A Dual-Channel UAV RF Direction Finding and Tracking Method and System Based on Angle Posterior Feedback Scanning

Abstract

This invention discloses a dual-channel UAV radio frequency direction finding and tracking method and system based on angle posterior feedback scanning, employing a dual-ring dual-channel architecture with a fixed reference array element and multiple scanning array elements. Through adaptive acquisition of long and short time slots, representative FFT block selection, and reference normalized cross-spectrum suppression of power perturbations, a three-dimensional observation quality matrix is ​​constructed and weighted to generate a broadband angle spectrum. The tracking status is determined using posterior entropy, main-side peak interval, and average coherence, and the scanning sequence is dynamically generated based on the array element discrimination capability calculated by the angle posterior. This invention significantly reduces the tracking computation load, improves phase observation stability and anti-interference capability, and achieves continuous high-precision radio frequency direction finding and tracking of UAVs at low cost.

Description

Technical Field

[0001] This invention relates to the fields of radio detection, array signal processing and UAV radio frequency direction finding technology, and discloses a dual-channel UAV radio frequency direction finding and tracking method and system based on angle posterior feedback scanning. Background Technology

[0002] With the widespread application of low-altitude drones, passive direction finding technology for drone image transmission and remote control radio frequency signals has become an important technical means in the fields of low-altitude security and drone monitoring. Currently, drone radio frequency direction finding is mainly based on array signal processing. Depending on the number of receiving channels and the acquisition method, it can be divided into several technical routes, including multi-channel synchronous array direction finding, single-channel switch-scan direction finding, and dual-channel switch-scan direction finding with a reference channel. Among these, the dual-channel switch-scan scheme, through the cooperation of one fixed reference channel and one switch-scan channel, can achieve phase difference observation at a relatively low hardware cost, making it the mainstream technology direction for low-cost, embedded direction finding equipment.

[0003] Multi-channel synchronous array direction finding schemes configure an independent receiving channel for each antenna element, enabling synchronous acquisition of complete array snapshot data. Combined with super-resolution algorithms, high direction finding accuracy can be achieved. However, the large number of channels, high hardware cost and power consumption, and the difficulty in synchronizing and calibrating multi-channel amplitude and phase make it unsuitable for low-cost, portable deployment scenarios. Single-channel switch scanning schemes poll multiple array elements using RF switches, multiplexing a single receiving channel in a time-division manner, resulting in lower hardware costs. However, the asynchronous acquisition of signals from different array elements leads to poor signal consistency, and the direction finding stability is susceptible to the burst characteristics of the signal.

[0004] To balance cost control and phase observation stability, the industry has developed a dual-channel switch-scan direction finding scheme with a fixed reference channel: the first receiving channel is fixedly connected to a reference element, and the second receiving channel is sequentially connected to multiple scanning elements via an RF switch. Each time slot synchronously acquires signals from the reference channel and the current scanning element. The reference channel serves as the phase reference to offset the effects of local oscillator drift and timing deviations. The direction of arrival is then estimated based on the phase observations of each element relative to the reference channel. This scheme improves the consistency of phase observations while controlling the number of channels and has been widely used in engineering.

[0005] However, in practical applications, existing dual-channel switch scanning direction finding schemes often use a fixed preset order for scanning sequences, without dynamically adjusting the acquisition priority based on the angle distribution of the current direction finding results. Limited short-slot acquisition resources cannot be concentrated on array elements with higher resolution, and there is room for improvement in direction finding update efficiency. Phase observations are mostly based on conventional cross-spectral calculations, which are easily affected by fluctuations in the power of UAV image transmission signals and uneven energy distribution at frequency points. The stability of phase observations needs to be improved in scenarios with sudden signal bursts. The observation data at different frequencies and time slots were not subject to differentiated quality control. Low-quality observations in noise and interference frequency bands participated in azimuth estimation together with high-quality observations, which affected the robustness of the direction finding results to some extent. The fixed-duration acquisition mode is usually adopted, which makes it difficult to simultaneously ensure the reliability of identification in the initial acquisition stage and the real-time update in the tracking stage; at the same time, it lacks a quantitative tracking status determination mechanism, and cannot adaptively adjust the working mode when the tracking quality deteriorates, resulting in insufficient robustness of continuous tracking. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a dual-channel UAV radio frequency direction finding and tracking method and system based on angle posterior feedback scanning. It solves the technical problems of fixed scanning sequence, inability to dynamically optimize the allocation of acquisition resources, susceptibility of conventional cross-spectrum to signal power fluctuations, lack of differentiated control over observation quality, difficulty in balancing the reliability of initial identification and real-time tracking with fixed acquisition duration, and lack of quantitative judgment mechanism for tracking status.

[0007] To achieve the above-mentioned technical effects, the technical solution adopted by this invention is: a dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning, comprising: The radio frequency signal output by the antenna array is synchronously sampled by a dual-channel receiver to obtain reference channel IQ data and scan channel IQ data; The reference channel IQ data and the scan channel IQ data are preprocessed to obtain a representative FFT block; Frequency point filtering is performed on the representative FFT blocks to obtain a set of effective image transmission frequency points; Based on the reference channel IQ data, the scan channel IQ data, and the effective image transmission frequency point set, the reference normalized cross spectrum is calculated to obtain the normalized coherence. Based on the normalized coherence, a three-dimensional observation quality matrix is ​​constructed; Based on the reference normalized cross spectrum and the three-dimensional observation quality matrix, the quality-weighted broadband angle spectrum is calculated to obtain the posterior angle distribution. Based on the posterior angle distribution, the posterior entropy, main-side peak interval and average coherence are calculated to obtain the direction finding confidence gating quantity. The current tracking stage is determined based on the direction finding confidence gating quantity, and the direction finding result is output.

[0008] As a preferred embodiment, obtaining the reference channel IQ data and the scan channel IQ data specifically includes: The radio frequency signal output by the fixed reference array element is sampled through the first receiving channel to obtain the original sampling data of the reference channel; The radio frequency signal output by the scanning array element switched by the radio frequency switch is sampled through the second receiving channel to obtain the original sampling data of the scanning channel; DC offset elimination is performed on the original sampled data of the reference channel and the original sampled data of the scan channel to obtain the reference channel IQ data and the scan channel IQ data.

[0009] As a preferred implementation, the representative FFT block is obtained specifically including: The reference channel IQ data and the scan channel IQ data are subjected to power normalization processing to obtain normalized IQ data; The normalized IQ data is divided into multiple 512-point candidate FFT blocks to obtain a candidate block set; The candidate block set is segmented in chronological order to obtain a time segment set; The temporal envelope power of the candidate blocks in each time segment is calculated to obtain the candidate block with the highest power in the segment. The candidate block with the highest power within the segment is extracted to obtain a representative FFT block.

