An unmanned aerial vehicle electromagnetic scattering detection method based on physical field disturbance information fusion

Abstract

The application discloses a kind of unmanned plane electromagnetic scattering detection methods based on physical field disturbance information fusion, belong to unmanned plane detection technical field, the method utilizes the background electromagnetic field formed by existing communication base station in monitoring area, broadcast television, navigation signal or industrial electromagnetic radiation source, collects background electromagnetic field data under unmanned plane target statelessness, and establishes multidimensional background electromagnetic field reference model;In monitoring stage, difference comparison is carried out to the current electromagnetic field data with the reference model, electromagnetic field disturbance tensor is generated, and amplitude disturbance, phase disturbance, spectral disturbance, polarization disturbance, angle of arrival disturbance and rotor micro-doppler feature are extracted from it;Further combined with fusion decision, the existence, position, trajectory and confidence of unmanned plane target are output.The application does not need to actively emit detection signal, can reduce deployment cost and system exposure risk, and improve the accuracy, robustness and anti-false alarm capability of unmanned plane detection in complex environment.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) detection technology, and more specifically, to a method for detecting electromagnetic scattering of UAVs based on the fusion of physical field perturbation information. Background Technology

[0002] Existing anti-drone detection and location technologies generally fall into two categories: passive detection and active detection. Passive detection technologies primarily detect drone targets through passive radar listening, electromagnetic signal acquisition, and visual recognition. Their working principle typically relies on the drone's own radiated or exposed characteristic information, such as 2.4GHz / 5.8GHz band communication signals generated by the drone's communication link, GPS / BeiDou navigation signals, or the imaging characteristics of the drone's body in visible light, infrared, and other optical sensors. This type of technology is applicable when the target has continuous communication, navigation, or obvious optical exposure characteristics. However, it is difficult to obtain effective detection data for "silent drones" that have shut down their communication or navigation links, fly covertly at low altitudes, or have inconspicuous optical characteristics, thus easily creating blind spots in prevention and control.

[0003] Active detection technologies, such as traditional active radar, typically detect and locate targets by actively emitting electromagnetic waves and receiving reflected signals. While these technologies do not entirely rely on the signals emitted by the UAV itself, most existing solutions are based on a single electromagnetic reflection signal for detection, lacking the ability to comprehensively model and correct for multi-physical factors such as meteorological environment, terrain and building environment, and background electromagnetic noise. In application scenarios such as complex cities, airports, and important venues, low-altitude, slow-moving, and small UAV targets have small radar cross-sections, fly at low altitudes and slow speeds, and are easily affected by ground clutter, building reflections, weather changes, and background electromagnetic interference. This leads to problems such as insufficient target recognition stability, weak resistance to environmental interference, and low positioning accuracy in existing active detection technologies.

[0004] Furthermore, existing anti-drone detection and positioning solutions generally lack an integrated system architecture design encompassing "acquisition-processing-positioning-feedback." Related algorithms often remain at a single stage of signal detection or target identification, failing to deeply integrate the physical mechanisms of electromagnetic scattering, multi-environment coupling models, and real-time AI extrapolation capabilities. Especially in silent drone detection scenarios, existing technologies struggle to accurately identify, calculate the location of, and trace the trajectory of drone targets based on the disturbance and scattering characteristics of the target on the background electromagnetic field. This makes it difficult to meet the practical needs of core areas such as airports, military restricted areas, nuclear power plants, and important conference venues for all-weather, high-reliability, and high-precision drone control. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting electromagnetic scattering of unmanned aerial vehicles (UAVs) based on the fusion of physical field disturbance information, so as to break through the dependence of existing technologies on the UAV's own signals or actively transmitted signals, and realize the accurate detection and positioning of silent UAVs.

[0006] According to a first aspect of the present invention, a method for detecting electromagnetic scattering by a UAV based on the fusion of physical field perturbation information is provided, comprising the following steps:

[0007] Acquire background electromagnetic field data of the monitoring area when there are no UAV targets, and establish a background electromagnetic field reference model covering the monitoring area based on the background electromagnetic field data;

[0008] Multiple electromagnetic monitoring points are deployed within the monitoring area to synchronously collect background electromagnetic fields within the monitoring area and obtain real-time electromagnetic field data.

[0009] The real-time electromagnetic field data is compared with the background electromagnetic field benchmark model to extract the electromagnetic scattering disturbance signal generated by the UAV target entering the background electromagnetic field.

