtained, forming a first countermeasures pool.

[0248] By applying environmental relationships, countermeasures containing environmental information are selected from the first countermeasures pool to obtain the second countermeasures pool;

[0249] By disabling environmental relationships, the third countermeasure pool is obtained by excluding countermeasures in the second countermeasure pool that disable environmental information.

[0250] By analyzing the relationship between environmental impact and countermeasure effectiveness, the association attributes between environmental information and various countermeasures in the third countermeasure pool (such as the effectiveness attenuation coefficient of airport airspace against 1.5GHz electromagnetic interference) are retrieved from the knowledge graph of UAV countermeasures. Combined with the original effectiveness of each countermeasure, the corrected effectiveness is calculated.

[0251] Based on the modified effectiveness of each means in the third countermeasure pool, all countermeasures are combined and the expected countermeasure effect of each combination is marked to obtain the initial countermeasure strategy set for the single-machine situation.

[0252] Specifically, combining all countermeasures means combining all possible countermeasures: including combinations of two countermeasures, and combinations of three or more countermeasures used in concert.

[0253] For example, combining the original effectiveness of each countermeasure, the corrected effectiveness can be calculated as follows: the original effectiveness of the 1.5GHz electromagnetic interference device is 90%, the attenuation coefficient is 0.8, and the corrected effectiveness is 72%.

[0254] For example, the expected countermeasure effect is: the countermeasure success rate is the average effectiveness of the two methods after correction.

[0255] In Specific Implementation 2, the target drone model is DJIMini3Pro, and the target area is a mountainous tourist area.

[0256] Countermeasureable Relationships: The DJIMini3Pro node has countermeasureable relationships with multiple countermeasure device nodes, including "1.8GHz electromagnetic jammer, GNSS deception device, acoustic wave repeller, and infrared jammer".

[0257] Applicable environments: 1.8GHz electromagnetic interference device (applicable environment: mountainous areas, suburbs); GNSS deception device (applicable environment: mountainous areas, open areas); acoustic repellent device (applicable environment: mountainous areas, scenic areas); infrared jammer (applicable environment: mountainous areas, forest areas).

[0258] Disabled environments: 1.8GHz electromagnetic interference device (disabled environment: none); GNSS spoofing device (disabled environment: none); acoustic repellent device (disabled environment: residential area); infrared jammer (disabled environment: none).

[0259] Environmental impact countermeasures effectiveness: In mountainous tourist areas, the effectiveness attenuation coefficient of 1.8GHz electromagnetic interference device is 0.9 (original effectiveness 92%), the effectiveness attenuation coefficient of GNSS deception device is 0.95 (original effectiveness 95%), the effectiveness attenuation coefficient of acoustic wave repellent device is 0.85 (original effectiveness 88%), and the effectiveness attenuation coefficient of infrared interference device is 0.8 (original effectiveness 85%).

[0260] Step (1) Construct the first countermeasures pool. The first countermeasures pool consists of: {1.8GHz electromagnetic jammer, GNSS deception device, acoustic deflector, and infrared jammer}.

[0261] Step (2) Select the second countermeasure pool by applying the environmental relationship.

[0262] The second countermeasure pool includes: {1.8GHz electromagnetic interference device, GNSS deception device, acoustic deterrent device, infrared jammer}.

[0263] Step (3) Filter the pool of third countermeasures by disabling environmental relationships.

[0264] The exclusion of acoustic deterrent devices from prohibited environments includes residential areas surrounding scenic spots (a small number of residential areas exist within the scenic area, triggering the prohibition condition). The third countermeasure pool includes: {1.8GHz electromagnetic jammers, GNSS deception devices, and infrared jammers}.

[0265] Step (4): Correct the effectiveness by using the relationship between environmental impact and countermeasure effectiveness.

[0266] Retrieve the association attributes between "mountainous tourist attractions" and various countermeasures in the third countermeasure pool from the drone countermeasure knowledge graph: namely, the effectiveness attenuation coefficient of mountainous tourist attractions against 1.8GHz electromagnetic jammers is 0.9, against GNSS spoofing devices is 0.95, and against infrared jammers is 0.8. Then, combine the original effectiveness of each countermeasure to calculate the corrected effectiveness.

