An adaptive ground moving target indication method

Abstract

The application is suitable for the field of radar signal processing, and provides a kind of adaptive ground moving target detection method, including channel equalization processing to multiple groups of echo data received by multiple channels, to obtain multiple groups of equalization data;Channel rearrangement and vectorization processing are carried out on the multiple groups of equalization data to obtain the original matrix;From the original matrix, the rearranged data with data uniformity less than a first preset value and amplitude probability density greater than or equal to a second preset value are selected to obtain multiple groups of screening data;Target detection is carried out according to the multiple groups of screening data to obtain a detection result reflecting whether there is a target;When the detection result reflects the existence of the target, the optimal weight vector is determined using the multiple groups of screening data;The optimal weight vector is used to determine the radial velocity estimate value of the target.The technical scheme based on the application can effectively improve the local non-uniform region coherence, greatly reduce the target false alarm rate, improve the solving efficiency, and effectively improve the target detection performance in the non-uniform strong clutter background.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of radar signal processing, and particularly relates to an adaptive ground moving target detection method. BACKGROUND

[0002] Synthetic Aperture Radar (SAR) is a kind of high-resolution imaging radar, which can obtain high-resolution radar images similar to optical photographs under extremely low visibility weather conditions. The radar is also called synthetic aperture radar, which utilizes the relative motion between the radar and the target to synthesize a larger equivalent antenna aperture from a smaller real antenna aperture through data processing. The synthetic aperture radar has the characteristics of high resolution, all-weather operation, and effective identification of camouflage and penetration of coverings. In actual work process, the radar transmits a detection signal, receives a return signal, and generates a SAR image after processing the return signal to identify the target. The SAR image is a two-dimensional image.

[0003] However, in actual use process, there are a large number of different and non-uniform clutters in the return signal received by the radar, and the existing radar signal processing method cannot effectively suppress the clutter under the background of non-uniform strong clutter, resulting in that the residual clutter and the target information are mixed, and multiple misidentified targets exist in the finally generated SAR image, which seriously affects the detection performance of the moving target. SUMMARY

[0004] In order to solve the problems of poor clutter suppression effect and high target detection false alarm rate of the existing radar signal processing method under the background of non-uniform strong clutter, the application provides an adaptive ground moving target detection method. The technical problem to be solved by the application is solved through the following technical scheme:

[0005] The application provides an adaptive ground moving target detection method, which comprises: performing channel equalization processing on multiple groups of echo data received by multiple channels to obtain multiple groups of equalization data; performing channel rearrangement and vectorization processing on the multiple groups of equalization data to obtain an original matrix; selecting rearranged data with data uniformity less than a first preset value and amplitude probability density greater than or equal to a second preset value from the original matrix to obtain multiple groups of screening data; performing target detection according to the multiple groups of screening data to obtain a detection result reflecting whether there is a target; when the detection result reflects that there is a target, determining an optimal weight vector by using the multiple groups of screening data; and determining a radial velocity estimation value of the target by using the optimal weight vector.

[0006] The application has the following beneficial technical effects:

[0007] The echo data received by multiple channels is subjected to channel equalization, so that the clutter data in the channels is consistent, and then the multiple equalized data is subjected to channel rearrangement and vectorization processing to obtain an original matrix, so as to effectively improve the interaction between the data, and then the data with high uniformity and large density, that is, the screened data, is selected from the original matrix, the row and column are selected by using the generalized inner product and the amplitude probability distribution characteristic criterion, the RPCA problem is solved by improving the pseudo-frame decomposition method to construct three sub-matrices to efficiently obtain a sparse matrix, and then the sparse matrix is subjected to target detection, in the scene of local strong non-uniform clutter and dense target data pollution, the interference caused by non-uniform samples can be effectively reduced, and the problems of difficult convergence, complex calculation model and large calculation amount are avoided, finally when the detection result shows that there is a target, the optimal weight vector is determined by using the screened data, and then the radial velocity estimation value of the target is determined, and the radial velocity estimation value of the target can be used to obtain the motion trajectory of the moving target. Based on the technical scheme provided in the application, the local non-uniform region coherence can be effectively improved, the sparse low-rank matrix reconstruction effect can be improved, the target false alarm rate can be greatly reduced, the solving efficiency can be improved, and the target detection performance in a non-uniform strong clutter background can be effectively improved.

