An Improved Incremental Wiener Filtering Super-Resolution Imaging Method Based on GNSS-R SAR

Abstract

The application belongs to the field of satellite navigation signal processing and radar signal processing, and specifically discloses an improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR. The method first uses fast Fourier transform to perform frequency domain remodeling on the observation signal and the system pulse response, and constructs a frequency domain nonlinear adaptive weight to suppress the frequency components in the system spectrum that are prone to cause noise amplification. Then, in each incremental Wiener iteration, the local sidelobe ratio of the current time domain temporary recovery result is extracted, and the regularization parameter of the next iteration is corrected. Finally, according to the amplitude median of the current time domain temporary recovery result, a candidate soft threshold is constructed, and the current time domain temporary recovery result is processed by a soft threshold operator for time domain sparse contraction to obtain a super-resolution imaging result. The application improves the distinguishability and reconstruction stability of 60m close-range targets in low signal-to-noise ratio echoes without changing the bandwidth of GNSS transmitted signals and the hardware structure.

Description

Technical Field

[0001] This invention belongs to the fields of satellite navigation signal processing and radar signal processing, and specifically relates to an improved incremental Wiener filtering super-resolution imaging method based on GNSS-RSAR. Background Technology

[0002] Synthetic Aperture Radar (SAR) systems based on navigation satellite reflected / scattered signals are integrated air-space-ground dual / multi-base SAR systems composed of navigation satellites, low-Earth orbit satellites, and airborne / ground-based receivers. These systems can fully utilize the vast existing navigation satellite network as external radiation sources, possessing significant advantages such as flexible and diverse spatial geometry configurations, strong concealment, short revisit periods, and high spatiotemporal coverage. They represent one of the important directions for the future development of distributed microwave remote sensing and radar networking.

[0003] However, since the original design purpose of the radio signals (GNSS signals) transmitted by the Global Navigation Satellite System was for positioning and velocity measurement rather than high-resolution imaging, the transmitted signal power is low and the ranging code bandwidth is narrow, which severely limits the resolution of radar echoes when performing traditional matching imaging, and makes them extremely susceptible to background noise and multipath interference.

[0004] For example, the ranging code bandwidth of a traditional GPS system is only 1.023 MHz, corresponding to an optimal physical distance resolution of about 150 m. When there are closely adjacent targets less than 150 m (e.g., 60 m) in the observation scene, the traditional correlation imaging method based on fast Fourier transform will suffer from severe main lobe aliasing, making it difficult to stably and clearly distinguish close-range targets.

[0005] Therefore, existing solutions introduce a super-resolution imaging mechanism to overcome the limitations of traditional imaging systems imposed by physical hardware such as the bandwidth of the transmitted signal (i.e., the Rayleigh criterion). This super-resolution imaging mechanism treats the low-resolution observation echo as a degraded model of the high-resolution target scattering distribution after being blurred by the system impulse response (point spread function), downsampling, and strong noise contamination. By constructing an accurate deconvolution mathematical model and combining it with advanced signal processing and optimization theory, it is possible to invert and recover faint targets and image details far exceeding the Nyquist sampling limit from overlapping aliased signals.

[0006] Currently, incremental Wiener filtering (IWF) is widely used for super-resolution imaging. Existing IWF methods typically estimate the signal-to-noise ratio (SNR) based on the peak and median noise levels of the current recovered results, and update the regularization parameters using this global statistic. While this approach can adjust the deconvolution intensity during iteration, its feedback primarily reflects the overall amplitude distribution and lacks awareness of the local structure of the target main lobe, residual side lobes, and inter-target aliasing. In low SNR and narrow bandwidth scenarios of Global Navigation Satellite System Reflectometer (GNSS-R) and SAR, when the distance between two targets is close to or smaller than a traditional range resolution cell, existing IWF imaging results still exhibit problems such as inter-target basal uplift, residual side lobes, and the amplitude of weak targets being affected by the side lobes of strong targets. Summary of the Invention

[0007] The purpose of this invention is to propose an improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR. This method improves the resolvability and reconstruction stability of 60 m close-range targets in low signal-to-noise ratio echoes without changing the GNSS signal bandwidth and hardware structure by introducing nonlinear frequency domain weighting, regularization parameter feedback based on local sidelobe ratio sensing, and temporal sparsity contraction processing based on soft threshold.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: An improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR includes the following steps: Step 1. Acquire GNSS echo data as the observation signal; Step 2. Use Fast Fourier Transform to reshape the frequency domain of the observed signal and the system impulse response to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and construct a frequency domain nonlinear adaptive weight. The frequency domain weights are obtained by setting a lower limit constraint on the frequency domain nonlinear adaptive weights; Step 3. Estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum and the frequency domain weights using weighted Wiener filtering. Step 4. Perform incremental Wiener iteration based on the initial frequency domain estimation solution to obtain the super-resolution imaging results; Step 4.1. In each incremental Wiener iteration, calculate the temporary recovery result in the time domain based on the current frequency domain estimated solution, extract its local sidelobe ratio, and then correct the regularization parameter for the next iteration; Step 4.2. Construct candidate soft thresholds based on the median amplitude of the current time-domain temporary recovery results, introduce a threshold upper limit ratio coefficient to restrict the candidate soft thresholds, obtain the updated soft thresholds, and construct a soft threshold operator based on the soft thresholds; Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution. When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the current time-domain sparse contraction result is used as the super-resolution imaging result; otherwise, the current output frequency-domain estimation solution is used as the input for the next iteration.

