Multi-dimensional super-resolution target sensing method based on 5G AeroMACS

Abstract

The invention belongs to communication perception integration, and provides a multidimensional super-resolution target perception method based on 5G AeroMACS. Through combination of a 5G communication technology and a communication and sensing integration technology, when a signal is transmitted, to-be-transmitted data is modulated into a plurality of OFDM symbols by adopting 5GAeroMACS, and the OFDM symbols are transmitted. When signals are received, OFDM symbols are received through a plurality of receiving antennas, data reduction is carried out, the data are stacked into three-dimensional data tensors, and the three-dimensional data tensors are stored in a three-dimensional complex matrix. And based on the three-dimensional complex matrix within the time window length, determining a signal subspace dimension, constructing a signal subspace, a noise subspace and a three-dimensional spectrum peak search function, and obtaining a multi-dimensional super-resolution spectrogram. And during target detection, a three-dimensional constant false alarm detection algorithm is adopted to obtain a target sensing result. According to the invention, super-resolution detection of airport multi-dimensional targets is realized, moving targets are accurately monitored and identified, potential safety risks are prevented, and the airport operation efficiency and the flight safety are improved at the same time.

Description

Technical Field

[0001] The present invention belongs to communication perception integration, and specifically relates to a multi-dimensional super-resolution target perception method based on 5G AeroMACS. Background Art

[0002] The rapid growth of the aviation industry has made airport security supervision essential. A key task of airport security monitoring is to achieve high-definition monitoring and identification of moving targets such as aircraft, vehicles and personnel inside the airport. This high-definition monitoring plays a decisive role in preventing runway incursions, improving airport operation efficiency and ensuring flight safety. Although traditional monitoring technologies, such as radar and sensor networks, can provide monitoring data, they are insufficient in terms of monitoring accuracy, data processing speed and real-time response. These deficiencies pose a challenge to the safe operation of airports and also limit the efficiency and safety of air transportation.

[0003] As an emerging aviation broadband communication technology, 5G AeroMACS technology can safely and efficiently transmit operational business information using dedicated frequencies within civil aviation airports. Combined with Beidou high-precision positioning technology, 5G AeroMACS technology can significantly improve the communication efficiency between airports, air traffic management and airlines, and enhance the ability to perceive the airport's operating situation. The application of this technology is of great significance for preventing vehicles from invading runways, and can achieve the coordinated operation of aircraft, vehicles, runways and facilities, thereby improving the safe operation level of airports.

[0004] Integrated communication and perception technology has opened up a new way to improve airport monitoring capabilities. The highly integrated system is based on a simultaneous beam system, mainly including linear frequency modulation waves, signal separation and orthogonal frequency division multiplexing technology. By modulating the linear frequency modulation wave to load communication information, this integrated signal can simultaneously realize radar detection and communication functions, but the challenge is that the communication rate is low and it is difficult to meet actual communication needs. Signal separation technology fuses radar signals and communication signals at the transmitting end and separates and processes them at the receiving end. It is divided into two methods: additive and multiplicative signal separation. The degree of integration of this method is limited, and the communication signal will occupy part of the transmission power of the radar signal, affecting the detection performance of the radar. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention proposes a multi-dimensional super-resolution target perception method based on 5G AeroMACS. Based on the 5G AeroMACS OFDM communication system, the method realizes super-resolution detection of multi-dimensional targets at the airport, which not only helps to achieve super-resolution perception, but also significantly improves the airport's safety monitoring capabilities. Applying this method, the airport can more accurately monitor and identify mobile targets, prevent potential safety risks, and improve the airport's operating efficiency and flight safety. In addition, it can also provide airports with richer data support, help airport managers make more accurate decisions, optimize resource allocation, improve service quality, and ultimately realize the intelligence and automation of airport operations.

[0006] The technical solution of the present invention is as follows: A multi-dimensional super-resolution target perception method based on 5G AeroMACS includes the following steps: Step S1, using 5G AeroMACS to modulate the data to be transmitted into multiple OFDM symbols and transmit them; Step S2, multiple receiving antennas receive OFDM symbols in the air, perform data restoration and stack them into a three-dimensional data tensor, and store them in a three-dimensional complex matrix; Step S3, constructing a signal subspace, a noise subspace and a three-dimensional spectrum peak search function based on the three-dimensional complex matrix to obtain a multi-dimensional super-resolution spectrum; Step S4: adopt a three-dimensional constant false alarm detection algorithm to obtain the final target perception result.

