Joint principal component analysis and subspace filtering based spaceborne sar scatter wave mutual interference detection and suppression method, system, storage medium and electronic device
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN UNIVERSITY
- Filing Date
- 2026-03-06
- Publication Date
- 2026-06-09
Smart Images
Figure CN122172187A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of signal processing technology, and in particular to a method, system, storage medium, and electronic device for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering. Background Technology
[0002] Currently, Synthetic Aperture Radar (SAR) is an active microwave imaging system capable of achieving high-resolution ground observations in all weather and at all times. SAR obtains high range resolution by transmitting broadband signals and combines this with platform motion to achieve high azimuth resolution. It is widely used in topographic mapping, disaster monitoring, agricultural surveillance, and military reconnaissance.
[0003] However, in complex electromagnetic environments, SAR systems are inevitably subject to various forms of radio frequency interference (RFI). Multiple scattering wave interference (MTSI) is a typical form of RFI, usually appearing in the echo signals of SAR images, causing a significant deterioration in image quality. The sources of MTSI mainly originate from other radar or communication equipment in the surrounding environment. The signals emitted by these devices may interfere with the SAR signal, producing abrupt changes or anomalies in the frequency or time domains during imaging.
[0004] Traditional interference suppression methods typically rely on techniques such as Robust Principal Component Analysis (RPCA) or Block Subspace Filtering (BSF) to separate interference from the signal. Interference suppression based on these methods can usually effectively extract the valid signal from image data and suppress or remove interference. However, existing RPCA and other methods often depend on preset thresholds and do not fully consider the dynamic characteristics of interference, which can easily lead to interference identification errors. Especially in complex scenarios with multiple interference sources and large variations in interference intensity, they often fail to provide stable suppression results. Summary of the Invention
[0005] The purpose of this application is to provide a method, system, storage medium, and electronic device for detecting and suppressing mutual interference of spaceborne SAR scattering waves by combining principal component analysis and subspace filtering, so as to solve or alleviate the problems existing in the prior art.
[0006] To achieve the above objectives, this application provides the following technical solution:
[0007] This application provides a method for detecting and suppressing mutual interference of spaceborne SAR scattered waves using a combination of principal component analysis and subspace filtering, comprising: step S101, preprocessing the SAR image data containing interference to extract time-domain interquartile range and frequency-domain fluctuation features; step S102, using robust principal component analysis (RPCA) to suppress the SAR image data and outputting the interference-free pseudo-clean SAR image data; step S103, postprocessing the interference-free pseudo-clean SAR image data to extract time-domain interquartile range and frequency-domain fluctuation features; step S104, comparing the interference-free image data with the original image. The characteristics and amplitude differences of the data are used to determine the specific location of the interference; Step S105: By comparing the amplitude differences between the pseudo-clean image and the original image in the range frequency domain, regions with similar intensity differences are divided into blocks; Step S106: Based on the pseudo-clean image data after RPCA interference removal, the threshold of the Robust Block Subspace Filtering (RBSF) method is applied; Step S107: According to the auxiliary threshold of each block, RBSF suppression is performed on the range frequency domain of the original image data to achieve more accurate interference suppression; Step S108: The clean SAR image data after interference removal is output, and the final interference suppression result is displayed.
[0008] Preferably, in step S101, the data containing interference echoes is processed pulse by pulse to extract the time-domain interquartile range and frequency-domain variability features, specifically as follows:
[0009] Interquartile Range (IQR) is a statistical metric used to describe the distribution range of signal amplitude, thereby identifying outliers and impulse interference. Its calculation formula is as follows:
[0010] First, regarding the first Amplitude sequence of pulses Calculate its interquartile range (IQR) characteristics:
[0011]
[0012] in, and They represent the first The third quartile and the first quartile of each pulse amplitude sequence are then used. The maximum interquartile range of each pulse is then taken as the characteristic value of that pulse, which can be expressed as:
[0013]
[0014] in, ,and Indicates the first The first pulse A distance-oriented sample.
[0015] By calculating the above interquartile range features pulse by pulse, the time-domain interquartile range feature sequence can be obtained:
[0016]
[0017] Frequency domain variability is a characteristic used to measure the degree of variation of a signal in the frequency domain. It reflects the degree of fluctuation in the amplitude distribution of a signal in the frequency domain, and is particularly suitable for identifying interference signals with prominent frequency components. For example, high variability in the frequency domain usually indicates that the frequency components of the signal are widely distributed, and there may be significant spectral interference. By calculating the frequency domain variance, this variation can be effectively identified, and further, a basis can be provided for interference detection and suppression.
[0018] Frequency domain volatility is quantified by calculating the variance of signal frequency domain samples, which quantifies the degree of variation in the frequency domain. The calculation formula is as follows:
[0019] For the Amplitude sequence of pulses Frequency components in the frequency domain, frequency domain wave properties It can be represented as:
[0020]
[0021] in, Indicates the first The first pulse One frequency domain sample, It is the first The frequency domain sample mean of each pulse It is the number of frequency domain samples.
[0022] By calculating the frequency domain variance of each pulse, the frequency domain wave characteristic sequence is obtained:
[0023]
[0024] in, Indicates the first Frequency domain variability of pulses.
[0025] Preferably, in step S102, robust principal component analysis (RPCA) is used to suppress the SAR image data, and the output of the interference-free pseudo-clean SAR image data is as follows:
[0026] Robust Principal Component Analysis (RPCA) is a low-rank matrix factorization-based method that aims to extract low-rank principal components and sparse outliers from a data matrix. Unlike traditional Principal Component Analysis (PCA), RPCA maintains high robustness even when sparse noise or outliers are present in the data. RPCA is widely used in image processing, signal recovery, and anomaly detection, and it demonstrates its superiority, especially when processing remote sensing data containing strong noise or interference.
