PRI jitter radar signal sorting method based on rattan moss algorithm, program, equipment and storage medium

Abstract

The invention belongs to the technical field of radar signal sorting, and particularly relates to a PRI jitter radar signal sorting method based on a moss algorithm, a program, equipment and a storage medium. According to the method, TOA difference distribution is reconstructed through an overlapped PRI box structure, and the adaptability of the algorithm is improved in combination with a dynamic time starting point calibration thought. According to the method, a moss algorithm is designed, Chebyshev mapping is adopted to construct population initialization position distribution, a Levy flight strategy is introduced to change a spore propagation search mode, an environment factor adaptive search mechanism is integrated, falling into a local optimal solution is avoided, the global search capability is improved, and the parameter space coverage rate is increased; therefore, PRI jitter radar signal sorting can be realized in an unknown environment. According to the method, the overlapping rate, the observation time adjustable coefficient, the sub-harmonic elimination adjustable coefficient and the noise elimination adjustable coefficient of the PRI box in the sorting process are obtained in a self-adaptive mode, so that the generalization and accuracy of sorting are improved, and PRI jitter radar signal sorting is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal sorting, and particularly relates to a method, program, device and storage medium for sorting PRI jitter radar signals based on the vine moss algorithm. Background Technique

[0002] With the development and evolution of radars, the electromagnetic environment has become more complex, and the requirements for radar signal sorting technology have also been continuously improved. Existing methods have problems such as low sorting accuracy and insufficient generalization. In radar signal sorting, the pulse sorting of jitter radar signals is a relatively complex problem. Therefore, how to accurately sort PRI jitter radar signals in an unknown environment has become crucial.

[0003] Existing methods for sorting PRI jitter radar signals include the dynamic extended association method, CDIF, SDIF, PRI transformation method, etc. When pulses overlap, these methods are prone to detecting harmonic components and generating a large number of false radiation sources, and setting parameters depends on prior knowledge, resulting in low accuracy and insufficient generalization.

[0004] Through the retrieval of existing technical literature, it is found that Xu Minliang et al. in the invention (patent number: CN202011643542.2) invented a "method for sorting jitter signals based on CDIF", which achieved the sorting of jitter radar signals through region division and variable histogram resolution. However, this method needs to specify the detection repetition period interval range for region division, and has poor sorting ability for pulse sequences with a large PRI jitter range. Some parameter settings need to be obtained according to experience, resulting in insufficient generalization. Zhang Chunjie et al. in the invention (patent number: CN202011072664.0) invented a "method for sorting PRI jitter radar signals in the case of pulse loss and aliasing", which judged the jitter rate of jitter signals by improving the PRI overlap box structure and utilized the correlation confidence to achieve the search and extraction of PRI jitter radar signals in the case of pulse loss and aliasing. However, some parameter settings in this method need to be obtained according to experience, resulting in insufficient generalization and affecting the accuracy of sorting results. When sorting PRI jitter radar signals in the case of pulse overlap, combining intelligent algorithms to automatically adjust parameters according to the signal environment and sorting results can effectively improve the generalization and accuracy of sorting. Summary of the Invention

[0005] The purpose of the present invention is to provide a method, program, device and storage medium for sorting PRI jitter radar signals based on the vine moss algorithm, aiming at the problem that existing sorting methods have insufficient generalization ability and poor sorting accuracy when sorting PRI jitter radar signal pulses in the case of pulse overlap because they cannot adaptively adjust parameters.

