Bistatic radar passive interference identification method based on HRRP characteristics

Abstract

The invention discloses a bistatic radar passive interference identification method based on HRRP characteristics. The method comprises the following steps: firstly, carrying out pre-processing operations such as normalization, alignment and support region range determination on HRRP features, and then extracting 18 features including peak point length, normalization variance, front-end structure ratio and the like; and a feature selection algorithm based on a Relief algorithm and a specificity coefficient is used to select features with large discrimination and low redundancy to construct a feature vector, and finally, an SVM is used to complete passive interference identification. The invention proves that the bistatic radar and the selected feature combination can effectively improve the interference identification performance.

Description

Technical Field

[0001] The invention belongs to the technical field of radar passive interference identification, and in particular relates to a dual-base radar passive interference identification method based on HRRP features. Background Art

[0002] With the continuous development of radar system and radar technology, bistatic radar can play an important role in electronic countermeasures because of its advantages of high concealment, strong anti-reconnaissance capability, and not vulnerable to anti-radiation weapons. When radar detects or identifies target signals, interference signals will destroy its normal function. Compared with active interference, passive interference represented by chaff and corner reflection has a larger radar cross section (RCS) and can scatter energy in different directions, so it poses a great threat to the detection performance of bistatic radar.

[0003] In order to reduce the interference effect of foil strips and corner reflectors on radar, many researchers have studied passive interference identification and found that radar can use the large signal bandwidth characteristics to achieve high-resolution imaging of the distance of the radial target scattering center and characterize the fine geometric structure characteristics and length information of the target, which is of great significance in passive interference identification. In the passive interference identification based on high-resolution range profile (HRRP), the main method is to extract HRRP features of different dimensions and use classifiers for identification. However, the existing passive interference identification methods based on HRRP are all aimed at single-base radars and single interference types, and the research on HRRP feature extraction and passive interference multi-classification of dual-base radars is not sufficient. Summary of the invention

[0004] In response to the above problems and current research trends, this paper proposes a dual-base radar passive interference identification method based on HRRP features, which can comprehensively evaluate feature discrimination and redundancy to select efficient identification features and improve recognition performance.

[0005] The present invention proposes a bistatic radar passive interference identification method based on HRRP features, comprising the following steps:

[0006] Step S1, simulate and construct the HRRP data set, and perform a series of preprocessing operations on the HRRP, aiming to make the features retain the effective information of the target as much as possible:

[0007] Step S1.1, normalizing HRRP to eliminate the influence of different target distances on HRRP amplitude;

[0008] Step S1.2, aligning the HRRP to move the center of gravity of the target to the center of the window;

[0009] Step S1.3, limiting the range of the distance unit occupied by the target to reduce the influence of noise outside the target unit on feature extraction;

[0010] Step S2, extracting a total of 18 features including peak point length, normalized variance, and front-end structure ratio, etc., as a basis for subsequent feature selection;

[0011] Step S3, using a feature selection algorithm based on a relief algorithm and a specificity coefficient to select features with high discrimination and low redundancy to construct a feature vector;

[0012] Step S4, dividing the constructed feature vector into a training set and a test set, and performing training and testing using the SVM classifier.

[0013] Furthermore, the preprocessing operation of HRRP mainly includes normalization, alignment and determination of the support area range. In order to eliminate the influence of different target distances on the HRRP amplitude, HRRP is first normalized. Let the sequence of HRRP modulus values ​​be x0, then after normalization, it can be expressed as:

[0014]

[0015] Where max(·) represents the maximum value, K is the HRRP length, and the HRRP sequence value range is mapped to 0 to 1 after normalization.

[0016] Different targets have different positions in the range window, so the HRRP is aligned to move the target's center of gravity to the center of the window. First, the center of gravity position G of the sequence is calculated, and then the HRRP is cyclically shifted to move the position corresponding to G to the midpoint of the sequence. The center of gravity G can be calculated by the following formula:

[0017]

[0018] Finally, it is necessary to limit the range of distance units occupied by the target to reduce the impact of noise outside the target unit on feature extraction. In HRRP, the radial distance interval where the target position is located is the support area. noise The sampling points are regarded as noise, their mean is calculated and multiplied by an adjustable bias value as the threshold, and the interval between the first and last sampling points above the threshold is taken as the support area S(k), k=1,2,……,N.