[0008] The application will be further described in detail below with reference to the drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a flowchart of the adaptive ground moving target detection method provided by the application;

[0010] Figure 2 is an example diagram of channel errors provided by an embodiment of the application;

[0011] Figure 3 is an interference result generated by the echo data collected in Figure 2

[0012] Figure 4 is an interference result generated by the echo data collected in Figure 2

[0013] Figure 5 is an interference result generated by the echo data collected in Figure 2

[0014] Figure 6 is a detection result obtained by detecting the echo data in Figure 2 DETAILED DESCRIPTION

[0015] ​​​​The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0016] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0018] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings, disclosure, and appended claims in carrying out the claimed invention. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0019] To address the problems of poor clutter suppression and high false alarm rate in existing radar signal processing methods under non-uniform strong clutter backgrounds, this invention provides an adaptive ground moving target detection method. This method can effectively suppress clutter, reduce the false alarm rate of target detection, and has the advantages of simple calculation and high efficiency.

[0020] Now combined Figure 1 The technical solution of the adaptive ground moving target detection method provided by the present invention is described. Figure 1 This is a flowchart illustrating the adaptive ground moving target detection method provided by the present invention. Figure 1 As shown, the method includes:

[0021] Step 110: Perform channel equalization processing on multiple sets of echo data received from multiple channels to obtain multiple sets of equalized data.

[0022] Here, multiple sets of echo data are received under conditions of strong local non-uniform clutter and dense target data contamination. The echo data includes clutter data, Gaussian white noise data, and target data. These multiple sets of echo data can be represented as: z(p,u) = s(p,u) + c(p,u) + n(p,u); where z(p,u) refers to the multiple sets of echo data received from multiple channels, with each channel corresponding to one set of echo data; s(p,u) refers to the target data; c(p,u) refers to the clutter data; n(p,u) refers to the Gaussian white noise; and (p,u) represents the pixel range-azimuth index of the SAR image. After receiving the echo data, due to the strong non-uniformity of the clutter data received from multiple channels, channel equalization processing is required to effectively suppress the clutter. This ensures that the clutter data in multiple channels is essentially consistent, allowing for subsequent cancellation using different channels to suppress the clutter data.

[0023] Here, multiple sets of echo data z1(p,u), z2(p,u), ..., z N After channel equalization processing of (p,u), multiple sets of equalized data z are obtained. r_1 (p,u),z r_2 (p,u),…,z r_N (p,u). Here, multiple sets of echo data are pixel data from SAR images.

[0024] It should be noted that channel equalization only applies to clutter data and does not modify the phase difference of the target data. Therefore, preprocessing of the echo data is required before performing channel equalization.

[0025] Here, the flight direction of the SAR radar is defined as the azimuth direction. The direction of the antenna aiming line is orthogonal to the flight path and is called the cross-track direction or range direction. The cross-track direction is the direction of pulse transmission, which provides the tilt range along the flight path to the target. These two directions provide the basis for the size required to generate an image from the area within the antenna beamwidth during the duration of the data collection window. The azimuth of the SAR radar defines the range of the radar platform along the flight path. Fast time defines the duration of each pulse operation. Slow time defines the time it takes for the pulse to travel along the flight path. Based on the angle θ between the beam center pointing and the flight path, the multi-channel synthetic aperture radar operates in a front-side-looking configuration (θ = 90 degrees), with the radar at an altitude of H and flying along the X-axis at a velocity v. N channels with a uniform spacing of d are distributed along the track direction, with the first channel serving as the reference channel for transmitting signals, and the N channels receiving echoes. Assume that at azimuth zero time, a moving target moves from a to b, and its azimuth velocity is expressed as v.a The radial projected velocity is expressed as v c If a linear frequency modulated signal is transmitted between channels, the echo data received by the nth channel can be expressed as:

[0026]

[0027] in, This refers to the echo data received by the nth channel at time t, where rect(·) represents the echo distance envelope, |·| represents the absolute value, and t represents the fast time. m It refers to slow time, t m =m / f PRF Where m is the pulse index, f PRF The pulse repetition frequency, The instantaneous slant distance refers to the distance between the target and the phase center of the nth channel, where λ is the signal wavelength, c is the speed of light, j is the complex part, and w a (·) refers to the directional window function, μ refers to the frequency modulation, and T p It refers to the duration of the transmitted pulse, and exp(·) is an exponential function.