[0267] 1.8GHz electromagnetic interference device: 92% × 0.9 = 82.8%;

[0268] GNSS spoofing device: 95% × 0.95 = 90.25%;

[0269] Infrared jammer: 85% × 0.8 = 68%.

[0270] Step (5): Generate an initial set of countermeasure strategies.

[0271] Based on the modified effectiveness of each countermeasure in the third countermeasure pool, all possible combinations of countermeasures (including two-measure combinations and three-measure combinations) are performed, and the expected countermeasure effect of each combination is marked, resulting in:

[0272] Initial countermeasure strategy set: {(GNSS deception device, 1.8GHz electromagnetic jammer), (GNSS deception device, infrared jammer), (1.8GHz electromagnetic jammer, infrared jammer), (GNSS deception device, 1.8GHz electromagnetic jammer, infrared jammer)}.

[0273] Expected results for each combination:

[0274] Combination 1: GNSS spoofing device and 1.8GHz electromagnetic interference device. Countermeasure success rate ≥88% (average effectiveness of both methods), response time ≤3.5s, no risk of interference with electronic devices of tourists in scenic areas.

[0275] Combination 2: GNSS spoofing device and infrared jammer. Countermeasure success rate ≥79.1% (average effectiveness of both methods), response time ≤4s, no risk of interference with electronic devices of tourists in scenic areas.

[0276] Combination 3: 1.8GHz electromagnetic interference device and infrared interference device. Countermeasure success rate ≥75.4% (average effectiveness of both methods), response time ≤4.5s, no risk of interference to electronic devices of tourists in scenic areas.

[0277] Combination 4: GNSS spoofing device, 1.8GHz electromagnetic jammer and infrared jammer.

[0278] Expected results: Countermeasure success rate ≥80.4%, response time ≤5s, no risk of interference from electronic devices of tourists in scenic areas, and more comprehensive coverage of countermeasure scenarios.

[0279] Analysis of the beneficial effects of this embodiment:

[0280] 1) The implementation does not directly invoke the preset fixed rule of "model-countermeasure", but instead uses a four-layer association filtering of "countermeasure-possible relationship → applicable environment relationship → prohibited environment relationship → environmental impact on countermeasure effectiveness relationship" to deeply bind the drone model with environmental information. This design can cope with scenarios where "the same model requires different countermeasures in different environments", avoiding the problem of existing technical rules being isolated and unable to associate multiple factors.

[0281] 2) The implementation plan explicitly outlines all possible combinations of countermeasures, including two, three, or more measures working together, covering different collaborative scenarios. Each combination strategy is clearly labeled with its expected effect. This labeling allows the advantages of the collaborative strategy to be quantified, avoiding the limitations of existing technologies that only output the means without explaining the effects. It also provides a basis for subsequent multi-objective optimization, ensuring the practicality of the collaborative solution.

[0282] 3) The core of the implementation relies on predefined semantic relationships in the knowledge graph: for example, the applicable environment relationship indicates that countermeasures can be used in a certain environment, and the environmental impact on countermeasure effectiveness relationship indicates the attenuation logic of the environment on the countermeasure effect. These are not based solely on numerical calculations as in the prior art. For example, the semantic relationship of "laser weapon - prohibited environment - airport" clearly states the logic that "laser weapons are prohibited in the airport environment," rather than judging it through surface numerical values; the environmental impact on countermeasure effectiveness relationship reflects the deep association that the airport environment will reduce the effectiveness of electromagnetic jammers, which is precisely the semantic connotation that the prior art cannot understand.

[0283] In an optional embodiment, step 400: The initial countermeasure strategy set is processed using a target optimization algorithm to obtain a target countermeasure strategy, including:

[0284] A multi-objective optimization set is defined, which includes benefit-oriented and cost-oriented optimization objectives. Benefit-oriented optimization objectives include countermeasure success rate and collaborative efficiency. Cost-oriented optimization objectives include response time, economic cost, and collateral damage. Countermeasure success rate is derived from the "environmental impact of countermeasure effectiveness" relationship in the knowledge graph. Response time is derived from the "response time" attribute of the "can be countermeasured" relationship. Collaborative efficiency is calculated only for cluster situations, i.e., it is normalized according to the temporal linkage deviation of countermeasures. In single-machine situations, the normalized value of collaborative efficiency is set to 1.

