CN112882011A - Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics - Google Patents

Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics Download PDF

Info

Publication number
CN112882011A
CN112882011A CN202110032916.5A CN202110032916A CN112882011A CN 112882011 A CN112882011 A CN 112882011A CN 202110032916 A CN202110032916 A CN 202110032916A CN 112882011 A CN112882011 A CN 112882011A
Authority
CN
China
Prior art keywords
sample set
training
peak
frequency
function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110032916.5A
Other languages
Chinese (zh)
Inventor
王鹏辉
杨浩蔚
刘宏伟
丁军
陈渤
徐一兼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN202110032916.5A priority Critical patent/CN112882011A/en
Publication of CN112882011A publication Critical patent/CN112882011A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/02—Systems using reflection of radio waves, e.g. primary radar systems; Analogous systems
    • G01S13/50—Systems of measurement based on relative movement of target
    • G01S13/52—Discriminating between fixed and moving objects or between objects moving at different speeds
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/28—Details of pulse systems
    • G01S7/285—Receivers
    • G01S7/292—Extracting wanted echo-signals
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/415—Identification of targets based on measurements of movement associated with the target

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a radar carrier frequency change steady target identification method based on frequency domain correlation characteristics, and mainly solves the problem that the classification performance of three types of airplane targets is reduced when radar carrier frequencies are changed in the prior art. The implementation scheme is as follows: 1) respectively carrying out airplane body compensation and fast Fourier change on the time domain echo signals of the training and testing sample set to obtain respective Doppler domain echo signals, and calculating respective frequency peak functions; 2) extracting variance characteristics, entropy characteristics, threshold peak value number characteristics and position characteristics of a first threshold peak of training and testing sample frequency-peak functions to form a training and testing characteristic matrix; 3) normalizing the training characteristic matrix and inputting the result into a classifier for training; 4) and normalizing the test feature matrix and inputting the result into a trained classifier to obtain a classification result. The method still has a good classification effect when the radar carrier frequency changes, and can be used for classifying different types of airplanes.

