阅读说明：本技术 一种基于箔条干扰实测数据的对抗方法 (Countermeasure method based on foil strip interference measured data ) 是由 陈伯孝 王湖升 叶倾知 于 2021-07-21 设计创作，主要内容包括：本发明公开了一种基于箔条干扰实测数据的对抗方法,包括：获取雷达的回波信号；对回波信号进行处理,得到距离-多普勒二维像；对距离-多普勒二维像进行二维恒虚警检测,得到过检测门限的点集；对过检测门限的点集进行聚类分组,得到目标信号和箔条信号分离的若干组数据；从距离-多普勒二维像中获取每组数据的若干特征信息；利用分类器和特征信息对若干组数据进行分类,得到目标信号和箔条信号的识别结果。本发明提供的基于箔条干扰实测数据的对抗方法由于采用的特征信息在箔条扩散的各个阶段表现出了与目标较高的区分度,从而使得该方法在箔条云整个扩散过程中都具备有效的对抗效果和极强的普适性。(The invention discloses a foil strip interference measured data-based countermeasure method, which comprises the following steps: acquiring an echo signal of a radar; processing the echo signal to obtain a range-Doppler two-dimensional image; performing two-dimensional constant false alarm detection on the distance-Doppler two-dimensional image to obtain a point set of an over-detection threshold; clustering and grouping the point set of the over-detection threshold to obtain a plurality of groups of data of the target signal and the foil signal; acquiring a plurality of characteristic information of each group of data from the range-Doppler two-dimensional image; and classifying the data of the plurality of groups by using the classifier and the characteristic information to obtain the identification results of the target signal and the foil strip signal. The countermeasure method based on foil strip interference actual measurement data provided by the invention has the advantages that the adopted characteristic information shows higher discrimination degree with the target at each stage of foil strip diffusion, so that the method has effective countermeasure effect and strong universality in the whole diffusion process of foil strip cloud.)

1. A countermeasure method based on foil strip interference measured data is characterized by comprising the following steps:

acquiring an echo signal of a radar;

processing the echo signal to obtain a range-Doppler two-dimensional image;

performing two-dimensional constant false alarm detection on the distance-Doppler two-dimensional image to obtain a point set of an over-detection threshold;

clustering and grouping the point sets passing the detection threshold to obtain a plurality of groups of data of the target signals separated from the foil signals;

acquiring a plurality of characteristic information of each group of data from the range-Doppler two-dimensional image;

and classifying the plurality of groups of data by using a classifier and the characteristic information to obtain the identification results of the target signal and the foil strip signal.

2. The foil strip interference measured data-based countermeasure method of claim 1, wherein the echo signal is processed to obtain a range-doppler two-dimensional image, comprising:

performing digital down-conversion processing on an echo signal of the radar to filter a carrier frequency component in the echo signal and obtain a baseband signal;

and performing pulse compression and coherent accumulation on the baseband signals to obtain a range-Doppler two-dimensional image.

3. The foil strip interference measured data-based countermeasure method of claim 2, wherein clustering and grouping the point set of the over-detection threshold to obtain a plurality of groups of data of the target signal separated from the foil strip signal comprises:

selecting a plurality of scattered points as initial clustering centers, and uniformly paving the initial clustering centers on the whole distance-Doppler plane;

calculating the offset of all the clustering centers, and performing corresponding offset on each clustering center;

if the distance between any two clustering centers is judged to be smaller than the clustering radius, the two clustering centers are merged into the middle point of the two clustering centers;

and updating the offset of the clustering centers, and clustering to separate the target signals from the foil signals until the offsets of all the clustering centers are converged to obtain a plurality of groups of data.

4. The foil strip interference measured data-based countermeasure method of claim 3, wherein the cluster center offset is calculated by the formula:

wherein x is_tIs a center point of a certain class at time t, x_t+1The offset of the central point at the moment of t +1, k is the number of points after coherent accumulation, S_hIs comprised in x_tIs the center, h is the set of points within the radius range, h is the cluster radius, x_iA point in the set of points representing an overdetection threshold.