[0027] The basic assumption of RPCA is: given an observation matrix This matrix can be approximated as a low-rank matrix. With sparse noise matrix The sum, that is:
[0028]
[0029] in: It is the observation matrix (e.g., SAR image data with interference). It is a low-rank matrix that contains the main structure and features of the data (e.g., the main components of the signal). It is a sparse matrix that contains outliers or noise in the data (e.g., interference signals in an image).
[0030] The goal of RPCA is to separate the low-rank matrix by optimizing the following objective function. sparse matrix :
[0031]
[0032] in: It is a low-rank matrix The nuclear norm (i.e., the sum of singular values) is used to ensure Maintain a low-rank structure. It is a sparse matrix of Norms are used to promote The number of non-zero elements in the data should be as small as possible. It is the weighting coefficient that balances the two terms, controlling the trade-off between low-rank components and sparse noise.
[0033] By solving this optimization problem, RPCA can effectively extract low-rank components from complex observation matrices. and noise or interference components The signal is extracted, thereby enabling signal recovery or interference suppression.
[0034] In SAR (Synthetic Aperture Radar) image processing, Reactive Radar Accuracy Coherence (RPCA) is commonly used to extract useful signal components from observation data containing interference and noise. SAR images are typically affected by various types of noise and interference, such as Mediatized Scattered Intrusion (MTSI), which can severely impact image quality and analysis results. Using the RPCA method, we can identify and separate these interferences from the raw data, recovering a clearer signal image.
[0035] Specifically, in the SAR image processing, we first represent the SAR image data containing interference as a matrix. Each row corresponds to a pixel in the image, and each column corresponds to different radar echo information. Then, the RPCA method is used for matrix decomposition to extract the low-rank matrix. and sparse noise matrix Low-rank matrix It contains the main signal components of the image, while the sparse matrix This includes interference information in the image.
[0036] Finally, the low-rank matrix obtained by the RPCA method will be... The suppression result is output as pseudo-clean image data for subsequent operations of this method.
[0037] Preferably, in step S104, the specific location of the interference is determined by comparing the feature and amplitude differences between the interference-removed image data and the original image data. Specifically:
[0038] During interference suppression, the feature differences between the original image and the de-interference image can effectively help us determine the specific location of the interference. By comparing the interquartile range (IQR) in the time domain and the frequency domain variability, as well as their amplitude differences in different regions of the image, we can identify areas with abnormal changes, which are usually the locations of interference.
[0039] First, the temporal interquartile range (IQR) and frequency domain variability characteristics of the original and de-interference images are calculated, and the differences between them are quantified:
[0040]
[0041]
[0042] in: and They represent the first Temporal interquartile range features of the original and de-interference images of each pulse. and These represent the frequency domain fluctuation (frequency domain variance) characteristics of the original image and the denoised image, respectively. and This indicates the differences between them.
[0043] Next, we also need to calculate the amplitude difference between the original image and the de-interference image, especially the amplitude difference in the distance frequency domain:
[0044]
[0045] in, and These represent the amplitude information of the original image and the de-interference image, respectively.
[0046] To determine the location of interference, we conduct a comprehensive analysis of characteristic differences and amplitude differences. Typically, interference regions exhibit significant variations in interquartile range, frequency domain variability, and amplitude. By setting appropriate thresholds... The difference values can be used to determine which areas are experiencing interference.
[0047]
[0048] in, This represents the difference in interquartile range in the time domain. This represents the difference in frequency domain volatility. Indicates the difference in magnitude.
[0049] Next, to pinpoint the exact location of the interference, we further determine its location in the range-frequency domain by comparing the differences between the pseudo-clean image and the original image. Specifically, the determination of interference relies not only on features in the time and frequency domains but also on calculating the amplitude differences between the original image and the pseudo-clean image in the range-frequency domain to identify the specific interference location.
[0050] For each interference pulse Calculate the amplitude difference in the distance frequency domain:
[0051] in, and These represent the frequency domain amplitude information of the original image and the de-interference image, respectively.
[0052] By setting a threshold Generate interference location mask :
[0053]
[0054] in, Indicates the first The first pulse Does each distance-frequency domain sample contain interference?
[0055] By masking all interference pulses Perform a maximization operation to obtain the final interference location mask. :
[0056]
[0057] Preferably, in step S105, by comparing the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the distance frequency domain, regions with similar intensity differences are divided into blocks, specifically:
[0058] Perform a distance-to-Fourier transform on the original image and the pseudo-clean image to obtain frequency domain data:
[0059]
[0060]
[0061] During the segmentation process, the interference intensity is estimated by comparing the amplitude difference between the pseudo-clean image and the original image in the range-frequency domain, and regions with similar intensity are segmented. First, the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the range-frequency domain is calculated:
[0062]
[0063] in, and The original image and the pseudo-clean image are respectively represented in the first... The first pulse The amplitude of each distance frequency domain sample.
[0064] By observing the amplitude difference, we can estimate the interference intensity of each interference pulse. The calculation formula is as follows:
[0065]
[0066] in, It is a mask, representing the first The interference locations of each pulse in the range-frequency domain are limited to these locations during calculation.
[0067] To divide regions with similar intensities into blocks, we calculate the intensity difference between every two regions. and set a threshold :
[0068]
[0069] if Then it is believed that the first and If the pulses containing interference belong to the same region, they should be processed uniformly.