[0028] By performing channel equalization processing on multiple sets of echo data received from multiple channels in step 110, the sensitivity of channel error can be effectively reduced, thereby weakening the interference of strong non-uniform clutter on target identification and improving the data coherence between channels in the global and local areas, making it easier to obtain independent and uniform echo data for target identification.

[0029] Step 120: Perform channel rearrangement and vectorization on multiple sets of balanced data to obtain the original matrix.

[0030] Here, for multiple sets of balanced data z r_1 (p,u),z r_2 (p,u),…,z r_N (p,u) Channel rearrangement refers to arranging the balanced data from a single channel into multiple rearranged data sets z. r (p,u). The multiple sets of rearranged data can be represented as: z r (p,u)=[z r_1 (p,u),z r_2 (p,u),…,z r_N (p,u)] Τ .in,(·) T This refers to the transpose operation. The rearranged data forms the original matrix D, where D = [vec(z...]. r_1 ),vec(z r_2 ),…,vec(z r_N )], vec(·) represents the vectorization operation.

[0031] Channel rearrangement can improve the interactivity between data, thereby improving target detection performance.

[0032] Step 130: Select rearranged data from the original matrix whose data uniformity is less than the first preset value and whose amplitude probability density is greater than or equal to the second preset value, to obtain multiple sets of filtered data.

[0033] Here, because the echo data is received in a scenario with strong non-uniform clutter and dense target data contamination, it is very easy to miss some target data and mistakenly identify some non-uniform clutter as target data during target identification. To address this situation, it is necessary to select uniform data from multiple sets of balanced data for identification, eliminating the identification interference caused by non-uniform data. Specifically, step 130 includes: calculating the data uniformity of each set of rearranged data using a generalized inner product algorithm to obtain multiple data uniformities; selecting data uniformities less than the first preset value from the multiple data uniformities to obtain the uniformity corresponding to each set of rearranged data; calculating the amplitude of each set of rearranged data to obtain the probability density distribution corresponding to the amplitude of each set of rearranged data; selecting amplitude probability densities greater than or equal to the second preset value from the multiple amplitude probability densities to obtain multiple target probability densities; and using the row and column index values ​​corresponding to the multiple target probability densities as the filtering data for the rearranged data.

[0034] Here, the basic principle of the generalized inner product algorithm is to use data uniformity as a discriminant statistic. First, multiple rearranged data sets are selected, and these rearranged data sets are used to estimate the clutter covariance matrix. Then, the data uniformity of each of the multiple rearranged data sets is calculated. Finally, a detection threshold for data uniformity is set, and rearranged data sets exceeding the detection threshold are removed. The remaining rearranged data sets are then used to estimate the covariance matrix until the data uniformity of the remaining rearranged data sets is within the detection threshold. The generalized inner product algorithm can eliminate non-uniform data in the rearranged data sets (which can also be understood as eliminating strong scattering points). For example, the generalized inner product algorithm can be expressed as: Among them, z r (p,u) refers to multiple sets of rearranged data. This refers to the covariance matrix. It refers to the complex real number space, and N refers to the total number of channels. Also, the target uniformity satisfies: GIP(p,u)≤δ, where δ is the preset detection threshold.

[0035] Here, a higher amplitude probability density corresponding to the rearranged data means more overall information and better uniformity. When calculating the amplitude probability density, the original matrix D is first constructed using the rearranged data corresponding to the target uniformity. The amplitude probability density for each row and column of the original matrix is ​​then calculated. The specific calculation formula can be expressed as:

[0036]

[0037] Where, D = [vec(z r_1 ),vec(z r_2 ),…,vec(z r_N )], vec(·) denotes the vectorization operation, his(·) denotes the power histogram distribution function, p(i s ) refers to the magnitude probability density corresponding to each row, i s This refers to the row index, p(j s ) refers to the magnitude probability density corresponding to each column, j s This refers to the column index, and lg(·) refers to the logarithmic function. This refers to one row of data in the original matrix. It refers to a column of data in the original matrix.

[0038] By selecting rearranged data with high uniformity and rich information in step 130, the target recognition accuracy can be significantly improved in the presence of local strong non-uniform clutter and dense target data contamination.