[0006] A PRI jitter radar signal sorting method based on a vine moss algorithm comprises the following steps:

[0007] Step 1: Receive the radar signal within the sampling time and convert it into an ordered pulse sequence according to the order of pulse arrival time;

[0008] Step 2: Pre-sort the pulse sequence to obtain the preliminary clustering result of the radar signal;

[0009] Step 3: For each pulse sequence in the preliminary clustering results of the radar signal, determine the minimum value of the PRI value range, the maximum value of the PRI value range and the number of PRI boxes; use the vine moss algorithm to take the overlap rate of the PRI overlap boxes in the pulse main sorting, the adjustable coefficient of observation time, the adjustable coefficient of eliminating subharmonics and the adjustable coefficient of eliminating noise as the optimization target, use Chebyshev mapping to construct the population initialization position distribution, introduce the Levy flight strategy to change the spore reproduction search method, and integrate the environmental factor adaptation search mechanism to enhance the global search capability and obtain the optimal parameter combination;

[0010] Step 4: According to the optimal parameter combination corresponding to each pulse sequence in the preliminary clustering results of radar signals, pulse primary sorting is performed on the preliminary clustering results of radar signals, the PRI value retrieval range of each pulse sequence in the preliminary clustering results of radar signals is determined, the corresponding pulse description words are removed from the preliminary clustering results of radar signals, the pulse sequence of each radar is separated, and radar signal sorting is completed.

[0011] Furthermore, in step 1, the radar signal is converted into an ordered pulse sequence pulse_list as follows:

[0012] pulse_list={PDW1,PDW2,...,PDW i ,...,PDW n}

[0013] Among them, PDW i is the ith pulse description word of the pulse sequence pulse_list, i=1,2,...,n, n is the total number of pulses in the pulse sequence; PDW i ={TOA i ,PW i ,BW i ,CF i ,PA i}, TOA i is the pulse arrival time of the ith pulse description word, PW i is the pulse width of the ith pulse description word, BW i is the bandwidth of the ith pulse description word, CF i is the carrier frequency of the ith pulse description word, PAi is the pulse amplitude of the i-th pulse descriptor word;

[0014] The pre-sorting process of the pulse sequence in step 2 is specifically as follows:

[0015] Step 2.1: Calculate the mean values of PW i , BW i and CF i for all pulse descriptor words;

[0016]

[0017] Step 2.2: Calculate the standard deviations of PW i , BW i and CF i for all pulse descriptor words;

[0018]

[0019]

[0020] Step 2.3: Calculate the standard scores of PW i , BW i and CF i for each pulse descriptor word;

[0021]

[0022] Step 2.4: Calculate the normalization results of PW i , BW i and CF i for each pulse descriptor word:

[0023]

[0024] Step 2.5: Cluster the normalization results PW_uni, BW_uni, and CF_uni through clustering, and then cluster the pulse descriptor words PDW in the pulse sequence pulse_list i to obtain the preliminary clustering result pre_res of the radar signal:

[0025] pre_res = {pulse_list1, pulse_list2,..., pulse_list ii ,..., pulse_list f}

[0026] where, pulse_list iiis the \(i\)-th pulse sequence in the preliminary clustering result of radar signals, where \(i = 1,\cdots,f\), and \(f\) is the total number of pulse sequences in the preliminary clustering result of radar signals; the set of reception times of each pulse sequence in the preliminary clustering result of radar signals is \(T\), \(T=\{T_1,T_2,\cdots,T ii ,\cdots,T f \}, and \(T ii is the reception time of the \(i\)-th pulse sequence in the preliminary clustering result of radar signals.

[0027] Furthermore, the specific method of constructing the initial position distribution of the population by using the Chebyshev mapping in step 3 is as follows:

[0028]

[0029] where \(X u (1)\) is the initial position of the \(u\)-th vine moss, \(u = 1,\cdots,NUM\), and \(NUM\) is the number of vine mosses in the population; the position of each vine moss represents a parameter combination of the overlapping rate \(e\) of a group of PRI overlapping boxes, the adjustable coefficient \(\alpha\) of the observation time, the adjustable coefficient \(\beta\) for eliminating sub-harmonics, and the adjustable coefficient \(\gamma\) for eliminating noise; \(R1\) is a random number in \([0,1]\); \(l_x\) is the lower bound of the position parameter range of the vine moss group, \(u_x\) is the upper bound of the position parameter range of the vine moss group, and \(h\) is a constant parameter.