5. The foil strip interference measured data-based countermeasure method of claim 1, wherein the step of obtaining a plurality of characteristic information of each set of data from the range-doppler two-dimensional image comprises:

acquiring the number of distance units, the number of Doppler channels and the total number of threshold-crossing points occupied by each group of data from the distance-Doppler two-dimensional image;

calculating the moment kurtosis and skewness of each distance unit frequency spectrum in the group, and respectively calculating the mean value of the moment kurtosis and skewness to obtain the moment kurtosis and skewness of the group;

and calculating the frequency deviation sum and the half-power width variance of the group according to the frequency spectrum difference of each distance unit in the group.

6. The foil strip interference measured data-based countermeasure method of claim 5, wherein calculating the kurtosis and skewness of each distance unit frequency spectrum in the group, and calculating the mean value thereof to obtain the kurtosis and skewness of the group respectively comprises:

taking the frequency spectrums which are i distance units away from the distance dimension peak point in the current group as probability density functions and counting characteristic information to obtain the probability density functions;

calculating the mean and variance of the probability density function;

calculating the moment kurtosis and skewness of the frequency spectrum which is away from the distance dimension peak value by i distance units according to the mean value and the variance of the probability density function;

and combining the moment kurtosis and skewness of 2 distance units around the peak value in the group, and respectively calculating the mean value of the moment kurtosis and skewness to obtain the moment kurtosis and skewness of the group.

7. The foil strip interference measured data-based countermeasure method of claim 6, wherein the distance dimension peak is separated by i distance units of the frequency spectrum of the moment kurtosis K_urtosis(i) The calculation formula of (2) is as follows:

skewness S of the frequency spectrum of the distance dimension peak value by i distance units_kewness(i) The calculation formula of (2) is as follows:

wherein M is a power of 2 which is greater than k and is closest to k, k is the number of points after coherent integration, and X'_i(k) For the frequency spectrum X at i range bins from the range dimension peak point_i(k) Probability density function of f_iIs X_i(k) Peak frequency of (a)_iIs X'_i(k) Standard deviation of (2).

8. The foil strip interference measured data-based countermeasure method of claim 7, wherein the frequency offset sum is calculated by the following formula:

wherein f is₀Representing peak frequencies of the spectrum corresponding to the peak distance units in the group;

the calculation formula of the half-power width variance is as follows:

wherein the content of the first and second substances,represents the 3dB width variance, B_iIs X_i(k) The calculated variance is represented by Var (·).

9. The foil strip interference measured data-based countermeasure method of claim 1, wherein the classifying of the plurality of sets of data by the classifier and the feature information to obtain the identification results of the target signal and the foil strip signal comprises:

acquiring echo signals of radars in different stages and corresponding characteristic information;

and disorganizing the acquired data according to the frame number, and sequentially inputting the corresponding characteristic information into a classifier for training and testing in a 10-fold cross validation mode to obtain a classification result of the target and the foil strips.

10. The foil strip interference measured data-based countermeasure method of claim 9, wherein the classifier comprises naive bayes, random forests, or support vector machines.

Technical Field

The invention belongs to the technical field of radars, and particularly relates to a countermeasure method based on foil strip interference measured data.

Background

Radar is a core component for accurately guiding weapons in modern war, and radar interference has become a conventional means in modern battlefields. Radar interference is a general term for all tactics and technical measures for destroying and disturbing local radar to detect the target information of the local radar. The foil strip interference is used as a common radar passive interference means for the first time in the sea warfare in 1973, and is widely applied to electronic countermeasure due to the advantages of obvious interference effect, convenient interference implementation and low interference cost. The foil strips are thrown out and scattered in the air to form a cloud-shaped interference object with certain length, width and thickness, which is called a foil strip corridor. The foil strips have no power and completely depend on natural diffusion, and the diffusion process is divided into a diffusion state and a complete diffusion state according to the stability of the corridor size and the spectrum width of the foil strips. The method is divided into a centroid type and a dilution type according to the fact that whether the position of the foil strip ejected out is in the same wave beam with a target or not, and centroid type foil strip interference can be used for covering and protecting the target by a foil strip corridor in a short distance; a dilute, disturbing foil corridor can form bait at a long distance. At present, researches prove that the interference function of a foil strip corridor is irrelevant to a radar system, and foil strip interference is an effective interference means for the existing pulse Doppler radar, phased array radar, synthetic aperture radar, millimeter wave radar and future radar of a new system, so that the research on foil strip interference resistance has very important significance on future electronic countermeasure.

