A method and system for quickly identifying specific drone models based on radio frequency fingerprint library

Abstract

The present application provides a method and system for quickly identifying specific drone models based on a radio frequency fingerprint library. The present application first receives the radio frequency signal generated by the target drone during flight, and performs dynamic noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal. The noise-reduced radio frequency signal is then separated from the combined data containing the signal intensity fluctuation pattern, phase jump frequency, and modulation distortion characteristics, and the combined data is matched layer by layer with the pre-built radio frequency fingerprint library to generate a matching result. The model-related parameters of the target drone are then extracted from the matching result. Finally, an identification decision containing a real-time environmental adaptation factor is generated based on the model-related parameters, and the target drone model result corresponding to the identification decision is output. The technical solution provided by the present application not only greatly improves the discrimination of subtle signal differences, but also achieves the stability and accuracy of drone model identification in complex electromagnetic interference and multi-obstacle reflection scenarios.

Description

Technical Field

[0001] The present application relates to the field of drone technology, and in particular to a method and system for quickly identifying specific drone models based on a radio frequency fingerprint library. Background Art

[0002] With the widespread application of drones in security monitoring, logistics, and other fields, there is an urgent need to quickly and accurately identify specific drone models in open environments with complex electromagnetic interference and multiple obstacle reflections, such as urban buildings and industrial plants. In such scenarios, drone RF signals are easily distorted by environmental interference, and the signals of different drone models vary slightly. Signal processing and model matching must be completed within millisecond response times to meet real-time monitoring and countermeasures.

[0003] Currently, a matching and identification solution based on a static RF fingerprint library is primarily used to meet these requirements. This solution pre-collects RF signals from different drone models in a fixed experimental environment, extracts signal strength and spectrum characteristics, and generates a baseline fingerprint library. During actual identification, the target drone's real-time signal is compared against the fingerprint library using a threshold value, and the model with the highest match is selected as the identification result.

[0004] The above solution has significant limitations in dynamic open scenarios: first, the static fingerprint library cannot adapt to the real-time changes in electromagnetic interference intensity and obstacle reflection paths in complex environments, resulting in the offset between signal characteristics and baseline fingerprints exceeding the preset threshold, causing misjudgment; second, the fixed threshold matching mechanism is difficult to cope with dynamic fluctuations in the degree of signal distortion, and multiple threshold adjustments or recalibration of the fingerprint library are required, seriously affecting the real-time performance of recognition; third, a coordinated mechanism between environmental parameters (such as interference source distribution and obstacle location) and RF fingerprint correction has not been established, resulting in huge differences in recognition efficiency in different scenarios. Summary of the Invention

[0005] The present application provides a method and system for rapid identification of specific drone models based on a radio frequency fingerprint library, which is used to solve the problems in the prior art, such as the inability of static radio frequency fingerprint libraries to adapt to signal offsets caused by electromagnetic interference and obstacle reflections in dynamic open scenes, the lack of real-time performance of fixed threshold matching mechanisms due to environmental fluctuations, and the unstable recognition efficiency caused by the lack of coordinated association between environmental parameters and radio frequency fingerprint correction.

[0006] In a first aspect, the present application provides a method for quickly identifying a specific drone model based on a radio frequency fingerprint library, comprising:

[0007] Receive the radio frequency signal generated by the target UAV during flight, and perform dynamic noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal;

[0008] Separate the combined data containing signal strength fluctuation pattern, phase hopping frequency and modulation distortion characteristics from the de-noised RF signal, and match the combined data with the pre-built RF fingerprint library layer by layer to generate a matching result;

[0009] Extracting model-related parameters of the target UAV from the matching results, the model-related parameters including the signal propagation delay difference corresponding to the reflection path of a specific obstacle and the frequency domain distortion correlation corresponding to the distribution of the electromagnetic interference source;

[0010] An identification decision including a real-time environment adaptation factor is generated based on the model association parameters, and a target UAV model result corresponding to the identification decision is output.

[0011] Optionally, the method is characterized in that receiving a radio frequency signal generated by a target UAV during flight and performing dynamic noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal includes:

[0012] The original radio frequency signal of the target UAV is captured in real time by a multi-channel signal acquisition device deployed in the monitoring area. The original radio frequency signal contains noise components caused by electromagnetic interference sources and obstacle reflections in the current scene;

[0013] Monitoring environmental parameters in the original radio frequency signal, wherein the environmental parameters include real-time interference intensity distribution and obstacle density distribution;

[0014] Selecting a signal filtering intensity level according to the environmental parameters; if the real-time interference intensity is higher than a set range and the obstacle density is higher than a set threshold, performing layer-by-layer noise suppression on the original radio frequency signal using multi-stage series filtering; and if the real-time interference intensity is lower than the set range, performing noise suppression on the original radio frequency signal using single-stage filtering to obtain a filtered radio frequency signal;

[0015] The filtered RF signal is processed for signal integrity restoration to generate a noise-reduced RF signal.

[0016] Optionally, combined data including signal strength fluctuation patterns, phase hopping frequencies, and modulation distortion features are separated from the noise-reduced RF signal, and the combined data is matched layer by layer with a pre-built RF fingerprint library to generate a matching result, including:

[0017] Dividing the denoised radio frequency signal into multiple signal segments according to a preset time window, and extracting the fluctuation pattern of signal strength as it changes with flight altitude, the frequency distribution of phase jump intervals, and the position characteristics of modulation waveform distortion from each signal segment;

[0018] The fluctuation pattern, frequency distribution and location characteristics are integrated into combined data according to a preset weight ratio;

[0019] Calculate similarity between the combined data and the first-layer signal strength fluctuation pattern benchmark set stored in the radio frequency fingerprint library, and select a candidate model set whose similarity is higher than a first threshold;

[0020] The frequency distribution and position features in the combined data are jointly matched with the second-layer phase jump frequency reference set and modulation distortion reference set corresponding to the candidate model set to generate a matching result including the candidate model matching degree ranking.

[0021] Optionally, model-related parameters of the target UAV are extracted from the matching results, where the model-related parameters include a signal propagation delay difference corresponding to a specific obstacle reflection path and a frequency domain distortion correlation corresponding to an electromagnetic interference source distribution, including:

[0022] Sorting the candidate models in the matching results according to their matching scores, extracting the radio frequency fingerprint reference data corresponding to the candidate model with the highest ranking;

[0023] Obtaining a signal propagation delay difference parameter associated with obstacle distribution in a current scene from the radio frequency fingerprint reference data;

[0024] Acquire a frequency domain distortion correlation parameter associated with the electromagnetic interference source distribution from the radio frequency fingerprint reference data;

[0025] The signal propagation delay difference parameter and the frequency domain distortion correlation parameter are integrated into the model correlation parameter of the target UAV.

[0026] Optionally, generating an identification decision including a real-time environment adaptation factor based on the model association parameter, and outputting a target UAV model result corresponding to the identification decision, including:

[0027] Obtain real-time tracking data of the dynamic changes in the position of obstacles in the current scene and real-time fluctuation data of the intensity of electromagnetic interference sources;

[0028] Adjusting the weight coefficient of the signal propagation delay difference parameter according to the real-time tracking data, and adjusting the weight coefficient of the frequency domain distortion correlation parameter according to the real-time fluctuation data;

[0029] Input the adjusted signal propagation delay difference parameter and the adjusted frequency domain distortion correlation parameter into the environmental adaptation factor calculation model to generate a real-time environmental adaptation factor for correcting the matching degree of candidate models in the radio frequency fingerprint library;

[0030] Dynamically compensating the matching degree of candidate models in the matching results according to the real-time environment adaptation factor to generate an identification decision including a final matching model;

[0031] The model identifier of the final matching model in the identification decision is checked for consistency with the model feature description pre-stored in the radio frequency fingerprint library. When the consistency verification passes, the final matching model is output as the target drone model result.

