A robust data domain space-time dimensionality reduction processing method

Abstract

The application provides a kind of robust data domain space-time dimension reduction processing method, comprising: establishing radar array echo signal model, to obtain radar echo data;The radar echo data is transformed to range-Doppler domain, to generate range-Doppler domain echo data;Establish the first optimization model, to generate the robust dimension reduction matrix coefficient corresponding to different Doppler channel;And construct dimension reduction matrix to different Doppler channel;According to dimension reduction matrix, the range-Doppler domain echo data is processed, and the correlation coefficient between the data after dimension reduction in different range units is calculated, for each Doppler channel and range unit data, select out non-homogeneous distribution sample;Establish the second optimization model, for each Doppler channel and range unit data, according to the selected non-homogeneous distribution sample, to solve the corresponding space-time processing weight coefficient;Through space-time processing weight coefficient, the Doppler channel and range unit data are linearly combined into single-channel output signal.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing, and in particular to a robust data domain spatiotemporal dimensionality reduction processing method. Background Technology

[0002] Space-Time Adaptive Processing (STAP) has garnered significant attention since its inception due to its ability to improve clutter suppression and target detection performance. According to the Reed-Mallet-Brennan (RMB) criterion, sample-averaged covariance-based space-time processing methods require at least twice the number of independent, identically distributed (IID) samples to achieve a desired signal-to-noise ratio (SINR) loss within 3 dB. Meeting this condition is impractical in most space-time signal processing scenarios, especially for medium to large-scale arrays. Furthermore, matrix inversion requires high computational complexity and necessitates storage space for filter computation.

[0003] To address the main challenges of space-time processing—namely, the few-sample problem and the need for low computational complexity—many STAP algorithms have been proposed over the past few decades.

[0004] Full-dimensional STAP utilizes all temporal and spatial array elements, but for large-scale arrays, the dimension of the received data is often very high, sometimes reaching tens of thousands, resulting in a huge computational burden for covariance matrix inversion, making real-time processing difficult in practical engineering. Furthermore, STAP estimates the covariance matrix using training samples that are adjacent to the target cell and whose statistical characteristics satisfy the Independent and Identical Distributed (IID) condition. According to the RMB criterion, to keep the performance loss due to inaccurate estimation within 3dB, the number of samples satisfying the IID condition must be at least twice the order of the covariance matrix. In addition, radar clutter exhibits severe non-stationarity at close range, causing the statistical characteristics of the training samples to differ from those of the target cell samples, making it difficult to obtain IID samples, which severely impacts STAP performance. In practical engineering applications, full-dimensional STAP faces the problems of high computational complexity and a severe shortage of training samples.

[0005] Dimensionality reduction STAP algorithms include coefficient algorithms or extended coefficient algorithms, joint processing schemes for multiple Doppler channels, joint domain localization algorithms, and spatiotemporal multibeam algorithms. The core of dimensionality reduction spatiotemporal processing lies in reducing the dimensionality of the spatiotemporal signal by employing linear mappings, thereby reducing computational complexity and the number of samples required. Therefore, the stability of dimensionality reduction algorithms depends on the design of the dimensionality reduction matrix. On the other hand, rank-reduced STAP algorithms, such as principal component analysis, transspectral metrics, and multi-level Wiener filters, can provide more stable performance by using samples with twice the clutter rank. While dimensionality reduction (rank) STAP reduces the need for training samples, it cannot be used for conditions where the signals are not independent and identically distributed.

[0006] In summary, fast and robust spatiotemporal signal processing under the condition of a small amount of non-independent and identically distributed sample training data is an engineering application problem that urgently needs to be solved. Summary of the Invention

[0007] Therefore, it is necessary to provide a robust data domain spatiotemporal dimensionality reduction method to address the aforementioned technical problems, so as to achieve fast and robust spatiotemporal signal processing under the condition of a small number of non-independent and identically distributed sample training data.