[0285] Formula used:

[0286] (2)

[0287] The original data of the benefit-oriented optimization objectives are normalized to obtain the normalized values ​​of each benefit-oriented optimization objective; among them, For the first The normalized value of a benefit-oriented optimization objective; For the first The original data for each benefit-oriented optimization objective; For the first The minimum baseline value for each benefit-oriented optimization objective; For the first The maximum baseline value of each benefit-oriented optimization objective;

[0288] Formula used:

[0289] (3)

[0290] The original data for cost-type optimization objectives are normalized to obtain normalized values ​​for each cost-type optimization objective; among them, For the first The normalized value of a cost-type optimization objective; For the first The original data for each cost-based optimization objective; For the first The minimum baseline value for a cost-based optimization objective; For the first The maximum baseline value for each cost-based optimization objective;

[0291] The weight coefficients of each optimization objective are determined based on the situation type information;

[0292] Formula used:

[0293] (4)

[0294] Calculate the overall score for each initial countermeasure strategy in the initial countermeasure strategy set; where, For the overall score; For the first The weight coefficients of the first optimization objective; if the first... If the optimization objective is a benefit-oriented optimization objective, then If the first If the optimization objective is a cost-based optimization objective, then ;

[0295] Select the initial set of countermeasures with the highest overall score as the target countermeasure.

[0296] Specifically, regarding the determination of weight coefficients for each optimization objective based on situation type information, a differentiated set of weight coefficients is preset to account for the characteristics of single-machine and cluster situations, with the sum of all weight coefficients in each set being 1. The weight coefficients are used to quantify the decision priority of different optimization objectives under corresponding situations. For example, the weight coefficients can be set according to the following rules:

[0297] For single-machine scenarios, the focus is on eliminating a single threat quickly, accurately, and at low cost. The weight coefficients for each optimization objective are set as follows:

[0298] Reverse conversion power Response time Economic costs Incidental damage costs Collaboration efficiency cost .

[0299] Regarding the cluster situation, the focus is on responding to complex coordinated attacks and ensuring the coordinated effectiveness of the countermeasure system itself. The weight coefficients of each optimization objective are set as follows:

[0300] Reverse conversion power Response time Economic costs Incidental damage costs Collaboration efficiency cost .

[0301] The beneficial effects of this embodiment are analyzed as follows: 1) The initial set of countermeasure strategies (multi-means combination) is used as the optimization object to select the optimal collaborative solution to replace the output of a single means. 2) The optimization set covers five core dimensions, including countermeasure success rate and response time, to achieve multi-dimensional balance. 3) The evaluation logic of countermeasure schemes in the prior art only focuses on whether the threat can be effectively countered (such as ensuring the timeliness of countermeasures through trajectory prediction deviation and matching the threat intensity through electromagnetic suppression priority), without considering the economic cost of the countermeasure process. This makes it unable to cope with the detailed requirements of controlling countermeasure costs in low-cost scenarios. This invention incorporates economic cost into the optimization dimension, which can dynamically balance the effect and cost: for example, in specific implementation method 2, although combination 3 has the lowest economic cost, it is eliminated due to its low countermeasure success rate; although combination 1 has a higher cost, it becomes the optimal solution due to its high success rate and no collateral damage. This decision-making logic can adapt to the differentiated scenarios where high-value areas (such as airports) can accept higher costs to ensure effectiveness, and low-value areas (such as suburban open land) need to control costs. 4) Current assessments of drone countermeasure technologies only focus on the "suppression effect of countermeasures on drone threats," such as avoiding interference in protected frequency bands through spectrum avoidance coefficients. They do not consider collateral damage to the surrounding environment, such as the impact on civilian equipment and personnel, and cannot address the detailed needs of sensitive environments like scenic spots and residential areas where secondary risks need to be mitigated. This invention, through the collateral damage dimension, can precisely control countermeasure safety, avoiding secondary problems such as potential interference with civilian equipment and personnel safety risks in sensitive environments. However, it cannot address the detailed needs of such safety control. 5) While existing multi-dimensional assessments of drone countermeasure solutions involve multiple parameters, they lack a logical correlation between dimensional weights and scenario situations. They can only be assessed according to fixed logic and cannot address the detailed needs of different countermeasure target priorities within the same scenario.

