Anti-unmanned aerial vehicle passive positioning method and system based on low-orbit satellite external radiation source

Abstract

The invention provides an anti-unmanned aerial vehicle passive positioning method and system based on an external radiation source of a low-orbit satellite, and relates to the technical field of satellite positioning, and the method comprises the steps: employing the external radiation source of the low-orbit satellite as an irradiation signal source under the same time reference, and carrying out the irradiation of each ground receiving station, performing synchronous sampling on the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel, performing frequency correction and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel, and performing feature comparison on the complex sampling sequence of the reference receiving channel and a historical robust complex sampling sequence; and obtaining a feature comparison result, and based on the feature comparison result, updating a background reference value used for delay-frequency shift image background normalization and a background reference value allowable fluctuation interval, so that the effectiveness of multi-station observation consistency constraint can be improved, a more stable unmanned aerial vehicle position solution can be obtained, and thus anti-unmanned aerial vehicle passive positioning is realized.

Description

Technical Field

[0001] This invention relates to the field of satellite positioning technology, and in particular to a passive anti-drone positioning method and system based on low-orbit satellite external radiation sources. Background Technology

[0002] In counter-drone scenarios, to detect, identify, and locate low-altitude drones, existing technologies typically focus on the radio frequency radiation characteristics exhibited by the drone and its control link. Specifically, one common approach uses the radio signals from the drone's control or image transmission link as the target. With multiple receiving nodes deployed, it collects time-delay observations such as signal arrival time or time difference, and obtains the signal source location through multi-point joint calculation. Furthermore, it can schedule the time interval between data acquisition and calculation using monitoring strategies to maintain continuous monitoring of the target area under certain resource constraints. Another approach focuses on acquiring the direction of arrival information through antenna arrays, estimating the signal incident direction using array signal processing methods, and, with multi-station deployment or multi-array coordination, using the intersection of azimuth and elevation angle constraints from different stations to locate the drone or its radiation source.

[0003] For example, Chinese invention patent CN119224806B discloses a grading evaluation method for centimeter-level PPK positioning results of unmanned aerial vehicles (UAVs). Through PPK positioning calculation, it obtains the state vectors and their variance-covariance matrices, positioning solution states, and position accuracy attenuation factor (PDOP) for each epoch of forward and reverse Kalman filtering. It performs common-view analysis at each epoch of forward and reverse Kalman filtering and conducts a two-tailed normality test on the cross-difference results of the forward and reverse filtered state vectors after common-view analysis. Combining the hypothesis test results, positioning solution states, and PDOP values, it performs a positioning grading evaluation and optimizes the output results.

[0004] For example, Chinese invention patent CN112558111B discloses a drone positioning method and device. This method determines the drone's reference coordinates based on first ranging information fed back from multiple base stations monitoring the drone's location. When a target base station with a positioning error record is identified among the multiple base stations, a correction function corresponding to the target base station is obtained. The reference coordinates are then calibrated using the correction function to obtain the drone's positioning information. This prevents inaccurate drone positioning due to base station failures. At the same time, even if the faulty base station has not been repaired, drone positioning can still be performed to execute drone operations.

[0005] The above-mentioned problem has the following technical issues:

[0006] In the passive anti-drone positioning process based on low-Earth orbit satellite signals, the receiver typically sets up a reference receiving channel and a monitoring receiving channel. The reference receiving channel is used to stably receive direct satellite signals and form a comparison benchmark, while the monitoring receiving channel is used to receive mixed signals from the protected airspace. The system performs time and frequency alignment on the two sampled data streams, and after background component suppression processing on the monitoring receiving channel data, it calculates a two-dimensional correlation result with relative time delay and relative frequency shift as independent variables. Then, it selects the peak position from this two-dimensional correlation result as the source of the time delay difference and frequency shift difference observations required for subsequent position calculation.

[0007] Because two-dimensional correlation results are calculated from discrete sampled data within a finite time interval, and because the actual receiving environment contains interference factors such as background reflection, multipath propagation, and noise, the two-dimensional correlation results are prone to phenomena such as energy diffusion, increased local peaks, or peak splitting in the relative time delay and relative frequency shift dimensions. Especially in low-altitude scenarios, the scattered signal from UAVs is usually weaker than the background component. The local peaks in the two-dimensional correlation results may originate from UAV scattering, background reflection, or the correlation response formed by multipath propagation. Furthermore, the latter may exhibit similar peak shape and amplitude characteristics to the former in a single calculation, making it difficult to distinguish the source of the peaks based solely on amplitude or simple threshold rules.

[0008] Therefore, existing processes often exhibit multiple local peaks simultaneously in two-dimensional correlation results, making it difficult to effectively distinguish their sources based solely on peak amplitude or simple threshold rules. This results in the peak positions used to characterize UAV scattering components being difficult to determine stably. Consequently, peak selection is prone to inconsistencies in calculation results across different time intervals, leading to a lack of continuity and consistency in the time delay and frequency shift differences derived from peak positions. Since there is a risk of misselection or instability in the observation extraction stage, observation bias will further propagate to subsequent position calculation stages, making the position calculation results prone to multiple solutions, jumps, or systematic deviations. This makes it difficult to accurately reflect the target's true spatial location, thereby reducing the reliability and availability of positioning outputs in anti-UAV applications. Summary of the Invention

[0009] Therefore, embodiments of the present invention provide a passive anti-drone positioning method and system based on low-orbit satellite external radiation sources, which can improve the reliability and availability of positioning output in anti-drone applications.

[0010] The technical solution of this invention is implemented as follows:

[0011] This invention provides a passive anti-drone positioning method based on a low-Earth orbit satellite external radiation source. The method includes: Step 1: Under the same time reference, using a low-Earth orbit satellite external radiation source as the illumination signal source, synchronously sampling the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel for each ground receiving station, and performing frequency correction and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel; Step 2: Performing feature comparison between the complex sampling sequence of the reference receiving channel and the historical robust complex sampling sequence, obtaining the feature comparison result, and updating the background reference value and the allowable fluctuation range of the background reference value for background normalization of the delay-frequency shift map based on the feature comparison result; Step 3: Obtaining the delay-frequency shift combination set. In the delay-frequency shift combination set, the complex sampling sequence of the monitoring receiving channel is corrected under each delay-frequency shift combination, and a correlation analysis is performed with the complex sampling sequence of the reference receiving channel. A delay-frequency shift map is generated based on the correlation analysis results. Based on the delay-frequency shift map, background reference value, and allowable fluctuation range of the background reference value, the high-confidence candidate peak points of each UAV are analyzed. Step four: The high-confidence candidate peak points of each UAV in each ground receiving station are statistically analyzed. Combining the low-orbit satellite position and the position of each ground receiving station, a weighted residual cost function is established to evaluate the consistency of the UAV position observed by multiple ground receiving stations. The weighted residual cost function is iteratively optimized to minimize the weighted residual cost function, thereby solving the position of each UAV and realizing passive anti-UAV positioning.

[0012] This application also provides a passive anti-drone positioning system based on a low-Earth orbit satellite external radiation source. This system applies a passive anti-drone positioning method based on a low-Earth orbit satellite external radiation source. The system includes: a sequence correction module, used to synchronously sample the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel for each ground receiving station, using the low-Earth orbit satellite external radiation source as the illumination signal source, and performing frequency and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel; and a feature comparison module, used to perform feature comparison between the complex sampling sequence of the reference receiving channel and historical robust complex sampling sequences, obtain the feature comparison results, and update the background reference value and the allowable fluctuation range of the background reference value for delay-frequency shift map background normalization based on the feature comparison results. The system comprises two modules: a peak point analysis module, which acquires a set of delay-frequency shift combinations, corrects the complex sampling sequence of the monitoring receiving channel under each delay-frequency shift combination, and performs correlation analysis with the complex sampling sequence of the reference receiving channel. A delay-frequency shift map is generated based on the correlation analysis results. High-confidence candidate peak points for each UAV are analyzed based on the delay-frequency shift map, background reference values, and the allowable fluctuation range of the background reference values. The location positioning module is used to statistically identify high-confidence candidate peak points for each UAV in various ground receiving stations. Combining the low-orbit satellite position and the positions of various ground receiving stations, a weighted residual cost function is established to evaluate the consistency of UAV positions observed by multiple ground receiving stations. This function is iteratively optimized to minimize the weighted residual cost function, thereby solving for the position of each UAV and achieving passive anti-UAV positioning.