[0008] This application provides a robust data domain spatiotemporal dimensionality reduction method, including:

[0009] Establish a radar array echo signal model, and obtain radar echo data based on the radar array echo signal model;

[0010] The radar echo data is transformed to the range-Doppler domain to generate range-Doppler domain echo data;

[0011] A first optimization model is established, and under the first preset constraint, robust dimension reduction matrix coefficients corresponding to different Doppler channels are generated through the first optimization model.

[0012] Based on the coefficients of the robust dimensionality reduction matrix, a dimensionality reduction matrix is ​​constructed for the different Doppler channels;

[0013] The distance-Doppler domain echo data is dimensionality reduced according to the dimensionality reduction matrix, and the correlation coefficient between different distance cells is calculated. For each Doppler channel and distance cell data, the data with a correlation coefficient greater than a preset threshold are selected as non-uniformly distributed samples.

[0014] A second optimization model is established. For each Doppler channel and distance cell data, the selected non-identically distributed samples are substituted into the second optimization model to solve for the corresponding spatiotemporal processing weight coefficients.

[0015] The Doppler channel and range cell data are linearly synthesized into a single-channel output signal using the spatiotemporal processing weighting coefficients.

[0016] In one embodiment, based on the established radar array echo signal model, the corresponding radar echo data is represented as follows:

[0017]

[0018] Among them, y l，c Indicates the received clutter signal, n l This represents noise, where L is the distance unit. The modulation parameters of the target reflection are represented, a0 represents the steering vector corresponding to the target, N represents N array elements, M represents M pulses, and l represents the echo signal of the l-th range element.

[0019] In one embodiment, the transformation of the radar echo data to the range-Doppler domain to generate range-Doppler domain echo data is expressed as follows:

[0020]

[0021] Among them, I N Represented as an identity matrix of dimension N, denoted as Kronecker product, and F denotes the DFT matrix.

[0022] In one embodiment, based on the radar array echo signal model, the Represented as:

[0023]

[0024] in,

[0025] In one embodiment, the first optimization model is represented as:

[0026] min W ||W|| st|W H～ a0-1|+δ||W||≤C0;

[0027] Where H represents the conjugate transpose, δ represents the pre-set robustness coefficient, and C0 represents the loss control parameter.

[0028] In one embodiment, the dimension reduction matrix, corresponding to the dimension reduction data matrix of the k-th Doppler channel, is represented as:

[0029]

[0030] in, For the robust dimensionality reduction matrix coefficients, I represents the identity matrix, T represents the matrix in [...], 0 represents the zero matrix, and m represents a natural number.

[0031] In one embodiment, the correlation coefficient is calculated using the following formula:

[0032]

[0033] in, Let represent the dimensionality reduction data in the t-th column of the dimensionality reduction matrix, and α represent the optimization variable.

[0034] In one embodiment, the second optimization model is represented as:

[0035]

[0036] in, σ represents the penalty applied to the average beamforming output, δ represents the penalty applied to noise, L represents the robustness coefficient, and L represents the selected set of range cell indices. Let C0 represent the dimensionality-reduced data corresponding to the selected l-th distance unit, and let C0 be the loss control parameter. Let l be the label of all elements in set L.

[0037] In one embodiment, the single-channel output signal is represented as:

[0038]

[0039] in, The space-time processing weight coefficient is represented by H, which is the conjugate transpose, and represents the dimensionality-reduced data of the selected t-th distance unit.