[0010] Acquire multiple environmental parameters related to the monitoring area, wherein the multiple environmental parameters include at least one or more of meteorological parameters, electromagnetic environment parameters, and terrain and building parameters;

[0011] The electromagnetic scattering disturbance signal is preprocessed to obtain disturbance feature data, which includes at least one or more of amplitude disturbance features, phase disturbance features, and frequency disturbance features.

[0012] The disturbance feature data, the multiple environmental parameters, and the background electromagnetic field benchmark model are input into the physical constraint-type artificial intelligence inference model, and the UAV target is inferred and reconstructed based on the electromagnetic scattering physical rules and the multi-environment coupling relationship.

[0013] Based on the results of the simulation and reconstruction, the detection results, position coordinates and motion trajectory of the UAV target are output.

[0014] Optionally, the background electromagnetic field data is derived from the existing background electromagnetic radiation field within the monitoring area, which includes one or more of the following: civilian communication base station signal field, broadcast television transmission signal field, navigation signal field, or industrial electromagnetic radiation field.

[0015] Optionally, establishing the background electromagnetic field reference model specifically includes:

[0016] In the absence of drone targets, the background electromagnetic field in the preset frequency band within the monitoring area is collected in terms of spectrum and intensity.

[0017] The collected background electromagnetic field data is processed for equipment noise calibration, background interference removal, and time synchronization.

[0018] The background electromagnetic field reference model is established according to the spatial location, frequency dimension, amplitude dimension, and phase dimension of the monitoring area.

[0019] Optionally, multiple electromagnetic monitoring points are deployed according to a sparse grid topology. The sparse grid topology is determined based on the background electromagnetic field intensity distribution, terrain and building obstruction in the monitoring area, and target positioning accuracy requirements, so as to reduce the number of electromagnetic monitoring points while meeting the monitoring coverage requirements.

[0020] Optionally, before extracting the electromagnetic scattering disturbance signal, the method further includes:

[0021] The real-time electromagnetic field data collected from the multiple electromagnetic monitoring points are synchronized by clock and spatially registered.

[0022] Based on the sparse gridded topology, the real-time electromagnetic field data between different electromagnetic monitoring points are interpolated and reconstructed to obtain continuous electromagnetic field disturbance distribution data within the monitoring area.

[0023] The suspected target disturbance region is determined based on the continuous electromagnetic field disturbance distribution data.

[0024] Optionally, preprocessing the electromagnetic scattering disturbance signal includes:

[0025] The electromagnetic scattering disturbance signal is filtered to remove power frequency interference, white noise, and electromagnetic interference from non-target background.

[0026] The filtered electromagnetic scattering disturbance signal is detrended to eliminate the slow drift of the background electromagnetic field.

[0027] The detrended electromagnetic scattering disturbance signal is converted from the time domain to the frequency domain, and the time domain to frequency domain conversion includes fast Fourier transform and / or wavelet transform;

[0028] The amplitude disturbance features, phase disturbance features, and frequency disturbance features are extracted based on the conversion results.

[0029] Optionally, the multiple environmental parameters include one or more of wind direction, wind speed, air pressure, temperature, humidity, background electric field distribution, background magnetic field distribution, electromagnetic interference intensity, three-dimensional terrain data, and building structure parameters; the physical constraint-based artificial intelligence inference model performs environmental compensation on the disturbance feature data based on the multiple environmental parameters to reduce the impact of meteorological changes, electromagnetic noise, terrain shading, and building reflections on the detection results.

[0030] Optionally, the physically constrained AI inference model performs inferences using an algorithm that combines physical constraints with deep learning, specifically including the following sub-steps:

[0031] The disturbance feature data is input into the signal feature extraction network, and the deep electromagnetic scattering features caused by the UAV target are extracted through convolution operation. The signal feature extraction network adopts a multi-scale residual convolution structure, which includes multiple dilated convolution branches arranged in parallel with increasing dilation rate. Each branch corresponds to the electromagnetic scattering feature extraction at different spatial scales. After adaptive weighting by the channel attention mechanism, the features are fused into multi-scale electromagnetic scattering deep features.

