Unmanned aerial vehicle detection countering method

Abstract

The invention relates to an unmanned aerial vehicle detection countering method, which comprises the following steps of: deploying a multi-dimensional sensing network to carry out full-band instantaneous scanning, synchronously acquiring noise characteristics of radio signals, material characteristics of a terahertz wave band and Doppler effect characteristics, mapping to a virtual twinborn body, reconstructing physical parameters, a communication protocol and a behavior mode of a target unmanned aerial vehicle, performing depth comparison by combining a disguise sample library generated by the antagonism generation network, and generating a holographic target file containing a three-dimensional motion vector, an energy feature and a potential threat intention; a target quantum encryption communication link is interfered through a quantum entanglement signal, a controllable plasma cloud cluster is generated, navigation and image transmission signals of the controllable plasma cloud cluster are selectively attenuated, the controllability of the target is judged, and an uncontrollable target is repelled through coded sound waves; and sending a control instruction containing a biological heuristic obstacle avoidance algorithm in stages, and guiding the control instruction to an intelligent recovery cabin. According to the invention, accurate detection, intelligent identification, safe countering and reliable takeover of the unmanned aerial vehicle are realized, and the airspace safety guarantee capability is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of server heat dissipation, and in particular relates to a method for detecting and countering unmanned aerial vehicles (UAVs). Background Art

[0002] Currently, drone detection and countermeasure technology has developed in parallel across multiple technology paths, but there are still many limitations in practical applications: Limitations of detection methods: Traditional drone detection relies on single-spectrum monitoring or optical imaging, which struggles to cover wide-band drone signals in complex electromagnetic environments. Furthermore, it has low accuracy for drones equipped with camouflage or stealth materials. For example, drones using frequency-hopping communications or radar-absorbing materials can effectively evade single-band monitoring, resulting in a high rate of missed detections.

[0003] Target Identification and Threat Assessment Deficiencies: Existing identification technologies often rely on matching against pre-set feature libraries. This lack of adaptive learning capabilities makes misjudgment prone to new drones or camouflaged targets. Furthermore, threat assessment relies heavily on location information, making it difficult to deeply analyze the drone's physical parameters and behavioral patterns, leading to inadequately targeted countermeasures.

[0004] The limited and invasive nature of countermeasures: Current countermeasures primarily rely on electromagnetic interference (EMI) or physical destruction. The former can easily disrupt surrounding civilian communications, while the latter can cause secondary damage after a drone crashes. Traditional EMI is ineffective against high-end drones that utilize quantum encryption. Physical destruction is also ineffective against targets requiring recovery and evidence collection.

[0005] Technical bottlenecks in takeover control: Existing drone takeover methods often rely on cracking model-specific communication protocols, resulting in poor versatility and a lack of accurate modeling of the dynamic response of the drone's flight control system. Takeover commands can easily fail when the drone is in an interference environment or has hardware discrepancies, making smooth guidance to the recovery area difficult. Obstacle avoidance is particularly limited in complex terrain.

[0006] Inefficiency in multi-technology collaboration: Detection, identification, countermeasures, and control are often independent systems, leading to delays in data transmission and decision-making, making it difficult to respond to high-speed or evasive drone targets. For example, information asynchrony between the detection and countermeasure modules can delay countermeasure timing, reducing interception success rates.

[0007] Therefore, there is an urgent need to improve the drone detection and countermeasure methods in the existing technology to solve the above-mentioned technical problems existing in the existing technology. Summary of the Invention

[0008] The purpose of the present invention is to provide a drone detection and countermeasure method to achieve full-process and efficient disposal of various types of drones, while ensuring airspace safety and taking into account the accuracy and safety of countermeasures.

[0009] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows: A method for detecting and countering drones, comprising the following steps: S1: Deploy a multi-dimensional perception network to instantly scan the entire frequency band from 0.1 to 18 GHz, and simultaneously collect multi-dimensional features in the airspace, including noise features of radio signals, material features of the terahertz band, and Doppler effect features; S2: Mapping multi-dimensional features to a virtual twin, reconstructing the target drone's physical parameters, communication protocols, and behavioral patterns through digital threading technology, and performing deep comparisons with a camouflage sample library generated by a generative adversarial network. When the confidence level exceeds a dynamic threshold, a holographic target profile containing 3D motion vectors, energy characteristics, and potential threat intent is generated. S3: By modulating the quantum entangled signal in the specified frequency band, the quantum encrypted communication link of the drone is non-invasively interfered with, a controllable plasma cloud is generated around the target drone, and its navigation and image transmission signals are selectively attenuated. The target drone is then judged to be controllable. If not, a coded sound wave matching the structural resonance frequency of the target drone is emitted to achieve forced expulsion without physical damage. If so, step S4 is executed. S4: Activate the takeover system based on brain-computer interface mapping, analyze the electrical signal characteristics of the drone flight control system, establish a neuromorphic mapping model of control instructions, and send control instructions containing bio-inspired obstacle avoidance algorithms in stages to guide the target to the intelligent recovery cabin.

[0010] Preferably, the multi-dimensional sensing network includes a metamaterial antenna array, a weak noise detection unit, and a terahertz imaging module. The specific process of the multi-dimensional sensing network instantaneously scanning the entire frequency band and synchronously collecting the noise characteristics of the radio signal in the airspace, the material characteristics of the terahertz band, and the Doppler effect characteristics in step S1 is as follows: S11: Initialize the multi-dimensional perception network. The metamaterial antenna array automatically adjusts the array aperture according to the preset monitoring range to form a three-dimensional beam topology that adapts to the full frequency band of 0.1-18 GHz. The weak noise detection unit completes the noise baseline calibration and establishes the recognition baseline of the -170 dBm weak signal. The terahertz imaging module is preheated to a stable working state, and the material feature sampling interval of the 0.3-3 THz frequency band is set. S12: Performing full-band instantaneous scanning: The metamaterial antenna array uses spatial diversity technology to divide the 0.1-18 GHz frequency band into 128 parallel detection channels. Synchronous excitation of each channel is achieved through photonic crystal delay lines, completing full-band instantaneous scanning. The real-time signal preprocessing unit performs real-time Fourier transform on the received signal to extract frequency agility characteristics and channel occupancy patterns. S13: Feature Acquisition and Synchronous Fusion: The metamaterial antenna array captures the noise spectrum of the signal, separates the target signal from the ambient noise through a vacuum noise comparison algorithm, records the signal's phase noise, amplitude fluctuation, and frequency drift characteristics, and generates a noise feature fingerprint library. The terahertz imaging module transmits coherent terahertz waves and receives the echo signal after penetrating the target drone. Time-domain terahertz spectroscopy is used to analyze the characteristic absorption peaks of different materials, reconstruct the dielectric constant distribution map of the target drone's internal structure, and distinguish between the payload and the body structure. The coherent Doppler radar principle is used to measure the frequency offset generated by the moving target. Combined with multi-antenna phase difference calculation, a three-dimensional motion vector is generated. The rotation frequency and number of blades of the rotor and propeller are analyzed through the micro-Doppler effect. S14: Perform real-time data association and calibration: The time synchronization module uses the atomic clock to align the timestamps of multi-dimensional feature acquisition. The spatial coordinate conversion unit maps the observation data of different sensors to a unified coordinate system. The Kalman filter algorithm fuses and reduces noise on multi-dimensional features.

