Self-adaptive control method and system for low-altitude target

Abstract

The invention belongs to the technical field of unmanned aerial vehicle countering, and discloses a self-adaptive control method and system for a low-altitude target. The method comprises the steps of collecting radar scanning information of a low-altitude target, and performing feature extraction on the radar scanning information to obtain position features; a target image of a low-altitude target is collected, feature extraction is conducted on the target image, and the number of rotors, the shape of a fuselage and color features are obtained; comparing with a pre-stored color library to obtain the color of the low-altitude target; acquiring a thermal radiation image and a signal frequency band; performing feature extraction on the thermal radiation image to obtain thermal features; performing feature extraction on the signal frequency band to obtain an electronic fingerprint; the number of rotors, the fuselage shape, the thermal characteristics and the electronic fingerprints serve as input of a model analysis model, and the model of the low-altitude target is obtained; according to the method, the threat score is dynamically calculated through the trajectory analysis model, and the multi-countering means collaboration is triggered through the threat score, so that the interference coverage efficiency and the target recognition accuracy are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV countermeasures. More specifically, the present invention relates to an adaptive control method and system for low-altitude targets. Background Art

[0002] With the popularization of UAV technology and the development of hypersonic weapons, low-altitude targets exhibit characteristics such as "low, slow, small", stealth, and intelligence.

[0003] The patent application with the publication number CN114911267A discloses an intelligent autonomous UAV countermeasure system, including an AI-empowered network attack module, an intelligent UAV capture net module, a GPS navigation deception module, and a multi-band radio interference suppression module. Among them, the AI-empowered network attack module is used to carry intelligent devices, call attack tools, and use software vulnerabilities to cut off the WIFI link between the controller and the UAV, realizing the takeover of the invaded UAV; the intelligent UAV capture net module is used to autonomously track the UAV and autonomously launch the UAV capture net to achieve automatic operation; the GPS navigation deception module is used to generate navigation deception signals to lure the UAV to a preset location; the multi-band radio interference suppression module is used to generate multi-band suppression signals to block the communication between the UAV and the user terminal. Through multi-mode deployment, this technical solution can achieve unattended operation, accurately drive away or force down the invaded UAV, and ensure the safety of low-altitude airspace.

[0004] Although the above method can meet most scenarios, through research and practical application of the above method and the existing technology, it is found that the above method and the existing technology have at least the following partial defects:

[0005] Lack of dynamic countermeasures based on threat assessment by identifying low-altitude targets.

[0006] In view of this, the present invention proposes an adaptive control method and system for low-altitude targets to solve the above problems. Summary of the Invention

[0007] In order to overcome the above defects of the existing technology and to achieve the above object, the present invention provides the following technical solution: An adaptive control method for low-altitude targets, including the following steps:

[0008] Collect the radar scan information of the low-altitude target, and extract features from the radar scan information to obtain position features;

[0009] Collect the target image of the low-altitude target, extract features from the target image to obtain the number of rotors, fuselage shape, and color features; and compare with the pre-stored color library to obtain the color of the low-altitude target;

[0010] Collect and analyze the signal frequencies of the low-altitude target received at preset N collection points, and calculate to obtain the position of the drone pilot;

[0011] Collect thermal radiation images; extract features from the thermal radiation images to obtain thermal features; extract features from the signal frequency band to obtain electronic fingerprints; use the number of rotors, fuselage shape, thermal features, and electronic fingerprints as the input of the model analysis model to obtain the model of the low-altitude target;

[0012] Perform matching analysis on the electronic fingerprints to obtain a matching result; if the matching result is an unknown target, then use the position feature, color, thermal feature, and model as the input of the trajectory analysis model to obtain a threat score;

[0013] Collect environmental data. If the threat score exceeds the preset threat threshold, then combine the environmental data, position feature, color, thermal feature, model, and pilot position to adopt a countermeasure mode through a dynamic defense strategy to counter the low-altitude target.

[0014] Furthermore, the method for obtaining the electronic fingerprint includes:

[0015] Passively receive the transmission signal of the low-altitude target, extract the serial number, MAC address, signal modulation method, and frame structure of the transmission signal of the low-altitude target, and detect whether the transmission signal of the low-altitude target is a forged signal based on the detection strategy. If it is a forged signal, directly activate the dynamic defense strategy to adopt a countermeasure mode. If it is not a forged signal, perform spectrum analysis and quantization on the transmission signal of the low-altitude target, extract the carrier frequency offset and phase noise, splice the carrier frequency offset and phase noise as the signal fingerprint, extract the heartbeat packet sending interval and the data packet length distribution, splice the heartbeat packet sending interval and the data packet length distribution as the behavior fingerprint, quantify the signal features through a statistical model, and use the signal fingerprint and the behavior fingerprint as the input of the fingerprint analysis model to obtain the electronic fingerprint; the detection strategy includes monitoring whether the carrier frequency of the transmission signal of the low-altitude target has a frequency drift, comparing whether the signal modulation method matches the pre-stored device protocol library, monitoring whether the signal strength exceeds the preset strength threshold, and monitoring whether the pilot position is within the preset authorized area. If any one of the detection results is yes, it is determined as a forged signal, otherwise it is not a forged signal.

