A method to resist chaff interference based on multi-domain features and LightGBM

Abstract

The present invention belongs to the technical field of electronic countermeasures, data classification and signal recognition, and in particular, relates to a method for resisting chaff interference based on multi-domain features and LightGBM. It includes: preprocessing the target echo data under the chaff interference of the horizontal polarization receiving channel of the broadband polarization radar to obtain the range image sequence of the horizontal polarization receiving channel; performing chaff interference type judgment and target segmentation to obtain the target range image sequence and the chaff interference range image sequence; extracting and obtaining feature vectors from the time domain, frequency domain and polarization domain, and taking the feature vector obtained from each sequence as a sample; labeling the samples; dividing the labeled samples into data to obtain training sets and test sets; and performing training and testing to realize the recognition and countermeasure of chaff interference. The method has good countermeasure effect against shielded chaff interference and high classification accuracy, good robustness to sea clutter and high anti-interference efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of electronic countermeasures, data classification and signal recognition, and in particular relates to an anti-chaff interference method based on multi-domain features and LightGBM. Background Art

[0002] As one of the main methods of terminal guidance, radar active homing guidance has been widely used due to its advantages of long range, high detection accuracy, and all-day and all-weather operation. However, since radar active guidance requires the emission of electromagnetic waves to complete the detection, identification, positioning and tracking of targets, the radar seeker also has the disadvantage of being easily affected by the battlefield electromagnetic environment.

[0003] Chaff interference is one of the most typical passive interferences against active radar seekers. This interference scatters radar electromagnetic waves by projecting a large number of chaff fibers, generating strong echoes with clutter characteristics in the radar receiver, achieving the purpose of electromagnetic deception or suppression of the radar, thereby effectively protecting its own platform or other targets of its own side, and playing an extremely important role in electronic countermeasures. With the optimization of chaff materials, the advancement of manufacturing processes and the improvement of delivery equipment, chaff interference has the advantages of simple manufacturing, low cost, easy use, easy to obtain wide-band characteristics and the ability to simultaneously interfere with multiple radars of different systems, different frequencies and different directions. It is one of the important threats faced by radar seekers. Therefore, how to improve the anti-chaff interference capability of radar seekers is still a hot and difficult problem in research at home and abroad.

[0004] According to the purpose of chaff projectile projection, chaff interference is mainly divided into shielded chaff interference and diluted or centroid chaff interference. At present, chaff anti-interference methods mainly include: signal feature-based recognition method, filter-based chaff interference countermeasure method and composite guidance technology, but most of the chaff anti-interference methods are proposed for diluted or centroid chaff interference (diluted chaff interference, centroid chaff interference, etc.), and the proposed features are only a single transform domain.

[0005] The existing method of dilute chaff jamming countermeasure based on polarization domain characteristics first estimates the polarization scattering matrix of the target and the chaff false target, then defines and calculates the polarization scattering parameters of the target and the chaff false target respectively; then calculates the absolute value of the correlation between the co-polarization and cross-polarization channels; finally, the support vector machine is used to realize the classification and recognition of chaff false targets and radar targets such as targets and aircraft. This algorithm has a certain anti-interference effect, but the disadvantage is that it is difficult to obtain polarization scattering parameters and cannot counter shielded chaff jamming.

[0006] There is also a method for identifying false targets from foil strips based on polarization domain features. This method first defines and calculates the mean polarization ratio of the horizontal channel and the polarization ratio of the vertical channel of the real target and the false target; then takes the mean of the smaller polarization ratios of multiple pulses as the characteristic parameter; finally, compares the characteristic parameters of the real and false targets, and the target with the larger value is judged as the target. Although the characteristic calculation complexity of this method is low, the method has poor environmental adaptability. When the sea clutter is large and the real and false target echoes are aliased, the stability of the characteristic parameter is poor.

[0007] There is also a method for identifying false targets from foil strips based on polarization domain features. First, the polarization ratio mean of the horizontal channel and the polarization ratio of the vertical channel of the real target and the false target from foil strips are defined and calculated respectively; then the mean of the smaller polarization ratios of multiple pulses is taken as the characteristic parameter; finally, the characteristic parameters of the real and false targets are compared, and the threshold is used for classification, and the target greater than the threshold is judged as the real target. Although the feature calculation complexity of this method is low, the method has poor environmental adaptability. When the sea clutter is large and the real and false target echoes are aliased, the stability of the characteristic parameter is poor.

[0008] In addition, there is also a method based on the range image to extract waveform entropy, correlation coefficient and scattering intensity ratio features, and identify chaff interference through FCM algorithm and cluster analysis. Although this method has high accuracy, it does not consider the overlap of the range image of the chaff false target and the real target, and the countermeasure effect against the shielded chaff interference is poor.

[0009] Therefore, based on the above-mentioned prior art, it can be known that the echo data of shielded foil interference is complex, classification based on the characteristics of a single transform domain is difficult, and the recognition rate is low. The present invention is committed to overcoming these defects and realizing efficient recognition of foil interference based on multi-domain features. Summary of the invention

[0010] The purpose of the present invention is to address the defects of complex echo data of targets under shielded foil interference, difficult classification according to features of a single transform domain, and low recognition rate. A method for resisting foil interference based on multi-domain features and LightGBM is proposed. This method not only has high accuracy, but also can resist various foil interferences including diluted or centroid foil interference and shielded foil interference.