[0039] Step 140: Perform target detection based on multiple sets of screened data to obtain detection results reflecting the presence or absence of targets.

[0040] Here, step 140 includes: according to the multiple sets of filtered data, performing improved pseudo-frame decomposition and soft thresholding on the multiple sets of vector data in sequence to obtain a sparse matrix containing target information; and using a constant false alarm rate (CFAR) detection algorithm to detect the sparse matrix to obtain the detection result.

[0041] Here, because moving targets exhibit sparse characteristics in the image domain and have strong correlations between channels, exhibiting low-rank characteristics, improved pseudo-frame decomposition is performed using the data selected in step 130 to generate low-rank and sparse matrices. This allows for the separation of sparse targets from low-rank background clutter, achieving moving target detection. Specifically, an online robust principal component analysis (RPCA) model is constructed using the original matrix D. By performing improved pseudo-frame decomposition and soft thresholding on the original matrix D, a sparse matrix is ​​obtained.

[0042] For example, the expression for the RPCA model can be represented as:

[0043]

[0044] Here, min refers to finding the minimum value. This refers to a K*N dimensional real space, where K is a positive integer representing the dimension of the filtered data, N is the total number of channels, S is a sparse matrix, L is a low-rank matrix, β is a constant related to noise, rank(·) represents the rank of the matrix, ||·||0 is the zero norm, and ||·|| F Let α be the F-norm, α be the hyperparameter, and st be the constraint condition.

[0045] During the calculation, the improved pseudo-frame decomposition of the original matrix based on the multiple sets of vector data involves multiple iterations. To accelerate the convergence of the low-rank matrix, improved pseudo-frame decomposition is used, breaking it down into column submatrices, row submatrices, and cross matrices. After obtaining the converged low-rank matrix, the difference between the original matrix and the low-rank matrix is ​​calculated. A soft thresholding operator is then applied to this difference for projection, yielding a sparse matrix. By using soft thresholding and improved pseudo-frame decomposition, the recovery effect is guaranteed while avoiding large-scale matrix operations, effectively reducing computational complexity.

[0046] The iteration process will now be illustrated using iterations q and q+1 as examples. Before each iteration begins, the row and column indices of the filtered data are obtained and used as the set of row and column indices for subsequent iterations. Furthermore, in the subscripts, q and q+1 represent the data from the q-th and q+1-th iterations, respectively.

[0047] (1) In the (q+1)th iteration, randomly select a specific number of rows and columns from the set of row and column indices of the filtered data, and update the index set I. q and J q , used for sparse matrix iteration.

[0048] (2) For DL q The selected rows and columns are soft-thresholded and projected onto a sparse matrix set to obtain the sparse matrix S. q+1 .

[0049] (3) Matrix DS q+1 Perform CUR decomposition to obtain row, column, and cross submatrices. Exit the iteration when the error meets the tolerance requirement, thus efficiently obtaining the low-rank matrix L. q+1 .

[0050] Here, after obtaining the sparse matrix, a constant false alarm rate (CFAR) detection algorithm is used to detect the sparse matrix, yielding detection results. These results include both the presence and absence of a target.

[0051] Step 150: When the detection results indicate the presence of a target, determine the optimal weight vector using multiple sets of filtered data.

[0052] Here, one set of filtered data corresponds to one channel, and multiple sets of filtered data correspond to N channels, where N is a positive integer greater than 2; step 150 includes: when the detection result reflects the presence of a target, obtaining the first set of filtered data from the multiple sets of filtered data; for the nth channel, generating the nth two-channel data covariance matrix using the first set of filtered data and the nth set of filtered data, where n takes values ​​from 1 to N; obtaining the target guidance vector corresponding to the nth channel; and calculating the optimal weight vector corresponding to the nth channel based on the target guidance vector corresponding to the nth channel and the nth two-channel data covariance matrix.

[0053] Here, the expression for the optimal weight vector corresponding to the N channels is:

[0054] w opt (v r ) = R -1 a(v r ) / (a H (v r )R -1 a(v r ));

[0055] a(v r )=[1,exp(-j2πv r Δt2 / λ),…,exp(-j2πv r Δt n / λ)] Τ ;

[0056] Among them, w opt (v r ) is a set consisting of the optimal weight vectors corresponding to N channels, R is a set consisting of N two-channel data covariance matrices, a(v r ) is a set consisting of target guidance vectors corresponding to N channels, Δt n = (n-1)d / (vv) a ), d refers to the preset channel spacing, v a It refers to the directional velocity, t refers to the fast time, v r This refers to the projected velocity of the preset target slant range, (·) H This refers to the conjugate transpose operation, (·) T This refers to the transpose operation, where exp(·) is the exponential function and λ is the signal wavelength.

