Pulse type radio frequency interference judgment method and system based on multi-domain feature entropy weight fusion, storage medium and electronic equipment

Abstract

The invention discloses a pulse type radio frequency interference judgment method based on multi-domain feature entropy weight fusion, a storage medium and electronic equipment, and the method comprises the following steps: carrying out pulse-by-pulse processing on interference-containing echo data, respectively extracting a time domain statistical center deviation degree feature and a time domain sharpness feature, and extracting a frequency domain kurtosis feature and a frequency domain skewness first-order difference feature, and carrying out pulse-by-pulse processing on the extracted feature; constructing a pulse-by-pulse multi-domain feature vector sequence; then, based on the information entropy theory, adaptive weighted fusion is carried out on the multi-domain features, pulse-by-pulse interference detection scores are obtained, statistical histogram distribution is constructed according to the detection scores, and Gaussian smoothing processing is carried out; further, an OTSU self-adaptive threshold judgment method is adopted for segmentation, and automatic judgment of pulse-containing interference pulses is achieved; and outputting a pulse type interference judgment result. According to the invention, the abnormal characteristics of the pulse-type interference in the multi-domain feature space can be effectively represented without manual threshold setting, and the judgment accuracy and stability under the weak pulse interference condition are improved.

Description

Technical Field

[0001] This application relates to the field of signal processing technology, and in particular to a pulsed radio frequency interference judgment method, system, computer-readable storage medium, and electronic device based on multi-domain feature entropy weight fusion. Background Technology

[0002] Synthetic Aperture Radar (SAR) is an active microwave imaging system capable of high-resolution ground observation in all weather and at all times. SAR achieves high range resolution by transmitting broadband signals and synthesizing them using platform motion to achieve high azimuth resolution. It is widely used in topographic mapping, disaster monitoring, agricultural surveillance, and military reconnaissance.

[0003] However, in increasingly complex electromagnetic environments, SAR systems are inevitably affected by radio frequency interference (RFI), among which pulse RFI (PRFI) is a typical form of RFI. PRFI usually originates from ground radar or communication equipment whose frequencies overlap with the SAR signal. It is characterized by unidirectional propagation, high power, short duration, and short pulse periodicity, resulting in bright lines or spike structures in the SAR echo, which seriously affects the effective reception of the echo signal and the subsequent imaging quality.

[0004] Currently, methods for identifying pulse interference mainly fall into two categories: time-domain and transform-domain methods. Time-domain methods characterize the transient characteristics of interference pulses through amplitude abrupt changes, peak values, or differential analysis, but they are prone to missed detections or false alarms in weak interference or noisy environments. Transform-domain methods rely on concentrated spectral energy, spectral asymmetry, or short-time Fourier transform analysis to identify interference characteristics, reflecting the concentration and non-stationary changes of interference in the frequency domain, but typically ignoring dynamic changes between pulses. Furthermore, existing methods often employ fixed thresholds, making them difficult to adapt to different interference intensities and complex scenarios, leading to unstable detection results or large errors. Especially in weak-energy interference scenarios, traditional methods are prone to missed detections, while in strong interference scenarios, they may generate numerous false alarms, directly hindering subsequent interference suppression and improvements in SAR imaging quality. Summary of the Invention

[0005] The purpose of this application is to provide a pulsed radio frequency interference judgment method, system, storage medium and electronic device based on multi-domain feature entropy weight fusion, 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 pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion, including: step S101, processing the echo data pulse by pulse to extract time-domain statistical center deviation and sharpness features; step S102, extracting frequency-domain kurtosis and skewness first-order difference features pulse by pulse to construct a multi-domain feature vector sequence; step S103, determining the adaptive weights of the pulse-by-pulse multi-domain features based on information entropy theory; step S104, performing weighted fusion of the pulse-by-pulse multi-domain features to obtain an interference detection score sequence; step S105, constructing a statistical histogram distribution based on the detection scores and performing Gaussian smoothing; step S106, segmenting using OTSU adaptive threshold decision to achieve automatic judgment of pulses containing pulse-type interference; step S107, outputting the pulse-type interference judgment result.

[0008] Preferably, in step S101, the echo data is processed pulse by pulse to extract the time-domain statistical center deviation and sharpness features, specifically as follows:

[0009] The formula for calculating the deviation of the statistical center in the time domain is expressed as follows:

[0010]

[0011] in, Indicates the first A sample of amplitude values, Indicates the sample median. Indicates the absolute median difference. To prevent the introduction of tiny positive numbers by dividing by zero.

[0012] Then, for the first The statistical center deviation of the amplitude sequence of each pulse is calculated, and the maximum value is taken as the characteristic value of the pulse, which can be expressed as:

[0013]

[0014] in , Indicates the first The first pulse A distance-oriented sample, Indicates the total number of azimuth pulses. It is the absolute deviation of the median.

[0015] By performing the above processing pulse by pulse, the time-domain statistical center deviation feature sequence can be obtained. :

[0016]

[0017] It is the first The statistical center deviation characteristic value corresponding to each pulse.