At present, the following three solutions are mainly provided for the problem of how to counteract the interference of the foil strips. The 1 st is J.Chuan Yin et al researches the difference between the radar reflection sectional area characteristic, the polarization scattering characteristic, the diffusion model and other physical characteristics and the target on the basis of the modeling of the physical characteristics of the foil strip, and realizes the identification of the target and a foil strip corridor; the 2 nd is that Ran LP and Hansoo Lee, etc. preprocess the echo signal, extract the time frequency characteristic to finish the separation of the target echo by adopting the time frequency analysis mode, and Liu people, etc. adopt the angle estimation optimization algorithm, the polarization enhances the target signal, etc. to restrain the interference signal; the 3 rd is Yanchun Zuo, and Utku Kaydok et al, starting from a Range-Doppler two-dimensional map (R-D) after conventional radar signal processing, analyze the number of distance units occupied by the target and the foil strip, and the number of Doppler channels in the R-D map, and distinguish the target from the foil strip.

However, although the idea of the 1 st concept can achieve very high recognition accuracy, the practical applicability of the method is not strong because the process indexes such as the size of the foil metal, the plating layer, the polarization scattering property and the like are not uniform, and the anti-interference party cannot acquire the key information; the 2 nd idea can effectively resist the foil strip interference with stable state, but because the echo characteristic of the whole process from non-diffusion to complete diffusion is unstable, the method can not stably and effectively resist the foil strip interference of the whole diffusion process; the 3 rd idea can only accurately distinguish the target from the foil strip in the complete diffusion state, and the decision threshold needs to be adjusted according to the diffusion process, so that the false target in the non-diffusion state cannot be distinguished.

Disclosure of Invention

In order to solve the problems in the prior art, the invention provides a countermeasure method based on foil strip interference measured data. The technical problem to be solved by the invention is realized by the following technical scheme:

a countermeasure method based on foil strip interference measured data comprises the following steps:

acquiring an echo signal of a radar;

processing the echo signal to obtain a range-Doppler two-dimensional image;

performing two-dimensional constant false alarm detection on the distance-Doppler two-dimensional image to obtain a point set of an over-detection threshold;

clustering and grouping the point sets passing the detection threshold to obtain a plurality of groups of data of the target signals separated from the foil signals;

acquiring a plurality of characteristic information of each group of data from the range-Doppler two-dimensional image;

and classifying the plurality of groups of data by using a classifier and the characteristic information to obtain the identification results of the target signal and the foil strip signal.

In an embodiment of the present invention, processing the echo signal to obtain a range-doppler two-dimensional image includes:

performing digital down-conversion processing on an echo signal of the radar to filter a carrier frequency component in the echo signal and obtain a baseband signal;

and performing pulse compression and coherent accumulation on the baseband signals to obtain a range-Doppler two-dimensional image.

In an embodiment of the present invention, clustering and grouping the point sets of the over-detection threshold to obtain a plurality of groups of data of the target signal and the foil signal, including:

selecting a plurality of scattered points as initial clustering centers, and uniformly paving the initial clustering centers on the whole distance-Doppler plane;

calculating the offset of all the clustering centers, and performing corresponding offset on each clustering center;

if the distance between any two clustering centers is judged to be smaller than the clustering radius, the two clustering centers are merged into the middle point of the two clustering centers;

and updating the offset of the clustering centers, and clustering to separate the target signals from the foil signals until the offsets of all the clustering centers are converged to obtain a plurality of groups of data.

In an embodiment of the present invention, the calculation formula of the cluster center offset is:

wherein x is_tIs a center point of a certain class at time t, x_t+1The offset of the central point at the moment of t +1, k is the number of points after coherent accumulation, S_hIs comprised in x_tIs the center, h is the set of points within the radius range, h is the cluster radius, x_iA point in the set of points representing an overdetection threshold.