[0070] For each interference pulse Allocate a block identifier This indicates which block it belongs to. Ultimately, the block identifier... It can be determined in the following ways:
[0071]
[0072] By using the above block division, we can generate an interference block mask for each pulse. :
[0073]
[0074] Preferably, in step S106, the threshold of the Robust Block Subspace Filtering (RBSF) method is adjusted based on the pseudo-clean image data after RPCA interference removal, as follows:
[0075] For each interference block We perform eigenvalue decomposition on the corresponding region of the pseudo-clean image. Suppose we have an image matrix for this region... Perform eigenvalue decomposition to obtain the eigenvalue sequence. ,in It is the largest eigenvalue.
[0076]
[0077] Select the largest eigenvalue of this block. As RBSF threshold .
[0078] Preferably, in step S107, RBSF suppression is performed on the distance-frequency domain of the original image data based on the auxiliary threshold of each block, thereby achieving more accurate interference suppression. The specific steps are as follows:
[0079] Robust Block Subspace Filtering (RBSF) is a signal processing method designed to suppress broadband radio frequency interference (RFI) in synthetic aperture radar (SAR) images. Its core principle lies in utilizing the differences in subspace characteristics between the signal and the interference, and ensuring the stability of the estimation through robust statistical techniques. Specifically, it can be divided into the following four key steps:
[0080] (1). Low-rank assumption: The effective signal resides in a low-rank subspace, while interference signals typically exhibit sparse behavior. The observed signal can be represented as:
[0081]
[0082] in, It is a low-rank matrix, representing the effective signal. It is a sparse matrix representing the interference signal.
[0083] (2). Robust covariance estimation: Robust methods such as MCD are used to estimate the covariance matrix to avoid the influence of outliers.
[0084]
[0085] in, Represents the total number of samples Representing a subset Includes One sample; Indicates the first indivual 3D complex eigenvectors; Representing a subset Mean vector:
[0086]
[0087] (3). Perform eigenvalue decomposition on the robust covariance matrix:
[0088]
[0089] in, It is the eigenvector matrix. .
[0090] (4). Subspace Removal: The signal is projected onto the signal subspace using a projection operator to remove interference signals.
[0091]
[0092]
[0093] In step S106, we remove the pseudo-clean image after interference using RPCA, and divide each interference block. Perform eigenvalue decomposition to obtain the largest eigenvalue of the region. and use it as the RBSF threshold. .
[0094]
[0095] For each block We will use the largest eigenvalue This threshold is used as the RBSF threshold, and is then used to suppress interference in the original image.
[0096] For each block, we obtain the eigenvalue sequence through eigenvalue decomposition. Then with RBSF threshold Compare:
[0097]
[0098] Finally, the result after RBSF inhibition Transform to the time domain to obtain the result.
[0099] Preferably, in step S108, the final result of the suppression of mutual interference of scattered waves is output, specifically as follows:
[0100] Based on the suppression results obtained in step S107, the final image data scattering suppression results are generated, which can be used for subsequent signal processing.
[0101] This step completes the closed-loop processing from RPCA preprocessing → multi-domain feature localization of interference → adaptive thresholding and segmentation → RBSF interference suppression, realizing automatic suppression and processing of image data containing mutual interference from scattered waves.
[0102] This application provides a system for detecting and suppressing mutual interference of spaceborne SAR scattered waves using a combined principal component analysis and subspace filtering method. The system includes: a feature extraction unit configured to extract time-domain interquartile range and frequency-domain variability features pulse-by-pulse from the interfering image data of the input system, and perform preliminary interference suppression using the RPCA method to obtain a pseudo-clean image; an interference judgment unit configured to accurately determine the specific location of the interference signal in the range-frequency domain based on feature comparison between the pseudo-clean image and the original image, as well as the difference in range-frequency domain amplitude; an adaptive block segmentation unit configured to divide the interference region into different blocks based on the difference in interference intensity and amplitude, and uniformly process similar regions based on the intensity difference between the blocks; a threshold calculation unit configured to calculate the maximum subspace eigenvalue of each block based on the amplitude difference in the range-frequency domain between the pseudo-clean image and the original image, and use it as the adaptive threshold for subsequent interference suppression; and an RBSF suppression unit configured to perform robust block subspace filtering (RBSF) suppression on the original image based on the adaptive threshold of each block, removing the interference signal and retaining the effective signal.
[0103] This application also provides a computer-readable storage medium storing a computer program for implementing the above-described method for detecting and suppressing mutual interference of spaceborne SAR scattering waves using joint principal component analysis and subspace filtering.
[0104] This application also provides an electronic device, including: a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the pulse radio frequency interference judgment method of multi-domain feature entropy weight fusion as described above.