[0016] Furthermore, the method for obtaining the threat score includes:

[0017] Compare the electronic fingerprint with the pre-stored database to obtain the matching result corresponding to the low-altitude target. The matching result includes unknown fingerprints and the white list and black list stored in the pre-stored database. When the matching result is the white list, let it pass. When the matching result is the black list, immediately trigger the dynamic defense strategy to adopt a countermeasure mode; when the matching result is an unknown fingerprint, store the electronic fingerprint corresponding to the low-altitude target in the temporary library, and use the position feature, color, thermal feature, and model as the input of the trajectory analysis model to obtain the threat score.

[0018] Further, the method for countering low-altitude targets includes:

[0019] Define the state space S: Obtain threat scores, signal frequency bands, and environmental data; the environmental data includes wind speed, temperature, precipitation, electromagnetic interference intensity, and terrain occlusion coefficient;

[0020] Define the action space A: Select the interference signal frequency band, power level, generate false GPS coordinates deviating from the preset no-fly zone, adjust the pan-tilt angle according to the low-altitude target angle, and turn on the corresponding frequency band for interception; among them, the actions in the action space also include combined actions;

[0021] Design the reward function: Obtain the total reward function by weighting the anti-aircraft success reward, energy consumption penalty, collateral damage penalty, and environmental adaptation reward;

[0022] Update the Q-value function: Use the Q-learning algorithm to learn the optimal policy, initialize the Q-value function as a table, where the rows of the table correspond to different environmental states , and the columns correspond to different actions , set the learning rate , discount factor and exploration rate ; at each time step , detect the current environmental state , according to the exploration rate , with probability randomly select an action , with probability select the action with the largest Q-value in the current environmental state ; execute the selected action to obtain a new environmental state , at the same time, calculate the total reward function value according to the reward function; update the Q-value function according to the update formula.

[0023] Repeat updating the Q-value function until the Q-value converges, obtain the corresponding optimal Q-value function, and select and execute the optimal action from the action space according to the environmental state.

[0024] Further, the method for obtaining the position of the drone pilot includes;

[0025] Obtain the coordinates of the collection points and the time of arrival of the received signals, calculate the time difference between any two collection points respectively, construct a hyperbolic equation based on the time difference and coordinates, and solve the hyperbolic equation through an iterative algorithm to obtain the coordinates of the drone pilot.

[0026] Further, the method for obtaining the thermal signature includes:

[0027] Convert the thermal radiation image into a grayscale or pseudo-color image, where the highlighted area corresponds to the high-temperature area. Segment the image based on a preset segmentation threshold, label each independent heat source based on connected component analysis, obtain the number of independent heat sources, traverse the temperature values corresponding to the pixel points of each independent heat source, record the highest temperature value of each independent heat source, and collect the highest temperature value of each independent heat source once every preset time interval to calculate the average temperature change rate of each independent heat source; Concatenate the number of independent heat sources, the highest temperature value of each independent heat source, and the average temperature change rate of each independent heat source as thermal features.

[0028] Further, the method for obtaining the number of rotors includes:

[0029] Perform edge detection on the target image to extract the low-altitude target contour, erode the low-altitude target contour with a linear structuring element and a square structuring element in sequence to obtain the eroded low-altitude target contour, count the number of connected regions in the eroded low-altitude target contour, and combine the rotor symmetry judgment to obtain the first rotor number;

[0030] Collect the Doppler frequency shift and obtain the second rotor number through spectrum analysis;

[0031] Cross-validate the first rotor number and the second rotor number. If the first rotor number and the second rotor number are the same, use the values of the first rotor number and the second rotor number as the rotor number. If the first rotor number and the second rotor number are different, trigger manual review or dynamic resampling to obtain the rotor number.

[0032] Further, the method for obtaining the color of the low-altitude target includes:

[0033] Extract the color features of the low-altitude target from the target image through RGB pixel analysis and compare them with the pre-stored color library to obtain the color of the low-altitude target.

[0034] Further, the method for obtaining the fuselage shape includes:

[0035] Perform object detection on the target image, separate and obtain the target foreground, and use the target foreground as the input of the shape analysis model to obtain the fuselage shape.

[0036] Further, the position features include three-dimensional position information, speed, and heading; the method for obtaining the position features includes:

[0037] Step 1: Collect the time difference between the transmitted electromagnetic wave and the received electromagnetic wave, and calculate the distance of the low-altitude target in combination with the electromagnetic wave transmission speed;

[0038] Step 2: Calculate the low-altitude target speed in combination with the radar transmission frequency and the Doppler frequency shift of the reflected signal;

[0039] Step 3: Control the mechanical rotating antenna or the electronically scanned beam to point in different directions, calculate and obtain the direction of the low-altitude target; combine the distance of the low-altitude target to obtain the three-dimensional position information of the low-altitude target; the three-dimensional position information includes longitude, latitude, and altitude.

[0040] Step 4: At intervals of a preset period, repeat Steps 1 - 3 to obtain the second three-dimensional position information of the low-altitude target, calculate the displacement vector between the second three-dimensional position information and the three-dimensional position information, and calculate the heading of the low-altitude target based on the arctangent function in combination with the displacement vector.