[0302] This invention achieves detailed priority adaptation of countermeasure targets in multiple scenarios by setting dimensional weights according to situation type and combining five major dimensions to realize dynamic priority adaptation.

[0303] Furthermore, regarding cluster situation, in one optional embodiment, when the situation type information is cluster situation, the target matching entity node is a tactical mode node, and the matching result is the target tactical mode node.

[0304] The situational characteristics of the clusters to be matched include quantitative and structural characteristics, formation characteristics, coordinated action characteristics, and communication and coordination characteristics;

[0305] Tactical mode nodes include at least cluster-coordinated attack mode nodes, multi-type hybrid flanking penetration mode nodes, and distributed cluster harassment mode nodes;

[0306] Based on the similarity scores of all target matching entity nodes, the matching results are determined, including:

[0307] The tactical mode node corresponding to the maximum similarity value is determined as the target tactical mode node.

[0308] It is important to emphasize that, unlike the matching process for single-machine situations, the matching process for cluster situations does not involve comparing with a threshold. Instead, after obtaining multiple similarity values, the tactical mode node corresponding to the maximum similarity value is directly identified as the target tactical mode node.

[0309] In Implementation Method 3, assuming a suspicious drone cluster is detected in a certain area, the multi-source sensors extract the following characteristics of the cluster to be matched: the quantity and structure characteristics are "12 drones: 8 attack drones + 4 jamming drones", the formation characteristics are "dense diamond formation with a spatial distance of ≤30 meters", the coordinated action characteristics are "synchronous turning, the jamming drones suppress the attack drones for 3 seconds and then the attack drones follow up and penetrate", and the communication coordination characteristics are "multiple drones use 2.4GHz synchronous frequency hopping signals, and there is one lead drone sending control commands to other drones".

[0310] The similarity scores of the aforementioned features to be matched with the tactical pattern nodes "cluster cooperative attack pattern node, multi-model mixed flanking penetration pattern node, and distributed cluster harassment pattern node" in the UAV countermeasure knowledge graph were calculated, and the similarity scores were 92%, 65%, and 48%, respectively. According to the rule of directly taking the node corresponding to the maximum similarity score, the target tactical pattern node was determined to be a cluster cooperative attack pattern node.

[0311] The beneficial effects of this embodiment are as follows: 1) It overcomes the limitations of single-machine model matching and adapts to the characteristics of cluster threats. Unlike single-machine situational awareness which focuses on UAV model identification, this embodiment extracts four types of exclusive features for cluster situations and matches them with tactical pattern nodes in the knowledge graph to accurately locate the attack intent of the cluster, avoiding the shortcomings of existing technologies that can only identify known single-machine models and cannot determine cluster tactics. 2) It simplifies the matching logic and improves the response speed to cluster threats. The embodiment adopts a matching rule that directly takes the maximum similarity (without threshold comparison), which is more suitable for the need for rapid decision-making in cluster threats compared to the complex process of single-machine situational awareness. This reduces decision-making time and buys time for subsequent countermeasures.

[0312] In one optional embodiment, the relationships in the UAV countermeasure knowledge graph also include tactical combination relationships; when the situation type information is a cluster situation, based on the matching results and environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated, including:

[0313] Invoke tactical combination relationships to obtain the countermeasure coordination framework associated with the target tactical mode node;

[0314] For each link in the collaborative framework of countermeasures, candidate countermeasures for each link are obtained from the drone countermeasures knowledge graph;

[0315] Based on the applicable environment, the candidate methods for each stage are selected, and the applicable environment includes environmental information, to obtain the set of first countermeasures for each stage;

[0316] Based on the relationship of the prohibited environment, the means that contain environmental information in the prohibited environment in the first countermeasures set of each stage are excluded, and the second countermeasures set of each stage is obtained;

[0317] Based on the relationship of the effectiveness of environmental impact countermeasures, the correlation attributes of each measure in the set of second countermeasures at each stage are retrieved, the original effectiveness of each measure is corrected, and the corrected effectiveness of each measure is obtained.