[0019] Furthermore, the feature extraction mainly extracts HRRP scale features and structural features. HRRP scale features mainly characterize the location, number and interval of the sampling points (referred to as peak points) where the peak value is located in the distance dimension, and mainly reflect the radial size-related properties of the target. HRRP structural features mainly characterize the characteristics of the proportional structure, and mainly reflect the dispersion, undulation and other structural characteristics of the target's strong scattering points in the radial direction. A total of 18 features are extracted, and the definitions of each feature are as follows:

[0020] 1) Peak point length (F1): It is defined as the distance between the two peak points with the longest interval in the target support area, which can be expressed as:

[0021] F1=f s ·[P k (N peak )-P k (1)] (3)

[0022] Among them, P k Indicates the position of the peak point in the target support area, f s Represents the radar sampling interval, N Peak is the number of peak points.

[0023] 2) Number of peak points (F2): defined as the number of peak points in the target support area, which can be expressed as:

[0024] F2=N Peak (4)

[0025] 3) 25% length (F3): It is defined as the length of the interval where the energy in the target support area drops by 25% from the maximum peak point, which can be expressed as:

[0026] F3=(N 25,b -N 25,a )·f s (5)

[0027] Among them, N 25,a ,N 25,b They represent the upper and lower bounds of the sampling point where the energy drops by 25% from the maximum peak energy.

[0028] 4) 50% length (F4): It is defined as the length of the interval where the energy in the target support area drops by 50% from the maximum peak point, which can be expressed as:

[0029] F4=(N 50,b -N 50,a )·f s (6)

[0030] Among them, N 50,a ,N 50,bThey represent the upper and lower bounds of the sampling point where the energy drops by 50% from the maximum peak energy.

[0031] 5) Number of 25% peak points (F5): defined as the number of peak points within the 25% length interval, which can be expressed as:

[0032] F5=len[P k (n)],N 25,a <n<N 25,b (7)

[0033] Where len(·) represents the sequence length.

[0034] 6) 50% peak point number (F6): defined as the number of peak points within the 50% length interval, which can be expressed as:

[0035] F6=len[P k (n)],N 50,a <n<N 50,b (8)

[0036] 7) 25% length ratio (F7): defined as the ratio of 25% length to peak point length, which can be expressed as:

[0037] F7=F3 / F1 (9)

[0038] 8) 50% length ratio (F8): defined as the ratio of 50% length to peak point length, which can be expressed as:

[0039] F8=F4 / F1 (10)

[0040] 9) Sorting energy ratio (F9): defined as the ratio of the top 5% points in the target area to the total target energy. The specific calculation process is:

[0041] a. Sort the energy of each sampling point in the target support area in descending order;

[0042] b. Sum the top 5% of the HRRP energy in descending order;

[0043] c. Calculate the ratio of the sum of the top 5% of HRRP to the total energy, which is the permutation energy ratio feature F9.

[0044] 10) Normalized variance (F10): It is defined as the maximum value of the variance of the sampling points in all windows divided by the mean value for all points in the support area. The specific calculation process is:

[0045] a. Extend the target HRRP support area by 3 sampling points on both sides;

[0046] b. Slide the window in the extended HRRP area with a window size of 4 sampling points, and calculate the mean and standard deviation of the amplitude of the sampling points in each window;

[0047] c. Find the maximum value of the ratio of the mean and standard deviation and convert it into a dB value, which is the normalized variance feature F of the current target 10 .

[0048] 11) Sparsity (F11): It is defined as the variance of the distances between all adjacent peak points within the target support area. The specific calculation process is as follows:

[0049] a. Find all peak points in the target HRRP support area;

[0050] b. Calculate the distances between adjacent points of these peak points;

[0051] c. Calculate the variance of these distances, which is the sparsity feature F of the current target 11 .

[0052] 12) Primary-secondary peak ratio (F12): defined as the energy ratio of the maximum peak to the second largest peak in the target support area, which can be expressed as:

[0053]

[0054] Where E peak1 Indicates the energy of the maximum peak, E peak2 Represents the energy of the next largest peak.

[0055] 13) Front-end structure ratio (F13): defined as the ratio of the distance from the left secondary peak to the highest peak to the length of the target support area. The left secondary peak refers to the maximum value on the left side of the highest peak of HRRP (excluding the target highest peak).

[0056] 14) Back-end structure ratio (F14): defined as the ratio of the distance from the right secondary peak to the highest peak to the length of the target support area. The right secondary peak refers to the maximum value on the right side of the highest peak of HRRP (excluding the highest peak of the target).

[0057] 15) Symmetry (F15): It is defined as the ratio of the energy sum of the first half of the target support area to the energy sum of the second half. It reflects the symmetry of the target scattering cross-sectional area distribution and can be expressed as:

[0058]

[0059] Among them, S(k) represents the energy value of the kth sampling point, and N represents the number of sampling points in the support area.

