Adaptive sparse constrained inverse synthetic aperture radar imaging method based on hybrid norm
By adopting the 120 norm adaptive sparse constraint method in ISAR high-resolution imaging, the problems of slow computing speed and cumbersome parameter search in the prior art are solved, and efficient and accurate signal reconstruction under short coherence processing time is achieved.
Patent Information
- Application Number
- CN202210711111.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-06-22
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2042-06-22
AI Technical Summary
The existing sparse constraint ISAR high-resolution imaging algorithm has slow computing speed and cumbersome parameters search, and the reconstruction accuracy is insufficient under noisy conditions, especially in short coherence processing time, it is difficult to ensure the accuracy of the signal.
Adaptive sparse constraint method based on l20 norm is adopted, and inverse synthesized aperture radar echo signal is processed by delineating line frequency tone and distance compression, combining noise and signal statistical information, and using the quasi-Newtonian method to establish and solve the optimal relationship to realize the signal High resolution imaging.
The calculation speed and reconstruction accuracy of the algorithm are improved, the regularization coefficient determination process is simplified, the number of iterations is reduced, and the processing efficiency and accuracy are significantly improved.
Smart Images

Figure CN115079171B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of communication technology, and further relates to an adaptive sparse constrained inverse synthetic aperture radar (ISAR) imaging method based on a hybrid norm for a multifunctional radar system in the field of radar signal processing. The present invention can be used for high-resolution imaging processing of targets under short coherent processing time of an inverse synthetic aperture radar. Background Art
[0002] Inverse Synthetic Aperture Radar (ISAR) imaging is a high-resolution imaging technology. Its range resolution is improved by transmitting broadband signals, and the improvement of azimuth resolution is achieved by forming a synthetic array through the relative motion between the radar and the target. In practical applications, due to environmental factors and noise interference, long coherent processing interval (CPI) observations lack stability and cannot ensure the accuracy and reliability of echo data. In comparison, short CPI sampling observations can regard the target motion as steady motion, and the data obtained by sampling is more stable and reliable, and the processing is simpler. Using limited short aperture data to achieve high-resolution ISAR imaging is the key to multi-functional radar systems.
[0003] Chen Qianqian, Zhang Lei, Xu Gang, Xing Mengdao, and Bao Zheng mentioned an ISAR high-resolution imaging method based on the l1 norm in their paper "Interferometric inverse synthetic aperture radar three-dimensional imaging algorithm using multi-channel joint sparse reconstruction" (Journal of Xi'an Jiaotong University). The specific steps of the method are: to obtain the complex matrix of the signal after performing range compression, motion compensation and cross-range unit migration correction on the inverse synthetic aperture radar (ISAR) echo data, then to perform dimension reduction processing on the complex matrix through a Gaussian random matrix, and finally to realize the inverse synthetic aperture radar (ISAR) imaging of the target by solving the optimization equation of the l1 norm. Although the method realizes the inverse synthetic aperture radar (ISAR) imaging processing of the target based on the sparse prior characteristics, the shortcomings of the method in the implementation process are that the operation speed of the algorithm is limited to a certain extent due to the poor convergence characteristics of the l1 norm, and the regularization coefficient λ in the optimization expression of signal reconstruction affects both the reconstruction accuracy and the sparsity of the solution, and the coefficient needs to be debugged according to the noise variance σ and the Laplace coefficient η. The process of finding the optimal coefficient is relatively cumbersome, time-consuming and long, and the accuracy cannot be guaranteed. Summary of the invention
[0004] Aiming at the problems of slow operation speed and cumbersome and inaccurate parameter search in the sparse constrained ISAR high-resolution imaging algorithm in the prior art, the present invention proposes a method based on l20 Norm-adaptive sparse constrained ISAR high-resolution imaging algorithm. This algorithm uses the statistical information of observation samples, noise and targets to deduce how to reconstruct the sparse aperture signal after precise motion compensation into a full aperture signal in the presence of noise, and achieve high-resolution inverse synthetic aperture radar ISAR imaging when the observation method is short CPI sampling, so as to achieve accurate and efficient recovery of target images.
[0005] To achieve the purpose of the present invention, the technical idea of the present invention is to perform range compression processing and motion compensation on the echo signal received by the inverse synthetic aperture radar ISAR by deinterlacing the frequency modulation, and combining l 20 The norm establishes an optimal relationship between the compensated signal construction and the final target to be reconstructed, and the quasi-Newton method is used to solve the cost function of the above optimal relationship to obtain the high-resolution imaging result of the target signal.