[0013] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0014] 1. This invention utilizes a low-Earth orbit satellite external radiation source as the illumination signal source under the same time reference. Ground receiving stations synchronously sample the complex sampling sequences of the reference receiving channel and the monitoring receiving channel, and implement frequency and phase corrections to reduce the broadening and drift effects of residual frequency deviation and phase drift on the delay-shift correlation results. The complex sampling sequence of the reference receiving channel is compared with historical robust complex sampling sequences. Based on the comparison results, the background reference value and allowable fluctuation range used for background normalization of the delay-shift map are updated, allowing the background reference to adaptively adjust with environmental changes, thereby reducing interference from direct leakage and ground clutter on candidate peak extraction. The monitoring receiving channel sequence is corrected based on the delay-shift combination set and correlated with the reference channel to generate a delay-shift map. This map is then combined with the background reference value and allowable fluctuation range to form a deviation characterization and screen high-confidence candidate peaks, avoiding misselection of non-target peaks and time-window jumps when multiple peaks coexist. Further, high-confidence candidate peaks from multiple stations are summarized, and a weighted residual cost function is constructed by combining the low-orbit satellite position and the position of each ground receiving station. This function is then iteratively minimized to improve the effectiveness of the multi-station observation consistency constraint and obtain a more stable UAV position solution, thereby achieving passive anti-UAV positioning.

[0015] 2. This invention compares the amplitude stability, phase continuity, frequency consistency, and peak shape coherence characteristics of the complex sampling sequence of the reference receiving channel with historical robust complex sampling sequences. It also uses a linkage rule between the number of inconsistent features and anomaly counts to hierarchically maintain the background reference value and its allowable fluctuation range, enabling differentiated handling in two scenarios: short-term disturbances and long-term anomalies in the reference channel. When the number of inconsistent features is low, maintaining the reference and converging anomaly counts prevents frequent rises in the background benchmark due to occasional noise, thus preventing passive drift of the threshold system in the deviation map and the introduction of missed candidate peaks. When the number of inconsistent features is in the middle range, anomaly count accumulation triggers a re-update, which can respond to persistent interference, expand the allowable fluctuation range to cover background changes, and suppress overcompensation caused by single-window anomalies, reducing the probability of multiple peaks coexisting and non-target peak misselection caused by stable background residues in the delay-shift map. When the number of inconsistent features reaches a high level, the reference receiving channel is directly marked as unavailable to avoid erroneous reference templates from entering the correlation analysis and generating false peaks. This reduces the time window jumps of candidate peak points and the transmission of observational biases to the position calculation stage from the source, thereby improving the consistency of UAV positioning results.

[0016] 3. This invention transforms the candidate uncertainty caused by the coexistence of multiple peaks in a single-station delay-frequency shift map into cross-station consistency constraints by pairwise comparison and screening of high-confidence candidate peaks from different ground receiving stations to form matching peaks. This reduces the probability of non-target peaks caused by clutter sidelobes or multipath propagation being mistakenly adopted in multi-station joint processing. Furthermore, matching peaks are marked with target attribution and aggregated into target observation groups, structurally aggregating and isolating multiple observations corresponding to the same UAV. This avoids cross-matching of candidate peaks from different UAVs in multi-target scenarios, preventing mismatched observations and multiple solutions for position calculation. Delay and frequency shift parameters are extracted from each matching peak within the target observation group, and delay difference and frequency difference observations are calculated. The delay difference observation is then converted into an equivalent ranging difference, transforming the observation from pixel coordinates into a measurement form that can directly participate in geometric constraints. This reduces the amplification effect on position calculation when candidate peaks are mistakenly selected. The average confidence scores of two candidate peaks in the matching peaks are used as the confidence scores of the matching peaks. This helps to weaken or eliminate low-confidence observations during the observation aggregation stage, suppress abrupt changes in observations caused by peak jumps across time windows, improve the continuity and consistency of subsequent position calculations, and meet the requirements of anti-drone applications for stable positioning output. Attached Figure Description

[0017] Figure 1 This is a flowchart of the passive anti-drone positioning method based on low-orbit satellite external radiation source provided in the embodiments of the present invention;

[0018] Figure 2 This is a schematic diagram of the structure of the anti-drone passive positioning system based on a low-orbit satellite external radiation source provided in an embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of the delay-frequency shift heatmap provided in an embodiment of the present invention;

[0020] Figure 4 This is a schematic diagram of the deviation degree provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0022] In this application, the term "at least one" means one or more, and the term "multiple" means two or more; for example, multiple devices means two or more devices. "At least two" means two or more. "At least three" means three or more.

[0023] With the increasing frequency and scale of UAV activities in low-altitude airspace, anti-UAV systems face higher demands on the coverage, concealment, and persistence of target positioning. Existing positioning methods relying on active emission systems typically require additional emission sources or external illumination signals, which easily expose deployment locations and are affected by emission permits, power constraints, and electromagnetic countermeasures environments. Furthermore, they are prone to instability and positioning result jumps in complex terrain backgrounds and under conditions of multiple concurrent targets. To obtain stable observations usable for positioning without adding new active radiation, this embodiment proposes a passive anti-UAV positioning method based on low-orbit satellite external radiation sources. Utilizing the wide-area coverage and persistent existence of low-orbit satellite downlink communication signals, delay and frequency shift observations are constructed and consistent calculations are performed through collaborative processing by multiple ground receiving stations. This improves the reliability of positioning results under complex electromagnetic backgrounds and meets the engineering application requirements of low-altitude security protection scenarios.

[0024] Example 1: This embodiment of the invention provides a passive anti-drone positioning method based on a low-Earth orbit satellite external radiation source. For example... Figure 1 The flowchart shown is for a passive anti-drone positioning method based on low-Earth orbit satellite external radiation sources. The processing flow of this method may include the following steps:

[0025] Under the same time reference, using the external radiation source of a low-Earth orbit satellite as the illumination signal source, for each ground receiving station, the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel are synchronously sampled. Each ground receiving station is equipped with a reference receiving channel and a monitoring receiving channel. The reference receiving channel is used to stably receive the direct downlink signal from the external radiation source of the low-Earth orbit satellite to form a time and frequency reference; the monitoring receiving channel is used to receive the mixed signal from the direction of the protected airspace. The two channels are synchronously sampled under the same time reference to obtain the corresponding complex sampling sequences.

[0026] For example, ground receiving station A performs in-phase and quadrature demodulation on the baseband signal obtained by down-conversion of the received signal. At discrete time n×T with sampling interval T, the in-phase component I[n] and the quadrature component Q[n] are collected respectively. Then the complex sampling sequence can be expressed as x[n]=I[n]+j×Q[n], where j is the imaginary unit, n represents the sampling time point number, n=1, 2, 3, ..., N, and N is the total number of sampling time points. The complex sampling sequence is used to represent the baseband in-phase component and quadrature component of the received signal at each sampling time, thereby simultaneously characterizing the amplitude and phase information of the signal at that time.

[0027] Complex sampling sequence x of the reference receiving channel belonging to ground receiving station A ref [n]=I ref [n]+j×Q ref [n], the complex sampling sequence x of the monitoring and receiving channel belonging to ground receiving station A.mon [n]=I mon [n]+j×Q mon [n].

[0028] Due to slight deviations between the local oscillator and the sampling clock at various ground receiving stations, the reference sequence may exhibit a continuous phase rotation over time within a time window, which corresponds to the residual frequency deviation of the sequence. Simultaneously, the reference sequence may show a slow shift in the overall phase reference between different time windows, which corresponds to the amount of sequence phase drift.

[0029] The residual frequency deviation of a sequence refers to the slight frequency inconsistency that remains after downconversion, which is directly manifested as a continuous phase rotation over time. To estimate this rotation rate, this embodiment calculates the phase increment of adjacent sampling points, using conjugate multiplication to avoid the jump problem caused by directly expanding the phase: u[n]=x ref [n]×x ref * [n-1], (·) * The conjugate is represented by u[n], which is a complex number. It means multiplying the complex sample of the current sampling point with the conjugate of the complex sample of the previous sampling point. The conjugate multiplication will cancel the amplitude influence between the two points, so that the phase of the result mainly reflects the phase difference between the two points. Therefore, u[n] can be used to stably extract the phase change of adjacent sampling points.

[0030] After obtaining u[n], we define Δϕ[n] = arg(u[n]), where Δϕ[n] represents the phase increment of the nth sampling point relative to the (n-1)th sampling point, and arg(·) represents taking a complex phase angle. We average Δϕ[n] over a time window to obtain avg_Δϕ[n], which can then be used to estimate the residual frequency deviation of the sequence. Δf represents the residual frequency deviation of the sequence, π is the mathematical constant pi, T is the sampling interval, and 2×π×T is used to convert the average phase increment of each sampling point into the frequency deviation per second.

[0031] The phase shift of a sequence refers to the slow shift of the phase reference of a reference sequence over a relatively long time scale. This manifests as inconsistencies in the phase zeros at different time points. ref [n] is summed, and S is the summation result. The summation is a vector superposition of the phase and amplitude of each sampling point within the time window, forming a composite vector representing the overall phase orientation of the time window. The phase angle of the composite vector is then taken to obtain the phase reference for the time window, ϕ0=arg(S), where arg(·) represents the phase angle of the complex number, i.e., the angle between the complex number and the real axis in the complex plane, used to characterize the overall phase reference direction of the time window. To measure the deviation of the current time window phase reference from the historical stable state, the database stores the historical robust phase reference ϕ. histLet represent the phase level of the historical steady state, and calculate the difference between the two as the sequence phase drift Δϕ0=wrap(ϕ0−ϕ). hist The wrap(·) operation represents the phase normalization operation, which limits the phase difference to the range of (−π, π] to avoid numerical jumps when the phase difference crosses ±π, which would affect subsequent compensation.