[0010] As a preferred implementation, obtaining the normalized coherence specifically includes: The reference channel IQ data and the scan channel IQ data are cross-spectrally accumulated at each frequency point of the effective image transmission frequency point set to obtain the cross-spectral accumulation amount; The reference channel IQ data and the scan channel IQ data are summed at each frequency point of the effective image transmission frequency point set to obtain the reference channel power accumulation amount and the scan channel power accumulation amount; The cross-spectral accumulation, the reference channel power accumulation, and the scan channel power accumulation are normalized to obtain the reference normalized cross-spectrum. The reference normalized cross-spectral expression: ; in, For reference, normalized cross spectrum, This is the cross-spectral accumulation. For reference channel power accumulation, This is the cumulative power of the scan channel. For the observation time slot numbering, For valid image transmission frequency point numbering, This is a preset regularization constant.

[0011] As a preferred implementation, constructing the three-dimensional observation quality matrix specifically includes: The observation quality weights are obtained by weighting the normalized coherence. The three-dimensional observation quality matrix is ​​obtained by rearranging the observation quality weights and the reference normalized cross spectrum in three dimensions.

[0012] As a preferred implementation, obtaining the posterior distribution of the angle specifically includes: The theoretical relative complex phase is calculated based on the candidate angle, effective frequency point number, and scan array element coordinates; The three-dimensional cross spectrum matrix is ​​obtained by three-dimensional rearrangement based on the reference normalized cross spectrum; The phase consistency score is obtained by performing conjugate multiplication on the elements of the three-dimensional cross-spectral matrix and the corresponding theoretical relative complex phase. The phase consistency score is truncated to obtain the effective matching score; The effective matching score is weighted and accumulated with the three-dimensional observation quality matrix to obtain the quality-weighted broadband angle spectrum; The expression for the quality-weighted broadband angular spectrum is: ; in, For quality-weighted broadband angular spectrum To scan the round number, Numbering of scan array elements The total number of frequency points in the effective image transmission frequency point set. This is a three-dimensional observation quality matrix. It is a three-dimensional cross-spectral matrix. For theoretical relative complex phase The conjugate of complex numbers, For candidate roles, For valid image transmission frequency point numbering, This is the number of the scan array elements.

[0013] As a preferred implementation, the direction-finding reliable gating quantity is obtained, specifically including: The quality-weighted broadband angle spectrum is normalized to obtain the posterior probability distribution of the angle. Calculate the normalized posterior entropy, main-side peak interval, and average coherence based on the posterior probability distribution of the angle. The expression for the normalized posterior entropy is: ; The expression for the main-side peak interval is: ; The expression for the average coherence is: ; in, For normalized posterior entropy, The total number of candidate angles. Let positivity be the posterior probability distribution of the angle. For the first Candidate angle index, As a candidate, Main and secondary peaks are separated. The maximum spatial spectral peak value, To exclude the maximum side peak value after excluding the neighborhood of the main peak, For average coherence, The total number of observation time slots participating in direction finding. The total number of frequency points in the effective image transmission frequency point set. For the first The first observation time slot, the first Normalized coherence corresponding to each effective frequency point.

[0014] As a preferred embodiment, the output direction finding results specifically include: When the normalized posterior entropy The main and secondary peak intervals And the average coherence If so, it is determined that the system is in a stable tracking phase, and the output is... The candidate angle corresponding to the maximum value is used as the direction finding result; When the normalized posterior entropy The main and secondary peak intervals or the average coherence If the signal is lost, it indicates that the system is in a phase of risk of losing lock, and a low-confidence direction finding result is output.

[0015] As a preferred embodiment, the method further includes: calculating the discrimination score of each scanning array element based on the posterior angle distribution to obtain the next round of short-slot radio frequency switching scanning sequence; For the posterior probability distribution of the angle Sort the candidates in descending order and select the top 10 to form a candidate angle set. ; According to the candidate angle set Any two candidate angles , Calculate the first The theoretical phase difference of each scanning array element at the effective frequency point; The theoretical phase difference is logarithmically weighted and summed, and then weighted again using the posterior probability of the candidate angles to obtain the discrimination score. ; Regarding the discrimination ability score Arrange the sequences in descending order to obtain the next round of short-slot RF switch scanning sequences; The discrimination ability score The expression: ; in, Candidates The posterior probability, Candidates The posterior probability, The total number of frequency points in the effective image transmission frequency point set. Candidates The corresponding theoretical relative complex phase, Candidates The corresponding theoretical relative complex phase.

[0016] Based on the above method, the present invention provides a dual-channel UAV radio frequency direction finding and tracking system based on angle posterior feedback scanning, comprising: A dual-channel receiver is used to synchronously sample the radio frequency signals output by the antenna array to obtain reference channel IQ data and scan channel IQ data; The preprocessing module is used to preprocess the reference channel IQ data and the scan channel IQ data to obtain a representative FFT block; The frequency point filtering module is used to filter the representative FFT blocks by frequency points to obtain a set of valid image transmission frequency points; The cross-spectrum calculation module is used to calculate the reference normalized cross-spectrum based on the reference channel IQ data, the scanning channel IQ data, and the effective image transmission frequency point set, and obtain the normalized coherence. A quality matrix construction module is used to construct a three-dimensional observation quality matrix based on the normalized coherence. An angle spectrum calculation module is used to calculate the quality-weighted broadband angle spectrum based on the reference normalized cross spectrum and the three-dimensional observation quality matrix, and obtain the posterior angle distribution. The confidence assessment module is used to calculate the posterior entropy, main-side peak interval and average coherence based on the posterior angle distribution to obtain the direction finding confidence gating quantity. The result output module is used to determine the current tracking stage based on the direction finding confidence gating quantity and output the direction finding result.

[0017] Compared with existing technologies, this invention utilizes an angle posterior feedback mechanism to dynamically calculate the directional discrimination capability of each array element based on the probability distribution of the current direction finding results, adaptively generating the next round of scanning sequences, and prioritizing the allocation of short-slot acquisition resources to high-discrimination array elements. With an 8-element configuration, only 4 core array elements need to be scanned during the stable tracking phase to maintain direction finding accuracy. The single-round acquisition time and azimuth calculation computation are reduced compared to the full-array fixed-polling scheme, resulting in an improved direction finding update rate and achieving a higher tracking refresh rate at the same hardware cost.