[0011] Preferably, in step S2, the multi-dimensional features are mapped to the virtual twin, and the physical parameters, communication protocol and behavior pattern of the target UAV are reconstructed through digital thread technology. The specific process is as follows: S21: Deploy a digital thread system consisting of a data center, a time series database, and distributed computing nodes. Use the Protobuf-based feature data encapsulation format to define data interaction standards and establish an end-to-end data link from perception features to virtual mapping and attribute reconstruction. S22: Synchronously access multi-dimensional feature data through the edge gateway interface of the digital thread system; S23: Multi-feature collaborative mapping based on digital thread system: Mapping the noise characteristics of radio signals to communication properties: converting the noise fingerprint vector in the noise characteristics into the electromagnetic characteristics of the virtual twin; The noise power spectrum is used to infer the working frequency band, transmission power and modulation mode of the communication module, and a noise interference model of the virtual communication link is constructed. Record the intermediate parameters of the feature mapping process to form a traceable communication feature mapping thread; Mapping terahertz material characteristics to physical parameters: A digital thread drives a reverse modeling engine, converting terahertz point cloud data into a 3D model of the target drone. The material properties of each component are determined by matching the material absorption spectrum to a virtual material library. Calculate physical parameters based on material properties: Calculate the mass of the entire machine through density and volume, infer the maximum load through structural strength parameters, and estimate the endurance through the motor's thermal radiation characteristics; Mapping Doppler effect characteristics to motion models: The motion parameters are input into the dynamics simulation module, and the real-time posture of the virtual twin is calculated using the six-degree-of-freedom equation to reproduce the flight trajectory of the physical target. Analyze the micro-Doppler modulation spectrum to extract propeller parameters and build an output characteristic model of the virtual power system; S24: Reconstruct the core attributes of the target drone. Physical parameter reconstruction: Combined with the material feature mapping results in the terahertz band, a set of physical properties including geometric parameters, mass parameters, and dynamic parameters is generated. The payload type is identified based on the structural point cloud data, and the payload weight is inferred from the volume to construct a physical performance boundary model of the virtual twin. Communication protocol reconstruction: Deeply analyze noise characteristics, identify frequency hopping protocols through frequency hopping patterns, and analyze data frame structures through noise burst periods. Combined with a designated drone communication protocol library, the target protocol type is inferred through feature matching, and an equivalent protocol stack is deployed in the virtual twin. Behavioral pattern reconstruction: The digital thread calls the timing analysis module to classify the Doppler trajectory data into behaviors and extract the characteristic parameters of each behavior.

[0012] Preferably, in step S2, a depth comparison is performed with the camouflage sample library generated by the adversarial generative network. When the confidence exceeds a dynamic threshold, a holographic target file containing three-dimensional motion vectors, energy characteristics, and potential threat intentions is generated. The specific process is as follows: S25: Deploy a multimodal adversarial generative network, which includes a generator and a discriminator. The generator uses the DCGAN architecture; the discriminator uses the ResNet-50 network to distinguish between real samples and disguised samples. S26: Generate camouflage sample library: Input dimension: Integrates the noise characteristics of radio signals, the material characteristics of the terahertz band, and the Doppler effect characteristics as conditional constraints; Generate camouflage samples: three-dimensional virtual samples including physical camouflage samples, signal camouflage samples, and behavioral camouflage samples; Dynamic expansion of the disguise sample library: After processing every 100 real target cases, the GAN model is incrementally trained to incorporate newly discovered disguise features into the generator; S27: Deeply compare the multi-dimensional features with the disguised sample library: Physical feature comparison: Calculate the geometric similarity between the target drone 3D model and the physical camouflage sample through IoU; Signal feature comparison: Cosine similarity is used to calculate the matching degree between the target noise fingerprint vector and the signal disguise sample, identifying frequency hopping pattern disguise and power hiding features; Behavioral feature comparison: The motion trajectory is converted into a behavioral embedding vector through the LSTM encoder and compared with the behavioral disguise samples in the sample library; S28: Holographic target file generation and threat intention analysis: 3D motion vector: Based on Doppler effect feature analysis, a 3D motion parameter set including instantaneous velocity, acceleration, and angular velocity is generated, along with a trajectory prediction curve. Energy characteristics: terahertz thermal imaging characteristics are used to infer battery capacity, remaining power, and battery life estimation, and mark the energy supply demand level; Potential threat intention assessment: Obtain behavioral parameters including the distance between the target drone's flight trajectory and the designated sensitive area, the length of stay, and the movement pattern; use graph neural networks to analyze the correlation between the target drone's behavior and preset threat cases, and output threat intention labels. Threat intention labels include reconnaissance, crossing the boundary, and suspicious payload delivery. For target drones with high threat intentions, automatically associate their physical parameters including payload volume to generate a potential hazard assessment and obtain an intention assessment report.

[0013] Preferably, the specific process of step S3 is as follows: S31: Target communication frequency band detection: Based on the UAV communication protocol parameters reconstructed by the virtual twin, the target frequency band of interference is determined and the quantum state characteristics of the target signal are synchronously collected; S32: Quantum entangled signal generation: The quantum signal generator generates entangled photon pairs through spontaneous parametric down-conversion, where one beam of photons retains a local reference, while the other beam is modulated to the target communication frequency band through a tunable filter. Based on the principle of quantum teleportation, the quantum state characteristics of the local reference photon are mapped to the transmitted photon, causing the interference signal to be partially entangled with the quantum state of the target communication link. The interference signal adopts a pulse modulation mode. S33: Non-intrusive jamming implementation: The directional transmitting antenna focuses the quantum entangled signal onto the target drone; the phase and amplitude of the jamming signal are adjusted in real time through quantum measurement feedback, reducing the quantum bit error rate of the target communication link from 10 -6 Level up to 10 -2 Level; synchronously monitor the adaptive adjustment of the target communication link and dynamically update the interference parameters; S34: Controllable plasma cloud generation and signal attenuation regulation: Plasma cloud parameter planning: Calculate the coverage of the plasma cloud based on the drone size from the holographic target archive; According to the type of signal to be attenuated, the electron density and collision frequency of the plasma are determined so that the attenuation rate of the cloud for the target frequency band is ≥30dB and the attenuation for other frequency bands is ≤5dB. Plasma cloud generation and maintenance: The microwave excitation device emits directional millimeter waves, focusing on the air area around the target drone, and generates plasma by ionizing air molecules; Real-time monitoring of cloud density distribution, automatically increasing microwave excitation power in the corresponding area when the local density falls below a threshold; Selective signal attenuation verification: The spectrum monitoring module synchronously collects the changes in the target drone's navigation signal and image transmission signal strength to verify the attenuation effect; S35: Determine the controllability of the target drone: Communication link response test: Under the synergistic effect of quantum interference and plasma attenuation, a standardized control command test packet is sent to the target drone; the target's response signal is monitored, and signal analysis is used to determine whether its flight control system still maintains the ability to receive commands; Dynamic state assessment: Analyze the target's motion stability in a disturbed environment based on Doppler effect characteristics and virtual twin models. Evaluate the power system state: Use terahertz thermal imaging to monitor motor temperature changes and, combined with speed fluctuations, determine whether the motor is still in a controllable operating state. Quantitative controllability scoring: Evaluates communication responsiveness, motion stability, and power system status; each indicator is scored out of 100 points, with a combined score of 70 or higher indicating controllability, 50-69 indicating limited controllability, and <50 indicating complete loss of control. For controllable targets, a takeover protocol corresponding to their model is further matched to provide a basis for subsequent takeover control.

[0014] Preferably, it also includes dynamically adjusting the countermeasure effect and setting a termination mechanism: Continuously monitor the target drone's status and automatically increase the quantum interference intensity or expand the plasma cloud coverage when the score drops below a preset threshold; If the target is judged to be completely out of control and has a low risk of falling, the countermeasure strength will be gradually reduced, linearly decaying to off within 10 seconds.

[0015] Preferably, in step S4, the takeover system based on brain-computer interface mapping is activated, and the specific process of establishing a neuromorphic mapping model of control instructions by analyzing the electrical signal characteristics of the drone flight control system is as follows: S41: Brain-computer interface mapping takes over system initialization: Flight control system electrical signal feature analysis: The real-time electrical signal features of the core control unit are extracted by capturing the UAV flight control system communication data during the quantum entangled signal interference phase; The electrical signal features are transformed into characteristic waves through wavelet transform, and a mapping relationship library between electrical signal features and physical actions is established as a benchmark for subsequent instruction generation.