In one embodiment of the present invention, acquiring a plurality of feature information of each set of data from the range-doppler two-dimensional image includes:

acquiring the number of distance units, the number of Doppler channels and the total number of threshold-crossing points occupied by each group of data from the distance-Doppler two-dimensional image;

calculating the moment kurtosis and skewness of each distance unit frequency spectrum in the group, and respectively calculating the mean value of the moment kurtosis and skewness to obtain the moment kurtosis and skewness of the group;

and calculating the frequency deviation sum and the half-power width variance of the group according to the frequency spectrum difference of each distance unit in the group.

In one embodiment of the present invention, calculating the moment kurtosis and skewness of each distance unit frequency spectrum in a group, and respectively calculating the mean value thereof to obtain the moment kurtosis and skewness of the group comprises:

taking the frequency spectrums which are i distance units away from the distance dimension peak point in the current group as probability density functions and counting characteristic information to obtain the probability density functions;

calculating the mean and variance of the probability density function;

calculating the moment kurtosis and skewness of the frequency spectrum which is away from the distance dimension peak value by i distance units according to the mean value and the variance of the probability density function;

and combining the moment kurtosis and skewness of 2 distance units around the peak value in the group, and respectively calculating the mean value of the moment kurtosis and skewness to obtain the moment kurtosis and skewness of the group.

In one embodiment of the invention, the distance dimension peak is separated by i distance units_urtosis(i) The calculation formula of (2) is as follows:

skewness S of the frequency spectrum of the distance dimension peak value by i distance units_kewness(i) The calculation formula of (2) is as follows:

wherein M is a power of 2 which is greater than k and is closest to k, k is the number of points after coherent integration, and X'_i(k) For the frequency spectrum X at i range bins from the range dimension peak point_i(k) Probability density function of f_iIs X_i(k) Peak frequency of (a)_iIs X'_i(k) Standard deviation of (2).

In an embodiment of the present invention, the calculation formula of the frequency offset sum is:

wherein f is₀Representing peak frequencies of the spectrum corresponding to the peak distance units in the group;

the calculation formula of the half-power width variance is as follows:

wherein the content of the first and second substances,represents the 3dB width variance, B_iIs X_i(k) The calculated variance is represented by Var (·).

In an embodiment of the present invention, classifying the sets of data by using a classifier and the feature information to obtain the identification results of the target signal and the foil strip signal, includes:

acquiring echo signals of radars in different stages and corresponding characteristic information;

and disorganizing the acquired data according to the frame number, and sequentially inputting the corresponding characteristic information into a classifier for training and testing in a 10-fold cross validation mode to obtain a classification result of the target and the foil strips.

In one embodiment of the invention, the classifier comprises a naive bayes, a random forest, or a support vector machine.

The invention has the beneficial effects that:

1. the countermeasure method based on foil strip interference actual measurement data provided by the invention processes and classifies radar echo signals, and combines various characteristic information and classifiers to complete effective separation of the target and the foil strip; the adopted characteristic information shows higher discrimination with the target at each stage of the foil strip diffusion, so that the method has effective countermeasure effect and strong universality in the whole diffusion process of the foil strip cloud;

2. the invention considers the realization of engineering application, uses a two-dimensional constant false alarm method which is convenient to realize the application, inhibits the broadening effect of foil strip clouds, completes the high-speed, accurate and effective separation of the target and the foil strips through a mean shift clustering algorithm, and finally completes the task of identifying the target and the foil strips through a classifier.

The present invention will be described in further detail with reference to the accompanying drawings and examples.