[0032] Optionally, adjusting a weight coefficient of a signal propagation delay difference parameter according to the real-time tracking data, and adjusting a weight coefficient of a frequency domain distortion correlation parameter according to the real-time fluctuation data, includes:

[0033] Dynamically adjusting a weight coefficient of a signal propagation delay difference parameter based on a dynamic rate of change of an obstacle position in the real-time tracking data; when the dynamic rate exceeds a preset rate threshold, linearly enhancing the weight coefficient based on a ratio of the dynamic rate to the preset rate threshold to generate an adjusted delay weight coefficient;

[0034] According to the fluctuation amplitude of the electromagnetic interference source intensity in the real-time fluctuation data, the weight coefficient of the frequency domain distortion correlation parameter is dynamically adjusted. When the fluctuation amplitude exceeds the preset amplitude threshold, the weight coefficient is increased in segments according to the difference between the fluctuation amplitude and the amplitude threshold to generate an adjusted frequency domain weight coefficient.

[0035] Optionally, dynamically compensating the matching degree of the candidate models in the matching result according to the real-time environment adaptation factor to generate an identification decision including a final matching model includes:

[0036] Decomposing the real-time environment adaptation factor into a time delay correction factor and a frequency domain correction factor;

[0037] Extracting an initial matching degree of the candidate model set from the matching results, and multiplying the delay correction factor by the signal propagation delay difference parameter corresponding to each candidate model in the initial matching degree item by item to generate a matching degree after delay compensation;

[0038] Multiplying the frequency domain correction factor by the frequency domain distortion correlation parameter corresponding to each candidate model in the matching degree after delay compensation item by item to generate the matching degree after frequency domain compensation;

[0039] The matching degree after delay compensation and the matching degree after frequency domain compensation are combined according to a preset ratio to generate a comprehensive compensation matching degree for each candidate model;

[0040] The candidate model set is sorted in descending order according to the comprehensive compensation matching degree, the candidate model with the highest ranking is selected as the final matching model, and an identification decision including the final matching model is generated.

[0041] In a second aspect, the present application provides a system for quickly identifying specific drone models based on a radio frequency fingerprint library, including:

[0042] A receiving module is used to receive the radio frequency signal generated by the target UAV during flight, and dynamically reduce the noise of the radio frequency signal to generate a noise-reduced radio frequency signal;

[0043] A matching module is used to separate the combined data containing the signal strength fluctuation pattern, phase hop frequency and modulation distortion characteristics from the noise-reduced RF signal, and match the combined data with the pre-built RF fingerprint library layer by layer to generate a matching result;

[0044] An extraction module is configured to extract model-related parameters of the target UAV from the matching results, wherein the model-related parameters include a signal propagation delay difference corresponding to a specific obstacle reflection path and a frequency domain distortion correlation corresponding to an electromagnetic interference source distribution;

[0045] The output module is used to generate an identification decision including a real-time environment adaptation factor based on the model association parameters, and output a target UAV model result corresponding to the identification decision.

[0046] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for quickly identifying a specific drone model based on a radio frequency fingerprint library as described in the first aspect above.

[0047] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements a method for quickly identifying a specific drone model based on a radio frequency fingerprint library as described in the first aspect.

[0048] The embodiments of this application effectively address the core challenge of distorted drone RF signal characteristics in complex electromagnetic interference and multi-obstacle reflection scenarios through the collaborative design of dynamic noise reduction processing and layer-by-layer matching of combined data. Specifically, dynamic noise reduction processing adaptively adjusts the filter strength based on real-time environmental parameters (interference intensity, obstacle density), significantly suppressing the impact of noise components on key features such as signal intensity fluctuation patterns and phase jump frequency. At the same time, by separating and fusing multi-dimensional features to generate combined data, combined with the layer-by-layer matching mechanism of the RF fingerprint library (progressive screening of intensity fluctuations → phase jumps → modulation distortion), the ability to distinguish subtle signal differences is greatly improved, thereby achieving high-precision, low-latency model recognition in open scenarios.

[0049] Furthermore, by decomposing the environmental adaptation factor into a delay correction factor (associated with the obstacle reflection path) and a frequency domain correction factor (associated with the electromagnetic interference distribution), two compensation corrections are applied to the candidate model's delay difference parameters and frequency domain distortion parameters, respectively. These are then combined with a preset ratio to generate a comprehensive matching degree, ensuring robust identification decisions under environmental fluctuations. This mechanism not only avoids the real-time limitations of traditional static threshold matching but also enables continuous optimization of scenario adaptation through incremental updates to the RF fingerprint library, ultimately improving both the stability and accuracy of drone model identification in complex and open environments.

[0050] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0052] Figure 1 A flowchart of a method for quickly identifying a specific drone model based on a radio frequency fingerprint library provided by the present application is shown;

[0053] Figure 2 The following is a schematic diagram showing the structure of a system for rapid identification of specific drone models based on a radio frequency fingerprint library provided by the present application;

[0054] Figure 3 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0055] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0056] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0057] In open scenarios with complex electromagnetic interference and multiple obstacle reflections, existing matching and identification solutions based on static RF fingerprint libraries face severe challenges: static fingerprint libraries rely on benchmark data in fixed experimental environments and are difficult to adapt to the real-time changing electromagnetic interference intensity and obstacle reflection paths in dynamic open scenarios, resulting in the offset between the target drone signal characteristics and the benchmark fingerprint exceeding the preset threshold, causing misjudgment; at the same time, the fixed threshold matching mechanism cannot respond to dynamic fluctuations in the degree of signal distortion, and requires repeated threshold calibration or updating of the fingerprint library, which seriously restricts the real-time performance of identification; in addition, the existing solutions lack a coordinated mechanism between environmental parameters (such as interference source distribution, obstacle location) and RF fingerprint correction, resulting in significant fluctuations in recognition efficiency in different scenarios, making it difficult to meet high-precision supervision requirements.

[0058] To address these issues, this application proposes a method for rapid identification of specific drone models based on an RF fingerprint library. Its core technology lies in a closed-loop process: dynamic noise reduction processing, layered matching of combined data, and dynamic compensation of environmental adaptation factors. This overcomes the inherent limitations of static fingerprint libraries and fixed threshold mechanisms. Specifically, dynamic noise reduction is driven by real-time environmental parameters (interference intensity, obstacle density), adaptively suppressing the impact of noise components on key features such as signal intensity fluctuation patterns and phase jump frequency. Furthermore, the de-noised RF signal is separated into multi-dimensional combined data. Candidate models are screened through layer-by-layer matching of the RF fingerprint library (intensity fluctuation → phase jump → modulation distortion), and the matching results are dynamically compensated using environmental adaptation factors. This solution effectively eliminates signal feature offsets through real-time coordination of environmental parameters and the RF fingerprint library (for example, adjusting delay difference weights based on the dynamic rate of obstacles and correcting frequency domain distortion correlation based on the amplitude of interference fluctuations). This significantly improves the real-time, accuracy, and environmental robustness of identification in dynamic, open scenarios, providing reliable technical support for drone monitoring and countermeasures.

[0059] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0060] Figure 1 The present invention provides a flowchart of a method for quickly identifying a specific drone model based on a radio frequency fingerprint library. Figure 1 As shown, the method includes:

[0061] Step 101: receiving a radio frequency signal generated by a target UAV during flight, and performing dynamic noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal;

[0062] In this step, dynamic noise reduction processing refers to the process of adaptively adjusting the noise suppression strategy according to real-time environmental parameters (such as interference intensity distribution and obstacle density distribution); among them, the real-time interference intensity distribution represents the electromagnetic interference energy distribution of different frequency bands in the monitoring area, and the obstacle density distribution represents the density of obstacle reflection paths in a unit space; multi-stage cascade filtering refers to the noise suppression operations of different frequency bands and amplitudes performed on the radio frequency signal in sequence based on environmental parameters, and single-stage filtering refers to the noise suppression of a single frequency band or amplitude; signal integrity recovery processing refers to the repair of key signal features related to the flight status of the drone that may be lost during the filtering process through signal segment reorganization and feature retention mechanism.