[0016] Take over system hardware activation: load the pre-trained UAV flight control system response model, initialize the communication interface, and activate the directional command transmission array; Simultaneously establish a communication link with the intelligent recovery capsule to obtain the recovery capsule coordinates, hatch status, and a three-dimensional map of the surrounding airspace; S42: Construction of neuromorphic mapping model for control instructions: Modeling the response characteristics of the flight control system: Send 10 sets of calibration commands to the drone, record the feedback characteristics of the electrical signal characteristics, and build a command-response dynamic model using a Bayesian network; Based on the physical parameters of the virtual twin, the model parameters are modified so that the deviation between the virtual simulation command response and the actual measurement is ≤3%; Setting up neuromorphic control command generation rules: Using a spiking neural network to simulate the discharge patterns of biological neurons, traditional control commands are converted into spatiotemporally encoded pulse sequences. Introducing the Hebbian learning rule ensures that the command pulse sequence resonates with the characteristic frequency bands of the drone's electrical signal characteristics. Activate the takeover system based on brain-computer interface mapping, analyze the electrical signal characteristics of the UAV flight control system, and establish a neuromorphic mapping model of control instructions. Preferably, the specific process of sending control instructions including the bio-inspired obstacle avoidance algorithm in stages in step S4 to guide the target to the intelligent recovery cabin is as follows: S43: Embedding the biologically inspired obstacle avoidance algorithm: Obstacle avoidance algorithm simulates biological mechanisms: Drawing on the swarm intelligence mechanism of bee swarm navigation, it identifies obstacles through residual optical sensor data from drones or environmental data from ground monitoring, and combines it with acceleration sensor data to predict collision risks; It simulates the principle of bat echolocation, generates virtual sound wave detection signals, calculates obstacle distance and avoidance angle, and outputs three basic obstacle avoidance actions: emergency pull-up, lateral avoidance, and deceleration. S44: Dynamically plan the recovery path: With the intelligent recovery capsule as the endpoint, the A* algorithm is used to generate an initial three-dimensional path. The path is updated based on real-time positioning data, and local replanning is triggered when the deviation error is greater than 3m. When the distance to the recovery capsule is ≤100m, the spiral descent path mode is automatically switched. S45: Staged control command transmission and status closed loop: The first stage is takeover confirmation: Send low-intensity neuromorphic commands and monitor the response characteristics of the drone's electrical signal signature; Gradually increase the complexity of instructions and preview the execution effects of instructions through virtual twins; If there is no response to the specified command for consecutive times, the command strength will be automatically increased while maintaining the quantum interference at a low intensity; The second stage is path guidance: Send instructions according to the planned path and waypoints. Each waypoint instruction includes three-dimensional coordinates and arrival time limit. Embedded bio-inspired obstacle avoidance algorithm: When an obstacle is detected, it automatically inserts obstacle avoidance instructions and automatically returns to the original path after avoiding it; The third stage is docking with the recovery capsule: when it reaches the specified distance from the recovery capsule, it sends a hovering command to activate the recovery capsule's laser positioning system; Based on the laser positioning data, refined control instructions are sent to gradually bring the drone into the capture range of the recovery capsule; After the drone enters the recovery cabin, it sends a power shutdown command, triggering the recovery cabin's electromagnetic buffer device to generate a gradient magnetic field to achieve contactless deceleration and complete a soft landing.

[0017] The beneficial effects of the present invention include: The drone detection and countermeasure method provided by this invention first deploys a multi-dimensional perception network consisting of a metamaterial antenna array, a weak noise detection unit, and a terahertz imaging module to achieve instantaneous scanning across the entire frequency band from 0.1 to 18 GHz, simultaneously collecting radio signal noise characteristics, terahertz material characteristics, and Doppler effect characteristics. Compared to traditional single-spectrum or optical detection methods, this method covers a wider frequency band and has richer feature dimensions. Combined with Kalman filtering and fusion noise reduction technology, it can effectively identify drones using frequency-hopping communications and material camouflage, improving target capture rates and significantly enhancing the early detection capability of low-detectability targets.

[0018] Secondly, digital threading technology is used to map multi-dimensional features to a virtual twin, reconstructing the target's physical parameters, communication protocols, and behavioral patterns. This is then combined with a camouflage sample library from a generative adversarial network for in-depth comparison. This approach transcends the limitations of traditional feature library matching and improves the accuracy of camouflaged target recognition by generating multiple camouflage samples, including physical, signal, and behavioral ones. Furthermore, based on the 3D motion vectors, energy signatures, and threat intent analysis of the holographic target archive, this allows for an upgrade from target identification to threat recognition, providing a decision-making basis for countermeasure strategies.

[0019] Thirdly, the coordinated countermeasure of using quantum entangled signals to interfere with quantum encrypted communication links and controllable plasma clouds is non-invasive, avoiding physical damage to drones compared to traditional electromagnetic suppression or physical destruction methods, facilitating subsequent evidence collection or recovery; selective attenuation technology reduces interference with surrounding civilian communications and improves electromagnetic compatibility; combined with coded acoustic expulsion, non-destructive forced expulsion is achieved, reducing the risk of secondary crashes.

[0020] Finally, the takeover system based on brain-computer interface mapping analyzes the characteristics of electrical signals and constructs a neuromorphic mapping model, so that the control instructions and the flight control system form a frequency band resonance and the takeover response delay is shortened; a bio-inspired obstacle avoidance algorithm is embedded to improve the success rate of obstacle avoidance in complex terrain; phased command sending and docking with the intelligent recovery cabin achieve a smooth soft landing of the drone, significantly improving the recovery success rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the flow of the drone detection and countermeasure method of the present invention.

[0022] Figure 2 Schematic diagram of the spatial diversity principle of the present invention.

[0023] Figure 3 Schematic diagram of the terahertz imaging module of the present invention. DETAILED DESCRIPTION

[0024] The following is combined with Figure 1~Figure 3 The present invention is described in further detail: Example 1 See attached Figure 1 As shown, a drone detection and countermeasure method includes the following steps: S1: Deploy a multi-dimensional perception network consisting of a metamaterial antenna array, a weak noise detection unit, and a terahertz imaging module. This network instantaneously scans the entire 0.1-18 GHz frequency band, simultaneously capturing multi-dimensional characteristics within the airspace, including the noise signature of the radio signal, the material characteristics of the terahertz band, and the Doppler effect. The metamaterial antenna array, based on artificially designed microstructures (metasurfaces, tunable elements, etc.), can precisely manipulate the propagation path and phase of electromagnetic waves at subwavelength scales, thereby dynamically adjusting the antenna's radiation characteristics.

[0025] S2: Map multi-dimensional features to a virtual twin, reconstruct the physical parameters, communication protocols, and behavior patterns of the target drone through digital thread technology, and perform deep comparison with the camouflage sample library generated by the adversarial generative network. When the confidence level exceeds the dynamic threshold, a holographic target profile containing three-dimensional motion vectors, energy characteristics, and potential threat intentions is generated.

[0026] S3: By modulating the quantum entangled signal in the specified frequency band, non-invasive interference is performed on the quantum encrypted communication link of the drone, generating a controllable plasma cloud around the target drone, selectively attenuating its navigation and image transmission signals, and judging whether the target drone is controllable. If not, a coded sound wave matching the structural resonance frequency of the target drone is emitted to achieve forced expulsion without physical damage. If so, execute step S4.

[0027] S4: Activate the takeover system based on brain-computer interface mapping, analyze the electrical signal characteristics of the drone flight control system, establish a neuromorphic mapping model of control instructions, and send control instructions containing bio-inspired obstacle avoidance algorithms in stages to guide the target to the intelligent recovery cabin.

[0028] In step S1, the multi-dimensional perception network instantaneously scans the entire frequency band from 0.1 to 18 GHz, and simultaneously collects noise characteristics of radio signals, material characteristics of the terahertz band, and Doppler effect characteristics in the airspace. The specific process is as follows: S11: Initialize the multi-dimensional perception network. The metamaterial antenna array automatically adjusts the array aperture according to the preset monitoring range. The array aperture is continuously adjustable within the range of 1-5 meters, forming a three-dimensional beam topology structure adapted to the 0.1-18GHz frequency band. The weak noise detection unit completes the noise baseline calibration and establishes the recognition baseline of the -170dBm level weak signal. The terahertz imaging module is preheated to a stable working state, and the material feature sampling interval of the 0.3-3THz frequency band is set.

[0029] S12: Perform instantaneous scanning of the entire frequency band: See Figure 2 As shown in the figure, the metamaterial antenna array uses spatial diversity technology to divide the 0.1-18 GHz frequency band into 128 parallel detection channels. The photonic crystal delay line is used to achieve synchronous excitation of each channel, completing 50ns instantaneous scanning of the entire frequency band. The real-time signal preprocessing unit performs real-time Fourier transform on the received signal to extract the frequency agility characteristics and channel occupancy patterns.

[0030] See also Figure 2 As shown in the figure, the signals received by the n antennas in the antenna array cover the full frequency band of 0.1-18 GHz. G1-Gn are broadband gain conditioning modules, which perform gain compensation on the full-band signals received by each antenna to balance the receiving gains of different antennas and perform preliminary filtering to remove out-of-band interference, but retain the full frequency band of 0.1-18 GHz. The modules output full-band, multi-antenna parallel signal streams to prepare for subsequent channelization.