[0105] Beneficial effects:
[0106] The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves, combining principal component analysis (RPCA) and subspace filtering, provided in this application, firstly extracts the interquartile range (IQR) and frequency variability features of the interfering SAR echo data pulse by pulse. Simultaneously, RPCA is used for preliminary interference suppression, resulting in a pseudo-clean image. Then, the amplitude difference between the pseudo-clean image and the original image in the range-frequency domain is calculated to determine the specific location of the interference. Furthermore, the interference region is segmented based on similarity using the amplitude difference in the range-frequency domain, merging areas with similar interference intensity for suppression. Subsequently, the pseudo-clean image is subspace decomposed, and its maximum eigenvalue is used as the dynamic threshold for RBSF. This method achieves high-precision detection and suppression even under weak mutual interference of scattered waves, while effectively preserving the integrity and reliability of the useful signal. Attached Figure Description
[0107] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. Wherein:
[0108] Figure 1 This is a schematic diagram of the process of the present invention;
[0109] Figure 2 The original image of the sentinel data described in this embodiment of the invention;
[0110] Figure 3 This is a comparison diagram showing the features of the original image and its corresponding pseudo-clean image in part a of section 2 of this invention;
[0111] Figure 4 for Figure 2 Schematic diagram of the frequency domain segmentation of the interference region in part a;
[0112] Figure 5 for Figure 2 The experimental results images for part a of the data;
[0113] Figure 6 for Figure 2 The experimental results images for part b in the middle;
[0114] Figure 7 This is a unit configuration diagram of the present invention. Detailed Implementation
[0115] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Various examples are provided by way of explanation and not by way of limitation. In fact, those skilled in the art will recognize that modifications and variations can be made to the present application without departing from the scope or spirit thereof. For example, a feature shown or described as part of one embodiment may be used in another embodiment to produce yet another embodiment. Therefore, it is desirable that the present application encompass such modifications and variations that fall within the scope of the appended claims and their equivalents.
[0116] Exemplary methods
[0117] like Figure 1 As shown, the method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering includes the following steps:
[0118] Step S101: Perform pulse-by-pulse processing on the data containing interference echoes to extract the time-domain interquartile range and frequency-domain variability features, specifically:
[0119] Interquartile Range (IQR) is a statistical metric used to describe the distribution range of signal amplitude, thereby identifying outliers and impulse interference. Its calculation formula is as follows:
[0120] First, regarding the first Amplitude sequence of pulses Calculate its interquartile range (IQR) characteristics:
[0121]
[0122] in, and They represent the first The third quartile and the first quartile of each pulse amplitude sequence are then used. The maximum interquartile range of each pulse is then taken as the characteristic value of that pulse, which can be expressed as:
[0123]
[0124] in, ,and Indicates the first The first pulse A distance-oriented sample.
[0125] By calculating the above interquartile range features pulse by pulse, the time-domain interquartile range feature sequence can be obtained:
[0126]
[0127] Frequency domain variability is a characteristic used to measure the degree of variation of a signal in the frequency domain. It reflects the degree of fluctuation in the amplitude distribution of a signal in the frequency domain, and is particularly suitable for identifying interference signals with prominent frequency components. For example, high variability in the frequency domain usually indicates that the frequency components of the signal are widely distributed, and there may be significant spectral interference. By calculating the frequency domain variance, this variation can be effectively identified, and further, a basis can be provided for interference detection and suppression.
[0128] Frequency domain volatility is quantified by calculating the variance of signal frequency domain samples, which quantifies the degree of variation in the frequency domain. The calculation formula is as follows:
[0129] For the Amplitude sequence of pulses Frequency components in the frequency domain, frequency domain wave properties It can be represented as:
[0130]
[0131] in, Indicates the first The first pulse One frequency domain sample, It is the first The frequency domain sample mean of each pulse It is the number of frequency domain samples.
[0132] By calculating the frequency domain variance of each pulse, the frequency domain wave characteristic sequence is obtained:
[0133]
[0134] in, Indicates the first Frequency domain variability of pulses.
[0135] In step S102, robust principal component analysis (RPCA) is used to suppress the SAR image data, and the output of the interference-free pseudo-clean SAR image data is as follows:
[0136] Robust Principal Component Analysis (RPCA) is a low-rank matrix factorization-based method that aims to extract low-rank principal components and sparse outliers from a data matrix. Unlike traditional Principal Component Analysis (PCA), RPCA maintains high robustness even when sparse noise or outliers are present in the data. RPCA is widely used in image processing, signal recovery, and anomaly detection, and it demonstrates its superiority, especially when processing remote sensing data containing strong noise or interference.
[0137] The basic assumption of RPCA is: given an observation matrix This matrix can be approximated as a low-rank matrix. With sparse noise matrix The sum, that is:
[0138]
[0139] in: It is the observation matrix (e.g., SAR image data with interference). It is a low-rank matrix that contains the main structure and features of the data (e.g., the main components of the signal). It is a sparse matrix that contains outliers or noise in the data (e.g., interference signals in an image).
[0140] The goal of RPCA is to separate the low-rank matrix by optimizing the following objective function. sparse matrix :
[0141]
[0142] in: It is a low-rank matrix The nuclear norm (i.e., the sum of singular values) is used to ensure Maintain a low-rank structure. It is a sparse matrix of Norms are used to promote The number of non-zero elements in the data should be as small as possible. It is the weighting coefficient that balances the two terms, controlling the trade-off between low-rank components and sparse noise.
[0143] By solving this optimization problem, RPCA can effectively extract low-rank components from complex observation matrices. and noise or interference components The signal is extracted, thereby enabling signal recovery or interference suppression.
[0144] In SAR (Synthetic Aperture Radar) image processing, Reactive Radar Accuracy Coherence (RPCA) is commonly used to extract useful signal components from observation data containing interference and noise. SAR images are typically affected by various types of noise and interference, such as Mediatized Scattered Intrusion (MTSI), which can severely impact image quality and analysis results. Using the RPCA method, we can identify and separate these interferences from the raw data, recovering a clearer signal image.
[0145] Specifically, in the SAR image processing, we first represent the SAR image data containing interference as a matrix. Each row corresponds to a pixel in the image, and each column corresponds to different radar echo information. Then, the RPCA method is used for matrix decomposition to extract the low-rank matrix. and sparse noise matrix Low-rank matrix It contains the main signal components of the image, while the sparse matrix This includes interference information in the image.
[0146] Finally, the low-rank matrix obtained by the RPCA method will be... The suppression result is output as pseudo-clean image data for subsequent operations of this method.