[0030] Furthermore, the Levy flight strategy is introduced in step 3 to change the spore reproduction search method, specifically as follows:

[0031] Vine mosses conduct spore reproduction search by dispersing spores with the wind. Due to different wind speeds, the search mechanisms of vine mosses are different:

[0032]

[0033] where is the position of the new vine moss after spore propagation of the \(u\)-th vine moss in the \(Ite\)-th iteration; \(K\) is a constant parameter; \(R2\) is a random number in \([0,1]\), and \(R3\) is a random number in \([0,1]\); \(WF\) is the wind intensity, MI is the maximum number of iterations; \(L\) is the spore propagation distance coefficient under the condition of gentle wind; \(B\) is the expansion parameter of the Levy distribution step size; \(z\) is a standard normal distribution random variable; \(\mu\) is the parameter of the Levy distribution; \(X best is the position of the global optimal moss individual.

[0034] Furthermore, the specific steps in step 3 are as follows:

[0035] Step 3.1: Set the parameters of the vine moss algorithm, including the number \(num\) of vine mosses in the population and the maximum number of iterations \(MI\); initialize the current iteration number \(Ite = 1\), and construct the initial position distribution \(X\) of the population by using the Chebyshev mappingu (1); Calculate the fitness of each vine moss according to its initial position X u (1) Select the position of the vine moss with the minimum corresponding fitness value as the global optimal position X best ;

[0036] Step 3.2: Randomly take the v-th dimensional value X best as the threshold, and compare it with the v-th dimensional value X best,v of all vine moss positions X u (Ite). If X u,v (Ite) > X u,v (Ite), then add X best,v (Ite) to the set area_m(Ite); if X u (Ite) ≤ X u,v (Ite), then add X best,v (Ite) to the set area_l(Ite); u (Ite);

[0037] Step 3.3: Compare the number of vine mosses in area_m(Ite) and area_l(Ite), and select the set with more vine mosses as area(Ite);

[0038] Step 3.4: Calculate the wind direction WD;

[0039]

[0040] where sum(area(Ite)) is the number of all vine mosses in the set area(Ite);

[0041] Step 3.5: Introduce the Levy flight strategy to change the spore reproduction search method, and obtain the positions of the new vine mosses after spore propagation of each vine moss The spore propagation distance coefficient L under wind conditions is:

[0042]

[0043] Step 3.6: Incorporate the environmental factor adaptation search mechanism to enhance the global search ability, and judge whether the environmental factor EV(Ite) meets the sexual reproduction search conditions of vine moss. If the condition EV(Ite) ≥ 0.5 is met, sexual reproduction search is carried out; otherwise, vegetative reproduction search is carried out to obtain the updated position X u (Ite + 1);

[0044]

[0045] Among them, the environmental factor EV(Ite) is a random number between [0, 1]; both rand and N are 1×4 random vectors, each element in rand is a random number between [0, 1], and each element in N is a random number with a standard Gaussian distribution; ⊙ is an operator for finding the product of two vector elements; R4 is a random number between [0, 1]; Q is a propagation factor, ; H is a constant parameter;

[0046] Step 3.7: Calculate the fitness of each vine moss according to the updated position X u (Ite + 1). If there is a vine moss with an updated position X u (Ite + 1) whose fitness is less than the fitness of the global optimal position X best , then update the global optimal position X best ;

[0047] Step 3.8: If Ite < MI, then set Ite = Ite + 1 and return to Step 3.2; otherwise, output the global optimal position X best corresponding parameter combination.