[0007] Preferably, the step S1 specifically includes: Step S1-1, data preparation: using 5G AeroMACS to map the data to be transmitted into M OFDM symbols; Step S1-2, IFFT operation: perform IFFT operation on each OFDM symbol to convert the frequency domain signal into a time domain signal; Step S1-3, adding a cyclic prefix: adding a cyclic prefix at the end of each OFDM symbol; Step S1-4, signal transmission: the modulated OFDM symbols are sent into the air through a transmitting antenna.

[0008] Preferably, each OFDM symbol in step S1-1 contains N subcarriers, and the subcarriers are within the aviation-specific frequency range of 5091-5150 MHz.

[0009] Preferably, the step S2 specifically includes: Step S2-1, H receiving antennas each receive an OFDM symbol in the air; Step S2-2, sequentially intercepting, reshaping, removing cyclic prefixes, matrixing, performing FFT transformation and matched filtering on the OFDM symbols of each receiving antenna; Step S2-3: Store the result of matched filtering into a three-dimensional complex matrix with dimensions of H×N×M.

[0010] Preferably, the step S2-2 specifically includes: (1) Extract the first F samples of the OFDM symbol of the receiving antenna with the same length as the transmitted OFDM symbol and store them in the variable Rx0; (2) Reshape the variable Rx0 into a (N, -1) matrix and then transpose it to obtain the variable Rx1, where -1 represents the automatically calculated matrix dimension to ensure that the total number of elements remains unchanged after reshaping; (3) Remove the cyclic prefix from the variable Rx1 and reshape it into a matrix Rx with dimension (N-CPsize,-1), where CPsize is the size of the cyclic prefix; (4) Perform FFT transformation on the matrix Rx to obtain the complex matrix Rx_dem, whose dimension is (N-CPsize,M); (5) Perform element-by-element multiplication of the complex matrix Rx_dem with the complex conjugate of the transmitted OFDM symbol to eliminate the influence of 5G AeroMACS transmission.

[0011] Preferably, the step S3 specifically includes: Step S3-1, dividing the three-dimensional complex matrix into The size is A sub-tensor of , , ; Step S3-2: Expand each sub-tensor and merge by column to obtain column vector; Step S3-3, calculating the covariance matrix; Step S3-4, based on the covariance matrix of the time window length Q, using the deep learning model to estimate the signal subspace dimension P; Step S3-5: Decompose the covariance matrix into eigenvalues ​​and construct the signal subspace using the eigenvectors corresponding to the largest P eigenvalues. , the remaining The noise subspace is constructed by using eigenvectors ; Step S3-6: construct a three-dimensional spectrum peak search function to obtain an RVA spectrum.

[0012] Preferably, the expression of the three-dimensional spectrum peak search function in step S3-6 is as follows:

[0013] in, Indicates the target delay, Indicates the Doppler frequency shift of the target, Represents the angle of the target, represents the three-dimensional delay-Doppler search column vector;

[0014]

[0015]

[0016]

[0017] in, represents the imaginary unit, represents the element spacing of the antenna array, Indicates the wavelength of the 5G AeroMACS signal.

[0018] Preferably, the step S4 specifically includes: Step S4-1: Setting the false alarm probability ; Step S4-2, preprocessing the RVA spectrum; Step S4-3, determine the current detection area, including the central detection unit, the protection area and the reference area: for each detection unit in the RVA spectrum, set the width around it to be , , The protection area is set symmetrically in the distance, speed and angle dimensions; the side lengths of the reference area in the distance, speed and angle dimensions are , , ; Step S4-4, noise amplitude estimation: Calculate the weighted value of the signal strength of all units in the current detection area within the time window length Q as the average background clutter level Z ; Step S4-5: According to the false alarm probability , the volume of the reference area and the time window length Q, determine the threshold factor ; Step S4-6: For each detection unit, use the threshold factor and the average background clutter level Z To calculate the detection threshold ; Step S4-7, target detection: compare the signal strength of each detection unit with the detection threshold If the signal strength of the detection unit exceeds the detection threshold, , it is considered that there is a target in the detection unit, and the position of the detection unit is recorded as the target position.