[0063] In this embodiment, the target drone's raw RF signal is first captured synchronously using multi-channel signal acquisition equipment deployed within the monitoring area. These equipment covers different spatial locations to capture multipath reflection signals. Subsequently, the energy distribution of the raw signal across different frequency bands (e.g., interference strength in the 2.4GHz / 5.8GHz bands) is analyzed in real time, and obstacle density parameters are simultaneously acquired using a lidar or infrared sensor. Based on the combination of interference intensity and obstacle density (e.g., high interference and high density scenarios), multi-stage cascade filtering is activated: the first stage removes broadband noise outside the primary frequency band, the second stage performs amplitude threshold truncation on impulse noise, and the third stage performs frequency domain notching on the remaining noise. If the interference intensity or obstacle density falls below the threshold, only the single-stage primary frequency band bandpass filtering is performed. Finally, the filtered signal undergoes time-domain waveform continuity testing to identify signal abrupt or phase jump breakpoints caused by filtering. Interpolation and repair are performed based on adjacent segment signal characteristics (e.g., amplitude slope and phase continuity), retaining segments of signal strength abrupt and phase jumps related to the drone's flight altitude and attitude, generating a de-noised RF signal.

[0064] For example, in an urban building complex monitoring scenario, multi-channel signal acquisition equipment deployed on a building rooftop captured the target drone's raw RF signal and detected strong interference from WiFi routers in the 2.4GHz band (the interference intensity distribution showed an energy peak exceeding a threshold). Simultaneously, the LiDAR feedback indicated a dense obstacle density distribution (density values ​​above a threshold) from high-rise building reflection paths. Based on this, the system activated multi-stage cascade filtering: the first stage removed spurious signals outside the 2.4GHz band, the second stage amplitude-clipped impulse noise caused by building reflections, and the third stage frequency-domain notched any remaining narrowband interference. After waveform continuity testing, the filtered signal identified two phase transition breakpoints. Linear interpolation was performed based on the phase slopes of the preceding and following segments to repair the signal. This ultimately generated a noise-reduced signal that retained the altitude-related signal intensity fluctuations and the complete phase transition characteristics for subsequent combined data separation and matching.

[0065] Step 102: Separate the combined data including the signal strength fluctuation pattern, phase hopping frequency, and modulation distortion characteristics from the noise-reduced RF signal, and perform layer-by-layer matching of the combined data with a pre-built RF fingerprint library to generate a matching result.

[0066] In this step, the signal strength fluctuation pattern refers to the temporal fluctuation pattern of the RF signal strength as the flight altitude and posture of the drone change; the phase jump frequency refers to the number of discontinuous jumps of the signal phase in unit time and the interval distribution characteristics; the modulation distortion feature refers to the morphological difference in the distortion position and distortion interval of the RF signal modulation waveform relative to the standard modulation method; the combined data is a multidimensional data set formed by the fusion of the above three types of features according to preset weights; the RF fingerprint library stores the baseline feature data of different types of drones in typical scenarios; layer-by-layer matching refers to the hierarchical screening of candidate models in the order of signal strength → phase jump → modulation distortion.

[0067] In this embodiment, the denoised RF signal is first divided into continuous signal segments according to a preset time window (such as 50ms), ensuring that each signal segment contains a complete flight state cycle (such as climbing, hovering). For each signal segment, the signal strength fluctuation pattern is extracted: the signal strength timing curve is aligned through a dynamic time warping algorithm, and its fluctuation similarity with the altitude change reference curve is calculated; the phase jump frequency is extracted: the number of phase jumps in the signal segment is counted, and the frequency distribution histogram of the jump interval is calculated through a sliding window; the modulation distortion feature is extracted: the modulation waveform of the signal segment is compared with the standard waveform point by point, the distortion position is marked, and the proportion of the distortion interval length is calculated. Subsequently, the fluctuation pattern, frequency distribution, and distortion features are fused into a combined data unit according to a preset weight (such as 5:3:2), and the combined data units of all signal segments are spliced ​​into a complete combined data. Next, the combined data is input into the RF fingerprint library. In the first-level matching, the similarity calculation method based on dynamic time warping is used to screen the candidate model set whose signal strength fluctuation pattern similarity is higher than the threshold (such as 80%). In the second-level matching, the overlapping area of ​​the phase hop frequency distribution and the weighted sum of the offset of the modulation distortion position are calculated to generate the comprehensive matching degree of the candidate model, and the matching results are generated in descending order.

[0068] For example, in an urban building complex monitoring scenario, the noise-reduced RF signal generated in step 101 is segmented into 50ms time windows. One signal segment corresponds to the drone's flight phase through a gap between high-rise buildings. When extracting the signal strength fluctuation pattern, the intensity curve exhibits periodic attenuation due to building obstruction. After time alignment with the baseline curve of model C in the fingerprint library, the peak-to-valley difference is less than 8%. Phase jump frequency statistics show 12 jumps per 50ms (due to increased phase mutations caused by multipath reflections), and the frequency distribution histogram overlaps 78% with the baseline histogram of model C. Modulation distortion feature analysis reveals that waveform distortion is concentrated on the falling edge of the modulation (caused by building reflection path delay), accounting for 15% of the distortion interval. After the combined data was fused according to the preset weights (signal strength fluctuation 5: phase jump 3: modulation distortion 2), the first-level matching screened models C and D as candidate sets. In the second-level matching, the weighted score of model C's phase jump overlap area (78%) and distortion offset (0.08) was 0.82, and model D scored 0.65, ultimately generating a result that model C had the highest match.

[0069] Step 103: extracting the model-related parameters of the target UAV from the matching results, wherein the model-related parameters include the signal propagation delay difference corresponding to the reflection path of a specific obstacle and the frequency domain distortion correlation corresponding to the distribution of the electromagnetic interference source;

[0070] In this step, the signal propagation delay difference refers to the deviation between the signal propagation time delay of the target UAV in the current obstacle distribution scenario and the benchmark delay of the reference model in the RF fingerprint library under the same reflection path; the frequency domain distortion correlation refers to the degree of overlap between the signal frequency domain distortion position of the target UAV in the electromagnetic interference source distribution scenario and the benchmark distortion position of the reference model under the same interference conditions.

[0071] In this embodiment, based on the matching results generated in step 102 (the candidate model set and the matching ranking), the RF fingerprint reference data corresponding to the highest-ranked candidate model is extracted. First, the baseline signal propagation delay for the model under a typical obstacle reflection path is obtained from the RF fingerprint reference data (e.g., the baseline delay for a high-rise building reflection path is 1.2ms). By comparing the target drone's actual delay along the same reflection path in the current scenario (e.g., 1.5ms), the difference (0.3ms) is calculated as the signal propagation delay difference parameter. Second, the baseline frequency domain distortion position for the model under a typical electromagnetic interference distribution (e.g., the distortion range in the 2.4GHz band caused by WiFi interference) is obtained from the RF fingerprint reference data. By analyzing the target drone's actual distortion position under the current interference source distribution (e.g., distortion in the 2.4-2.42GHz range), the overlap ratio (e.g., 85%) with the baseline distortion range is calculated as the frequency domain distortion correlation parameter. Finally, the delay difference parameter and the frequency domain correlation parameter are combined to form the model correlation parameter for subsequent calculation of the environmental adaptation factor.

[0072] For example, in the urban building complex monitoring scenario, the matching result of step 102 determines that model C is the candidate model with the highest matching degree. The benchmark delay (1.2ms) of model C under the high-density building reflection path is extracted from the RF fingerprint reference data of model C, and compared with the actual delay of the same path in the current scenario (due to multi-path reflection enhancement, the measured delay is 1.55ms), and the signal propagation delay difference parameter of 0.35ms is calculated. At the same time, the benchmark distortion range (2.4-2.41GHz) of model C under 2.4GHz frequency band WiFi interference is extracted from the reference data. The actual distortion range of the target drone in the current scenario is 2.4-2.415GHz (due to changes in the distribution of interference sources, the range is expanded by 5MHz). The overlap ratio is calculated to be 80% (overlapping range 10MHz / total distortion range 15MHz), which is used as the frequency domain distortion correlation parameter. After the above parameters are integrated, the model correlation parameters include a delay difference of 0.35ms and a frequency domain correlation of 80%.