[0048] Furthermore, in Step 3, for the ii-th pulse sequence pulse_list in the preliminary clustering result of radar signals ii , determine the minimum value τ min (ii) of the PRI value range, the maximum value τ max (ii) of the PRI value range, and the number F(ii) of PRI bins, and τ min (ii) is not greater than the moment corresponding to the start point of pulse_list ii , and τ max (ii) is not less than the moment corresponding to the end point of pulse_list ii ; then the bin width wid of the PRI bin is: The central coordinate τ x of the x-th PRI bin is: τ x = (x - 0.5)·wid + τ min , and the overlapping width ov_wid x of the x-th PRI bin is: ov_wid x = e·τ x ; the range [C 1x , C 2x of the x-th PRI bin is:

[0049] When obtaining the optimal parameter combination through the vine moss algorithm, the method for calculating the fitness according to the position X u of the vine moss is specifically as follows:

[0050] Step 4.1: According to the position Xu ={e, α, β, γ}, obtain the overlapping rate e of the PRI overlapping box, the adjustable coefficient α of the observation time, the adjustable coefficient β for eliminating sub-harmonics, and the adjustable coefficient γ for eliminating noise;

[0051] Step 4.2: Initialize the cumulative spectral values G of all PRI boxes = {G1, G2,..., G x ,..., G F} = {0, 0,..., 0,..., 0}, the time starting points O of all PRI boxes = {O1, O2,..., O x ,..., O F} = {0, 0,..., 0,..., 0}, the usage times flag of all PRI boxes = {flag1, flag2,..., flag x ,..., flag F} = {0, 0,..., 0,..., 0}; Initialize j = 2 and i = 1;

[0052] Step 4.3: Calculate the arrival time difference ε between the i-th pulse descriptor and the j-th pulse descriptor;

[0053] ε = TOA j - TOA i

[0054] Step 4.4: If ε ≥ τ max , then let j = j + 1 and execute Step 4.5; if ε ≤ τ min , then let i = i - 1 and execute Step 4.6; if τ min < ε < τ max , then execute Step 4.7;

[0055] Step 4.5: If j > num(pulse_list ii ), then execute Step 4.8; if j ≤ num(pulse_list ii ), then let i = j - 1 and return to Step 4.3; num(pulse_list ii ) is the number of pulses in the ii-th pulse sequence in the pre-sorting result;

[0056] Step 4.6: If i > 1, then return to Step 4.3; if i ≤ 1, then let j = j + 1 and return to Step 4.5;

[0057] Step 4.7: Determine the PRI box range [C 1x , C 2x where ε falls, and obtain the corresponding PRI box index x; Since [C 1x , C 2x includes the overlapping parts of other PRI boxes, for [C1x , C 2x Update the cumulative spectral value G for all PRI bins within the range x After that, let i = i - 1, and return to step 4.6;

[0058] Update the cumulative spectral value G x The method is as follows:

[0059] Step 4.7.1: If the usage count flag 1x , C 2x of a certain PRI bin within the range x = 0, then let flag x = flag x + 1, O x = TOA j ;

[0060] Step 4.7.2: For all PRI bins within the range 1x , C 2x calculate their initial phase η0, decomposition parameters U and W;

[0061] U = η0 + 0.5,

[0062] Step 4.7.3: If a certain PRI bin within the range 1x , C 2x satisfies U = 1 and O x = TOA i or U ≥ 2 and |W| ≤ ξ, then update its time start point to O x = TOA j ; ξ is a constant coefficient for determining the time start point;

[0063] Step 4.7.4: Update the cumulative spectral value G 1x , C 2x for all PRI bins within the range x ;

[0064] G x = G x + D x

[0065] where,

[0066] Step 4.8: Calculate the pulse flow density ρ ii ;

[0067]

[0068] Step 4.9: For each PRI bin, calculate the detection threshold A x ; Plot Gx —τ x Discrete points, and connect each discrete point to form curve G x (t); Plot A x —τ x Discrete points, and connect each discrete point to form curve A x (t); If the hump peak G x in curve G x (τ x ) > A x (τ x ), then obtain the abscissa of the hump peak, that is, the τ x corresponding to the hump peak G x (τ x );

[0069]

[0070] Step 4.10: Calculate the position X u of the vine moss and the corresponding fitness fit;

[0071]

[0072] where s is the total number of abscissas of the obtained hump peaks; c q is the abscissa value corresponding to the position of the q-th hump peak; a q is the abscissa value corresponding to the intersection of curve G x (t) and curve A x (t) on the left side of the q-th hump; b q is the abscissa value corresponding to the intersection of curve G x (t) and curve A x (t) on the right side of the q-th hump.