Drawings

Fig. 1 is a flowchart of a countermeasure method based on foil strip interference measured data according to an embodiment of the present invention;

fig. 2 is a schematic structural diagram of a two-dimensional constant false alarm processor according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a countermeasure experiment procedure relating to the processing of the foil strip interference measured data according to the embodiment of the present invention;

FIG. 4 is a schematic diagram of the physical meaning of the foil strip features provided by an embodiment of the present invention;

FIG. 5 is a range-Doppler two-dimensional graph during a simulation experiment;

FIG. 6 is a graph illustrating the processing results of the mean shift algorithm;

FIG. 7 shows the accumulated results of pulse pressure processing of data in different frame segments by the method of the present invention;

FIG. 8 is a graph of the clustering results for a first and second foiled sliver bullet processed with the present invention in different frame segments;

FIG. 9 is a schematic representation of the differentiation of the stages resulting from the analysis of the diffusion process of the foil strip by the method of the invention;

FIG. 10 is a graph comparing the difference between the new characteristics of two individual foil strips and the target after processing the measured data using the method of the present invention;

fig. 11-13 are classification results obtained using na iotave bayes, random forests, and support vector machine classifiers, respectively.

Detailed Description

The present invention will be described in further detail with reference to specific examples, but the embodiments of the present invention are not limited thereto.

Example one

Referring to fig. 1, fig. 1 is a flowchart of a countermeasure method based on measured data of foil strip interference according to an embodiment of the present invention, including:

step 1: and acquiring an echo signal of the radar.

In this embodiment, the radar echo signal s (t) of the single-point target and the single foil strip wire can be represented as:

wherein the content of the first and second substances,T_ris a repetition period; t is_eIs the pulse width; mu-B/T_eFrequency modulation slope and frequency modulation bandwidth. The modulation method of the measured data is known as chirp, and each frame of data includes N ═ 32 echoes. Tau and f_dRespectively delay and doppler frequency.

R₀And R_iRespectively the distance between the single-point target and the ith foil strip; v. of₀And v_iThe speeds of the single-point target and the ith foil strip are respectively; c is the speed of light, f_cIs the carrier frequency; f. of_diThe speed fluctuation caused to the ith foil strip for the whole foil strip cloud cluster.

Assuming that there are a total of m foil strip filaments in the foil strip cloud, the total echo signal of the target and the foil strip can be expressed as:

wherein s is₀(t) represents a target echo signal, s_i(t) represents the foil strip echo signal.

Step 2: and processing the echo signals to obtain a range-Doppler two-dimensional image.

21) And carrying out digital down-conversion processing on the echo signal to filter out a carrier frequency component in the echo signal so as to obtain a baseband signal.

Specifically, in this embodiment, a conventional digital down-conversion processing method is adopted to implement carrier frequency removal processing on an echo signal, and specific steps are not described in detail here, and exp [ j2 π f is removed by the carrier frequency removal processing_c(t-τ-nT_r)]Component, the resulting baseband signal being denoted s_ddc(t)。

22) And performing pulse compression and coherent accumulation on the baseband signals to obtain a range-Doppler two-dimensional image.

Firstly, designing a filter with impulse response h (t), and carrying out pulse compression processing, wherein,

pulse compression result s_pcThe modulus value of (t) is:

|s_pc(t) | will be inA sinc function-like peak is formed.

Then, s is_pcThe resulting sampled signal x (n) sampled every period of (t) is represented as:

x(n)＝aexp(j2πf_dnT_r),n＝0,1,…,N-1

a represents the sampling amplitude, zero padding is carried out on the sampling signal x (n) to M points for FFT to obtain a coherent accumulation result:

where M is the nearest power of 2 greater than k, which represents the number of points after coherent integration.

The coherent accumulation result X (k) will be f_dT_rAt M position forms oneThe peak of the function shape.

After pulse compression and coherent accumulation, a distance-Doppler two-dimensional image of actually measured data is obtained.

And step 3: and performing two-dimensional constant false alarm detection on the distance-Doppler two-dimensional image to obtain a point set of an over-detection threshold.

Referring to fig. 2, fig. 2 is a schematic structural diagram of a two-dimensional constant false alarm processor according to an embodiment of the present invention. In the embodiment, 32 protection units and 16 reference units are respectively adopted in the distance dimension and the Doppler dimension.