[0019] Preferably, the expression of the average background clutter level Z is as follows:

[0020] in, represents the time-weighted forgetting factor, Indicates q The distance dimension index within the current detection area at the moment i , speed dimension index j , Angle dimension index k The signal strength of the unit, Represents the distance dimension index within the current detection area i , speed dimension index j , Angle dimension index k The weight value of whether the unit is located in the reference area. When the unit is a detection unit, , when the unit is within the protection area, ; When the cell is within the reference region, , represents the distance-weighted standard deviation, represents the speed-weighted standard deviation, represents the angle-weighted standard deviation;

[0021]

[0022]

[0023] in, represents the speed of light, Represents the 5G AeroMACS signal subcarrier spacing, Represents 5G AeroMACS signal OFDM symbol time.

[0024] Preferably, the threshold factor The expression is:

[0025] in, Represents the expectation of computing the weight value tensor.

[0026] Preferably, the detection threshold The expression is as follows: .

[0027] Compared with the prior art, the present invention has the following advantages: 1. The present invention proposes a multi-dimensional super-resolution target perception method based on 5G AeroMACS, which applies 5G AeroMACS technology and OFDM technology, as well as the characteristics of synaesthesia integration. The high-bandwidth and low-latency data transmission capabilities provided by 5G AeroMACS ensure real-time transmission and processing of monitoring data. By mapping the data to be transmitted to OFDM symbols, using subcarriers in the aviation-specific frequency range of 5091-5150MHz, IFFT operations are performed to convert frequency domain signals into time domain signals, and cyclic prefixes are added to prevent inter-symbol interference caused by multipath effects. Finally, the modulated OFDM symbols are sent to the air through the transmitting antenna, making full use of the low latency, high reliability, and large bandwidth characteristics of 5G technology.

[0028] 2. The present invention proposes a multi-dimensional super-resolution target perception method based on 5G AeroMACS, which divides the three-dimensional complex matrix into multiple sub-tensors, and performs covariance matrix calculation and eigenvalue decomposition to construct signal subspace and noise subspace. By constructing a three-dimensional delay Doppler search column vector and a three-dimensional spectrum peak search function, the target is accurately positioned. In the target detection stage, a three-dimensional constant false alarm detection algorithm is adopted. By setting the false alarm probability, preprocessing the RVA spectrum, determining the protection area and the reference area, estimating the noise amplitude, and calculating the detection threshold, the target is detected with high accuracy.

[0029] 3. The multi-dimensional super-resolution target perception method based on 5G AeroMACS proposed in the present invention improves the accuracy and reliability of target perception through multi-dimensional data processing and super-resolution technology. At the same time, by optimizing the signal transmission and reception process and adopting advanced signal processing algorithms, the present invention can achieve rapid and accurate detection of targets in complex airport environments, providing strong technical support for airport safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. The features and advantages of the present invention can be more clearly understood by referring to the drawings. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0031] Figure 1 It is a flow chart of the multi-dimensional super-resolution target perception method based on 5G AeroMACS of the present invention.

[0032] Figure 2 It is a deep learning model diagram used in the present invention to estimate the number of signal subspace dimensions.

[0033] Figure 3 It is a schematic diagram of a three-dimensional complex matrix and a three-dimensional subtensor in the present invention.

[0034] Figure 4 It is a schematic diagram of the division of the data tensor detection area, protection area, and reference area in the present invention.

[0035] Figure 5 This is the estimation result diagram for a single target scenario (900m, -20m / s, -32 degrees).

[0036] Figure 6 This is the estimation result diagram for the dual-target scenario (900m, -20m / s, -32 degrees; 600m, -30m / s, 9 degrees; 500m, 40m / s, 20 degrees; 300m, 50m / s, 38 degrees). DETAILED DESCRIPTION

[0037] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0039] Since the size of targets at airports varies greatly, super-resolution technology is particularly needed. It is necessary to jointly identify targets from multiple dimensions such as distance, speed, and angle. Figure 1 As shown in the figure, in 5G AeroMACS, the signal transmission and reception processes follow the physical layer signal transmission process of 5G communication, while combining the characteristics of OFDM technology and synaesthesia integration.

[0040] 1. 5G AeroMACS signal transmission (1) Data preparation: In 5G AeroMACS, the data preparation phase involves mapping the data to be transmitted (such as high-precision digital maps of airports, runway occupancy, real-time aircraft positions, etc.) onto M OFDM symbols. Each OFDM symbol contains N subcarriers, which are in the aviation-specific frequency range of 5091-5150 MHz.