[0318] By combining the effectiveness of each method, select methods from the second set of countermeasures in each stage and combine them in sequence according to the countermeasure coordination framework, and mark the expected countermeasure effect of each combination to generate an initial set of countermeasure strategies for the cluster situation.

[0319] Specifically, the countermeasure coordination framework defines the multi-stage, time-sequential combination logic of countermeasures against specific tactical patterns. For example, the coordination framework for a "cluster coordinated attack pattern" can be three stages: "suppression → guidance → cleanup," where the "suppression" stage focuses on using large-scale electromagnetic interference to disrupt communications.

[0320] The "guidance" phase uses navigation deception to lure the cluster to a safe area; the "cleanup" phase uses net-capture drones or laser weapons to physically eliminate the remaining or penetrating drones.

[0321] For example, the collaborative framework for countermeasures is: suppression → guidance → cleanup.

[0322] For example, the collaborative framework for countermeasures is: key protection → interference and decoy → residual investigation.

[0323] For example, the coordinated framework for countermeasures is: low-altitude interception → electronic suppression → physical destruction.

[0324] For example, the initial countermeasures set for a cluster situation includes the following three combinations:

[0325] Combination 1: Short-range high-power electromagnetic jammer (suppression) + navigation deception device (guidance) + net-capture drone (cleanup).

[0326] Combination 2: Directional radio frequency jammer (suppression) + navigation deception device (guidance) + net-capture drone (cleanup).

[0327] Combination 3: Directional radio frequency jammer (suppression) + signal decoy device (guidance) + net-capture drone (cleanup).

[0328] For example, the optimization process of the initial countermeasure strategy set for the cluster situation is illustrated by way of example.

[0329] After obtaining the initial set of countermeasure strategies for the cluster situation, the target optimization algorithm is used to process the initial set of countermeasure strategies to obtain the target countermeasure strategy for the cluster situation. This process can be referred to the optimization process for single-machine situations, and will not be elaborated here.

[0330] The beneficial effects of this embodiment are as follows: 1) The embodiment ensures that the countermeasure logic matches the cluster tactics by calling a collaborative framework of countermeasures bound to the target tactical mode through tactical combination relationships. This solves the shortcomings of existing technologies in terms of single countermeasure strategies and lack of collaboration. 2) The embodiment ensures that the means of each link in the collaborative framework are adapted to the scenario through a three-layer screening of "applicable environment → disabled environment → environmental impact on countermeasure effectiveness". This solves the problem that existing countermeasure technologies only focus on the threat situation and ignore the impact of the environment on cluster countermeasures.

[0331] In one optional embodiment, the intelligent generation method for drone countermeasure strategies further includes:

[0332] Based on the target countermeasure strategy, the corresponding countermeasure execution equipment is controlled to perform countermeasure tasks against suspicious drones, and the countermeasure process data is recorded in real time; including but not limited to the execution results of the countermeasure task (e.g., whether the suspicious drone was successfully driven away or captured), the countermeasure task time, the operating parameters of the countermeasure equipment (e.g., the operating frequency band of the electromagnetic jammer, the signal strength of the navigation deception device), and the dynamic feedback of the suspicious drone during the countermeasure process (e.g., changes in flight trajectory, communication signal interruption).

[0333] If the countermeasure process data shows that the effectiveness of the target countermeasure strategy is higher than a preset threshold, then the attributes of the countermeasure effectiveness relationship between the target countermeasure strategy and the corresponding drone model in the drone countermeasure knowledge graph are updated, such as increasing the attribute value of the effectiveness probability and correcting the attribute value of the average response time.