[0009] Furthermore, based on the aforementioned improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR, this invention also proposes a corresponding improved incremental Wiener filtering super-resolution imaging system based on GNSS-R SAR, which adopts the following technical solution: An improved incremental Wiener filtering super-resolution imaging system based on GNSS-R SAR includes the following modules: The signal acquisition module is used to acquire GNSS echo data as observation signals. The frequency domain reshaping module is used to reshape the observed signal and the system impulse response in the frequency domain using fast Fourier transform, so as to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and to construct frequency domain nonlinear adaptive weights. The frequency domain weights are obtained by setting a lower limit constraint on the frequency domain nonlinear adaptive weights; The parameter initialization module is used to estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum and the frequency domain weights using weighted Wiener filtering. And a super-resolution imaging module, used to obtain super-resolution imaging results by incremental Wiener iteration based on the initial frequency domain estimation solution; In each incremental Wiener iteration, the time-domain temporary recovery result is calculated based on the current frequency domain estimated solution, and its local sidelobe ratio is extracted. Then, the regularization parameter for the next iteration is corrected. Candidate soft thresholds are constructed based on the median amplitude of the current time-domain temporary recovery results. A threshold upper limit proportional coefficient is introduced to restrict the candidate soft thresholds, resulting in updated soft thresholds. A soft threshold operator is then constructed based on the soft thresholds. Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution. When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the current time-domain sparse contraction result is used as the super-resolution imaging result; otherwise, the current output frequency-domain estimation solution is used as the input for the next iteration.

[0010] Furthermore, based on the aforementioned improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR, this invention also proposes a computer device comprising a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, it implements the steps of the aforementioned improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR.

[0011] Furthermore, based on the aforementioned improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR, this invention also proposes a computer-readable storage medium storing a program that, when executed by a processor, implements the steps of the aforementioned improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR.

[0012] The present invention has the following advantages: As described above, this invention proposes an improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR. The method first constructs a frequency-domain nonlinear adaptive weighting for the one-dimensional observation echo data (observation signal) and the frequency domain representation of the system impulse response obtained along the range direction. This prevents the incremental Wiener update from treating all frequency components equally, instead selectively suppressing frequency bands with weaker system responses and those prone to noise amplification during deconvolution, thus reducing the risk of noise amplification during ill-conditioned inversion. Furthermore, this invention does not rely solely on the global signal-to-noise ratio (SNR) index composed of peak and median noise, but introduces a local sidelobe ratio. This allows the regularization parameter to dynamically change with the degree of sidelobe elevation and the aliasing state between targets in the current image; solving the problems of sidelobe oscillation caused by excessively weak regularization and main lobe broadening and weak target loss caused by excessively strong regularization. Furthermore, for close-range dual-target scenarios at 60 m, this invention constrains inter-target sidelobe residue through local sidelobe ratio feedback and reduces background fluctuations through temporal sparsity contraction using soft thresholding. This allows the two target peaks to maintain clear boundaries and a low inter-target basis in the reconstruction results, improving resolution stability under low signal-to-noise ratio conditions. Because this invention embeds soft-threshold-based temporal sparsity contraction into each round of incremental Wiener updates, rather than performing simple filtering only after iteration, it can continuously reduce amplitude noise, sidelobe artifacts, and spurious scattering points during iterative evolution. Attached Figure Description

[0013] Figure 1 This is a flowchart of the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR in an embodiment of the present invention; Figure 2 This is a schematic diagram of the GNSS-R SAR system configuration in an embodiment of the present invention; Figure 3 This is a range profile of a traditional RD imaging system. Figure 4 This is an RD imaging profile processed by the original incremental Wiener filtering method; Figure 5 This is an RD imaging profile processed by the improved incremental Wiener filtering method of this invention. Detailed Implementation

[0014] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1 To address the issues of low signal-to-noise ratio (SNR), narrow bandwidth, and overlapping main lobes of near-range targets in GNSS reflected signals, this embodiment 1 describes an improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR. This method first applies Kaiser window weighting to the range-compressed GNSS echo data and performs an azimuth Fourier transform to obtain an RD imaging profile with lower sidelobes. Then, for the selected one-dimensional range observation data and system impulse response, a frequency-domain nonlinear adaptive weight is constructed to suppress frequency components in the system spectrum that easily cause noise amplification. During the incremental Wiener iteration process, the local sidelobe ratio of the current time-domain temporary recovery result is extracted, and the regularization parameter for the next iteration is dynamically corrected using the local sidelobe ratio. Simultaneously, based on the sparsity of the target scattering distribution, soft thresholding is used to continuously suppress weak noise components, residual sidelobes, and spurious scattering components, thereby obtaining a one-dimensional range super-resolution imaging result with lower sidelobes and clearer target peaks. This invention can improve the resolvability and reconstruction stability of 60 m near-range targets in low SNR echoes without changing the GNSS transmit signal bandwidth and hardware structure.