[0032] Based on the residual frequency deviation of the sequence, a frequency compensation factor g is constructed. f [n]=e -j×2×π×Δf×n×T Where n×T is the sampling time corresponding to the nth sampling point, 2×π is the angular frequency conversion coefficient, j is the imaginary unit, and e is the natural constant. The principle of this compensation factor is to apply a phase rotation opposite to the direction of the residual frequency deviation, so that the phase rotation caused by the frequency inconsistency over time is canceled out, thus reducing x... ref [n]×g f [n] is the updated x ref [n].

[0033] In performing phase correction, this embodiment constructs a phase compensation factor g based on the sequence phase drift Δϕ0. ϕ =e −j×Δϕ The principle behind this is to apply a fixed phase shift to the complex sampling sequence throughout the entire time window, aligning the phase reference of the current time window with a historical robust phase reference. Based on the phase compensation factor, phase correction is performed on the reference sequence, and x... ref [n]×g ϕ As the updated x ref [n].

[0034] Since the reference receiving channel and the monitoring receiving channel are located in the same ground receiving station and share the local oscillator and sampling clock, the residual frequency deviation and phase drift experienced by the two channels are statistically consistent. To ensure that the two channels are aligned under the same frequency and phase references, this embodiment applies the same set of compensation factors synchronously to the complex sampling sequence x of the monitoring receiving channel. mon [n], to obtain the corrected monitoring complex sampling sequence x mon [n]×g f [n]×g ϕ As the updated x mon [n].

[0035] After completing the above processing, the impact of residual frequency and phase drift on delay-frequency shift heatmap calculation and candidate peak stability can be reduced.

[0036] Ground receiving station A pre-stores a set of historical robust complex sampling sequences during its historical operation. The historical robust complex sampling sequences refer to reference sequences that are collected and filtered under relatively stable electromagnetic environment conditions without obvious abnormal interference and are stored as a benchmark. They are used to characterize the typical amplitude and phase characteristics of the reference receiving channel under normal background conditions.

[0037] The complex sampling sequence of the reference receiving channel is compared with historical robust complex sampling sequences in terms of features. This comparison includes amplitude stability feature comparison, phase continuity feature comparison, frequency consistency feature comparison, and peak shape coherence feature comparison. Each of these features is calculated from the complex sampling sequence of the reference channel within the current time window and compared with the corresponding historical benchmark features in the database. The comparison results are marked as consistent or inconsistent depending on whether they exceed the allowable fluctuation range.

[0038] In amplitude stability feature comparison, amplitude refers to the magnitude of a complex sampling point, which reflects the signal strength at that sampling point. It is calculated as |x|. ref [n]∣. Amplitude stability describes whether the amplitude exhibits normal fluctuation levels within a time window, avoiding significant gain jitter, sudden pulse interference, or receiver link saturation. This embodiment calculates the mean amplitude within the current time window and compares it with the allowable fluctuation range of the mean amplitude determined by historical robust sequences in the database. If the current mean amplitude does not fall within the allowable fluctuation range, the amplitude stability characteristic is considered inconsistent; otherwise, it is considered consistent.

[0039] In phase continuity feature alignment, phase refers to the angle of the complex sampling points in the complex plane, which reflects the phase state of the signal. It is calculated as arg(x) ref [n]). Phase continuity is used to describe whether the phase changes smoothly over time, avoiding anomalies such as phase jumps or significant increases in phase noise. In this embodiment, the variance of Δϕ[n] is calculated and compared with the allowable fluctuation range of the variance of Δϕ[n] determined by historical robust sequences in the database. If the variance of Δϕ[n] does not belong to the allowable fluctuation range of the variance of Δϕ[n], the phase continuity characteristics are determined to be inconsistent; otherwise, the phase continuity characteristics are determined to be consistent.

[0040] In the frequency consistency feature comparison, Δf is compared with the allowable fluctuation range of Δf determined by historical robust sequences in the database. If Δf does not belong to the allowable fluctuation range of Δf, the frequency consistency features are determined to be inconsistent; otherwise, the frequency consistency features are determined to be consistent.

[0041] In peak shape coherence feature comparison, peak shape refers to the main peak response formed on the delay axis after the correlation degree calculation of the complex sampling sequence of the reference receiving channel and the historical robust complex sampling sequence. This embodiment uses a cross-correlation algorithm to calculate the two sequences. Specifically, within the current time window, for multiple preset delay values, the complex sampling sequence of the reference receiving channel and the historical robust complex sampling sequence aligned to the corresponding delay are multiplied by their complex conjugates and accumulated to obtain a correlation degree sequence that varies with delay. Then, the amplitude of this correlation degree sequence is taken to obtain a correlation degree amplitude sequence, and the point with the largest amplitude is searched as the main peak point. The amplitude value at this main peak point is extracted as the main peak value. The database pre-stores the allowable fluctuation range of the main peak value determined by the historical robust complex sampling sequence. The current main peak value is compared with the allowable fluctuation range of the main peak value. If the main peak value does not belong to the allowable fluctuation range of the main peak value, the peak shape coherence feature is determined to be inconsistent; otherwise, the peak shape coherence feature is determined to be consistent.

[0042] The results of determining whether the above four types of features are consistent or inconsistent are given, and the number of inconsistent features is counted.

[0043] If the number of inconsistent features in the statistical feature comparison results is not higher than the first defined number, then the current background reference value and the allowable fluctuation range of the background reference value are maintained, and the anomaly count is reduced by one, and the anomaly count is not less than zero.

[0044] The first and second definition quantities are preset judgment thresholds in the database, used to classify the status of the reference receiving channel into three categories: basically stable, slightly abnormal, and severely abnormal. The anomaly count is a counting variable used to reflect the persistence of the anomaly, used to avoid triggering frequent adjustments of background parameters by a single, occasional fluctuation; the definition anomaly count is the trigger threshold for the anomaly count, used to determine whether the anomaly has reached the level that requires adjustment of background parameters, and the first definition quantity is less than the second definition quantity.

[0045] When the number of inconsistent features is not higher than the first defined number, it is considered that the current reference receiving channel can still stably characterize the background state. Therefore, the current background reference value and the allowable fluctuation range of the background reference value remain unchanged, and the anomaly count is decremented by one and is not less than zero. The background reference value is the background baseline intensity level used for background normalization of the delay and frequency shift map, representing the normal level of pixel values ​​in the delay and frequency shift map under conditions of no target disturbance or weak disturbance. The allowable fluctuation range of the background reference value is the range of natural fluctuations allowed around the background reference value, used to characterize the normal fluctuation scale of the background under time-varying and noise conditions, so that normal fluctuations and abnormal deviations can be distinguished when converting pixel values ​​into deviation levels.

[0046] When the number of inconsistent features is higher than the first threshold but lower than the second threshold, a slight anomaly is considered to have occurred in the reference receiving channel. The anomaly count is then incremented and compared with the threshold anomaly count. If the anomaly count is greater than the threshold anomaly count, the anomaly is considered persistent. Based on the feature comparison results, the background reference value is increased and the allowable fluctuation range of the background reference value is expanded, and the anomaly count is reset to zero. If the anomaly count is not greater than the threshold anomaly count, the anomaly is considered insufficient to change the background model, and the current background reference value and allowable fluctuation range remain unchanged. The meaning of increasing the background reference value and expanding the allowable fluctuation range is that when the reference channel background is no longer sufficiently quiet, increasing the background benchmark and widening the normal fluctuation range can reduce the sensitivity to occasional background disturbances during subsequent delay and frequency shift map normalization processes, avoiding misjudging background fluctuations as target correlation peaks.

[0047] Specifically, increasing the background reference value and expanding the allowable fluctuation range of the background reference value based on the feature comparison results refers to querying the corresponding background reference value increment from the inconsistency feature quantity-background reference value increment mapping table in the database based on the number of inconsistent features, and accumulating it with the current background reference value to increase the background reference value. At the same time, querying the corresponding background reference value allowable fluctuation range change from the inconsistency feature quantity-background reference value allowable fluctuation range change amount mapping table stored in the database based on the number of inconsistent features, subtracting the background reference value allowable fluctuation range change amount from the minimum value of the background reference value allowable fluctuation range, and adding the background reference value allowable fluctuation range change amount to the maximum value of the background reference value allowable fluctuation range, thereby expanding the allowable fluctuation range of the background reference value.