[0040] In this embodiment, a robust data domain spatiotemporal dimensionality reduction method is provided, comprising: establishing a radar array echo signal model; acquiring radar echo data based on the radar array echo signal model; transforming the radar echo data to the range-Doppler domain to generate range-Doppler domain echo data; establishing a first optimization model; generating robust dimensionality reduction matrix coefficients corresponding to different Doppler channels through the first optimization model under a first preset constraint; constructing a dimensionality reduction matrix for the different Doppler channels based on the robust dimensionality reduction matrix coefficients; performing dimensionality reduction processing on the range-Doppler domain echo data based on the dimensionality reduction matrix, and calculating the correlation coefficient between the dimensionality-reduced data and different range cells; for each Doppler channel and range cell data, selecting data with a correlation coefficient greater than a preset threshold as non-identically distributed samples; establishing a second optimization model; for each Doppler channel and range cell data, substituting the selected non-identically distributed samples into the second optimization model to solve for the corresponding spatiotemporal processing weight coefficients; and linearly synthesizing the Doppler channel and range cell data into a single-channel output signal through the spatiotemporal processing weight coefficients. In this embodiment, by introducing robust dimensionality reduction matrix coefficients, the robustness to array steering vector errors is improved, which can overcome the difficulty of training data with a small number of non-independent and identically distributed samples, and can reduce computational complexity, thus meeting the needs of engineering applications. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating a robust data domain spatiotemporal dimensionality reduction method provided in one embodiment of the present invention;

[0043] Figure 2 This is a range-Doppler map of a single array element provided in one embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the target signal and the original echo signal added by the 180-range cell in one embodiment of the present invention on a single array element;

[0045] Figure 4 This is a range-Doppler spectrum output of beamforming using the method described in one embodiment of the present invention.

[0046] Figure 5This is a range-Doppler spectrum output of beamforming using the 3DT method in one embodiment of the present invention.

[0047] Figure 6 This is a schematic diagram of the Doppler channel outputs of a 200-distance unit provided in one embodiment of the present invention;

[0048] Figure 7 This is a schematic diagram of the output of each Doppler channel of a 350-meter distance unit provided in one embodiment of the invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] In one embodiment, such as Figure 1 As shown, a robust data domain spatiotemporal dimensionality reduction method is provided, including the following steps:

[0051] In step S110, a radar array echo signal model is established, and radar echo data is obtained based on the radar array echo signal model.

[0052] In this embodiment of the application, assuming the radar array contains N array elements and M pulses, the echo signal of the l-th range element can be represented by the radar array echo signal model as follows:

[0053]

[0054] Among them, y l，c Indicates the received clutter signal, n l This represents noise, where L is the distance unit. The modulation parameters for target reflection are represented by a0, the preset target-correlation vector is represented by N, the number of array elements is represented by M, and l is represented by the echo signal of the l-th range element.

[0055] In step S120, the radar echo data is transformed to the range-Doppler domain to generate range-Doppler domain echo data;

[0056] In this embodiment, the radar echo data is transformed to the range-Doppler domain, and the echo signal of the l-th range cell can be transformed as follows:

[0057]

[0058] Among them, IN Represented as an identity matrix of dimension N, The product of Kronecker products is represented by F, and the DFT matrix is ​​represented by F.

[0059] Furthermore, based on the radar array echo signal model established above, the echo signal of the l-th range cell... This can be further expressed as:

[0060]

[0061] in,

[0062] In step S130, a first optimization model is established, and under the first preset constraint, robust dimension reduction matrix coefficients corresponding to different Doppler channels are generated through the first optimization model.

[0063] In this embodiment of the application, the first optimization model is represented as:

[0064] min W ||W|| st|W H～ a0-1|+δ||W||≤C0;

[0065] Where H represents the conjugate transpose, δ represents the pre-set robustness coefficient, and C0 represents the loss control parameter.

[0066] The first preset constraint condition st|W H～ a0-1|+δ||W||≤C0 is specifically used for robust constraints on w. This constraint can ensure that w that satisfies this constraint can retain the target signal even when there is a certain disturbance in a0.

[0067] In this embodiment, W represents the optimization problem of the first optimization model, and the optimal solution obtained through the first optimization model is the coefficient of the robust dimensionality reduction matrix. in, These represent the coefficients of the robust dimensionality reduction matrix corresponding to different Doppler channels.

[0068] In step S140, a dimension reduction matrix is ​​constructed for different Doppler channels based on the robust dimension reduction matrix coefficients;

[0069] In this embodiment, the radar echo signal can correspond to multiple different target signal steering vectors a0 in the Doppler domain, i.e., multiple different Doppler channels. Some positions in a radar echo signal y can be represented as a Doppler channel. Through robust dimensionality reduction matrix coefficients, some Doppler channels can be merged, while other Doppler channels can be retained without dimensionality reduction.