[0015] like Figure 1 As shown, the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR includes the following steps: Step 1. Collect GNSS echo data as the observation signal.

[0016] Figure 2 The configuration of the GNSS-R SAR system is shown; in which Indicates transmitter / satellite, Indicates receiver, Indicates the reference antenna. Indicates the detection antenna. Indicates the observation target. Indicates the direction of the line of sight. Indicates the observation angle. Transmitter / Satellite Transmit GNSS signals, receiver via reference antenna Receive direct signals and detect them through the antenna. Receive from target The reflected signals formed by reflection or scattering constitute the observation geometry of bistatic / multistatic GNSS-R SAR.

[0017] Complex envelope form of GNSS signal received at the receiver in a GNSS-R SAR system Represented as: ;in, Indicates GNSS ranging code, Indicates navigation message data code, Indicates a fast time. Indicates the GNSS carrier frequency. This represents the signal phase. To avoid confusion with fast-time variables in subsequent two-dimensional signal models, we will consistently use [this term]. To indicate a fast time, use It indicates a slow time.

[0018] Furthermore, direct signal and reflected signals The two-dimensional forms are represented as follows: ; ; in, and These represent the Doppler frequency shifts of the direct signal and the reflected signal, respectively. and These represent the phases of the direct signal and the reflected signal, respectively. and These represent the direct signal and the reflected signal in slow time, respectively. The corresponding propagation delay.

[0019] The imaging process of GNSS echo data is as follows: First, the data is processed using a software receiver (Rx) to generate a noise-free copy of the direct signal as a local reference signal. Range compression is then achieved by cross-correlation between the reference signal and the echo signal in a fast time dimension. The result of the range compression is derived as follows: ; in, This represents the range compression result obtained by cross-correlating the reference signal and the reflected echo signal. This represents the envelope of the cross-correlation function of the ranging code. This represents the bistatic path difference between the propagation paths of reflected and direct signals. Represents the speed of light. This represents the noise term after distance compression.

[0020] Then, a Fourier transform is performed on the range compression result in the azimuth direction in slow time, yielding: ; in, This represents the range-Doppler imaging (RD imaging) result obtained after azimuth-to-Fourier transform. Indicates slow time The corresponding Doppler frequency variable, Indicates the bistatic Doppler frequency. This indicates that the target energy is focused at the bistatic Doppler frequency. This represents the noise term in the RD domain.

[0021] The imaged GNSS echo data is then represented as a convolutional form of the signal: ; in For the echo data after imaging, The target scattering coefficient, For noise; The observation matrix is ​​a cyclic matrix of the normalized autocorrelation function of the ranging code, expressed as: ; in This represents the number of sampling points for the one-dimensional time delay profile observation signal. Indicates the system impulse response The Each sample value, The problem of overcoming the Rayleigh criterion's limitation on transmitted signal bandwidth and achieving fine resolution of closely overlapping targets has been transformed into solving the problem in the deconvolution equation (i.e. The solution to the target space distribution is obtained by solving the problem in the form of a typical ill-conditioned inverse problem.

[0022] Step 2. Use Fast Fourier Transform to reshape the frequency domain of the observed signal and the system impulse response to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and construct frequency domain nonlinear adaptive weights; by setting a lower limit constraint on the frequency domain nonlinear adaptive weights, the frequency domain weights are obtained, which are used to suppress the noise amplification effect in noise-sensitive frequency bands.

[0023] First, the observed signal conduct Point Fast Fourier Transform yields the frequency domain representation of the observed signal: ; in This represents the observed signal spectrum, i.e., the actual observed spectrum; This represents the Fast Fourier Transform.

[0024] At the same time, the system impulse response conduct Point Fast Fourier Transform yields the frequency domain representation of the system's impulse response: ; in This represents the system's impulse response spectrum.

[0025] Subsequently, to measure the strength of the system's response at different frequency positions, the normalized amplitude spectrum of the system's spectrum is calculated, and the calculation formula is as follows: ; in This represents the normalized system spectral amplitude. This represents the amplitude of the system's impulse response spectrum. This represents the maximum value in the system's impulse response spectrum amplitude. Indicates taking the absolute value; To prevent extremely small positive numbers with a denominator of 0.