[0018] The formula for calculating the time-domain sharpness feature is expressed as follows:

[0019]

[0020] in, and These represent the minimum and maximum values ​​within the sliding window to the left of the sample point, respectively. and These represent the minimum and maximum values ​​within the sliding window to the right of the sample point, respectively. The squaring operation is used to enhance the response to sharp peaks and abrupt changes.

[0021] Then, for the first The sharpness sequence corresponding to each pulse amplitude sequence is calculated and normalized. To highlight the extreme peak characteristics, the high quantile statistic is selected from the normalized sharpness sequence as the time-domain sharpness feature of the pulse. The result can be expressed as follows:

[0022]

[0023] in, , Indicates the first The sharpness sequence corresponding to each pulse.

[0024] The high quantile percentage is close to 1; The value range is [0.9, 0.999]; by performing the above processing pulse by pulse, the temporal sharpness feature sequence of all pulses can be obtained. Its form can be expressed as:

[0025]

[0026] By calculating the above features pulse by pulse, the complete time-domain statistical center deviation feature sequence SCD and the time-domain sharpness feature sequence Sharp can be obtained, providing input basis for subsequent multi-domain feature entropy weight fusion.

[0027] Preferably, in step S102, the first-order difference features of frequency domain kurtosis and skewness are extracted pulse by pulse to construct a multi-domain feature vector sequence, specifically as follows:

[0028] The formula for calculating kurtosis is expressed as:

[0029]

[0030] in, Represents the first value in the spectral amplitude sequence Sampling points ,

[0031] This represents the total number of frequency domain sampling points. The mean of the spectral amplitude sequence is given; the frequency domain kurtosis coefficient is the ratio of the fourth central moment to the square of the second central moment.

[0032] Then, for the first Perform a frequency domain transformation on each pulse to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain kurtosis features are calculated. The result can be expressed as:

[0033]

[0034] in, , This indicates the total number of azimuth pulses.

[0035] By performing the above processing pulse by pulse, the frequency domain kurtosis feature sequence of all pulses can be obtained, which can be expressed as:

[0036]

[0037] The formula for calculating the frequency domain skewness characteristic can be expressed as:

[0038]

[0039] in, Represents the first value in the spectral amplitude sequence One sampling point, , representing the total number of frequency domain sampling points, The mean of the spectral amplitude sequence; the frequency domain skewness is obtained by the third and second central moments. The power ratio characterizes the asymmetry of the spectral distribution, making this structure highly sensitive to frequency domain energy shifts and asymmetric changes in spectral profile.

[0040] Then, for the first The frequency domain transform of the time-domain echo of each pulse is performed to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain skewness is calculated, and the result can be expressed as:

[0041]

[0042] in, , This indicates the total number of azimuth pulses.

[0043] To further characterize the degree of variation of frequency domain statistical properties between adjacent pulses, a first-order difference feature of frequency domain skewness is introduced, which is defined as the absolute value of the difference in frequency domain skewness between adjacent pulses:

[0044]

[0045] in, , when When the frequency domain skewness first-order difference is defined as zero.

[0046] By performing the above processing pulse by pulse, the first-order difference feature sequence of frequency domain skewness corresponding to all pulses is obtained. Its form can be expressed as:

[0047]

[0048] After obtaining the time-domain statistical center deviation feature, time-domain sharpness feature, frequency-domain kurtosis feature, and frequency-domain skewness first-order difference feature of each pulse, in order to achieve a multi-angle joint characterization of pulse interference, the features of different domains and different behavioral mechanisms mentioned above are modeled in a unified manner to construct a pulse-by-pulse multi-domain feature vector sequence.

[0049] Specifically, for the first For each pulse, the corresponding time-domain statistical center deviation characteristic is... Temporal sharpness characteristics Frequency domain kurtosis characteristics and frequency domain skewness first-order difference characteristics By performing joint characterization, a multi-domain feature vector corresponding to the pulse is formed, which can be expressed as:

[0050]

[0051] in, , This indicates the total number of azimuth pulses.

[0052] By performing the above feature construction process on all pulses, a complete pulse-by-pulse multi-domain feature vector sequence can be obtained, the overall form of which can be expressed as:

[0053]

[0054] Preferably, in step S103, adaptive weight determination is performed on the pulse-by-pulse multi-domain features based on information entropy theory, specifically as follows:

[0055] The pulse-by-pulse multi-domain feature vector sequences obtained in steps S101 and S102 are jointly represented by pulse indices to form a pulse-by-pulse multi-domain feature matrix:

[0056]

[0057] in, Represents the feature dimension. This indicates the total number of pulses.

[0058] To eliminate differences in dimensions, amplitude range, and statistical distribution among different features, and to avoid the adverse effects of feature numerical scale on weight allocation results, the feature matrix is ​​normalized according to its feature dimension, as expressed by:

[0059]

[0060] in, , To prevent the introduction of tiny positive numbers with a denominator of zero, It is the result of normalization.

[0061] Based on the normalized features, the values ​​of each feature on the pulse sequence are treated as discrete random variables, and the probability distributions corresponding to each feature are constructed:

[0062]

[0063] Furthermore, based on information entropy theory, the uncertainty and dispersion of each feature in the impulse dimension are quantitatively evaluated, and the 1st... Information entropy corresponding to each feature:

[0064]

[0065] It is a logarithm with base 10.