[0031] The switching logic or demodulator integrates a channelizer to perform two key operations: frequency band segmentation, which divides the 0.1-18 GHz frequency range into sub-bands, and spatial-frequency parallel processing, which feeds the signal streams from n antennas into a 128-channel processing chain. This means that through channelization within the switching logic and demodulator, the full-band signal from each antenna is divided into multiple sub-bands (channels). This leverages the spatial parallelism of multiple antennas to ultimately achieve parallel detection across all 128 channels.

[0032] S13: Feature acquisition and synchronous fusion: The metamaterial antenna array captures the noise spectrum of the signal, separates the target signal from the ambient noise through a vacuum noise comparison algorithm, records the phase noise, amplitude fluctuation and frequency drift characteristics of the signal, and generates a noise feature fingerprint library. The terahertz imaging module transmits coherent terahertz waves and receives the echo signal after penetrating the target drone. The characteristic absorption peaks of materials including plastics, metals, and composite materials are analyzed through time-domain terahertz spectroscopy technology, and the dielectric constant distribution map of the internal structure of the target drone is reconstructed to distinguish between the payload and the body structure. The coherent Doppler radar principle is used to perform high-precision measurement of the frequency offset generated by the moving target. Combined with multi-antenna phase difference calculation, a three-dimensional motion vector including velocity, acceleration, and angular velocity is generated. The rotation frequency and blade number characteristics of the rotor and propeller are analyzed through the micro-Doppler effect.

[0033] The principle diagram of the terahertz imaging module can be found in Figure 3 As shown, a laser generates ultrashort laser pulses, providing a pump source for terahertz wave excitation. The laser is focused onto the emitter, where the photoconductivity or optical rectification effect converts the optical pulse energy into terahertz pulses, instantaneously exciting free carriers or polarized electric fields, radiating terahertz waves onto the sample. The terahertz detector, through electro-optical sampling, converts the electric field changes of the terahertz waves into measurable optical signal changes. A current preamplifier amplifies the weak current signal output by the detector. An A / D converter and digital signal processing convert the analog electrical signal into a digital signal. Using algorithms such as Fourier transform, denoising, and spectrum analysis, the amplitude, phase, and spectral information of the terahertz wave are extracted, ultimately inverting the sample's optical parameters (refractive index, absorption coefficient) and structural image. A DC bias applies a DC voltage to the terahertz emitter to optimize terahertz wave excitation.

[0034] S14: Perform real-time data association and calibration: The time synchronization module uses an atomic clock to achieve 1ns-level timestamp alignment for multi-dimensional feature acquisition. The spatial coordinate conversion unit maps the observation data of different sensors to a unified coordinate system. The Kalman filter algorithm fuses and reduces noise on multi-dimensional features to improve the detection sensitivity of low-speed / hovering targets.

[0035] In step S2, the multi-dimensional features are mapped to the virtual twin, and the physical parameters, communication protocols, and behavior patterns of the target UAV are reconstructed through digital thread technology. The specific process is as follows: S21: Deploy a digital thread system that includes a data middle platform, a time series database, and distributed computing nodes. Use the Protobuf-based feature data encapsulation format to define data interaction standards and establish an end-to-end data link of perception features, virtual mapping, and attribute reconstruction.

[0036] S22: Synchronously access multi-dimensional feature data through the edge gateway interface of the digital thread system; Noise characteristics of radio signals: noise fingerprint vector, phase noise power spectrum, frequency drift coefficient time series data; Material characteristics in the terahertz band: three-dimensional structure point cloud, material absorption spectrum, and dielectric constant distribution matrix; Doppler effect characteristics: including three-dimensional motion parameters of velocity, acceleration, and angular velocity, micro-Doppler modulation spectrum, and trajectory coordinate sequence.

[0037] S23: Multi-feature collaborative mapping based on digital thread system: The noise characteristics of the radio signal are mapped to the communication properties: the digital thread system calls the electromagnetic simulation module to convert the noise fingerprint vector into the electromagnetic characteristics of the virtual twin, reproducing the hardware characteristics of the signal transmitter in the virtual space, including the oscillator phase noise and the nonlinear distortion of the power amplifier.

[0038] The noise power spectrum is used to infer the operating frequency band, transmission power and modulation methods including AM / FM / FSK of the communication module, and a noise interference model of the virtual communication link is constructed; the digital thread system records the intermediate parameters of the feature mapping process, including the simulation step size and electromagnetic environment coefficient, to form a traceable communication feature mapping thread.

[0039] Mapping of material characteristics in the terahertz band to physical parameters: The digital thread drives the reverse modeling engine, converting terahertz point cloud data into a 3D model of the target drone, including the geometric parameters of internal structures such as the motor, battery, and circuit board. Material absorption spectra are matched against a virtual material library to determine the material properties of each component. Physical parameters are calculated based on these material properties: The vehicle's mass is calculated from density and volume, the maximum load is inferred from structural strength parameters, and flight time is estimated from the motor's thermal radiation characteristics.

[0040] Doppler effect characteristics are mapped to the motion model: The digital thread system inputs motion parameters into the dynamics simulation module and calculates the real-time attitude of the virtual twin, including pitch, roll, and yaw angles, through six-degree-of-freedom equations, driving the digital model to reproduce the flight trajectory of the physical target. The micro-Doppler modulation spectrum is analyzed to extract propeller parameters: number of blades, speed, and pitch, and to construct an output characteristic model of the virtual power system. The digital thread system records the time-series correlation data of motion characteristics, including the causal relationship of speed change → thrust change → acceleration change, providing a dynamic basis for behavioral pattern reconstruction; S24: Reconstruct the core attributes of the target drone. Physical Parameter Reconstruction: The digital thread integrates the mapping results of material characteristics in the terahertz band to generate a set of physical properties including geometric parameters, mass parameters, and power parameters. Geometric parameters include length, wingspan, and altitude; mass parameters include total weight and center of gravity; and power parameters include maximum lift and endurance.

[0041] Based on the structural point cloud data, the payload types including cameras, batteries, and special equipment are identified, and the load weight is inferred through volume to construct the physical performance boundary model of the virtual twin. The physical performance boundaries include the maximum climb rate and wind resistance level.

[0042] Communication Protocol Reconstruction: The digital thread system deeply analyzes the noise characteristics of radio signals, identifying frequency-hopping protocols through frequency hopping patterns and analyzing data frame structures, including frame length and interval time, based on noise burst cycles. By combining known drone communication protocol libraries, including DJI Lightbridge and WiFi 802.11n, the target protocol type is inferred through feature matching, and an equivalent protocol stack is deployed in the virtual twin.

[0043] Behavioral pattern reconstruction: The digital thread system calls the timing analysis module to classify the Doppler trajectory data into behaviors, including cruising, hovering, circling, and diving, and extract the characteristic parameters of each behavior, including the cruising speed range and the fluctuation amplitude of the hovering position.

[0044] Reproduce typical behavior patterns in the virtual twin: simulate low-battery return behavior by adjusting the virtual power system parameters, and simulate no-fly zone avoidance behavior by modifying the navigation logic.

[0045] In step S2, a depth comparison is performed with the camouflage sample library generated by the adversarial generative network. When the confidence exceeds the dynamic threshold, a holographic target file containing three-dimensional motion vectors, energy characteristics, and potential threat intentions is generated. The specific process is as follows: S25: Deploy a multimodal adversarial generative network, which includes a generator and a discriminator. The generator uses the DCGAN architecture; the discriminator uses the ResNet-50 network to distinguish between real samples and disguised samples. S26: Generate camouflage sample library: Input dimension: Fusion of the noise feature vector of the radio signal, the material characteristics of the terahertz band, and the Doppler motion parameter sequence as conditional constraints; Camouflage sample library generation: This generates three-dimensional virtual samples, including physical, signal, and behavioral camouflage samples. Physical camouflage samples include bird-like shapes and camouflage coatings; signal camouflage samples include frequency hopping and noise camouflage; and behavioral camouflage samples include civil aviation flight path imitation and hovering camouflage. The camouflage sample library is categorized and stored by camouflage type, including physical, signal, and behavioral camouflage libraries. The total capacity is dynamically maintained at around 100,000, supporting fast retrieval.