[0026] Because GNSS ranging codes have limited bandwidth, frequency bands with weaker system response are prone to noise amplification during deconvolution. Therefore, based on the normalized system spectral amplitude... Constructing frequency domain nonlinear adaptive weights for: ; in This represents the frequency domain weighting intensity adjustment parameter; This represents a nonlinear exponent. Adjustment is made... and The weight attenuation rate and noise suppression intensity of different frequency bands can be controlled. In this embodiment, , .

[0027] To avoid the complete loss of effective frequency information due to excessively small weights, further adaptive weighting in the frequency domain nonlinearity is applied. Set a lower bound constraint to obtain the frequency domain weights. Its expression is as follows: ; in Indicates the lower bound of the frequency domain weights. This indicates taking the minimum value. This indicates taking the maximum value.

[0028] Through the above processing, frequency bands with stronger system responses and more concentrated effective target information are assigned larger weights, while frequency bands with weaker system responses and where noise is easily amplified are assigned smaller weights. The resulting frequency domain weights can provide a more stable update direction for subsequent incremental Wiener iterations.

[0029] This invention, based on the existing incremental Wiener filter deconvolution framework, constructs nonlinear frequency domain adaptive weights based on the normalized system spectral amplitude and embeds these weights into the residual update term of the incremental Wiener filter. This enables the algorithm to perform differentiated updates based on the response strength of different frequency components in the GNSS-R SAR system. For frequency components with weak system impulse responses that are prone to noise amplification during deconvolution, the frequency domain weights can selectively suppress them, thereby reducing the risk of noise amplification in weak response frequency bands and improving the stability of range-directed deconvolution recovery.

[0030] Step 3. Estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum, and the frequency domain weights using weighted Wiener filtering.

[0031] Regarding the observed signal spectrum obtained in step 2 System response spectrum and frequency domain weights We employ peak-noise statistical estimation and regularization initialization techniques to initialize the iterative parameters, regularization parameters, and initial frequency domain estimation solutions in incremental Wiener filtering.

[0032] First, define the maximum number of iterations as... Define regularization parameters The dynamic adjustment range is ; in This indicates the lower bound of the regularization parameter. This indicates the upper bound of the regularization parameter; the purpose of setting this dynamically adjustable range is to prevent... Too small a value will cause the noise to be amplified excessively, and also prevent... An excessively large value can lead to overly smoothed recovery results.

[0033] In this embodiment, For example, 10 -8 , For example, the value is 10. 2 .

[0034] Then, based on the observed signal The initial peak signal-to-noise ratio is estimated by the peak amplitude and median amplitude, and the calculation formula is as follows: ; in This represents the initial peak signal-to-noise ratio. The peak value represents the maximum amplitude of the observed signal. This represents the median amplitude of the observed signal, used to approximate the background noise level.

[0035] according to The initialization parameters for regularization are: ; in This represents the initial regularization parameter.

[0036] Then, Limited to the range of dynamic adjustment within, that is Updated to: .

[0037] Finally, the spectrum of the observed signal will be... System impulse response spectrum and frequency domain weights Substituting both into the weighted Wiener filter expression, we obtain the initial frequency domain estimation solution as follows: ; in This represents the initial frequency domain estimate solution. The initial frequency domain estimate before the start of the iteration is the starting point of the iterative algorithm, providing initial values ​​for subsequent incremental Wiener iterations. express conjugate, It represents the square of the amplitude of the system's impulse response spectrum.

[0038] This step yields an initial frequency domain estimation solution with preliminary deconvolution and noise suppression capabilities. This provides a stable starting point for subsequent residual-driven incremental updates.

[0039] Step 4. Perform incremental Wiener iteration based on the initial frequency domain estimation solution to obtain the super-resolution imaging results.

[0040] Step 4.1. In each incremental Wiener iteration, calculate the temporary recovery result in the time domain based on the current frequency domain estimated solution, extract its local sidelobe ratio, and then correct the regularization parameter for the next iteration.

[0041] Regarding the first The next incremental Wiener iteration estimates the solution in the current frequency domain. and current regularization parameters Using residual-driven frequency-domain weighted incremental Wiener update, local sidelobe ratio dynamic feedback, time-domain adaptive sparsity contraction, and frequency-domain write-back as input, the target scattering distribution is successively corrected. By gradually reducing the residual between the actual observed spectrum and the predicted observed spectrum, the main lobe broadening, sidelobe residue, and noise amplification caused by the system impulse response are suppressed, resulting in intermediate frequency-domain estimation results with progressively improved resolution. Indicates the current iteration number. , This indicates the maximum number of iterations.

[0042] In the In the next incremental Wiener iteration, the solution is first estimated based on the current frequency domain. Calculate the corresponding predicted observation spectrum: ; in, Indicates by the first The predicted observation spectrum is obtained by predicting the frequency domain estimation solution in the next iteration.