[0041] Step 5: Calculate the modulus of the displacement vector, calculate the ratio of the modulus to the preset period to obtain the speed of the low-altitude target.

[0042] An adaptive control system for low-altitude targets, implementing the above-mentioned adaptive control method for low-altitude targets, includes:

[0043] Position analysis module: Collect the radar scan information of the low-altitude target, and extract features from the radar scan information to obtain position features.

[0044] Color analysis module: Collect the target image of the low-altitude target, extract features from the target image to obtain the number of rotors, fuselage shape, and color features; and compare with the pre-stored color library to obtain the color of the low-altitude target.

[0045] Traceability analysis module: Analyze the signal frequency bands of the low-altitude target received at preset N collection points, and calculate and obtain the position of the drone pilot.

[0046] Model analysis module: Collect the thermal radiation image; extract features from the thermal radiation image to obtain thermal features; extract features from the signal frequency bands to obtain electronic fingerprints; use the number of rotors, fuselage shape, thermal features, and electronic fingerprints as the input of the model analysis model to obtain the model of the low-altitude target.

[0047] Threat analysis module: Conduct matching analysis on the electronic fingerprints to obtain a matching result; if the matching result is an unknown target, use the position features, color, thermal features, and model as the input of the trajectory analysis model to obtain a threat score.

[0048] Intelligent countermeasure module: Collect environmental data. If the threat score exceeds the preset threat threshold, combine the environmental data, position features, color, thermal features, model, and the position of the drone pilot to adopt a countermeasure mode through a dynamic defense strategy to counter the low-altitude target.

[0049] The technical effects and advantages of the adaptive control method and system for low-altitude targets of the present invention:

[0050] The present invention constructs an intelligent defense system with full-spectrum feature fusion by integrating four-dimensional perception data of radar, vision, thermal infrared, and electromagnetic signals: radar scanning identifies centimeter-level position features, vision recognition realizes precise classification of the number of rotors and the fuselage shape, and unique device-level identifiers are formed through thermal features and electronic fingerprints to achieve precise identification of low-altitude targets and at the same time achieve precise positioning of the drone pilot's position; the system dynamically calculates threat scores through a trajectory analysis model and triggers the coordination of multiple countermeasures through threat scores, effectively improving the interference coverage efficiency and target recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a schematic flow chart of an adaptive control method for a low-altitude target of the present invention;

[0052] Figure 2 is a schematic flow chart of a method for countering low-altitude targets of the present invention;

[0053] Figure 3 is a schematic flow chart of a dynamic defense strategy optimization method of the present invention;

[0054] Figure 4 is a structural diagram of an adaptive control system for a low-altitude target of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0056] Embodiment 1

[0057] Please refer to Figure 1 As shown, the adaptive control method for a low-altitude target in this embodiment includes the following steps:

[0058] Collect radar scanning information of the low-altitude target, and extract features from the radar scanning information to obtain position features; the radar scanning information includes the electromagnetic waves emitted towards the low-altitude target and the corresponding emission times, and the reflected signals of the low-altitude target received and the corresponding reception times; the position features include three-dimensional position information, speed, and heading;

[0059] The method for obtaining the position features includes:

[0060] Step 1. Collect the time difference between the emitted electromagnetic waves and the received electromagnetic waves, and calculate the distance of the low-altitude target in combination with the emission speed of the electromagnetic waves;

[0061] Step 2: Calculate the speed of the low-altitude target by combining the radar transmission frequency and the Doppler shift of the reflected signal;

[0062] Step 3: Control the mechanical rotating antenna or the electronically scanned beam to point in different directions, and calculate the direction of the low-altitude target; combine the distance of the low-altitude target to obtain the three-dimensional position information of the low-altitude target; the three-dimensional position information includes longitude, latitude and altitude;

[0063] Step 4: At intervals of a preset time period, repeat Steps 1 - 3 to obtain the second three-dimensional position information of the low-altitude target, calculate the displacement vector between the second three-dimensional position information and the three-dimensional position information, and calculate the heading of the low-altitude target based on the arctangent function combined with the displacement vector;

[0064] Step 5: Calculate the modulus of the displacement vector, calculate the ratio of the modulus to the preset time period to obtain the speed of the low-altitude target.

[0065] Position features play a core supporting role in low-altitude defense and disposal: The position features such as the target position and speed obtained through radar scanning not only provide a spatio-temporal reference for multi-sensor data fusion to achieve continuous tracking and prediction of the target track, but also serve as input parameters for the pilot positioning algorithm to achieve precise positioning of the emission source in combination with signal fingerprint analysis. At the same time, the position features participate in the construction of the threat assessment model, calculate the threat score together with the trajectory features and environmental data, and provide a quantitative basis for the trigger threshold of the dynamic defense strategy. In addition, in distributed collaborative defense, the position features can support the optimal scheduling of countermeasure resources and achieve precise interference coverage through the multi-node interference power distribution model.