Description

Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics
Technical Field
The invention belongs to the technical field of radars, and particularly relates to an aircraft target classification method which can be used for classifying different types of aircrafts under the condition that radar carrier frequencies are changed.
Background
The micro-Doppler characteristics of the target reflect the unique geometric characteristics of the target, and are one of the important ways to realize the classification of the target. Because the micro-motion parts of the helicopter, the propeller aircraft and the jet aircraft have obvious differences in the aspects of structure, motion characteristics and the like, the micro-Doppler modulation characteristics of the three types of aircrafts are different, and classification of targets of the three types of aircrafts can be realized by extracting the micro-Doppler characteristics.
The echo parameter model of the vertical rotor is established in a Master paper 'airplane target classification method research based on JEM effect' issued in 2014 of Li-Weiluo, and four-dimensional time domain correlation characteristics are extracted, so that the classification of three types of airplanes is realized.
Classification of airplane targets is realized according to differences of envelope fluctuation degrees and frequency domain spectral line dispersion degrees in three types of airplane Doppler domains in the doctor paper 'micro Doppler echo simulation and micro motion feature extraction technical research' published in 2010 in Daohu.
Du L, Wang B, Li Y et al published in 2013 in an article "Robust Classification Scheme for aircraft Targets With Low Resolution radio Based on EMD-CLEAN Feature Extraction Method" of IEEE outputs Journal analyzed the micro-Doppler spectra of helicopters, propellers and jets, and proposed a Feature Extraction Method Based on EMD and CLEAN algorithms to realize the Classification of three types of Airplane Targets.
The classification methods proposed in the above documents all have a premise that the radar carrier frequencies in the training samples and the test samples are kept unchanged. However, under the actual working condition of the radar, in order to improve the anti-interference performance of the radar system and solve the problems of blind speed and fuzzy distance, the radar carrier frequency changes, and the change of the radar carrier frequency can form the mismatching of the radar carrier frequency in the training sample and the radar carrier frequency in the test sample, and the mismatching of the radar carrier frequency can influence the micro-doppler modulation effect of three types of airplane targets, so that the scattering condition of the original characteristics in the characteristic space is changed, and therefore, the classification of the airplane is carried out by using the classification methods, and the classification performance is reduced.
Disclosure of Invention
The invention aims to provide a radar carrier frequency change steady target identification method based on frequency domain correlation characteristics aiming at the defects of the prior art, so as to ensure that the classification performance of three types of airplane targets cannot be reduced when the radar carrier frequency changes.
In order to achieve the above object, the method comprises the following steps:
(1) extracting N groups of time domain echo signals from a radar single carrier frequency echo database of three types of targets, namely a helicopter, a propeller plane and a jet plane, as a training sample set, and extracting M groups of time domain echo signals from three types of target echo signals received by the radar as a test sample set;
(2) sequentially and respectively carrying out aircraft fuselage compensation and fast Fourier transform of P points on N groups of time domain echo signals of the training sample set and M groups of time domain echo signals of the testing sample set to obtain N groups of Doppler domain echo signals U of the training sample setn(k) And M groups of Doppler domain echo signals U of test sample setm(k) Wherein N is 1,2, the sample sequence number of the training sample set, M is 1,2, the sample sequence number of the test sample set;
(3) calculating N groups of Doppler domain echo signals U of training sample setn(k) Frequency-peak function fpn(l) And M groups of Doppler domain echo signals U of test sample setm(k) Frequency-peak function fpm(l) Wherein N is a training sample set sample number, M is a testing sample set sample number, l is 1,2, and fix (P/2) is a doppler domain translation variable, fix (·) represents a rounding operation towards zero, and P represents a total number of points of a fast fourier transform;
(4) respectively calculating N groups of frequency peak functions fp of training sample setn(l) And testing the sample setM sets of frequency-peak functions fpm(l) The variance characteristic, the entropy characteristic, the threshold-crossing peak value number characteristic and the position characteristic of the first threshold-crossing peak of the first vector are combined to construct an N multiplied by 4 dimensional training characteristic matrix F1And M4 dimensional test feature matrix F2And for the training feature matrix F1And the test feature matrix F2Respectively carrying out normalization processing to obtain normalized training feature matrixes
Figure BDA0002893080670000022
And normalized test feature matrix
Figure BDA0002893080670000021
(5) The normalized training feature matrix is
Figure BDA0002893080670000023
Inputting the data into a classifier for training to obtain a trained classifier;
(6) testing characteristic matrix after normalization
Figure BDA0002893080670000024
And sending the data into a trained classifier to obtain a classification result.
When the radar carrier frequency changes, the method can ensure that the classification performance of three types of airplane targets cannot be reduced compared with the traditional characteristic extraction method because the method firstly calculates the frequency peak function of the time domain echo signal of the airplane target, extracts the variance characteristic, the entropy characteristic, the threshold-crossing peak value number characteristic and the position characteristic of the first threshold-crossing peak of the frequency peak function and then sends the extracted characteristics to a trained classifier to output the airplane category information.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a flow chart of an implementation of the feature extraction and normalization operations of the present invention;
FIG. 3 is a comparison graph of the performance of the classification of three types of aircraft targets using the method of the present invention and a conventional method.
Detailed Description
Embodiments and effects of the present invention will be described in further detail below with reference to the accompanying drawings.
Referring to fig. 1, the implementation steps of this example are as follows:
step 1: a training sample set and a test sample set are obtained.
(1.1) extracting N groups of time domain echo signals from a radar single carrier frequency echo database of three targets, namely a helicopter, a propeller plane and a jet plane, as a training sample set, wherein N is an integer with the value larger than 900, and the number of the time domain echo signals of the three planes is an integer larger than 300;