[0048] When the number of inconsistent features is not less than the second defined number, the reference receiving channel is considered to have a serious anomaly risk. In this embodiment, the reference receiving channel is anomaly detected and marked as unusable. Unusable means that the reference receiving channel will no longer be used as a background benchmark and normalization reference within the current time window, in order to avoid introducing systematic deviations by allowing abnormal reference sequences to enter the subsequent delay and frequency shift map generation and deviation calculation processes. Through the above mechanism, this embodiment can progressively and adaptively adjust abnormal backgrounds while maintaining the stability of background reference values, and promptly isolate reference channels in the event of serious anomalies, thereby providing a reliable reference basis for subsequent background normalization and candidate peak extraction of delay and frequency shift maps.

[0049] The delay-frequency shift combination set is extracted from the database. The delay-frequency shift combination set refers to a set of discrete Cartesian combinations of delay and frequency shift hypotheses. It is used to exhaustively test in the computer which delays and frequency shifts better match the complex sampling sequence of the current monitoring receiving channel and the complex sampling sequence of the reference receiving channel.

[0050] The processor pre-determines a set of delay candidates, where each delay τ represents the delay used to align the complex sampling sequence of the monitoring and receiving channel; it also pre-determines a set of frequency shift candidates, where each frequency shift f represents the frequency shift used to compensate the frequency of the complex sampling sequence of the monitoring and receiving channel. The set of delay-frequency shift combinations is all combinations of (τ, f), and each combination corresponds to a grid point in the delay-frequency shift map. First, x... mon [n] multiplied by e point by point -j×2×π×f×n×T This means that the phase is rotated in the opposite direction according to the frequency shift assumption f, thereby canceling out the phase rotation under this frequency shift assumption. After the correction is completed, x is updated. mon [n], then with x ref [n] Perform correlation analysis, C(τ, f) is a complex value representing the degree of matching, or correlation, between two sequences when the delay is τ and the frequency shift is f. ∗ represents complex conjugation, and n−τ represents the alignment sampling of the complex sampled sequence of the reference receiving channel according to the delay assumption.

[0051] The complex sampling sequence of the reference receiving channel can be considered as the time-frequency reference and matching template of the direct signal. The complex sampling sequence of the monitoring receiving channel contains the component to be detected and may have a phase that rotates continuously over time due to frequency shift. To avoid energy cancellation during correlation accumulation due to this phase rotation, the monitoring sequence is rotated in reverse point by point for each frequency shift assumption to cancel the phase rotation under that frequency shift assumption. Subsequently, the reference sequence is shifted accordingly under the delay assumption, and the similarity between the two sequences under this combination of delay and frequency shift is cumulatively evaluated by using cross-correlation operation of complex conjugate multiplication and summation. If the delay and frequency shift assumptions are consistent with the actual path parameters, the product terms of the two sequences under the conditions of time alignment and phase alignment have strong consistency in the complex plane, resulting in a large correlation response after accumulation. Conversely, the phase difference causes the accumulated terms to cancel each other out, resulting in a smaller response.

[0052] Based on the correlation of each delay frequency shift combination, the pixel values ​​corresponding to each delay frequency shift combination are matched from the correlation-pixel value mapping table in the database.

[0053] Using delay as the first dimension and frequency shift as the second dimension, the pixel values ​​corresponding to each delay-frequency shift combination are filled into the corresponding two-dimensional grid positions to form a delay-frequency shift heatmap.

[0054] Figure 3This is a schematic diagram of a delay-frequency shift heatmap provided in an embodiment of the present invention. The horizontal axis represents the delay axis, indicating the delay enumerated when generating the heatmap, i.e., the displacement by which the monitoring receiving channel sequence is aligned forward or backward relative to the reference receiving channel sequence. The scale on the horizontal axis can be understood as the number of candidate delay sampling points, which can be converted into time delay according to the sampling interval. The conversion relationship is that the time delay equals the number of sampling points multiplied by the sampling interval. In engineering, microseconds are commonly used as the unit of time. The vertical axis represents the frequency shift axis, indicating the frequency shift enumerated when generating the heatmap, i.e., the possible frequency offset of the monitoring sequence relative to the reference sequence, in Hertz (Hz). Each grid point in the figure corresponds to a set of delay and frequency shift assumptions. The pixel value of the grid point is the pixel value matched by the correlation degree obtained by aligning and calculating under the set of assumptions. The peak size of the color bar is used to represent the size of the pixel value. The brighter the color, the higher the correlation degree under the combination of delay and frequency shift. The brighter clump-like areas in the figure correspond to possible target perturbation peaks, while the weak bright lateral bands near zero frequency shift usually correspond to the concentrated manifestation of relatively stable background components or residual direct components in the frequency shift dimension.

[0055] The pixel values ​​of each grid point in the delay-frequency shift heatmap are compared with the background reference value. Then, the range width of the allowable fluctuation range of the background reference value is used as the normalization scale. The difference between the pixel value of each grid point and the background reference value is divided by the range width to obtain the background deviation value of each grid point. Here, the range width refers to the maximum value of the allowable fluctuation range of the background reference value minus the minimum value of the allowable fluctuation range of the background reference value. The background deviation value refers to the normalized result of the degree of deviation of the pixel value of a certain grid point from the background reference value of that grid point on the allowable fluctuation range width, which is used to characterize the degree to which the pixel value exceeds or approaches the normal fluctuation range of the background.

[0056] By establishing a deviation map under the same delay and frequency shift coordinate axes, the coordinate definitions and grid resolutions are kept consistent, thus ensuring that each grid point can correspond one-to-one.

[0057] Figure 4 This is a schematic diagram of the deviation degree provided in the embodiment of the present invention. Its horizontal axis is still the delay axis, and its vertical axis is still the frequency shift axis. The meanings of the two are the same as those of the present invention. Figure 3 Consistent, meaning that the position of each pixel is consistent with Figure 3 The delay coordinates and frequency shift coordinates in the graph correspond one-to-one, and use the same sampling point scale and Hertz scale. The difference lies in the physical meaning of the pixel values: the pixel values ​​in the deviation map are not the direct correlation magnitude, but rather... Figure 1The background deviation value is obtained by subtracting the pixel value at the same coordinate position from the background reference value, and then normalizing it using the width of the allowable fluctuation range of the background reference value as the normalization scale. The range width is the maximum value minus the minimum value of the allowable fluctuation range. If a negative value appears after normalization, it is set to zero. Therefore, the degree of deviation of the color bar indicates how much the correlation at that coordinate position is higher than the robust background by the width of the allowable fluctuation range; the larger the value, the more significant the deviation. Through this transformation that keeps the coordinates unchanged but changes the meaning of the values, the stable background can be suppressed without changing the correspondence between the delay coordinates and the frequency shift coordinates, making the peaks that truly exceed the background fluctuations more prominent, thus providing a more stable input for subsequent candidate peak point detection and screening in the deviation degree map.

[0058] Subsequently, the background deviation values ​​of each grid point calculated above are filled into the same pixel position on the deviation map according to their corresponding delay coordinates and frequency shift coordinates, so that the deviation map directly represents the degree of deviation under that delay-frequency shift combination at each pixel. To avoid interference with subsequent peak detection when the background deviation value is negative, the negative values ​​in the deviation map are uniformly set to zero to suppress grid points below the background reference level.

[0059] To identify prominent response points potentially corresponding to target disturbances from the deviation map, local maximum detection is performed. This involves taking a neighboring window centered on each grid point. The neighboring window is a local area defined by a predetermined number of grid points in both the delay and frequency shift directions, used to evaluate whether the grid point exhibits a peak-like prominence within this local area. The central grid point is judged to meet two conditions: its background deviation value is not less than the background deviation values ​​of all other grid points within the neighboring window, and it is greater than the background deviation value of at least one grid point within the neighboring window. Central grid points meeting these conditions are marked as candidate peak points for the UAV. Further, they are required to be strictly greater than the background deviation value of at least one grid point within the neighboring window. This is to exclude flat areas where all grid points in the neighboring window have identical values ​​or uniformly raised areas caused by pure noise, avoiding misclassification of flat areas without peak characteristics as candidate peak points. Specifically, a candidate peak point for the UAV refers to a grid point exhibiting a local peak-like prominence in the deviation map, representing a possible delay-frequency shift combination corresponding to a target disturbance.

[0060] After selecting candidate peaks for each UAV from the deviation map, a candidate threshold is further introduced to eliminate weak peaks caused by noise fluctuations or background residue. The candidate threshold is a value determined by statistically analyzing the high quantile of the background deviation values ​​of all grid points in the deviation map. It serves as a screening criterion for distinguishing between weak and significant deviation points. The high quantile refers to the value that is located at the percentile position preset by relevant technical personnel after sorting the background deviation values ​​from small to large. It is used to characterize the typical level of large deviation within the current time window.