[0070] In this embodiment of the application, a dimension reduction matrix is ​​constructed for different Doppler channels based on the robust dimension reduction matrix coefficients, including:

[0071] 7. The dimension-reduced data matrix corresponding to the k-th Doppler channel is represented as follows:

[0072]

[0073] in, Let i be the coefficients of the robust dimensionality reduction matrix calculated above, i is 1....m, and m is the given weight coefficient, which means that the data from other Doppler channels are combined into a single channel data according to a certain fixed weight. I represents the identity matrix, T represents the matrix in [...], 0 represents the 0 matrix, and m represents the Doppler channel m.

[0074] For example, the corresponding target guidance vector

[0075] In this embodiment, taking the k-th Doppler channel as an example, the dimensionality-reduced data after the above dimensionality reduction matrix processing has 2m+1 adjacent auxiliary Doppler channels as auxiliary channels, for example, from mk to m+k. Simultaneously, the remaining unselected channel data are summed according to a fixed weight to form a composite channel data.

[0076] Where m represents a natural number.

[0077] The auxiliary channels can be reserved to preserve the target signal as much as possible. The remaining unselected channel data are added together with a fixed weight to form a composite channel data, since the other channels have almost no target signal components, thus reducing the amount of subsequent calculations.

[0078] In step S150, the range-Doppler domain echo data is dimensionality reduced according to the dimensionality reduction matrix, and the correlation coefficient between different range cells is calculated. For each Doppler channel and range cell data, the data with a correlation coefficient greater than a preset threshold are selected as non-uniformly distributed samples.

[0079] In the embodiments of this application, non-uniformly distributed samples are non-independent, identically distributed data.

[0080] In this embodiment of the application, the dimensionality-reduced data has many columns, and each column corresponds to a distance unit. The distance unit data is the data in the column corresponding to each distance unit.

[0081] In this application embodiment, the commonly used correlation coefficient is the Pearson coefficient. The Pearson coefficient characterizes the degree of linear correlation between data at different distance units. The larger the Pearson coefficient, the more similar the two data points are. This correlation coefficient, Pearson, can be calculated using the following formula:

[0082]

[0083] in, This represents the dimensionality reduction data in the t-th column of the generated dimensionality reduction matrix, where t represents the t-th distance unit, and α represents the optimization variable. This represents the target guidance vector. Then, you can select... The coefficient is greater than the preset threshold. As non-uniformly distributed samples. The preset threshold can be set according to the actual situation, and this application does not limit it.

[0084] In step S160, a second optimization model is established. For each Doppler channel and distance cell data, the selected non-uniformly distributed samples are substituted into the second optimization model to solve the corresponding spatiotemporal processing weight coefficients.

[0085] In this embodiment of the application, the second optimization model can be expressed as:

[0086]

[0087] in, This indicates a penalty applied to the average beamforming output, where σ is the penalty for noise, δ represents the robustness coefficient, and L represents the selected set of range cell indices. Let C0 represent the dimensionality-reduced data corresponding to the selected l-th distance unit, and let C0 be the loss control parameter. Let l be the label of all elements in set L.

[0088] By solving the above optimization model under the second preset constraint, the space-time processing weight coefficients can be obtained. The second preset constraint can be a robust constraint on w. This ensures that even with a certain disturbance in a0, w satisfying this constraint can still retain the target signal. This indicates that the output energy of other range cells is limited to less than epsilon to ensure the clutter suppression effect of w. Therefore, by solving for this space-time processing weight coefficient... It can effectively suppress clutter components in the single-channel output z.

[0089] In step S170, the Doppler channel and range cell data are linearly synthesized into a single-channel output signal using the space-time processing weighting coefficients.