Specifically, firstly, 8 units are respectively selected from the upper, lower, left and right adjacent positions of a detection unit (CUT) as protection units;

then selecting 4 units from the CUT crossing protection unit in four directions as reference units respectively, and calculating the amplitude mean value of all the reference units;

amplitude mean multiplied by a threshold factor K₀Obtaining a detection threshold, wherein the threshold factor K₀Expressed as:

K₀＝-log_G(P_fa)-1

wherein, P_faFor false alarm probability, take P_fa＝10^-6。

And comparing the detection threshold with the CUT amplitude value, and representing the point which passes through the threshold as 1 and the point which does not pass through the threshold as 0, thereby obtaining a two-dimensional CFAR result, namely a point set S which passes through the detection threshold.

And 4, step 4: and clustering and grouping the point sets passing the detection threshold to obtain a plurality of groups of data of the target signal and the foil signal.

Specifically, assuming that there are a total of L points in the threshold point set S, the coordinates of the points can be expressed as:

x_i＝(r_i,d_i)i＝1,2,…,L

wherein r is_iAnd d_iRespectively, the range bin and doppler channel of the ith point.

S_h＝{y:(y-x_t)^T(y-x_t)＜h²}

h is the clustering radius, and in this embodiment, h is 21, x_tIs the center point of a certain class at the time t; s_hIs comprised in x_tIs the center, h is the set of points within the radius.

Then, the points which pass the threshold are classified according to a mean shift clustering method.

41) And selecting a plurality of scattered points as initial clustering centers, and uniformly paving the initial clustering centers on the whole range-Doppler plane.

In this embodiment, a scatter point far greater than the sum of the target number and the foil number may be selected as the center point, for example, when the target number is p and the foil number is q, a scatter point l > 10 × (p + q) may be selected as the center point.

42) And calculating the offset of all the clustering centers, and performing corresponding offset on each clustering center.

The calculation formula of the offset of the clustering center is as follows:

43) if the distance between any two clustering centers is judged to be smaller than the clustering radius h, the two clustering centers are merged into the middle point of the two clustering centers;

44) updating the offset of the clustering centers, clustering to separate the target signals from the foil signals until the offsets of all the clustering centers are converged, and obtaining a plurality of groups of data.

Specifically, steps 42) -43) are repeated until the offset of all cluster centers tends to 0, and each cluster center is classified into one type, so that a plurality of groups of data are obtained.

And 5: and acquiring a plurality of characteristic information of each group of data from the range-Doppler two-dimensional image.

Referring to fig. 3, fig. 3 is a schematic diagram of a countermeasure experiment procedure related to foil strip interference actual measurement data processing according to an embodiment of the present invention; wherein, the area of the R-D graph is the total number of the threshold-crossing points, and the spectrum width is the number of the Doppler channels.

Specifically, step 5 comprises:

51) obtaining the number N of range units occupied by each group of data from the range-Doppler two-dimensional image_rangeNumber of Doppler channels N_DopplerAnd the total number of threshold-crossing points N_pointsThese three items of basic feature information.

52) And calculating the moment kurtosis and skewness of each distance unit frequency spectrum in each group, and respectively calculating the mean value of the moment kurtosis and skewness to obtain the moment kurtosis and skewness of each group.

Firstly, the frequency spectrum of i distance units away from a distance dimension peak point in a current group is used as a probability density function, and characteristic information is counted to obtain the probability density function.

Specifically, the frequency spectrum for the intra-group peak distance unit is X₀(k) The frequency spectrum of i distance units from the peak of the distance dimension is X_i(k) In that respect To be X_i(k) Taking the probability density function as a probability density function and counting the characteristics, firstly integrating and summing and normalizing to obtain a distribution function F_i(k) And then the difference is differentiated to obtain a standard probability density function X_i'(k)。

Then, the mean and variance of the probability density function are calculated, and the calculation formula is as follows:

then, the moment kurtosis and skewness of the distance unit frequency spectrum which is separated from the distance dimension peak value by i distance units are calculated according to the mean value and the variance of the probability density function.