[0032] The deep electromagnetic scattering features are input into a spatiotemporal correlation analysis network to analyze the perturbation correlation relationships of multiple software-defined radio (SDR) monitoring points in the temporal and spatial dimensions, and to construct a multi-node joint feature vector. The spatiotemporal correlation analysis network is a spatiotemporal graph attention network, whose graph structure uses each SDR monitoring node as a vertex and the electromagnetic coupling strength between nodes as the edge weight. The spatiotemporal graph attention network contains a structure of alternating stacked spatial graph attention layers and temporally gated recurrent layers. The spatial graph attention layer learns the dynamic correlation weights between nodes through a multi-head attention mechanism, and the temporally gated recurrent layer captures the temporal evolution of the feature sequences of each node through gated recurrent units. Finally, a multi-node joint feature vector that integrates spatiotemporal correlation information is output.

[0033] Optionally, based on the results of the simulation and reconstruction, the output of the UAV target's detection results, position coordinates, and trajectory includes:

[0034] Based on the degree of matching between the disturbance feature data and the electromagnetic scattering characteristics of the UAV, it is determined whether there is a UAV target in the monitoring area;

[0035] If the presence of a UAV target is determined, the position coordinates of the UAV target are calculated based on one or more of the following: disturbance arrival time, disturbance phase difference, disturbance amplitude distribution, and disturbance frequency offset corresponding to multiple electromagnetic monitoring points.

[0036] The motion trajectory of the UAV target is generated based on its position coordinates at consecutive time points.

[0037] Optionally, after outputting the detection results, position coordinates, and trajectory of the UAV target, the following may also be included:

[0038] The detection results, location coordinates, and movement trajectory are sent to the prevention and control terminal.

[0039] Based on the location coordinates, speed and trajectory of the UAV target, one or more of the following: the acquisition frequency band, sampling frequency, acquisition channel and data upload strategy of the multiple electromagnetic monitoring points are dynamically adjusted:

[0040] Based on the adjusted acquisition strategy, real-time electromagnetic field data will continue to be collected to form a closed-loop detection process that optimizes UAV target detection, localization, trajectory tracking, and monitoring strategies.

[0041] The UAV electromagnetic scattering detection method based on physical field perturbation information fusion disclosed herein has the following technical advantages:

[0042] This invention utilizes the existing background electromagnetic radiation field within the monitoring area as the detection medium, establishes a baseline model of the background electromagnetic field, and extracts electromagnetic scattering disturbance signals by comparing real-time data with the baseline model. It no longer relies on the target's own radiation, but instead utilizes the scattering effect of the UAV's metal structure on the existing electromagnetic field in the environment to induce disturbances for detection. This completely eliminates the dependence on the UAV's own signals, solves the "blind spot" problem in the background technology, enables accurate perception of non-cooperative, zero-radiation targets, and improves the reliability of core area defense.

[0043] This invention introduces multiple environmental parameters and inputs them into a physically constrained artificial intelligence inference model to perform environmental compensation for disturbed signals, further defining the physical constraint rules based on Maxwell's equations. This invention quantifies the attenuation of radio waves by weather and the obstruction and reflection of signals by terrain through a multi-environment physical AI model, and introduces physical equations as regularization constraints into the AI ​​inference, forcing the model output to conform to the laws of electromagnetic physics. This effectively filters out false signals caused by environmental noise and multipath effects, ensuring high stability in detection performance even in complex urban and mountainous environments.

[0044] Electromagnetic monitoring points are deployed using a sparse grid topology, and the sparse data is interpolated and reconstructed based on the topology to generate continuous electromagnetic field disturbance distribution data. By optimizing the monitoring point layout, AI algorithms are used to compensate for the information loss caused by the reduction in physical points. While ensuring monitoring coverage and positioning accuracy, the number of required SDR monitoring devices is reduced, thereby lowering hardware procurement costs, transmission link pressure, and subsequent maintenance costs.

[0045] This invention constructs a deep learning architecture comprising multi-scale residual convolution and a spatiotemporal graph attention network, analyzing the perturbation correlations of multiple SDR monitoring points in the temporal and spatial dimensions. By learning the dynamic electromagnetic coupling weights between nodes through the spatial graph attention layer and capturing the evolutionary patterns of target motion through a temporal gating layer, deep extraction of target features is achieved. It can accurately calculate the position coordinates of the UAV and generate a continuous and smooth motion trajectory, solving the problem of "insufficient stability in target recognition" in the background technology and providing accurate data support for subsequent countermeasures.