[0147] In step S104, by comparing the differences in features and amplitudes between the de-interference image data and the original image data, the specific location of the interference is determined, specifically as follows:
[0148] During interference suppression, the feature differences between the original image and the de-interference image can effectively help us determine the specific location of the interference. By comparing the interquartile range (IQR) in the time domain and the frequency domain variability, as well as their amplitude differences in different regions of the image, we can identify areas with abnormal changes, which are usually the locations of interference.
[0149] First, the temporal interquartile range (IQR) and frequency domain variability characteristics of the original and de-interference images are calculated, and the differences between them are quantified:
[0150]
[0151]
[0152] in: and They represent the first Temporal interquartile range features of the original and de-interference images of each pulse. and These represent the frequency domain fluctuation (frequency domain variance) characteristics of the original image and the denoised image, respectively. and This indicates the differences between them.
[0153] Next, we also need to calculate the amplitude difference between the original image and the de-interference image, especially the amplitude difference in the distance frequency domain:
[0154]
[0155] in, and These represent the amplitude information of the original image and the de-interference image, respectively.
[0156] To determine the location of interference, we conduct a comprehensive analysis of characteristic differences and amplitude differences. Typically, interference regions exhibit significant variations in interquartile range, frequency domain variability, and amplitude. By setting appropriate thresholds... The difference values can be used to determine which areas are experiencing interference.
[0157]
[0158] in, This represents the difference in interquartile range in the time domain. This represents the difference in frequency domain volatility. Indicates the difference in magnitude.
[0159] Next, to pinpoint the exact location of the interference, we further determine its location in the range-frequency domain by comparing the differences between the pseudo-clean image and the original image. Specifically, the determination of interference relies not only on features in the time and frequency domains but also on calculating the amplitude differences between the original image and the pseudo-clean image in the range-frequency domain to identify the specific interference location.
[0160] For each interference pulse Calculate the amplitude difference in the distance frequency domain:
[0161]
[0162] By setting a threshold Generate interference location mask :
[0163]
[0164] in, Indicates the first The first pulse Does each distance-frequency domain sample contain interference?
[0165] By masking all interference pulses Perform a maximization operation to obtain the final interference location mask. :
[0166]
[0167] In step S105, by comparing the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the distance frequency domain, regions with similar intensity differences are divided into blocks, specifically:
[0168] Perform a distance-to-Fourier transform on the original image and the pseudo-clean image to obtain frequency domain data:
[0169]
[0170]
[0171] During the segmentation process, the interference intensity is estimated by comparing the amplitude difference between the pseudo-clean image and the original image in the range-frequency domain, and regions with similar intensity are segmented. First, the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the range-frequency domain is calculated:
[0172]
[0173] in, and The original image and the pseudo-clean image are respectively represented in the first... The first pulse The amplitude of each distance frequency domain sample.
[0174] By observing the amplitude difference, we can estimate the interference intensity of each interference pulse. The calculation formula is as follows:
[0175]
[0176] in, It is a mask, representing the first The interference locations of each pulse in the range-frequency domain are limited to these locations during calculation.
[0177] To divide regions with similar intensities into blocks, we calculate the intensity difference between every two regions. and set a threshold :
[0178]
[0179] if Then it is believed that the first and If the pulses containing interference belong to the same region, they should be processed uniformly.
[0180] For each interference pulse Allocate a block identifier This indicates which block it belongs to. Ultimately, the block identifier... It can be determined in the following ways:
[0181]
[0182] By using the above block division, we can generate an interference block mask for each pulse. :
[0183]
[0184] In step S106, based on the pseudo-clean image data after RPCA interference removal, the threshold of the Robust Block Subspace Filtering (RBSF) method is adjusted as follows:
[0185] For each interference block We perform eigenvalue decomposition on the corresponding region of the pseudo-clean image. Suppose we have an image matrix for this region... Perform eigenvalue decomposition to obtain the eigenvalue sequence. ,in It is the largest eigenvalue.
[0186]
[0187] Select the largest eigenvalue of this block. As RBSF threshold .
[0188] In step S107, based on the auxiliary threshold of each block, RBSF suppression is performed on the distance frequency domain of the original image data to achieve more accurate interference suppression. The specific steps are as follows:
[0189] Robust Block Subspace Filtering (RBSF) is a signal processing method designed to suppress broadband radio frequency interference (RFI) in synthetic aperture radar (SAR) images. Its core principle lies in utilizing the differences in subspace characteristics between the signal and the interference, and ensuring the stability of the estimation through robust statistical techniques. Specifically, it can be divided into the following four key steps:
[0190] (1). Low-rank assumption: The effective signal resides in a low-rank subspace, while interference signals typically exhibit sparse behavior. The observed signal can be represented as:
[0191]
[0192] in, It is a low-rank matrix, representing the effective signal. It is a sparse matrix representing the interference signal.
[0193] (2). Robust covariance estimation: Enabling robust methods such as MCD to estimate the covariance matrix and avoid the influence of outliers.
[0194]
[0195] in,
[0196] (3). Perform eigenvalue decomposition on the robust covariance matrix:
[0197]
[0198] in, It is the eigenvector matrix. .
[0199] (4). Subspace Removal: The signal is projected onto the signal subspace using a projection operator to remove interference signals.
[0200]
[0201]
[0202] In step S106, we remove the pseudo-clean image after interference using RPCA, and divide each interference block. Perform eigenvalue decomposition to obtain the largest eigenvalue of the region. and use it as the RBSF threshold. .