[0043] Subsequently, the actual observed spectrum With predicted observation spectrum By taking the difference, we get the first... Frequency domain residuals before the next iteration update: ; in, Indicates the first The frequency domain residual before the frequency domain weighted incremental Wiener update in the next iteration is used to characterize the observational components that are not yet explained by the current estimation results.

[0044] Then, As an incremental correction amount, combined with frequency domain weights System impulse response spectrum and current regularization parameters That is, the first The regularization parameter of the next iteration Perform a frequency-domain weighted incremental Wiener update to obtain the first... Intermediate frequency domain estimation results of the next iteration: ; in, Indicates the first The intermediate frequency domain estimation result obtained by the frequency domain weighted incremental Wiener update in the next iteration.

[0045] The updated predicted observation spectrum is calculated based on the current intermediate frequency domain estimation results. for: ; Then the first The intermediate frequency domain residual of the next iteration is: ; in, Indicates the first The intermediate frequency domain residuals when the frequency domain weighted incremental Wiener update is completed in the next iteration and before the time domain sparsity contraction is performed.

[0046] To determine whether the current frequency-domain weighted incremental Wiener update still has an effective improvement effect, the calculation of the... The residual improvement amount corresponding to the next iteration : ; in, Used to indicate the first The relative decrease in frequency domain residual energy before and after the second-frequency domain weighted incremental Wiener update. This represents the L2 norm.

[0047] Define a preset convergence threshold ; when When the time is reached, it indicates that the residual improvement brought by this update has been small, and the current iteration is marked as meeting the convergence condition, but the local sidelobe ratio extraction, regularization parameter correction, time-domain sparsity shrinkage and frequency-domain write-back of this iteration are still completed.

[0048] To extract the temporal structure features from the current intermediate frequency domain estimation results, the current intermediate frequency domain estimation results are... Perform an inverse Fourier transform to obtain the first... Temporal recovery results of the next iteration for: ; in, Indicates the inverse Fourier transform. Indicates the number of transformation points.

[0049] Temporary recovery results for the current time domain Analyze the amplitude and extract the amplitude of the current main peak. : ; in Indicates the first The maximum amplitude of the current temporary recovery result in the time domain in the next iteration is the amplitude of the current main peak. To take the absolute value.

[0050] To avoid misidentifying the main lobe region near the main peak as a side lobe, a main peak protection region is established. A preset number of sampling points are extended to the left and right of the sampling point at the current distance from the main peak, forming the main peak protection region. In this embodiment, the preset number is 3 sampling points.

[0051] The maximum amplitude is obtained by searching for the maximum amplitude value outside the main peak protection region. .

[0052] Based on the current main peak amplitude and maximum sidelobe amplitude Calculate the first The local sidelobe ratio corresponding to the next iteration is obtained as follows:

[0053] ; in, Indicates the first The local sidelobe ratio corresponding to the next iteration.

[0054] Meanwhile, based on the current temporary recovery results in the time domain Re-estimation of peak signal-to-noise ratio: ; in, Indicates the first The peak signal-to-noise ratio corresponding to each iteration is used to characterize the strength of the main peak amplitude relative to the background noise level in the current temporal temporary recovery result; This indicates taking the median value.

[0055] Based on peak signal-to-noise ratio Calculate candidate regularization parameters: ; in, This represents the candidate regularization parameter obtained solely from the current peak signal-to-noise ratio.

[0056] To enable the regularization parameters to further perceive the local sidelobe state in the current temporal temporary recovery result, the local sidelobe ratio is used. The candidate regularization parameters are dynamically adjusted, and the adjusted regularization parameters are restricted to the dynamically adjusted range. Inside, the first The regularization parameter for the next iteration : ; in, This represents the sidelobe feedback adjustment coefficient, used to control the influence of the local sidelobe ratio on the regularization parameter. In this embodiment, .

[0057] when A higher value indicates a significant increase in the sidelobes in the current temporal temporary recovery result, and the regularization parameter... This increases accordingly to suppress sidelobe oscillations and noise amplification; when A lower value indicates a more prominent main lobe and weaker side lobe remnants; the regularization parameter... The size is reduced accordingly to retain more target details.

[0058] In each round of incremental Wiener iteration, the present invention transforms the intermediate estimation result in the frequency domain to the time domain, extracts the main peak amplitude of the current temporary recovery result in the time domain, and searches for the maximum sidelobe amplitude outside the main peak protection region to calculate the local sidelobe ratio. Then, the local sidelobe ratio is used to dynamically correct the candidate regularization parameter obtained from the peak signal-to-noise ratio, thereby constructing a structure-aware feedback closed loop.

[0059] Compared with methods that rely solely on peak signal-to-noise ratio or peak noise statistics to adjust regularization parameters, this invention enables regularization parameters to simultaneously sense background noise levels, sidelobe remnants, inter-target floor uplift status, and near-range target aliasing, thereby improving the stability of range-oriented super-resolution recovery in low signal-to-noise ratio GNSS-R SAR scenarios.