[0046] After outputting the results, the acquisition frequency band, sampling frequency and uploading strategy of the electromagnetic monitoring point are dynamically adjusted according to the target's motion state. A feedback link between the detection results and the acquisition strategy is established. The working parameters can be adaptively adjusted according to the target's maneuvering characteristics, forming a complete closed loop of "detection-location-tracking-optimization", which further improves the system's real-time performance and intelligence level.

[0047] Other features and advantages of the invention will become clear from the following detailed description of exemplary embodiments of the invention with reference to the accompanying drawings. Attached Figure Description

[0048] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0049] Figure 1 This is a flowchart illustrating the UAV electromagnetic scattering detection method based on physical field perturbation information fusion provided in an embodiment of the present invention. Detailed Implementation

[0050] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0051] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0052] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0053] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0054] This invention proposes an embodiment of an unmanned aerial vehicle (UAV) electromagnetic scattering detection method based on physical field perturbation information fusion. Specifically, as follows: Figure 1 As shown, it includes the following steps:

[0055] The background electromagnetic field data of the monitoring area is acquired when there are no UAV targets, and a background electromagnetic field benchmark model covering the monitoring area is established based on the background electromagnetic field data. In this embodiment of the invention, the background electromagnetic field data comes from the background electromagnetic radiation field that already exists in the monitoring area, and the background electromagnetic radiation field includes one or more of the following: civilian communication base station signal field, broadcast television transmission signal field, navigation signal field, or industrial electromagnetic radiation field.

[0056] In this embodiment of the invention, establishing the background electromagnetic field reference model specifically includes: acquiring the spectrum and intensity of the background electromagnetic field in a preset frequency band within the monitoring area under no-UAV target conditions; performing equipment noise calibration, background interference removal, and time synchronization processing on the acquired background electromagnetic field data; and establishing the background electromagnetic field reference model according to the spatial location, frequency dimension, amplitude dimension, and phase dimension of the monitoring area.

[0057] Furthermore, in another embodiment of the present invention, the background electromagnetic field reference model is a multidimensional background field tensor model. This multidimensional background field tensor model includes at least spatial location dimension, time dimension, frequency dimension, amplitude dimension, phase dimension, and polarization dimension. The spatial location dimension characterizes the field strength distribution at different receiving nodes or grid points within the monitoring area; the time dimension characterizes the periodic changes in the background field caused by day / night cycles, weather, and communication load variations; and the polarization dimension characterizes the electromagnetic field response in different polarization directions. This reduces false alarms caused by periodic environmental changes and provides a more refined reference for subsequent disturbance detection.

[0058] Multiple electromagnetic monitoring points are deployed within the monitoring area to synchronously collect background electromagnetic fields within the monitoring area, thereby obtaining real-time electromagnetic field data.

[0059] Furthermore, this embodiment of the invention also performs a quality assessment on multiple existing electromagnetic radiation sources within the monitoring area, calculating an availability score for each radiation source. The availability score is determined based on at least two of the following indicators: signal stability, frequency occupancy, coverage angle, signal-to-noise ratio, and historical disturbance sensitivity. During the detection process, radiation sources with availability scores higher than a preset threshold are selected as opportunistic illumination sources. This proactive screening of background sources most suitable for detecting drone scattering disturbances improves stability and anti-interference capabilities in passive detection scenarios.

[0060] Furthermore, corresponding sub-background electromagnetic field benchmark models are established for multiple opportunity illumination sources, and corresponding sub-perturbation features are extracted respectively. If the sub-perturbation features from multiple opportunity illumination sources in the same spatial region meet the consistency condition in time, space or motion trajectory, the confidence of UAV target decision is improved. If only a single opportunity illumination source has isolated perturbation, its decision weight is reduced.

[0061] In this embodiment of the invention, multiple electromagnetic monitoring points are arranged according to a sparse grid topology. The sparse grid topology is determined based on the background electromagnetic field intensity distribution, terrain and building obstruction in the monitoring area, and target positioning accuracy requirements, so as to reduce the number of electromagnetic monitoring points while meeting the monitoring coverage requirements.

[0062] The real-time electromagnetic field data is compared with the background electromagnetic field reference model to extract the electromagnetic scattering disturbance signal generated by the UAV target entering the background electromagnetic field; wherein, the electromagnetic scattering disturbance signal includes at least one of amplitude disturbance, phase disturbance, spectral disturbance, multipath structure change, polarization state change and micro-Doppler disturbance.