[0203]
[0204] For each block We will use the largest eigenvalue This threshold is used as the RBSF threshold, and is then used to suppress interference in the original image.
[0205] For each block, we obtain the eigenvalue sequence through eigenvalue decomposition. Then with RBSF threshold Compare:
[0206]
[0207] Finally, the result after RBSF inhibition Transform to the time domain to obtain the result.
[0208] In step S108, the final result of the scattered wave mutual interference suppression is output, specifically as follows:
[0209] Based on the suppression results obtained in step S107, the final image data scattering suppression results are generated, which can be used for subsequent signal processing.
[0210] This step completes the closed-loop processing from RPCA preprocessing → multi-domain feature localization of interference → adaptive thresholding and segmentation → RBSF interference suppression, realizing automatic suppression and processing of image data containing mutual interference from scattered waves.
[0211] This invention employs RPCA (Robust Principal Component Analysis) for initial interference removal, which is then used as a reference threshold for RBSF (Robust Block Subspace Filtering) suppression. Utilizing information entropy theory and a dynamic threshold adaptive adjustment mechanism, it accurately identifies and locates interference signals based on varying interference intensities and complex scene characteristics, thereby improving the accuracy and stability of interference suppression. Particularly when interference signals are weak, this method maintains high-efficiency interference detection and suppression capabilities while ensuring the integrity and reliability of useful signals.
[0212] The combined principal component analysis and subspace filtering method for detecting and suppressing cross-interference in spaceborne SAR scattering waves comprehensively considers multiple characteristics, including temporal amplitude anomalies, frequency-domain energy differences, and spectral non-stationarity, and suppresses interference through adaptive thresholding. Furthermore, the method employs adaptive weighted fusion and dynamic threshold selection techniques to flexibly adapt to interference environments of varying intensities. Under strong interference signals, the method accurately locates the interference region by dynamically adjusting the threshold, effectively removing interference while preserving the integrity of the useful signal. This adaptive thresholding mechanism not only improves the accuracy of cross-interference suppression but also provides solid data support for subsequent image processing and SAR imaging quality enhancement.
[0213] like Figure 2 The figure shows the raw experimental data used in this method. Both sets of experimental data are from the Munich area of Germany. The images in part a and part b were acquired on January 8, 2021 and February 1, 2021, respectively, and the data type is spaceborne SAR single-look complex image data.
[0214] like Figure 3 As shown, the comparison results of the original image and the pseudo-clean image preprocessed by Robust Principal Component Analysis (RPCA) in the experimental examples of this application are presented in terms of multi-domain statistical features, including time-domain interquartile range and distance-frequency domain variability. The comparison shows that the abnormal fluctuations of the relevant statistical features are significantly reduced after preprocessing, providing a stable feature foundation for subsequent difference modeling and adaptive block segmentation.
[0215] like Figure 4 As shown, the azimuth-adaptive block segmentation results based on the differential feature model are presented. By analyzing the multi-domain feature differences between the original image and the pseudo-clean image, automatic segmentation of the interference region is achieved, providing a basis for the adaptive determination of subsequent subspace filtering parameters.
[0216] like Figure 5 and Figure 6 The images shown depict the experimental results of this application, with the original images sourced from [source name missing]. Figure 2 a and b. Figure 5 and Figure 6The interference suppression effects from different datasets are shown. In the figures, the upper half of the image is the original image with interference, and the lower half is the suppression result after applying this method. By comparison, it is clear that this method can effectively suppress mutual interference of scattered waves while preserving the effective information of the original signal.
[0217] Exemplary System
[0218] Figure 7 This application also provides a spaceborne SAR scattering wave mutual interference detection and suppression system combining principal component analysis and subspace filtering, comprising: a feature extraction unit configured to extract time-domain interquartile range and frequency-domain fluctuation features pulse-by-pulse from the interference-containing image data of the input system, and perform preliminary interference suppression using the RPCA method to obtain a pseudo-clean image; an interference judgment unit configured to accurately determine the specific location of the interference signal in the range-frequency domain based on the feature comparison between the pseudo-clean image and the original image and the difference in range-frequency domain amplitude; an adaptive block division unit configured to divide the interference region into different blocks based on the difference in interference intensity and amplitude, and uniformly process similar regions based on the intensity difference between blocks; a threshold calculation unit configured to calculate the maximum subspace eigenvalue of each block based on the difference in amplitude between the pseudo-clean image and the original image in the range-frequency domain, and use it as the adaptive threshold for subsequent interference suppression; and an RBSF suppression unit configured to perform robust block subspace filtering (RBSF) suppression on the original image based on the adaptive threshold of each block, removing the interference signal and retaining the effective signal.
[0219] The method and system for detecting and suppressing mutual interference of spaceborne SAR scattered waves based on joint principal component analysis and subspace filtering provided in this application can implement any of the above-mentioned interference suppression steps and processing procedures, and achieve the same technical effect. Therefore, the specific steps and processing procedures will not be described in detail here.
[0220] Exemplary device
[0221] This application provides an electronic device, including a storage device and a processor. The processor is suitable for executing various programs; the memory is used to store multiple programs. The feature is that when the memory executes the programs on the processor, it implements any of the aforementioned methods for detecting and suppressing mutual interference of spaceborne SAR scattered waves using joint principal component analysis and subspace filtering.
[0222] Since the steps of the method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering have been described in detail in the specific implementation examples, they will not be repeated here.
[0223] The processor includes a Central Processing Unit (CPU), a Network Processor (NP), etc., and can also be a digital signal processor, an application-specific integrated circuit, an off-the-shelf programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.