[0060] Step 4.2. Construct candidate soft thresholds based on the median amplitude of the current time-domain temporary recovery results, introduce a threshold upper limit ratio coefficient to restrict the candidate soft thresholds, obtain the updated soft thresholds, and construct a soft threshold operator based on the soft thresholds; Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution.

[0061] Current temporal domain temporary recovery results Construct candidate soft thresholds using the median amplitude: ; in, Indicates the first The candidate soft threshold corresponding to the next iteration. This represents the soft threshold shrinkage coefficient. In this embodiment, .

[0062] To avoid the false suppression of weak target components due to excessively large candidate soft thresholds, a threshold upper limit scaling factor is introduced. For candidate soft thresholds By applying restrictions, the updated soft threshold is obtained. : ; in, This represents the upper limit of the threshold proportional coefficient, satisfying... .

[0063] Define the soft threshold operator as follows: ; in, Represents a symbolic function. This represents any time-domain sample to be subjected to soft thresholding.

[0064] The soft threshold operator Effect on the current temporal domain temporary recovery result , obtained the The time-domain sparsity contraction results of the next iteration : ; After completing the temporal sparsity contraction, Transform to the frequency domain to obtain the first... The output frequency domain estimation solution of the next iteration: ; in, Indicates the first The output frequency domain estimation solution is obtained after the time-domain sparsity contraction and frequency domain write-back in the next iteration; When the preset convergence condition is met or the maximum number of iterations is reached, the current time-domain sparse shrinkage result will be applied. As the result of super-resolution imaging; otherwise, the current output frequency domain estimation solution. and output regularization parameters The incremental Wiener iteration continues as input for the next iteration.

[0065] This step processes the data to obtain the final GNSS-R SAR range super-resolution imaging results. Figure 3 This shows a range profile of a traditional RD imaging system. Figure 4 The image shows the RD imaging profile processed by the original incremental Wiener filtering method; Figure 5 This image shows an RD imaging profile processed by the improved incremental Wiener filtering method of this invention. From Figure 3-5 As can be seen, compared with the traditional RD imaging method and the original incremental Wiener filtering method, the improved incremental Wiener filtering method of this invention can further reduce noise amplification and sidelobe residue in low signal-to-noise ratio GNSS-R SAR scenarios, and improve the resolvability and recovery stability of dual targets at close range of 60m.

[0066] Example 2 This embodiment 2 describes an improved incremental Wiener filter super-resolution imaging system based on GNSS-R SAR, which is based on the same inventive concept as the improved incremental Wiener filter super-resolution imaging method based on GNSS-R SAR in embodiment 1 above.

[0067] An improved incremental Wiener filtering super-resolution imaging system based on GNSS-R SAR includes the following modules: The signal acquisition module is used to acquire GNSS echo data as observation signals. The frequency domain reshaping module is used to reshape the observed signal and the system impulse response in the frequency domain using fast Fourier transform, so as to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and to construct frequency domain nonlinear adaptive weights. The frequency domain weights are obtained by setting a lower limit constraint on the frequency domain nonlinear adaptive weights; The parameter initialization module is used to estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum and the frequency domain weights using weighted Wiener filtering. And a super-resolution imaging module, used to obtain super-resolution imaging results by incremental Wiener iteration based on the initial frequency domain estimation solution; In each incremental Wiener iteration, the time-domain temporary recovery result is calculated based on the current frequency domain estimated solution, and its local sidelobe ratio is extracted. Then, the regularization parameter for the next iteration is corrected. Candidate soft thresholds are constructed based on the median amplitude of the current time-domain temporary recovery results. A threshold upper limit proportional coefficient is introduced to restrict the candidate soft thresholds, resulting in updated soft thresholds. A soft threshold operator is then constructed based on the soft thresholds. Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution. When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the current time-domain sparse contraction result is used as the super-resolution imaging result; otherwise, the current output frequency-domain estimation solution is used as the input for the next iteration.

[0068] It should be noted that any content not mentioned in the above-described functional modules of the system described in Embodiment 2 can be referred to the step description of the corresponding method in Embodiment 1 above, and will not be repeated in detail here.

[0069] Example 3 This embodiment 3 describes a computer device including a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, it implements the steps of the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR described in embodiment 1 above.

[0070] Example 4 This embodiment 4 describes a computer-readable storage medium storing a program that, when executed by a processor, is used to implement the steps of the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR in embodiment 1 above.

[0071] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc.

[0072] Of course, the above description is only a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. It should be noted that any equivalent substitutions or obvious modifications made by those skilled in the art under the guidance of this specification fall within the scope of this specification and should be protected by the present invention.

Claims

1. An improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR, characterized in that, Includes the following steps: Step 1. Acquire GNSS echo data as the observation signal; Step 2. Use Fast Fourier Transform to reshape the frequency domain of the observed signal and the system impulse response to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and construct a frequency domain nonlinear adaptive weight. The frequency domain weights are obtained by setting a lower limit constraint on the frequency domain nonlinear adaptive weights; Step 3. Estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum and the frequency domain weights using weighted Wiener filtering. Step 4. Perform incremental Wiener iteration based on the initial frequency domain estimation solution to obtain the super-resolution imaging results; Step 4.