[0066] Based on the information entropy results, a difference coefficient is introduced to characterize the degree of contribution of each feature to the effective information of interference discrimination:

[0067]

[0068] The difference coefficients are then normalized to obtain the adaptive weights corresponding to each feature:

[0069]

[0070] Preferably, in step S104, the pulse-by-pulse multi-domain features are weighted and fused to obtain an interference detection score sequence, specifically as follows:

[0071] The multi-domain features are linearly weighted and fused according to their adaptive weights to obtain the pulse-by-pulse comprehensive interference discrimination score:

[0072]

[0073] in, These are the adaptive weights corresponding to each feature.

[0074] The comprehensive discrimination score Used to characterize the The significance of interference from each pulse is determined and used as the input for subsequent Gaussian smoothing histograms.

[0075] Preferably, in step S105, a statistical histogram distribution is constructed based on the detection scores, and Gaussian smoothing is performed, specifically as follows:

[0076] First, the pulse-by-pulse interference detection score sequence is analyzed. After linear normalization, the expression is as follows:

[0077]

[0078] in, A tiny positive number introduced to prevent the denominator from being zero.

[0079] To facilitate subsequent threshold analysis based on statistical distribution, the normalized detection score is mapped to a finite discrete grayscale space, specifically, it is linearly mapped to an 8-bit grayscale range. Its expression is:

[0080]

[0081] Based on this, a statistical histogram is constructed for the mapped detection score sequence, and its histogram counting function can be expressed as:

[0082]

[0083] in, This indicates that the test score falls into the first category. The number of pulses within each grayscale range .

[0084] Gaussian smoothing of a histogram counting sequence yields the following result:

[0085]

[0086] in, The standard deviation of the Gaussian kernel is used to control the smoothing intensity. This represents the mean.

[0087] Preferably, in step S106, OTSU adaptive threshold decision is used for segmentation to achieve automatic determination of pulse-type interference pulses, specifically:

[0088] The Gaussian-smoothed grayscale histogram is considered as the statistical distribution of pulse-by-pulse detection scores in the grayscale space. By normalizing this histogram, it is converted into a probability distribution to eliminate the influence of pulse number and data size variations on the threshold calculation results, resulting in:

[0089]

[0090] Based on this, candidate thresholds are introduced. The detected score samples are divided into two categories: low score and high score, which correspond to the potential set of normal pulses and the set of pulses containing interference, respectively.

[0091] Further calculate the cumulative probability and cumulative mean corresponding to the threshold:

[0092]

[0093] Represents the cumulative probability. This represents the cumulative mean.

[0094] Meanwhile, the global mean of all samples is defined as:

[0095]

[0096] Based on the above statistics, an inter-class variance function is constructed to measure the threshold. Separation capability for two types of samples:

[0097]

[0098] By searching the entire grayscale range for the threshold that maximizes the inter-class variance. This allows us to obtain the optimal adaptive segmentation position.

[0099] The optimal grayscale threshold is mapped back to the normalized detection score domain to obtain the final decision threshold:

[0100]

[0101] in, The optimal grayscale threshold that maximizes the inter-class variance is obtained through the OTSU method; finally, the pulse-by-pulse interference detection score is... Greater than the adaptive threshold When the pulse is not in the specified state, it is determined to be a pulse containing interference; otherwise, it is determined to be a normal pulse, thus completing the automatic interference judgment at the pulse-by-pulse level.

[0102] Preferably, in step S107, the pulse interference judgment result is output, specifically as follows:

[0103] Based on the obtained pulse-by-pulse adaptive decision results, the final pulse-type interference judgment sequence is generated, which can be directly used for subsequent interference suppression.

[0104] This step completes the closed-loop processing from pulse-by-pulse multi-domain feature extraction → information entropy weighted fusion → adaptive threshold decision → interference pulse determination, realizing the automatic judgment and labeling of pulse-type interference pulses.

[0105] This application embodiment also provides a pulse interference judgment system with multi-domain feature entropy weight fusion and adaptive threshold, including: a feature extraction unit configured to extract time-domain statistical center deviation and sharpness features pulse by pulse from the interference echo data of the input system, and calculate frequency-domain kurtosis and frequency-domain skewness first-order difference features pulse by pulse, thereby constructing a multi-domain feature vector for each pulse; and a feature weight calculation unit configured to calculate the difference coefficient of each feature based on information entropy theory for the multi-domain feature vector sequence, and normalize it to generate corresponding adaptive weights;

[0106] The interference judgment unit is configured to perform weighted fusion of the multi-domain features according to their adaptive weights to obtain a pulse-by-pulse interference detection score sequence, normalize and map it to a grayscale space to construct a histogram, and after Gaussian smoothing, use OTSU adaptive threshold to judge the pulses so as to realize the automatic judgment of pulses containing pulse-type interference; the result output unit is configured to generate a pulse-type interference judgment sequence based on the pulse-by-pulse judgment result for use in subsequent interference suppression processing.

[0107] This application also provides a computer-readable storage medium storing a computer program thereon, the program being a pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion as described above.