[0011] In order to achieve the above object, the present invention adopts the following technical scheme:

[0012] The anti-foil interference method comprises the following steps:

[0013] S1, the preprocessing module preprocesses the target echo data under the foil interference of the horizontal polarization receiving channel of the broadband polarization radar to obtain the range image sequence of the horizontal polarization receiving channel;

[0014] The preprocessing includes pulse compression, target range image alignment and normalization operations;

[0015] S2, the target segmentation module performs chaff interference type judgment and target segmentation on the range image sequence of the horizontal polarization receiving channel obtained by S1 to obtain a target range image sequence and a chaff interference range image sequence;

[0016] S2 specifically includes the following sub-steps:

[0017] S21, calculating the average power of the range image sequence of the horizontal polarization receiving channel, and setting a threshold to determine the type of chaff interference, if the power is less than the threshold, it is diluted or centroid chaff interference, otherwise it is shielded chaff interference;

[0018] S22, when the chaff interference type is diluted or centroid chaff interference, the range image sequence of the horizontal polarization receiving channel is detected by CA-CFAR, and the detected signal is segmented, wherein the target segmentation is specifically to take n range resolution units with the midpoint of the range image of each detected target as the center as the range image sequence of the target, and obtain a target range image sequence and a chaff interference range image sequence; when the chaff interference type is shielded chaff interference, if the target distance is unknown, divide one target by every n range resolution units of the range image, otherwise take n range resolution units with the range resolution unit of the known target as the center as the target range image sequence, and take n range resolution units in the remaining area without target information as the chaff interference range image sequence, and obtain a target range image sequence and a chaff interference range image sequence;

[0019] S3, the feature extraction module extracts the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity of the target range image sequence and the chaff interference range image sequence obtained in S2 from the time domain, extracts the spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain, extracts the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain, combines the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity, obtains the feature vectors corresponding to all target range image sequences and chaff interference range image sequences, and takes the feature vector obtained from each sequence as a sample to obtain all samples;

[0020] It includes the following sub-steps:

[0021] S31, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain, specifically:

[0022] S311, extracting a pulse range image of the target and the chaff interference from the target range image sequence and the chaff interference range image sequence obtained in S2;

[0023] S312, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain according to the range image of a pulse of the target and the chaff interference obtained in S311;

[0024] The extracting of the range image concentration from the time domain is specifically as follows: sorting the amplitudes of all the range resolution units of the range image of a pulse of the target and the chaff interference, and calculating the ratio of the upper quartile to the peak value based on the range resolution unit amplitude sequence sorted from large to small, to obtain the range image concentration;

[0025] The energy proportion is extracted from the time domain, specifically, the amplitudes of all distance resolution units of the range image of a pulse of the target and the chaff interference are sorted, and the ratio of the sum of the amplitudes of the first m distance resolution units to the sum of the amplitudes of all n distance resolution units is calculated based on the range resolution unit amplitude sequence sorted from large to small, to obtain the energy proportion, and m is less than n;

[0026] The energy mean is extracted from the time domain, specifically: according to the radar cross section of the chaff cloud being generally larger than the radar cross section of the target, the mean of the amplitudes of the range images of the two is calculated to obtain the energy mean of the two;

[0027] The energy variance is extracted from the time domain, specifically: according to the difference in the range image distribution of the target and the chaff cloud, the variance of the amplitude of the range image of the two is calculated to obtain the energy variance of the two;

[0028] The waveform width is extracted from the time domain, specifically, according to the difference in length between the target and the chaff cloud in the radar radial direction, the product of the number of range resolution units occupied by the range image of a pulse of the target and the chaff interference when the main peak amplitude is attenuated by 90% and the range resolution is calculated to obtain the waveform width;

[0029] The extracting of the average similarity from the time domain is specifically as follows: according to the geometric distribution of the target and the chaff cloud and the difference in volume transformation during movement, the average similarity of two adjacent range images is calculated;

[0030] S32, extracting spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain;

[0031] S321, performing fast Fourier transform on the range image of a pulse of the target and the chaff interference obtained in S311 to obtain the frequency spectrum of the target and the chaff interference;

[0032] S322, extracting spectrum broadening, frequency domain skewness, and frequency domain kurtosis from the frequency domain according to the frequency spectrum of the target and the chaff interference obtained in S321;

[0033] The extracting spectrum broadening from the frequency domain specifically comprises: calculating the widths of the spectrum of the target and the chaff cloud according to the difference in motion characteristics of the target and the chaff cloud to obtain the widths of the spectrum of the target and the chaff cloud;

[0034] The extracting of frequency domain skewness from the frequency domain is specifically as follows: according to the difference in the frequency spectrum shapes of the target and the chaff cloud, the skewness coefficient of the frequency spectrum of the two is calculated to obtain the frequency domain skewness;

[0035] The extracting of frequency domain kurtosis from the frequency domain is specifically as follows: according to the difference in the frequency spectrum shapes of the target and the foil cloud, the kurtosis coefficient of the frequency spectrum of the two is calculated to obtain the frequency domain kurtosis;

[0036] S323, calculating the frequency spectrum of the target and chaff interference obtained in S321 to obtain the power spectrum of the target and chaff interference;

[0037] S324, extracting power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain according to the power spectrum of the target and the chaff interference;

[0038] The extracting of the power spectrum envelope fluctuation from the frequency domain is specifically as follows: according to the difference in the degree of change of the power spectrum envelope amplitude of the target and the chaff cloud, the envelope fluctuation of the power spectrum of the two is calculated to obtain the power spectrum envelope fluctuation;

[0039] The power spectrum Shannon entropy is extracted from the frequency domain, specifically: the probability value in the entropy function is replaced by the normalized power value of each frequency resolution unit of the power spectrum of the target and the chaff cloud, and the power spectrum Shannon entropy of the two is calculated respectively;

[0040] S33, extracting the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain;

[0041] S331, extracting the co-polarization range image and cross-polarization range image of the target and the chaff interference from the target range image sequence and the chaff interference range image sequence obtained in S2, and extracting the polarization similarity from the polarization domain;