[0073] Step 104: generating an identification decision including a real-time environment adaptation factor based on the model association parameters, and outputting a target UAV model result corresponding to the identification decision;

[0074] In this step, the real-time environmental adaptation factor refers to the weight correction coefficient generated based on the dynamic parameters of the current scene (the rate of change of the obstacle position, the fluctuation amplitude of the interference source intensity), which is used to compensate for the environmental offset of the RF fingerprint library matching result; the dynamic compensation mechanism refers to the process of secondary adjustment of the matching degree of the candidate model according to the environmental adaptation factor; the consistency verification mechanism refers to the step of logically verifying the consistency of the identification of the final matching model with the model features pre-stored in the RF fingerprint library.

[0075] In this embodiment, the signal propagation delay difference and frequency domain distortion correlation are first extracted from the model-related parameters and input into the environmental adaptation factor calculation model. Based on real-time tracking data (the obstacle position changes at a rate of 5 m / s) and fluctuation data, this model generates a delay correction factor and a frequency domain correction factor using a preset rate-weight mapping table and an amplitude-weight piecewise function. Next, the correction factors are dynamically compensated for the candidate model's matching degree: the delay correction factor is multiplied by the initial matching degree and delay difference parameter of model C to obtain the delay compensation matching degree; the frequency domain correction factor is multiplied by the frequency domain correlation parameter to obtain the frequency domain compensation matching degree. The two compensation results are combined according to a preset ratio to generate a comprehensive compensation matching degree. Model C is still the highest matching result based on the comprehensive matching degree ranking. Finally, the identification of model C (such as serial number UA-C-001) is verified for consistency with the model characteristics (flight altitude-signal strength curve, standard modulation waveform) pre-stored in the RF fingerprint library to confirm that its signal strength fluctuation pattern error is less than 5% and the modulation distortion overlap is greater than 75%. After the verification is passed, the result of model C being the target drone model is output.

[0076] For example, in an urban building complex monitoring scenario, the model-related parameters extracted in step 103 (delay difference 0.35ms, frequency domain correlation 80%) are input into the environmental adaptation factor calculation model. Based on a real-time obstacle position change rate of 5m / s (exceeding the threshold of 4m / s), a linear increase coefficient of 1.2 is triggered for the delay weight, and a base weight of 1.0 is added to generate a delay correction factor of 1.2. Furthermore, based on the electromagnetic interference source fluctuation amplitude of ±3dB (exceeding the threshold of 2dB), a 1dB difference triggers a stepwise increase of 0.1 / dB in the frequency domain weight, generating a frequency domain correction factor of 1.1. The initial matching score of 0.82 for model C was compensated: the delay compensation matching score = 0.82 × 1.2 × (1 - 0.35 ms / baseline delay 1.2 ms) = 0.82 × 1.2 × 0.708 ≈ 0.698; the frequency domain compensation matching score = 0.698 × 1.1 × (80% / 100%) = 0.698 × 1.1 × 0.8 ≈ 0.614. The combined overall matching score of 0.614 still surpassed the other candidate models. Model C's identification was verified against pre-stored features in the fingerprint database: the measured signal strength fluctuation pattern error was 3% (<5%), and the modulation distortion overlap was 78% (>75%). After verification, model C was output as the final identification result, which was consistent with the actual drone model in the scene.

[0077] In order to solve the problem of interference from RF signal noise components caused by complex electromagnetic interference and obstacle reflections and improve the signal noise reduction quality in dynamic open scenes, in some embodiments, according to step 102, receiving the RF signal generated by the target UAV during flight and performing dynamic noise reduction processing on the RF signal to generate a noise-reduced RF signal includes:

[0078] Step 201: Using a multi-channel signal acquisition device deployed in a monitoring area, the original radio frequency signal of the target UAV is captured in real time. The original radio frequency signal contains noise components caused by electromagnetic interference sources and obstacle reflections in the current scene.

[0079] In this step, the multi-channel signal acquisition equipment refers to multiple RF receiving modules distributedly deployed in the monitoring area, which are used to synchronously collect RF signals transmitted by drones from different spatial locations; the noise components in the original RF signals include broadband noise caused by electromagnetic interference sources (such as WiFi routers and base stations), and pulse noise caused by multipath interference caused by obstacle reflections.

[0080] In this example, in an urban building complex monitoring scenario, multi-channel signal acquisition devices deployed on three adjacent rooftops (covering the east, west, and south directions) were simultaneously activated to capture the RF signals transmitted by the target drone in flight. The east-facing device, due to its proximity to a commercial Wi-Fi hotspot, received signals mixed with strong broadband noise in the 2.4 GHz band. The west-facing device, reflecting off a high-rise glass curtain wall, received multipath impulse noise (manifested as sudden amplitude spikes). The south-facing device, with less obstruction, received signals with less noise. After synchronization and clock calibration, the signals from each channel were aggregated and sent to a central processing unit to form a raw RF signal dataset that incorporates spatially varying characteristics.

[0081] Step 202: monitoring environmental parameters in the original radio frequency signal, wherein the environmental parameters include real-time interference intensity distribution and obstacle density distribution;

[0082] In this step, the real-time interference intensity distribution refers to the spatial distribution characteristics of electromagnetic interference energy in different frequency bands, which is extracted through spectrum analysis; the obstacle density distribution refers to the number and spatial density of obstacle reflection paths in a unit area, which is measured by lidar or infrared sensors.

[0083] In this embodiment, the raw RF signals collected in step 201 are subjected to channel-by-channel spectrum analysis. Fast Fourier transform (FFT) of the east channel signal generates an energy distribution map in the 2.4-2.48 GHz frequency band, revealing a peak Wi-Fi interference intensity concentration of -22 dBm at 2.412 GHz. Time-domain amplitude statistics of the west channel signal identify dense pulse noise intervals (20-30 mutations per second). Simultaneously, a LiDAR scan of the glass curtain wall area west of the building complex reports a density of 5 obstacle reflection paths per square meter. These data are integrated to generate environmental parameters: the real-time interference intensity distribution is labeled {2.4 GHz band: strong, 5.8 GHz band: weak}, and the obstacle density distribution is labeled {west area: high, east / south areas: medium}.

[0084] Step 203: Selecting a signal filtering strength level based on the environmental parameters. If the real-time interference intensity is higher than a set range and the obstacle density is higher than a set threshold, performing layer-by-layer noise suppression on the original RF signal using multi-stage series filtering. If the real-time interference intensity is lower than the set range, performing noise suppression on the original RF signal using single-stage filtering to obtain a filtered RF signal.

[0085] In this step, multi-stage cascade filtering refers to the sequential execution of frequency band isolation, amplitude truncation, and frequency domain notching based on different noise types (broadband noise, impulse noise, etc.); single-stage filtering refers to bandpass or notch processing for a single noise type (such as main frequency band interference); and the signal filtering intensity level refers to the strategy of dynamically selecting the filtering level and suppression amplitude based on environmental parameters.

[0086] In this embodiment, based on the environmental parameters monitored in step 202 (the real-time interference intensity distribution is marked as strong interference in the 2.4 GHz band, and the obstacle density distribution is marked as high density in the west), it is determined that the multi-stage cascade filtering conditions (interference intensity > threshold, obstacle density > threshold) are met. First, the original RF signal from the east channel undergoes a first-stage frequency band isolation filter: spurious signals outside the 2.4 GHz main band are truncated (retaining 2.4-2.48 GHz) to suppress broadband noise caused by WiFi. Second, the west channel signal undergoes a second-stage amplitude truncation filter: pulse noise amplitude mutation intervals (>30 dB) are detected and replaced with the average of adjacent intervals to suppress multipath reflection noise. Finally, a third-stage frequency domain notch filter is applied to residual narrowband interference: notch windows are generated at WiFi hotspot frequencies such as 2.412 GHz and 2.422 GHz to further filter out residual interference. Due to the low interference intensity (<threshold) on the south channel, only a single-stage main band bandpass filter is applied.

[0087] Step 204: Perform signal integrity restoration processing on the filtered RF signal to generate a noise-reduced RF signal;

[0088] In this step, signal integrity restoration processing refers to repairing signal waveform distortion caused by filtering operations, including phase jump break repair, amplitude mutation loss compensation and waveform continuity reconstruction; key feature retention refers to restoring signal strength mutation segments and phase jump segments that are strongly related to the UAV flight status (such as altitude and attitude).