1. In each incremental Wiener iteration, calculate the temporary recovery result in the time domain based on the current frequency domain estimated solution, extract its local sidelobe ratio, and then correct the regularization parameter for the next iteration; Step 4.

2. Construct candidate soft thresholds based on the median amplitude of the current time-domain temporary recovery results, introduce a threshold upper limit ratio coefficient to restrict the candidate soft thresholds, obtain the updated soft thresholds, and construct a soft threshold operator based on the soft thresholds; Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution. When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the current temporal sparse contraction result is used as the super-resolution imaging result. Otherwise, the current output frequency domain estimate is used as the input for the next iteration.

2. The improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR according to claim 1, characterized in that, Observing signals in step 1 Represented as: ; in For the imaged GNSS echo data, The target scattering coefficient, For noise, This is the observation matrix.

3. The improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR according to claim 2, characterized in that, Step 2 specifically involves: First, the observed signal conduct Point Fast Fourier Transform yields the frequency domain representation of the observed signal: ； in This represents the observed signal spectrum, i.e., the actual observed spectrum; Indicates Fast Fourier Transform; At the same time, the system impulse response conduct Point Fast Fourier Transform yields the frequency domain representation of the system's impulse response: ； in Represents the system's impulse response spectrum; Subsequently, the normalized amplitude spectrum of the system spectrum is calculated using the following formula: ； in This represents the normalized system spectral amplitude. This represents the amplitude of the system's impulse response spectrum. This represents the maximum value in the system's impulse response spectrum amplitude. To prevent extremely small positive numbers with a denominator of 0; Based on the normalized system spectrum amplitude Construct frequency domain nonlinear adaptive weights for: ； in This represents the frequency domain weighting intensity adjustment parameter. Indicates a nonlinear exponent; Adaptive weighting for frequency domain nonlinearity Set a lower bound constraint to obtain the frequency domain weights. Its formula is expressed as follows: ； in Indicates the lower bound of the frequency domain weights. This indicates taking the minimum value. This indicates taking the maximum value.

4. The improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR according to claim 3, characterized in that, Step 3 specifically involves: First, define the maximum number of iterations as... Define regularization parameters The dynamic adjustment range is ;in This indicates the lower bound of the regularization parameter. Indicates the upper bound of the regularization parameter; Then, based on the observed signal The initial peak signal-to-noise ratio is estimated by the peak amplitude and median amplitude, and the calculation formula is as follows: ； in This represents the initial peak signal-to-noise ratio. The peak value represents the maximum amplitude of the observed signal. This represents the median amplitude of the observed signal; according to The initialization parameters for regularization are: ； in Indicates the initial regularization parameters; Then, Limited to the range of dynamic adjustment within, that is Updated to: ； Finally, the spectrum of the observed signal will be... System impulse response spectrum and frequency domain weights Substituting both into the weighted Wiener filter expression, we obtain the initial frequency domain estimation solution as follows: ； in This represents the initial frequency domain estimation solution. express conjugate, It is the square of the amplitude of the system's impulse response spectrum.

5. The improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR according to claim 4, characterized in that, Step 4.1 specifically involves: In the In the next incremental Wiener iteration, the solution is first estimated based on the current frequency domain. Calculate the corresponding predicted observation spectrum: ； in, Indicates by the first The predicted observation spectrum is obtained by predicting the frequency domain estimation solution in the next iteration; Subsequently, the actual observed spectrum With predicted observation spectrum By taking the difference, we get the first... Frequency domain residuals before the next iteration update: ； in, Indicates the first The frequency domain residual before the frequency domain weighted incremental Wiener update is performed in the next iteration; Then, As an incremental correction amount, combined with frequency domain weights System impulse response spectrum and current regularization parameters That is, the first The regularization parameter of the next iteration Perform a frequency-domain weighted incremental Wiener update to obtain the first... Intermediate frequency domain estimation results of the next iteration: ； in, Indicates the first The intermediate frequency domain estimation result obtained by frequency domain weighted incremental Wiener update in the next iteration; The updated predicted observation spectrum is calculated based on the current intermediate frequency domain estimation results. for: ； Then the first The intermediate frequency domain residual of the next iteration is: ； in, Indicates the first The intermediate frequency domain residual when the frequency domain weighted incremental Wiener update is completed in the next iteration and before the time domain sparsity contraction is performed; Calculate the first The residual improvement amount corresponding to the next iteration : ； in, Used to indicate the first The relative decrease in frequency domain residual energy before and after the second-frequency domain weighted incremental Wiener update. Represents the L2 norm; Define a preset convergence threshold ;when When the current iteration satisfies the convergence condition, mark it as such. For the current intermediate frequency domain estimation results Perform an inverse Fourier transform to obtain the first... Temporal recovery results of the next iteration for: ； in, Indicates the inverse Fourier transform; Temporary recovery results for the current time domain Analyze the amplitude and extract the amplitude of the current main peak. : ； in To take the absolute value; Using the current location of the main peak as the center, expand the sampling points to the left and right by a preset number of sampling points to form the main peak protection area; The maximum amplitude is obtained by searching for the maximum amplitude value outside the main peak protection region. ; Based on the current main peak amplitude and the maximum sidelobe amplitude, calculate the first... The local sidelobe ratio corresponding to the next iteration is obtained as follows: ； in, Indicates the first The local sidelobe ratio corresponding to the next iteration; Meanwhile, based on the current temporary recovery results in the time domain Re-estimation of peak signal-to-noise ratio: ； in, Indicates the first Peak signal-to-noise ratio corresponding to the next iteration Indicates taking the median value; Based on peak signal-to-noise ratio Calculate candidate regularization parameters: ； in, This represents the candidate regularization parameter obtained solely from the current peak signal-to-noise ratio; use The candidate regularization parameters are dynamically adjusted, and the adjusted regularization parameters are restricted to the dynamically adjusted range. Inside, the first The regularization parameter for the next iteration : ； in, This represents the sidelobe feedback adjustment coefficient.