[0042] The extracting polarization similarity from the polarization domain specifically comprises: calculating the similarity of the co-polarization and cross-polarization range images of the target and the foil according to the co-polarization range image and the cross-polarization range image of the target and the foil interference, so as to obtain the polarization similarity;

[0043] S332, obtaining a polarization ratio range image of the target and the chaff interference according to the co-polarization range image and the cross-polarization range image of the target and the chaff interference obtained in S331;

[0044] S333, extracting the polarization ratio mean and the polarization ratio variance sum from the polarization domain according to the polarization ratio range images of the target and the chaff interference obtained in S331;

[0045] The extracting the polarization ratio mean from the polarization domain is specifically: calculating the mean of the polarization ratio range images of the target and the foil interference to obtain the polarization ratio mean;

[0046] The extracting of the polarization ratio variance from the polarization domain is specifically: calculating the variance of the polarization ratio range images of the target and the foil interference to obtain the polarization ratio variance;

[0047] S34. Merge the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity to obtain the feature vector.

[0048] S4, the labeling module labels all the samples obtained in S3 to obtain labeled samples;

[0049] The labeling is specifically: setting the feature vector label of the target range image sequence to 1 and the feature vector label of the foil strip interference range image sequence to 0, so as to obtain a labeled sample;

[0050] S5, the data partitioning module partitions the labeled samples obtained in S4 according to a certain ratio to obtain a training set and a test set;

[0051] S6, the LightGBM training module uses the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters, and then inputs the training set into the LightGBM classifier with the optimal hyperparameters for training to obtain the LightGBM classifier with the lowest classification cross entropy;

[0052] S6 specifically includes the following sub-steps:

[0053] S61, use the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters;

[0054] S62, using the training set obtained in S5, the leaf-based growth algorithm with depth restriction, and the histogram-based sparse feature optimization algorithm to train the LightGBM classifier with the optimal hyperparameters, to obtain the LightGBM classifier with the lowest classification cross entropy;

[0055] S7, the LightGBM classification module inputs the test set obtained in S5 into the LightGBM classifier with the lowest classification cross entropy obtained in S6 for classification, so as to realize the recognition and countermeasure of chaff interference;

[0056] So far, from S1 to S7, a method to resist foil interference based on multi-domain features and LightGBM has been completed.

[0057] Beneficial Effects

[0058] Compared with the existing classification methods, the anti-foil interference method based on multi-domain features and LightGBM described in the present invention has the following beneficial effects:

[0059] 1. The anti-chaff interference system benefits from the target segmentation module, feature extraction module and LightGBM training module, which can not only counteract the dilution or centroid type chaff interference, but also counteract the shielding type chaff interference, and has high classification accuracy and stable countermeasure effect;

[0060] 2. The anti-chaff interference method benefits from the fact that the feature extraction module in S3 uses features in multiple transform domains, including time domain, frequency domain and polarization domain, to avoid the defect of poor applicability caused by the influence of the environment on the features of a single transform domain. It has good robustness to sea clutter and strong environmental adaptability.

[0061] 3. The anti-foil interference method benefits from the LightGBM training module adopting a lightweight and efficient LightGBM classifier, which avoids the defects of the anti-foil interference method based on threshold classification, that the threshold is greatly affected by the environment and difficult to determine. The binary tree structure and feature segmentation points of the LightGBM classifier can be adaptively learned. The training time of the LightGBM classifier is much shorter than the training time of the support vector machine used in other inventive methods, and the anti-interference efficiency is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is the composition and connection relationship of the anti-chaff interference system based on multi-domain features and LightGBM of the present invention;

[0063] Figure 2 It is a fast-time-slow-time image of the full-range image sequence after preprocessing in the anti-chaff interference method based on multi-domain features and LightGBM of the present invention;

[0064] Figure 3 It is a flow chart of feature extraction in an anti-chaff interference method based on multi-domain features and LightGBM of the present invention;

[0065] Figure 4 It is a system block diagram of signal processing when a specific implementation of an anti-chaff interference system based on multi-domain features and LightGBM of the present invention is implemented;

[0066] Figure 5It is the result of the anti-chaff interference method based on multi-domain features and LightGBM in the present invention against the diluted or centroidal chaff interference in different sea conditions;

[0067] Figure 6 It is the countermeasure result of shielded foil strip interference with different interference-to-signal ratios when the anti-foil strip interference method based on multi-domain features and LightGBM of the present invention is specifically implemented. DETAILED DESCRIPTION

[0068] The anti-foil interference method based on multi-domain features and LightGBM described in the present invention is further illustrated and described in detail below in conjunction with the accompanying drawings and embodiments.

[0069] Example 1

[0070] This embodiment describes the process of anti-interference of target echo data received by broadband polarization radar under the background of foil strip interference using the method described in the present invention. The broadband polarization radar transmits electromagnetic waves with a period of 90 pulses of horizontal polarization waves and then 70 vertical polarization waves. The data set is the target echo data received by the radar horizontal polarization channel under three backgrounds. The first background is a background of dilution or centroid foil strip interference without sea clutter. The 7 groups of data under this background have a pulse number of 3200 in each group, including 1600 pulses in the diffusion stage and 1600 pulses in the stable stage. The second background is a background of dilution or centroid foil strip interference with different sea conditions. The 6 groups of signal-to-clutter ratios (SCR) under this background are 14.03dB, 9.58dB, 6.84dB, 4.97dB, 3.57dB, and 2.49dB, respectively. The pulse number of each group is 12800, including 1600 pulses in the growth period and 11200 pulses in the stable period. The third background is the shielded chaff interference background with different interference-to-signal ratios under low sea conditions. The interference-to-signal ratios of the 10 background groups are -20dB, -13.98dB, -6.02dB, -3.10dB, -0.92dB, 0dB, 6.02dB, 9.54dB, 12.04dB, 13.98dB, 14.03dB, 12800 pulses, including 1600 pulses in the growth period and 11200 pulses in the stable period. The dimension of a pulse data is 1x2048, and a pulse data contains both targets and chaff interference. The target category is 2.