[0018] A reference-normalized cross-spectrum calculation method is adopted, which normalizes the cross-spectrum results by the square root of the dual-channel power product. This effectively offsets the interference of UAV image transmission signal power fluctuations, uneven frequency energy distribution, and amplitude fluctuations between time slots on phase observation, ensuring that the cross-spectrum phase stably reflects the true path difference between array elements. For sudden, power-jumping image transmission signals, the consistency of phase estimation is significantly improved compared to conventional cross-spectrum schemes, and the reliability of direction finding in low signal-to-noise ratio scenarios is improved.

[0019] A three-dimensional observation quality matrix is ​​constructed, consisting of scan cycles, scan elements, and effective frequencies. Differential weights are assigned to each observation data point based on normalized coherence. During angle spectrum fusion, the contribution weights of noise frequencies, out-of-band interference, and transient anomalies from switching are automatically suppressed. This mechanism filters out misleading low-quality observations, making the posterior angle distribution more concentrated in the true direction of arrival. The anomalous jump amplitude of the direction finding results is significantly reduced compared to the unweighted scheme, resulting in stronger anti-interference capabilities.

[0020] The system employs a two-tiered adaptive operating mode: long-slot identification and short-slot tracking. In long-slot mode, it performs full-element scanning and full-band frequency point search, ensuring the accuracy of effective image transmission frequency point acquisition and initial azimuth estimation in the initial stage. Once the target is locked, it automatically switches to short-slot mode, directly reusing the locked effective frequency point set without repeated frequency point search. The delay of single-round signal acquisition and processing is compressed to less than 1 / 8 of that in long-slot mode, improving the real-time performance of continuous tracking without reducing the initial acquisition success rate.

[0021] A reliable gating mechanism for direction finding is constructed based on three quantization dimensions: normalized posterior entropy, main-side peak interval, and average coherence. This mechanism can automatically determine whether the current tracking state is stable or at risk of loss of lock. When consecutive lock-losses occur during short-slot tracking, the system autonomously reverts to long-slot mode to re-execute frequency identification and azimuth acquisition, avoiding continuous output of low confidence or incorrect azimuth. The entire process requires no manual intervention and can adapt to complex scenarios such as target maneuvering and signal discontinuity, improving the stability and autonomy of continuous tracking. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the logic of the dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to the present invention. Figure 2 This is a schematic diagram of the dual-ring antenna array structure of the present invention; Figure 3 This is a schematic diagram illustrating the principle of selecting representative FFT blocks in the long time slot mode of this invention; Figure 4 This is a schematic diagram of the effective image transmission frequency point selection results in the long time slot mode of the present invention; Figure 5 This is a schematic diagram illustrating the principle of effective frequency point reuse in the short time slot mode of the present invention; Figure 6 This is a schematic diagram showing the matching comparison between the theoretical phase and the measured phase under different candidate directions according to the present invention; Figure 7 This is a schematic diagram of the mass-weighted broadband angle spectrum and posterior angle distribution of the present invention; Figure 8 This is a schematic diagram of the module structure of the dual-channel UAV radio frequency direction finding and tracking system based on angular posterior feedback scanning according to the present invention. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to the embodiments and accompanying drawings. However, this should not be construed as limiting the scope of the above-described subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0024] Example 1: The overall hardware and system infrastructure of the present invention. In this example, the hardware adopts a dual-channel receiver and a dual-ring array architecture. refer to Figure 2 , Figure 2 This diagram illustrates the positions of the dual-ring array and the fixed reference element. The red dot at the center of the array represents the fixed reference element. The inner ring has a radius of 0.38λ, and the outer ring has a radius of 0.72λ. A total of eight blue scanning elements are evenly distributed across the dual rings. The first receiving channel is directly connected to the central reference element, while the second receiving channel uses an RF switch to time-division multiplex the eight scanning elements. The system has a baseband bandwidth of 61.44MHz, a sampling rate of 61.44MSPS, and operates in the 2.4GHz image transmission band commonly used by UAVs. Before deployment, static amplitude and phase offline calibration was performed on both channels and the RF switch throughout the entire path to eliminate inherent hardware phase deviations.

[0025] Complete algorithm logic flow reference Figure 1 All signal processing, adaptive control, and feedback scheduling steps follow Figure 1 Serial processing logic: IQ acquisition → preprocessing and representative block selection → effective frequency point extraction → normalized cross spectrum and three-dimensional quality matrix construction → weighted broadband angular spectrum calculation → credibility assessment → state determination → posterior feedback to generate scan sequence, closed-loop iterative operation.

[0026] Step S1: IQ Data Preprocessing and Representative FFT Block Generation refer to Figure 1 , Figure 3 This step preprocesses the raw IQ data output by the dual-channel receiver, selecting the time-domain segment with the best signal quality for subsequent frequency point analysis and cross-spectrum calculation.

[0027] ① Data preprocessing The system first discards the first 100 protection samples in each scan slot to avoid transient interference caused by RF switch switching; then it acquires valid IQ data, with a sampling length of [missing information] in long slot mode. Point, sampling length in short time slot mode is For valid IQ data, DC removal, power normalization, and amplitude limiting are performed sequentially: DC removal eliminates DC bias in the receiving channel; power normalization normalizes the signal power of each time slot to a uniform level, reducing the impact of overall power differences between different time slots; amplitude limiting suppresses sudden abnormal large value sampling points, avoiding single-point anomalies from interfering with subsequent statistics.

[0028] ② Candidate FFT block partitioning The preprocessed normalized IQ data is segmented into 512-point segments to obtain several candidate FFT blocks, forming a candidate block set. In long-slot mode, each observation slot can obtain... A candidate FFT block; in short time slot mode, it can be obtained Candidate FFT blocks.

[0029] ③ Time segmentation and representative block selection Figure 3 This diagram illustrates the selection of candidate FFT blocks and representative blocks for long time slots. The horizontal axis represents the FFT block number, and the vertical axis represents the signal envelope power. All candidate blocks are divided into 64 segments according to time. Only the candidate block with the highest time domain envelope power is retained in each segment as the representative FFT block, which greatly filters out noiseless idle data and reduces the amount of computation. Divide the candidate block set evenly according to time order. In this embodiment, time segments are used to divide the time into segments. Each segment contains multiple consecutive candidate blocks. For all candidate blocks within each time segment, calculate their temporal envelope power:

[0030] ; in, For the first The first observation time slot, the first Temporal envelope power of candidate FFT blocks; The observation time slot number corresponds to a single acquisition process after a radio frequency switch switching; The sequence number of the candidate FFT block; This is a dual-channel power weighting coefficient used to adjust the proportion of the reference channel and scan channel power in the envelope calculation. In this embodiment, it is taken as... That is, representative blocks are selected only by reference channel power; In this embodiment, the number of sampling points contained in a single FFT block is... ; The sample point number within the FFT block, with a value range of... ~ ; For the first The first observation time slot, the first Within the candidate block, the first The reference channel IQ sampled value corresponding to each sampling point; For the first The first observation time slot, the first Within the candidate block, the first The scan channel IQ sample value corresponding to each sampling point.