[0041] (2) IFFT operation: Perform IFFT operation on each OFDM symbol to convert the frequency domain signal into a time domain signal. The output of the IFFT operation is a time domain OFDM symbol of length N.

[0042] (3) Adding a cyclic prefix (CP): In order to prevent inter-symbol interference (ISI) caused by multipath effects, a cyclic prefix (CP) is added to the end of each OFDM symbol.

[0043] (4) Signal transmission: The modulated OFDM symbols are sent into the air through the transmitting antenna, taking advantage of the low latency, high reliability, and large bandwidth characteristics of 5G technology.

[0044] 2. 5G AeroMACS signal reception H receiving antennas are used to receive signals. Due to the multipath effect, the signals received by each receiving antenna may be different.

[0045] Each receiving antenna will get a data matrix of dimension N×M, which contains the data of all subcarriers received by the receiving antenna on M OFDM symbols. All H N×M data matrices are stacked to form a three-dimensional data tensor with dimension H×N×M.

[0046] Wherein, H: the number of receiving antennas; N: the number of subcarriers in each OFDM symbol; M: the number of OFDM symbols.

[0047] (1) Initialize the three-dimensional complex matrix CIM_3dfft: Create a three-dimensional complex matrix CIM_3dfft with dimensions (H, M, N).

[0048] The real part of the three-dimensional complex matrix CIM_3dfft is initialized to 0, and the imaginary part is also initialized to 0, and then the imaginary unit 1j is added, so that each element in the three-dimensional complex matrix CIM_3dfft is a complex number 0.

[0049] (2) Traverse each receiving antenna: Use a loop for ii in range(H): to iterate over each receive antenna. For each receive antenna, perform the following steps.

[0050] a) Intercept the received signal of the same length as the transmitted signal: The first TxSignal_cp.shape[0] samples are intercepted from the received signal RxSignal[ii] of the ii-th receiving antenna. This part has the same length as the transmitted signal TxSignal_cp and is stored in the variable Rx0.

[0051] b) Reshape the received signal: Reshape the intercepted variable Rx0 into a matrix of (N, -1) and then transpose it to get the variable Rx1. The -1 here indicates the automatically calculated matrix dimension to ensure that the total number of elements remains unchanged after reshaping.

[0052] c) Remove the cyclic prefix (CP) and reshape into a matrix: The cyclic prefix (CP) is removed from the variable Rx1 to combat the multipath effect. After removing the cyclic prefix (CP), the signal is reshaped into a matrix Rx with dimension (N-CPsize, -1), where CPsize is the size of the cyclic prefix.

[0053] d) Perform FFT transformation on the signal after removing the cyclic prefix (CP): Perform FFT transformation on the matrix Rx to obtain the complex matrix Rx_dem, whose dimension is (N-CPsize, M). FFT transformation converts the time domain signal into the frequency domain signal for the next step of processing.

[0054] e) Transmitted information elimination (matched filtering): Multiply the complex matrix Rx_dem element-wise with the complex conjugate of the transmitted signal np.conj(TxData) to remove the effect of 5G AeroMACS transmission.

[0055] (3) Put the processed data into the corresponding layer of the 3D matrix: The result CIM_2dfft after matched filtering of each receiving antenna is placed in the three-dimensional complex matrix CIM_3dfft, where the result of the ii-th receiving antenna is stored in the ii-th layer of the three-dimensional complex matrix CIM_3dfft. In this way, the processing result of each receiving antenna is stored in the three-dimensional complex matrix CIM_3dfft, which is convenient for subsequent analysis and processing.

[0056] 3. Multidimensional super-resolution spectrogram estimation (1) If Figure 2-Figure 3 As shown, the three-dimensional complex matrix CIM_3dfft is divided into pieces of size The number of rows , number of columns , number of layers If C, S, and D are larger, the sub-tensors are larger and their number is smaller. The number of sub-tensors is .

[0057] The process of segmentation starts from the first layer and the first row, and shifts to the right along the column. When it reaches the rightmost end, it changes to the second row. When a layer is completed, the shift of the next layer begins. According to this process, all sub-tensors are segmented.

[0058] Among them, the first sub-tensor Expanded by layer, it can be expressed as:

[0059] in, Represents the first sub-tensor Middle Layer data, , defined as:

[0060] in, Represents the first c Row, No. s Column, No. d The data values ​​in the layer.