[0061] The ratio of the number of inconsistent features to the maximum allowed number of inconsistent features stored in the database is used to refine the candidate threshold. When the stability of the reference receiving channel decreases, the background deviation value in the deviation map is more easily amplified by non-target disturbances. Using a fixed candidate threshold would increase the risk of false detections. By comparing the number of inconsistent features with their maximum allowed value, a normalized coefficient reflecting the current anomaly level of the reference channel can be obtained. This coefficient is then multiplied by the candidate threshold, causing the candidate threshold to adaptively increase with the anomaly level. This tightens the candidate peak admission criteria when the reference channel is subjected to increased disturbances, reducing the number of false peaks caused by background fluctuations entering the subsequent deep screening process.

[0062] Several candidate peaks of UAVs with background deviation values ​​less than the candidate threshold are screened out, thereby updating each candidate peak of UAVs and extracting high-confidence candidate peaks of each UAV from each candidate peak.

[0063] A depth discrimination analysis was performed on the aforementioned candidate peak points for each UAV, and the deviation map has already provided several candidate peak points. Each candidate peak point corresponds to a set of delays and frequency shifts. The meaning of this set of parameters is that when the complex sampling sequence of the monitoring receiving channel is phase-inversely compensated according to the frequency shift and aligned with the complex sampling sequence of the reference receiving channel according to the delay, the coherence superposition of the two sequences reaches a local maximum. Therefore, the differential phase characteristics, coherence attenuation characteristics, and spectral energy transfer characteristics are not arbitrarily extracted from the monitoring receiving channel, but are extracted from the aligned coherent calculation results and their neighborhood using the delay and frequency shift corresponding to the UAV candidate peak point as an index. These characteristics are used to characterize the degree of perturbation of the candidate peak point relative to the historical robust background state.

[0064] Differential phase features are used to characterize the phase shift of a candidate peak point of a UAV relative to the robust background. For example, for the delay and frequency shift corresponding to a candidate peak point of a UAV, after frequency shift compensation and delay alignment of the complex sampling sequence of the monitoring receiving channel, a correlation calculation is performed with the complex sampling sequence of the reference receiving channel. The correlation calculation result is marked as the complex correlation output value at the candidate peak point. The phase angle of the complex correlation output value is used as the current phase measurement value. For example, if the complex correlation output is 0.80 plus 0.60 multiplied by the imaginary unit, its phase angle is approximately 0.64 radians. The phase reference value of the same delay and frequency shift position under historical robust conditions is read from the database, for example, 0.40 radians. The current phase measurement value is subtracted from the phase reference value to obtain the phase difference, for example, 0.24 radians. When the phase difference crosses the ±3.1416 radian boundary, the phase difference is phase normalized to fall within the agreed ±3.1416 radian boundary, thus forming the differential phase feature. Using the above method, the phase changes caused by environmental or target disturbances at the candidate peak points of the UAV can be extracted from the complex sampling sequence of the monitoring and receiving channel in a calculable numerical form, and used to subsequently determine the credibility of the candidate peak points.

[0065] The coherence attenuation characteristic is used to characterize the degree of consistency reduction between the complex sampling sequence of the reference receiving channel and the aligned complex sampling sequence of the monitoring receiving channel under the delay and frequency shift corresponding to the candidate peak of the UAV. For example, at the candidate peak of the UAV, the complex correlation output value is first obtained and its energy value is calculated. The energy value can be obtained by the sum of the squares of the in-phase component and the quadrature component of the complex correlation output. For example, if the complex correlation output is 0.80 plus 0.60 multiplied by the imaginary unit, then the energy value is the square of 0.80 plus the square of 0.60, which is 1.00. At the same time, the sequence energy of the reference receiving channel and the monitoring receiving channel are counted within the same time window and used to normalize the energy value. For example, if the reference sequence energy is 1.50 and the monitoring sequence energy is 1.60, multiplying 1.50 by 1.60, the current coherence can be taken as 1 divided by the product result, which is approximately 0.417. The database stores the baseline coherence value for the corresponding position under historical robust conditions, for example, 0.60. The coherence attenuation is obtained by subtracting the current coherence value from the baseline coherence value, for example, 0.183, and this attenuation is used as the coherence attenuation feature. In this way, the reduction in consistency near the candidate peak point of the UAV caused by clutter residue, multipath, or non-robust background can be numerically characterized, thereby improving the stability of candidate peak point selection.

[0066] The spectral domain energy migration feature is used to characterize whether the energy distribution in the neighborhood of a candidate peak point in the frequency shift direction is relatively robust and shifts or diffuses relative to the background. In this embodiment, the delay of the candidate peak point of the UAV is kept constant. Several frequency shift grid points are selected near the frequency shift of the candidate peak point, and the associated energy value at each frequency shift grid point is calculated. The energy centroid in the frequency shift direction is calculated using the energy value as a weight. For example, the energy values ​​at the three frequency shift grids of 300 Hz, 350 Hz, and 400 Hz are 0.20, 1.00, and 0.50, respectively. Then the energy centroid is the sum of 300 multiplied by 0.20, 350 multiplied by 1.00, and 400 multiplied by 0.50, divided by the sum of 0.20, 1.00, and 0.50, resulting in approximately 358.82 Hz. The database stores the centroid reference value at the corresponding position under historical robust states, for example, 340 Hz. Subtracting the centroid reference value from the current centroid yields the energy transfer, for example, 18.82 Hz, which is then used as the spectral domain energy transfer feature. In this way, the degree of shift in the frequency shift dimension energy distribution relative to the background near candidate peaks can be extracted in a calculable form. This is used to suppress false peak selection caused by sidelobes or background fluctuations and to provide a basis for subsequent high-confidence candidate peak determination.

[0067] Taking a candidate peak point of a certain UAV as an example, its differential phase feature is 0.24 radians, its coherence attenuation feature is 0.183, and its spectral energy transfer feature is 18.82 Hz. The reference perturbation feature vector corresponding to this delay and frequency shift position in the database can be taken as follows: differential phase feature 0.00 radians, coherence attenuation feature 0.00, and spectral energy transfer feature 0.00 Hz. The allowable fluctuation scales for each of the three types of features are given, for example, the allowable fluctuation scale for differential phase is 0.50 radians, for coherence attenuation is 0.30, and for spectral energy transfer is 50 Hz. The deviation degree comparison result is calculated for each of the three types of features of the current candidate peak point, that is, the absolute value of the difference between each feature and its reference value is divided by the corresponding allowable fluctuation scale to obtain the dimensionless deviation degree. For example, the differential phase deviation is 0.24 divided by 0.50, resulting in 0.48; the coherence attenuation deviation is 0.183 divided by 0.30, resulting in 0.61; and the spectral energy transfer deviation is 18.82 divided by 50, resulting in 0.38. This deviation comparison result characterizes the perturbation intensity of the current candidate peak point relative to the robust background. A smaller value indicates a closer approximation to the robust background, while a larger value indicates a more significant deviation.

[0068] To ensure that the degree of deviation can be directly converted into a confidence score suitable for ranking, the deviation comparison results are normalized and weighted. This embodiment normalizes the deviation by mapping it to a similarity score; that is, for each deviation, the deviation is subtracted by one, and the result is truncated to the range of 0 to 1, thus obtaining a normalized consistency score for each feature. A higher consistency score indicates a more robust match to the background features. In the example above, the differential phase consistency score is 1 minus 0.48, resulting in 0.52; the coherence attenuation consistency score is 1 minus 0.61, resulting in 0.39; and the spectral energy transfer consistency score is 1 minus 0.38, resulting in 0.62. Weights for the three feature classes are pre-defined in the database to reflect the contribution of different features to reliability; for example, the differential phase weight is 0.40, the coherence attenuation weight is 0.35, and the spectral energy transfer weight is 0.25. The three consistency scores are multiplied by their respective weights and summed; the reciprocal of the sum is the confidence score. For example, 0.52 multiplied by 0.40, plus 0.39 multiplied by 0.35, plus 0.62 multiplied by 0.25, yields approximately 0.50, with a confidence level of 1 / 0.5 = 2. Here, the confidence level characterizes the overall consistency between the candidate peak point of the UAV and the reference perturbation feature vector in terms of the three types of perturbation features. The higher the confidence level, the less likely the candidate peak point is to be caused by background anomalous broadening, sidelobe enhancement, or clutter residue, and the more suitable it is as an observation source for subsequent multi-station matching and positioning calculation.

[0069] Within the same time window, the confidence scores of all candidate peaks for UAVs are calculated using the method described above, and the confidence scores are sorted in descending order. A pre-defined limit is set in the database to restrict the number of peaks entering subsequent processing. This limit is a pre-set upper limit for retaining candidate peaks that can enter subsequent multi-station matching and location calculation stages. For example, if the limit is 3, the confidence scores and their delay and frequency shift coordinates of the top 3 candidate peaks after sorting are taken, and these 3 candidate peaks are marked as high-confidence candidate peaks for UAVs. If there are fewer than 3 candidate peaks, all candidate peaks are output as high-confidence candidate peaks for UAVs. Through the above normalized weighted aggregation and sorting filtering process, when multiple candidate peaks exist in the two-dimensional correlation output of delay and frequency shift, peaks that better conform to robust perturbation characteristics and have higher confidence levels are prioritized, thereby reducing observational bias caused by false peak selection and providing more stable input for subsequent multi-station matching and location calculation.