[0060] 16) Dispersion (F16): reflects the dispersion of the target HRRP radial amplitude distribution, which can be expressed as:

[0061]

[0062] Among them, S(k) represents the energy value of the kth sampling point, and N represents the number of sampling points in the support area.

[0063] 17) Descaled structural feature (F17): reflects the HRRP amplitude feature that is independent of scale transformation and can be expressed as:

[0064]

[0065] 18) Fluctuation structure characteristics (F18): reflects the strength of the target HRRP amplitude fluctuation, which can be expressed as:

[0066]

[0067] Among them, S max Indicates the maximum energy of the peak point in the support area, S peak and M represent all energies greater than 0.25S max The energy and number of the peak points, b1(k) and b2(k) represent the positions of the nearest peak points on the left and right sides of the kth peak point among the M peak points, S b1 (k) and S b2 (k) represents their respective energies.

[0068] Furthermore, the feature selection is mainly based on the Relief algorithm and the specificity coefficient to select efficient identification features with large feature discrimination and low redundancy to construct a feature vector. The specific method of the Relief algorithm is as follows:

[0069] The Relief algorithm is used to evaluate the discrimination of features to indicate the ability of this feature to distinguish different categories. The core of the Relief algorithm is to measure the distance between a sample and its neighboring samples of the same category and neighboring samples of different categories. When the average distance between a sample and samples of the same category is greater than the average distance between a sample and samples of different categories, it means that the feature has a higher discrimination for different categories and a larger weight value. The steps are as follows: First, randomly select a sample x from the data set, and then select k samples from each category of the data set, where samples of the same category as sample x form set H, and samples of different categories form set M. Then calculate the sample distance to update the discrimination of each feature, as shown in the following formula:

[0070]

[0071] Among them, max(x a ) and min(x a) respectively represent the maximum and minimum eigenvalues ​​of feature a in all samples, P(c) is the proportion of samples of category c in all samples, and q is the number of times samples are drawn. class(x) represents the category to which sample x belongs.

[0072] The specificity coefficient can characterize the degree of consistency between the distribution of a feature and other features in the sample. The more consistent the distribution, the weaker the uniqueness of the feature and the higher the redundancy. Let the number of features in a feature set be F, and define the specificity coefficient of the pth feature as:

[0073]

[0074] Among them, C is the number of categories, K is the number of samples, and X i,j,k is the kth sample value of the jth feature in the ith category, m i,j is the mean of the j-th feature in the i-th category. p When it is small, it means that the specificity of the pth feature is weak and it can be considered as a redundant feature and removed.

[0075] Beneficial effects:

[0076] 1. Feature extraction for bistatic radar: The present invention focuses on the HRRP feature extraction of bistatic radar. It aims to solve the problem that the existing methods are not sufficiently studied under bistatic radar and can more effectively identify various interference types.

[0077] 2. Efficient feature selection: By combining the relief algorithm and the specificity coefficient, the present invention comprehensively evaluates the discrimination and redundancy of features, thereby selecting efficient recognition features and improving the accuracy and efficiency of recognition.

[0078] 3. Verification and performance improvement: The support vector machine is used to verify the effect of feature selection, which further improves the recognition performance of passive interference and provides stronger technical support for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 The overall framework diagram of the bistatic radar passive interference identification method based on HRRP features;

[0080] Figure 2 It is a schematic diagram of the HRRP support area;

[0081] Figure 3 This is a graph showing the change in recognition rate under different numbers of features;

[0082] Figure 4 The comparison chart of single and double base recognition rate under different feature numbers and interference-noise ratios;

[0083] Figure 5This is a comparison chart of interference recognition rate using different feature combination methods. DETAILED DESCRIPTION

[0084] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is clearly and completely described below in conjunction with the embodiments of the present application and the accompanying drawings.

[0085] The detailed process of this implementation method is as follows:

[0086] Figure 1 This is the overall framework diagram of the dual-base radar passive interference identification method based on HRRP features. The overall method is divided into four parts: preprocessing, feature extraction, feature selection, and SVM:

[0087] Step S1, simulate and construct the HRRP data set, and perform a series of preprocessing operations on the HRRP features:

[0088] Through simulation, HRRP datasets for single-base and dual-base radars are constructed, including HRRPs for three types of targets: chaff cloud, corner reflector array, and ship, with 600 samples for each type. The HRRP of targets under different observation azimuth conditions is simulated by changing the relative position of the radar and the target. For single-base radar HRRP, the incident and receiving azimuths are the same, both 50° to 130°; for dual-base radar HRRP, the incident azimuth is 0° and the receiving azimuth is 50° to 130°. All HRRPs are normalized, aligned, and preprocessed to determine the support area range. The schematic diagram of the support area range is shown in the figure. Figure 2 shown.