[0066] Collect the target image of the low-altitude target, extract features from the target image to obtain the number of rotors, fuselage shape and color features; and compare with the pre-stored color library to obtain the color of the low-altitude target. The target image feature extraction plays a multi-dimensional supporting role in low-altitude defense and disposal: The type of multi-rotor and fixed-wing UAVs can be quickly distinguished by identifying the number of rotors, and the target model can be further locked by combining the fuselage shape analysis; The comparison of the color features with the pre-stored library can assist in identifying camouflage or specific attribution, and at the same time provide a visual verification benchmark for multi-sensor data fusion; These features are jointly input into the model analysis model, which can effectively improve the accuracy of target identity discrimination, and update the threat assessment model through the dynamic matching of color features to achieve cross-verification of target attributes and behavior patterns in complex environments, providing a key basis for the precise trigger of countermeasure strategies.

[0067] The methods for obtaining the number of rotors include:

[0068] Perform edge detection on the target image to extract the contour of the low-altitude target. Then, use a linear structuring element and a square structuring element to erode the contour of the low-altitude target in sequence to obtain the eroded contour of the low-altitude target. Count the number of connected regions in the eroded contour of the low-altitude target, and combine the rotor symmetry judgment to obtain the number of the first rotors.

[0069] Collect the Doppler frequency shift and obtain the number of the second rotors through spectrum analysis.

[0070] Cross-verify the number of the first rotors and the number of the second rotors. If the number of the first rotors is the same as the number of the second rotors, use the numerical values of the number of the first rotors and the number of the second rotors as the number of rotors. If the number of the first rotors is different from the number of the second rotors, trigger manual review or dynamic resampling to obtain the number of rotors.

[0071] The methods for obtaining the fuselage shape include:

[0072] Perform target detection on the target image, separate and obtain the target foreground, and use the target foreground as the input of the shape analysis model to obtain the fuselage shape.

[0073] The training method of the shape analysis model includes:

[0074] Pre-collect F groups of shape analysis data, where the shape analysis data includes the target foreground and the corresponding fuselage shape.

[0075] Use the target foreground as the input of the shape analysis model and the corresponding fuselage shape as the output of the shape analysis model. With the goal of minimizing the error between the output corresponding fuselage shape and the actual corresponding fuselage shape, optimize the network parameters of the shape analysis model through a nature-inspired optimization algorithm, obtain the network parameters corresponding to minimizing the error between the output corresponding fuselage shape and the actual corresponding fuselage shape of the shape analysis model, and use the shape analysis model constructed with the corresponding network parameters as the trained shape analysis model.

[0076] The methods for obtaining the color of the low-altitude target include:

[0077] Extract the color features of the low-altitude target from the target image through RGB pixel analysis and compare them with the pre-stored color library to obtain the color of the low-altitude target.

[0078] Collect and analyze the signal frequency bands of low-altitude targets received at N preset collection points, and calculate the position of the drone operator; by accurately locking the position of the drone operator, it can provide a spatial reference for countermeasure operations; combining the position information of the drone operator with the target trajectory characteristics can dynamically evaluate the threat radiation range (such as the coverage radius of sensitive areas) and optimize the interference resource allocation strategy; at the same time, the accurate positioning result provides a key evidence chain when evidence collection is required, and combines electronic fingerprint matching to achieve a triple association of "person-drone-signal", reducing the probability of misidentification and significantly improving the compliance of operations while ensuring the countermeasure effectiveness.

[0079] The methods for obtaining the position of the drone operator include:

[0080] Obtain the coordinates of the collection points and the time of arrival of the received signals, calculate the time difference between any two collection points respectively, construct a hyperbolic equation based on the time difference and coordinates, and solve the hyperbolic equation through an iterative algorithm to obtain the coordinates of the drone operator.

[0081] Collect thermal radiation images; extract features from the thermal radiation images to obtain thermal features; extract features from the signal frequency bands to obtain electronic fingerprints; use the number of rotors, fuselage shape, thermal features, and electronic fingerprints as inputs to the model analysis model to obtain the model of the low-altitude target; the thermal radiation images and signal frequency band features provide key discrimination bases in low-altitude defense and disposal: the type of drone can be effectively distinguished through thermal features and the position of the heat source can be located. Combining the device-level uniqueness identification ability of electronic fingerprints with the number of rotors and fuselage shape to form multi-modal feature fusion can effectively improve the accuracy of the model analysis model; among them, electronic fingerprint matching not only realizes the rapid identification of known models, but also triggers the threat assessment process through the detection of unknown targets, provides model traceability and threat grading support for dynamic defense strategies, improves the target identification accuracy by more than 30% compared with traditional single-feature identification, and provides data support for the accurate selection of countermeasure means at the same time.

[0082] The methods for obtaining thermal features include:

[0083] Convert the thermal radiation image into a grayscale or pseudo-color image. Among them, the highlighted area (such as the corresponding temperature value is greater than the preset temperature threshold) corresponds to the high-temperature area. Segment the image based on the preset segmentation threshold, mark each independent heat source based on connected component analysis, obtain the number of independent heat sources, traverse the temperature values corresponding to the pixel points of each independent heat source, record the highest temperature value of each independent heat source, collect the highest temperature value of each independent heat source once every preset time, and calculate the average temperature change rate of each independent heat source; splice the number of independent heat sources, the highest temperature value of each independent heat source, and the average temperature change rate of each independent heat source as thermal features.