[0057] It should be noted that the optimal weight vector for each channel can be calculated using the above formula. For example, when calculating the optimal weight vector for channel 1, a(v r The value of ) is 1, and R is the covariance matrix of the two channels obtained by calculating the data from channel 1 and channel 2. When calculating the optimal weight vector corresponding to channel 2, a(vr The value of ) is exp(-j2πv r Δt2 / λ), R is the covariance matrix of the two channels calculated from the data of channel 1 and channel 2, Δt2=(2-1)d / (vv) a ).

[0058] Step 160: Determine the radial velocity estimate of the target using the optimal weight vector.

[0059] Here, step 160 includes: calculating the nth velocity estimate based on the optimal weight vector corresponding to the nth channel and the nth two-channel data covariance matrix; selecting the maximum value among the N velocity estimates as the radial velocity estimate of the target.

[0060] Here, multiple echo data are acquired from multiple channels each time, and a set of radial velocity estimates of the target are calculated using the multiple sets of echo data. By acquiring echo data multiple times during the time the target moves from position a to position b, multiple sets of radial velocity estimates of the target are obtained, and thus the target's trajectory is obtained.

[0061] Here, the expression for the radial velocity estimate of the target is:

[0062]

[0063] in, This refers to the estimated radial velocity of the target; argmax(·) refers to the operation of finding the maximum value; v r This refers to the projected velocity of the preset target slant range, w opt (v r ) is a set consisting of the optimal weight vectors corresponding to N channels, z(p,u) refers to the set consisting of multiple sets of echo data, R is a set consisting of N covariance matrices of two-channel data, (·) H This refers to the conjugate transpose operation.

[0064] The process of obtaining equalization data in step 110 is now described in further detail. Specifically, it includes: step 111: calibrating multiple sets of echo data received from multiple channels to obtain multiple sets of corrected data; step 112: performing channel equalization processing on the multiple sets of corrected data to obtain multiple sets of equalized data.

[0065] It should be noted that the clutter data in echo data is complex and variable and cannot be described by specific expressions. Therefore, the data processing described below is essentially aimed at the target data in the echo data.

[0066] Step 111 includes:

[0067] (1) Multiple sets of echo data received from multiple channels are sequentially processed by range compression, range migration correction and azimuth focusing to obtain multiple sets of preprocessed data.

[0068] Here, the distance compression process is equivalent to the channel matched filtering process, the distance migration correction process is equivalent to the channel distance alignment process, and the azimuth focusing process is equivalent to the azimuth matched filtering process.

[0069] Here, the multiple sets of echo data are first subjected to range compression, then range migration correction, and finally azimuth focusing. The preprocessed data of the target corresponding to the nth channel can be represented as follows:

[0070]

[0071] in, This refers to the preprocessed data corresponding to the nth channel, sinc[·] refers to the sigma function, B refers to the carrier bandwidth, λ refers to the signal wavelength, R0 refers to the nearest slant range at the target beam center time, and t m This refers to slow time, x0 refers to the target's starting position on the X-axis, and v a It refers to the directional velocity, t refers to the fast time, v r This refers to the projected velocity of the preset target slant range, B. D This refers to the target Doppler bandwidth, v e d refers to the relative velocity between the target and the radar, v refers to the preset channel spacing, v refers to the radar's flight speed along the X-axis, exp(·) is an exponential function, and j refers to the complex part.

[0072] Furthermore, the target's orientation is directed towards the velocity v. a and radial projection velocity v c Substituting the preprocessed data into the above expression, the preprocessed data of the target corresponding to the nth channel can be expressed as:

[0073]

[0074] in, This refers to the preprocessed data of the target corresponding to the nth channel, sinc[·] refers to the singer function, and B refers to the carrier bandwidth. D This refers to the target Doppler bandwidth, t m This refers to slow time, x0 refers to the target's initial position on the X-axis, d refers to the preset channel spacing, v refers to the radar's flight speed along the X-axis, λ refers to the signal wavelength, R0 refers to the closest slant range at the target beam center, exp(·) is an exponential function, j refers to the complex part, v r This refers to the projected slant distance velocity of the preset target, where c refers to the speed of light.