[0063] Specifically, the real-time electromagnetic field data is differentially analyzed with the background electromagnetic field reference model to obtain an electromagnetic field perturbation tensor. This tensor includes amplitude perturbation components, phase perturbation components, spectral perturbation components, polarization perturbation components, and angle-of-arrival perturbation components. Time-frequency analysis is performed on the electromagnetic field perturbation tensor to extract micro-Doppler perturbation features caused by the periodic motion of the UAV rotor. Specifically, short-time Fourier transform, wavelet transform, or time-frequency ridge extraction are performed on the spectral perturbation components of the electromagnetic field perturbation tensor to obtain periodic micro-Doppler components. Based on the frequency interval, harmonic quantity, energy distribution, and duration of the periodic micro-Doppler components, it is determined whether the perturbed target exhibits UAV rotor scattering characteristics.

[0064] In another embodiment of the present invention, in addition to the electromagnetic scattering disturbance signal, at least one of the following is further acquired: acoustic field disturbance signal, airflow disturbance signal, magnetic field disturbance signal, or optical / infrared disturbance signal; the electromagnetic scattering disturbance signal is input into the fusion decision model along with other physical field disturbance signals, and the fusion decision model outputs the UAV target confidence level based on the time synchronization relationship, spatial consistency relationship, and target motion continuity relationship, which can solve the problem of high false alarm rate of a single electromagnetic field in complex urban environments.

[0065] In this embodiment of the invention, before extracting the electromagnetic scattering disturbance signal, the method further includes: clock synchronization and spatial registration of the real-time electromagnetic field data collected by the plurality of electromagnetic monitoring points; interpolation and reconstruction of the real-time electromagnetic field data between different electromagnetic monitoring points based on the sparse gridded topology to obtain continuous electromagnetic field disturbance distribution data within the monitoring area; and determining the suspected target disturbance area based on the continuous electromagnetic field disturbance distribution data.

[0066] Multiple environmental parameters related to the monitoring area are acquired. These multiple environmental parameters include at least one or more of meteorological parameters, electromagnetic environment parameters, and terrain and building parameters. In this embodiment of the invention, the multiple environmental parameters include one or more of wind direction, wind speed, air pressure, temperature, humidity, background electric field distribution, background magnetic field distribution, electromagnetic interference intensity, three-dimensional terrain data, and building structure parameters. The physical constraint-based artificial intelligence inference model performs environmental compensation on the disturbance feature data based on the multiple environmental parameters to reduce the impact of meteorological changes, electromagnetic noise, terrain shading, and building reflections on the detection results.

[0067] The electromagnetic scattering disturbance signal is preprocessed to obtain disturbance feature data, which includes at least one or more of amplitude disturbance features, phase disturbance features, and frequency disturbance features. In this embodiment of the invention, the preprocessing of the electromagnetic scattering disturbance signal includes: filtering the electromagnetic scattering disturbance signal to remove power frequency interference, white noise, and non-target background electromagnetic interference; detrending the filtered electromagnetic scattering disturbance signal to eliminate the slow drift of the background electromagnetic field; performing time-domain to frequency-domain transformation on the detrending electromagnetic scattering disturbance signal, which includes fast Fourier transform and / or wavelet transform; and extracting the amplitude disturbance features, phase disturbance features, and frequency disturbance features based on the transformation result.

[0068] The disturbance feature data, the multiple environmental parameters, and the background electromagnetic field benchmark model are input into a physical constraint-based artificial intelligence inference model, and the UAV target is inferred and reconstructed based on the physical rules of electromagnetic scattering and the coupling relationship of multiple environments.

[0069] In this embodiment of the invention, the physically constrained artificial intelligence deduction model performs deductions using an algorithm that combines physical constraints with deep learning, specifically including the following sub-steps:

[0070] The disturbance feature data is input into the signal feature extraction network, and the deep electromagnetic scattering features caused by the UAV target are extracted through convolution operation. The signal feature extraction network adopts a multi-scale residual convolution structure, which includes multiple dilated convolution branches arranged in parallel with increasing dilation rate. Each branch corresponds to the electromagnetic scattering feature extraction at different spatial scales. After adaptive weighting by the channel attention mechanism, the features are fused into multi-scale electromagnetic scattering deep features.