[0084] The methods for obtaining electronic fingerprints include:

[0085] Passively receive the transmission signals of low-altitude targets, extract the serial numbers, MAC addresses, signal modulation methods, and frame structures of the transmission signals of low-altitude targets, detect whether the transmission signals of low-altitude targets are forged signals based on a detection strategy. If it is a forged signal, directly activate the dynamic defense strategy to adopt a countermeasure mode. If it is not a forged signal, perform spectrum analysis and quantization on the transmission signals of low-altitude targets, extract the carrier frequency offset and phase noise, splice the carrier frequency offset and phase noise as the signal fingerprint, extract the heartbeat packet sending interval and packet length distribution, splice the heartbeat packet sending interval and packet length distribution as the behavior fingerprint, quantify the signal characteristics through a statistical model, use the signal fingerprint and behavior fingerprint as the input of the fingerprint analysis model to obtain the electronic fingerprint; The detection strategy includes monitoring whether the carrier frequency of the transmission signal of the low-altitude target has frequency drift, comparing whether the signal modulation method matches the pre-stored device protocol library, monitoring whether the signal intensity exceeds the preset intensity threshold, and monitoring whether the position of the drone pilot is within the preset authorized area. If any one of the detection results is yes, it is determined as a forged signal, otherwise it is not a forged signal.

[0086] The training method of the model analysis model includes:

[0087] Pre-collect a group of B training data. The training data includes model analysis data and the corresponding models; The model analysis data includes the number of rotors, fuselage shape, thermal characteristics, and electronic fingerprints;

[0088] Use the model analysis data as the input of the model analysis model, use the model as the output of the model analysis model, aiming at minimizing the error between the output model and the actual model, optimize the network parameters of the model analysis model through a natural inspiration optimization algorithm, obtain the network parameters corresponding to minimizing the error between the model output by the model analysis model and the actual model, and use the model analysis model constructed with the corresponding network parameters as the trained model analysis model.

[0089] The training method of the fingerprint analysis model includes:

[0090] Pre-collect a group of C fingerprint training data. The fingerprint training data includes signal fingerprints and behavior fingerprints, and the corresponding electronic fingerprints;

[0091] Use the fingerprint training data as the input of the fingerprint analysis model, use the electronic fingerprint as the output of the fingerprint analysis model, aiming at minimizing the error between the output electronic fingerprint and the actual electronic fingerprint, optimize the network parameters of the fingerprint analysis model through a natural inspiration optimization algorithm, obtain the network parameters corresponding to minimizing the error between the electronic fingerprint output by the fingerprint analysis model and the actual electronic fingerprint, and use the fingerprint analysis model constructed with the corresponding network parameters as the trained fingerprint analysis model.

[0092] Perform matching analysis on the electronic fingerprint to obtain a matching result; if the matching result is an unknown target, use the location feature, color, thermal feature, and model as the input of the trajectory analysis model to obtain a threat score; the electronic fingerprint matching analysis and threat score calculation form an intelligent decision-making closed loop in low-altitude defense disposal: the device-level uniqueness identification of the electronic fingerprint realizes the rapid classification of known targets and provides an accurate model basis for the countermeasure strategy; when the matching fails, the location feature, color feature, thermal feature, and model analysis results are used to dynamically calculate the threat score through the trajectory analysis model, and the threat level is quantified by combining the target track and environmental data, providing a scientific basis for the countermeasure threshold decision, reducing the misjudgment rate, and at the same time optimizing the allocation of countermeasure resources through threat priority ranking to maximize the defense efficiency in a complex electromagnetic environment.

[0093] The methods for obtaining the threat score include:

[0094] Compare the electronic fingerprint with the pre-stored database to obtain the matching result corresponding to the low-altitude target. The matching result includes unknown fingerprints, as well as the whitelist and blacklist stored in the pre-stored database. When the matching result is the whitelist, let it pass; when the matching result is the blacklist, immediately trigger the dynamic defense strategy to adopt the countermeasure mode; when the matching result is an unknown fingerprint, store the electronic fingerprint corresponding to the low-altitude target in the temporary library, and use the location feature, color, thermal feature, and model as the input of the trajectory analysis model to obtain the threat score.

[0095] The training method of the trajectory analysis model includes:

[0096] Collect C groups of trajectory analysis data in advance. The trajectory analysis data includes input data and the corresponding threat score; the input data includes location feature, color, thermal feature, and model.

[0097] Use the input data as the input of the trajectory analysis model and the threat score as the output of the trajectory analysis model. With the goal of minimizing the error between the output threat score and the actual threat score, optimize the network parameters of the trajectory analysis model through the natural inspiration optimization algorithm, obtain the network parameters corresponding to minimizing the error between the threat score output by the trajectory analysis model and the actual threat score, and use the trajectory analysis model constructed by the corresponding network parameters as the trained trajectory analysis model.