[0108] 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.

[0109] Beneficial effects:

[0110] The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion provided in this application firstly extracts multi-domain features such as time-domain statistical center deviation, sharpness, frequency-domain kurtosis, and first-order difference of frequency-domain skewness from the pulse-by-pulse SAR echo containing interference, and constructs a pulse-by-pulse multi-domain feature vector sequence. Then, based on information entropy theory, adaptive weight allocation is performed on the multi-domain features, and the pulse features are weighted and fused to obtain a pulse-by-pulse comprehensive interference discrimination score sequence. Subsequently, the detection scores are subjected to statistical histogram construction and smoothing, and automatic identification of pulse-type interference pulses is achieved through adaptive threshold decision. Finally, the pulse-type interference judgment result is output, which can be directly used for subsequent interference suppression or signal recovery. This invention enables high-precision judgment even under weak pulse-type interference conditions, while ensuring the integrity and reliability of useful signals, providing a reliable data foundation for subsequent interference suppression. Attached Figure Description

[0111] 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:

[0112] Figure 1 This is a schematic diagram of the process of the present invention;

[0113] Figure 2 These are comparative images of data experiments described in embodiments of the present invention;

[0114] Figure 3 This is a unit configuration diagram according to this application. 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 pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion includes:

[0118] In step S101, pulse-by-pulse processing is performed on the echo data to extract the time-domain statistical center deviation and sharpness features;

[0119] Specifically, the formula for calculating the deviation of the time-domain statistical center is as follows:

[0120]

[0121] in, Indicates the first A sample of amplitude values, Indicates the sample median. Indicates the absolute median difference. To prevent the introduction of tiny positive numbers by dividing by zero.

[0122] Then, for the first The statistical center deviation of the amplitude sequence of each pulse is calculated, and the maximum value is taken as the characteristic value of the pulse, which can be expressed as:

[0123]

[0124] in , Indicates the first The first pulse A distance-oriented sample, Indicates the total number of azimuth pulses. It is the absolute deviation of the median.

[0125] By performing the above processing pulse by pulse, the time-domain statistical center deviation feature sequence can be obtained. :

[0126]

[0127] It is the first The statistical center deviation characteristic value corresponding to each pulse.

[0128] The formula for calculating the time-domain sharpness feature is expressed as follows:

[0129]

[0130] in, and These represent the minimum and maximum values ​​within the sliding window to the left of the sample point, respectively. and These represent the minimum and maximum values ​​within the sliding window to the right of the sample point, respectively. The squaring operation is used to enhance the response to sharp peaks and abrupt changes.

[0131] Then, for the first The sharpness sequence corresponding to each pulse amplitude sequence is calculated and normalized. To highlight the extreme peak characteristics, the high quantile statistic is selected from the normalized sharpness sequence as the time-domain sharpness feature of the pulse. The result can be expressed as follows:

[0132]

[0133] in, , Indicates the first The sharpness sequence corresponding to each pulse.

[0134] The high quantile percentage is close to 1; The value range is [0.9, 0.999]; by performing the above processing pulse by pulse, the temporal sharpness feature sequence of all pulses can be obtained. Its form can be expressed as:

[0135]

[0136] By calculating the above features pulse by pulse, the complete time-domain statistical center deviation feature sequence SCD and the time-domain sharpness feature sequence Sharp can be obtained, providing input basis for subsequent multi-domain feature entropy weight fusion.

[0137] In step S102, the first-order difference features of frequency domain kurtosis and skewness are extracted pulse by pulse to construct a multi-domain feature vector sequence;

[0138] Specifically, the formula for calculating kurtosis is as follows:

[0139]

[0140] in, Represents the first value in the spectral amplitude sequence Sampling points ,

[0141] This represents the total number of frequency domain sampling points. The mean of the spectral amplitude sequence is given; the frequency domain kurtosis coefficient is the ratio of the fourth central moment to the square of the second central moment.

[0142] Then, for the first Perform a frequency domain transformation on each pulse to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain kurtosis features are calculated. The result can be expressed as:

[0143]

[0144] in, , This indicates the total number of azimuth pulses.

[0145] By performing the above processing pulse by pulse, the frequency domain kurtosis feature sequence of all pulses can be obtained, which can be expressed as:

[0146]

[0147] The formula for calculating the frequency domain skewness characteristic can be expressed as:

[0148]

[0149] in, Represents the first value in the spectral amplitude sequence One sampling point, , representing the total number of frequency domain sampling points, The mean of the spectral amplitude sequence; the frequency domain skewness is obtained by the third and second central moments. The power ratio characterizes the asymmetry of the spectral distribution, making this structure highly sensitive to frequency domain energy shifts and asymmetric changes in spectral profile.

[0150] Then, for the first The frequency domain transform of the time-domain echo of each pulse is performed to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain skewness is calculated, and the result can be expressed as:

[0151]

[0152] in, , This indicates the total number of azimuth pulses.

[0153] To further characterize the degree of variation of frequency domain statistical properties between adjacent pulses, a first-order difference feature of frequency domain skewness is introduced, which is defined as the absolute value of the difference in frequency domain skewness between adjacent pulses:

[0154]

[0155] in, , when When the frequency domain skewness first-order difference is defined as zero.