[0090] In this embodiment of the application, the single-channel output signal can be represented as:

[0091]

[0092] in, Here, H is the conjugate transpose, t represents the distance to the t-th unit, and y is the space-time processing weight coefficient. t This represents the dimensionality-reduced data of the distance between the selected t-th cell and the selected cell.

[0093] In this embodiment of the application, yt can be the above-mentioned

[0094] In this embodiment, by solving the spatiotemporal weight coefficients, the selected dimensionality-reduced data is synthesized into a single-channel output signal using the spatiotemporal weight coefficients. This effectively suppresses clutter components in the single-channel output z. Furthermore, the introduction of robust dimensionality reduction matrix coefficients improves the robustness to array steering vector errors. This overcomes the difficulty of training data with a small number of non-independent and identically distributed samples, and reduces computational complexity, thus meeting the needs of engineering applications.

[0095] In the embodiments of this application, single-channel signal synthesis can be achieved by performing calculations on different Doppler channels and distance unit data through the above steps 150-170.

[0096] In this embodiment, a robust data domain spatiotemporal dimensionality reduction method is provided, comprising: establishing a radar array echo signal model; acquiring radar echo data based on the radar array echo signal model; transforming the radar echo data to the range-Doppler domain to generate range-Doppler domain echo data; establishing a first optimization model; generating robust dimensionality reduction matrix coefficients corresponding to different Doppler channels through the first optimization model under a first preset constraint; constructing a dimensionality reduction matrix for the different Doppler channels based on the robust dimensionality reduction matrix coefficients; performing dimensionality reduction processing on the range-Doppler domain echo data based on the dimensionality reduction matrix, and calculating the correlation coefficient between the dimensionality-reduced data and different range cells; for each Doppler channel and range cell data, selecting data with a correlation coefficient greater than a preset threshold as non-identically distributed samples; establishing a second optimization model; for each Doppler channel and range cell data, substituting the selected non-identically distributed samples into the second optimization model to solve for the corresponding spatiotemporal processing weight coefficients; and linearly synthesizing the Doppler channel and range cell data into a single-channel output signal through the spatiotemporal processing weight coefficients. In this embodiment, by introducing robust dimensionality reduction matrix coefficients, the robustness to array steering vector errors is improved, which can overcome the difficulty of training data with a small number of non-independent and identically distributed samples, and can reduce computational complexity, thus meeting the needs of engineering applications.

[0097] To verify the effectiveness of the robust data domain spatiotemporal dimensionality reduction method proposed in this embodiment, the following simulation experiments are conducted to further demonstrate its effectiveness.

[0098] For echo data with 300 elements, 128 pulses, and a range element count of 600, the echo distance of a single element is plotted using the Doppler graph. Figure 2 As shown, the Doppler domain is loaded with an 80dB Chebyshev window.

[0099] Due to the presence of amplitude and phase errors in the array channels, a 6*50 planar array was used. Virtual target signals (corresponding to normalized Doppler frequencies of 0.35, 0.7, and 0.9) were added at range cells 180, 200, and 350, respectively, with a signal-to-noise ratio (SNR) of 0 dB. The amplitude response error of each array element followed a zero-mean Gaussian distribution with a standard deviation of 0.05. The phase error followed a zero-mean Gaussian distribution with a standard deviation of 0.02. Spatial-time processing was performed on adjacent Doppler channels, reducing the 300 array elements to 12 subarrays. Data from the 30 adjacent cells of each range cell were used as sample data.

[0100] First, the spatiotemporal processing effects in the noise floor region are compared. A schematic diagram of the target signal added at a 180-range cell and the original echo signal on a single array element is shown below. Figure 3As shown in the figure. In the figure, target Signal represents the target signal, and Clutter Signal represents the original echo signal.

[0101] The output of the space-time processing beamforming using this method and the traditional 3DT clutter suppression method (3DT) is as follows: Figures 4-5 As shown, the outputs of each Doppler channel corresponding to distance units 200 and 350 are respectively as follows: Figures 6-7 As shown, the range-Doppler spectrum results from the beamforming output show that this method has a better overall effect in suppressing sidelobe clutter. Comparing the Doppler cross-sections of each range cell also shows that both methods effectively preserve the target signal, but this method has a better effect in suppressing clutter in other Doppler channels.