[0089] In this embodiment, waveform integrity testing is performed on the multi-channel signals filtered in step 203. The east channel filtered signal, due to frequency band truncation, suffers a phase jump at the 2.48 GHz boundary. A linear interpolation signal is generated by calculating the phase slopes of adjacent frequency bands to repair the phase jump. The west channel filtered signal, due to amplitude truncation, loses some amplitude jumps. The peaks and valleys of these jumps are reconstructed based on the amplitude trends of the preceding and following segments. The south channel signal has relatively good integrity, requiring only smooth filtering of the transition region. After processing, intensity jumps associated with flight altitude (e.g., a 20 dB drop in intensity during hovering) and complete phase jumps (e.g., a shortened phase jump interval to 5 ms during climb) are extracted from each channel signal and combined to generate a de-noised RF signal.

[0090] In order to improve the matching robustness of multi-dimensional signal features in complex scenarios, in some embodiments, according to step 102, combined data including signal strength fluctuation patterns, phase jump frequency, and modulation distortion features are separated from the denoised RF signal, and the combined data is matched layer by layer with a pre-built RF fingerprint library to generate a matching result, including:

[0091] Step 301: Divide the noise-reduced RF signal into multiple signal segments according to a preset time window, and extract from each signal segment the fluctuation pattern of signal strength as a function of flight altitude, the frequency distribution of phase jump intervals, and the location characteristics of modulation waveform distortion;

[0092] In this step, time window division refers to dividing the continuous RF signal into signal segments of fixed duration, which is used to analyze the periodic characteristics of the flight status; the signal strength fluctuation pattern refers to the temporal peak and valley changes in signal strength as the UAV's flight altitude rises and falls; the phase jump frequency distribution represents the number of discontinuous phase jumps in the signal segment and the statistical characteristics of its time interval; the modulation waveform distortion position feature refers to the position and length ratio of the distorted segment in the signal modulation waveform relative to the standard waveform on the time axis.

[0093] In this embodiment, the denoised RF signal is first divided into continuous signal segments of a preset time window. Each signal segment corresponds to a different stage of the drone's flight state. The signal strength curve is matched with the flight altitude change curve through a timing alignment algorithm, and the similarity between the two in peak and valley positions, amplitude differences, and change rates is analyzed to extract the fluctuation pattern characteristics; then, the number of phase jumps in each signal segment is counted and the distribution characteristics of the jump intervals are calculated through a sliding window to generate a frequency distribution feature; at the same time, the modulated waveform of the signal segment is compared with the standard waveform point by point, and the position and length ratio of the distorted segment are marked to extract the position feature. Finally, each signal segment outputs a feature vector containing the fluctuation pattern, frequency distribution, and distortion position.

[0094] Step 302: The fluctuation pattern, frequency distribution and location characteristics are integrated into combined data according to a preset weight ratio;

[0095] In this step, the preset weight ratio refers to assigning different importance weights to the three types of features according to scenario requirements (such as signal strength fluctuation weight 5, phase hop frequency weight 3, modulation distortion weight 2); the combined data is a multidimensional data set obtained by weighted fusion of the three types of feature vectors, which is used for matching calculations in the RF fingerprint library.

[0096] In this embodiment, the extracted fluctuation pattern, frequency distribution and position features are normalized separately and weightedly fused according to a preset weight ratio. The fluctuation pattern is matched with the reference curve through similarity difference calculation, the frequency distribution is determined by overlapping area analysis to determine the degree of fit with the reference distribution, and the position feature is evaluated for consistency of the distorted segment through deviation tolerance. The three are weighted and summed to generate a combined data unit. The combined data units of all signal segments are spliced ​​into complete combined data in chronological order and input into the RF fingerprint library for matching.

[0097] Step 303: perform similarity calculations on the combined data and the first-layer signal strength fluctuation pattern benchmark set stored in the radio frequency fingerprint library, and select a candidate model set whose similarity is higher than a first threshold;

[0098] In this step, the first-layer signal strength fluctuation pattern benchmark set refers to the benchmark fluctuation curves of signal strength changes with flight altitude for various types of drones in typical scenarios pre-stored in the RF fingerprint library; similarity calculation refers to quantifying the degree of matching between the target signal fluctuation pattern and the benchmark curve through dynamic time warping or waveform difference accumulation algorithm.

[0099] In this embodiment, the fluctuation pattern feature vectors of all signal segments in the combined data generated in step 302 are sequentially aligned with the first-layer reference fluctuation curves of each model in the RF fingerprint library and the difference accumulation calculation is performed to generate a similarity score for each model; models with similarity scores higher than a preset threshold are screened to form a candidate set, and their corresponding second-layer phase hop frequency reference set and third-layer modulation distortion reference set are loaded into the next matching stage.

[0100] Step 304: jointly match the frequency distribution and position features in the combined data with the second-layer phase hopping frequency reference set and modulation distortion reference set corresponding to the candidate model set to generate a matching result including a ranking of the candidate model matching degrees;

[0101] In this step, the second-layer phase jump frequency reference set refers to the phase jump interval distribution histogram of the candidate model in a typical scenario; the modulation distortion reference set refers to the reference data of the modulation waveform distortion position and length ratio of the candidate model under the same interference conditions; joint matching refers to the weighted comprehensive evaluation through frequency distribution overlapping area calculation and distortion position offset analysis.

[0102] In this embodiment, the second-layer benchmark data (phase jump frequency distribution histogram) and the third-layer benchmark data (modulation distortion position) of each model are extracted from the candidate model set, the overlapping area of ​​the frequency distribution characteristics in the target combination data and the phase jump benchmark histogram is calculated, and the offset analysis is performed on the position characteristics and the modulation distortion benchmark. The two results are weighted and summed according to the preset weights to generate the comprehensive matching degree of each candidate model, and the final matching results are formed by arranging them in descending order.

[0103] In order to address the differential impact of obstacle reflection paths and electromagnetic interference source distribution on signal characteristics in dynamic scenarios, in some embodiments, according to step 103, the model-related parameters of the target drone are extracted from the matching results. The model-related parameters include the signal propagation delay difference corresponding to the specific obstacle reflection path and the frequency domain distortion correlation corresponding to the electromagnetic interference source distribution, including:

[0104] Step 401, sorting the candidate models in the matching results according to their matching degrees, and extracting the radio frequency fingerprint reference data corresponding to the candidate model with the highest ranking;

[0105] In this step, the candidate model matching ranking refers to the model list sorted in descending order of comprehensive matching degree generated through the hierarchical matching process (signal strength fluctuation → phase jump → modulation distortion); the RF fingerprint reference data refers to the benchmark feature data of the target model in typical scenarios pre-stored in the RF fingerprint library, including signal propagation delay, phase jump frequency and modulation distortion position parameters.

[0106] In this embodiment, based on the matching results generated in step 304 (e.g., Model C > Model E > Model F), the RF fingerprint reference data corresponding to the highest-ranked candidate model (Model C) is extracted, including the reference signal propagation delay, phase jump interval distribution histogram, and modulation waveform distortion position reference interval of the model in a typical obstacle distribution scenario, providing a benchmark comparison basis for the subsequent extraction of model-related parameters.

[0107] Step 402: Acquire a signal propagation delay difference parameter associated with obstacle distribution in the current scene from the radio frequency fingerprint reference data;

[0108] In this step, the signal propagation delay difference parameter refers to the deviation between the actual signal propagation delay of the target UAV under the current obstacle distribution and the benchmark delay under the same scenario in the RF fingerprint reference data. It is used to quantify the impact of dynamic environmental changes on the signal propagation path.

[0109] In this embodiment, the baseline signal propagation delay (e.g., baseline delay) of model C in a typical high-rise building dense scenario is obtained from the RF fingerprint reference data. By comparing the actual delay (e.g., actual delay) under the same obstacle distribution conditions in the current scenario, the difference between the two is calculated as the delay difference parameter, reflecting the impact of dynamic changes in the current obstacle position (e.g., the addition of a temporary obstacle) on the signal propagation path.