[0075] (2) Channel registration is performed on the multiple sets of preprocessed data to obtain the multiple sets of correction data. The vectorized representation of the correction data for the multiple targets is s′=[s′1,s′2,…,s′]. N ] T Similarly, the vectorized representation of multiple sets of clutter correction data is: c′=[c′1,c′2,…,c′ N ] T Furthermore, multiple sets of correction data can be vectorized as z′1(p,u), z′2(p,u), ..., z′ N (p,u).

[0076] Step 112 includes:

[0077] (1) Using the first set of data in the multiple sets of correction data, calculate with each set of correction data in the multiple sets of correction data to obtain multiple sets of joint data; use the multiple sets of joint data to solve the Wiener filtering problem to obtain the data reconstruction weights.

[0078] Here, the first channel (i.e., channel 1) is the reference channel. We first assume that the data in each channel includes H0 and H1, where H0 contains only clutter data, and H1 is a mixture of clutter and target data. Taking channel 1 as the reference channel and channel n as the channel to be reconstructed, the joint data can be represented as: z 1,n (p,u)=[z′1(p,u),z′ n Τ (p,u)] Τ , where z 1,n (p,u) represents the nth set of joint data, or in other words, the joint pixel vector of channel 1 and channel n, where n takes values ​​from 1 to N. z′1(p,u) refers to the first set of data in the multiple sets of correction data. n (p,u) refers to the nth set of correction data. When the data from the two channels are well balanced, the clutter information is basically the same, which means that there exists a set of weights that can make the clutter data between channels consistent. Therefore, the equalization process can be understood as a weight solution process, and thus can be summarized as solving the Wiener filtering problem. The formula for the Wiener filtering problem can be expressed as:

[0079]

[0080] in, R 1,n It is a set consisting of N two-channel data covariance matrices, (·) H This refers to the conjugate transpose operation, where E(·) represents the expectation. (·) T This refers to the transpose operation, w n Reconstruct the weights for the data.

[0081] (2) Multiply the reconstructed weights of the data with multiple sets of corrected data to obtain multiple sets of balanced data.

[0082] After several iterations, the weighted covariance matrix can be obtained, and then the weights w can be reconstructed based on the weighted covariance. n This process is applied across all channels to obtain multiple sets of equalization data. The calculation formula for each set of equalization data can be expressed as:

[0083] To address the problems of poor clutter suppression and high false alarm rate in existing radar signal processing methods under non-uniform strong clutter backgrounds, this invention provides an adaptive ground moving target detection method. This method performs channel equalization on echo data received from multiple channels to ensure consistency of clutter data within each channel. Then, it performs channel rearrangement and vectorization on the equalized data to obtain an original matrix, effectively improving data interactivity. Data with high uniformity and density is selected from the original matrix (i.e., filtered data) for target detection. In scenarios with strong local non-uniform clutter and dense target data contamination, this method effectively reduces interference from non-uniform samples, avoiding problems such as difficult convergence, complex computational models, and high computational load. Finally, when the detection result indicates the presence of a target, the optimal weight vector is determined using the filtered data, thereby determining the target's radial velocity estimate. This radial velocity estimate can be used to obtain the moving target's trajectory. Based on the technical solution proposed in this invention, local non-uniform clutter can be effectively suppressed, the false alarm rate can be greatly reduced, the solution efficiency can be improved, and the target detection performance under non-uniform strong clutter backgrounds can be significantly enhanced.

[0084] To clearly illustrate the effects of the technical solution proposed in this invention, the following description is provided in conjunction with the appendix. Figure 2 The attached diagram provides an example.

[0085] Figure 2 This is an example diagram of channel error provided in an embodiment of the present invention. Figure 2 The channel error in the data is obtained after acquiring echo data from the SAR radar. The SAR radar operates in frontal side-looking mode, with three channels evenly distributed along the track direction. The carrier frequency is 8.85 GHz, the bandwidth is 40 MHz, the sampling rate is 60 MHz, the pulse repetition frequency is 1000 Hz, the channel spacing is 0.559 m, and the platform speed is 115 m / s. Figure 2 As shown, the green area represents global error 1, the blue area represents strong non-uniform local error 2, and the red area represents local error 3. The existence of local errors is due to the environment in which the data was collected, which is a scene with strong local non-uniform clutter and dense target data contamination.