and (1.2) extracting M groups of time domain echo signals from three types of target echo signals received by the radar as a test sample set, wherein M is an integer with the value larger than 0.
Step 2: and preprocessing the training sample set and the testing sample set.
(2.1) carrying out Fourier transform on N groups of time domain echo signals of the training sample set and M groups of time domain echo signals of the testing sample set respectively to obtain Doppler spectrums of corresponding echo signals, and taking the obtained Doppler spectrums as a body component range;
(2.2) searching a maximum value in the range of the fuselage component, and recording the amplitude U, the phase theta and the position of the maximum value, wherein the position of the maximum value is the Doppler frequency f;
(2.3) obtaining a maximum reconstructed signal according to the following formula:
Figure BDA0002893080670000031
wherein K is the number of pulse accumulations, exp represents the exponential operation with natural numbers as the base;
(2.4) respectively subtracting the signals reconstructed by the respective maximum values from the N groups of time domain echo signals of the training sample set and the M groups of time domain echo signals of the test sample set to obtain N groups of training sample set time domain echo signals and M groups of test sample set time domain echo signals after the airplane body compensation;
(2.5) performing time domain echo signal summation on the obtained N groups of training sample sets after airplane fuselage compensationFast Fourier transform of P points is respectively carried out on M groups of test sample set time domain echo signals to obtain N groups of Doppler domain echo signals U of the training sample setn(k) And M groups of Doppler domain echo signals U of test sample setm(k) Where N is 1,2,., N is a sample number of the training sample set, N is a total number of samples of the training sample set, M is 1,2,. wherein M is a sample number of the test sample set, and M is a total number of samples of the test sample set.
And step 3: respectively calculating frequency peak function fp of training sample setn(l) And the frequency-peak function fp of the test sample setm(l)。
(3.1) Doppler domain echo signal U to training sample setn(k) Modulus taking and normalization processing are carried out to obtain a modulus taking normalization signal X of the training sample setn(k)：
Figure BDA0002893080670000041
Where k1, 2,., P represents the total number of points of the fast fourier transform;
(3.2) calculating a normalized signal X of the training sample setn(k) Of a cyclic autocorrelation function phin(l) Sum-cycle average amplitude difference function
Figure BDA0002893080670000042
Figure BDA0002893080670000043
Figure BDA0002893080670000044
Wherein, l is 1, 2., fix (P/2) is a doppler domain shift variable, fix (·) is a rounding operation towards zero, mod (k + l, P) is a remainder operation, | · | is an absolute value operation;
(3.3) solving the cyclic autocorrelation function phi of the training sample setn(l) Sum-cycle average amplitude difference function
Figure BDA0002893080670000045
Of the union function Fn(l)：
Figure BDA0002893080670000046
(3.4) Joint function F from training sample setn(l) Constructing a frequency peak function fp of a training sample setn(l)：
Figure BDA0002893080670000047
(3.5) Doppler domain echo signal U for test sample setm(k) Modulus taking and normalization processing are carried out to obtain a modulus taking normalization signal X of the test sample setm(k)：
Figure BDA0002893080670000048
Where k1, 2,., P represents the total number of points of the fast fourier transform;
(3.6) calculating the normalized signal X of the test sample setm(k) Of a cyclic autocorrelation function phim(l) Sum-cycle average amplitude difference function
Figure BDA0002893080670000049
Figure BDA00028930806700000410
Figure BDA0002893080670000051
Wherein, l is 1, 2., fix (P/2) is a doppler domain shift variable, fix (·) is a rounding operation towards zero, mod (k + l, P) is a remainder operation, | · | is an absolute value operation;
(3.7) solving the cyclic autocorrelation function phi of the test sample setm(l) Sum-cycle average amplitude difference function
Figure BDA0002893080670000052
Of the union function Fm(l)：
Figure BDA0002893080670000053
(3.8) Joint function F from test sample setm(l) Constructing a frequency-peak function fp of a test sample setn(l)：
Figure BDA0002893080670000054
And 4, step 4: and respectively extracting the frequency domain correlation characteristics of the training sample set and the test sample set, and respectively carrying out normalization processing on the frequency domain correlation characteristics.
And the frequency domain correlation characteristics comprise a variance characteristic, an entropy characteristic, an over-threshold peak value number characteristic and a position characteristic of a first over-threshold peak of the frequency peak function.
Referring to fig. 2, the following is embodied:
(4.1) calculating a training sample set frequency peak function fpn(l) Variance feature c ofn1：
Figure BDA0002893080670000055
Wherein L is 1,2, and L is a doppler domain translation variable, L is fix (P/2), and P represents the total number of points of the fast fourier transform;
(4.2) calculating a training sample set frequency peak function fpn(l) Entropy characteristic c ofn2：
Figure BDA0002893080670000056
(4.3) computational trainingSample set frequency peak function fpn(l) The number characteristic c of the threshold-crossing peak valuen3：
cn3＝length{fpn(l)≥η1×max(fpn(l))}，
Where max (fp)n(l) To find fpn(l) Peak operation of the first highest peak, η1Is a set threshold parameter, and the value is 0.7; the function of the length {. is to find L1, 2n(l)≥η1×max(fpn(l) Point number of conditions);
(4.4) calculating the frequency-peak function fp of the training sample setn(l) C of the first threshold peak crossingn4：
cn4＝find_first{fp(l)＞η2×max(fp(l))}，
Wherein eta2Is a set threshold parameter, and the value is 0.8; the function find _ first {. is to find L ═ 1,2n(l)≥η2×max(fpn(l) A doppler domain shift variable l for the condition;
(4.5) calculating a test sample set frequency peak function fpm(l) Variance feature c ofm1：
Figure BDA0002893080670000061
Wherein L is 1,2, and L is a doppler domain translation variable, L is fix (P/2), and P represents the total number of points of the fast fourier transform;
(4.6) calculating the frequency-peak function fp of the test sample setm(l) Entropy characteristic c ofm2：
Figure BDA0002893080670000062
(4.7) calculating a test sample set frequency peak function fpm(l) The number characteristic c of the threshold-crossing peak valuem3：
cm3＝length{fpm(l)≥η1×max(fpm(l))}，
Where max (fp)m(l) To find fpm(l) Peak operation of the first highest peak, η1Is a set threshold parameter, and the value is 0.7; the function of the length {. is to find L1, 2m(l)≥η1×max(fpm(l) Point number of conditions);
(4.8) calculating a test sample set frequency peak function fpm(l) C of the first threshold peak crossingm4：
cm4＝find_first{fpm(l)＞η2×max(fpm(l))}，