[0071] Figure 1 The composition and connection relationship of the anti-chaff interference system based on multi-domain features and LightGBM; Figure 1It can be seen that the classification system includes a preprocessing module, a target segmentation module, a feature extraction module, a label annotation module, a data partitioning module, a LightGBM training module and a LightGBM classification module; and the preprocessing module is connected to the target segmentation module, the target segmentation module is connected to the feature extraction module, the feature extraction module is connected to the label annotation module, the label annotation module is connected to the data partitioning module, the data partitioning module is connected to the LightGBM training module, and the LightGBM training module is connected to the LightGBM classification module.

[0072] S1, the preprocessing module preprocesses the target echo data under the foil interference of the horizontal polarization receiving channel of the broadband polarization radar to obtain the range image sequence of the horizontal polarization receiving channel;

[0073] The preprocessing includes pulse compression, target range image alignment and normalization operations;

[0074] Figure 2 It is a fast-time-slow-time diagram of the range image sequence of the horizontal polarization receiving channel after preprocessing in the anti-chaff interference method based on multi-domain features and LightGBM. Figure 2 It can be seen that the range image sequence contains information in three dimensions: time domain, frequency domain and polarization domain. The data can clearly reflect the differences between chaff interference and targets in time domain, frequency domain and polarization domain.

[0075] S2, the target segmentation module performs chaff interference type judgment and target segmentation on the range image sequence of the horizontal polarization receiving channel obtained by S1 to obtain a target range image sequence and a chaff interference range image sequence;

[0076] S2 specifically includes the following sub-steps:

[0077] S21, calculating the average power of the range image sequence of the horizontal polarization receiving channel, and setting a threshold to determine the type of chaff interference, if the power is less than the threshold, it is diluted or centroid chaff interference, otherwise it is shielded chaff interference;

[0078] S22, when the chaff interference type is diluted or centroid chaff interference, the range image sequence of the horizontal polarization receiving channel is detected by CA-CFAR, and the detected signal is segmented, the target segmentation is specifically to take n (n=320) range resolution units as the range image sequence of the target with the midpoint of the range image of each detected target as the center, to obtain a target range image sequence and a chaff interference range image sequence; when the chaff interference type is shielded chaff interference, if the target distance is unknown, every n range resolution units of the range image are spaced to divide a target, otherwise, the range image sequence is taken with the range resolution unit of the known target as the center, and n range resolution units are taken as the target range image sequence in the remaining area without target information as the chaff interference range image sequence, to obtain a target range image sequence and a chaff interference range image sequence;

[0079] S3, the feature extraction module extracts the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity of the target range image sequence and the chaff interference range image sequence obtained in S2 from the time domain, extracts the spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain, extracts the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain, combines the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity, obtains the feature vectors corresponding to all target range image sequences and chaff interference range image sequences, and takes the feature vector obtained from each sequence as a sample to obtain all samples;

[0080] Figure 3 It is a flow chart of feature extraction in the anti-chaff interference method based on multi-domain features and LightGBM;

[0081] It includes the following sub-steps:

[0082] S31, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain, specifically:

[0083] S311, extracting a pulse range image s of the target and chaff interference from the target range image sequence and chaff interference range image sequence obtained in S2 = [s(1), s(2), ..., s(j), ..., s(n)];

[0084] S312, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain according to the range image of a pulse of the target and the chaff interference obtained in S311;

[0085] The extraction of the range image concentration from the time domain is specifically as follows: the amplitudes of all the range resolution units of the range image of a pulse of the target and the chaff interference are sorted, and based on the range resolution unit amplitude sequence o=[o(1), o(2), ..., o(j), ..., o(n)] sorted from large to small, the ratio of the upper quartile Q3 to the peak value o(1) is calculated to obtain the range image concentration

[0086] The energy proportion is extracted from the time domain, specifically, the amplitudes of all distance resolution units of the range image of a pulse of the target and the chaff interference are sorted, and the ratio of the sum of the amplitudes of the first m distance resolution units to the sum of the amplitudes of all n distance resolution units is calculated based on the range resolution unit amplitude sequence sorted from large to small, to obtain the energy proportion Where n is the number of distance resolution units, and m is less than n;

[0087] The energy mean is extracted from the time domain, specifically: according to the radar cross section of the chaff cloud is generally larger than the radar cross section of the target, the mean amplitude of the range image of a pulse of the target and the chaff interference is calculated to obtain the energy mean

[0088] The energy variance is extracted from the time domain, specifically: according to the difference in the range image distribution of the target and the chaff cloud, the variance of the amplitude of the range image of a pulse of the target and the chaff interference is calculated to obtain the energy variance

[0089] The waveform width is extracted from the time domain, specifically, according to the difference in length between the target and the chaff cloud in the radar radial direction, the product of the number of distance resolution units occupied when the main peak amplitude of a pulse of the range image of the target and the chaff interference is attenuated by 90% and the range resolution is calculated to obtain the waveform width W=|k2-k1|×Δr, wherein k1 and k2 are the subscripts at the main peak attenuation of 90%, and Δr is the range resolution;

[0090] The method of extracting the average similarity from the time domain is as follows: according to the geometric distribution of the target and the chaff cloud and the difference in volume transformation during motion, the average similarity of two adjacent range images s(j) and s′(j) is calculated.