[0086] Figure 3 This is the utilization provided by the embodiments of the present invention. Figure 2Interference results generated from echo data collected in the middle. Figure 4 The conventional radar signal processing method provided in this embodiment of the invention is for... Figure 2 The interference result generated after processing the echo data collected in the middle. Figure 5 The method of the present invention provided in the embodiments of the present invention is used to... Figure 2 The interference result generated after processing the echo data collected in the middle. For example... Figure 3 As shown, due to channel errors, the interference results exhibit strong non-uniformity, resulting in poor target detection performance. Figure 4 As shown, the interference results obtained using the two-dimensional frequency domain equalization algorithm in conventional radar signal processing methods show that the global error is well handled, but the local error still exhibits non-uniformity. Figure 5 As shown, both global and local errors are well balanced, and the interference phase uniformity between channels is good.

[0087] Figure 6 The embodiments of the present invention provide the use of various methods to... Figure 2 The detection results were obtained by detecting echo data. It is known that there are actually 7 targets. Figure 6 (a) in the figure represents the results obtained based on the technical solution proposed in this invention, and the results include nine items. Figure 6 (b) in the figure shows the result of detection based on the alternating direction multiplier method algorithm. Figure 6 (c) in the table represents the results detected using the GoDec algorithm. Figure 6 (d) in the text is based on l 1,ε The results of the algorithm detection Figure 6 (e) in the figure represents the result detected using the AP algorithm. Figure 6 (f) in the figure represents the result of detection based on the HQF algorithm. Figure 6 The number of detected results in (b) to (f) all exceed nine. That is, the technical solution proposed in this invention has a lower false alarm rate and superior detection performance compared to existing radar signal processing methods.

[0088] Table 1 shows the different algorithm pairs. Figure 2 Table 1 shows the channel correlation coefficients after echo data processing. Table 2 shows the measured channel correlation coefficients after data processing. As can be seen from Tables 1 and 2, the technical solution provided by this invention can significantly improve the coherence and uniformity between data. Table 3 provides processing methods for different data processing methods. Figure 2 The running time of the echo data is shown in Table 3. As can be seen from Table 3, the running time of the technical solution proposed in this invention is significantly shorter than that of other methods.

[0089] Table 1 Channel correlation coefficients after processing with different algorithms

[0090]

[0091] Table 2 Correlation coefficients of measured data processing channels

[0092]

[0093] Table 3 Running times for different methods

[0094]

[0095]

[0096] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. An adaptive ground moving target detection method, characterized in that, The method includes: Channel equalization processing is performed on multiple sets of echo data received from multiple channels to obtain multiple sets of equalized data. The multiple sets of balanced data are rearranged by channels and vectorized to obtain the original matrix; From the original matrix, rearranged data with data uniformity less than a first preset value and amplitude probability density greater than or equal to a second preset value are selected to obtain multiple sets of filtered data. Target detection is performed based on the multiple sets of screening data to obtain detection results reflecting the presence or absence of a target; When the detection results indicate the presence of a target, the optimal weight vector is determined using the multiple sets of filtered data. The radial velocity estimate of the target is determined using the optimal weight vector.

2. The adaptive ground moving target detection method according to claim 1, characterized in that, The process of selecting rearranged data from the original matrix whose data uniformity is less than a first preset value and whose amplitude probability density is greater than or equal to a second preset value yields multiple sets of filtered data, including: The generalized inner product algorithm is used to calculate the data uniformity of each group of rearranged data, resulting in multiple data uniformities. From the plurality of data uniformities, select the data uniformity that is less than the first preset value to obtain the uniformity corresponding to each group of rearranged data; Calculate the magnitude of each group of rearranged data to obtain the probability density distribution corresponding to the magnitude of each group of rearranged data. From multiple amplitude probability densities, select amplitude probability densities that are greater than or equal to the second preset value to obtain multiple target probability densities; The row and column index values ​​corresponding to the probability densities of the multiple targets are used as the filtering data for the rearranged data.