[0006] To achieve the above object, the present invention relates to a 20 The hybrid norm adaptive sparse constrained ISAR high-resolution imaging method comprises the following steps:
[0007] (1) Distance compression:
[0008] Performing range compression processing on the sparse echo signals received by the inverse synthetic aperture radar ISAR to obtain range compressed echo signals;
[0009] (2) Envelope alignment:
[0010] Using the adjacent correlation method to perform envelope alignment processing on the echo signal processed by the range compression, so as to obtain the echo signal processed by the envelope alignment;
[0011] (3) Phase compensation:
[0012] The echo signal after envelope alignment is subjected to phase correction processing by using a multi-feature point synthesis method to obtain an echo signal after phase compensation;
[0013] (4) Reconstructing short-aperture echo signal:
[0014] Based on the sparsity of ISAR, the optimal cost function of signal reconstruction is established by combining the Gaussian probability distribution density of noise and the Laplace probability density of signal using maximum a posteriori probability estimation. The quasi-Newton algorithm is used to solve the optimal cost function and obtain the reconstructed full-aperture echo signal.
[0015] (5) Range Doppler imaging:
[0016] The echo signal reconstructed in step 4 is subjected to fast Fourier transform to realize high-resolution range Doppler imaging of inverse synthetic aperture radar.
[0017] Compared with the prior art, the present invention has the following advantages:
[0018] First, the present invention uses the mixed norm to establish a sparse constraint optimization model. 20 The hybrid norm combines the advantages of the l2 norm and the l0 norm, optimizes the convergence of the model, and improves the calculation speed of the model solution.
[0019] Second, in the optimization model of the present invention, under arbitrary noise conditions, the regularization coefficient (constraint coefficient value λ) can be adaptively adjusted according to the sparsity of each iteration, and a relatively stable coefficient can be determined in just a few iteration cycles. This method avoids the complex process of estimating the optimal coefficient based on the statistical characteristics of noise and signal, and improves the efficiency of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a flow chart of the present invention;
[0021] Figure 2 This is the aircraft point target model used in the experiment;
[0022] Figure 3 This is the imaging result diagram of the existing sparse constrained ISAR high-resolution imaging technology;
[0023] Figure 4 This is an imaging result diagram of the simulation data by the technology of the present invention;
[0024] Figure 5 The comparison result of the computing time of 50 iterations of the present invention and the existing sparse constrained ISAR high-resolution imaging technology is shown;
[0025] Figure 6 This is the imaging result of the measured data using the existing sparse constrained ISAR high-resolution imaging technology.
[0026] Figure 7 This is the imaging result diagram of the measured data using the technology of the present invention. DETAILED DESCRIPTION
[0027] The present invention will be further described below in conjunction with the accompanying drawings.
[0028] Refer to the attached Figure 1 The specific implementation steps of the present invention are as follows:
[0029] Step 1, distance compression.
[0030] The first step is to perform linear frequency modulation processing on the echo signal of the target scene received by the inverse synthetic aperture radar ISAR; generally, the radar transmission signal is a linear frequency modulation signal s(t r )
[0031]
[0032] Where, t r Indicates fast time; T p is the pulse width; f c is the carrier frequency; α is the modulation frequency; rect[·] represents the window function.
[0033] The echo signal is pulse compressed by demultiplexing to obtain the formula
[0034]
[0035] In the formula, β p represents the target scattering coefficient, which can be regarded as a constant when the target's rotation angle is small; t m represents slow time, T represents the length of coherent processing time, c represents the speed of light, and 0≤t m ≤T; r0 is the distance between the radar and the target rotation center O; r(t m ) represents the distance from the scattering point p(x,y) on the target to the radar; γ is the wavelength of the radar transmitting signal; C is the transmission speed of electromagnetic waves, and j represents the imaginary unit.
[0036] In the second step, Fourier transform is performed on the signal after the demultiplexing frequency modulation processing to obtain the echo signal after the range compression processing.
[0037] Step 2, envelope alignment.
[0038] The first step is to perform maximum correlation processing between adjacent pulses of echo signals processed by range compression, sum all aligned envelopes, take the average value first, and use the average value as the standard range delay for alignment processing;
[0039] The second step is to perform Fourier transform on the echo signal processed by range compression to obtain the frequency domain signal processed by range compression;
[0040] In the third step, the frequency domain signal processed by the range compression in the second step is multiplied by the linear phase function corresponding to the relative range delay in the first step to obtain the echo signal processed by envelope alignment.
[0041] Step 3, phase compensation.
[0042] Comprehensively process multiple prominent points with strong signal-to-noise ratio in the envelope aligned image to increase the equivalent signal-to-noise ratio, which can improve the estimation accuracy of the initial phase error.
[0043] The first step is to select L prominent point units from the envelope aligned image, remove the phase components that vary with slow time due to different Doppler frequencies, and move the lateral image peak of each range unit to the center of the image. The selection criteria are: select prominent point units from the range units, and each unit has only one obvious target point.