[0156] By performing the above processing pulse by pulse, the first-order difference feature sequence of frequency domain skewness corresponding to all pulses is obtained. Its form can be expressed as:

[0157]

[0158] After obtaining the time-domain statistical center deviation feature, time-domain sharpness feature, frequency-domain kurtosis feature, and frequency-domain skewness first-order difference feature of each pulse, in order to achieve a multi-angle joint characterization of pulse interference, the features of different domains and different behavioral mechanisms mentioned above are modeled in a unified manner to construct a pulse-by-pulse multi-domain feature vector sequence.

[0159] Specifically, for the first For each pulse, the corresponding time-domain statistical center deviation characteristic is... Temporal sharpness characteristics Frequency domain kurtosis characteristics and frequency domain skewness first-order difference characteristics By performing joint characterization, a multi-domain feature vector corresponding to the pulse is formed, which can be expressed as:

[0160]

[0161] in, , This indicates the total number of azimuth pulses.

[0162] By performing the above feature construction process on all pulses, a complete pulse-by-pulse multi-domain feature vector sequence can be obtained, the overall form of which can be expressed as:

[0163]

[0164] In step S103, adaptive weights are determined for pulse-by-pulse multi-domain features based on information entropy theory;

[0165] Specifically: The pulse-by-pulse multi-domain feature vector sequences obtained in steps S101 and S102 are jointly represented by pulse indices to form a pulse-by-pulse multi-domain feature matrix:

[0166]

[0167] in, Represents the feature dimension. This indicates the total number of pulses.

[0168] To eliminate differences in dimensions, amplitude range, and statistical distribution among different features, and to avoid the adverse effects of feature numerical scale on weight allocation results, the feature matrix is ​​normalized according to its feature dimension, as expressed by:

[0169]

[0170] in, , To prevent the introduction of tiny positive numbers with a denominator of zero, It is the result of normalization.

[0171] Based on the normalized features, the values ​​of each feature on the pulse sequence are treated as discrete random variables, and the probability distributions corresponding to each feature are constructed:

[0172]

[0173] Furthermore, based on information entropy theory, the uncertainty and dispersion of each feature in the impulse dimension are quantitatively evaluated, and the 1st... Information entropy corresponding to each feature:

[0174]

[0175] It is a logarithm with base 10.

[0176] Based on the information entropy results, a difference coefficient is introduced to characterize the degree of contribution of each feature to the effective information of interference discrimination:

[0177]

[0178] The difference coefficients are then normalized to obtain the adaptive weights corresponding to each feature:

[0179]

[0180] In step S104, the pulse-by-pulse multi-domain features are weighted and fused to obtain the interference detection score sequence;

[0181] Specifically: Multi-domain features are linearly weighted and fused according to their adaptive weights to obtain a pulse-by-pulse comprehensive interference discrimination score.

[0182]

[0183] in, These are the adaptive weights corresponding to each feature.

[0184] The comprehensive discrimination score Used to characterize the The significance of interference from each pulse is determined and used as the input for subsequent Gaussian smoothing histograms.

[0185] In step S105, a statistical histogram distribution is constructed based on the detection scores, and Gaussian smoothing is performed.

[0186] Specifically: First, the pulse-by-pulse interference detection score sequence After linear normalization, the expression is as follows:

[0187]

[0188] in, A tiny positive number introduced to prevent the denominator from being zero.

[0189] To facilitate subsequent threshold analysis based on statistical distribution, the normalized detection score is mapped to a finite discrete grayscale space, specifically, it is linearly mapped to an 8-bit grayscale range. Its expression is:

[0190]

[0191] Based on this, a statistical histogram is constructed for the mapped detection score sequence, and its histogram counting function can be expressed as:

[0192]

[0193] in, This indicates that the test score falls into the first category. The number of pulses within each grayscale range .

[0194] Gaussian smoothing of a histogram counting sequence yields the following result:

[0195]

[0196] in, The standard deviation of the Gaussian kernel is used to control the smoothing intensity. This represents the mean.

[0197] In step S106, the OTSU adaptive threshold decision is used for segmentation to achieve automatic determination of pulse-type interference pulses;

[0198] Specifically: The Gaussian-smoothed grayscale histogram is considered as the statistical distribution of pulse-by-pulse detection scores in the grayscale space; by normalizing this histogram, it is converted into a probability distribution to eliminate the influence of pulse number and data size variations on the threshold calculation results, resulting in:

[0199]

[0200] Based on this, candidate thresholds are introduced. The detected score samples are divided into two categories: low score and high score, which correspond to the potential set of normal pulses and the set of pulses containing interference, respectively.

[0201] Further calculate the cumulative probability and cumulative mean corresponding to the threshold:

[0202]

[0203] Represents the cumulative probability. This represents the cumulative mean.

[0204] Meanwhile, the global mean of all samples is defined as:

[0205]

[0206] Based on the above statistics, an inter-class variance function is constructed to measure the threshold. Separation capability for two types of samples:

[0207]

[0208] By searching the entire grayscale range for the threshold that maximizes the inter-class variance. This allows us to obtain the optimal adaptive segmentation position.