[0031] The candidate block with the highest envelope power within each time segment is selected as the representative block for that segment. All representative blocks from all segments together constitute a representative FFT block set. A 512-point FFT transform is performed on each representative FFT block to obtain the spectrum of the representative block of the reference channel. With scan channel representative block spectrum ,in To represent the segment numbering of the block, This is the FFT frequency point number.

[0032] By selecting high-power representative blocks in segments, effective observation segments of burst signals from image transmission are preserved, the amount of data for subsequent cross-spectrum calculations is significantly reduced, and low-energy pure noise segments are filtered out, indirectly improving the signal-to-noise ratio of phase estimation.

[0033] Step S2: Filtering of valid image transmission frequencies refer to Figure 4 , Figure 5 This step selects image transmission signal frequency points with significantly higher energy than the noise floor from the FFT frequency points to form an effective image transmission frequency point set, which is used for subsequent cross-spectrum and direction finding calculations.

[0034] Perform the complete frequency selection process in long time slot mode: Figure 4 This is a schematic diagram showing the selection results of effective frequency points for long time slots. Figure 4 The above figure shows the average power spectrum of the representative block. Figure 4 The following figure shows the selected normalized effective frequency points; ① Calculate the average power of the representative block at each FFT frequency point: ; in, For the first The first observation time slot, the first The average power of the reference channel corresponding to each FFT frequency point; The total number of time segments, i.e., the total number of representative FFT blocks, in this embodiment... ; The time segment number is the value of the time segment. ~ ; For the first The first observation time slot, the first The first representative block in the Reference channel spectrum values ​​at each frequency point; This is the FFT frequency point number.

[0035] ② Set the candidate frequency range. In this embodiment, we select... The baseband frequency points are used as the candidate range to exclude DC and near-DC low-frequency interference.

[0036] ③ Calculate the frequency point selection threshold: ; in, A threshold is selected for the frequency point; frequency points with power higher than the threshold are determined to be valid image transmission frequency points. To select the larger of the two input values; The noise threshold coefficient is taken in this embodiment. ; The noise floor estimate is the median power of all candidate frequency points within the candidate frequency range. The peak ratio threshold coefficient is taken in this embodiment. ; The maximum power value across all frequency points within the candidate frequency range; This is the set of frequency point indices corresponding to the candidate frequency range.

[0037] ④ Include frequency points with power exceeding the threshold into the set of valid image transmission frequency points. If the number of frequency points in the set is less than 48, supplement them to 48 by power from highest to lowest; if it exceeds 192, retain the top 192 with the highest power to control the amount of subsequent calculations. The final output is the set of valid image transmission frequency point numbers. With the number of effective frequency points .

[0038] In short time slot mode, the set of valid image transmission frequency points locked in long time slot mode is directly reused, and frequency point search is not repeatedly performed to shorten processing latency.

[0039] Figure 5 This diagram illustrates the reuse of effective frequency points in short time slots. Once locked, there is no need to search the entire frequency band again; the effective frequency point set output by long time slots can be directly reused, saving the complete spectrum traversal process and compressing the processing delay of a single round.

[0040] Step S3: Calculate the reference normalized cross spectrum and normalized coherence. This step calculates the normalized cross spectrum of the scanning channel relative to the reference channel based on the effective set of image transmission frequencies, and obtains the normalized coherence degree to quantify the reliability of the observation.

[0041] ① Cross-spectrum and power accumulation For each valid frequency point, traverse all representative FFT blocks and accumulate to calculate the cross spectrum and dual-channel power: ; ; ; in, For the first The first observation time slot, the first The cross-spectral accumulation corresponding to each effective frequency point is a complex number; This is the number of the valid image transmission frequency point, with a value range of [value range missing]. ~ ; For the first The original FFT frequency point number corresponding to each valid frequency point, i.e. ; For the first The first observation time slot, the first The first representative block in the The spectral values ​​of the scanning channels at each frequency point; For the first The first observation time slot, the first The first representative block in the Reference channel spectrum values ​​at each frequency point; For the first The first observation time slot, the first The accumulated power of the reference channel corresponding to each effective frequency point; For the first The first observation time slot, the first The cumulative power of the scanning channel corresponding to each effective frequency point.

[0042] ②Reference normalized cross-spectral calculation Dividing the cross-spectral sum by the square root of the product of the two-channel power yields the reference normalized cross-spectrum: ; in, For the first The first observation time slot, the first The reference normalized cross spectrum value corresponding to each effective frequency point is a complex number; This is a regularization constant used to prevent the denominator from being zero. In this embodiment, it is taken as... .

[0043] ③ Calculation of normalized coherence Normalized coherence is the square of the modulus of the reference normalized cross spectrum: ; in, For the first The first observation time slot, the first The normalized coherence degree corresponding to each effective frequency point ranges from 0 to 1. The higher the value, the stronger the phase correlation and the more reliable the quality of the observation.

[0044] Step S4: Construction of the three-dimensional observation quality matrix refer to Figure 6 , Figure 7 This step involves rearranging the two-dimensional observation quality into three dimensions according to the scanning rounds, scanning array elements, and effective frequency points to construct a three-dimensional observation quality matrix, which is used for subsequent weighted fusion of angle spectra.

[0045] ① Observation quality weight mapping Based on the normalized coherence, the observation quality weights are mapped to: ; ; in, For the first The first observation time slot, the first The original quality values ​​corresponding to each effective frequency point; To ensure the quality value is non-negative, the input value is taken as the larger of the input value and 0. For the first The first observation time slot, the first The final observation quality weights corresponding to each effective frequency point; The weighting factor for the quality baseline is taken in this embodiment. ; The median of all observed raw mass values ​​is used as the mass floor. The coherence weighting coefficient is taken in this embodiment. .

[0046] By setting a quality threshold, the weights of low-quality observations can be prevented from being completely zeroed, thus preserving a certain degree of information fusion capability.

[0047] ② Three-dimensional rearrangement Based on the scan cycle number and scan element number corresponding to each observation time slot, the two-dimensional observation quality and reference normalized cross-spectrum are rearranged into a three-dimensional structure: , when and ; , when and ; in, For the elements of the three-dimensional observation quality matrix, corresponding to the first... Scan wheel, first The first scanning array element, the first Quality weight of each effective frequency point; The elements of the three-dimensional cross-spectral matrix correspond to the first... Scan wheel, first The first scanning array element, the first Normalized cross-spectral values ​​of each effective frequency point; For each scan round number, one round corresponds to a complete scan of a specified array element sequence; The number of the scan array element; For the first The scan cycle number to which each observation time slot belongs; For the first The scanning array element number corresponding to each observation time slot.