3. The adaptive ground moving target detection method according to claim 1, characterized in that, The process of performing channel equalization on multiple sets of echo data received from multiple channels to obtain multiple sets of equalized data includes: The multiple sets of echo data received from the multiple channels are calibrated to obtain multiple sets of correction data. Channel equalization processing is performed on the multiple sets of correction data to obtain multiple sets of equalized data.

4. The adaptive ground moving target detection method according to claim 3, characterized in that, The calibration process is performed on the multiple sets of echo data received from the multiple channels to obtain multiple sets of correction data, including: Multiple sets of echo data received from multiple channels are sequentially processed by range compression, range migration correction and azimuth focusing to obtain multiple sets of preprocessed data. Channel registration is performed on the multiple sets of preprocessed data to obtain the multiple sets of corrected data.

5. The adaptive ground moving target detection method according to claim 3, characterized in that, The process of performing channel equalization on the multiple sets of correction data to obtain multiple sets of equalized data includes: Using the first set of data from the multiple sets of correction data, calculations are performed with each set of correction data from the multiple sets of correction data to obtain multiple sets of joint data; The Wiener filtering problem is solved using multiple sets of joint data to obtain the data reconstruction weights; The reconstructed weights of the data are multiplied by the multiple sets of corrected data to obtain the multiple sets of balanced data.

6. The adaptive ground moving target detection method according to claim 1, characterized in that, The step of performing target detection based on the multiple sets of screening data to obtain detection results reflecting the presence or absence of a target includes: Based on the multiple sets of filtered data, the original matrix is ​​sequentially subjected to improved pseudo-frame decomposition and soft thresholding to obtain a sparse matrix containing target information. The sparse matrix is ​​detected using a constant false alarm rate (CFAR) detection algorithm to obtain the detection results.

7. The adaptive ground moving target detection method according to claim 1, characterized in that, One set of filtered data corresponds to one channel, and the multiple sets of filtered data correspond to N channels, where N is a positive integer greater than 2; When the detection result indicates the presence of a target, determining the optimal weight vector using the multiple sets of screening data includes: When the detection result indicates the presence of a target, the first set of filtered data from the multiple sets of filtered data is obtained; For the nth channel, the first set of filtered data and the nth set of filtered data are used to generate the nth two-channel data covariance matrix, where n takes the value from 1 to N; Obtain the target guidance vector corresponding to the nth channel; The optimal weight vector corresponding to the nth channel is calculated based on the target steering vector corresponding to the nth channel and the covariance matrix of the nth two-channel data.

8. The adaptive ground moving target detection method according to claim 7, characterized in that, The step of determining the radial velocity estimate of the target using the optimal weight vector includes: Calculate the nth velocity estimate based on the optimal weight vector corresponding to the nth channel and the covariance matrix of the nth two-channel data. The maximum value among the N velocity estimates is selected as the radial velocity estimate of the target.

9. The adaptive ground moving target detection method according to claim 7, characterized in that, The expression for the optimal weight vector corresponding to N channels is: wopt(vr)=R- 1 a(vr)(a H (vr)R- 1 a(vr)); a(v r )=[1,exp(-j2πv r Δt2λ),…,exp(-j2πv r Δt n l)] Τ ; Among them, w opt (v r ) is a set consisting of the optimal weight vectors corresponding to the N channels, R is a set consisting of N two-channel data covariance matrices, a(v r ) is a set consisting of the target guidance vectors corresponding to the N channels, Δt n = (n-1)d(vv) a ), d refers to the preset channel spacing, v a It refers to the directional velocity, t refers to the fast time, v r This refers to the projected velocity of the preset target slant range, (·) H This refers to the conjugate transpose operation, (·) T This refers to the transpose operation, where exp(·) is the exponential function and λ is the signal wavelength.

10. The adaptive ground moving target detection method according to claim 7, characterized in that, The expression for the radial velocity estimate of the target is: in, This refers to the estimated radial velocity of the target; argmax(·) refers to the operation of finding the maximum value; v r This refers to the projected velocity of the preset target slant range, w opt (v r ) is a set consisting of the optimal weight vectors corresponding to N channels, z(p,u) refers to the set consisting of the multiple sets of echo data, R is a set consisting of N covariance matrices of two-channel data, (·) H This refers to the conjugate transpose operation.

Images