[0098] Collect environmental data. If the threat score exceeds the preset threat threshold, countermeasures will be taken against low-altitude targets in a countermeasure mode through a dynamic defense strategy by combining environmental data, location features, color, thermal features, model, and the position of the drone operator. The environmental data and the dynamic defense strategy form an intelligent response closed-loop in low-altitude defense disposal: by collecting environmental parameters such as terrain, population density, and obstacle distribution in real time and combining with the threat score, the system can dynamically select the optimal countermeasure mode to improve the countermeasure efficiency; the environmental data supports the countermeasure path planning (such as bypassing residential areas), and combines with the position of the drone operator to achieve precise strikes and reduce the probability of collateral damage; at the same time, the dynamic strategy maximizes the defense efficiency under the premise of ensuring safety through the linkage of threat levels and environmental sensitivity thresholds. For example, full-band interference is triggered around airports, while the decoy mode is adopted in commercial areas, significantly improving the compliance and pertinence of countermeasure actions.

[0099] Please refer to Figure 2 As shown, the method for countering low-altitude targets includes:

[0100] Define the state space S: Obtain the threat score, signal frequency band, and environmental data; the environmental data includes wind speed, temperature, precipitation, electromagnetic interference intensity, and terrain occlusion coefficient;

[0101] Define the action space A: Select the interference signal frequency band, power level, generate false GPS coordinates deviating from the preset no-fly zone, and adjust the gimbal angle according to the low-altitude target angle and turn on the corresponding frequency band for interception; among them, the actions in the action space also include combined actions; such as the superposition of selecting the interference signal frequency band and generating false GPS coordinates deviating from the preset no-fly zone;

[0102] Design the reward function: Obtain the total reward function by weighting the countermeasure success reward, energy consumption penalty, collateral damage penalty, and environmental adaptation reward;

[0103] Update the Q-value function: Use the Q-learning algorithm to learn the optimal strategy, initialize the Q-value function as a table, where the rows of the table correspond to different environmental states , and the columns correspond to different actions , set the learning rate , discount factor and exploration rate ; at each time step , detect the current environmental state , according to the exploration rate , with probability randomly select an action , with probability select the action with the maximum Q-value in the current environmental state ; execute the selected action to obtain a new environmental state , meanwhile, calculate the total reward function value according to the reward function; update the Q-value function according to the update formula;

[0104] Repeat updating the Q-value function until the Q-value converges, obtain the corresponding optimal Q-value function, and according to the environmental state, use the optimal Q-value function to select the optimal action from the action space and execute it.

[0105] Embodiment 2

[0106] Please refer to Figure 3 As shown, this embodiment provides a method for optimizing a dynamic defense strategy, including the following steps:

[0107] Take the model of the low-altitude target as an antibody, each antibody stores the corresponding electronic fingerprint and thermal feature, and calculate the feature affinity ; The calculation formula is as follows:

[0108] ;

[0109] ;

[0110] ;

[0111] Among them, and are the physical affinity and network affinity corresponding to the antibody respectively; is the preset temperature adjustment coefficient; is the highest temperature value; is the preset temperature threshold; is the included angle between the electronic fingerprint and the preset electronic fingerprint;

[0112] Calculate the threat level according to the feature affinity; if the threat level ;

[0113] Calculate the pheromone reference value according to the threat level, and iteratively update the threat pheromone based on the pheromone reference value; if the pheromone reference value ; Among them, is the distance between the low-altitude target and the interference device;

[0114] Calculate the device selection probability based on the threat pheromone, and select the device whose device selection probability exceeds the preset probability threshold as the action basis for selection in the action space. Detailed action data, such as the interference signal frequency band and power level, etc., can be further obtained based on the dynamic defense strategy.

[0115] Through the above steps, the system defines the UAV model as an antibody and stores its electronic fingerprint and thermal characteristics. The system calculates the characteristic affinity of the antibody in real time, combines the temperature adjustment coefficient with the preset threshold to achieve multi-modal threat level quantification. The threat level is further converted into a pheromone reference value, and the threat pheromone concentration is dynamically updated in combination with the target distance to guide the ant colony algorithm to perform probabilistic selection of interference devices. Finally, the countermeasure actions (such as frequency band, power) are determined through the dynamic defense strategy to achieve optimal resource scheduling. This improves the accuracy of target recognition, shortens the response time, and at the same time optimizes the spatial distribution of threat pheromones to improve the interference coverage efficiency, significantly enhancing the low-altitude defense effectiveness in complex electromagnetic environments.

[0116] Embodiment 3

[0117] Please refer to Figure 4 As shown, this embodiment discloses an adaptive control system for low-altitude targets, including:

[0118] Position analysis module: Collect the radar scan information of the low-altitude target, and extract the position features from the radar scan information.

[0119] Color analysis module: Collect the target image of the low-altitude target, extract the features from the target image to obtain the number of rotors, fuselage shape and color features; and compare with the pre-stored color library to obtain the color of the low-altitude target.

[0120] Traceability analysis module: Analyze the signal frequency bands of the low-altitude target received at the preset N collection points, and calculate to obtain the position of the drone pilot.