[0061] Among them, the second subtensor Expanded by layer, it can be expressed as:

[0062] in, Represents the second sub-tensor Middle The data of the layer is defined as:

[0063] The last sub-tensor Expanded by layer, it can be expressed as:

[0064] in, Represents the last sub-tensor Middle The data of the layer is defined as:

[0065] (2) Merge all sub-tensors by column, that is, convert the sub-tensors into a column vector.

[0066]

[0067] (3) Therefore, the result of the covariance matrix calculation can be obtained as:

[0068] Where R represents the covariance matrix, Represents the first c Row, No. s Column, No. d The data values ​​in the layer are converted to a column vector.

[0069] (4) Estimate the signal subspace dimension P based on the deep learning model.

[0070] The deep learning model considers the covariance matrix of the continuous time window length Q as input. For the covariance matrix input at each moment, the features of the input covariance matrix are extracted through convolution-activation function-batch normalization operations, and the low-dimensional features are converted into vectors through flattening operations. The vectors at each moment are fused through the long short-term memory network layer, and finally the directly corresponding signal subspace dimension P is output.

[0071] The deep learning model needs to be pre-trained. The data set required for pre-training is generated based on different signal-to-noise ratio conditions, different numbers of targets, or based on 5G AeroMACS synaesthesia echo generation based on actual airport scenes.

[0072] (5) Decompose the eigenvalues ​​of the correlation matrix:

[0073] in, represents the eigenvector matrix, represents a diagonal matrix of eigenvalues.

[0074] The signal subspace is constructed using the eigenvectors corresponding to the largest P eigenvalues , the remaining The corresponding noise subspace is constructed by using the eigenvectors .

[0075] (6) Construct a three-dimensional spectrum peak search function. First, we need to construct a corresponding three-dimensional delay-Doppler search column vector, which can be expressed as:

[0076] in,

[0077]

[0078]

[0079] Therefore, the three-dimensional peak search function can be written as:

[0080] in, Indicates the target delay, Indicates the Doppler frequency shift of the target, Represents the angle of the target, represents the three-dimensional delay-Doppler search column vector, represents the imaginary unit, represents the element spacing of the antenna array, Indicates the wavelength of the 5G AeroMACS signal.

[0081] 4. Multi-dimensional CFAR target detection The final target perception result is obtained through the three-dimensional constant false alarm detection algorithm.

[0082] (1) Setting the false alarm probability .

[0083] (2) Perform necessary preprocessing on the RVA spectrum and fill zeros around it to reduce edge effects.

[0084] (3) If Figure 4 As shown in the figure, determine the current detection area, including the central detection unit and the protection area and reference area: for each detection unit (Cell Under Test, CUT) in the RVA spectrum, set the width around it to , , The protection zone is set symmetrically in the distance, speed and angle dimensions.

[0085] (4) Determine the side lengths of the reference area in the three dimensions of distance, speed and angle: , , .

[0086] (5) Noise amplitude estimation: Calculate the weighted value of the signal strength of all units in the current detection area within the time window length Q, and use this as the average background clutter level Z The formula is:

[0087] in, represents the time-weighted forgetting factor, Indicates q The distance dimension index within the current detection area at the moment i , speed dimension index j , Angle dimension index k The signal strength of the unit, Represents the distance dimension index within the current detection area i , speed dimension index j , Angle dimension index k The weight value of whether the unit is located in the reference area. When the unit is a detection unit, , when the unit is within the protection area, ; When the cell is within the reference region, , represents the distance-weighted standard deviation, represents the speed-weighted standard deviation, represents the angle-weighted standard deviation;

[0088]

[0089]

[0090] in, represents the speed of light, Represents the 5G AeroMACS signal subcarrier spacing, Represents 5G AeroMACS signal OFDM symbol time.

[0091] (6) Based on the false alarm probability and the volume of the reference area to determine the threshold factor : In the case of average weighting, the threshold factor The expression is:

[0092] otherwise,

[0093] in, Represents the expectation of computing the weight value tensor.

[0094] (7) For each detection unit, use the threshold factor And the corresponding average background clutter level Z to calculate the detection threshold The formula is:

[0095] (8) Target detection: Compare the signal strength of each detection unit with the corresponding detection threshold. If the signal strength of the detection unit exceeds the detection threshold, it is considered that there is a target in the unit, and the position of the detection unit is recorded as the target position.