[0334] The beneficial effects of this embodiment are that it solves the shortcomings of existing drone countermeasure technologies, which rely on human experience and cannot be dynamically optimized. 1) It autonomously collects empirical countermeasure data, replacing human experience input. This embodiment does not rely on experts manually recording cases, but automatically records the entire countermeasure process data in real time, including: execution results, such as whether the drone was successfully driven away / captured, time consumption; equipment parameters; target feedback (such as changes in drone flight trajectory, communication signal interruption). This data directly reflects the actual effect of the countermeasure strategy, providing objective evidence for system learning, eliminating the need for experts to manually compile cases, and breaking the limitations of experience dependence. 2) It automatically updates knowledge graph attributes, enabling the system to evolve autonomously. When the countermeasure process data shows that "the effectiveness of the countermeasure is higher than the preset threshold" (e.g., successful capture and less than 3 minutes), the system automatically updates the association attribute of "target countermeasure strategy - corresponding drone model" in the knowledge graph without manual intervention. For example, if the countermeasure success rate of "1.5GHz electromagnetic interference combined with navigation deception" against "DJIMini3" reaches 95%, which is higher than the threshold of 90%, then the "effective probability" attribute value of the "can be countermeasured" relationship between the two is increased, for example, from the original 88% to 92%. This dynamic update allows the system to autonomously optimize its knowledge reserves as countermeasure practices progress, cope with the rapid iteration of drone technology, and avoid the problem of the existing rule base becoming outdated and invalid.

[0335] See Figure 2 The present invention also provides an intelligent generation device for drone countermeasure strategies, which may include:

[0336] The acquisition module 210 is used to acquire multi-source sensor data of the target area and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation;

[0337] Matching module 220 is used to extract features to be matched corresponding to situation type information from multi-source sensor data, and to perform similarity matching between the features to be matched and the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the features to be matched are multimodal features to be matched corresponding to single-unit situation or cluster situation features to be matched corresponding to cluster situation.

[0338] The generation module 230 is used to obtain at least one countermeasure from the UAV countermeasure knowledge graph based on the matching results and environmental information and generate an initial set of countermeasure strategies.

[0339] The optimization module 240 is used to process the initial countermeasure strategy set using a target optimization algorithm to obtain the target countermeasure strategy.

[0340] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, the disclosure, and the appended claims in carrying out the claimed invention. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results. Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Obviously, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A method for intelligently generating anti-drone strategies, characterized in that, include: Acquire multi-source sensor data of the target area, and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation; The matching features corresponding to the situation type information are extracted from the multi-source sensor data, and the matching features are matched with the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the matching features are either multimodal features corresponding to the single-UAV situation or cluster situation features corresponding to the cluster situation. Based on the matching results and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated. The initial countermeasure strategy set is processed using a target optimization algorithm to obtain the target countermeasure strategy; The entity nodes in the drone countermeasures knowledge graph include at least tactical pattern nodes; When the situation type information is a cluster situation, the target matching entity node is a tactical mode node, and the matching result is the target tactical mode node; The situational characteristics of the cluster to be matched include quantitative and structural characteristics, formation characteristics, cooperative action characteristics, and communication and coordination characteristics; The tactical mode nodes include at least cluster-coordinated attack mode nodes, multi-type hybrid flanking penetration mode nodes, and distributed cluster harassment mode nodes; Based on the similarity scores of all target matching entity nodes, the matching results are determined, including: The tactical mode node corresponding to the maximum similarity value is determined as the target tactical mode node.

2. The intelligent generation method for UAV countermeasure strategies according to claim 1, characterized in that, The features to be matched are compared with the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching results, including: According to the similarity calculation formula: ； Calculate the similarity score for each target matching entity node; where, Similarity; The feature vector to be matched; ; Match the feature vectors of entity nodes to the target; ; The first feature vector to be matched Individual feature terms; The first feature vector of the target entity node is matched to the feature vector of the target entity node. Individual feature terms; For the first The weight coefficients of each sub-feature term; It is a positive integer greater than 2; The matching result is determined based on the similarity values ​​of all target matching entity nodes.

3. The intelligent generation method for UAV countermeasure strategies according to claim 2, characterized in that, The entity nodes in the drone countermeasure knowledge graph also include drone model nodes; when the situation type information is the single-drone situation, the target matching entity node is the drone model node, and the matching result is the target drone model node that matches the suspected drone; the multimodal features to be matched include radio frequency features, flight features, and optical features; Based on the similarity scores of all target matching entity nodes, the matching results are determined, including: Compare the similarity values ​​corresponding to all drone model nodes with the first preset similarity threshold; If there is at least one drone model node whose similarity value is greater than or equal to the first preset similarity threshold, then the drone model node corresponding to the maximum similarity value is determined as the target drone model node. If the similarity values ​​corresponding to all drone model nodes are less than the first preset similarity threshold, the model of the suspicious drone is determined to be an unknown model, and matching continues according to a preset priority order to determine the target drone model node; the preset priority order is as follows: communication protocol node, sub-communication protocol node, hardware feature type node, and software feature type node.