[0121] Model analysis module: Collect the thermal radiation image; extract the features from the thermal radiation image to obtain the thermal features; extract the features from the signal frequency band to obtain the electronic fingerprint; use the number of rotors, fuselage shape, thermal features and electronic fingerprint as the input of the model analysis model to obtain the model of the low-altitude target.

[0122] Threat analysis module: Perform matching analysis on the electronic fingerprint to obtain the matching result; if the matching result is an unknown target, use the position features, color, thermal features and model as the input of the trajectory analysis model to obtain the threat score.

[0123] Intelligent countermeasure module: Collect the environmental data. If the threat score exceeds the preset threat threshold, take countermeasure actions against the low-altitude target through the dynamic defense strategy in combination with the environmental data, position features, color, thermal features, model and the position of the drone pilot.

[0124] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the said claims.

[0125] Finally: The above description is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should all be included within the protection scope of the present invention.

Claims

1. An adaptive control method for low-altitude targets, characterized in that It includes the following steps: Collect the radar scanning information of low-altitude targets, and extract features from the radar scanning information to obtain position features; Collect the target images of low-altitude targets, extract features from the target images to obtain the number of rotors, fuselage shape and color features; And compare with the pre-stored color library to obtain the color of the low-altitude target; Collect and analyze the signal frequency bands of low-altitude targets received at preset N collection points, and calculate to obtain the position of the drone pilot; Collect thermal radiation images; Extract features from the thermal radiation images to obtain thermal features; Extract features from the signal frequency bands to obtain electronic fingerprints; use the number of rotors, fuselage shape, thermal features and electronic fingerprints as the input of the model analysis model to obtain the model of the low-altitude target; Conduct matching analysis on the electronic fingerprints to obtain a matching result; if the matching result is an unknown target, then use the position features, color, thermal features and model as the input of the trajectory analysis model to obtain a threat score; Collect environmental data. If the threat score exceeds the preset threat threshold, then combine the environmental data, position features, color, thermal features, model and the position of the drone pilot to adopt a countermeasure mode through a dynamic defense strategy to counter the low-altitude target.

2. The adaptive control method for low-altitude targets according to claim 1, wherein The method for obtaining the electronic fingerprint includes: Passively receive the transmission signals of low-altitude targets, extract the serial number, MAC address, signal modulation method and frame structure of the transmission signals of low-altitude targets, and detect whether the transmission signals of low-altitude targets are forged signals based on the detection strategy. If it is a forged signal, directly activate the dynamic defense strategy to adopt a countermeasure mode. If it is not a forged signal, conduct spectrum analysis and quantization on the transmission signals of low-altitude targets, extract the carrier frequency offset and phase noise, splice the carrier frequency offset and phase noise as the signal fingerprint, extract the heartbeat packet sending interval and packet length distribution, splice the heartbeat packet sending interval and packet length distribution as the behavior fingerprint, quantify the signal features through a statistical model, and use the signal fingerprint and behavior fingerprint as the input of the fingerprint analysis model to obtain the electronic fingerprint; the detection strategy includes monitoring whether the carrier frequency of the transmission signal of the low-altitude target has frequency drift, comparing whether the signal modulation method matches the pre-stored device protocol library, monitoring whether the signal intensity exceeds the preset intensity threshold, and monitoring whether the position of the drone pilot is within the preset authorized area. If any one of the detection results is yes, it is judged as a forged signal, otherwise it is not a forged signal.

3. The adaptive control method for low-altitude targets according to claim 1, wherein, The method for obtaining the threat score includes: Compare the electronic fingerprint with the pre-stored database to obtain the matching result corresponding to the low-altitude target. The matching result includes unknown fingerprints and the whitelist and blacklist stored in the pre-stored database. When the matching result is the whitelist, then let it pass. When the matching result is the blacklist, immediately trigger the dynamic defense strategy to adopt a countermeasure mode; when the matching result is an unknown fingerprint, then store the electronic fingerprint corresponding to the low-altitude target in the temporary library, and use the position features, color, thermal features and model as the input of the trajectory analysis model to obtain the threat score.

4. An adaptive control method for low-altitude targets according to claim 1, characterized in that The method for countering the low-altitude target includes: Define the state space S: obtain the threat score, signal frequency band and environmental data; the environmental data includes wind speed, temperature, precipitation, electromagnetic interference intensity and terrain occlusion coefficient; Define the action space A: Select the interference signal frequency band and power level, generate false GPS coordinates deviating from the preset no-fly zone, adjust the gimbal angle according to the low-altitude target angle, and turn on the corresponding frequency band for interception; among them, the actions in the action space also include combined actions; Design the reward function: Obtain the total reward function by weighting the anti-jamming success reward, energy consumption penalty, collateral damage penalty, and environmental adaptation reward; Update the Q-value function: Use the Q-learning algorithm to learn the optimal policy. Initialize the Q-value function as a table, where the rows of the table correspond to different environmental states , and the columns correspond to different actions . Set the learning rate , discount factor and exploration rate ; At each time step , detect the current environmental state , according to the exploration rate , with probability , randomly select an action , and with probability , select the action with the maximum Q-value in the current environmental state ; Execute the selected action to obtain a new environmental state . At the same time, calculate the total reward function value according to the reward function; Update the Q-value function according to the update formula; Repeat updating the Q-value function until the Q-value converges, obtain the corresponding optimal Q-value function, and select and execute the optimal action from the action space according to the environmental state.