[0209] The optimal grayscale threshold is mapped back to the normalized detection score domain to obtain the final decision threshold:

[0210]

[0211] in, The optimal grayscale threshold that maximizes the inter-class variance is obtained through the OTSU method; finally, the pulse-by-pulse interference detection score is... Greater than the adaptive threshold When the pulse is not in the specified state, it is determined to be a pulse containing interference; otherwise, it is determined to be a normal pulse, thus completing the automatic interference judgment at the pulse-by-pulse level.

[0212] In step S107, output the pulse interference judgment result;

[0213] Specifically: Based on the obtained pulse-by-pulse adaptive decision results, the final pulse-type interference judgment sequence is generated, which can be directly used for subsequent interference suppression.

[0214] This step completes the closed-loop processing from pulse-by-pulse multi-domain feature extraction → information entropy weighted fusion → adaptive threshold decision → interference pulse determination, realizing the automatic judgment and labeling of pulse-type interference pulses.

[0215] This invention is of great significance for improving the detection accuracy and stability of SAR systems under complex interference environments. The method comprehensively characterizes the differences in interference pulses across multiple dimensions, including temporal amplitude anomalies, spike abrupt changes, frequency domain energy concentration, and non-stationary spectral variations. It adaptively determines the weights of each feature using information entropy theory and achieves pulse-level interference detection based on an adaptive threshold. This enables high-precision detection even under weak pulse interference conditions, while ensuring the integrity and reliability of useful signals, providing a reliable data foundation for subsequent interference suppression.

[0216] like Figure 2 As shown, the experimental comparison results of this application are presented. The upper left corner shows the experimental data containing pulse interference under different signal-to-interference ratios. The upper right and lower left corners show the processing results of existing methods, and the lower right corner shows the results processed by this invention. The comparison of the above results shows that, under the same scenario data conditions, the judgment effect of this invention on data containing pulse interference is significantly better than that of existing methods.

[0217] Exemplary System

[0218] Figure 3 This application also provides a pulsed radio frequency interference judgment system based on multi-domain feature entropy weight fusion, comprising: a feature extraction unit configured to extract time-domain statistical center deviation and sharpness features pulse by pulse from the interference echo data of the input system, and calculate frequency-domain kurtosis and frequency-domain skewness first-order difference features pulse by pulse, thereby constructing a multi-domain feature vector for each pulse; a feature weight calculation unit configured to calculate the difference coefficient of each feature based on information entropy theory for the multi-domain feature vector sequence, and normalize it to generate corresponding adaptive weights; an interference judgment unit configured to perform weighted fusion of the multi-domain features according to their adaptive weights to obtain a pulse-by-pulse interference detection score sequence, normalize it and map it to grayscale space to construct a histogram, perform Gaussian smoothing, and use OTSU adaptive threshold to judge the pulse, so as to realize automatic judgment of pulses containing pulsed interference; and a result output unit configured to generate a pulsed interference judgment sequence based on the pulse-by-pulse judgment result for subsequent interference suppression processing.

[0219] The pulsed radio frequency interference judgment system with multi-domain feature entropy weight fusion provided in this application embodiment can realize any of the above-mentioned pulsed interference judgment steps and processes, and achieve the same technical effect, which 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; when the memory executes the programs on the processor, it implements the pulse radio frequency interference judgment method based on multi-domain feature entropy weight fusion.

[0222] Since the pulse-type radio frequency interference judgment steps based on multi-domain feature entropy weight fusion have been described in detail in the specific implementation method 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 (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. 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 to: extract time-domain statistical center deviation features, time-domain sharpness features, and frequency-domain kurtosis and first-order difference of frequency-domain skewness from pulse-by-pulse synthetic aperture radar echo data from the input system, and construct a pulse-by-pulse multi-domain feature vector sequence; perform adaptive weight calculation on the multi-domain features based on information entropy theory, and perform weighted fusion of each feature according to the weight to obtain a pulse-by-pulse comprehensive interference discrimination score sequence; normalize the interference discrimination score sequence and construct a statistical histogram distribution, and combine Gaussian smoothing to enhance the stability of the score distribution; on this basis, use the OTSU adaptive threshold decision method to segment the pulse-by-pulse interference discrimination score to realize the automatic determination of pulses containing pulse interference and normal pulses; finally, output the pulse interference judgment result for subsequent interference suppression or imaging processing.

[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 pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion, characterized in that, include: Step S101: Perform pulse-by-pulse processing on the echo data to extract the time-domain statistical center deviation and sharpness features; Step S102: Extract the first-order difference features of frequency domain kurtosis and skewness pulse by pulse to construct a multi-domain feature vector sequence; Step S103: Adaptive weight determination of pulse-by-pulse multi-domain feature vector sequence based on information entropy theory; Step S104: Perform weighted fusion on the pulse-by-pulse multi-domain feature vector sequence to obtain the interference detection score sequence; Step S105: Construct a statistical histogram distribution based on the detection scores and perform Gaussian smoothing. Step S106: Use OTSU adaptive threshold decision to perform segmentation, thereby achieving automatic determination of pulses containing pulse interference. Step S107: Output the pulse interference judgment result.

2. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S101, the echo data is processed pulse by pulse to extract the time-domain statistical center deviation and sharpness features, specifically: The formula for calculating the deviation of the statistical center in the time domain is expressed as follows: ； in, Indicates the first A sample of amplitude values, Indicates the sample median. Indicates the absolute median difference. To prevent the introduction of tiny positive numbers by division by zero; Then, for the first The statistical center deviation of the amplitude sequence of each pulse is calculated, and the maximum value is taken as the characteristic value of the pulse, which can be expressed as: ； in , Indicates the first The first pulse A distance-oriented sample, This indicates the total number of azimuth pulses. It is the absolute deviation of the median; By performing the above processing pulse by pulse, the time-domain statistical center deviation feature sequence can be obtained. : ； It is the first The statistical center deviation characteristic value corresponding to each pulse; The formula for calculating the time-domain sharpness feature is expressed as follows: ； in, and These represent the minimum and maximum values ​​within the sliding window to the left of the sample point, respectively. and These represent the minimum and maximum values ​​within the sliding window to the right of the sample point, respectively. The squaring operation is used to enhance the response to sharp peaks and abrupt changes. Then, for the first The sharpness sequence corresponding to each pulse amplitude sequence is calculated and normalized. To highlight the extreme peak characteristics, the high quantile statistic is selected from the normalized sharpness sequence as the time-domain sharpness feature of the pulse. The result can be expressed as follows: ； in, , Indicates the first The sharpness sequence corresponding to each pulse. The high quantile percentage is close to 1; The value range is [0.9, 0.999]; by performing the above processing pulse by pulse, the temporal sharpness feature sequence of all pulses can be obtained. Its form can be expressed as: ； By calculating the above features pulse by pulse, the complete time-domain statistical center deviation feature sequence SCD and the time-domain sharpness feature sequence Sharp can be obtained, providing input basis for subsequent multi-domain feature entropy weight fusion.

3. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S102, the first-order difference features of frequency domain kurtosis and skewness are extracted pulse by pulse to construct a multi-domain feature vector sequence, specifically as follows: The formula for calculating kurtosis is expressed as: ； in, Represents the first value in the spectral amplitude sequence Sampling points , ; This represents the total number of frequency domain sampling points. The mean of the spectral amplitude sequence is given; the frequency domain kurtosis coefficient is the ratio of the fourth central moment to the square of the second central moment. Then, for the first Perform a frequency domain transformation on each pulse to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain kurtosis features are calculated. The result can be expressed as: ； in, , Indicates the total number of azimuth pulses; By performing the above processing pulse by pulse, the frequency domain kurtosis feature sequence of all pulses can be obtained, which can be expressed as: ； The formula for calculating the frequency domain skewness characteristic can be expressed as: ； in, Represents the first value in the spectral amplitude sequence One sampling point, , representing the total number of frequency domain sampling points, The mean of the spectral amplitude sequence; the frequency domain skewness is obtained by the third and second central moments. The power ratio characterizes the asymmetry of the spectral distribution, and this structure makes it highly sensitive to frequency domain energy shifts and asymmetric changes in spectral profile. Then, for the first The frequency domain transform of the time-domain echo of each pulse is performed to obtain its spectrum. And obtain the spectrum amplitude sequence by taking the amplitude value of the spectrum. Based on this, the corresponding frequency domain skewness is calculated, and the result can be expressed as: ； in, , Indicates the total number of azimuth pulses; To further characterize the degree of variation of frequency domain statistical properties between adjacent pulses, a first-order difference feature of frequency domain skewness is introduced, which is defined as the absolute value of the difference in frequency domain skewness between adjacent pulses: ； in, , when At that time, the first-order difference of the frequency domain skewness is defined as zero; By performing the above processing pulse by pulse, the first-order difference feature sequence of frequency domain skewness corresponding to all pulses is obtained. Its form can be expressed as: ； After obtaining the time-domain statistical center deviation feature, time-domain sharpness feature, frequency-domain kurtosis feature, and frequency-domain skewness first-order difference feature of each pulse, in order to achieve a multi-angle joint characterization of pulse interference, the features of different domains and different behavioral mechanisms mentioned above are modeled in a unified manner to construct a pulse-by-pulse multi-domain feature vector sequence. Specifically, for the first For each pulse, the corresponding time-domain statistical center deviation characteristic is... Temporal sharpness characteristics Frequency domain kurtosis characteristics and frequency domain skewness first-order difference characteristics By performing joint characterization, a multi-domain feature vector corresponding to the pulse is formed, which can be expressed as: ； in, , Indicates the total number of azimuth pulses; By performing the above feature construction process on all pulses, a complete pulse-by-pulse multi-domain feature vector sequence can be obtained, the overall form of which can be expressed as: 。 4. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S103, adaptive weights are determined for the pulse-by-pulse multi-domain features based on information entropy theory, specifically as follows: The pulse-by-pulse multi-domain feature vector sequences obtained in steps S101 and S102 are jointly represented by pulse indices to form a pulse-by-pulse multi-domain feature matrix: ； in, Represents the feature dimension. Indicates the total number of pulses; To eliminate differences in dimensions, amplitude range, and statistical distribution among different features, and to avoid the adverse effects of feature numerical scale on weight allocation results, the feature matrix is ​​normalized according to its feature dimension, as expressed by: ； in, , To prevent the introduction of tiny positive numbers with a denominator of zero, It is the result of normalization; Based on the normalized features, the values ​​of each feature on the pulse sequence are treated as discrete random variables, and the probability distributions corresponding to each feature are constructed: ； Furthermore, based on information entropy theory, the uncertainty and dispersion of each feature in the impulse dimension are quantitatively evaluated, and the 1st... Information entropy corresponding to each feature: ； It is a logarithm with base 10; Based on the information entropy results, a difference coefficient is introduced to characterize the degree of contribution of each feature to the effective information of interference discrimination: ； The difference coefficients are then normalized to obtain the adaptive weights corresponding to each feature: 。 5. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S104, the pulse-by-pulse multi-domain features are weighted and fused to obtain the interference detection score sequence, specifically as follows: The multi-domain features are linearly weighted and fused according to their adaptive weights to obtain the pulse-by-pulse comprehensive interference discrimination score: ； in, These are the adaptive weights corresponding to each feature; The comprehensive discrimination score Used to characterize the The significance of interference from each pulse is determined and used as the input for subsequent Gaussian smoothing histograms.

6. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S105, a statistical histogram distribution is constructed based on the detection scores, and Gaussian smoothing is performed, specifically as follows: First, the pulse-by-pulse interference detection score sequence is analyzed. After linear normalization, the expression is as follows: ； in, A tiny positive number introduced to prevent the denominator from being zero; To facilitate subsequent threshold analysis based on statistical distribution, the normalized detection score is mapped to a finite discrete grayscale space, specifically, it is linearly mapped to an 8-bit grayscale range. Its expression is: ； Based on this, a statistical histogram is constructed for the mapped detection score sequence, and its histogram counting function can be expressed as: ； in, This indicates that the test score falls into the first category. The number of pulses within each grayscale range ; Gaussian smoothing of a histogram counting sequence yields the following result: ； in, The standard deviation of the Gaussian kernel is used to control the smoothing intensity. This represents the mean.

7. The pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion according to claim 1, characterized in that, In step S106, OTSU adaptive threshold decision is used for segmentation to achieve automatic determination of pulse-type interference pulses, specifically: The Gaussian-smoothed grayscale histogram is considered as the statistical distribution of pulse-by-pulse detection scores in the grayscale space. By normalizing this histogram, it is converted into a probability distribution to eliminate the influence of pulse number and data size variations on the threshold calculation results, resulting in: ； Based on this, candidate thresholds are introduced. The detected score samples are divided into two categories: low score and high score, which correspond to the potential set of normal pulses and the set of pulses containing interference, respectively. Further calculate the cumulative probability and cumulative mean corresponding to the threshold: ； Represents the cumulative probability. This represents the cumulative mean; Meanwhile, the global mean of all samples is defined as: ； Based on the above statistics, an inter-class variance function is constructed to measure the threshold. Separation capability for two types of samples: ； By searching the entire grayscale range for the threshold that maximizes the inter-class variance. To obtain the optimal adaptive segmentation position; The optimal grayscale threshold is mapped back to the normalized detection score domain to obtain the final decision threshold: ； in, The optimal grayscale threshold that maximizes the inter-class variance is obtained through the OTSU method; finally, the pulse-by-pulse interference detection score is... Greater than the adaptive threshold When the pulse is not in the specified state, it is determined to be a pulse containing interference; otherwise, it is determined to be a normal pulse, thus completing the automatic interference judgment at the pulse-by-pulse level.

8. A pulse-type radio frequency interference judgment system based on multi-domain feature entropy weight fusion, characterized in that, include: The feature extraction unit is configured to extract time-domain statistical center deviation and sharpness features pulse by pulse from the interference echo data of the input system, and calculate frequency-domain kurtosis and frequency-domain skewness first-order difference features pulse by pulse, thereby constructing a multi-domain feature vector for each pulse. The feature weight calculation unit is configured to calculate the difference coefficient of each feature based on the information entropy theory for the multi-domain feature vector sequence, and normalize it to generate the corresponding adaptive weight. The interference judgment unit is configured to perform weighted fusion of the multi-domain features according to their adaptive weights to obtain a pulse-by-pulse interference detection score sequence, normalize and map it to a gray space to construct a histogram, and after Gaussian smoothing, use OTSU adaptive threshold to judge the pulses so as to realize the automatic judgment of pulses containing pulse-type interference. The result output unit is configured to generate a pulsed interference judgment sequence based on the pulse-by-pulse judgment result for use in subsequent interference suppression processing.

9. A storage device storing a plurality of programs, characterized in that, The program application is loaded and executed by the processor to implement the pulsed radio frequency interference judgment method of multi-domain feature entropy weight fusion 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 pulse-type radio frequency interference judgment method based on multi-domain feature entropy weight fusion as described in any one of claims 1-7.

Images