[0091] S32, extracting spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain;

[0092] S321, performing fast Fourier transform on the range image of a pulse of the target and the chaff interference obtained in S311 to obtain a frequency spectrum W(f) of the target and the chaff interference;

[0093] S322, extracting spectrum broadening, frequency domain skewness, and frequency domain kurtosis from the frequency domain according to the frequency spectrum W(f) of the target and chaff interference obtained in S321;

[0094] The spectrum broadening is extracted from the frequency domain, specifically: according to the difference in the motion characteristics of the target and the chaff cloud, the width of the spectrum of the two is calculated to obtain the spectrum width W of the two f =|f2-f1|×Δf, where f1 and f2 are the subscripts of the main peak of the spectrum attenuated by 90%, and Δf is the frequency resolution;

[0095] The method of extracting the frequency domain skewness from the frequency domain is as follows: according to the difference in the spectrum shapes of the target and the foil cloud, the skewness coefficient of the spectrum of the two is calculated to obtain the frequency domain skewness Where E(W(f)) is the mean of the spectrum, and D(W(f)) is the standard deviation of the spectrum;

[0096] The method of extracting the frequency domain kurtosis from the frequency domain is specifically as follows: according to the difference in the spectrum shapes of the target and the foil cloud, the kurtosis coefficient of the spectrum of the two is calculated to obtain the frequency domain kurtosis

[0097] S323, calculating the frequency spectrum of the target and chaff interference obtained in S321 to obtain the power spectrum of the target and chaff interference Where N is the number of points in the spectrum;

[0098] S324, extracting power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain according to the power spectrum of the target and the chaff interference;

[0099] The power spectrum envelope fluctuation is extracted from the frequency domain, specifically: according to the difference in the degree of change of the power spectrum envelope amplitude of the target and the foil cloud, the envelope fluctuation of the power spectrum of the two is calculated to obtain the power spectrum envelope fluctuation Where E(S(f)) is the mean of the power spectrum, and D(S(f)) is the standard deviation of the power spectrum;

[0100] The power spectrum Shannon entropy is extracted from the frequency domain by replacing the probability value in the entropy function with the normalized power value of each frequency resolution unit of the power spectrum of the target and the foil cloud, and calculating the power spectrum Shannon entropy of the two respectively.

[0101] S33, extracting the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain;

[0102] S331, extract the co-polarization range image s of the target and the chaff interference from the target range image sequence and the chaff interference range image sequence obtained in S2 hh and the cross-polarization distance image s hv , extract polarization similarity from the polarization domain;

[0103] The polarization similarity is extracted from the polarization domain, specifically: according to the co-polarization range image s of the target and the foil interference hh and the cross-polarization distance image s hv , calculate the similarity of the co-polarization and cross-polarization range images of the target and the foil strip, and obtain the polarization similarity Where E(s hh (j)) is the mean value of the co-polarization range image, E(s hv (j)) is the mean value of the cross-polarization range image;

[0104] S332, co-polarization range image s of the target and chaff interference obtained according to S331 hh and the cross-polarization distance image s hv Get the polarization ratio range image of the target and the foil interference

[0105] S333, extracting the polarization ratio mean and the polarization ratio variance sum from the polarization domain according to the polarization ratio range images of the target and the chaff interference obtained in S331;

[0106] The extracting of the polarization ratio mean from the polarization domain is specifically as follows: calculating the mean of the polarization ratio range images of the target and the foil interference to obtain the polarization ratio mean

[0107] The extracting of polarization ratio variance from the polarization domain is specifically as follows: calculating the variance of the polarization ratio range image of the target and the foil interference to obtain the polarization ratio variance

[0108] S34. Merge the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity to obtain the feature vector.

[0109] S4, the labeling module labels all the samples obtained in S3 to obtain labeled samples;

[0110] The labeling is specifically: setting the feature vector label of the target range image sequence to 1 and the feature vector label of the foil strip interference range image sequence to 0, so as to obtain a labeled sample;

[0111] S5, the data partitioning module partitions the labeled samples obtained in S4 according to a certain ratio to obtain a training set and a test set;

[0112] S6, the LightGBM training module uses the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters, and then inputs the training set into the LightGBM classifier with the optimal hyperparameters for training to obtain the LightGBM classifier with the lowest classification cross entropy;

[0113] S6 specifically includes the following sub-steps:

[0114] S61, use the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters;

[0115] S62, using the training set obtained in S5, the leaf-based growth algorithm with depth restriction, and the histogram-based sparse feature optimization algorithm to train the LightGBM classifier with the optimal hyperparameters, to obtain the LightGBM classifier with the lowest classification cross entropy;

[0116] S7, the LightGBM classification module inputs the test set obtained in S5 into the LightGBM classifier with the lowest classification cross entropy obtained in S6 for classification, so as to realize the recognition and countermeasure of foil interference.

[0117] So far, from S1 to S7, a method to resist foil interference based on multi-domain features and LightGBM has been completed.

[0118] Figure 4 It is a system block diagram of signal processing when the anti-chaff interference system based on multi-domain features and LightGBM is implemented.

[0119] In the specific implementation, after preprocessing the broadband polarization radar target echo data of three backgrounds, the features of the chaff data and target data of each pulse are extracted, and finally the feature vector is obtained, which is the input sample of the LightGBM recognizer. The first background: no sea clutter dilution or centroid chaff interference background, the target data and chaff interference data have 22,400 samples each. The second background: different sea conditions dilution or centroid chaff interference background, the 6 groups of SCR under this background are 14.03dB, 9.58dB, 6.84dB, 4.97dB, 3.57dB, 2.49dB data, and each sea condition has 12,800 samples of target data and chaff interference data. The third background: shielded chaff interference background with different interference-to-signal ratios under low sea conditions. The interference-to-signal ratios of the 10 groups of this background are -20dB, -13.98dB, -6.02dB, -3.10dB, -0.92dB, 0dB, 6.02dB, 9.54dB, 12.04dB, 13.98dB using simulated interference-to-signal ratios, and the signal-to-noise ratio is 14.03dB. There are 12,800 samples for each of the target data and chaff interference data.