Wherein eta2Is a set threshold parameter, and the value is 0.8; the function find _ first {. is to find L ═ 1,2m(l)≥η2×max(fpm(l) A doppler domain shift variable l for the condition;
(4.9) constructing a feature vector p of each sample in the training sample setnAnd a feature vector p for each sample in the set of test samplesmRespectively, as follows:
pn＝[cn1,cn2,cn3,cn4]，
pm＝[cm1,cm2,cm3,cm4]，
wherein N is 1,2, and N, N is the total number of samples in the training sample set, and M is 1,2, and M, M is the total number of samples in the test sample set;
(4.10) respectively constructing an N multiplied by 4 dimensional training feature matrix F by the feature vector of each sample in the training sample set and the test sample set obtained in the step (4.9)1And M4 dimensional test feature matrix F2Respectively, as follows:
F1＝[p1；p2；...；pn；...；pN]，
F2＝[p1；p2；...；pm；...；pM]；
(4.11) calculating a training feature matrix F1The mean and standard deviation of the 4 features in (a),the mean vector μ and the standard deviation vector σ were constructed separately and expressed as follows:
μ＝[μ1,μ2,μ3,μ4]，
σ＝[σ1,σ2,σ3,σ4]；
(4.12) computing an N x 4 dimensional training feature matrix F1Each of the characteristic values cniNormalized result of (2)
Figure BDA00028930806700000713
Figure BDA0002893080670000071
Wherein muiIs the i-th component, σ, of the mean vector μiIs the i-th component of the standard deviation vector σ, i 1,2,3,4N 1, 2.
(4.13) use of F1The normalized result of each feature value
Figure BDA0002893080670000072
Constructing a normalized training feature matrix
Figure BDA0002893080670000073
Figure BDA0002893080670000074
Wherein
Figure BDA0002893080670000075
The normalization result of the characteristic vector of the nth sample in the training sample set is N, which is 1, 2.
(4.14) computing the Mx 4 dimensional test feature matrix F2Each of the characteristic values cmiNormalized result of (2)
Figure BDA0002893080670000076
Figure BDA0002893080670000077
Wherein muiIs the i-th component, σ, of the mean vector μiIs the i-th component of the standard deviation vector σ, i 1,2,3,4M 1, 2.
(4.15) use of F2The normalized result of each feature value
Figure BDA0002893080670000078
Constructing a normalized test feature matrix
Figure BDA0002893080670000079
Is represented as follows:
Figure BDA00028930806700000710
wherein
Figure BDA00028930806700000711
For the normalized result of the mth sample feature vector in the test sample set, M is 1, 2.
And 5: the normalized training feature matrix is
Figure BDA00028930806700000712
Inputting the data into a classifier for training to obtain the trained classifier.
The classifier adopts a Support Vector Machine (SVM), and the parameters adopt default parameter values.
Step 6: testing characteristic matrix after normalization
Figure BDA0002893080670000081
And sending the data into a trained classifier to obtain a classification result.
The effect of the present invention will be further described with reference to simulation experiments.
First, simulation experiment conditions
The aircraft rotor parameters are shown in table 1, the pulse repetition frequency PRF is 5KHz, and the dwell time tres100 ms. Gaussian white noise with a signal-to-noise ratio of 15dB was added to the experiment, where the signal-to-noise ratio was defined as the signal-to-noise ratio of the micromotion component to the noise. In a comparison experiment, the frequency domain correlation characteristics of the invention are the variance and entropy of a frequency peak function, the number of threshold peaks and the position 4-dimensional characteristics of a first threshold peak; the frequency domain characteristics of the traditional method are 4-dimensional characteristics of frequency domain second moment, fourth moment, waveform entropy and amplitude variance.
Training a sample set: training radar carrier frequency f0Each model of aircraft generated 300 samples for 10GHz, for a total of 3600 training samples.
Testing a sample set: the test radar carrier frequencies are divided into the following 9 groups, respectively f08GHz, 8.5GHz, 9GHz, 9.5GHz, 10GHz, 10.5GHz, 11GHz, 11.5GHz and 12GHz, with the test groups consistent with the training radar carrier frequency as control groups, the number of test samples for each group being 1200, i.e., 100 samples are generated for each model of aircraft.
TABLE 1 three types of aircraft rotor physical parameters
Target model of airplane Number of blades L1(m) L2(m) Rotating speed (rmin)
Helicopter BK17 4 0 5.5 383
Helicopter rice-17 5 0 10.645 185
Helicopter AS350 3 0 5.345 394
Helicopter bell 212 2 0 7.315 324
Propeller SAAB2000 6 0.28 1.905 950
Propeller L-420 5 0.12 1.15 1650
Propeller L-610G 4 0.23 1.675 1150
Propeller F406 3 0.23 1.18 1690
Jet plane A 30 0.3 1.0 3000
Jet B 38 0.38 1.1 3520
Jet C 27 0.18 0.51 8615
Jet D 33 0.2 0.6 5000
Second, simulation experiment contents
According to the above conditions and parameters, the frequency domain correlation features provided by the present invention and the frequency domain features of the conventional method are extracted from the training sample set by the present invention method and the conventional method, respectively, and then they are input into the classifier for training, and then the frequency domain correlation features and the frequency domain features of 9 groups of test sample sets are extracted, respectively, and then are input into the trained classifier, so as to obtain 9 groups of classification accuracy results of the present invention method and 9 groups of classification accuracy results of the conventional method, and the results are shown in fig. 3.
Third, simulation result analysis
From the aspect of classification accuracy, the frequency domain correlation characteristics of the method are superior to the traditional characteristics, and the classification accuracy is improved by nearly three percent.
From the perspective of radar carrier frequency robustness, when the difference between the test radar carrier frequency and the training radar carrier frequency is not large, the classification performance of the frequency domain correlation characteristic and the classification performance of the traditional characteristic of the invention have no obvious change; when the difference between the test radar carrier frequency and the training radar carrier frequency is large, the classification performance of the frequency domain correlation characteristics of the invention is not obviously reduced, but the classification performance of the traditional characteristics is obviously reduced.
The change is that the frequency domain correlation characteristic reflects the spectral line interval of the frequency domain, the spectral line interval is only related to the rotating speed of the rotor wing, when the radar carrier frequency changes, the frequency-peak function for extracting the characteristic is influenced to a lower degree, the influence on the characteristic dispersion condition is further smaller, and the original characteristic dispersion condition is maintained. Therefore, the radar carrier frequency robust classification is realized based on the frequency domain correlation characteristics, and the problem of the radar carrier frequency mismatch of the training sample and the test sample can be solved under the condition of not retraining the model.