[0044] The second step is to estimate and compensate the starting phase in each prominent point.
[0045] The third step is to use the weighted least square method (WLS) to do the weighted sum of the signal equation after the prominent point compensation, and give a large weight to the small fluctuation component; on the contrary, give a small weight. The above can get a good effect.
[0046] Step 4: reconstruct the short aperture echo signal.
[0047] For the echo signal after phase compensation, an optimization cost function for inverse synthetic aperture radar imaging based on hybrid norm adaptive sparse constraints is established.
[0048]
[0049] Where S is the observation matrix obtained during short CPI observation, and its size is is the total number of sampling pulses, N is the number of sampling points of the range image; A is the observation target to be reconstructed, the size is M×N, M is the number of pulses emitted by the signal under full aperture observation; F is the partial Fourier transform matrix, the size is and A m is the mth row vector of matrix A; β>0; λ n represents the constraint coefficient of the nth iteration adaptive estimation, ||·||2 represents the l2 norm of the vector, Represents the squared value of the vector norm.
[0050] λ n+1 =|A n | k+1 / u
[0051] Where |A| k is the modulus of the kth largest component of A, |A n | k A n The modulus of the kth largest component, A n represents the observation target that needs to be reconstructed in the nth iteration, u controls the convergence of the iterative process and
[0052] The quasi-Newton algorithm is used to solve the above optimization cost function to obtain the reconstructed ISAR echo signal.
[0053] Step 5, range Doppler imaging.
[0054] The reconstructed full-aperture echo signal is subjected to fast Fourier transform to realize high-resolution range Doppler imaging of inverse synthetic aperture radar.
[0055] Verification of invention results
[0056] In this section, we use simulation data and measured data to verify the operation effect of the algorithm. During the verification process, the algorithm proposed in this patent is compared with the imaging algorithm based on the l1 norm. The above two algorithms are used to calculate the data under different noise conditions and the results are compared.
[0057] 1. Simulation data processing experiment:
[0058] In the simulation experiment, the aircraft model used is 70×60×9, consisting of 330 discrete points. The aircraft point target model used in the experiment is as follows: Figure 2 shown.
[0059] The radar system simulation parameters are as follows: carrier frequency f c is 9.6GHz, T is 0.64s, r0 is 10000m, β p is 1.
[0060] By adding Gaussian white noise to the echo signal, different signal-to-noise ratios (4-20dB) are set. The pulse echo signal with an M value of 64 is range compressed, and then envelope alignment and phase compensation are performed. Subsequently, the echo signal after phase compensation is discretized to establish an optimal cost function, and the quasi-Newton algorithm is used to complete the reconstruction of the full-aperture echo signal with M = 256. Finally, fast Fourier transform is performed to achieve super-resolution imaging of ISAR.
[0061] The following correlation coefficient can be used as an indicator to illustrate the accuracy of algorithm reconstruction.
[0062]
[0063] in, is the image obtained by imaging the full-aperture echo data. is the ISAR image reconstructed by the algorithm, |·| represents the modulus, and ||·|| represents the l2 norm. coef The larger the value of is, the closer the full-aperture data reconstructed by the algorithm is to the original data, that is, the higher the reconstruction accuracy of the algorithm is, and vice versa.
[0064] The two algorithms were tested and compared by adding Gaussian white noise. Taking a 5dB signal-to-noise ratio as an example, the imaging results obtained by the two algorithms are as follows: Figure 3 and Figure 4 As shown. The correlation coefficients calculated through 50 reconstruction experiments show that the correlation coefficients of the two methods are not much different. Figure 3 C coef The value is 0.7851, Figure 4 Ccoef The value is 0.7831, which indicates that the present invention also has high reconstruction accuracy.
[0065] However, in Figure 3 In the process, the number of iterations reached 500 and the loop was exited, and the convergence threshold set for the iteration loop was not reached, which took 326.19 seconds. Figure 4 In the example, after 26 iterations, a stable constraint coefficient value λ=0.0045 is obtained, and after 131 iterations, the set loop convergence threshold is reached, which takes 209.5215 seconds. It can be seen that the reconstruction accuracy of the patent algorithm is not much different from that of the l1-norm sparse constraint algorithm, but the computational efficiency is higher than that of the l1-norm.
[0066] Two high-resolution imaging algorithms were used under different noise conditions. The operation time comparison results of 50 iterations are shown in Figure 2. Figure 5 As shown in the figure, the black dotted line is the running time required by the original algorithm, and the red dotted line is the running time required by the algorithm of this patent. In general, the running time of the adaptive sparse constraint algorithm based on the hybrid norm proposed in this patent is shorter than that of the classic original algorithm under different signal-to-noise ratios.