[0046] S27: Deep comparison of multi-dimensional features and disguised sample library: Physical feature comparison: The geometric similarity between the target 3D model and the physical camouflage samples in the camouflage sample library is calculated through IoU (Intersection over Union), focusing on the camouflaged parts, including the radar stealth design of the wing edges; Signal feature comparison: Cosine similarity is used to calculate the matching degree between the target noise fingerprint vector and the signal disguise sample, identifying features such as frequency hopping pattern disguise and power hiding; Behavioral feature comparison: The motion trajectory is converted into a behavioral embedding vector using an LSTM encoder and compared with behavioral camouflage samples in the sample library, including reconnaissance camouflage and evasion camouflage. The LSTM encoder is the encoding part of an autoencoder implemented using a long short-term memory network. It is used to compress the input motion trajectory data into a fixed-length vector representation. The specific process is as follows: The LSTM encoder primarily consists of an embedding layer and an LSTM layer. The embedding layer converts each coordinate point in the motion trajectory into an embedding vector of fixed dimension, typically 128 or 256. The LSTM layer processes the embedding vector sequence, extracting long-term dependency features through the coordinated action of forget gates, input gates, and output gates, generating hidden states and cell states. The encoder's final output of hidden states and cell states is combined into a behavioral embedding vector. This vector contains the trajectory's global motion features, including velocity and acceleration.

[0047] S28: Holographic target file generation and threat intention analysis: 3D motion vector: Based on Doppler effect feature analysis, a 3D motion parameter set including instantaneous velocity, acceleration, and angular velocity is generated, along with a trajectory prediction curve. Energy characteristics: The battery capacity, remaining power, and battery life are estimated through terahertz thermal imaging characteristics, and the energy supply demand level is marked, including high, medium, and low demand levels.

[0048] Potential Threat Intent Assessment: This system collects behavioral parameters, including the target drone's flight path, distance from designated sensitive areas, duration of stay, and movement patterns. These include military bases and nuclear power plants, and movement patterns such as low-altitude penetration and high-altitude reconnaissance. A graph neural network (GNN) analyzes the correlation between the target drone's behavior and pre-defined threat scenarios, outputting threat intent labels such as reconnaissance, boundary violations, and suspicious payload delivery. For high-threat-intent target drones, the system automatically correlates their physical parameters, including payload size, to generate a potential hazard assessment and a threat intent assessment report.

[0049] The specific process of step S3 is as follows: S31: Target communication frequency band detection: Based on the UAV communication protocol parameters reconstructed by the virtual twin, such as the operating frequency band and modulation mode of the quantum encryption link, the target interference frequency band is determined, usually the 1-6 GHz commonly used frequency band for quantum encryption communication, and the quantum state characteristics of the target signal, including the photon polarization direction and entanglement degree parameters, are synchronously collected; S32: Quantum entangled signal generation: The quantum signal generator generates entangled photon pairs through spontaneous parametric down-conversion (SPDC), where one beam of photons retains a local reference, while the other beam is modulated to the target communication frequency band through a tunable filter. Based on the principle of quantum teleportation, the quantum state characteristics of the local reference photon are mapped to the transmitted photon, causing the interference signal to partially entangle with the quantum state of the target communication link. The entanglement degree is controlled at 0.3-0.5 to avoid complete disruption of communication. The interference signal adopts pulse modulation mode, and the pulse width (10-100ns) matches the target communication frame structure to ensure that the interference energy is concentrated during the data transmission period. S33: Non-intrusive jamming implementation: The directional transmitting antenna focuses the quantum entangled signal onto the target drone with a beam width of ≤3° to avoid interference with surrounding equipment; the phase and amplitude of the jamming signal are adjusted in real time through quantum measurement feedback, reducing the quantum bit error rate of the target communication link from 10 -6 Level up to 10 -2 level, only affecting data integrity without physically damaging equipment; synchronously monitoring the adaptive adjustment of the target communication link, including frequency hopping and improving encryption strength, and dynamically updating interference parameters.

[0050] S34: Controllable plasma cloud generation and signal attenuation regulation: Plasma cloud parameter planning: Based on the size of the drone in the holographic target file, such as a wingspan of 0.5-2m, the coverage of the plasma cloud is calculated with a diameter of 1.5 times the target size to ensure complete coverage of the target.

[0051] Based on the signal type to be attenuated, including navigation signals (1.575GHz GPS, 1.602GHz Beidou) and image transmission signals (2.4 / 5.8GHz), the electron density and collision frequency of the plasma are determined to ensure that the cloud's attenuation rate for the target frequency band is ≥30dB and its attenuation for other frequency bands is ≤5dB.

[0052] The attenuation of electromagnetic waves by plasma is mainly due to absorption loss and reflection loss. Its attenuation characteristics are related to the frequency of electromagnetic waves. f , plasma electron density n e , the collision frequency of electrons and neutral particles n Closely related.

[0053] The relative complex dielectric constant of plasma is :e r = 1 −ω p 2 / ( n 2 + oh 2 ) −jν·ω p 2 / [ oh; ( n 2 + oh 2 )]; in, ω=2πf , oh is the angular frequency of the electromagnetic wave, oh p is the plasma angular frequency, f is the electromagnetic wave frequency, n e is the plasma electron density, n is the collision frequency between electrons and neutral particles.

[0054] The attenuation rate is determined by the attenuation coefficient of electromagnetic waves propagating in plasma. α calculate: Decay rate = 20 log 10 (E0 / E)=20log10(e αL )≈8.686 αL ; Among them, E0 represents the electric field strength before the electromagnetic wave enters the plasma cloud, that is, the initial electric field strength, and E represents the electric field strength after the electromagnetic wave passes through the plasma cloud, that is, the electric field strength after transmission. When the electromagnetic wave propagates in a lossy medium, the electric field strength satisfies E=E0e −αL, after deformation, we get E0 / E=e αL .

[0055] Plasma cloud generation and maintenance: The microwave excitation device emits directional millimeter waves in the 30-100 GHz range, focusing them on the air around the target drone. This ionizes air molecules to generate plasma. Magnetic confinement technology (magnetic field strength 0.1-0.5 T) is used to maintain the cloud's shape and prevent rapid spread.

[0056] The cloud density distribution is monitored in real time through terahertz imaging feedback. When the local density is lower than the threshold, the microwave excitation power in the corresponding area is automatically enhanced.

[0057] Selective signal attenuation verification: The spectrum monitoring module simultaneously collects changes in the target drone's navigation signal and image transmission signal strength to verify the attenuation effect, ensuring that the attenuation rate for civil communication frequency bands (900MHz GSM, 1.8GHz LTE) is ≤10% to avoid secondary interference.

[0058] S35: Determine the controllability of the target drone: Communication link response test: Using the synergistic effects of quantum interference and plasma attenuation, standardized control command test packets are sent to the target drone. These can be commands for hovering, lowering altitude, and other tasks that conform to common drone command formats. The target's response signals, including ACK frames and attitude adjustments, are monitored. Signal analysis is used to determine whether the flight control system is still capable of receiving commands.

[0059] Dynamic state assessment: Based on the Doppler effect and virtual twin models, the target's motion stability in a disturbed environment is analyzed, such as whether there is irregular drift with a displacement greater than 5m / 10s or attitude loss of control with a yaw angle fluctuation greater than 15°. Power system status assessment: Terahertz thermal imaging is used to monitor motor temperature changes and, combined with speed fluctuations, determine whether the system remains in a controllable operating state. Speed ​​stability ≥80% is considered normal.

[0060] Quantitative Controllability Scoring: Communication responsiveness (40% weight), motion stability (30% weight), and powertrain status (30% weight) are scored. Each metric is scored out of 100, and a weighted summation is used to arrive at a comprehensive score. A score of 70 or higher is considered controllable, 50-69 indicates limited controllability, and a score of less than 50 indicates complete loss of control. For controllable targets, we further match their model with the corresponding takeover protocols, including the DJI SDK interface and the universal MAVLink protocol, to provide a basis for subsequent takeover control.

[0061] It also includes dynamic adjustment of counter-effects and setting of termination mechanisms: Continuously monitor the target drone's status. When the score drops below a threshold, the system automatically increases the quantum interference intensity or expands the plasma cloud's coverage. If the target is deemed completely out of control and has a low crash risk, the countermeasure intensity is gradually reduced, linearly decaying to zero within 10 seconds to avoid wasting resources.