[0048] Step S5: Calculation of mass-weighted broadband angular spectrum and posterior angular distribution This step obtains a broadband angular spectrum by phase matching between the array manifold and the measured cross spectrum, combined with mass weighting, and further transforms it into an angular posterior probability distribution.

[0049] ① Theoretical relative complex phase calculation Based on the array's geometric coordinates, calculate the theoretical complex phase of each scanned array element relative to the reference array element at the candidate angle. Let the coordinates of the reference array element be... , No. The coordinates of each scan array element are The candidate incoming wave direction unit vector is , No. The wavelengths corresponding to the effective frequency points are Then the theoretical relative complex phase is:

[0050] ; in, The direction of the incoming wave is At that time, the first The scanning element is relative to the reference element at the nth scan array element. Theoretical relative complex phase at each effective frequency point; The azimuth angle of the incoming wave to be estimated is the candidate angle. The imaginary unit satisfies ; For the first The spatial position coordinate vectors of each scan array element; The spatial position coordinate vector of the reference array element; Azimuth The corresponding unit vector of the incoming wave direction; For the first The electromagnetic wave wavelength corresponding to each effective frequency point.

[0051] In this embodiment, the reference array element is located at the origin, and the eight scanning array elements are distributed at radii of [missing information]. and On the inner and outer double rings, The wavelength of the electromagnetic wave corresponding to the center frequency.

[0052] ② Phase consistency matching refer to Figure 6 A heatmap comparing the theoretical and measured phase residuals for multiple candidate directions. The brightness of color blocks at different angles represents the degree of matching. For each candidate angle... The elements of the three-dimensional cross-spectral matrix are multiplied by the conjugate of the corresponding theoretical complex phase, and the real part is taken as the phase consistency score. The score is then truncated non-negatively to filter out anomalous observations with significantly reversed phases. ; in, The phase consistency score for a single observation; For theoretical relative complex phase .

[0053] ③ Quality-weighted broadband angular spectrum Figure 7 The upper curve represents the normalized angle spectrum, with the peak corresponding to the target's true azimuth of 137.3°. The lower bar chart compares the matching scores and phase residuals of each candidate angle. The phase consistency scores for all scanning cycles, scanning elements, and effective frequency points are accumulated using the three-dimensional observation quality matrix as weights to obtain the quality-weighted broadband angle spectrum. ; in, Candidates The corresponding quality-weighted broadband angular spectrum value; Sum over all valid frequencies. This represents the total number of effective image transmission frequency points.

[0054] ④ Posterior probability distribution of angle Softmax normalization of the quality-weighted broadband angle spectrum yields the posterior probability distribution of the angles: ; in, For the first The posterior probabilities corresponding to each candidate angle; The index number of the candidate angle; For the first The azimuth values ​​of the candidate angles; For the first The angle spectrum values ​​corresponding to each candidate angle; The total number of candidate angles is 360. In this embodiment, the angle search range is 0°~360°, the search step size is 1°, and there are a total of 360 candidate angles. The traversal index number is the candidate angle.

[0055] Step S6: Calculation of direction finding reliable gating quantity refer to Figure 1 This step extracts three quantitative indicators from the posterior angle distribution to form a direction finding confidence gating quantity, which is used to determine the tracking status.

[0056] ① Normalized posterior entropy Normalized posterior entropy is used to measure the degree of centrality of the posterior distribution of angles. The lower the entropy value, the more central the distribution, and the more reliable the direction finding results. ; in, The normalized posterior entropy has a value range of 0 to 1. When the posterior distribution is perfectly uniform, the normalized posterior entropy is 1; when the distribution is completely concentrated on a single candidate angle, the entropy is 0.

[0057] ② Main and secondary peak intervals The main-side-peak interval is used to measure the dominance of the main peak relative to the side peaks in the angular spectrum, and is measured in dB. ; in, The interval between the main and secondary peaks is expressed in dB. This is the global maximum value of the angular spectrum, i.e., the value of the main peak spectrum; To exclude the maximum peak value after excluding the main peak's neighborhood, in this embodiment, the main peak's neighborhood is taken as a range of 10° to the left and right of the main peak.

[0058] ③ Average coherence The average coherence is the average of the coherence at all observation slots and effective frequencies involved in direction finding, reflecting the overall phase correlation of the signal. ; in, The average coherence; This represents the total number of observation slots participating in the current direction finding. For the first The first observation time slot, the first Normalized coherence corresponding to each effective frequency point.

[0059] Step S7: Tracking Status Judgment and Result Output This step determines the current tracking stage based on the direction finding confidence gating value, outputs the direction finding result with the corresponding confidence level, and controls the switching of working modes.

[0060] The locking condition is set as: normalized posterior entropy And the main and secondary peaks are spaced apart. and average coherence .

[0061] ①State switching in long time slot mode If the current system is in long time-slot mode and the locking condition is met, the initial locking is considered successful, and the system switches to short time-slot tracking mode, outputting the candidate angle corresponding to the maximum posterior probability as the direction finding result. If the locking condition is not met, the system remains in long time-slot mode, continues to acquire data, and executes the next round of recognition.

[0062] ②State switching in short time slot mode If the current state is in short time slot mode and the locking condition is met, it is determined to be in a stable tracking phase. The candidate angle corresponding to the maximum posterior probability is output as a high-confidence direction finding result, and the continuous loss of lock count is cleared.

[0063] If the locking condition is not met, the system is identified as being in a lock-loss risk phase, and a low-confidence direction finding result is output. At the same time, the consecutive lock-loss count is incremented by 1. When the consecutive lock-loss count reaches 5 times, the system switches back to long time slot mode and re-executes frequency point identification and initial azimuth estimation.

[0064] Step S8: Generation of Angular Posterior Feedback Scan Sequence This step calculates the ability of each scanning array element to distinguish candidate angles based on the current posterior distribution of the angle, generates the next round of short-slot RF switching scanning sequence, and realizes feedback control of the direction finding results on the acquisition process.

[0065] ① Selection of candidate angle set posterior probability of angle Sort the candidate angles in descending order and select the top 10 with the highest probability to form a candidate angle set. , representing the most likely direction of incoming wave in the current posterior distribution.