[0071] The deep electromagnetic scattering features are input into a spatiotemporal correlation analysis network to analyze the perturbation correlation relationships of multiple software-defined radio (SDR) monitoring points in the temporal and spatial dimensions, and to construct a multi-node joint feature vector. The spatiotemporal correlation analysis network is a spatiotemporal graph attention network, whose graph structure uses each SDR monitoring node as a vertex and the electromagnetic coupling strength between nodes as the edge weight. The spatiotemporal graph attention network contains a structure of alternating stacked spatial graph attention layers and temporally gated recurrent layers. The spatial graph attention layer learns the dynamic correlation weights between nodes through a multi-head attention mechanism, and the temporally gated recurrent layer captures the temporal evolution of the feature sequences of each node through gated recurrent units. Finally, a multi-node joint feature vector that integrates spatiotemporal correlation information is output.

[0072] The detection results, position coordinates, and trajectory of the UAV target are output based on the results of the deduction and reconstruction. In this embodiment of the invention, outputting the detection results, position coordinates, and trajectory of the UAV target based on the results of the deduction and reconstruction includes: determining whether a UAV target exists in the monitoring area based on the degree of matching between the disturbance feature data and the electromagnetic scattering features of the UAV; if a UAV target is determined to exist, calculating the position coordinates of the UAV target based on one or more of the disturbance arrival time, disturbance phase difference, disturbance amplitude distribution, and disturbance frequency offset corresponding to multiple electromagnetic monitoring points; and generating the trajectory of the UAV target based on the position coordinates at consecutive time points.

[0073] In this embodiment of the invention, after outputting the detection results, position coordinates, and movement trajectory of the UAV target, the method further includes: sending the detection results, position coordinates, and movement trajectory to the control terminal; dynamically adjusting one or more of the acquisition frequency band, sampling frequency, acquisition channel, and data upload strategy of the multiple electromagnetic monitoring points according to the position coordinates, movement speed, and movement trajectory of the UAV target; and continuing to acquire real-time electromagnetic field data based on the adjusted acquisition strategy to form a closed-loop detection process of UAV target detection, positioning, trajectory tracking, and monitoring strategy optimization.

[0074] In one specific embodiment, this invention is applied to an urban low-altitude monitoring scenario. Four electromagnetic field receiving nodes are deployed at the perimeter and rooftops of a certain park. Each node can receive the existing background electromagnetic radiation field formed by surrounding civilian communication base stations, broadcast television transmission sources, and BeiDou / GNSS navigation signals. This invention first selects multiple time windows during both nighttime and daytime to confirm the absence of UAV targets, collecting the spectrum, amplitude, phase, polarization, and angle of arrival of the background electromagnetic field within the monitoring area ranging from 700MHz to 6GHz. Combined with temperature, humidity, weather conditions, and communication base station load information, a multi-dimensional background electromagnetic field benchmark model covering the monitoring area is established. Upon entering real-time monitoring, when a quadcopter UAV enters from the east side of the park at a height of approximately 30 meters, its body and rotor scatter and block signals from different opportunity illumination sources, causing multiple receiving nodes to detect local amplitude disturbances, phase disturbances, and spectral disturbances within similar time windows. This invention differs the current electromagnetic field data from the background benchmark model to form an electromagnetic field disturbance tensor, and extracts micro-Doppler harmonic features related to the periodic rotation of the rotor from the spectral disturbance component. Subsequently, consistency discrimination is performed on the sub-disturbance characteristics from the communication base station signal field and the navigation signal field, confirming that multiple opportunistic illumination sources within the same spatial area exhibit disturbances with temporal synchronization relationships. Simultaneously, the UAV's position is estimated using the disturbance arrival time difference, phase difference, and intensity difference between the four receiving nodes, and the trajectories of disturbance centers within multiple consecutive time windows are correlated to obtain a continuous flight trajectory that satisfies the UAV's speed and acceleration constraints. If only short-term communication load changes or single-node field strength fluctuations caused by vehicle passage occur in the same area, they lack rotor micro-Doppler characteristics and cannot form spatial and temporal consistency among multiple opportunistic illumination sources and multiple receiving nodes; therefore, they are judged as environmental interference and eliminated by the system. This embodiment demonstrates that the present invention can achieve passive UAV detection using existing environmental electromagnetic radiation without actively emitting detection signals. Through multi-dimensional background modeling, disturbance tensor analysis, micro-Doppler identification, multi-source consistency discrimination, and trajectory continuity constraints, it effectively improves the accuracy and anti-false alarm capability of UAV detection in complex urban electromagnetic environments.