[0062] In step S4, the takeover system based on brain-computer interface mapping is activated. By analyzing the electrical signal characteristics of the UAV flight control system, the specific process of establishing the neuromorphic mapping model of the control instructions is as follows: S41: Brain-computer interface mapping takes over system initialization: Analysis of electrical signal characteristics of flight control system: By capturing the drone flight control system communication data during the quantum entangled signal interference phase, the real-time electrical signal characteristics of the core control unit, including the MCU and sensor fusion module, are extracted. These include pulse width modulation (PWM) signal sequences, sensor data interaction frequencies, and control command checksum patterns. The characteristic waves of these electrical signal features are extracted through wavelet transform, including the 10-20Hz characteristic frequency band corresponding to altitude adjustment and the 20-50Hz characteristic frequency band for direction control. A mapping relationship library between electrical signal features and physical actions is established, such as a specific pulse sequence corresponding to a climbing action, which serves as a benchmark for subsequent instruction generation.

[0063] Take over system hardware activation: Load the pre-trained UAV flight control system response model, initialize the communication interface, and activate the directional command transmission array, ensuring that the command transmission beam and the UAV receiving antenna are aligned with an accuracy of ≤0.5° to reduce the signal attenuation caused by the residual plasma cloud. Simultaneously establish a communication link with the intelligent recovery capsule to obtain the capsule's coordinates, hatch status, and a three-dimensional map of the surrounding airspace, including no-fly zones and obstacle data.

[0064] S42: Construction of neuromorphic mapping model for control instructions: Modeling the flight control system's response characteristics: 10 sets of calibration commands are sent to the drone, such as for small climbs and left turns. The feedback characteristics of the electrical signal characteristics, including response delay and signal amplitude changes, are recorded, and a command-response dynamic model is constructed using a Bayesian network. Set the neuromorphic control instruction generation rules: A spiking neural network (SNN) is used to simulate the discharge pattern of biological neurons. The traditional control instruction for height = 10m is replaced by a spatiotemporal coded pulse sequence, where three pulses within 50ms correspond to an increase in height. The Hebbian learning rule is introduced to resonate the command pulse sequence with the characteristic frequency band of the UAV's electrical signal characteristics. For example, the command pulse frequency of the steering control system matches the directional control characteristic frequency band of the flight control system. Activate the brain-computer interface-based takeover system. By analyzing the electrical signal characteristics of the drone's flight control system, a neuromorphic mapping model for control commands is established. The input layer receives preprocessed flight control electrical signal characteristics, and each characteristic corresponds to an input neuron in the model. The output layer corresponds to the drone's control commands. Each command corresponds to an output neuron, and the activation state of the neuron indicates the trigger probability of the command.

[0065] The hidden layer is designed based on the hierarchical structure of biological neural networks. Hidden layer neurons simulate the information integration function of the cerebral cortex, receiving input layer signals through synaptic connections and performing nonlinear processing on the signal characteristics. Synaptic weights are initially assigned based on the correlation between electrical signal characteristics and control commands, such as whether a certain frequency is more likely to trigger a steering command. These weights are then dynamically adjusted through learning.

[0066] In step S4, the specific process of sending control instructions containing the bio-inspired obstacle avoidance algorithm in stages to guide the target to the intelligent recovery capsule is as follows: S43: Embedding the biologically inspired obstacle avoidance algorithm: Biological mechanism simulation of obstacle avoidance algorithm: Drawing on the collective intelligence mechanism of swarm navigation, obstacles are identified through the residual optical sensor data of the drone or the environmental data monitored on the ground, and collision risks are predicted in combination with acceleration sensor data.

[0067] It simulates the principle of bat echolocation, generates virtual sound wave detection signals, and calculates obstacle distance and avoidance angle based on the principle of minimum energy consumption.

[0068] Referencing the biological characteristics of bat sound waves, the key parameters of the virtual sound wave detection signal are designed: Frequency range: Choose 10kHz~100kHz, taking into account both detection accuracy and propagation distance. High-frequency signals are more accurate but attenuate quickly, while low-frequency signals travel farther. Waveform type: Pulsed ultrasonic waves are used, with a typical pulse width of 0.5~5ms. The pulse interval is dynamically adjusted according to the environment, with a long interval in open areas and a short interval in complex environments. Amplitude modulation: The initial transmitted signal amplitude is fixed, and the echo signal amplitude decays with the propagation distance, which conforms to the spherical wave attenuation law: the amplitude is inversely proportional to the distance.

[0069] Simulates the ability of bats to adaptively adjust signals according to the environment: when there are no obstacles, they transmit low-frequency, long-pulse-interval signals to reduce energy consumption; when a suspected obstacle is detected, they automatically switch to high-frequency, short-pulse-interval signals to improve positioning accuracy.

[0070] By analyzing the difference between the echo and the transmitted signal, the distance to the obstacle is calculated. The core principle is to exploit the relationship between time difference and wave speed. In a virtual scenario, a detector (simulating a bat's ear) receives the echo signal and simultaneously records the echo's arrival time, frequency offset, and amplitude.

[0071] Distance calculation principle: Based on the core formula of bat echolocation: obstacle distance = sound wave propagation speed × echo time difference / 2. Among them, echo time difference = echo arrival time - signal transmission time.

[0072] The speed of sound wave propagation is about 343m / s in air. If the time difference between the transmitted signal and the echo is 10ms, the distance = (343m / s × 0.01s) / 2≈1.715m.

[0073] Calculate the avoidance angle based on the principle of minimum energy consumption: When bats avoid obstacles, they tend to choose the path with the lowest energy consumption, such as the smallest turning angle and the shortest detour distance. They need to optimize the angle by combining the direction of the obstacle and their own movement state.

[0074] Determination of obstacle azimuth: Analyze the differences in echoes using a multi-channel virtual auditory sensor (simulating the binaural effect of bats).

[0075] Time difference positioning: The time difference Δt between the echoes reaching the left and right sensors is combined with the speed of the sound wave to calculate the azimuth angle θ: θ = arcsin (sound wave speed × Δt / distance between the two sensors).

[0076] Intensity difference positioning: When an obstacle deviates from the centerline, the echo intensity received by the near-side sensor is higher. The intensity difference is used to assist in correcting the azimuth, ultimately obtaining the azimuth of the obstacle relative to the direction of movement of the detector, such as 30° to the left and 5° directly in front.

[0077] Calculation logic for the minimum energy consumption avoidance angle: Energy consumption is positively correlated with the complexity of the motion trajectory. For example, sharp turns consume more energy than small angle turns. The following conditions must be met: Constraints: The avoidance path must be far away from obstacles, that is, the safety distance ≥ obstacle radius + detector size; Optimization goal: Minimize the steering angle, that is, minimize Δθ, and return to the original target direction after detouring. For example, if the original plan is to fly to point A, it can still arrive efficiently after avoiding it.

[0078] Specific process: Set the detector's current moving direction as the reference (0°) and the obstacle azimuth as θ, which can be set to 20° on the left. Calculate the minimum steering angle for safe detour: If the obstacle is on the left, turn to the right by Δθ. Δθ must be ≥ θ + safety margin angle, and Δθ must be as small as possible. For example, if the obstacle is 20° to the left, the minimum steering angle can be 25°, which avoids the obstacle and reduces energy consumption.

[0079] If the obstacle is directly in front, that is, θ≈0°, choose the minimum left / right turning angle, such as ±30°, and give priority to the side with the smaller deviation from the original target direction. If the original target is on the right, give priority to turning right.

[0080] S44: Perform dynamic planning of the recovery path: With the intelligent recovery capsule as the end point, use the A* algorithm to generate an initial three-dimensional path containing 5-8 waypoints. The path safety redundancy is ≥ 5m, that is, the minimum distance to obstacles is 5m. The route is updated every 500ms based on real-time positioning data, which is derived from residual ground radar or drone GPS signals. Local replanning is triggered when the deviation exceeds 3m. When the distance to the recovery capsule is ≤100m, the aircraft automatically switches to a spiral descent path with a pitch of 20m and a radius of 10m to minimize airflow interference.

[0081] S45: Staged control command transmission and status closed loop: The first stage is takeover confirmation: Send low-intensity neuromorphic commands, such as maintaining the current altitude, monitor the response characteristics of the drone's electrical signal characteristics, and confirm that the command reception success rate is ≥90%; By gradually increasing the complexity of commands, it is possible to perform small-angle steering and preview the command execution effect through the virtual twin to ensure that the actual action deviation is ≤5°; If there is no response to three consecutive commands, the command intensity will be automatically increased, and the pulse amplitude will be increased by 20%. At the same time, the quantum interference will be maintained at a low intensity to avoid complete blockage of communication.