[0110] Step 403: Acquire a frequency domain distortion correlation parameter associated with the electromagnetic interference source distribution from the radio frequency fingerprint reference data;

[0111] In this step, the frequency domain distortion correlation parameter refers to the overlap ratio between the actual signal frequency domain distortion position of the target UAV under the current electromagnetic interference source distribution and the baseline distortion position under the same interference conditions in the RF fingerprint reference data. It is used to quantify the degree of interference of the dynamic changes of electromagnetic interference on the signal frequency domain characteristics.

[0112] In this embodiment, the reference frequency domain distortion position (e.g., a specific frequency band interval) of model C in a typical electromagnetic interference scenario (e.g., densely distributed WiFi hotspots) is extracted from the RF fingerprint reference data. By comparing the actual distortion position (e.g., frequency band expansion or offset) caused by the distribution of electromagnetic interference sources in the current scenario (e.g., newly added temporary communication base stations), the overlap ratio between the actual distortion interval and the reference interval is calculated, and a frequency domain distortion correlation parameter is generated to reflect the offset effect of the current interference fluctuation on the signal characteristics.

[0113] Step 404: Integrate the signal propagation delay difference parameter and the frequency domain distortion correlation parameter into a model correlation parameter of the target UAV;

[0114] In this step, the model correlation parameter refers to a composite parameter that integrates the signal propagation delay difference (reflecting the dynamic impact of obstacles) and the frequency domain distortion correlation (reflecting the dynamic impact of electromagnetic interference). It is used to quantify the combined effect of the current scene environment fluctuations on the target drone signal characteristics.

[0115] In this embodiment, the signal propagation delay difference parameter extracted in step 402 (e.g., the delay deviation caused by the addition of an obstacle) and the frequency domain distortion correlation parameter generated in step 403 (e.g., the decrease in overlap caused by interference expansion) are integrated in a preset format (e.g., a vector or matrix) to form a composite parameter set including delay difference weights and frequency domain correlation weights. This provides input for the subsequent dynamic calculation of the real-time environment adaptation factor and is associated with the benchmark data correction logic of the target model in the RF fingerprint library.

[0116] In order to improve the real-time performance and accuracy of identification decisions under dynamic environmental fluctuations, in some embodiments, according to step 104, generating an identification decision including a real-time environmental adaptation factor based on the model-related parameters and outputting a target drone model result corresponding to the identification decision includes:

[0117] Step 501: Acquire real-time tracking data of dynamic changes in obstacle positions and real-time fluctuation data of electromagnetic interference source intensity in the current scene;

[0118] In this step, real-time tracking data refers to the information on obstacle position movement rate and distribution density changes collected in real time by multi-source sensors (such as lidar and cameras); real-time fluctuation data refers to the dynamic change characteristics of the energy intensity of the electromagnetic interference source in the frequency domain and time domain monitored by spectrum analysis equipment.

[0119] In this embodiment, a lidar array deployed in the monitoring area scans the positions of obstacles in real time to generate a thermal distribution map that includes movement rate (such as the amount of obstacle displacement per unit time) and density changes. At the same time, a distributed spectrum monitoring network samples the frequency band energy of electromagnetic interference sources (such as WiFi routers and base stations) to generate the fluctuation amplitude (such as the peak energy variation range) and temporal fluctuation trend of the interference intensity at each frequency point. After the above data are aligned in time and space, a set of environmental parameters that are strongly correlated with the current scene dynamics is formed, providing input for weight coefficient adjustment.

[0120] Step 502: adjusting the weight coefficient of the signal propagation delay difference parameter according to the real-time tracking data, and adjusting the weight coefficient of the frequency domain distortion correlation parameter according to the real-time fluctuation data;

[0121] In this step, weight coefficient adjustment refers to the adaptive correction of the decision weights of model-related parameters (delay difference, frequency domain correlation) based on environmental dynamic parameters (obstacle movement rate, interference fluctuation amplitude) to enhance the response sensitivity to scene changes.

[0122] In this embodiment, based on the comparative relationship between the obstacle movement rate in the real-time tracking data and the preset rate threshold, a linear enhancement rule is used to adjust the weight coefficient of the delay difference parameter: if the rate exceeds the threshold, the weight is increased according to the ratio of the rate to the threshold; at the same time, based on the difference between the fluctuation amplitude of the interference intensity in the real-time fluctuation data and the preset amplitude threshold, a segmented increasing rule is used to adjust the weight coefficient of the frequency domain correlation parameter: if the amplitude exceeds the threshold, the weight coefficient gradient is increased segmentedly according to the difference, and finally an updated weight adapted to the environment is generated.

[0123] Step 503: Input the adjusted signal propagation delay difference parameter and the adjusted frequency domain distortion correlation parameter into the environment adaptation factor calculation model to generate a real-time environment adaptation factor for correcting the matching degree of candidate models in the radio frequency fingerprint library;

[0124] In this step, the environmental adaptation factor calculation model refers to the calculation logic that generates a dynamic correction factor by weightedly fusing the signal propagation delay difference parameters and the frequency domain distortion correlation parameters based on dynamic weight coefficients (delay difference weight, frequency domain correlation weight). The real-time environmental adaptation factor is a correction coefficient that quantifies the impact of current environmental fluctuations on the matching results.

[0125] In this embodiment, the delay difference weight coefficient and the frequency domain correlation weight coefficient adjusted in step 502 are weighted and superimposed with the corresponding delay difference parameters (such as the delay deviation caused by obstacle movement) and the frequency domain distortion correlation parameters (such as the decrease in overlap caused by interference expansion) to generate a delay correction factor and a frequency domain correction factor. The two are combined into a real-time environment adaptation factor using a preset fusion rule (such as 60% for the delay correction factor and 40% for the frequency domain correction factor), which is used to dynamically correct the matching degree of candidate models in the RF fingerprint library.

[0126] Step 504: dynamically compensating the matching degree of the candidate models in the matching result according to the real-time environment adaptation factor, and generating a recognition decision including a final matching model;

[0127] In this step, dynamic compensation refers to the correlation calculation between the real-time environmental adaptation factor and the initial matching degree of the candidate model to eliminate the matching degree deviation caused by environmental fluctuations; the recognition decision is the final model determination result generated by the comprehensive matching degree ranking after compensation.

[0128] In this embodiment, the initial matching degree of the candidate model is extracted from the matching results (e.g., the initial matching degree of model C is 0.82). The real-time environment adaptation factor is multiplied by the initial matching degree, the delay difference parameter, and the frequency domain correlation parameter for compensation step by step to generate the delay compensation matching degree and the frequency domain compensation matching degree. After the two are combined into a comprehensive compensation matching degree according to a preset ratio, the candidate models are sorted in descending order, and the model with the highest matching degree is selected as the final result. The model is then checked for consistency with the model features in the RF fingerprint library to ensure output reliability.

[0129] Step 505: Verify the consistency of the model identifier of the final matching model in the identification decision with the model feature description pre-stored in the radio frequency fingerprint library. If the consistency verification passes, output the final matching model as the target drone model result;

[0130] In this step, consistency verification refers to the process of comparing the measured feature data of the final matching model (such as signal strength fluctuation pattern, modulation distortion position) with the benchmark features of the model pre-stored in the RF fingerprint library in multiple dimensions to verify their logical consistency; model feature description refers to the unique identifier of the drone model stored in the RF fingerprint library and its corresponding benchmark feature parameter set (such as standard fluctuation curve, modulation waveform template).

[0131] In this embodiment, the pre-stored model feature description of the final matching model is extracted from the RF fingerprint library, including the standard signal strength fluctuation curve, the phase jump interval distribution histogram and the modulation waveform template; the measured signal strength fluctuation pattern of model C is time-aligned with the standard fluctuation curve and the peak-valley position error is calculated, and the interval overlap analysis of the measured modulation distortion position and the template is performed; if the fluctuation pattern error is less than the preset threshold and the modulation distortion overlap is higher than the minimum requirement, the consistency verification is determined to be passed and the model result is output; otherwise, an abnormal alarm is triggered and the matching process is re-executed.