[0089] Step S2, extracting a total of 18 features including peak point length, normalized variance, and front-end structure ratio.

[0090] Step S3, based on the feature selection algorithm of the Relief algorithm and the specificity coefficient, the features with high discrimination and low redundancy are selected to construct the feature vector. First, the HRRP features are sorted from high to low according to the discrimination according to the Relief algorithm, and then the specificity coefficient value is calculated to evaluate the redundancy of each feature, and the features are eliminated according to the number of feature selections. After selecting a certain number of the remaining features according to the sorting, the feature vector is constructed.

[0091] The features of the monostatic and bistatic radar HRRP features are selected based on the proposed method, and the results are shown in Tables 1 and 2. For the monostatic radar HRRP features, 50% length (F4), 25% length ratio (F7) and sorting energy ratio (F9) have high discrimination for different types of targets, while the high discrimination features of the bistatic radar HRRP are mainly the undulating structure feature (F18), sorting energy ratio (F9) and the number of peak points (F2). This shows that the recognition features of the passive interference HRRP under monostatic and bistatic radars are not exactly the same.

[0092] Table 1

[0093]

[0094]

[0095] Table 2

[0096]

[0097] Step S4, divide the constructed feature vector into training set and test set according to 8:2, and use SVM classifier for training and testing:

[0098] The training set samples are input into the SVM classifier for training. After the training is completed, the test set is extracted according to the feature selection results to construct a feature vector, and then sent to the trained classifier to complete the passive interference identification. Since SVM is a two-classification algorithm, and the passive interference identification of the present invention is a three-classification problem, it is necessary to use SVM for two-level classification: first train a classifier to classify "category A" and "category B+C", and then train a classifier to further classify all "category B+C" samples output by the previous level into "category B" and "category C".

[0099] Without loss of generality, the accuracy metric is used to evaluate the recognition performance.

[0100] Figure 3 This is a graph showing the variation of recognition rates under different numbers of features. 1 to 18 HRRP features sorted by discrimination under dual bases are selected to construct feature vectors and the recognition performance is tested using SVM.

[0101] from Figure 3It can be seen that when the feature F18 with the highest discrimination is used alone as input, the interference recognition rate is only 62.3%. This shows that it is difficult to distinguish different categories by considering only a single feature. As the number of features increases, the interference recognition rate shows a trend of first increasing and then decreasing. When the number of input features is 6 to 7, the interference recognition rate is as high as 92.7%. When all 18 features are input, the interference recognition rate is 77.3%, indicating that some features with low discrimination have a negative impact on the construction of the SVM hyperplane due to the similar or large overlap of the feature value distributions for different targets, resulting in a decrease in model performance. Comparison Figure 3 As shown in Table 2, when the number of features increases from 4 to 10, more redundant features are added, the interference recognition performance does not change much and consumes more computation. These results prove that the proposed feature selection method can comprehensively characterize the effectiveness of features and provide key information for constructing feature vectors.

[0102] Next, the interference recognition performance of the single-base and dual-base radars is compared. The same number of feature selections is set to 4, 8, and 16, respectively, and the recognition rates of the two are compared under different JNR conditions. The results are as follows: Figure 4 As shown in Figure 2, it can be seen that the recognition effect is not good when the number of features is low at low JNR, which is because the strong noise covers up most of the discriminative HRRP features.

[0103] The average recognition rate performance comparison is shown in Table 3. The average recognition rate of the bistatic radar under different JNRs is 2.4% to 5.4% higher than that of the monostatic radar, and the number of features required for the bistatic radar to achieve the highest average recognition rate is less than that of the monostatic radar. This shows that the bistatic radar is a better passive interference recognition system after introducing the receiving direction degree of freedom.

[0104] Table 3

[0105]

[0106] Based on the bistatic radar, the feature vector constructed by the feature selection method proposed in this paper is compared with the dimension, feature length and distribution entropy features of the equivalent scattering center under different JNRs. Among them, the number of feature selections in the present invention is set to 3, and the results are as follows Figure 5 The recognition rate of the proposed method is higher than that of the comparison method under different JNR conditions, which proves that the proposed feature selection method has better recognition performance.

[0107] In summary, the passive interference identification method based on SSM-Transformer proposed in this paper can comprehensively evaluate the feature discrimination and redundancy to select efficient identification features and improve the recognition performance.