[0073] Furthermore, according to the optimal parameter combination corresponding to each pulse sequence pulse_list ii in the preliminary clustering result pre_res of the radar signal, according to Steps 4.1 to 4.9, use the horizontal axis coordinates of the detected hump peaks as the PRI estimation value, set the deviation range ψ above and below the PRI estimation value as the PRI value retrieval range, and remove the corresponding pulse descriptor words from the pulse sequence pulse_list ii ; For each pulse sequence pulse_list ii in the preliminary clustering result pre_res of the radar signal, perform the above operations, separate the pulse sequences of each radar, distinguish different radar signals, and complete the sorting of radar signals.

[0074] A computer device / system / apparatus, comprising a memory, a processor, and a computer program stored on the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned PRI jitter radar signal sorting method based on the vine moss algorithm.

[0075] A computer-readable storage medium, on which a computer program / instructions are stored, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned PRI jitter radar signal sorting method based on the vine moss algorithm are implemented.

[0076] A computer program product, comprising a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned PRI jitter radar signal sorting method based on the vine moss algorithm are implemented.

[0077] The beneficial effects of the present invention are as follows:

[0078] The present invention reconstructs the TOA difference distribution through an overlapping PRI bin structure, combines the idea of dynamic time origin calibration, and improves the adaptability of the algorithm. The present invention designs a vine moss algorithm, constructs the population initialization position distribution by using Chebyshev mapping, introduces the Levy flight strategy to change the spore reproduction search method, integrates the environmental factor adaptation search mechanism, avoids falling into local optimal solutions, improves the global search ability, and enhances the parameter space coverage rate, enabling the present method to realize the sorting of PRI jitter radar signals in an unknown environment. The present invention adaptively obtains the overlapping rate of PRI bins, the adjustable coefficient of the observation time, the adjustable coefficient for eliminating sub-harmonics, and the adjustable coefficient for eliminating noise during the sorting process, thereby enhancing the generalization and accuracy of sorting and realizing the sorting of PRI jitter radar signals. Description of the Drawings

[0079] Figure 1 is the overall flow schematic diagram of the present invention.

[0080] Figure 2 is the schematic diagram of the vine moss algorithm in the present invention. Detailed Embodiments

[0081] The following further describes the present invention with reference to the drawings.

[0082] The present invention designs a method for sorting PRI jitter radar signals based on the vine moss algorithm, mainly to solve the problem that in the case of pulse overlap in the existing sorting methods, when sorting the pulses of PRI jitter radar signals, the generalization ability is insufficient and the sorting accuracy is poor due to the inability to adaptively adjust parameters. The present invention uses the Chebyshev mapping to complete the initialization position distribution of the population, combines the Lévy flight strategy, changes the reproduction search method of vine moss spores, incorporates the environmental factor adaptation search mechanism, enhances the global search ability, and adaptively obtains the overlap rate of the PRI bins, the adjustable coefficient of the observation time, the adjustable coefficient for eliminating sub-harmonics, and the adjustable coefficient for eliminating noise during the sorting process, thereby improving the generalization and accuracy of sorting and realizing the sorting of PRI jitter radar signals.

[0083] As shown in the attached Figure 1 figure, it is a schematic flow diagram of the method for sorting PRI jitter radar signals based on the vine moss algorithm of the present invention.