[0096] Example 1 Assume that the 5G AeroMACS base station operates at 5.1GHz, and the transmitted OFDM signal frame contains 64 OFDM symbols, each OFDM symbol contains 64 subcarriers, the subcarrier spacing is 60kHz, the OFDM symbol time is 16.67μs, the cyclic prefix length accounts for 25%, and the cyclic prefix time is 4.17μs. The transmitted data is modulated using 16QAM, and the array antenna is a linear array containing 16 array elements with an array element spacing of half a wavelength. When calculating the spectrum, the number of spectrum points in the distance dimension is set to 32, the number of spectrum points in the velocity dimension is set to 32, and the number of spectrum points in the angle dimension is set to 32.

[0097] Single target scenario: There is one target in the signal coverage area, and the parameters are: distance 900m, radial speed -20m / s, angle -32 degrees.

[0098] The target number obtained by covariance matrix inference is 1, and the obtained three-dimensional super-resolution spectrum is as follows Figure 5 , showing the profiles of the three-dimensional spectrum along the angle dimension, which are -37.74 degrees, -31.94 degrees and -26.13 degrees respectively. It can be seen that in the profile (b) where the angle value is closest to the target, a sharp spectral peak is formed, while the peaks in other profiles (a) and (c) are lower than (b).

[0099] The point cloud distance index obtained after target detection is 11, the speed index is 15, and the angle index is 10.

[0100] Multi-target scenario: There are 4 targets in the signal coverage area, and the parameters are: distance 900m, radial speed -20m / s, angle -32 degrees Distance 600m, radial speed -30m / s, angle 9 degrees Distance 500m, radial speed 40m / s, angle 20 degrees Distance 300m, radial speed 50m / s, angle 38 degrees The number of targets obtained by covariance matrix inference is 4, and the obtained three-dimensional super-resolution spectrum is as follows Figure 6 The profiles of the three-dimensional spectrum along the angle dimension are shown, which are -60.97 degrees, -31.94 degrees, 8.71 degrees, 20.32 degrees, 37.74 degrees and 60.97 degrees. There is no obvious peak in the angle value diagram (a) and (f) of the target principle, but in the angle value profiles (b), (c), (d) and (e) where the target is closest, sharp spectrum peaks are formed at the correct distance and correct speed position of the target.

[0101] The point cloud distance indexes obtained after target detection are 11, 7, 6, 4, the speed indexes are 15, 15, 16, 17, and the angle indexes are 10, 17, 19, 22.

[0102] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-dimensional super-resolution target perception method based on 5G AeroMACS, characterized in that: The following steps are involved: Step S1, using 5G AeroMACS to modulate the data to be transmitted into multiple OFDM symbols and transmit them; Step S2, multiple receiving antennas receive OFDM symbols in the air, perform data restoration and stack them into a three-dimensional data tensor, and store them in a three-dimensional complex matrix; Step S3, constructing a signal subspace, a noise subspace and a three-dimensional spectrum peak search function based on the three-dimensional complex matrix to obtain a multi-dimensional super-resolution spectrum; Step S4: adopt a three-dimensional constant false alarm detection algorithm to obtain the final target perception result.

2. The multi-dimensional super-resolution target perception method according to claim 1, characterized in that: The step S1 specifically includes: Step S1-1, data preparation: using 5G AeroMACS to map the data to be transmitted into M OFDM symbols; Step S1-2, IFFT operation: perform IFFT operation on each OFDM symbol to convert the frequency domain signal into a time domain signal; Step S1-3, adding a cyclic prefix: adding a cyclic prefix at the end of each OFDM symbol; Step S1-4, signal transmission: the modulated OFDM symbols are sent into the air through a transmitting antenna.

3. The multi-dimensional super-resolution target perception method according to claim 2, characterized in that: In step S1-1, each OFDM symbol includes N subcarriers, and the subcarriers are within the aviation-specific frequency range of 5091-5150 MHz.

4. The multi-dimensional super-resolution target perception method according to claim 3, characterized in that: The step S2 specifically includes: Step S2-1, H receiving antennas each receive an OFDM symbol in the air; Step S2-2, sequentially intercepting, reshaping, removing cyclic prefixes, matrixing, performing FFT transformation and matched filtering on the OFDM symbols of each receiving antenna; Step S2-3: Store the result of matched filtering into a three-dimensional complex matrix with dimensions of H×N×M.