[0108] The above description shows and describes several preferred embodiments of the present invention, but as mentioned above, it should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the invention concept described herein through the above teachings or the technology or knowledge of the relevant field. Changes and variations made by those skilled in the art do not depart from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A bistatic radar passive interference identification method based on HRRP features, characterized in that: The method comprises the following steps: Step S1, constructing a HRRP data set by simulation, and performing a series of preprocessing operations on the HRRP, such as normalization, alignment, and determination of the support area range; Step S2, extracting a total of 18 features including peak point length, normalized variance, and front-end structure ratio; Step S3, using a feature selection algorithm based on a relief algorithm and a specificity coefficient to select features with high discrimination and low redundancy to construct a feature vector; Step S4, dividing the constructed feature vector into a training set and a test set, and performing training and testing using the SVM classifier.

2. A bistatic radar passive interference identification method based on HRRP features as claimed in claim 1, characterized in that: The step S1 comprises the following steps: Step S1.1, normalizing HRRP to eliminate the influence of different target distances on HRRP amplitude; Step S1.2, aligning the HRRP to move the center of gravity of the target to the center of the window; Step S1.3, limit the range of the distance unit occupied by the target to reduce the impact of noise outside the target unit on feature extraction.

3. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: The method for normalizing HRRP in step S1 is: Assume that the sequence of HRRP modulus values ​​is x0, which can be expressed as follows after normalization: Where max(·) represents the maximum value, K is the HRRP length, and the HRRP sequence value range is mapped to 0 to 1 after normalization.

4. A bistatic radar passive interference identification method based on HRRP features as claimed in claim 1, characterized in that: The calculation method of the center of gravity G in step S1 is: Wherein, x0 is a sequence of HRRP module values, and K is the HRRP length.

5. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: In step S1, the radial distance interval where the target position is located in HRRP is the support area, and the first N noise The sampling points are regarded as noise, their mean is calculated and multiplied by an adjustable bias value as the threshold, and the interval between the first and last sampling points above the threshold is taken as the support area S(k), k = 1, 2, ..., N.

6. A bistatic radar passive interference identification method based on HRRP features as claimed in claim 1, characterized in that: The 18 features extracted in step S2 are: peak point length (F1), number of peak points (F2), 25% length (F3), 50% length (F4), 25% number of peak points (F5), 50% number of peak points (F6), 25% length ratio (F7), 50% length ratio (F8), sorting energy ratio (F9), normalized variance (F10), sparsity (F11), primary and secondary peak ratio (F12), front-end structure ratio (F13), back-end structure ratio (F14), symmetry (F15), dispersion (F16), descaled structural features (F17), and undulating structural features (F18).

7. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: The specific method of the Relief algorithm in step S3 is as follows: The Relief algorithm is used to evaluate the discrimination of a feature to indicate the ability of this feature to distinguish different categories. The steps are as follows: first, a sample x is randomly selected from the data set, and then k samples are selected from each category of the data set, where samples of the same category as sample x form a set H, and samples of different categories form a set M; then the sample distance is calculated to update the discrimination of each feature, as shown in the following formula: Among them, max(x a ) and min(x a ) represent the maximum and minimum eigenvalues ​​of feature a in all samples, P(c) is the proportion of samples of category c in all samples, q is the number of times samples are drawn; class(x) represents the category to which sample x belongs.

8. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: In step S3, the method for defining the specificity coefficient of the pth feature in the feature set is: Among them, F is the number of features, C is the number of categories, K is the number of samples, and X is the number of i,j,k is the kth sample value of the jth feature in the ith category, m i,j is the mean of the j-th feature in the ith category.

9. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: In step S3, when the specificity coefficient SC p When it is small, it means that the specificity of the pth feature is weak, and it can be considered as a redundant feature and eliminated. The specific method is as follows: first, the HRRP features are sorted from high to low according to the Relief algorithm, and then the specificity coefficient value is calculated to evaluate the redundancy of each feature, and features are eliminated according to the number of feature selections; a certain number of the remaining features are selected according to the sorting and then the feature vector is constructed.

10. The method for identifying bistatic radar passive interference based on HRRP features as claimed in claim 1, characterized in that: The specific method of step S4 is as follows: The constructed feature vector is divided into training set and test set according to 8:

2. The training set samples are input into the SVM classifier for training. After the training is completed, the corresponding features of the test set are extracted according to the feature selection results to construct the feature vector, which is sent to the trained classifier to complete the passive interference identification.

11. The method for identifying bistatic radar passive interference based on HRRP features according to claim 1, characterized in that: In step S4, SVM is used for two-level classification: first, a classifier is trained to classify "category A" and "category B+C", and then a classifier is trained to further classify all "category B+C" samples output by the previous level into "category B" and "category C".

Images