[0120] The first experiment: Use the training set without sea clutter dilution or centroid chaff interference background to train the LightGBM classifier, and test the adversarial results of the trained LightGBM classifier on the test set. For dilution or centroid chaff interference, recognition is adversarial. Therefore, the indicators are the commonly used indicators for evaluating the recognition effect of the classifier: accuracy, precision, recall, F1 score. In order to evaluate the radar target detection capability, the false alarm rate is also used as an evaluation indicator. After 100 experiments, the adversarial results of the trained LightGBM classifier on the test set without sea clutter dilution or centroid chaff interference background are shown in Table 1.

[0121] Table 1 Countermeasure results of non-sea clutter dilution or centroid chaff jamming

[0122] Evaluation Metrics Mean variance Accuracy 99.98％ 1.399e-08 Accuracy 99.99％ 7.8859e-09 Recall 99.99％ 7.8859e-09 F1 Score 99.98％ 1.4006e-08 False alarm rate 0.07％ 4.4176e-08

[0123] All indicators except the false alarm rate in Table 1 are higher than 99%, indicating that the anti-foil interference method based on multi-domain features and LightGBM of the present invention has a good counteracting effect on the diluted or centroid foil interference in the background without sea clutter; the variance of each indicator in Table 1 is less than 0.0001, indicating that the anti-foil interference method based on multi-domain features and LightGBM of the present invention has good stability.

[0124] The second experiment: Use the training set with different sea conditions dilution or centroid chaff interference background to train the LightGBM classifier, and test the adversarial results of the trained LightGBM classifier on the test set.

[0125] Figure 5 This is the result of the anti-chaff interference method based on multi-domain features and LightGBM in the present invention against the dilute or centroid chaff interference under different sea conditions. Figure 5 It can be seen that the confrontation results of the trained LightGBM classifier on the test set change with the change trend of SCR. The recall rate of the LightGBM classifier on the test data set gradually decreases with the increase of sea clutter power, and the false alarm probability of the LightGBM classifier increases with the decrease of SCR, indicating that sea clutter has a certain influence on the recognition effect of the model; all indicators of the LightGBM classifier on the test data set (except the false alarm rate) are above 99%, and the classification accuracy is relatively high, that is to say, the anti-chaff interference method based on multi-domain features and LightGBM of the present invention can well identify and counteract the diluted or centroid foil interference under different sea conditions; the recognition accuracy under the highest sea conditions is less than 1 percentage point lower than the recognition accuracy without sea clutter data, indicating that the anti-chaff interference method based on multi-domain features and LightGBM of the present invention has good robustness under the background of sea clutter.

[0126] The third experiment: Use the training set of obscured foil interference background with different interference-to-signal ratios to train the LightGBM classifier, and test the adversarial results of the trained LightGBM classifier on the test set of obscured foil interference background with the corresponding interference-to-signal ratio.

[0127] Figure 6 This is the result of the anti-chaff interference method based on multi-domain features and LightGBM in the present invention against shielded chaff interference with different interference-to-signal ratios. For shielded chaff interference, the mixed echo is used to identify whether there is a target in the range image of the fixed range resolution unit, that is, the target is detected and roughly measured. Figure 6 It can be seen that the confrontation results of the trained LightGBM classifier on the test set change with the interference-to-signal ratio. The larger the interference-to-signal ratio, the lower the recognition effect, but the decrease is less than 1%, indicating that the anti-foil interference method based on multi-domain features and LightGBM of the present invention has a good countermeasure effect against shielded foil interference.

[0128] The fourth experiment: Use the training sets extracted based on the method of the present invention with three backgrounds to train the LightGBM classifier respectively, and test the confrontation results of the trained LightGBM classifier on the test set. The invention patent "Millimeter wave fuze foil interference identification method based on range image feature extraction" is referred to as the existing method 1. According to the existing method 1, waveform entropy, correlation coefficient and scattering intensity ratio features are extracted to form a training set. The LightGBM classifier is trained respectively using the training sets of the three backgrounds, and the confrontation results of the trained LightGBM classifier on the test set are tested. The method of the paper "Multi-feature target anti-interference detection technology based on support vector machine (SVM)" is referred to as the existing method 2. According to the existing method 2, the echo scattering energy mean, waveform fluctuation characteristics, effective echo waveform width, scattering center dimension and target distribution characteristics are extracted to form a training set. The SVM is trained respectively using the training sets of the three backgrounds, and the confrontation results of the trained SVM on the test set are tested. The confrontation results of the two methods are shown in Table 2.

[0129] Table 2. Adversarial results of three methods under three backgrounds

[0130]

[0131] In Table 2, the recognition rates of the method of the present invention for foil interference and targets under three backgrounds are all higher than 99%, indicating that the anti-foil interference method based on multi-domain features and LightGBM of the present invention has good countermeasure effects against various types of foil interference; the recognition effects using the features of the existing method 1 and the LightGBM classifier are also both at 92%, but are lower than those of the method of the present invention, especially the recognition accuracy under the third background is 7.42% lower than that of the method of the present invention, indicating that the recognition effect of this method on shielded foil interference is slightly worse than that of the method of the present invention; the recognition effect of the existing method 2 on foil interference under the first and second backgrounds has the lowest accuracy among the three methods, and the accuracy under the third background is 57.90%, indicating that this method has almost no recognition ability for shielded foil interference.

[0132] The fifth experiment: Use the training sets under three backgrounds to train the LightGBM classifier and SVM, and record the time required for training and the accuracy of the trained LightGBM classifier and SVM on the test sets under three backgrounds. The results of the training time record are shown in Table 3. The results of the accuracy record are shown in Table 4.