[0082] The second phase involves path guidance, with duration dynamically adjusted based on distance. Commands are issued at each waypoint along the planned path, with each command including 3D coordinates and a time limit. A bio-inspired obstacle avoidance algorithm is embedded: when an obstacle is detected, an avoidance command is automatically inserted, and the system automatically returns to the original path after avoiding it. The effectiveness of command execution is evaluated in real time: the Doppler effect is used to calculate the deviation between the actual position and the waypoint, and correction commands are issued when the deviation exceeds 10m.

[0083] The third stage is docking with the recovery capsule: When the drone is at a specified distance from the capsule, a hover command is sent, activating the capsule's laser positioning system. Based on this laser positioning data, refined control commands are issued, such as a descent rate of 0.5m / s and lateral adjustments of 0.3m, gradually bringing the drone within the capsule's capture range. Once inside the capsule, a power-off command is sent, triggering the capsule's electromagnetic buffer device to generate a gradient magnetic field for contactless deceleration and a soft landing.

Claims

1. A drone detection and countermeasure method, characterized in that: The following steps are involved: S1: Deploy a multi-dimensional perception network to instantly scan the entire frequency band from 0.1 to 18 GHz, and simultaneously collect multi-dimensional features in the airspace, including noise features of radio signals, material features of the terahertz band, and Doppler effect features; S2: Mapping multi-dimensional features to a virtual twin, reconstructing the target drone's physical parameters, communication protocols, and behavioral patterns through digital threading technology, and performing deep comparisons with a camouflage sample library generated by a generative adversarial network. When the confidence level exceeds a dynamic threshold, a holographic target profile containing 3D motion vectors, energy characteristics, and potential threat intent is generated. S3: By modulating the quantum entangled signal in the specified frequency band, the quantum encrypted communication link of the drone is non-invasively interfered with, a controllable plasma cloud is generated around the target drone, and its navigation and image transmission signals are selectively attenuated. The target drone is then judged to be controllable. If not, a coded sound wave matching the structural resonance frequency of the target drone is emitted to achieve forced expulsion without physical damage. If so, step S4 is executed. S4: Activate the takeover system based on brain-computer interface mapping, analyze the electrical signal characteristics of the drone flight control system, establish a neuromorphic mapping model of control instructions, and send control instructions containing bio-inspired obstacle avoidance algorithms in stages to guide the target to the intelligent recovery cabin.

2. The method for detecting and countering a drone according to claim 1, characterized in that: The multi-dimensional sensing network includes a metamaterial antenna array, a weak noise detection unit, and a terahertz imaging module. In step S1, the multi-dimensional sensing network instantaneously scans the entire frequency band and simultaneously collects the noise characteristics of the radio signal in the airspace, the material characteristics of the terahertz band, and the Doppler effect characteristics. The specific process is as follows: S11: Initialize the multi-dimensional perception network. The metamaterial antenna array automatically adjusts the array aperture according to the preset monitoring range to form a three-dimensional beam topology that adapts to the full frequency band of 0.1-18 GHz. The weak noise detection unit completes the noise baseline calibration and establishes the recognition baseline of the -170 dBm weak signal. The terahertz imaging module is preheated to a stable working state, and the material feature sampling interval of the 0.3-3 THz frequency band is set. S12: Performing full-band instantaneous scanning: The metamaterial antenna array uses spatial diversity technology to divide the 0.1-18 GHz frequency band into 128 parallel detection channels. Synchronous excitation of each channel is achieved through photonic crystal delay lines, completing full-band instantaneous scanning. The real-time signal preprocessing unit performs real-time Fourier transform on the received signal to extract frequency agility characteristics and channel occupancy patterns. S13: Feature Acquisition and Synchronous Fusion: The metamaterial antenna array captures the noise spectrum of the signal, separates the target signal from the ambient noise through a vacuum noise comparison algorithm, records the signal's phase noise, amplitude fluctuation, and frequency drift characteristics, and generates a noise feature fingerprint library. The terahertz imaging module transmits coherent terahertz waves and receives the echo signal after penetrating the target drone. Time-domain terahertz spectroscopy is used to analyze the characteristic absorption peaks of different materials, reconstruct the dielectric constant distribution map of the target drone's internal structure, and distinguish between the payload and the body structure. The coherent Doppler radar principle is used to measure the frequency offset generated by the moving target. Combined with multi-antenna phase difference calculation, a three-dimensional motion vector is generated. The rotation frequency and number of blades of the rotor and propeller are analyzed through the micro-Doppler effect. S14: Perform real-time data association and calibration: The time synchronization module uses the atomic clock to align the timestamps of multi-dimensional feature acquisition. The spatial coordinate conversion unit maps the observation data of different sensors to a unified coordinate system. The Kalman filter algorithm fuses and reduces noise on multi-dimensional features.

3. The method for detecting and countering drones according to claim 1, characterized in that: In step S2, the multi-dimensional features are mapped to the virtual twin, and the physical parameters, communication protocols, and behavior patterns of the target UAV are reconstructed through digital thread technology. The specific process is as follows: S21: Deploy a digital thread system consisting of a data center, a time series database, and distributed computing nodes. Use the Protobuf-based feature data encapsulation format to define data interaction standards and establish an end-to-end data link from perception features to virtual mapping and attribute reconstruction. S22: Synchronously access multi-dimensional feature data through the edge gateway interface of the digital thread system; S23: Multi-feature collaborative mapping based on digital thread system: Mapping the noise characteristics of radio signals to communication properties: converting the noise fingerprint vector in the noise characteristics into the electromagnetic characteristics of the virtual twin; The noise power spectrum is used to infer the working frequency band, transmission power and modulation mode of the communication module, and a noise interference model of the virtual communication link is constructed. Record the intermediate parameters of the feature mapping process to form a traceable communication feature mapping thread; Mapping terahertz material characteristics to physical parameters: A digital thread drives a reverse modeling engine, converting terahertz point cloud data into a 3D model of the target drone. The material properties of each component are determined by matching the material absorption spectrum to a virtual material library. Calculate physical parameters based on material properties: Calculate the mass of the entire machine through density and volume, infer the maximum load through structural strength parameters, and estimate the endurance through the motor's thermal radiation characteristics; Mapping Doppler effect characteristics to motion models: The motion parameters are input into the dynamics simulation module, and the real-time posture of the virtual twin is calculated using the six-degree-of-freedom equation to reproduce the flight trajectory of the physical target. Analyze the micro-Doppler modulation spectrum to extract propeller parameters and build an output characteristic model of the virtual power system; S24: Reconstruct the core attributes of the target drone. Physical parameter reconstruction: Combined with the material feature mapping results in the terahertz band, a set of physical properties including geometric parameters, mass parameters, and dynamic parameters is generated. The payload type is identified based on the structural point cloud data, and the payload weight is inferred from the volume to construct a physical performance boundary model of the virtual twin. Communication protocol reconstruction: Deeply analyze noise characteristics, identify frequency hopping protocols through frequency hopping patterns, and analyze data frame structures through noise burst periods. Combined with a designated drone communication protocol library, the target protocol type is inferred through feature matching, and an equivalent protocol stack is deployed in the virtual twin. Behavioral pattern reconstruction: The digital thread calls the timing analysis module to classify the Doppler trajectory data into behaviors and extract the characteristic parameters of each behavior.

4. The method for detecting and countering a drone according to claim 3, wherein: In step S2, a depth comparison is performed with the camouflage sample library generated by the adversarial generative network. When the confidence exceeds the dynamic threshold, a holographic target file containing three-dimensional motion vectors, energy characteristics, and potential threat intentions is generated. The specific process is as follows: S25: Deploy a multimodal adversarial generative network, which includes a generator and a discriminator. The generator uses the DCGAN architecture; the discriminator uses the ResNet-50 network to distinguish between real samples and disguised samples. S26: Generate camouflage sample library: Input dimension: Integrates the noise characteristics of radio signals, the material characteristics of the terahertz band, and the Doppler effect characteristics as conditional constraints; Generate camouflage samples: three-dimensional virtual samples including physical camouflage samples, signal camouflage samples, and behavioral camouflage samples; Dynamic expansion of the disguise sample library: After processing every 100 real target cases, the GAN model is incrementally trained to incorporate newly discovered disguise features into the generator; S27: Deeply compare the multi-dimensional features with the disguised sample library: Physical feature comparison: Calculate the geometric similarity between the target drone 3D model and the physical camouflage sample through IoU; Signal feature comparison: Cosine similarity is used to calculate the matching degree between the target noise fingerprint vector and the signal disguise sample, identifying frequency hopping pattern disguise and power hiding features; Behavioral feature comparison: The motion trajectory is converted into a behavioral embedding vector through the LSTM encoder and compared with the behavioral disguise samples in the sample library; Adversarial Verification: The discriminator classifies a mixed set of multi-dimensional features and disguised samples, and outputs a authenticity score for the multi-dimensional features; When the score is less than 70 points, the target is judged to be possibly disguised, and the enhanced comparison mode is activated, increasing the sample library search volume by 30%; S28: Holographic target file generation and threat intention analysis: 3D motion vector: Based on Doppler effect feature analysis, a 3D motion parameter set including instantaneous velocity, acceleration, and angular velocity is generated, along with a trajectory prediction curve. Energy characteristics: terahertz thermal imaging characteristics are used to infer battery capacity, remaining power, and battery life estimation, and mark the energy supply demand level; Potential threat intention assessment: Obtain behavioral parameters including the distance between the target drone's flight trajectory and the designated sensitive area, the length of stay, and the movement pattern; use graph neural networks to analyze the correlation between the target drone's behavior and preset threat cases, and output threat intention labels. Threat intention labels include reconnaissance, crossing the boundary, and suspicious payload delivery. For target drones with high threat intentions, automatically associate their physical parameters including the payload volume to generate a potential hazard assessment and obtain an intention assessment report.