Claims (8)

1. A radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics is characterized by comprising the following steps:
(1) extracting N groups of time domain echo signals from a radar single carrier frequency echo database of three types of targets, namely a helicopter, a propeller plane and a jet plane, as a training sample set, and extracting M groups of time domain echo signals from three types of target echo signals received by the radar as a test sample set;
(2) respectively sequentially setting N groups of time domain echo signals of the training sample set and M groups of time domain echo signals of the testing sample setPerforming airplane fuselage compensation and fast Fourier transform of P points to obtain N groups of Doppler domain echo signals U of a training sample setn(k) And M groups of Doppler domain echo signals U of test sample setm(k) Wherein N is 1,2, the sample sequence number of the training sample set, M is 1,2, the sample sequence number of the test sample set;
(3) calculating N groups of Doppler domain echo signals U of training sample setn(k) Frequency-peak function fpn(l) And M groups of Doppler domain echo signals U of test sample setm(k) Frequency-peak function fpm(l) Wherein N is a training sample set sample number, M is a testing sample set sample number, l is 1,2, and fix (P/2) is a doppler domain translation variable, fix (·) represents a rounding operation towards zero, and P represents a total number of points of a fast fourier transform;
(4) respectively calculating N groups of frequency peak functions fp of training sample setn(l) And M sets of frequency-peak functions fp of the test sample setm(l) The variance characteristic, the entropy characteristic, the threshold-crossing peak value number characteristic and the position characteristic of the first threshold-crossing peak of the first vector are combined to construct an N multiplied by 4 dimensional training characteristic matrix F1And M4 dimensional test feature matrix F2And for the training feature matrix F1And the test feature matrix F2Respectively carrying out normalization processing to obtain normalized training feature matrixes
Figure FDA0002893080660000011
And normalized test feature matrix
Figure FDA0002893080660000012
(5) The normalized training feature matrix is
Figure FDA0002893080660000013
Inputting the data into a classifier for training to obtain a trained classifier;
(6) testing characteristic matrix after normalization
Figure FDA0002893080660000014
And sending the data into a trained classifier to obtain a classification result.
2. The method of claim 1, wherein in (2), the aircraft fuselage compensation is performed on N sets of time domain echo signals in the training sample set and M sets of time domain echo signals in the test sample set by using a CLEAN method, and the steps are as follows:
(2a) respectively carrying out Fourier transform on N groups of time domain echo signals of the training sample set and M groups of time domain echo signals of the testing sample set to obtain Doppler spectrums of corresponding echo signals, and taking the obtained Doppler spectrums as a body component range;
(2b) searching a maximum value in the range of the components of the airplane body, and recording the amplitude U, the phase theta and the position of the maximum value, wherein the position of the maximum value is the Doppler frequency f;
(2c) the maximum reconstructed signal is obtained as follows:
Figure FDA0002893080660000021
wherein K is the number of pulse accumulations, exp represents the exponential operation with natural numbers as the base;
(2d) and respectively subtracting the signals reconstructed by the maximum values from the N groups of time domain echo signals of the training sample set and the M groups of time domain echo signals of the test sample set to obtain N groups of time domain echo signals of the training sample set and M groups of time domain echo signals of the test sample set after the airplane body compensation.
3. The method of claim 1, wherein (3) N sets of Doppler domain echo signals U of the training sample set are calculatedn(k) Frequency-peak function fpn(l) The method comprises the following steps:
(3a) doppler domain echo signal U to training sample setn(k) Modulus taking and normalization processing are carried out to obtain a modulus taking normalization signal X of the training sample setn(k)：
Figure FDA0002893080660000022
Where k1, 2,., P represents the total number of points of the fast fourier transform;
(3b) calculating a modulus normalization signal X of a training sample setn(k) Of a cyclic autocorrelation function phin(l) Sum-cycle average amplitude difference function
Figure FDA0002893080660000023
Figure FDA0002893080660000024
Figure FDA0002893080660000025
Wherein, l is 1, 2., fix (P/2) is a doppler domain shift variable, fix (·) is a rounding operation towards zero, mod (k + l, P) is a remainder operation, | · | is an absolute value operation;
(3c) solving a cyclic autocorrelation function phi of a training sample setn(l) Sum-cycle average amplitude difference function
Figure FDA0002893080660000026
Of the union function Fn(l)：
Figure FDA0002893080660000031
(3d) Joint function F from a set of training samplesn(l) Frequency-peak function fp for constructing training sample setn(l)：
Figure FDA0002893080660000032
4. The method of claim 1, wherein (3) M groups of Doppler domain echo signals U of the test sample set are calculatedm(k) Frequency-peak function fpm(l) The method comprises the following steps:
(3e) doppler domain echo signal U to test sample setm(k) Modulus taking and normalization processing are carried out to obtain a modulus taking normalization signal X of the test sample setm(k)：
Figure FDA0002893080660000033
Where k1, 2,., P represents the total number of points of the fast fourier transform;
(3f) calculating a modulus normalized signal X of a test sample setm(k) Of a cyclic autocorrelation function phim(l) Sum-cycle average amplitude difference function
Figure FDA0002893080660000034
Figure FDA0002893080660000035
Figure FDA0002893080660000036
Wherein, l is 1, 2., fix (P/2) is a doppler domain shift variable, fix (·) is a rounding operation towards zero, mod (k + l, P) is a remainder operation, | · | is an absolute value operation;
(3g) solving a cyclic autocorrelation function phi of a test sample setm(l) Sum-cycle average amplitude difference function
Figure FDA0002893080660000037
Of the union function Fm(l)：
Figure FDA0002893080660000038
(3h) Union function F from test sample setm(l) Frequency-peak function fp for constructing test sample setn(l)：
Figure FDA0002893080660000039
5. The method of claim 1, wherein the N sets of frequency-peak functions fp of the training sample set extracted in (4)n(l) Is characterized by comprising the following steps:
(4a) calculating a training sample set frequency peak function fpn(l) Variance feature c ofn1：
Figure FDA0002893080660000041
Wherein L is 1,2, and L is a doppler domain translation variable, L is fix (P/2), and P represents the total number of points of the fast fourier transform;
(4b) calculating a training sample set frequency peak function fpn(l) Entropy characteristic c ofn2：
Figure FDA0002893080660000042
(4c) Calculating a training sample set frequency peak function fpn(l) The number characteristic c of the threshold-crossing peak valuen3：
cn3＝length{fpn(l)≥η1×max(fpn(l))}，
Where max (fp)n(l) To find fpn(l) Peak operation of the first highest peak, η1Is a set threshold parameter, and the value is 0.7; the function of the length {. is to find L1, 2n(l)≥η1×max(fpn(l) Point number of conditions);
(4d) Calculating a training sample set frequency peak function fpn(l) C of the first threshold peak crossingn4：
cn4＝find_first{fp(l)＞η2×max(fp(l))}，
Wherein eta2Is a set threshold parameter, and the value is 0.8; the function find _ first {. is to find L ═ 1,2n(l)≥η2×max(fpn(l) ) the doppler domain shift variable l of the condition.
6. The method of claim 1, wherein the M sets of frequency-peak functions fp of the test sample set are extracted in (4)m(l) Is characterized by comprising the following steps:
(4e) calculating a test sample set frequency peak function fpm(l) Variance feature c ofm1：
Figure FDA0002893080660000043
Wherein L is 1,2, and L is a doppler domain translation variable, L is fix (P/2), and P represents the total number of points of the fast fourier transform;
(4f) calculating a test sample set frequency peak function fpm(l) Entropy characteristic c ofm2：
Figure FDA0002893080660000051
(4g) Calculating a test sample set frequency peak function fpm(l) The number characteristic c of the threshold-crossing peak valuem3：
cm3＝length{fpm(l)≥η1×max(fpm(l))}，
Where max (fp)m(l) To find fpm(l) Peak operation of the first highest peak, η1Is a set threshold parameter, and the value is 0.7; the function of the length {. is to find L1, 2m(l)≥η1×max(fpm(l) Point number of condition)Counting;
(4h) calculating a test sample set frequency peak function fpm(l) C of the first threshold peak crossingm4：
cm4＝find_first{fpm(l)＞η2×max(fpm(l))}，
Wherein eta2Is a set threshold parameter, and the value is 0.8; the function find _ first {. is to find L ═ 1,2m(l)≥η2×max(fpm(l) ) the doppler domain shift variable l of the condition.
7. The method of claim 1, wherein the N x 4 dimensional training feature matrix F constructed in (4)1And M4 dimensional test feature matrix F2Respectively, as follows:
F1＝[p1；p2；...；pn；...；pN]，
F2＝[p1；p2；...；pm；...；pM]，
wherein p isn＝[cn1,cn2,cn3,cn4]N is the total number of samples in the training sample set, p is the feature vector of the nth sample in the training sample set, N is 1,2m＝[cm1,cm2,cm3,cm4]M is the feature vector of the mth sample in the test sample set, and M is the total number of samples in the test sample set.
8. The method of claim 1, wherein the training feature matrix normalized in (4)
Figure FDA0002893080660000052
And normalized test feature matrix
Figure FDA0002893080660000053
Respectively, as follows:
Figure FDA0002893080660000054
Figure FDA0002893080660000055
wherein
Figure FDA0002893080660000056
Figure FDA0002893080660000057
As a result of normalizing the ith feature value of the nth sample in the training sample set,
Figure FDA0002893080660000058
the calculation formula of (2) is as follows:
Figure FDA0002893080660000061
μ＝[μ1,μ2,μ3,μ4]and σ ═ σ [ σ ]1,σ2,σ3,σ4]Is a training feature matrix F1A vector consisting of the mean values of the 4 medium features and a vector consisting of the standard deviation;
Figure FDA0002893080660000062
Figure FDA0002893080660000063
for the normalized result of the i-th feature value of the m-th sample in the test sample set,
Figure FDA0002893080660000064
the calculation formula is as follows:
Figure FDA0002893080660000065
Figure FDA0002893080660000066
CN202110032916.5A 2021-01-12 2021-01-12 Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics Pending CN112882011A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110032916.5A CN112882011A (en) 2021-01-12 2021-01-12 Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110032916.5A CN112882011A (en) 2021-01-12 2021-01-12 Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics

Publications (1)

Publication Number Publication Date
CN112882011A true CN112882011A (en) 2021-06-01

Family

ID=76044427

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110032916.5A Pending CN112882011A (en) 2021-01-12 2021-01-12 Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics

Country Status (1)

Country Link
CN (1) CN112882011A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113408420A (en) * 2021-06-18 2021-09-17 西安电子科技大学 Spread spectrum signal identification method based on multi-dimensional parameter extraction and support vector machine
CN114070679A (en) * 2021-10-25 2022-02-18 中国电子科技集团公司第二十九研究所 Frequency-phase characteristic analysis method for intelligent pulse classification
CN114065810A (en) * 2021-11-07 2022-02-18 西安电子工程研究所 Spectral peak number-based unmanned aerial vehicle and armed individual soldier classification and identification algorithm

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104077787A (en) * 2014-07-08 2014-10-01 西安电子科技大学 Plane target classification method based on time domain and Doppler domain
CN104239894A (en) * 2014-09-03 2014-12-24 西安电子科技大学 Airplane target classification method based on time domain correlation characteristics
CN106814351A (en) * 2017-01-10 2017-06-09 西安电子科技大学 Aircraft Targets sorting technique based on three rank LPC techniques
CN107506794A (en) * 2017-08-23 2017-12-22 西安电子工程研究所 Ground moving object sorting algorithm based on decision tree
CN109975780A (en) * 2019-04-17 2019-07-05 西安电子工程研究所 Helicopter model recognizer based on pulse Doppler radar time domain echo
EP3722830A1 (en) * 2019-04-12 2020-10-14 Thales Device for identifying a type of aircraft, associated identification method and computer program
CN111812598A (en) * 2020-07-30 2020-10-23 中国电波传播研究所（中国电子科技集团公司第二十二研究所） Time domain and frequency domain multi-feature-based ground and sea clutter classification method