[0132] In order to address the differentiated requirements for weight coefficients due to dynamic changes in obstacle positions and fluctuations in interference source strength, in some embodiments, according to step 502, adjusting the weight coefficient of the signal propagation delay difference parameter based on the real-time tracking data, and adjusting the weight coefficient of the frequency domain distortion correlation parameter based on the real-time fluctuation data, includes:

[0133] Step 601: Dynamically adjust the weight coefficient of the signal propagation delay difference parameter based on the dynamic rate of change of the obstacle position in the real-time tracking data. When the dynamic rate exceeds a preset rate threshold, linearly enhance the weight coefficient according to the ratio of the dynamic rate to the preset rate threshold to generate an adjusted delay weight coefficient.

[0134] In this embodiment, the dynamic rate of change of the obstacle position (such as the displacement of the obstacle per unit time) is extracted through real-time tracking data and compared with a preset rate threshold. If the dynamic rate exceeds the threshold, the weight coefficient of the delay difference parameter is linearly enhanced according to the ratio of the rate to the threshold. For example, when the rate exceeds the threshold, the weight coefficient is positively adjusted according to the proportion of the rate excess; if it does not exceed the threshold, the original weight coefficient is maintained unchanged, ensuring that the weight of the delay difference parameter dynamically reflects the actual impact of the obstacle movement on the signal propagation path.

[0135] Step 602: dynamically adjust the weight coefficient of the frequency domain distortion correlation parameter based on the fluctuation amplitude of the electromagnetic interference source intensity in the real-time fluctuation data. When the fluctuation amplitude exceeds a preset amplitude threshold, the weight coefficient is incrementally increased in sections according to the difference between the fluctuation amplitude and the amplitude threshold to generate an adjusted frequency domain weight coefficient.

[0136] In this embodiment, the fluctuation amplitude of the electromagnetic interference source intensity (such as the difference between the peak and valley of the interference energy) is extracted through real-time fluctuation data and compared with the preset amplitude threshold. If the fluctuation amplitude exceeds the threshold, the weight coefficient of the frequency domain distortion correlation parameter is increased in segments according to the difference between the amplitude and the threshold. For example, every time a certain amplitude interval exceeds the threshold, the weight coefficient increases according to the gradient; if it does not exceed the threshold, the original weight coefficient is maintained to ensure that the frequency domain parameter weight is adaptively adjusted with the degree of interference fluctuation.

[0137] To address the limitation of the environmental adaptation factor in providing a one-dimensional correction to the matching results, in some embodiments, according to step 504, the matching degree of the candidate models in the matching results is dynamically compensated according to the real-time environmental adaptation factor to generate a recognition decision including the final matching model, including:

[0138] Step 701: Decompose the real-time environment adaptation factor into a time delay correction factor and a frequency domain correction factor;

[0139] In this embodiment, the real-time environmental adaptation factor is decomposed into a delay correction factor and a frequency domain correction factor through a preset weight separation rule. The delay correction factor is generated by the dynamic weight adjustment result of the signal propagation delay difference parameter, reflecting the impact of the dynamic change of the obstacle position on the matching degree; the frequency domain correction factor is generated by the dynamic weight adjustment result of the frequency domain distortion correlation parameter, reflecting the correction requirement for the matching degree due to electromagnetic interference fluctuations. Both are decoupled from the environmental adaptation factor through a weighted proportional allocation mechanism to form independent correction parameters to support the step-by-step compensation logic.

[0140] Step 702: extracting the initial matching degree of the candidate model set from the matching results, and multiplying the delay correction factor by the signal propagation delay difference parameter corresponding to each candidate model in the initial matching degree item by item to generate a matching degree after delay compensation;

[0141] In this embodiment, the initial matching degree of the candidate model is extracted from the matching results (a hierarchical matching score based on the signal strength fluctuation pattern and phase jump frequency), and the delay correction factor is multiplied by the signal propagation delay difference parameter corresponding to each candidate model (such as the delay deviation value caused by obstacle dynamics) to generate the matching degree after delay compensation. The larger the delay difference parameter (the more significant the impact of obstacle movement), the stronger the compensation effect of the delay correction factor on the matching degree, thereby increasing the matching priority of models that are sensitive to environmental dynamics.

[0142] Step 703: multiply the frequency domain correction factor by the frequency domain distortion correlation parameter corresponding to each candidate model in the matching degree after delay compensation, to generate the matching degree after frequency domain compensation;

[0143] In this embodiment, the frequency domain distortion correlation parameter (reflecting the degree of overlap between signal distortion and the benchmark under electromagnetic interference fluctuations) of each model is extracted from the candidate model set, and the frequency domain correction factor (the weight adjustment result triggered by the interference fluctuation amplitude) is multiplied by the frequency domain correlation parameter of each model item by item to generate the matching degree after frequency domain compensation; the lower the frequency domain correlation parameter (the interference causes the distortion overlap to decrease), the stronger the compensatory effect of the frequency domain correction factor on the matching degree, thereby giving priority to improving the matching weight of the anti-interference model.

[0144] Step 704: Combine the delay-compensated matching degree and the frequency-domain-compensated matching degree according to a preset ratio to generate a comprehensive compensation matching degree for each candidate model.

[0145] In this embodiment, the combined weight ratio of delay compensation and frequency domain compensation is preset according to the dynamic characteristics of the scene (for example, 60% for delay and 40% for frequency domain), and the matching degree after delay compensation (reflecting the impact of obstacle movement) and the matching degree after frequency domain compensation (reflecting the impact of interference fluctuations) are weighted and combined to generate a comprehensive compensated matching degree. The weighted merging mechanism is used to enhance the joint correction capability for complex environmental dynamics, and finally the candidate models are sorted in descending order according to the comprehensive matching degree to output the optimal recognition result.

[0146] Step 705 , sorting the candidate model set in descending order according to the comprehensive compensation matching degree, selecting the candidate model with the highest ranking as the final matching model, and generating an identification decision including the final matching model;

[0147] In this step, the comprehensive compensation matching degree refers to the comprehensive score of the candidate model generated by combining the preset ratio after the step-by-step compensation of the time delay correction factor and the frequency domain correction factor; descending sorting refers to arranging the candidate models from high to low according to the comprehensive compensation matching degree; the recognition decision is the judgment result data set containing the final matching model identification and matching degree verification information.

[0148] In this embodiment, the system numerically compares the comprehensive compensation matching degree of each model in the candidate model set, generates a sorted list in descending order, and selects the model with the highest ranking as the final matching model; at the same time, the identification, comprehensive compensation matching degree, time delay / frequency domain compensation parameters and environmental adaptation factors of the final matching model are integrated into structured recognition decision data for subsequent consistency verification and result output to ensure the traceability and reliability of the decision-making process.

[0149] Figure 2 The present invention provides a schematic diagram of a system for quickly identifying specific drone models based on a radio frequency fingerprint library. Figure 2 As shown, the system includes:

[0150] A receiving module is used to receive the radio frequency signal generated by the target UAV during flight, and dynamically reduce the noise of the radio frequency signal to generate a noise-reduced radio frequency signal;

[0151] A matching module is used to separate the combined data containing the signal strength fluctuation pattern, phase hop frequency and modulation distortion characteristics from the noise-reduced RF signal, and match the combined data with the pre-built RF fingerprint library layer by layer to generate a matching result;

[0152] An extraction module is configured to extract model-related parameters of the target UAV from the matching results, wherein the model-related parameters include a signal propagation delay difference corresponding to a specific obstacle reflection path and a frequency domain distortion correlation corresponding to an electromagnetic interference source distribution;

[0153] The output module is used to generate an identification decision including a real-time environment adaptation factor based on the model association parameters, and output a target UAV model result corresponding to the identification decision.

[0154] Figure 2 The specific drone model rapid identification system based on the radio frequency fingerprint library can be executed Figure 1 The implementation principles and technical effects of the method for rapidly identifying specific drone models based on a radio frequency fingerprint library described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the aforementioned method for rapidly identifying specific drone models based on a radio frequency fingerprint library has been described in detail in the relevant embodiments and will not be further elaborated here.