5. The method for detecting and countering drones according to claim 1, characterized in that: The specific process of step S3 is as follows: S31: Target communication frequency band detection: Based on the UAV communication protocol parameters reconstructed by the virtual twin, the target frequency band of interference is determined and the quantum state characteristics of the target signal are synchronously collected; S32: Quantum entangled signal generation: The quantum signal generator generates entangled photon pairs through spontaneous parametric down-conversion, where one beam of photons retains a local reference and the other is modulated to the target communication frequency band through a tunable filter; Based on the principle of quantum teleportation, the quantum state characteristics of the local reference photon are mapped to the transmitted photon, so that the quantum state of the interference signal and the target communication link are partially entangled; the interference signal adopts pulse modulation mode; S33: Non-intrusive jamming implementation: The directional transmitting antenna focuses the quantum entangled signal onto the target drone; the phase and amplitude of the jamming signal are adjusted in real time through quantum measurement feedback, reducing the quantum bit error rate of the target communication link from 10 -6 Level up to 10 -2 Level; synchronously monitor the adaptive adjustment of the target communication link and dynamically update the interference parameters; S34: Controllable plasma cloud generation and signal attenuation regulation: Plasma cloud parameter planning: Calculate the coverage of the plasma cloud based on the drone size from the holographic target archive; According to the type of signal to be attenuated, the electron density and collision frequency of the plasma are determined so that the attenuation rate of the cloud for the target frequency band is ≥30dB and the attenuation for other frequency bands is ≤5dB. Plasma cloud generation and maintenance: The microwave excitation device emits directional millimeter waves, focusing on the air area around the target drone, and generates plasma by ionizing air molecules; Real-time monitoring of cloud density distribution, automatically increasing microwave excitation power in the corresponding area when the local density falls below a threshold; Selective signal attenuation verification: The spectrum monitoring module synchronously collects the changes in the target drone's navigation signal and image transmission signal strength to verify the attenuation effect; S35: Determine the controllability of the target drone: Communication link response test: Under the synergistic effect of quantum interference and plasma attenuation, a standardized control command test packet is sent to the target drone; the target's response signal is monitored, and signal analysis is used to determine whether its flight control system still maintains the ability to receive commands; Dynamic state assessment: Analyze the target's motion stability in a disturbed environment based on Doppler effect characteristics and virtual twin models. Evaluate the power system state: Use terahertz thermal imaging to monitor motor temperature changes and, combined with speed fluctuations, determine whether the motor is still in a controllable operating state. Quantitative controllability scoring: Evaluates communication responsiveness, motion stability, and power system status; each indicator has a maximum score of 100 points, and a comprehensive score of 70 points or higher is considered controllable, 50-69 points is considered limited controllable, and <50 points is considered completely out of control. For controllable targets, the takeover protocol corresponding to their model is further matched to provide a basis for subsequent takeover control.

6. The method for detecting and countering a drone according to claim 5, characterized in that: It also includes dynamic adjustment of counter-effects and setting of termination mechanisms: Continuously monitor the target drone's status and automatically increase the quantum interference intensity or expand the plasma cloud coverage when the score drops below a preset threshold; If the target is judged to be completely out of control and has a low risk of falling, the countermeasure strength will be gradually reduced, linearly decaying to off within 10 seconds.

7. The method for detecting and countering drones according to claim 1, characterized in that: In step S4, the takeover system based on brain-computer interface mapping is activated. By analyzing the electrical signal characteristics of the UAV flight control system, the specific process of establishing the neuromorphic mapping model of the control instructions is as follows: S41: Brain-computer interface mapping takes over system initialization: Flight control system electrical signal feature analysis: The real-time electrical signal features of the core control unit are extracted by capturing the UAV flight control system communication data during the quantum entangled signal interference phase; Extract characteristic waves from the electrical signal features through wavelet transform, and establish a mapping relationship library between electrical signal features and physical actions as a benchmark for subsequent instruction generation; Take over system hardware activation: load the pre-trained UAV flight control system response model, initialize the communication interface, and activate the directional command transmission array; Simultaneously establish a communication link with the intelligent recovery capsule to obtain the recovery capsule coordinates, hatch status, and a three-dimensional map of the surrounding airspace; S42: Construction of neuromorphic mapping model for control instructions: Modeling the response characteristics of the flight control system: Send 10 sets of calibration commands to the drone, record the feedback characteristics of the electrical signal characteristics, and build a command-response dynamic model using a Bayesian network; Based on the physical parameters of the virtual twin, the model parameters are modified so that the deviation between the virtual simulation command response and the actual measurement is ≤3%; Setting up neuromorphic control command generation rules: Using a spiking neural network to simulate the discharge patterns of biological neurons, traditional control commands are converted into spatiotemporally encoded pulse sequences. Introducing the Hebbian learning rule ensures that the command pulse sequence resonates with the characteristic frequency bands of the drone's electrical signal characteristics. Activate the takeover system based on brain-computer interface mapping, and establish a neuromorphic mapping model of control instructions by analyzing the electrical signal characteristics of the drone flight control system.

8. The method for detecting and countering a drone according to claim 7, characterized in that: In step S4, the specific process of sending control instructions containing the bio-inspired obstacle avoidance algorithm in stages to guide the target to the intelligent recovery capsule is as follows: S43: Embedding the biologically inspired obstacle avoidance algorithm: Obstacle avoidance algorithm simulates biological mechanisms: Drawing on the swarm intelligence mechanism of bee swarm navigation, it identifies obstacles through residual optical sensor data from drones or environmental data from ground monitoring, and combines it with acceleration sensor data to predict collision risks; It simulates the principle of bat echolocation, generates virtual sound wave detection signals, calculates obstacle distance and avoidance angle, and outputs three basic obstacle avoidance actions: emergency pull-up, lateral avoidance, and deceleration. S44: Dynamically plan the recovery path: With the intelligent recovery capsule as the endpoint, the A* algorithm is used to generate an initial three-dimensional path. The path is updated based on real-time positioning data, and local replanning is triggered when the deviation error is greater than 3m. When the distance to the recovery capsule is ≤100m, the spiral descent path mode is automatically switched. S45: Staged control command transmission and status closed loop: The first stage is takeover confirmation: Send low-intensity neuromorphic commands and monitor the response characteristics of the drone's electrical signal signature; Gradually increase the complexity of instructions and preview the execution effects of instructions through virtual twins; If there is no response to the specified command for consecutive times, the command strength will be automatically increased while maintaining the quantum interference at a low intensity; The second stage is path guidance: Send instructions according to the planned path and waypoints. Each waypoint instruction includes three-dimensional coordinates and arrival time limit. Embedded bio-inspired obstacle avoidance algorithm: When an obstacle is detected, it automatically inserts obstacle avoidance instructions and automatically returns to the original path after avoiding it; The third stage is the docking of the recovery capsule: When it is at a specified distance from the recovery capsule, it sends a hovering command to activate the recovery capsule's laser positioning system; Based on the laser positioning data, refined control instructions are sent to gradually bring the drone into the capture range of the recovery capsule; After the drone enters the recovery cabin, it sends a power shutdown command, triggering the recovery cabin's electromagnetic buffer device to generate a gradient magnetic field to achieve contactless deceleration and complete a soft landing.

Images