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104077787A (en) * 2014-07-08 2014-10-01 西安电子科技大学 Plane target classification method based on time domain and Doppler domain
CN104239894A (en) * 2014-09-03 2014-12-24 西安电子科技大学 Airplane target classification method based on time domain correlation characteristics
CN106814351A (en) * 2017-01-10 2017-06-09 西安电子科技大学 Aircraft Targets sorting technique based on three rank LPC techniques
CN107506794A (en) * 2017-08-23 2017-12-22 西安电子工程研究所 Ground moving object sorting algorithm based on decision tree
EP3722830A1 (en) * 2019-04-12 2020-10-14 Thales Device for identifying a type of aircraft, associated identification method and computer program
CN109975780A (en) * 2019-04-17 2019-07-05 西安电子工程研究所 Helicopter model recognizer based on pulse Doppler radar time domain echo
CN111812598A (en) * 2020-07-30 2020-10-23 中国电波传播研究所（中国电子科技集团公司第二十二研究所） Time domain and frequency domain multi-feature-based ground and sea clutter classification method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
LAN DU 等: "Noise Reduction Method Based on Principal Component Analysis With Beta Process for Micro-Doppler Radar Signatures", 《 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING》 *
LAN DU 等: "Noise Robust Radar HRRP Target Recognition Based on Scatterer Matching Algorithm", 《IEEE SENSORS JOURNAL 》 *
戴永鹏: "基于微多普勒特性的警戒雷达目标分类技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *
赖陈潇: "基于窄带雷达数据的飞机目标分类方法", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113408420A (en) * 2021-06-18 2021-09-17 西安电子科技大学 Spread spectrum signal identification method based on multi-dimensional parameter extraction and support vector machine
CN113408420B (en) * 2021-06-18 2022-12-06 西安电子科技大学 Spread spectrum signal identification method based on multi-dimensional parameter extraction and support vector machine
CN114070679A (en) * 2021-10-25 2022-02-18 中国电子科技集团公司第二十九研究所 Frequency-phase characteristic analysis method for intelligent pulse classification
CN114065810A (en) * 2021-11-07 2022-02-18 西安电子工程研究所 Spectral peak number-based unmanned aerial vehicle and armed individual soldier classification and identification algorithm
CN114065810B (en) * 2021-11-07 2024-05-03 西安电子工程研究所 Unmanned aerial vehicle and armed individual soldier classification and identification method based on number of spectral peaks

Similar Documents

Publication Publication Date Title
CN112882011A (en) Radar carrier frequency variation robust target identification method based on frequency domain correlation characteristics
CN107728142B (en) Radar high-resolution range profile target identification method based on two-dimensional convolutional network
Du et al. Robust classification scheme for airplane targets with low resolution radar based on EMD-CLEAN feature extraction method
CN107728143B (en) Radar high-resolution range profile target identification method based on one-dimensional convolutional neural network
Chen et al. Target classification with low-resolution radar based on dispersion situations of eigenvalue spectra
CN107390193B (en) Frequency modulation continuous wave radar aircraft target classification method based on multi-distance unit fusion
CN104077787A (en) Plane target classification method based on time domain and Doppler domain
CN108594195A (en) Low repetition frequency modulated continuous wave radar Aircraft Targets sorting technique based on segmentation mixing
CN109164429B (en) Multi-fractal characteristic aircraft target classification method based on principal component analysis
CN112835003B (en) Radar repetition frequency variation steady target recognition method based on resampling preprocessing
CN112666533B (en) Repetition frequency change steady target identification method based on spatial pyramid pooling network
Li et al. Research on analysis of aircraft echo characteristics and classification of targets in low-resolution radars based on EEMD
CN107192993B (en) Frequency modulation continuous wave radar aircraft target classification method based on image entropy characteristics
CN106483513B (en) A kind of aircraft class target micro-doppler texture characteristic extracting method
Jiang et al. Specific radar emitter identification based on a digital channelized receiver
Zhu et al. Radar HRRP group-target recognition based on combined methods in the backgroud of sea clutter
CN112784916B (en) Air target micro-motion parameter real-time extraction method based on multitask convolutional network
Li et al. Classification of aircraft targets with surveillance radars based on fuzzy fractal features
CN112731331B (en) Micro-motion target noise steady identification method based on signal-to-noise ratio adaptive network
Dong Hybrid electromagnetic signal feature extraction
CN114662527A (en) Repetition frequency parameter change robust classification method based on repetition frequency correlation characteristic extraction
Luo et al. Automatic target recognition of radar HRRP based on high order central moments features
Ren et al. Radar emitter signals classification using kernel principle component analysis and fuzzy support vector machines
Zhang et al. Denoising of UAV micro-Doppler signatures based on BEMD and wavelet transform
CN113311407B (en) Air plane target classification online learning method based on support vector machine

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210601