[0155] In one possible design, Figure 2 The embodiment shown is a system for quickly identifying a specific drone model based on a radio frequency fingerprint library, which can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0156] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0157] The processing component 32 is used for the above Figure 1 The embodiment provides a method for quickly identifying a specific drone model based on a radio frequency fingerprint library.

[0158] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0159] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0160] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0161] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0162] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0163] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0164] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment provides a method for quickly identifying specific drone models based on a radio frequency fingerprint library.

[0165] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0166] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0167] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for quickly identifying specific drone models based on a radio frequency fingerprint library, characterized in that: include: Receive the radio frequency signal generated by the target UAV during flight, and perform dynamic noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal; Separate the combined data containing signal strength fluctuation pattern, phase hopping frequency and modulation distortion characteristics from the de-noised RF signal, and match the combined data with the pre-built RF fingerprint library layer by layer to generate a matching result; Extracting model-related parameters of the target UAV from the matching results, the model-related parameters including the signal propagation delay difference corresponding to the reflection path of a specific obstacle and the frequency domain distortion correlation corresponding to the distribution of the electromagnetic interference source; Generating an identification decision including a real-time environment adaptation factor based on the model association parameters, and outputting a target UAV model result corresponding to the identification decision; Generating an identification decision including a real-time environment adaptation factor based on the model association parameters, and outputting a target UAV model result corresponding to the identification decision, including: Obtain real-time tracking data of the dynamic changes in the position of obstacles in the current scene and real-time fluctuation data of the intensity of electromagnetic interference sources; Adjusting the weight coefficient of the signal propagation delay difference parameter according to the real-time tracking data, and adjusting the weight coefficient of the frequency domain distortion correlation parameter according to the real-time fluctuation data; Input the adjusted signal propagation delay difference parameter and the adjusted frequency domain distortion correlation parameter into the environmental adaptation factor calculation model to generate a real-time environmental adaptation factor for correcting the matching degree of candidate models in the radio frequency fingerprint library; Dynamically compensating the matching degree of candidate models in the matching results according to the real-time environment adaptation factor to generate an identification decision including a final matching model; The model identifier of the final matching model in the identification decision is checked for consistency with the model feature description pre-stored in the radio frequency fingerprint library. When the consistency verification passes, the final matching model is output as the target drone model result.

2. The method according to claim 1, characterized in that Receiving a radio frequency signal generated by a target UAV during flight and dynamically performing noise reduction processing on the radio frequency signal to generate a noise-reduced radio frequency signal, including: The original radio frequency signal of the target UAV is captured in real time by a multi-channel signal acquisition device deployed in the monitoring area. The original radio frequency signal contains noise components caused by electromagnetic interference sources and obstacle reflections in the current scene; Monitoring environmental parameters in the original radio frequency signal, wherein the environmental parameters include real-time interference intensity distribution and obstacle density distribution; Selecting a signal filtering intensity level according to the environmental parameters; if the real-time interference intensity is higher than a set range and the obstacle density is higher than a set threshold, performing layer-by-layer noise suppression on the original radio frequency signal using multi-stage series filtering; and if the real-time interference intensity is lower than the set range, performing noise suppression on the original radio frequency signal using single-stage filtering to obtain a filtered radio frequency signal; The filtered RF signal is processed for signal integrity restoration to generate a noise-reduced RF signal.

3. The method according to claim 1, characterized in that Separate the combined data containing signal strength fluctuation patterns, phase hopping frequency, and modulation distortion features from the de-noised RF signal, and match the combined data layer by layer with the pre-built RF fingerprint library to generate matching results, including: Dividing the denoised radio frequency signal into multiple signal segments according to a preset time window, and extracting the fluctuation pattern of signal strength as it changes with flight altitude, the frequency distribution of phase jump intervals, and the position characteristics of modulation waveform distortion from each signal segment; The fluctuation pattern, frequency distribution and location characteristics are integrated into combined data according to a preset weight ratio; Calculate similarity between the combined data and the first-layer signal strength fluctuation pattern benchmark set stored in the radio frequency fingerprint library, and select a candidate model set whose similarity is higher than a first threshold; The frequency distribution and position features in the combined data are jointly matched with the second-layer phase jump frequency reference set and modulation distortion reference set corresponding to the candidate model set to generate a matching result including the candidate model matching degree ranking.

4. The method according to claim 1, wherein Extracting the model-related parameters of the target UAV from the matching results, the model-related parameters include the signal propagation delay difference corresponding to the specific obstacle reflection path and the frequency domain distortion correlation corresponding to the electromagnetic interference source distribution, including: Sorting the candidate models in the matching results according to their matching scores, extracting the radio frequency fingerprint reference data corresponding to the candidate model with the highest ranking; Obtaining a signal propagation delay difference parameter associated with obstacle distribution in a current scene from the radio frequency fingerprint reference data; Acquire a frequency domain distortion correlation parameter associated with the electromagnetic interference source distribution from the radio frequency fingerprint reference data; The signal propagation delay difference parameter and the frequency domain distortion correlation parameter are integrated into the model correlation parameter of the target UAV.

5. The method according to claim 1, wherein Adjusting the weight coefficient of the signal propagation delay difference parameter according to the real-time tracking data, and adjusting the weight coefficient of the frequency domain distortion correlation parameter according to the real-time fluctuation data, including: Dynamically adjusting a weight coefficient of a signal propagation delay difference parameter based on a dynamic rate of change of an obstacle position in the real-time tracking data; when the dynamic rate exceeds a preset rate threshold, linearly enhancing the weight coefficient based on a ratio of the dynamic rate to the preset rate threshold to generate an adjusted delay weight coefficient; According to the fluctuation amplitude of the electromagnetic interference source intensity in the real-time fluctuation data, the weight coefficient of the frequency domain distortion correlation parameter is dynamically adjusted. When the fluctuation amplitude exceeds the preset amplitude threshold, the weight coefficient is increased in segments according to the difference between the fluctuation amplitude and the amplitude threshold to generate an adjusted frequency domain weight coefficient.

6. The method according to claim 1, characterized in that Dynamically compensating the matching degree of the candidate models in the matching results according to the real-time environment adaptation factor to generate an identification decision including the final matching model, including: Decomposing the real-time environment adaptation factor into a time delay correction factor and a frequency domain correction factor; Extracting an initial matching degree of the candidate model set from the matching results, and multiplying the delay correction factor by the signal propagation delay difference parameter corresponding to each candidate model in the initial matching degree item by item to generate a matching degree after delay compensation; Multiplying the frequency domain correction factor by the frequency domain distortion correlation parameter corresponding to each candidate model in the matching degree after delay compensation item by item to generate the matching degree after frequency domain compensation; The matching degree after delay compensation and the matching degree after frequency domain compensation are combined according to a preset ratio to generate a comprehensive compensation matching degree for each candidate model; The candidate model set is sorted in descending order according to the comprehensive compensation matching degree, the candidate model with the highest ranking is selected as the final matching model, and an identification decision including the final matching model is generated.

7. A system for quickly identifying specific drone models based on a radio frequency fingerprint library, used to execute a method for quickly identifying specific drone models based on a radio frequency fingerprint library according to any one of claims 1 to 6, characterized in that: include: A receiving module is used to receive the radio frequency signal generated by the target UAV during flight, and dynamically reduce the noise of the radio frequency signal to generate a noise-reduced radio frequency signal; A matching module is used to separate the combined data containing the signal strength fluctuation pattern, phase hop frequency and modulation distortion characteristics from the noise-reduced RF signal, and match the combined data with the pre-built RF fingerprint library layer by layer to generate a matching result; An extraction module is configured to extract model-related parameters of the target UAV from the matching results, wherein the model-related parameters include a signal propagation delay difference corresponding to a specific obstacle reflection path and a frequency domain distortion correlation corresponding to an electromagnetic interference source distribution; The output module is used to generate an identification decision including a real-time environment adaptation factor based on the model association parameters, and output a target UAV model result corresponding to the identification decision.

8. A computing device, characterized in that The method comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for quickly identifying a specific drone model based on a radio frequency fingerprint library as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a method for quickly identifying a specific drone model based on a radio frequency fingerprint library as described in any one of claims 1 to 6 is implemented.

Images