[0084] A method for sorting PRI jitter radar signals based on the vine moss algorithm includes the following steps:

[0085] Step 1: Receive the radar signals within the sampling time, and convert the radar signals into an ordered pulse sequence pulse_list in the order of the arrival time of the pulses;

[0086] pulse_list = {PDW1, PDW2,..., PDW i ,..., PDW n}

[0087] where PDW i is the i-th pulse descriptor of the pulse sequence pulse_list, i = 1, 2,..., n, and n is the total number of pulses in the pulse sequence; PDW i = {TOA i , PW i , BW i , CF i , PA i}, TOA i is the pulse arrival time of the i-th pulse descriptor, PW i is the pulse width of the i-th pulse descriptor, BW i is the bandwidth of the i-th pulse descriptor, CF i is the carrier frequency of the i-th pulse descriptor, and PA i is the pulse amplitude of the i-th pulse descriptor;

[0088] Step 2: Perform pre-sorting processing on the pulse sequence to obtain the preliminary clustering result of the radar signals;

[0089] Step 2.1: Calculate the PW i , BWi and the mean value of CF i ;

[0090]

[0091] Step 2.2: Calculate the PW, BW i , BW i and the standard deviation of CF i ;

[0092]

[0093] Step 2.3: Calculate the standard score of the PW, BW i , BW i and CF i for each pulse descriptor;

[0094]

[0095] Step 2.4: Calculate the normalization results of the PW, BW i , BW i and CF i for each pulse descriptor:

[0096]

[0097] Step 2.5: Cluster the normalization results PW_uni, BW_uni, and CF_uni through clustering, and cluster the pulse descriptors PDW in the pulse sequence pulse_list according to the clustering results to obtain the preliminary clustering result pre_res of the radar signal: i pre_res = {pulse_list1, pulse_list2,..., pulse_list

[0098] pre_res = {pulse_list1, pulse_list2,..., pulse_list ii ,..., pulse_list f}

[0099] where pulse_list ii is the ii-th pulse sequence in the preliminary clustering result of the radar signal, ii = 1,..., f, and f is the total number of pulse sequences in the preliminary clustering result of the radar signal; the set of reception times of each pulse sequence in the preliminary clustering result of the radar signal is T, T = {T1, T2,..., T ii ,..., T f}, and T ii is the reception time of the ii-th pulse sequence in the preliminary clustering result of the radar signal;

[0100] Step 3: For each pulse sequence in the preliminary clustering results of the radar signal, determine the minimum value of the PRI value range, the maximum value of the PRI value range and the number of PRI boxes; use the vine moss algorithm to take the overlap rate of the PRI overlap boxes in the pulse main sorting, the adjustable coefficient of observation time, the adjustable coefficient of eliminating subharmonics and the adjustable coefficient of eliminating noise as the optimization target, use Chebyshev mapping to construct the population initialization position distribution, introduce the Levy flight strategy to change the spore reproduction search method, and integrate the environmental factor adaptation search mechanism to enhance the global search capability and obtain the optimal parameter combination;

[0101] Step 3.1: Set the parameters of the vine moss algorithm, including the number of vine mosses in the population num and the maximum number of iterations MI; initialize the current number of iterations Ite = 1, and use Chebyshev mapping to construct the population initialization position distribution X u (1); According to the initial position X of each vine moss u (1) Calculate the fitness of each vine moss and select the position of the vine moss with the smallest fitness value as the global optimal position X best ;

[0102]

[0103] Among them, X u (1) is the initial position of the u-th vine moss, u = 1, ..., NUM, NUM is the number of vine mosses in the population; the position of each vine moss represents a parameter combination of the overlap rate e of a set of PRI overlap boxes, the adjustable coefficient of observation time α, the adjustable coefficient of subharmonic elimination β and the adjustable coefficient of noise elimination γ; R1 is a random number in [0,1]; l_x is the lower limit of the vine moss group position parameter range, u_x is the upper limit of the vine moss group position parameter range, and h is a constant parameter;

[0104] In one embodiment of the present invention, h may be set to 4, the number of groups NUM may be set to 30, the population dimension dim may be set to 3, and the maximum number of iterations MI may be set to 500.