[0075] This invention proposes a UAV electromagnetic scattering detection method based on physical field perturbation information fusion. It collects background electromagnetic field data from existing radiation sources such as communication base stations, broadcast television, navigation signals, and industrial electromagnetic radiation within a monitored area when no UAV target is present. This data establishes a multi-dimensional background electromagnetic field benchmark model, including dimensions such as spatial location, time, frequency, amplitude, phase, and polarization. During the monitoring phase, real-time electromagnetic field data is collected and compared differentially with the benchmark model to generate an electromagnetic field perturbation tensor containing amplitude perturbation, phase perturbation, spectral perturbation, polarization perturbation, angle-of-arrival perturbation, and rotor micro-Doppler perturbation. Furthermore, it combines multi-opportunity illumination source consistency discrimination, multi-receiver node collaborative positioning, adaptive background updating within a reliable targetless window, trajectory continuity false alarm suppression, and fusion judgment of other physical field perturbation information such as sound field, airflow field, magnetic field, optical, or infrared fields to output the existence, spatial location, trajectory, and confidence level of the UAV target. Through the above scheme, the present invention can perform passive detection using the existing electromagnetic radiation field in the environment without actively emitting detection signals, reducing the system exposure risk and deployment cost; improve the benchmark stability in complex environments through multi-dimensional background modeling and adaptive updates; enhance the ability of UAVs to distinguish from ordinary environmental interference, vehicles, personnel or bird targets through perturbation tensors and rotor micro-Doppler features; and significantly reduce the false alarm rate through multi-source consistency, multi-physics field fusion and trajectory continuity constraints, thereby improving the accuracy, robustness and long-term deployment adaptability of UAV detection.

[0076] The above description of the structure, features, and effects of the present invention is based on the embodiments shown in the figures. However, the above are only preferred embodiments of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched by those skilled in the art to form a variety of equivalent solutions without departing from or changing the design concept and technical effects of the present invention. Therefore, the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments, that do not exceed the spirit covered by the specification and figures, should be within the protection scope of the present invention.

Claims

1. A method for detecting electromagnetic scattering by unmanned aerial vehicles (UAVs) based on the fusion of physical field perturbation information, characterized in that, Includes the following steps: Acquire background electromagnetic field data of the monitoring area when there are no UAV targets, and establish a background electromagnetic field reference model covering the monitoring area based on the background electromagnetic field data; Multiple electromagnetic monitoring points are deployed within the monitoring area to synchronously collect background electromagnetic field data within the monitoring area, thereby obtaining real-time electromagnetic field data. The real-time electromagnetic field data is compared with the background electromagnetic field benchmark model to extract the electromagnetic scattering disturbance signal generated by the UAV target entering the background electromagnetic field. Acquire multiple environmental parameters related to the monitoring area, wherein the multiple environmental parameters include at least one or more of meteorological parameters, electromagnetic environment parameters, and terrain and building parameters; The electromagnetic scattering disturbance signal is preprocessed to obtain disturbance feature data, which includes at least one or more of amplitude disturbance features, phase disturbance features, and frequency disturbance features. The disturbance feature data, the multiple environmental parameters, and the background electromagnetic field benchmark model are input into the physical constraint-type artificial intelligence inference model, and the UAV target is inferred and reconstructed based on the electromagnetic scattering physical rules and the multi-environment coupling relationship. Based on the results of the simulation and reconstruction, the detection results, position coordinates and motion trajectory of the UAV target are output.

2. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, The background electromagnetic field data comes from the existing background electromagnetic radiation field within the monitoring area, which includes one or more of the following: civilian communication base station signal field, radio and television transmission signal field, navigation signal field, or industrial electromagnetic radiation field.

3. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, Establishing the background electromagnetic field reference model specifically includes: In the absence of drone targets, the background electromagnetic field in the preset frequency band within the monitoring area is collected in terms of spectrum and intensity. The collected background electromagnetic field data is processed for equipment noise calibration, background interference removal, and time synchronization. The background electromagnetic field reference model is established according to the spatial location, frequency dimension, amplitude dimension, and phase dimension of the monitoring area.

4. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, Multiple electromagnetic monitoring points are deployed according to a sparse grid topology. The sparse grid topology is determined based on the background electromagnetic field intensity distribution, terrain and building obstruction in the monitoring area, and target positioning accuracy requirements, so as to reduce the number of electromagnetic monitoring points while meeting the monitoring coverage requirements.

5. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 4, characterized in that, Before extracting the electromagnetic scattering disturbance signal, the process also includes: The real-time electromagnetic field data collected from the multiple electromagnetic monitoring points are synchronized by clock and spatially registered. Based on the sparse gridded topology, the real-time electromagnetic field data between different electromagnetic monitoring points are interpolated and reconstructed to obtain continuous electromagnetic field disturbance distribution data within the monitoring area. The suspected target disturbance region is determined based on the continuous electromagnetic field disturbance distribution data.

6. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, Preprocessing the electromagnetic scattering disturbance signal includes: The electromagnetic scattering disturbance signal is filtered to remove power frequency interference, white noise, and electromagnetic interference from non-target background. The filtered electromagnetic scattering disturbance signal is detrended to eliminate the slow drift of the background electromagnetic field. The detrended electromagnetic scattering disturbance signal is converted from the time domain to the frequency domain, and the time domain to frequency domain conversion includes fast Fourier transform and / or wavelet transform; The amplitude disturbance features, phase disturbance features, and frequency disturbance features are extracted based on the conversion results.

7. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, The multiple environmental parameters include one or more of the following: wind direction, wind speed, air pressure, temperature, humidity, background electric field distribution, background magnetic field distribution, electromagnetic interference intensity, three-dimensional terrain data, and building structure parameters; the physical constraint-based artificial intelligence inference model performs environmental compensation on the disturbance feature data based on the multiple environmental parameters to reduce the impact of meteorological changes, electromagnetic noise, terrain occlusion, and building reflections on the detection results.

8. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, The physical constraint-based artificial intelligence deduction model performs deductions using an algorithm that combines physical constraints with deep learning, specifically including the following sub-steps: The disturbance feature data is input into the signal feature extraction network, and the deep electromagnetic scattering features caused by the UAV target are extracted through convolution operation. The signal feature extraction network adopts a multi-scale residual convolution structure, which includes multiple dilated convolution branches arranged in parallel with increasing dilation rate. Each branch corresponds to the electromagnetic scattering feature extraction at different spatial scales. After adaptive weighting by the channel attention mechanism, the features are fused into multi-scale electromagnetic scattering deep features. The deep electromagnetic scattering features are input into a spatiotemporal correlation analysis network to analyze the perturbation correlation relationships of multiple software-defined radio monitoring points in the temporal and spatial dimensions, and to construct a multi-node joint feature vector. The spatiotemporal correlation analysis network is a spatiotemporal graph attention network, whose graph structure uses each software-defined radio monitoring node as a vertex and the electromagnetic coupling strength between nodes as the edge weight. The spatiotemporal graph attention network contains a structure of alternating stacked spatial graph attention layers and temporally gated recurrent layers. The spatial graph attention layer learns the dynamic correlation weights between nodes through a multi-head attention mechanism, and the temporally gated recurrent layer captures the temporal evolution of the feature sequences of each node through gated recurrent units. Finally, a multi-node joint feature vector integrating spatiotemporal correlation information is output.

9. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to claim 1, characterized in that, Based on the results of the simulation and reconstruction, the output includes the detection results, position coordinates, and trajectory of the UAV target, specifically including: Based on the degree of matching between the disturbance feature data and the electromagnetic scattering characteristics of the UAV, it is determined whether there is a UAV target in the monitoring area; If the presence of a UAV target is determined, the position coordinates of the UAV target are calculated based on one or more of the following: disturbance arrival time, disturbance phase difference, disturbance amplitude distribution, and disturbance frequency offset corresponding to multiple electromagnetic monitoring points. The motion trajectory of the UAV target is generated based on its position coordinates at consecutive time points.

10. The UAV electromagnetic scattering detection method based on physical field perturbation information fusion according to any one of claims 1 to 9, characterized in that, After outputting the detection results, location coordinates, and trajectory of the UAV target, it also includes: The detection results, location coordinates, and movement trajectory are sent to the prevention and control terminal. Based on the location coordinates, speed and trajectory of the UAV target, one or more of the following: the acquisition frequency band, sampling frequency, acquisition channel and data upload strategy of the multiple electromagnetic monitoring points are dynamically adjusted: Based on the adjusted acquisition strategy, real-time electromagnetic field data will continue to be collected to form a closed-loop detection process that optimizes the UAV target detection, localization, trajectory tracking, and monitoring strategies.

Images