[0102] Furthermore, a quantitative comparison of the signal-to-clutter and noise ratio (SCNR) of the beamforming output was employed. The average values ​​of the target's range cell and the three range cells and six Doppler channels surrounding the Doppler channel were calculated and used as the estimated clutter and noise power. The beamforming output SCNR of each method under different SNR conditions is shown in the table below, where target 1 corresponds to a range cell of 180, target 2 corresponds to a range cell of 200, and target 3 corresponds to a range cell of 350.

[0103]

[0104] As can be seen from the table above, the scheme (prop) provided in this application can achieve a higher SCNR. Therefore, the overall effect of this application in suppressing sidelobe clutter is better, and the robustness to array steering vector error is improved. It can overcome the difficulty of training data with a small number of non-independent and identically distributed samples, and can be applied to phased array radar.

[0105] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When executed, these computer-readable instructions can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0106] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0107] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A robust data domain spatiotemporal dimensionality reduction method, characterized in that, The method Establish a radar array echo signal model, and obtain radar echo data based on the radar array echo signal model; The radar echo data is transformed to the range-Doppler domain to generate range-Doppler domain echo data; A first optimization model is established, and under a first preset constraint, robust dimensionality reduction matrix coefficients corresponding to different Doppler channels are generated using the first optimization model. The first optimization model is expressed as follows: ； Where H represents the conjugate transpose, and δ represents a pre-defined robustness coefficient. Represented as loss control parameters; Based on the coefficients of the robust dimensionality reduction matrix, a dimensionality reduction matrix is ​​constructed for the different Doppler channels; The range-Doppler domain echo data is dimensionality reduced according to the dimensionality reduction matrix, and the correlation coefficient between the dimensionality-reduced data and the dimensionality-reduced target steering vector is calculated. For each Doppler channel and range cell data, the data with a correlation coefficient greater than a preset threshold are selected as non-uniformly distributed samples. A second optimization model is established. For each Doppler channel and range cell data, selected non-identically distributed samples are input into the second optimization model to solve for the corresponding spatiotemporal processing weight coefficients. The second optimization model is expressed as: ； in, This indicates a penalty applied to the average beamforming output, and σ represents a penalty applied to noise. For the selected number Dimensionally reduced data corresponding to distance units Represented as all labels in set L L is the number of distance units; The Doppler channel and range cell data are linearly synthesized into a single-channel output signal using the spatiotemporal processing weighting coefficients.

2. The robust data domain spatiotemporal dimensionality reduction method as described in claim 1, characterized in that, Based on the established radar array echo signal model, the corresponding radar echo data is represented as follows: = + +(α )∈ =1,2,......L, in, This indicates the received clutter signal. Let α represent noise, and let α represent the modulation parameter of the target reflection. This represents the guidance vector corresponding to the target. N express N Each array element, M This represents M pulses.

3. The robust data domain spatiotemporal dimensionality reduction method as described in claim 2, characterized in that, The process of transforming the radar echo data to the range-Doppler domain to generate range-Doppler domain echo data is expressed as follows: ； in, Represented as dimension The identity matrix, denoted as Kronecker product, and F denotes the DFT matrix.

4. The robust data domain spatiotemporal dimensionality reduction method as described in claim 3, characterized in that, According to the radar array echo signal model, ； in, , , .

5. The robust data domain spatiotemporal dimensionality reduction method as described in claim 1, characterized in that, The correlation coefficient is calculated using the following formula: = ； in, This represents the dimensionality-reduced data in the t-th column of the dimensionality-reduced matrix.

6. The robust data domain spatiotemporal dimensionality reduction method as described in claim 5, characterized in that, The single-channel output signal is represented as: ; in, Indicates the space-time processing weight coefficient. This represents the dimensionality-reduced data of the selected t-th distance unit.

Images