[0066] ② Calculation of array element discrimination capability For each scanned array element, its discriminative power for pairwise combinations of all candidate angles is calculated, and then weighted using the product of posterior probabilities as weights to obtain the discriminative power score for that element: ; in, For the first The discrimination score of each scan array element; the higher the score, the stronger the discrimination ability of the array element for the current candidate direction. , For each of the two distinct candidate angles in the candidate angle set, This indicates that all pairwise combinations are iterated to avoid duplicate calculations. This is a set of high-probability candidate angles; Candidates The corresponding posterior probability; Candidates The corresponding posterior probability; Candidates The corresponding number The scanning element is at the... Theoretical relative complex phase at each effective frequency point; Candidates The corresponding number The scanning element is at the... Theoretical relative complex phase at each effective frequency point; This represents the total number of effective image transmission frequency points.

[0067] The greater the theoretical phase difference between the two candidate angles on the array element, the easier it is for the array element to distinguish between the two easily confused directions, and the higher the discrimination ability score.

[0068] ③Scan sequence generation All scanning array elements are sorted in descending order of their discrimination ability score, and the top four elements are selected to form the next round of short-slot RF switching scanning sequence. The scanning channel switches array elements sequentially according to this sequence for acquisition, thereby prioritizing the allocation of limited short-slot time to the array elements with the strongest discrimination ability, reducing the number of scanning array elements and improving the update rate while ensuring direction finding accuracy.

[0069] When the posterior angle distribution is relatively dispersed and the candidate angle coverage is large, the system can switch to the balanced scanning mode to ensure that all scanning elements are scanned at least once in several rounds, thus avoiding the loss of large-scale targets due to over-focusing.

[0070] refer to Figure 8 Example 2 describes a dual-channel UAV RF direction finding and tracking system based on angle posterior feedback scanning. The hardware employs a dual-channel RF receiver, RF switch array, and embedded heterogeneous processing platform architecture. The software utilizes a modular pipeline scheduling mechanism to achieve adaptive scanning, weighted broadband direction finding, quantized reliability judgment, and closed-loop tracking scheduling for 2.4GHz / 5.8GHz UAV image transmission and remote control RF signals. This addresses the shortcomings of traditional dual-channel equipment, such as fixed scanning sequences, undifferentiated weighted observations, lack of quantized tracking status determination, and poor real-time performance. The overall signal processing chain of the system is as follows: antenna array RF acquisition → dual-channel synchronous IQ sampling → IQ preprocessing and representative FFT block extraction → effective image transmission frequency point selection → normalized cross-spectrum and coherence solution → three-dimensional observation quality matrix construction → quality-weighted broadband angle spectrum solution and angle posterior distribution generation → three-dimensional reliability gating index evaluation → tracking status judgment and direction finding result output. Simultaneously, the result output module feeds back the next frame of the RF switch scanning sequence, forming a complete closed-loop direction finding and tracking system.

[0071] ① The dual-channel receiver is the front-end signal acquisition hardware unit of this system. The front-end is connected to an antenna array containing one fixed reference array element and multiple switchable scanning array elements. Internally, it integrates two synchronous radio frequency receiving links and a radio frequency switch switching subunit.

[0072] Channel hardware division: The first receiving channel is fixedly connected to the central reference element of the array, continuously acquiring the reference RF signal as a global phase reference to eliminate phase errors caused by local oscillator drift and timing offset; the second receiving channel is cascaded with multiple single-pole multi-throw RF switches at the back end. The RF switches are driven by the embedded control signal at the back end, switching between different scanning elements in a time-division manner. In each observation time slot, the reference channel and the current scanning element's RF signals are acquired synchronously.

[0073] Sampling preprocessing function: The receiver's internal hardware circuit completes the down-conversion and analog-to-digital synchronous sampling of the original RF signal, and performs DC offset cancellation operation on the two original sampling data channels to filter out the inherent DC component interference of the hardware. Finally, it outputs time-domain aligned and phase-synchronized reference channel IQ data and scan channel IQ data stream, which are transmitted to the back-end preprocessing module.

[0074] Timing control characteristics: The receiver can distinguish between two sampling duration configurations: long time slot acquisition mode and short time slot tracking mode. The long time slot uses a large number of sampling points to ensure complete acquisition of weak signals, while the short time slot compresses the sampling length to improve the tracking refresh rate. In both modes, strict dual-channel synchronous sampling is maintained to ensure that the time dimension of the two IQ data channels corresponds one-to-one.

[0075] ② The preprocessing module receives two IQ data streams from the dual-channel receiver, performs standardized preprocessing, and selects the representative FFT block with the best signal-to-noise ratio to reduce invalid noise data from participating in subsequent calculations. The specific processing flow is as follows: Power normalization processing: Global power normalization is performed on the reference channel IQ data and the scan channel IQ data respectively to offset the amplitude difference caused by the jump in UAV signal power between different observation time slots and to eliminate the impact of the imbalance between strong and weak signals on cross-spectrum calculation. Candidate FFT block partitioning: The normalized continuous IQ data stream is uniformly divided into candidate FFT blocks of fixed length (512 points) to generate a complete candidate block set; Segmented selection: All candidate FFT blocks are equally divided into several time segments along the time axis. The temporal envelope power of each block in the segment is calculated, and the candidate block with the highest power in a single segment is selected as the effective segment. Output representative FFT block: Summarize all high-power candidate blocks obtained from segmented filtering, form a representative FFT block set and output it to the frequency point filtering module. Only retain time domain segments containing valid UAV image transmission signals, filter out pure noise and idle data, and significantly reduce the computing power overhead of back-end frequency point and cross-spectrum calculation.

[0076] ③ The frequency point filtering module takes the representative FFT block output by the preprocessing module as input, performs energy filtering on the full-band spectrum after FFT transformation, and outputs only the set of valid image transmission frequency points containing UAV image transmission signals, distinguishing between two working modes to achieve differentiated processing: Long time slot full-band screening mode (initial acquisition / lost-lock reacquisition stage): Calculate the average spectral power of all FFT frequency points, set a noise low threshold and peak ratio dual screening threshold, eliminate low frequency DC interference and environmental noise frequency points, retain only signal frequency points with power higher than the threshold, and constrain the number of effective frequency points within a reasonable range to avoid too many frequency points causing computational explosion; Short time slot reuse mode (stable tracking stage): directly reuse the effective image transmission frequency point set obtained from the previous round of long time slot screening, without repeating the full spectrum traversal search, saving the large-scale spectrum statistics process and shortening the single-round signal processing delay; Output content: Output two types of parameters: valid frequency point number and total number of valid frequency points. These are synchronously transmitted to the cross-spectrum calculation module, limiting subsequent cross-spectrum calculations to be performed only within the valid frequency points.