4. The intelligent generation method for UAV countermeasure strategies according to claim 3, characterized in that, The entity nodes in the drone countermeasures knowledge graph also include multiple communication protocol nodes. The matching process continues according to a preset priority order to determine the target drone model node, including: Multiple communication protocol sub-features are extracted from the multimodal features to be matched, and a communication protocol feature vector is constructed; the multiple communication protocol sub-features include signal frequency band, frequency hopping pattern, and frame structure identifier; Using the aforementioned similarity calculation formula, the communication protocol feature vector is matched with the preset feature vectors of multiple communication protocol nodes to obtain the first similarity result for each communication protocol node. If there is at least one communication protocol node whose first similarity result is greater than or equal to the second preset similarity threshold, then the communication protocol node corresponding to the maximum value of the first similarity result is determined as the target communication protocol node, and the node of the known drone model corresponding to the target communication protocol node is determined as the target drone model node through the adoption protocol relationship between the communication protocol node and the drone model node. If the first similarity results of all communication protocol nodes are less than the second preset similarity threshold, then the similarity protocol relationship between communication protocol nodes in the UAV countermeasure knowledge graph is invoked, and multiple sub-communication protocol nodes associated with the target communication protocol node are selected based on the filtering conditions to continue similarity matching, resulting in multiple second similarity results; the filtering conditions are that the protocol similarity attribute value is greater than or equal to the third preset similarity threshold. If at least one second similarity result is greater than or equal to the second preset similarity threshold, then the corresponding target UAV model node is obtained; If all the second similarity results are less than the second preset similarity threshold, then continue matching according to the preset priority order to determine the target UAV model node.

5. The intelligent generation method for UAV countermeasure strategies according to claim 2, characterized in that, The relationships in the drone countermeasures knowledge graph include at least the relationships that can be countered, the relationships that are applicable to the environment, the relationships that are prohibited in the environment, and the relationships that affect the effectiveness of countermeasures in terms of environmental impact. When the situation type information is a single-machine situation, based on the matching result and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated, including: Based on the countermeasureable relationship, all countermeasures associated with the matching result are obtained to form a first countermeasure pool; Based on the applicable environmental relationship, countermeasures containing the environmental information are selected from the first countermeasures pool to obtain a second countermeasures pool; By using the disabled environment relationship, the third countermeasure pool is obtained by excluding countermeasures in the second countermeasure pool whose disabled environment contains the environment information. By using the aforementioned environmental impact countermeasure effectiveness relationship, the association attributes between the environmental information in the UAV countermeasure knowledge graph and each countermeasure in the third countermeasure pool are retrieved, and the corrected effectiveness is calculated by combining the original effectiveness of each countermeasure. Based on the corrected effectiveness of each means in the third countermeasure pool, all countermeasures are combined and the expected countermeasure effect of each combination is marked to obtain the initial countermeasure strategy set for the single-machine situation.

6. The intelligent generation method for UAV countermeasure strategies according to claim 5, characterized in that, The relationships in the drone countermeasures knowledge graph also include tactical combination relationships; When the situation type information is a cluster situation, based on the matching result and the environmental information, at least one countermeasure is obtained from the UAV countermeasure knowledge graph and an initial countermeasure strategy set is generated, including: Invoke the tactical combination relationship to obtain the countermeasure coordination framework associated with the target tactical mode node; For each link of the aforementioned countermeasures coordination framework, candidate countermeasures for each link are obtained from the UAV countermeasures knowledge graph; Based on the applicable environment relationship, the candidate countermeasures for each stage are selected, and the applicable environment contains the environmental information, to obtain the first set of countermeasures for each stage. Based on the aforementioned disabled environment relationship, the means whose disabled environment contains the aforementioned environmental information are excluded from the first countermeasures set of each stage, thereby obtaining the second countermeasures set of each stage; Based on the aforementioned environmental impact countermeasure effectiveness relationship, the correlation attributes of each means in the second countermeasure means set at each stage are retrieved, the original effectiveness of each means is corrected, and the corrected effectiveness of each means is obtained. Based on the effectiveness of each modified method, methods are selected from the second set of countermeasures in each stage and combined in sequence according to the countermeasures coordination framework. The expected countermeasure effect of each combination is marked to generate an initial set of countermeasure strategies for the cluster situation.