[0070] In this embodiment, three ground receiving stations are set up within the same time window: receiving station 1, receiving station 2, and receiving station 3. After completing peak selection and confidence ranking on the deviation map, each receiving station obtains a set of high-confidence candidate peak points for its UAV. High-confidence candidate peak points for UAVs refer to grid points on the deviation map that satisfy the local maximum condition and rank highly in confidence. Each peak point includes a delay parameter and a frequency shift parameter, where the delay parameter corresponds to the position on the delay coordinate axis, and the frequency shift parameter corresponds to the position on the frequency shift coordinate axis. To facilitate subsequent tracking, each receiving station assigns a receiving station identifier and a peak identifier to each peak point. The receiving station identifier is used to distinguish the source station, and the peak identifier is used to distinguish different peak points within the same station.

[0071] For example, within a certain time window, receiving station 1 obtains two high-confidence candidate peaks: peak identifier 1 has a delay parameter of 12.0 microseconds, a frequency shift of 350 Hz, and a confidence level of 0.86; peak identifier 2 has a delay parameter of 3.5 microseconds, a frequency shift of 120 Hz, and a confidence level of 0.81. Receiving station 2 obtains two high-confidence candidate peaks: peak identifier 1 has a delay parameter of 15.0 microseconds, a frequency shift of 348 Hz, and a confidence level of 0.82; peak identifier 2 has a delay parameter of 6.5 microseconds, a frequency shift of 123 Hz, and a confidence level of 0.79. Receiving station 3 obtains two high-confidence candidate peaks: peak identifier 1 has a delay parameter of 9.5 microseconds, a frequency shift of 352 Hz, and a confidence level of 0.84; peak identifier 2 has a delay parameter of 2.0 microseconds, a frequency shift of 118 Hz, and a confidence level of 0.77.

[0072] Then, pairwise comparisons are performed to form matching peaks. Matching peaks refer to a pair of high-confidence candidate peaks from two different receiving stations that are determined to be the same type of peak response corresponding to the same target under preset matching conditions. Matching conditions can be preset by the database or set by relevant technical personnel. For example, the frequency shift difference between the two peaks should not exceed 10 Hz, and the difference in delay parameters should fall within the allowable range. A confidence lower limit can be superimposed to avoid low-confidence peaks from participating in the matching. In the example above, the frequency shift difference between peak identifier 1 of receiving station 1 and peak identifier 1 of receiving station 2 is 350 minus 348, which equals 2 Hz, satisfying the frequency shift proximity condition, thus forming a pair of matching peaks. The receiving station identifier and peak identifier corresponding to the matching peak are recorded. The frequency shift difference between peak identifier 1 of receiving station 1 and peak identifier 1 of receiving station 3 is 350 minus 352, which equals -2 Hz, also satisfying the condition. The frequency shift difference between peak identifier 1 of receiving station 2 and peak identifier 1 of receiving station 3 is 348 minus 352, which equals -4 Hz, also satisfying the condition. Another set of matching peaks can be formed around the peaks near 120 Hz using the same rules.

[0073] After obtaining several pairs of matching peaks, target attribution is applied to these peaks, and they are aggregated to form target observation groups. Target attribution refers to assigning the same target number to matching peaks that consistently point to the same target; a target observation group is a set of matching peaks with consistent attribution labels and their derived observations, used for subsequent location calculations. Taking this embodiment as an example, three pairs of matching peaks with frequency shifts concentrated around 350 Hz are consistent and can be classified into the target observation group for target 1; several pairs of matching peaks with frequency shifts concentrated around 120 Hz are consistent and can be classified into the target observation group for target 2. This attribution process can be implemented using consistency rules, such as requiring peaks within the same target observation group to have frequency shifts within the same frequency shift band, and requiring that delay and frequency shift changes show a continuous trend within a continuous time window, thereby reducing the probability of different target peaks being mistakenly merged into the same group.

[0074] Within each target observation group, an observation set is further generated. A target observation group refers to a set of several pairs of matching peaks that have been identified as belonging to the same UAV through target attribution markers. A matching peak is a pair of high-confidence candidate peaks from two different ground receiving stations, which meet preset matching conditions and are considered to correspond to the same type of peak response of the same UAV within that time window. Each peak has delay and frequency shift parameters.

[0075] For each pair of matching peaks in the target observation group, the delay parameters of the two peaks in that pair are taken and subtracted to obtain the delay difference observable. The delay difference observable characterizes the time offset difference between the two ground receiving stations and the same UAV, facilitating its subsequent conversion into spatial geometric constraints. Then, the frequency shift parameters of the two peaks in that pair are subtracted to obtain the frequency difference observable. The frequency difference observable characterizes the frequency offset difference between the two ground receiving stations and the same UAV, and can be used to help suppress mismatches and improve solution stability. Furthermore, the delay difference observable is converted into an equivalent range difference using the speed of light constant. The equivalent range difference is the result of converting the time difference into a distance difference, and is used directly in position calculation.

[0076] For example, within the target observation group of target 1, there exists a pair of matched peaks formed by peak identifier 1 of receiving station 1 and peak identifier 1 of receiving station 2. The delay parameters of this pair of peaks are 12.0 microseconds and 15.0 microseconds respectively. Therefore, the delay difference observation is 12.0 microseconds minus 15.0 microseconds, resulting in -3.0 microseconds. The frequency shift parameters of this pair of peaks are 350 Hz and 348 Hz respectively. Therefore, the frequency difference observation is 350 Hz minus 348 Hz, resulting in 2 Hz. Converting the delay difference observation of -3.0 microseconds to seconds is -3.0 multiplied by 10 to the power of -6 seconds, and then multiplied by the speed of light constant 3 multiplied by 10 to the power of 8 meters per second, yielding an equivalent ranging difference of -900 meters for this pair of matched peaks. The negative sign only indicates the directionality of the difference, i.e., the sign of the result obtained when receiving station 1 is the minuend and receiving station 2 is the subtrahend, used to maintain a consistent expression of the observation under the same calculation convention.

[0077] Furthermore, to ensure that the impact of different observations on the location calculation matches their reliability, observation weights are assigned to each pair of matched peaks. These weights are derived from the confidence level of the matched peaks, where confidence level is a quantification of the likelihood that a peak is a real UAV peak; a higher value indicates greater reliability. In this embodiment, the confidence level of a matched peak can be the average of the confidence levels of the two peaks in the pair. For example, the average of 0.86 and 0.82 is 0.84, and this average is used as the weighting basis for that observation in subsequent calculations.

[0078] Following the same method, delay difference observations, frequency difference observations, and equivalent ranging differences are calculated for each of the other matching peak points within the target observation group for target 1. A corresponding observation weight is assigned to each observation. Finally, all observations and their weights within the target observation group for target 1 are summarized and organized to form the observation set for target 1. For target 2, the same rule is used to form the observation set for target 2. Subsequently, combining the low-Earth orbit satellite positions and the positions of various ground receiving stations, the observation set is input into the position calculation process to obtain the corresponding UAV's position result.

[0079] For any target observation group, the observation weight of each pair of matching peaks is retrieved from the confidence-observation weight mapping table in the database based on the confidence level of each pair of matching peaks in the target observation group. The observation weight is a non-negative number used to represent the reliability of the observation formed by the pair of matching peaks. The higher the confidence level, the larger the mapping observation weight is usually, thus exerting a stronger constraint on the positioning results in the subsequent residual summarization.

[0080] Subsequently, the central processing unit obtains the current spatial position of the low-Earth orbit satellite and the spatial positions of various ground receiving stations participating in the target observation group; the spatial positions can be obtained from ephemeris data and measurement data installed at the receiving stations. To ensure consistency in distance calculation, the central processing unit unifies the coordinates of the low-Earth orbit satellite spatial positions and the spatial positions of various ground receiving stations, ensuring that all three are in the same coordinate system and use the same unit of length.

[0081] After coordinate unification, the central processing unit uses the UAV spatial position corresponding to the target observation group as the unknown to be solved, and constructs a geometric model for calculating the reference equivalent range difference. Let the UAV spatial position be vector p, the low-orbit satellite spatial position be vector s, and the i-th ground receiving station spatial position be vector r. i All three are three-dimensional coordinate vectors, with units of meters. For any ground receiving station i, where i represents the station number (i = 1, 2, 3, ..., k), and k is the overall station number, the reference two-way distance L for that station is defined. i (p) = ||ps|| + ||p - rᵢ|| − ||s - rᵢ||, where ||·|| denotes the Euclidean distance operation, L i (p) represents the equivalent distance measured by the i-th ground receiving station when the UAV is assumed to be located at p. For any pair of matching peaks in the target observation group, let the two corresponding receiving stations be i and j, where i and j have the same meaning. Then, the observation equivalent distance difference obtained from this pair of matching peaks is denoted as D. ij The unit is meters; assuming the UAV is located at p, the reference equivalent ranging difference between the receiving stations is denoted as C_D. i,j (p)=L i (p)−L j (p), from which the ranging residual e of the pair of matching peaks is obtained. i,j (p)=C_D i,j (p)−D i,j The ranging residual is used to characterize the degree of inconsistency between the predicted observation and the actual observation at candidate position p. It is assumed that all matching peak points form a set M={(i,j)}.