[0067] 2. Measured data processing experiment:
[0068] The radar carrier frequency is 9 GHz, the signal bandwidth is 400 MHz, the pulse repetition frequency is 100 Hz, the target speed is 100 m / s, and the rotation angular velocity is 0.013 rad / s. The parameter setting and data processing are the same as the simulation data processing.
[0069] Figure 6 This is the result of running a high-resolution imaging algorithm based on l1-norm sparse constraints. 2,0 The running results of the improved high-resolution imaging algorithm based on the mixed norm are as follows: Figure 7 As shown in the figure, the operation time of the l1-norm algorithm is 542.6968s, and the operation time of the patent algorithm is 411.1321s. It can be seen that under the measured data, the operation efficiency of the patent algorithm is higher than that of the original algorithm.
Claims
1. An adaptive sparse constrained inverse synthetic aperture radar imaging method based on a hybrid norm, characterized in that: The steps include: (1) Distance compression: Performing range compression processing on the sparse echo signals received by the inverse synthetic aperture radar ISAR to obtain range compressed echo signals; (2) Envelope alignment: Using the adjacent correlation method to perform envelope alignment processing on the echo signal processed by the range compression, so as to obtain the echo signal processed by the envelope alignment; (3) Phase compensation: The echo signal after envelope alignment is subjected to phase correction processing by using a multi-feature point synthesis method to obtain an echo signal after phase compensation; (4) Reconstructing short-aperture echo signal: Combined 20 The hybrid norm uses the sparsity of ISAR to establish a sparse dictionary to construct the relationship between the sampled signal and the signal to be reconstructed. The maximum a posteriori probability estimation is used to establish the optimal cost function for signal reconstruction by combining the Gaussian probability distribution density of the noise and the Laplace probability density of the signal. The quasi-Newton method is used to optimize the cost function and solve it to obtain the reconstructed full-aperture echo signal. (5) Range Doppler imaging: The full-aperture echo signal obtained above is subjected to fast Fourier transform to realize high-resolution range Doppler imaging of inverse synthetic aperture radar.
2. The method of adaptive sparse constrained inverse synthetic aperture radar imaging based on hybrid norm according to claim 1, characterized in that: The specific steps of the distance compression process in step (1) are as follows: The first step is to perform frequency modulation processing on the echo signal of the target scene received by the inverse synthetic aperture radar ISAR; In the second step, Fourier transform is performed on the signal after the demultiplexing frequency modulation processing to obtain the echo signal after the range compression processing.
3. The adaptive sparse constrained inverse synthetic aperture radar imaging method based on hybrid norm according to claim 1, characterized in that: The specific steps of the envelope alignment process in step (2) are as follows: The first step is to perform maximum correlation processing between adjacent pulses of echo signals processed by range compression, sum all aligned envelopes, take the average value first, and use the average value as the standard range delay for alignment processing; The second step is to perform Fourier transform on the echo signal processed by range compression to obtain the frequency domain signal processed by range compression; In the third step, the frequency domain signal processed by the range compression in the second step is multiplied by the linear phase function corresponding to the relative range delay in the first step to obtain the echo signal processed by envelope alignment.
4. The adaptive sparse constrained inverse synthetic aperture radar imaging method based on hybrid norm according to claim 1, characterized in that: The specific steps of the phase compensation process in step (3) are as follows: The first step is to select L prominent point units from the envelope-aligned image and remove the phase components that vary with slow time due to different Doppler frequencies, so as to move the lateral image peak of each range unit to the center of the image. The second step is to estimate and compensate the starting phase in each prominent point; The third step is to use the weighted least square method to make a weighted sum of the signal equation after the prominent point compensation. The smaller the fluctuation component, the larger the weight; otherwise, the smaller the weight.
5. The method of adaptive sparse constrained inverse synthetic aperture radar imaging based on hybrid norm according to claim 1, characterized in that: The specific steps of reconstructing the short aperture echo signal processing in step (4) are as follows: For the echo signal after phase compensation, an optimization cost function for inverse synthetic aperture radar imaging based on hybrid norm adaptive sparse constraints is established. Where S is the observation matrix obtained during short CPI observation, and its size is is the total number of sampling pulses, N is the number of sampling points of the range image; A is the observation target to be reconstructed, the size is M×N, M is the number of pulses emitted by the signal under full aperture observation; F is the partial Fourier transform matrix, the size is and A m is the mth row vector of matrix A; β>0; λ n represents the constraint coefficient of the nth iteration adaptive estimation, ||·||2 represents the l2 norm of the vector, represents the square value of the vector norm, λ n+1 =|A n | k+1 / u,|A n | k A n The modulus of the kth largest component, A n represents the observation target that needs to be reconstructed in the nth iteration, u controls the convergence of the iterative process and The quasi-Newton algorithm is used to solve the above optimization cost function to obtain the reconstructed ISAR echo signal.