[0133] Table 3 Training time (seconds) of LightGBM and SVM under three backgrounds

[0134]

[0135] Table 4 Accuracy of LightGBM and SVM under three backgrounds

[0136]

[0137] It can be seen from Table 3 that the training time of LightGBM is much shorter than that of SVM, indicating that the anti-foil interference method based on multi-domain features and LightGBM of the present invention has a fast training time, and the binary tree structure and feature segmentation points of the LightGBM classifier can be adaptively learned, and the anti-interference efficiency is high. It can be seen from Table 4 that the accuracy of the LightGBM classifier in the first two backgrounds is similar to that of the SVM classifier, both greater than 99%, but the accuracy of the LightGBM classifier in the third background is much greater than that of the SVM classifier, indicating that the use of the LightGBM classifier has a better recognition effect on shielded foil interference.

[0138] The above is only a preferred embodiment of the present invention, and the present invention should not be limited to the contents disclosed in the embodiment and the drawings. Any equivalent or modification completed without departing from the spirit disclosed in the present invention shall fall within the scope of protection of the present invention.

Claims

1. An anti-chaff jamming system based on multi-domain features and LightGBM, characterized by: It includes preprocessing module, target segmentation module, feature extraction module, label annotation module, data partitioning module, LightGBM training module and LightGBM classification module; The preprocessing module is connected to the target segmentation module, the target segmentation module is connected to the feature extraction module, the feature extraction module is connected to the label marking module, the label marking module is connected to the data partitioning module, the data partitioning module is connected to the LightGBM training module, and the LightGBM training module is connected to the LightGBM classification module; The preprocessing module first performs pulse compression and then performs range alignment on the signal collected by the broadband polarization radar to obtain a range image sequence of the horizontal polarization receiving channel; The target segmentation module determines the type of chaff interference on the range image sequence of the horizontal polarization receiving channel, segments the determined shielded chaff interference at equal intervals, and segments the determined diluted or centroid chaff interference according to CA-CFAR to obtain a target range image sequence and a chaff interference range image sequence; The feature extraction module extracts range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity based on the target range image sequence and the chaff interference range image sequence, and combines all features to obtain all feature vectors; The label marking module takes each feature vector as a sample and labels it to obtain a labeled sample; The data partitioning module partitions the labeled samples according to a certain ratio to obtain a training set and a test set; The LightGBM training module uses a training set and a grid search method to obtain the optimal hyperparameters of the LightGBM classifier, and then inputs the training set into the LightGBM classifier with the optimal hyperparameters for training to obtain a LightGBM classifier with the lowest classification cross entropy; The LightGBM classification module test set inputs the LightGBM classifier with the lowest classification cross entropy for classification, and outputs the classification result, which is 1 or 0, which specifically means whether it is a real target or not.

2. A method for anti-chaff interference based on multi-domain features and LightGBM, characterized in that: The steps include: S1, the preprocessing module preprocesses the target echo data under the foil interference of the horizontal polarization receiving channel of the broadband polarization radar to obtain the range image sequence of the horizontal polarization receiving channel; S2, the target segmentation module performs chaff interference type judgment and target segmentation on the range image sequence of the horizontal polarization receiving channel obtained by S1 to obtain a target range image sequence and a chaff interference range image sequence; S3, the feature extraction module extracts the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity of the target range image sequence and the chaff interference range image sequence obtained in S2 from the time domain, extracts the spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain, extracts the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain, combines the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity, obtains the feature vectors corresponding to all target range image sequences and chaff interference range image sequences, and takes the feature vector obtained from each sequence as a sample to obtain all samples; S4, the labeling module labels all the samples obtained in S3 to obtain labeled samples; S5, the data partitioning module partitions the labeled samples obtained in S4 according to a certain ratio to obtain a training set and a test set; S6, the LightGBM training module uses the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters, and then inputs the training set into the LightGBM classifier with the optimal hyperparameters for training to obtain the LightGBM classifier with the lowest classification cross entropy; S7, the LightGBM classification module inputs the test set obtained in S5 into the LightGBM classifier with the lowest classification cross entropy obtained in S6 for classification, so as to realize the recognition and countermeasure of foil interference.

3. The anti-chaff interference method as claimed in claim 2, characterized in that: The preprocessing described in S1 includes pulse compression, target range image alignment and normalization operations.

4. The anti-chaff interference method as claimed in claim 2, characterized in that: S2 specifically includes the following sub-steps: S21, calculating the average power of the range image sequence of the horizontal polarization receiving channel, and setting a threshold to determine the type of chaff interference, if the power is less than the threshold, it is diluted or centroid chaff interference, otherwise it is shielded chaff interference; S22. When the foil interference type is diluted or centroid foil interference, the range image sequence of the horizontal polarization receiving channel is detected by CA-CFAR, and the detected signal is segmented. The target segmentation is specifically to take n range resolution units with the midpoint of the range image of each detected target as the center as the range image sequence of the target, and obtain a target range image sequence and a foil interference range image sequence; when the foil interference type is shielded foil interference, if the target distance is unknown, divide one target at every n range resolution units of the range image, otherwise take n range resolution units with the range resolution unit of the known target as the center as the target range image sequence, and take n range resolution units in the remaining area without target information as the foil interference range image sequence, and obtain a target range image sequence and a foil interference range image sequence.