[0082] Furthermore, the central processing unit constructs a weighted residual cost function using all matching peak points within the target observation group as residual terms, and uses the minimum of this cost function as the localization solution objective. Let w be the observation weight of the pair of matching peak points obtained from the database mapping. i,j Then the weighted residual cost function is Σ represents the summation of all receiving station pairs corresponding to matching peak points within the target observation group. The physical meaning of the weighted residual cost function J(p) is: at the candidate position p, the weighted sum of squares of the observation residuals of each station pair; the smaller the value, the better the consistency interpretation of the multi-station observations at that position. The central processing unit can use an iterative method to search for the UAV's spatial position. For example, using the UAV position of the previous time window as the initial value p0, the position estimate is updated according to the direction of change of the cost function in each iteration until the decrease of the cost function is less than the minimum allowable decrease value preset in the database or the number of iterations reaches the upper limit, thus obtaining the UAV position solution for the current time window. If there are multiple target observation groups in the same time window, a corresponding set of equivalent range difference, set of observation weights, and cost function are constructed for each target observation group, and solved iteratively to output the positioning results of multiple UAVs.

[0083] A second aspect of this invention provides a passive anti-drone positioning system based on a low-Earth orbit satellite external radiation source. Figure 2 This is a schematic diagram of the structure of a passive anti-drone positioning system based on low-orbit satellite external radiation sources, including a sequence correction module, a feature comparison module, a peak point analysis module, a location positioning module, and a database.

[0084] The sequence correction module is connected to the feature comparison module, the feature comparison module is connected to the peak point analysis module, the peak point analysis module is connected to the location module, and the sequence correction module, feature comparison module, peak point analysis module, and location module are all connected to the database.

[0085] The database is used to store the parameters involved in the passive anti-drone positioning system based on low-Earth orbit satellite external radiation source. In view of the application requirements of the passive anti-drone positioning system based on low-Earth orbit satellite external radiation source, an appropriate database storage scheme is formulated to standardize the storage and management of various parameters involved in the entire process of the system.

[0086] The sequence correction module is used to synchronously sample the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel for each ground receiving station, using the external radiation source of the low-orbit satellite as the illumination signal source under the same time reference, and to perform frequency correction and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel.

[0087] The feature comparison module is used to compare the complex sampling sequence of the reference receiving channel with the historical robust complex sampling sequence to obtain the feature comparison result. Based on the feature comparison result, the background reference value used for delay-frequency shift map background normalization and the allowable fluctuation range of the background reference value are updated.

[0088] The peak point analysis module is used to obtain the delay-frequency shift combination set. Under each delay-frequency shift combination in the delay-frequency shift combination set, the complex sampling sequence of the monitoring receiving channel is corrected, and the correlation analysis is performed with the complex sampling sequence of the reference receiving channel. The delay-frequency shift map is generated through the correlation analysis results. Based on the delay-frequency shift map, the background reference value, and the allowable fluctuation range of the background reference value, the high confidence candidate peak points of each UAV are analyzed.

[0089] The location module is used to statistically identify high-confidence candidate peaks for each UAV at various ground receiving stations. Combining the low-orbit satellite position with the position of each ground receiving station, a weighted residual cost function is established to evaluate the consistency of UAV position observations from multiple ground receiving stations. The module is then iteratively optimized to minimize the weighted residual cost function, thereby solving for the position of each UAV and achieving passive anti-UAV positioning.

[0090] The delay and frequency shift ranges vary with changes in the low-Earth orbit satellite position, the protection sector setting, sampling rate, and time window. Relying solely on pre-set combinations in the database often requires storing a large number of grids for various parameter combinations, resulting in high maintenance costs and difficulty in timely adaptation. Therefore, Example 2 is proposed. Building upon Example 1, the delay and frequency shift combination sets are not directly extracted from the database but are generated online according to the physical and sampling constraints of the current scenario. First, the delay search range is determined. Since delay characterizes the additional propagation time of the UAV path relative to the direct path, this additional path length has an upper bound within the low-altitude protection sector. The system can provide a conservative upper bound in three-dimensional space based on the maximum horizontal radius and maximum height of the protection sector and convert this upper bound into the maximum delay using the speed of light. For example, if the maximum horizontal radius of the protection sector is taken as 10,000 meters, the upper bound of the additional path length can be taken as approximately 20,000 meters, corresponding to a maximum delay of approximately 20,000 meters divided by 3 × 10⁸ meters per second, which is approximately 66.7 microseconds. Therefore, the delay candidates can be set from 0 microseconds to 70 microseconds, and converted into discrete delay grid points according to the sampling rate. For example, when the sampling rate is 2MHz, one sampling point corresponds to 0.5 microseconds, and the delay grid points can be integer multiples of 0.5 microseconds, thus obtaining the delay candidate set.

[0091] Next, the frequency shift search range is determined. Frequency shift characterizes the Doppler effect caused by the UAV's motion and the frequency offset exhibited within the time window. Its range can be conservatively bounded by the wavelength corresponding to the UAV's maximum speed and downlink center frequency, and can be supplemented with the allowable range of the receiver link's residual frequency offset as a safety margin. For example, if the maximum speed is 40 meters per second, and the downlink center frequency is given by the system configuration and can be converted to wavelength, the system calculates the frequency shift upper bound and rounds it to a range suitable for gridding, for example, from -600Hz to +600Hz. The interval of the frequency shift grid points can be determined according to the time window length. For example, if the time window length is 0.02s, the base resolution can be 50Hz, and the frequency shift grid points can be integer multiples of 50Hz, thus obtaining the frequency shift candidate set.

[0092] After obtaining the candidate sets for delay and frequency shift, the combination set is the Cartesian product of the two. That is, each delay grid point and each frequency shift grid point form a delay-frequency shift combination, and these combinations are directly mapped one-to-one to the two-dimensional grid coordinates, following the rule that delay is the first dimension and frequency shift is the second dimension. When traversing the combination set to calculate the correlation degree, the system writes the correlation degree value of the same combination into the grid position corresponding to that combination, thereby generating a delay-frequency shift heatmap.

[0093] Online generation based on physical and sampling constraints can ensure that the search area covers the physically reachable delay and frequency shift range in the current scene, while avoiding the computational burden caused by invalid combinations. This reduces the probability of false peaks entering the candidate set from the source and improves the stability of subsequent peak selection and localization solutions.

[0094] The above description is only an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A passive anti-drone positioning method based on low-orbit satellite external radiation sources, characterized in that, The method includes: Step 1: Under the same time reference, using the external radiation source of the low-orbit satellite as the illumination signal source, for each ground receiving station, synchronously sample the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel, and perform frequency correction and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel. Step 2: Perform feature comparison between the complex sampling sequence of the reference receiving channel and the historical robust complex sampling sequence to obtain the feature comparison results. Based on the feature comparison results, update the background reference value used for delay-frequency shift map background normalization and the allowable fluctuation range of the background reference value. Step 3: Obtain the delay-frequency shift combination set. Under each delay-frequency shift combination in the delay-frequency shift combination set, correct the complex sampling sequence of the monitoring receiving channel and perform correlation analysis with the complex sampling sequence of the reference receiving channel. Generate a delay-frequency shift map based on the correlation analysis results. Analyze the high-confidence candidate peak points of each UAV based on the delay-frequency shift map, background reference value, and allowable fluctuation range of the background reference value. Step 4: Statistically identify the high-confidence candidate peaks of each UAV at each ground receiving station. Combine the low-orbit satellite position with the position of each ground receiving station to establish a weighted residual cost function for evaluating the consistency of UAV position observations from multiple ground receiving stations. Iteratively optimize the function to minimize the weighted residual cost function, thereby solving for the position of each UAV and achieving passive anti-UAV positioning.

2. The passive anti-drone positioning method based on low-orbit satellite external radiation sources as described in claim 1, characterized in that, The frequency and phase corrections are performed on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel. The specific correction process is as follows: Based on the complex sampling sequence of the reference receiving channel, the residual frequency deviation and phase drift of the sequence at the corresponding ground receiving station are analyzed, and the frequency compensation and phase compensation are generated. Frequency correction is performed on the complex sampling sequence of the reference receiving channel based on the frequency compensation amount; Phase correction is performed on the complex sampling sequence of the reference receiving channel based on the phase compensation amount; After calibration, update the complex sampling sequence of the reference receiving channel.