In particular, the statistical mean kurtosis K_urtosisDegree of deviation S_kewnessThe calculation formula of (2) is as follows:

calculating X with reference to the above formula_i(k) The calculation formula of the moment kurtosis and skewness is as follows:

wherein, X_i(k) Has a peak frequency of f_iHalf power width of B_i。

Finally, the mean value of the moment kurtosis and skewness of 2 distance units around the peak value in the combined group is respectively calculated to obtain the moment kurtosis and skewness of the group, which are specifically expressed as follows:

53) and calculating the group of frequency deviation sum and half-power width variance according to the frequency spectrum difference of each distance unit in the group, wherein the calculation formula is as follows:

referring to fig. 4, fig. 4 is a schematic diagram illustrating physical meanings of foil features provided in an embodiment of the present invention.

To this end, a total of seven features are obtained: number of occupied distance units N_rangeNumber of Doppler channels N_DopplerTotal number of threshold crossing points N_pointsDegree of kurtosis K_urtosisDegree of deviation S_kewnessFrequency deviation and F_offset-3dB half power width variance

In the embodiment, new features are found by combining the frequency spectrum features of the distance units around the peak value on the basis of the original feature information, and the new features show higher discrimination with a target at each stage of foil strip diffusion, so that the method is not only suitable for foil strip interference in a stable state, but also suitable for the whole foil strip cloud diffusion process.

Step 6: and classifying the data of the plurality of groups by using the classifier and the characteristic information to obtain the identification results of the target signal and the foil strip signal.

Firstly, data and characteristic information of radar echo signals at different stages are obtained.

The actual measurement data of different stages in the whole foil cloud diffusion process can be processed by the method from step 1 to step 5 to obtain the characteristic change conditions of the target and the foil at different stages and the difference conditions between different foils. Each frame of acquired data can be divided into a plurality of groups after being processed, and each group of data has corresponding characteristic information.

And then, disordering the acquired data according to the frame number, sequentially inputting the corresponding characteristic information into a classifier by adopting a 10-fold cross validation mode for training and testing, and obtaining a classification result of the target and the foil strips by utilizing the classification result.

Specifically, with reference to machine learning methods, the shuffled data is divided into a training set and a test set, where 90% of the data is used for training and 10% of the data is used for testing. Then, classification recognition is carried out by using a classifier. The classifier can adopt any one of naive Bayes, random forests or support vector machines, and can also adopt other classifiers.

The invention considers the realization of engineering application, uses a two-dimensional constant false alarm method which is convenient to realize the application, inhibits the broadening effect of foil strip clouds, completes the high-speed, accurate and effective separation of the target and the foil strips through a mean shift clustering algorithm, and finally completes the task of identifying the target and the foil strips through a classifier.

In the embodiment, the range-doppler two-dimensional signal is obtained by pulse compression and coherent accumulation processing of a conventional radar. Then starting from a distance-Doppler two-dimensional graph, combining an analysis method of distance and frequency distribution characteristics, firstly introducing a mean shift clustering algorithm to separate a target and a foil strip point set, and then providing frequency deviation and other new characteristic information to assist in completing antagonistic recognition of the whole diffusion process. The whole set of method is applied to a large amount of foil strip interference resistance actual measurement data of a certain phase-coherent terminal-guided radar, the change condition and the stability of various characteristics of foil strip interference in the whole process from the ejection to the complete diffusion of foil strips are analyzed, and finally, a machine learning classifier is used for completing the classification and identification of targets and foil strips. Theoretical analysis and actually measured data confrontation results show that the foil interference confrontation method can accurately identify the foil and the target, and has effective confrontation effect and strong universality in the whole spreading process of the foil cloud.

Example two

The beneficial effects of the present invention are further explained by the simulation test below.

1. Test conditions and parameter settings

The simulation test is carried out by taking a target and two foil strips as examples, and particularly, data from a Ku-band pulse Doppler radar are collected. The radar monitors the entire process of only the presence of the target → the beating of the first sliver bullet → the beating of the second sliver bullet, observing the entire diffusion process of the foil bullets. Each frame of the batch of data comprises 32 echoes; 32 protection units and 16 reference units are taken by the two-dimensional constant false alarm in the range Doppler; the clustering radius of the mean shift is 21, and three classifiers of a support vector machine, naive Bayes and a random forest are respectively adopted for target identification.

2. Test platform

Software: windows10 professional edition 64 bit, MATLAB2020b and its Classification leader toolbox.