5. The anti-chaff interference method as claimed in claim 2, characterized in that: S3 is one of the feature vectors, which is obtained by the following sub-steps: S31, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain, specifically: S311, extracting a pulse range image of the target and the chaff interference from the target range image sequence and the chaff interference range image sequence obtained in S2; S312, extracting the range image concentration, energy proportion, energy mean, energy variance, waveform width and average similarity from the time domain according to the range image of a pulse of the target and the chaff interference obtained in S311; The extracting of the range image concentration from the time domain is specifically as follows: sorting the amplitudes of all the range resolution units of the range image of a pulse of the target and the chaff interference, and calculating the ratio of the upper quartile to the peak value based on the range resolution unit amplitude sequence sorted from large to small, to obtain the range image concentration; The energy proportion is extracted from the time domain, specifically, the amplitudes of all distance resolution units of the range image of a pulse of the target and the chaff interference are sorted, and the ratio of the sum of the amplitudes of the first m distance resolution units to the sum of the amplitudes of all n distance resolution units is calculated based on the range resolution unit amplitude sequence sorted from large to small, to obtain the energy proportion, and m is less than n; The energy mean is extracted from the time domain, specifically: according to the radar cross section of the chaff cloud being generally larger than the radar cross section of the target, the mean of the amplitude of the range image of a pulse of the target and the chaff interference is calculated to obtain the energy mean; The energy variance is extracted from the time domain, specifically: according to the difference in the range image distribution of the target and the chaff cloud, the variance of the amplitude of the range image of a pulse of the target and the chaff interference is calculated to obtain the energy variance; The waveform width is extracted from the time domain, specifically, according to the difference in length between the target and the chaff cloud in the radar radial direction, the product of the number of range resolution units occupied by the range image of a pulse of the target and the chaff interference when the main peak amplitude is attenuated by 90% and the range resolution is calculated to obtain the waveform width; The extracting of the average similarity from the time domain is specifically as follows: according to the geometric distribution of the target and the chaff cloud and the difference in volume transformation during movement, the average similarity of two adjacent range images is calculated; S32, extracting spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain; S321, performing fast Fourier transform on the range image of a pulse of the target and the chaff interference obtained in S311 to obtain the frequency spectrum of the target and the chaff interference; S322, extracting spectrum broadening, frequency domain skewness, and frequency domain kurtosis from the frequency domain according to the frequency spectrum of the target and the chaff interference obtained in S321; The extracting spectrum broadening from the frequency domain specifically comprises: calculating the widths of the spectrum of the target and the chaff cloud according to the difference in motion characteristics of the target and the chaff cloud to obtain the widths of the spectrum of the target and the chaff cloud; The extracting of frequency domain skewness from the frequency domain is specifically as follows: according to the difference in the frequency spectrum shapes of the target and the chaff cloud, the skewness coefficient of the frequency spectrum of the two is calculated to obtain the frequency domain skewness; The extracting of frequency domain kurtosis from the frequency domain is specifically as follows: according to the difference in the frequency spectrum shapes of the target and the foil cloud, the kurtosis coefficient of the frequency spectrum of the two is calculated to obtain the frequency domain kurtosis; S323, calculating the frequency spectrum of the target and chaff interference obtained in S321 to obtain the power spectrum of the target and chaff interference; S324, extracting power spectrum envelope fluctuation and power spectrum Shannon entropy from the frequency domain according to the power spectrum of the target and the chaff interference; The extracting of the power spectrum envelope fluctuation from the frequency domain is specifically as follows: according to the difference in the degree of change of the power spectrum envelope amplitude of the target and the chaff cloud, the envelope fluctuation of the power spectrum of the two is calculated to obtain the power spectrum envelope fluctuation; The power spectrum Shannon entropy is extracted from the frequency domain, specifically: the probability value in the entropy function is replaced by the normalized power value of each frequency resolution unit of the power spectrum of the target and the chaff cloud, and the power spectrum Shannon entropy of the two is calculated respectively; S33, extracting the polarization ratio mean, polarization ratio variance and polarization similarity from the polarization domain; S331, extracting the co-polarization range image and cross-polarization range image of the target and the chaff interference from the target range image sequence and the chaff interference range image sequence obtained in S2, and extracting the polarization similarity from the polarization domain; The extracting polarization similarity from the polarization domain specifically comprises: calculating the similarity of the co-polarization and cross-polarization range images of the target and the foil according to the co-polarization range image and the cross-polarization range image of the target and the foil interference, so as to obtain the polarization similarity; S332, obtaining a polarization ratio range image of the target and the chaff interference according to the co-polarization range image and the cross-polarization range image of the target and the chaff interference obtained in S331; S333, extracting the polarization ratio mean and the polarization ratio variance sum from the polarization domain according to the polarization ratio range images of the target and the chaff interference obtained in S331; The extracting the polarization ratio mean from the polarization domain is specifically: calculating the mean of the polarization ratio range images of the target and the foil interference to obtain the polarization ratio mean; The extracting of the polarization ratio variance from the polarization domain is specifically: calculating the variance of the polarization ratio range images of the target and the foil interference to obtain the polarization ratio variance; S34. Merge the range image concentration, energy proportion, energy mean, energy variance, waveform width, average similarity, spectrum broadening, frequency domain skewness, frequency domain kurtosis, power spectrum envelope fluctuation, power spectrum Shannon entropy, polarization ratio mean, polarization ratio variance and polarization similarity to obtain the feature vector.

6. The method for resisting foil interference as claimed in claim 2, characterized in that: S4: labeling, specifically: the feature vector label of the target range image sequence is 1, the feature vector label of the foil strip interference range image sequence is 0, and a labeled sample is obtained.

7. The method for resisting foil interference as claimed in claim 2, characterized in that: S6 specifically includes the following sub-steps: S61, use the training set obtained in S5 and the grid search method to obtain the LightGBM classifier with the optimal hyperparameters; S62. Use the training set obtained in S5, the leaf-based growing algorithm with depth limit, and the histogram-based sparse feature optimization algorithm to train the LightGBM classifier with the optimal hyperparameters to obtain the LightGBM classifier with the lowest classification cross entropy.

Images