5. The multi-dimensional super-resolution target perception method according to claim 4, characterized in that: The step S2-2 specifically includes: (1) Extract the first F samples of the OFDM symbol of the receiving antenna with the same length as the transmitted OFDM symbol and store them in the variable Rx0; (2) Reshape the variable Rx0 into a (N, -1) matrix and then transpose it to obtain the variable Rx1, where -1 represents the automatically calculated matrix dimension to ensure that the total number of elements remains unchanged after reshaping; (3) Remove the cyclic prefix from the variable Rx1 and reshape it into a matrix Rx with dimension (N-CPsize,-1), where CPsize is the size of the cyclic prefix; (4) Perform FFT transformation on the matrix Rx to obtain the complex matrix Rx_dem, whose dimension is (N-CPsize,M); (5) Perform element-by-element multiplication of the complex matrix Rx_dem with the complex conjugate of the transmitted OFDM symbol to eliminate the influence of 5G AeroMACS transmission.

6. The multi-dimensional super-resolution target perception method according to claim 5, characterized in that: The step S3 specifically includes: Step S3-1, dividing the three-dimensional complex matrix into The size is A sub-tensor of , , ; Step S3-2: Expand each sub-tensor and merge by column to obtain column vector; Step S3-3, calculating the covariance matrix; Step S3-4, based on the covariance matrix of the time window length Q, using the deep learning model to estimate the signal subspace dimension P; Step S3-5: Decompose the covariance matrix into eigenvalues ​​and construct the signal subspace using the eigenvectors corresponding to the largest P eigenvalues. , the remaining The noise subspace is constructed by using eigenvectors ; Step S3-6: construct a three-dimensional spectrum peak search function to obtain an RVA spectrum.

7. The multi-dimensional super-resolution target perception method according to claim 6, characterized in that: The expression of the three-dimensional spectrum peak search function in step S3-6 is as follows: in, Indicates the target delay, Indicates the Doppler frequency shift of the target, Represents the angle of the target, represents the three-dimensional delay-Doppler search column vector; in, represents the imaginary unit, represents the element spacing of the antenna array, Indicates the wavelength of the 5G AeroMACS signal.

8. The multi-dimensional super-resolution target perception method according to claim 7, characterized in that: The step S4 specifically includes: Step S4-1: Setting the false alarm probability ; Step S4-2, preprocessing the RVA spectrum; Step S4-3, determine the current detection area, including the central detection unit and the protection area and reference area: for each detection unit in the RVA spectrum, set the width around it to be , , The protection area is set symmetrically in the distance, speed and angle dimensions; the side lengths of the reference area in the distance, speed and angle dimensions are , , ; Step S4-4, noise amplitude estimation: Calculate the weighted value of the signal strength of all units in the current detection area within the time window length Q as the average background clutter level Z ; Step S4-5: According to the false alarm probability , the volume of the reference area and the time window length Q, determine the threshold factor ; Step S4-6: For each detection unit, use the threshold factor and the average background clutter level Z To calculate the detection threshold ; Step S4-7, target detection: compare the signal strength of each detection unit with the detection threshold If the signal strength of the detection unit exceeds the detection threshold, , it is considered that there is a target in the detection unit, and the position of the detection unit is recorded as the target position.

9. The multi-dimensional super-resolution target perception method according to claim 8, characterized in that: The expression of the average background clutter level Z is as follows: in, represents the time-weighted forgetting factor, Indicates q The distance dimension index within the current detection area at the moment i , speed dimension index j , Angle dimension index k The signal strength of the unit, Represents the distance dimension index within the current detection area i , speed dimension index j , Angle dimension index k The weight value of whether the unit is located in the reference area. When the unit is a detection unit, , when the unit is within the protection area, ; When the cell is within the reference region, , represents the distance-weighted standard deviation, represents the speed-weighted standard deviation, represents the angle-weighted standard deviation; in, represents the speed of light, Represents the 5G AeroMACS signal subcarrier spacing, Represents the OFDM symbol time of 5G AeroMACS signal; The threshold factor The expression is: in, Represents the expectation of computing the weight value tensor.

10. The multi-dimensional super-resolution target perception method according to claim 9, characterized in that: Detection threshold The expression is as follows: 。

Images