6. The improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR according to claim 5, characterized in that, Step 4.2 specifically involves: Based on the current temporary recovery results in the time domain Construct candidate soft thresholds using the median amplitude: ； in, Indicates the first The candidate soft threshold corresponding to the next iteration. Indicates the soft threshold shrinkage coefficient; Introducing a threshold upper limit ratio coefficient For candidate soft thresholds By applying restrictions, we obtain the updated soft threshold: ； Define the soft threshold operator as follows: ； in, Represents a symbolic function. This represents any time-domain sample to be subjected to soft thresholding; the soft thresholding operator... Effect on the current temporal domain temporary recovery result , obtained the The time-domain sparsity contraction results of the next iteration : ； After completing the temporal sparsity contraction, Transform to the frequency domain to obtain the first... The output frequency domain estimation solution of the next iteration: ； in, Indicates the first The output frequency domain estimation solution is obtained after the time-domain sparsity contraction and frequency domain write-back in the next iteration; When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the 1st iteration will be... The temporal sparsity contraction result obtained in the second iteration As a result of super-resolution imaging; Otherwise, the current output frequency domain estimated solution and output regularization parameters The incremental Wiener iteration continues as input for the next iteration.

7. An improved incremental Wiener filtering super-resolution imaging system based on GNSS-R SAR, characterized in that, Includes the following modules: The signal acquisition module is used to acquire GNSS echo data as observation signals. The frequency domain reshaping module is used to reshape the observed signal and the system impulse response in the frequency domain using fast Fourier transform, so as to obtain the spectrum of the observed signal and the spectrum of the system impulse response, and to construct frequency domain nonlinear adaptive weights. The frequency domain weights are obtained by setting a lower limit constraint on the frequency domain nonlinear adaptive weights; The parameter initialization module is used to estimate the initial peak signal-to-noise ratio based on the peak and median amplitudes of the observed signal. and according to The regularization parameters are initialized, and then the initial frequency domain estimation solution is calculated by combining the observed signal spectrum, the system impulse response spectrum and the frequency domain weights using weighted Wiener filtering. And a super-resolution imaging module, used to obtain super-resolution imaging results by incremental Wiener iteration based on the initial frequency domain estimation solution; In each incremental Wiener iteration, the time-domain temporary recovery result is calculated based on the current frequency domain estimated solution, and its local sidelobe ratio is extracted. Then, the regularization parameter for the next iteration is corrected. Candidate soft thresholds are constructed based on the median amplitude of the current time-domain temporary recovery results. A threshold upper limit proportional coefficient is introduced to restrict the candidate soft thresholds, resulting in updated soft thresholds. A soft threshold operator is then constructed based on the soft thresholds. Using the constructed soft threshold operator, the current time-domain temporary recovery result is subjected to time-domain sparsity contraction processing to obtain the time-domain sparsity contraction result. Then, the time-domain sparsity contraction result is transformed into the frequency domain to obtain the output frequency domain estimation solution. When the current iteration satisfies the convergence condition or reaches the maximum number of iterations, the current temporal sparse contraction result is used as the super-resolution imaging result. Otherwise, the current output frequency domain estimate is used as the input for the next iteration.

8. A computer device, comprising a memory and one or more processors; characterized in that, The memory stores executable code, which, when executed by the processor, implements the steps of the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a program stored thereon; characterized in that, When executed by a processor, the program is used to implement the steps of the improved incremental Wiener filtering super-resolution imaging method based on GNSS-R SAR as described in any one of claims 1 to 6.

Images