5. The adaptive control method for low-altitude targets according to claim 1, characterized in that The method for obtaining the position of the drone pilot includes; Obtain the coordinates of the collection points and the time of arrival of the received signal, calculate the time difference between any two collection points respectively, construct a hyperbolic equation based on the time difference and coordinates, and solve the hyperbolic equation through an iterative algorithm to obtain the coordinates of the drone pilot.

6. The adaptive control method for a low-altitude target according to claim 1, wherein, The method for obtaining the thermal characteristics includes: Convert the thermal radiation image into a grayscale or pseudo-color image, where the highlighted area corresponds to the high-temperature area, segment the image based on a preset segmentation threshold, mark each independent heat source based on connected component analysis, obtain the number of independent heat sources, traverse the temperature values corresponding to the pixel points of each independent heat source, record the highest temperature value of each independent heat source, collect the highest temperature value of each independent heat source once every preset time, and calculate the average temperature change rate of each independent heat source; splice the number of independent heat sources, the highest temperature value of each independent heat source, and the average temperature change rate of each independent heat source as the thermal characteristics.

7. An adaptive control method for low-altitude targets according to claim 1, characterized in that The method for obtaining the number of rotors includes: Perform edge detection on the target image to extract the low-altitude target contour, erode the low-altitude target contour with a linear structuring element and a square structuring element in sequence, obtain the eroded low-altitude target contour, count the number of connected regions in the eroded low-altitude target contour, and combine the rotor symmetry judgment to obtain the first rotor number; Collect the Doppler frequency shift and obtain the second rotor number through spectrum analysis; Cross-validate the first rotor number and the second rotor number. If the first rotor number and the second rotor number are the same, use the numerical values of the first rotor number and the second rotor number as the rotor number. If the first rotor number and the second rotor number are different, trigger manual review or dynamic resampling to obtain the rotor number.

8. The adaptive control method for low-altitude targets according to claim 1, characterized in that The method for obtaining the color of the low-altitude target includes: Extract the color features of the low-altitude target from the target image through RGB pixel analysis and compare them with the pre-stored color library to obtain the color of the low-altitude target.

9. The adaptive control method for a low-altitude target according to claim 1, wherein The method for obtaining the fuselage shape includes: Perform object detection on the target image, separate and obtain the target foreground, and use the target foreground as the input of the shape analysis model to obtain the fuselage shape.

10. The adaptive control method for a low-altitude target according to claim 1, characterized in that, The position features include three-dimensional position information, speed, and heading; The method for obtaining the position features includes: Step 1: Collect the time difference between the transmitted electromagnetic wave and the received electromagnetic wave, and calculate the distance of the low-altitude target in combination with the electromagnetic wave transmission speed; Step 2: Calculate the speed of the low-altitude target in combination with the radar transmission frequency and the Doppler frequency shift of the reflected signal; Step 3: Control the mechanical rotating antenna or the beam direction of electronic scanning to point in different directions, calculate and obtain the direction of the low-altitude target; combine the distance of the low-altitude target to obtain the three-dimensional position information of the low-altitude target; the three-dimensional position information includes longitude, latitude, and altitude. Step 4: At intervals of a preset time period, repeat Steps 1 - 3 to obtain the second three-dimensional position information of the low-altitude target, calculate the displacement vector between the second three-dimensional position information and the three-dimensional position information, and calculate the heading of the low-altitude target based on the arctangent function combined with the displacement vector. Step 5: Calculate the modulus of the displacement vector, calculate the ratio of the modulus to the preset time period to obtain the speed of the low-altitude target.

11. An adaptive control system for low-altitude targets, implementing the adaptive control method for low-altitude targets according to any one of claims 1-10, characterized in that, It includes: Position analysis module: Collect the radar scanning information of the low-altitude target, and extract features from the radar scanning information to obtain position features. Color analysis module: Collect the target image of the low-altitude target, extract features from the target image to obtain the number of rotors, fuselage shape, and color features; compare with the pre-stored color library to obtain the color of the low-altitude target. Traceability analysis module: Analyze the signal frequency bands of the low-altitude target received at preset N collection points, and calculate and obtain the position of the drone operator. Model analysis module: Collect the thermal radiation image. Extract features from the thermal radiation image to obtain thermal features. Extract features from the signal frequency band to obtain an electronic fingerprint; use the number of rotors, fuselage shape, thermal features, and electronic fingerprint as the input of the model analysis model to obtain the model of the low-altitude target. Threat analysis module: Conduct a matching analysis on the electronic fingerprint to obtain a matching result; if the matching result is an unknown target, use the position features, color, thermal features, and model as the input of the trajectory analysis model to obtain a threat score. Intelligent countermeasure module: Collect environmental data. If the threat score exceeds the preset threat threshold, combine the environmental data, position features, color, thermal features, model, and the position of the drone operator to take a countermeasure mode against the low-altitude target through a dynamic defense strategy.

Images