7. The intelligent generation method for UAV countermeasure strategies according to claim 1, characterized in that, The initial countermeasure strategy set is processed using a target optimization algorithm to obtain the target countermeasure strategy, including: A multi-objective optimization set is determined; the multi-objective optimization set includes benefit-oriented optimization objectives and cost-oriented optimization objectives; the benefit-oriented optimization objectives include reaction success rate and collaborative efficiency; the cost-oriented optimization objectives include response time, economic cost, and collateral damage; Formula used: ； The original data of the benefit-oriented optimization objectives are normalized to obtain the normalized values ​​of each benefit-oriented optimization objective; among them, For the first The normalized value of a benefit-oriented optimization objective; For the first The original data for each benefit-oriented optimization objective; For the first The minimum baseline value for each benefit-oriented optimization objective; For the first The maximum baseline value of each benefit-oriented optimization objective; Formula used: ； The original data for cost-type optimization objectives are normalized to obtain normalized values ​​for each cost-type optimization objective; among them, For the first The normalized value of a cost-type optimization objective; For the first The original data for each cost-based optimization objective; For the first The minimum baseline value for a cost-based optimization objective; For the first The maximum baseline value for each cost-based optimization objective; The weight coefficients of each optimization objective are determined based on the situation type information. Formula used: ； Calculate the overall score for each initial countermeasure strategy in the initial countermeasure strategy set; where, The overall score; For the first The weight coefficients of the first optimization objective; if the first... If the optimization objective is a benefit-oriented optimization objective, then If the first If the optimization objective is a cost-based optimization objective, then ; The initial countermeasure strategy with the highest overall score is selected as the target countermeasure strategy.

8. The intelligent generation method for UAV countermeasure strategies according to claim 1, characterized in that, Also includes: Based on the target countermeasure strategy, the corresponding countermeasure execution device is controlled to perform countermeasure tasks against the suspicious drone, and the countermeasure process data is recorded in real time during the execution of the countermeasure task. If the countermeasure process data shows that the countermeasure effectiveness of the target countermeasure strategy is higher than a preset threshold, then the attribute of the countermeasure effectiveness relationship between the target countermeasure strategy and the corresponding UAV model in the UAV countermeasure knowledge graph is updated.

9. An intelligent generation device for unmanned aerial vehicle (UAV) countermeasure strategies, characterized in that, include: The acquisition module is used to acquire multi-source sensor data of the target area, and determine the environmental information of the target area and the situation type information of the suspicious UAV based on the multi-source sensor data; the situation type information includes single-UAV situation and cluster situation; The matching module is used to extract the features to be matched corresponding to the situation type information from the multi-source sensor data, and to perform similarity matching between the features to be matched and the features of multiple target matching entity nodes in the UAV countermeasure knowledge graph to determine the matching result; the features to be matched are multimodal features to be matched corresponding to the single-UAV situation or cluster situation features to be matched corresponding to the cluster situation. The generation module is used to obtain at least one countermeasure from the UAV countermeasure knowledge graph and generate an initial set of countermeasure strategies based on the matching results and the environmental information. The optimization module is used to process the initial countermeasure strategy set using a target optimization algorithm to obtain the target countermeasure strategy; The entity nodes in the drone countermeasures knowledge graph include at least tactical pattern nodes; When the situation type information is a cluster situation, the target matching entity node is a tactical mode node, and the matching result is the target tactical mode node; The situational characteristics of the cluster to be matched include quantitative and structural characteristics, formation characteristics, cooperative action characteristics, and communication and coordination characteristics; The tactical mode nodes include at least cluster-coordinated attack mode nodes, multi-type hybrid flanking penetration mode nodes, and distributed cluster harassment mode nodes; Based on the similarity scores of all target matching entity nodes, the matching results are determined, including: The tactical mode node corresponding to the maximum similarity value is determined as the target tactical mode node.

Images