[0077] ④ The cross-spectrum calculation module, relying on dual-channel IQ data and the effective image transmission frequency point set, completes the solution of normalized cross-spectrum and observation quality index normalized coherence, realizing stable signal phase extraction and avoiding power fluctuation interference: Multi-representative block accumulation operation: Traverse each valid frequency point, perform conjugate multiplication and accumulation on the spectral values ​​of all representative FFT blocks to obtain the cross-spectral accumulation, and simultaneously accumulate the total power of each frequency point of the reference channel and the scanning channel respectively; Reference normalized cross spectrum solution: A regularization constant is introduced to prevent division by zero in the denominator. The square root of the dual-channel power product is used to normalize the cross spectrum accumulation, resulting in a reference normalized cross spectrum that is resistant to power fluctuations and stably preserves the true phase difference information between array elements. Normalized coherence calculation: The square of the normalized cross spectrum is taken as the normalized coherence. The coherence value is 0~1. The higher the value, the stronger the phase correlation between the two signals at that time slot and frequency point and the higher the observation quality. Data output: The normalized cross spectrum and normalized coherence of the two-dimensional dimensions (observation time slots and effective frequency points) are uniformly output to the quality matrix construction module as the original basis for observation quality assessment.

[0078] ⑤ The quality matrix construction module, based on the normalized coherence output by the cross-spectrum calculation module, completes the observation quality weighting and three-dimensional data reconstruction, realizing differentiated weight labeling of observation data from different rounds, different array elements, and different frequency points: Observation quality weight mapping: Combine the global coherence limit and the coherence of a single frequency point for weighted fusion to generate non-negative observation quality weights. Observations with low coherence and noise are automatically assigned lower weights. At the same time, a minimum weight is set to avoid completely discarding low-quality data and losing information. Three-dimensional rearrangement: Using the scan round, scan array element, and effective frequency point as the three-dimensional coordinate axis, the two-dimensional observation weights and normalized cross-spectral data are reorganized according to the scan round number and array element number corresponding to each observation time slot to generate a three-dimensional observation quality matrix and a three-dimensional cross-spectral matrix respectively. Matrix cache output: The completed 3D observation quality matrix and the matching 3D cross-spectrum matrix are stored in the high-speed cache and simultaneously sent to the angle spectrum calculation module to provide weight support for broadband angle spectrum weighted fusion.

[0079] ⑥ The angle spectrum calculation module integrates the three-dimensional observation quality matrix and the three-dimensional cross-spectral matrix, combines array geometry theory to solve for the quality-weighted broadband angle spectrum, and transforms it into the posterior angle distribution for confidence assessment: Theoretical complex phase pre-calculation: Based on the antenna array coordinates, the angles of arrival of each candidate wave, and the wavelength of the effective frequency point, the theoretical relative complex phase and its conjugate value corresponding to each array element, each frequency point, and each candidate angle are solved in batches. Phase matching score calculation: Multiply the elements of the three-dimensional cross-spectral matrix with the corresponding theoretical complex phase conjugate, take the real part and perform non-negative truncation, and remove invalid matching results with phase reversal to obtain the phase consistency score of a single observation; Quality-weighted summation: Using the three-dimensional observation quality matrix as weights, the phase consistency scores of all scanning wheels, scanning array elements, and effective frequency points are weighted and summed to generate the quality-weighted broadband angle spectrum corresponding to each candidate angle. Angle posterior distribution generation: The global angle spectrum is normalized and transformed into a probabilistic posterior distribution. The posterior probability corresponding to each candidate angle represents the credibility of the incoming wave direction. The posterior probability data is output to the credibility assessment module.

[0080] ⑦ Credibility Assessment Module The reliability assessment module takes the posterior angle distribution as input, calculates three types of quantitative evaluation indicators in parallel, and combines them to form a unified direction-finding reliability gating quantity, thereby realizing digital evaluation of tracking quality. Normalized posterior entropy calculation: The normalized entropy value is calculated based on the posterior probabilities of all candidate angles. The lower the entropy value, the more concentrated the posterior distribution and the more certain the target orientation. Main-side peak spacing calculation: Extract the peak value of the main peak in the angular spectrum and the peak value of the largest side peak after removing the neighborhood of the main peak, and convert it into dB difference to characterize the distinction between the target signal and the interference side lobes; Average coherence calculation: The average normalized coherence of all observation time slots and all effective frequency points is calculated to characterize the overall dual-channel phase synchronization quality of the signal; Gating quantity packaged output: The normalized posterior entropy, main-side peak interval, and average coherence are integrated into a direction-finding reliable gating quantity, which is passed to the result output module as the basis for judgment in the tracking stage.

[0081] ⑧ The result output module is the core unit of the system's closed-loop control, and it has three functions: tracking state determination, direction finding result output, and adaptive scan sequence feedback generation. Tracking phase judgment logic: preset fixed threshold, synchronously verify three indicators: normalized posterior entropy, main-side peak interval, and average coherence. If all three indicators meet the standard, it is judged as a stable tracking phase; if any one of them fails to meet the standard, it is judged as a lock-out risk phase. Layered output of direction finding results: During the stable tracking phase, the candidate angle corresponding to the maximum posterior probability is output as the high-confidence standard direction finding result; during the lock loss risk phase, the low-confidence direction finding angle is output and the lock loss status is marked; when there are multiple consecutive lock losses, a command is issued to switch back to the long time slot full array element acquisition mode. Adaptive scanning sequence feedback generation: The high-probability candidate angle set is selected by sorting the posterior probability of the angles in descending order. The direction discrimination ability score is calculated for each array element. The higher the score, the stronger the discrimination ability of the array element for the current suspicious target direction. All scanning array elements are sorted in descending order of discrimination ability score. The high-scoring array elements are extracted to generate the next round of short time slot RF switch scanning sequence, which is sent to the RF switch control unit inside the dual-channel receiver. This realizes the reverse optimization acquisition process of the direction finding results, forming a complete closed-loop tracking of acquisition-solution-evaluation-optimization acquisition.

[0082] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning, characterized in that, include: The radio frequency signal output by the antenna array is synchronously sampled by a dual-channel receiver to obtain reference channel IQ data and scan channel IQ data. The reference channel IQ data and scan channel IQ data are preprocessed to obtain a representative FFT block. Frequency point filtering is performed on the representative FFT blocks to obtain a set of effective image transmission frequency points; Based on the reference channel IQ data, the scan channel IQ data, and the effective image transmission frequency point set, the reference normalized cross spectrum is calculated to obtain the normalized coherence. Based on the normalized coherence, a three-dimensional observation quality matrix is ​​constructed; Based on the reference normalized cross spectrum and the three-dimensional observation quality matrix, the quality-weighted broadband angle spectrum is calculated to obtain the posterior angle distribution. Based on the posterior angle distribution, the posterior entropy, main-side peak interval and average coherence are calculated to obtain the direction finding confidence gating quantity. The current tracking stage is determined based on the direction finding confidence gating quantity, and the direction finding result is output.

2. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, Obtaining the reference channel IQ data and the scan channel IQ data specifically includes: The radio frequency signal output by the fixed reference array element is sampled through the first receiving channel to obtain the original sampling data of the reference channel; The radio frequency signal output by the scanning array element switched by the radio frequency switch is sampled through the second receiving channel to obtain the original sampling data of the scanning channel; DC offset elimination is performed on the original sampled data of the reference channel and the original sampled data of the scan channel to obtain the reference channel IQ data and the scan channel IQ data.

3. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, Obtaining the representative FFT block specifically includes: The reference channel IQ data and the scan channel IQ data are subjected to power normalization processing to obtain normalized IQ data; The normalized IQ data is divided into multiple 512-point candidate FFT blocks to obtain a candidate block set; The candidate block set is segmented in chronological order to obtain a time segment set; The temporal envelope power of the candidate blocks in each time segment is calculated to obtain the candidate block with the highest power in the segment. The candidate block with the highest power within the segment is extracted to obtain a representative FFT block.

4. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, The normalized coherence degree is obtained by means of: The reference channel IQ data and the scan channel IQ data are cross-spectrally accumulated at each frequency point of the effective image transmission frequency point set to obtain the cross-spectral accumulation amount; The reference channel IQ data and the scan channel IQ data are summed at each frequency point of the effective image transmission frequency point set to obtain the reference channel power accumulation amount and the scan channel power accumulation amount; The cross-spectral accumulation, the reference channel power accumulation, and the scan channel power accumulation are normalized to obtain the reference normalized cross-spectrum. The reference normalized cross-spectral expression: ； in, For reference, normalized cross spectrum, This is the cross-spectral accumulation. For reference channel power accumulation, This is the cumulative power of the scan channel. For the observation time slot numbering, For valid image transmission frequency point numbering, This is a preset regularization constant.

5. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, Constructing the three-dimensional observation quality matrix specifically includes: The observation quality weights are obtained by weighting the normalized coherence. The three-dimensional observation quality matrix is ​​obtained by rearranging the observation quality weights and the reference normalized cross spectrum in three dimensions.

6. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, The posterior distribution of the angle is obtained, specifically including: The theoretical relative complex phase is calculated based on the candidate angle, effective frequency point number, and scan array element coordinates; The three-dimensional cross-spectrum matrix is ​​obtained by three-dimensional rearrangement based on the reference normalized cross-spectrum. The phase consistency score is obtained by performing conjugate multiplication on the elements of the three-dimensional cross-spectral matrix and the corresponding theoretical relative complex phase. The phase consistency score is truncated to obtain the effective matching score; The effective matching score is weighted and accumulated with the three-dimensional observation quality matrix to obtain the quality-weighted broadband angle spectrum; The expression for the quality-weighted broadband angular spectrum is: ； in, For quality-weighted broadband angular spectrum, To scan the round number, Numbering of scan array elements The total number of frequency points in the effective image transmission frequency point set. This is a three-dimensional observation quality matrix. It is a three-dimensional cross-spectral matrix. For theoretical relative complex phase The conjugate of complex numbers, For candidate roles, For valid image transmission frequency point numbering, This is the number of the scan array elements.

7. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, The direction-finding reliable gating quantity is obtained, specifically including: The quality-weighted broadband angle spectrum is normalized to obtain the posterior probability distribution of the angle. Calculate the normalized posterior entropy, main-side peak interval, and average coherence based on the posterior probability distribution of the angle. The expression for the normalized posterior entropy is: ； The expression for the main-side peak interval is: ； The expression for the average coherence is: ； in, For normalized posterior entropy, The total number of candidate angles. Let positivity be the posterior probability distribution of the angle. For the first Candidate angle index, As a candidate, Main and secondary peaks are separated. The maximum spatial spectral peak value, To exclude the maximum side peak value after excluding the neighborhood of the main peak, For average coherence, The total number of observation time slots participating in direction finding. The total number of frequency points in the effective image transmission frequency point set. For the first The first observation time slot, the first Normalized coherence corresponding to each effective frequency point.

8. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, The output direction finding results specifically include: When the normalized posterior entropy The main and secondary peak intervals And the average coherence If so, it is determined that the system is in a stable tracking phase, and the output is... The candidate angle corresponding to the maximum value is used as the direction finding result; When the normalized posterior entropy The main and secondary peak intervals or the average coherence If the signal is lost, it indicates that the system is in a phase of risk of losing lock, and a low-confidence direction finding result is output.

9. The dual-channel UAV radio frequency direction finding and tracking method based on angle posterior feedback scanning according to claim 1, characterized in that, Also includes: Based on the posterior angle distribution, the discrimination score of each scanning element is calculated to obtain the next round of short-slot RF switching scanning sequence; For the posterior probability distribution of the angle Sort the candidates in descending order and select the top 10 to form a candidate angle set. ; According to the candidate angle set Any two candidate angles , Calculate the first The theoretical phase difference of each scanning array element at the effective frequency point; The theoretical phase difference is logarithmically weighted and summed, and then weighted again using the posterior probability of the candidate angles to obtain the discrimination score. ; Regarding the discrimination ability score Arrange the sequences in descending order to obtain the next round of short-slot RF switch scanning sequences; The distinguishing ability score The expression: ； in, Candidates The posterior probability, Candidates The posterior probability, The total number of frequency points in the effective image transmission frequency point set. Candidates The corresponding theoretical relative complex phase, Candidates The corresponding theoretical relative complex phase.

10. A dual-channel UAV radio frequency direction finding and tracking system based on posterior angle feedback scanning, characterized in that, include: A dual-channel receiver is used to synchronously sample the radio frequency signals output by the antenna array to obtain reference channel IQ data and scan channel IQ data; The preprocessing module is used to preprocess the reference channel IQ data and the scan channel IQ data to obtain a representative FFT block; The frequency point filtering module is used to filter the representative FFT blocks by frequency points to obtain a set of valid image transmission frequency points; The cross-spectrum calculation module is used to calculate the reference normalized cross-spectrum based on the reference channel IQ data, the scanning channel IQ data, and the effective image transmission frequency point set, and obtain the normalized coherence. A quality matrix construction module is used to construct a three-dimensional observation quality matrix based on the normalized coherence. An angle spectrum calculation module is used to calculate the quality-weighted broadband angle spectrum based on the reference normalized cross spectrum and the three-dimensional observation quality matrix, and obtain the posterior angle distribution. The confidence assessment module is used to calculate the posterior entropy, main-side peak interval and average coherence based on the posterior angle distribution to obtain the direction finding confidence gating quantity. The result output module is used to determine the current tracking stage based on the direction finding confidence gating quantity and output the direction finding result.

Images