Hardware: a CPU: i9-10980 XE; memory: 256GB DDR 43200 MHz.

3. Test results and analysis

Referring to fig. 5, fig. 5 is a two-dimensional range-doppler plot during a simulation experiment; wherein, (a) the diagram shows a two-dimensional signal diagram formed on a distance-Doppler plane after pulse compression and coherent accumulation; (b) the graph (c) and the graph (d) respectively represent the processing results of the one-dimensional constant false alarm and the two-dimensional constant false alarm, and the two graphs are contrasted. As can be seen from the graph (a) in fig. 5, the pulse pressure accumulation result shows that the foil strips are widened to form a widening effect of a doppler dimension around the peak, the widening effect cannot be removed by the one-dimensional constant false alarm processing result in (b), and the spectrum widening effect is effectively suppressed by the two-dimensional constant false alarm processing in (c).

Fig. 6 is a diagram illustrating the processing results of the mean shift algorithm. As can be seen from FIG. 6, after the mean shift clustering, the constant false alarm scatter results are divided into three groups, and the three groups do not have aliasing, and the clustering result is accurate.

FIG. 7 shows the accumulated results of pulse pressure processing of data in different frame segments by the method of the present invention; wherein, (a) (d) (g) respectively shows the original echo signals of the preparation stage, the first foiling stripe bomb and the second foiling stripe bomb, (b) (e) (h) respectively shows the pulse compression processing results of the preparation stage, the first foiling stripe bomb and the second foiling stripe bomb, and (c) (f) (i) respectively shows the coherent accumulation processing results of the preparation stage, the first foiling stripe bomb and the second foiling stripe bomb.

FIG. 8 is a graph of the clustering results for a first and second foiled sliver bullet processed with the present invention in different frame segments; wherein, the (a) diagram and the (f) diagram respectively show the constant false alarm, the clustering result and the range-Doppler dimension characteristics of the targets, the foil strips 1 and the foil strips 2 of the frames 917, 941, 1000, 1426, 3812 and 5428. It can be seen from figure 8 that the target remains stationary at the range-doppler plane all the time, while the foil strips 1, 2 are punched out to a size that is smaller than the target, then approximates the target, and finally is much larger than the target. FIG. 9 is a schematic representation of the differentiation of the stages resulting from the analysis of the diffusion process of the foil strip by the method of the invention.

FIG. 10 is a graph comparing the difference between the new characteristics of two individual foil strips and the target after processing the measured data using the method of the present invention; wherein, the graphs (a) - (c) are respectively the frequency distribution of the distance units around the peak points of the target, the foil strip bullet 1 and the foil strip bullet 2. As can be seen from fig. 10, the frequency spectrum of the distance unit around the target peak always keeps the standard characteristics of center alignment, bilateral symmetry and uniform width, because the target is essentially composed of rigid bodies; the peak frequency spectrums of the foil strips 1 and 2 may be similar to the target, but after the analysis of the distance units around the peak values, the frequency spectrum centers cannot be aligned, are asymmetric left and right, have obvious width fluctuation and have huge difference with the target. The feature here is that the foil strip of the invention interferes with the novel feature.

Fig. 11-13 are classification results obtained using na iotave bayes, random forests, and support vector machine classifiers, respectively, after seven features are obtained using the method of the present invention. Wherein, fig. 11 is a naive bayes classification result, fig. 12 is a random forest classification result, and fig. 13 is a support vector machine classification result. As can be seen from FIGS. 11-13, the data of the three classifier confusion matrices are centered on the left diagonal, indicating that the classification results are substantially correct. In the radar guidance process, the confusion between the foil strips 1 and 2 does not influence the radar work, so the accuracy in brackets is the accuracy which tolerates the error, and the accuracy outside the brackets is the accuracy which does not tolerate the error. Obviously, in terms of accuracy, the random forest > support vector machine > naive bayes, in terms of prediction speed, naive bayes > support vector machine > random forest, and in terms of training rate, naive bayes > random forest > support vector machine. The three classifiers have advantages, if offline learning neglect, namely training speed is considered, the support vector machine has the advantages of high accuracy and high prediction speed.

The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.