[0224] Specifically, the processor can be configured as follows: It extracts time-domain interquartile range and frequency-domain fluctuation features pulse-by-pulse from the synthetic aperture radar image data of the input system, and performs preliminary interference suppression using the RPCA method to obtain a pseudo-clean image; based on the feature comparison between the pseudo-clean image and the original image, and the difference in range-frequency domain amplitude, it accurately determines the specific location of the interference signal in the range-frequency domain; based on the difference in interference intensity and amplitude, it divides the interference region into different blocks, and processes similar regions uniformly based on the intensity difference between the blocks; furthermore, based on the amplitude difference between the pseudo-clean image and the original image in the range-frequency domain, it calculates the maximum eigenvalue of the subspace of each block, and uses it as the adaptive threshold for subsequent interference suppression; finally, based on the adaptive threshold of each block, it performs robust block subspace filtering (RBSF) suppression on the original image to remove the interference signal, retain the effective signal, and output a clean image.
[0225] It should be noted that, depending on the implementation needs, the various components / steps described in the embodiments of this application can be broken down into more components / steps, or two or more components / steps or parts of the operation of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of this application.
[0226] The methods described in the embodiments of this application can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code downloaded over a network that is originally stored in a remote recording medium or a non-transitory machine storage medium and will be stored in a local recording medium. Thus, the methods described herein can be processed by software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the single-track synthetic aperture radar jamming source localization method described herein is implemented. Furthermore, when a general-purpose computer accesses the code used to implement the methods shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for executing the methods shown herein.
[0227] Those skilled in the art will recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application of the technical solution and the constraints involved. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this application.
[0228] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments.
[0229] The device and system embodiments described above are merely illustrative. The units referred to as separate entities may or may not be physically separate. The entities mentioned as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0230] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering, characterized in that, Includes the following steps: Step S101: Preprocess the SAR image data containing interference, i.e., the original image data, and extract the time-domain interquartile range and frequency-domain fluctuation characteristics. Step S102: Use Robust Principal Component Analysis (RPCA) to suppress the original image data and output approximately interference-free pseudo-clean SAR image data after interference removal. Step S103: Post-process the interference-free pseudo-clean SAR image data to extract the time-domain interquartile range and frequency-domain fluctuation characteristics. Step S104: By comparing the characteristics and amplitude differences between the de-interference pseudo-clean SAR image data and the original image data, determine the specific location of the interference; Step S105: By comparing the amplitude difference between the pseudo-clean image and the original image in the distance frequency domain, regions with similar intensity differences are divided into blocks; Step S106: Based on the pseudo-clean SAR image data after RPCA interference removal, the threshold of the Robust Block Subspace Filter (RBSF) method is adjusted. Step S107: Based on the auxiliary threshold of each block, perform RBSF suppression on the distance frequency domain of the original image data to achieve more accurate interference suppression. Step S108: Output the clean SAR image data after interference removal and display the final interference suppression results.
2. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S101, pulse-by-pulse processing is performed on the data containing interference echoes to extract the time-domain interquartile range and frequency-domain variability features, specifically: First, regarding the first Amplitude sequence of pulses Calculate its interquartile range (IQR) characteristics: ； in, and They represent the first The third quartile and the first quartile of each pulse amplitude sequence are used; then, the maximum interquartile range of each pulse is taken as the characteristic value of that pulse, which can be expressed as: ； in, ,and Indicates the first The first pulse One distance-oriented sample; By calculating the above interquartile range features pulse by pulse, the time-domain interquartile range feature sequence can be obtained: ； Frequency domain volatility is quantified by calculating the variance of signal frequency domain samples, which quantifies the degree of variation in the frequency domain. The calculation formula is as follows: For the Amplitude sequence of pulses Frequency components in the frequency domain, frequency domain wave properties It can be represented as: ； in, Indicates the first The first pulse One frequency domain sample, It is the first The frequency domain sample mean of each pulse It is the number of frequency domain samples; By calculating the frequency domain variance of each pulse, the frequency domain wave characteristic sequence is obtained: ； in, Indicates the first Frequency domain variability of pulses.
3. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S102, Robust Principal Component Analysis (RPCA) is used to suppress SAR image data, and the output of the interference-free pseudo-clean SAR image data is as follows: Given an observation matrix This matrix can be approximated as a low-rank matrix. With sparse noise matrix The sum, that is: ； in: It is the observation matrix. It is a low-rank matrix, containing the main structure and features of the data. It is a sparse matrix that contains outliers or noise in the data; The goal of RPCA is to separate the low-rank matrix by optimizing the following objective function. sparse matrix : ； in: It is a low-rank matrix The nuclear norm, which is the sum of singular values, is used to ensure Maintain a low-rank structure. It is a sparse matrix of Norms are used to promote The number of non-zero elements in the data should be as small as possible. It is the weighting coefficient that balances the two terms, controlling the trade-off between low-rank components and sparse noise; By solving this optimization problem, RPCA can effectively extract low-rank components from complex observation matrices. and noise or interference components The signal is extracted, thereby enabling signal recovery or interference suppression.
4. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S104, the specific location of the interference is determined by comparing the feature and amplitude differences between the interference-removed image data and the original image data. Specifically: First, the temporal interquartile range (IQR) and frequency domain variability characteristics of the original and de-interference images are calculated, and the differences between them are quantified: ； ； in: and They represent the first Temporal interquartile range features of the original and de-interference images of each pulse. and These represent the frequency domain fluctuations, or frequency domain variance characteristics, of the original image and the de-interference image, respectively. and Indicate the differences between them; Next, we also need to calculate the amplitude difference between the original image and the de-interference image, especially the amplitude difference in the distance frequency domain: ； in, and These represent the amplitude information of the original image and the de-interference image, respectively; By setting an appropriate threshold Based on the difference values, determine which areas are affected by interference: ； in, This represents the difference in interquartile range in the time domain. This represents the difference in frequency domain volatility. Indicates the difference in magnitude; Next, to pinpoint the exact location of the interference, we further determine the interference location in the range-frequency domain by comparing the differences between the pseudo-clean image and the original image; specifically: For each interference pulse Calculate the amplitude difference in the distance frequency domain: ； in, and These represent the frequency domain amplitude information of the original image and the de-interference image, respectively; By setting a threshold Generate interference location mask : in, Indicates the first The first pulse Does each distance-frequency domain sample contain interference? By masking all interference pulses Perform a maximization operation to obtain the final interference location mask. : 。 5. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S105, by comparing the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the distance frequency domain, regions with similar intensity differences are divided into blocks, specifically: Perform a distance-to-Fourier transform on the original image and the pseudo-clean image to obtain frequency domain data: ； ； First, calculate the amplitude difference between the pseudo-clean image and the original image in the interference-containing region in the distance frequency domain: ； in, and The original image and the pseudo-clean image are respectively represented in the first... The first pulse The amplitude of each distance frequency domain sample; By observing the amplitude difference, we can estimate the interference intensity of each interference pulse. The calculation formula is as follows: ； in, It is a mask, representing the first The interference locations of each pulse in the range-frequency domain are limited to these locations during calculation; To divide regions with similar intensities into blocks, we calculate the intensity difference between every two regions. and set a threshold : ； if Then it is believed that the first and If the regions containing interference from individual pulses belong to the same region, they should be processed uniformly. For each interference pulse Allocate a block identifier This indicates which block it belongs to; ultimately, the block identifier... It can be determined in the following ways: ； By using the above block division, we can generate an interference block mask for each pulse. : 。 6. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S106, based on the pseudo-clean image data after RPCA interference removal, the threshold of the Robust Block Subspace Filtering (RBSF) method is adjusted as follows: For each interference block We perform eigenvalue decomposition on the corresponding region of the pseudo-clean image. Suppose we have an image matrix for this region... Perform eigenvalue decomposition to obtain the eigenvalue sequence. ,in It is the largest eigenvalue; ； Select the largest eigenvalue of this block. As RBSF threshold .
7. The method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering according to claim 1, characterized in that, In step S107, based on the auxiliary threshold of each block, RBSF suppression is performed on the distance frequency domain of the original image data to achieve more accurate interference suppression. The specific steps are as follows: (1). Low-rank assumption: The effective signal resides in a low-rank subspace, while interference signals typically exhibit sparse behavior; the observed signal can be represented as: ； in, It is a low-rank matrix, representing the effective signal. It is a sparse matrix representing the interference signal; (2). Robust covariance estimation: Robust methods such as MCD are used to estimate the covariance matrix to avoid the influence of outliers. ;in Represents the total number of samples Representing a subset Includes One sample; Indicates the first indivual 3D complex eigenvectors; Representing a subset Mean vector: ； (3). Perform eigenvalue decomposition on the robust covariance matrix: ； in, It is the eigenvector matrix. , Represents the robust covariance matrix or robust subspace reconstruction matrix; (4). Subspace Removal: The signal is projected onto the signal subspace using a projection operator to remove interference signals. ;in, Represents the robust block subspace filtering matrix; Represents the identity matrix; Indicates the first The feature vector matrix of each block that is determined to be an interference subspace; This represents the original observation vector in the range frequency domain; This indicates the suppression result after RBSF filtering; In step S106, we remove the pseudo-clean image after interference using RPCA, and divide each interference block. Perform eigenvalue decomposition to obtain the largest eigenvalue of the region. and use it as the RBSF threshold. ; ； For each block We will use the largest eigenvalue This threshold is used as the RBSF threshold, and is then used to suppress interference in the original image. For each block, we obtain the eigenvalue sequence through eigenvalue decomposition. Then with RBSF threshold Compare: ； Finally, the result after RBSF inhibition Transform to the time domain to obtain the result.
8. A system for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering, characterized in that, include: The feature extraction unit is configured to extract time-domain interquartile range and frequency-domain variability features pulse-by-pulse from the interference-containing image data of the input system, and perform preliminary interference suppression using the RPCA method to obtain a pseudo-clean image; The interference judgment unit is configured to accurately determine the specific location of the interference signal in the range-frequency domain based on the feature comparison between the pseudo-clean image and the original image and the difference in range-frequency domain amplitude. The adaptive segmentation unit is configured to divide the interference region into different segments based on the differences in interference intensity and amplitude, and to process similar regions uniformly based on the intensity differences between segments. The threshold calculation unit is configured to calculate the maximum subspace feature value of each block based on the amplitude difference between the pseudo-clean image and the original image in the distance frequency domain, and use it as the adaptive threshold for subsequent interference suppression. The RBSF suppression unit is configured to perform robust block subspace filtering RBSF suppression on the original image based on an adaptive threshold for each block, thereby removing interference signals and retaining valid signals.
9. A storage medium storing a plurality of programs, characterized in that, The program application is loaded and executed by the processor to implement the method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering as described in any one of claims 1-7.
10. An electronic device, comprising a storage device and a processor; the processor being adapted to execute various programs; the memory being used to store multiple programs; characterized in that, When the memory executes the program on the processor, it implements the method for detecting and suppressing mutual interference of spaceborne SAR scattered waves by combining principal component analysis and subspace filtering as described in claim 9.