3. The passive anti-drone positioning method based on low-orbit satellite external radiation sources as described in claim 1, characterized in that, The feature comparison of the complex sampling sequence of the reference receiving channel with the historical robust complex sampling sequence specifically refers to: The feature comparison includes amplitude stability feature comparison, phase continuity feature comparison, frequency consistency feature comparison, and peak shape coherence feature comparison. If the number of inconsistent features in the statistical feature comparison results is not higher than the first defined number, then the current background reference value and the allowable fluctuation range of the background reference value are maintained, and the anomaly count is reduced by one, and the anomaly count is not less than zero. If the number of inconsistent features in the feature comparison results is higher than the first limit number but lower than the second limit number, the anomaly count is incremented by one and the anomaly count is collected. If the anomaly count is greater than the limit anomaly count, the background reference value is increased and the allowable fluctuation range of the background reference value is expanded based on the number of inconsistent features, and the anomaly count is cleared to zero. If the anomaly count is not greater than the limit anomaly count, the current background reference value and the allowable fluctuation range of the background reference value are maintained. If the number of inconsistent features in the feature comparison results is not less than the second defined number, then anomaly detection is performed on the reference receiving channel, and the reference receiving channel is marked as unusable.

4. The passive anti-drone positioning method based on low-orbit satellite external radiation sources as described in claim 1, characterized in that, The process of generating the delay-frequency shift map based on the correlation analysis results is as follows: The complex sampling sequence of the monitoring receiving channel is corrected by each delay frequency shift combination in the delay frequency shift combination set. After correction, the correlation with the complex sampling sequence of the reference receiving channel is analyzed and marked as the correlation of each delay frequency shift combination. Based on the correlation of each delay frequency shift combination, the pixel value corresponding to each delay frequency shift combination is matched; Using delay as the first dimension and frequency shift as the second dimension, the pixel values ​​corresponding to each delay-frequency shift combination are filled into the corresponding two-dimensional grid positions to form a delay-frequency shift heatmap.

5. Convert the delay-frequency shift heatmap into a deviation map.

6. The passive anti-drone positioning method based on low-orbit satellite external radiation sources as described in claim 4, characterized in that, The specific process of converting the delay-frequency shift heatmap into a deviation map is as follows: The pixel values ​​of each grid point in the delay-frequency shift heatmap are compared with the background reference value. The result is then compared with the width of the allowable fluctuation range of the background reference value, and finally marked as the background deviation value of each grid point. A deviation map is established using the same delay and frequency shift coordinates as the delay-frequency shift heatmap, and the background deviation values ​​of each grid point are filled into the same positions of the deviation map according to the coordinates. Set the negative values ​​in the deviation graph to zero; Candidate peak points for each drone were initially selected from the deviation map.

7. The passive anti-UAV positioning method based on low-orbit satellite external radiation sources as described in claim 5, characterized in that, The preliminary screening of candidate peak points for each UAV from the deviation map is as follows: Local maxima detection is performed on the deviation map. For each grid point in the deviation map, an adjacent neighborhood window is taken, and it is determined whether the grid point simultaneously meets the following conditions: the background deviation value of the grid point is greater than or equal to the background deviation values ​​of all other grid points in the neighborhood window, and is greater than the background deviation value of at least one grid point in the neighborhood window. If both conditions are met, then the grid point is marked as a candidate peak point for the UAV; Candidate peaks for each UAV are selected from the deviation map, and the high quantiles in the deviation map are marked as candidate thresholds. The candidate thresholds are then corrected based on the number of inconsistent features. Several candidate peaks of UAVs with background deviation values ​​less than the candidate threshold are filtered out, thereby updating each candidate peak of UAVs. High-confidence candidate peaks for each UAV are extracted from the candidate peaks of each UAV.

8. The passive anti-drone positioning method based on low-orbit satellite external radiation source as described in claim 6, characterized in that, The specific extraction process for extracting high-confidence candidate peaks from each UAV candidate peak is as follows: The differential phase characteristics, coherence attenuation characteristics, and spectral energy transfer characteristics of each UAV candidate peak point are obtained from the complex sampling sequence of the monitoring and receiving channel, and the perturbation feature vector of each UAV candidate peak point is constructed. The deviation of the perturbation feature vector of each candidate peak point of the UAV is compared with the reference perturbation feature vector; Obtain the deviation comparison results, normalize and weight the deviation comparison results to aggregate them, and thus obtain the confidence level of each candidate peak point of the UAV; Sort the confidence levels in descending order, extract the candidate peak points of each UAV corresponding to a predetermined number of confidence levels, and mark them as high-confidence candidate peak points for each UAV.

9. The passive anti-UAV positioning method based on low-orbit satellite external radiation sources as described in claim 1, characterized in that, The process of statistically analyzing the high-confidence candidate peaks of each UAV at various ground receiving stations also includes: For several high-confidence candidate peak points of UAVs at different ground receiving stations, pairwise comparison and screening are performed. Peak pairs that meet the matching conditions are retained and marked as matching peak points. The receiving station identifier and peak identifier corresponding to each pair of matching peak points are clearly defined. Each pair of matching peaks is marked with a target affiliation label. Matching peaks with consistent affiliation labels are aggregated to obtain several pairs of matching peaks belonging to the same UAV, which are then marked as the target observation group. For each pair of matching peaks in the target observation group, obtain the delay parameters corresponding to the two high-confidence candidate peaks of the UAVs in the matching peaks. Subtract the delay parameter of the other high-confidence candidate peak from the delay parameter of one UAV high-confidence candidate peak to calculate the delay difference measurement corresponding to the matching peak. Obtain the frequency shift parameters of the two corresponding high-confidence candidate peaks of the UAV, and calculate the frequency difference observation corresponding to the matching peak by subtracting the frequency shift parameter of the other high-confidence peak from the frequency shift parameter of one high-confidence peak. By multiplying the constant speed of light with the delay difference observable corresponding to each pair of matching peaks, the delay difference observable is converted into an equivalent range difference. The confidence scores of two high-confidence candidate peaks of UAVs in the matching peaks are averaged, and the result is marked as the confidence score of the matching peak. The delay difference observations, frequency difference observations, and equivalent distance difference corresponding to all matching peak points in the target observation group are summarized and organized to form a complete set of observations for the target observation group.

10. The passive anti-drone positioning method based on low-orbit satellite external radiation source as described in claim 1, characterized in that, The weighted residual cost function for evaluating the consistency of UAV positions observed by multiple ground receiving stations is established as follows: For any target observation group, the observation weights of each pair of matching peaks are determined by the confidence level of each pair of matching peaks in the target observation group. Obtain the current spatial position of low-Earth orbit satellites and the spatial positions of ground receiving stations in various locations; The coordinates of the spatial positions of low-orbit satellites and the spatial positions of ground receiving stations are unified so that they are in the same coordinate system; Using the spatial position of the UAV in the target observation group as the unknown to be solved, for the equivalent ranging difference of each pair of matching peak points in the target observation group, the corresponding reference equivalent ranging difference is obtained, and the difference is processed to calculate the ranging residual of each pair of matching peak points. Based on the observation weights of each pair of matching peaks, the ranging residuals of each pair of matching peaks are weighted and summarized to form a weighted residual cost function. An iterative solution method is used to search for the UAV's spatial location, minimizing the weighted residual cost function to obtain the current UAV position solution.

11. A passive anti-drone positioning system based on low-orbit satellite external radiation sources, characterized in that, The system includes: The sequence correction module is used to synchronously sample the complex sampling sequences corresponding to the reference receiving channel and the monitoring receiving channel for each ground receiving station, using the external radiation source of the low-orbit satellite as the illumination signal source under the same time reference, and to perform frequency correction and phase correction on the complex sampling sequences of the reference receiving channel and the monitoring receiving channel. The feature comparison module is used to compare the complex sampling sequence of the reference receiving channel with the historical robust complex sampling sequence, obtain the feature comparison result, and update the background reference value and the allowable fluctuation range of the background reference value for background normalization of the delay-frequency shift map based on the feature comparison result. The peak point analysis module is used to obtain the delay-frequency shift combination set. Under each delay-frequency shift combination in the delay-frequency shift combination set, the complex sampling sequence of the monitoring receiving channel is corrected, and the correlation analysis is performed with the complex sampling sequence of the reference receiving channel. The delay-frequency shift map is generated through the correlation analysis results. Based on the delay-frequency shift map, the background reference value, and the allowable fluctuation range of the background reference value, the high confidence candidate peak points of each UAV are analyzed. The location module is used to statistically identify high-confidence candidate peaks for each UAV at various ground receiving stations. Combining the low-orbit satellite position with the position of each ground receiving station, a weighted residual cost function is established to evaluate the consistency of UAV position observations from multiple ground receiving stations. The module is then iteratively optimized to minimize the weighted residual cost function, thereby solving for the position of each UAV and achieving passive anti-UAV positioning.

Images