[0105] Step 3.2: From X best Randomly take the vth dimension value X best,v As a threshold, compare it with all vine moss positions X u (Ite) the vth dimension value X u,v (Ite) comparison, if X u,v (Ite)>X best,v (Ite), then X u (Ite) is added to the set area_m(Ite); if X u,v (Ite)≤X best,v (Ite), then X u (Ite) joins the set area_l(Ite);

[0106] Step 3.3: Compare the number of vine mosses in area_m(Ite) and area_l(Ite), and select the set with more vine mosses as area(Ite);

[0107] Step 3.4: Calculate the wind direction WD;

[0108]

[0109] Among them, sum(area(Ite)) is the number of all vine mosses in the set area(Ite);

[0110] Step 3.5: Introduce the Levy flight strategy to change the spore reproduction search method, and obtain the positions of the new vine mosses after the spores of each vine moss are spread. The spore propagation distance coefficient L under the condition of gentle wind is:

[0111] Vine mosses conduct spore reproduction search by spore dispersion with the wind. Due to different wind speeds, the search mechanisms of vine mosses are different:

[0112]

[0113] Among them, is the position of the new vine moss after the spores of the u-th vine moss are spread in the Ite-th iteration; K is a constant parameter; R2 is a random number in [0, 1], and R3 is a random number in [0, 1]; WF is the wind intensity, MI is the maximum number of iterations; B is the expansion parameter of the Levy distribution step size; z is a standard normal distribution random variable; μ is the parameter of the Levy distribution; X best is the position of the globally optimal moss individual;

[0114] In an embodiment of the present invention, K can be set to 3.

[0115] L is the spore propagation distance coefficient under the condition of gentle wind:

[0116]

[0117] Step 3.6: Incorporate the environmental factor adaptation search mechanism to enhance the global search ability, and judge whether the environmental factor EV(Ite) meets the sexual reproduction search condition of vine moss. If the condition EV(Ite)≥0.5 is met, sexual reproduction search is carried out; otherwise, vegetative reproduction search is carried out to obtain the updated position X u (Ite + 1);

[0118]

[0119] Among them, the environmental factor EV(Ite) is a random number between [0, 1]; both rand and N are 1×4 random vectors, each element in rand is a random number between [0, 1], and each element in N is a random number with a standard Gaussian distribution; ⊙ is an operator for finding the product of two vector elements; R4 is a random number between [0, 1]; Q is a propagation factor, ; H is a constant parameter;

[0120] In an embodiment of the present invention, H can be set to 2.

[0121] Step 3.7: Calculate the fitness of each vine moss according to the updated position X u (Ite + 1) of each vine moss. If there is a vine moss with an updated position X u (Ite + 1) whose fitness is less than the fitness of the global optimal position X best , then update the global optimal position X best ;

[0122] Step 3.8: If Ite < MI, then set Ite = Ite + 1 and return to Step 3.2; otherwise, output the global optimal position X best The corresponding parameter combination;

[0123] Step 4: According to the optimal parameter combination corresponding to each pulse sequence in the preliminary clustering result of radar signals, perform pulse primary sorting on the preliminary clustering result of radar signals, determine the PRI value retrieval range of each pulse sequence in the preliminary clustering result of radar signals, remove the corresponding pulse descriptor from the preliminary clustering result of radar signals, and separate the pulse sequences of each radar to complete radar signal sorting;

[0124] For the ii-th pulse sequence pulse_list in the preliminary clustering result of radar signals ii , determine the minimum value τ min (ii) of the PRI value range, the maximum value τ max (ii) of the PRI value range, and the number F(ii) of PRI bins, and τ min (ii) is not greater than the moment corresponding to the start point of pulse_list ii , and τ max (ii) is not less than the moment corresponding to the end point of pulse_list ii ; then the bin width wid of the PRI bin is: The central coordinate τ x of the x-th PRI bin is: τ x = (x - 0.5)·wid + τ min , and the overlapping width ov_wid x of the x-th PRI bin is: ov_wid x = e·τx ; The range of the x-th PRI bin is [C 1x , C 2x is:

[0125] Step 4.1: According to the position X of the rattan moss u = {e, α, β, γ}, obtain the overlapping rate e of the PRI overlapping bin, the adjustable coefficient α of the observation time, the adjustable coefficient β of eliminating sub-harmonics, and the adjustable coefficient γ of eliminating noise;

[0126] Step 4.2: Initialize the cumulative spectral values G of all PRI bins = {G1, G2,..., G x ,..., G F} = {0, 0,..., 0,..., 0}, the time starting points O of all PRI bins = {O1, O2,..., O x ,..., O F} = {0, 0,..., 0,..., 0}, the usage times flag of all PRI bins = {flag1, flag2,..., flag x ,..., flag F} = {0, 0,..., 0,..., 0}; Initialize j = 2 and i = 1;

[0127] Step 4.3: Calculate the arrival time difference ε between the i-th pulse descriptor and the j-th pulse descriptor;

[0128] ε = TOA j - TOA i

[0129] Step 4.4: If ε ≥ τ max , then let j = j + 1 and execute Step 4.5; if ε ≤ τ min , then let i = i - 1 and execute Step 4.6; if τ min < ε < τ max , then execute Step 4.7;

[0130] Step 4.5: If j > num(pulse_list ii ), then execute Step 4.8; if j ≤ num(pulse_list ii ), then let i = j - 1 and return to Step 4.3; num(pulse_list ii ) is the number of pulses in the ii-th pulse sequence in the pre-sorting result;

[0131] Step 4.6: If i > 1, then return to Step 4.3; if i ≤ 1, then let j = j + 1 and return to Step 4.5;

[0132] Step 4.7: Determine the PRI bin range [C 1x , C 2x , and obtain the corresponding PRI bin index x; Since [C 1x , C 2x includes the overlapping parts of other PRI bins, update the cumulative spectrum value G 1x , C 2x for all PRI bins within the range, then set i = i - 1, and return to Step 4.6; x

[0133] The method for updating the cumulative spectrum value G x is as follows:

[0134] Step 4.7.1: If the usage count flag 1x , C 2x of a certain PRI bin within the range is 0, then set flag x = flag x + 1, and O x = TOA x ; j

[0135] Step 4.7.2: For all PRI bins within the range [C 1x , C 2x , calculate their initial phase η0, decomposition parameters U and W;

[0136] U = η0 + 0.5,

[0137] Step 4.7.3: If a certain PRI bin within the range [C 1x , C 2x satisfies U = 1 and O x = TOA i or U ≥ 2 and |W| ≤ ξ, then update its time start point to O x = TOA j ; ξ is a constant coefficient for determining the time start point;

[0138] In an embodiment of the present invention, ξ can be taken as 0.02.

[0139] Step 4.7.4: Update the cumulative spectrum value G 1x , C 2x for all PRI bins within the range; x

[0140] G x = G x + D x

[0141] Where,

[0142] Step 4.8: Calculate the pulse current density ρ ii ;

[0143]

[0144] Step 4.9: For each PRI bin, calculate the detection threshold A x ; Draw G x —τ x Discrete points and connect them to form a curve G x (t); draw A x —τ x Discrete points and connect them to form curve A x (t); if the curve G x The hump peak G in (t) x (τ x )>A x (τ x ), then obtain the horizontal coordinate of the hump peak value, that is, the hump peak value G x (τ x ) corresponding to τ x ;

[0145]

[0146] According to the radar signal preliminary clustering result pre_res each pulse sequence pulse_list ii The corresponding optimal parameter combination, according to step 4.1 to step 4.9, takes the horizontal axis coordinate of the detected hump peak as the PRI estimated value, sets the upper and lower deviation range ψ of the PRI estimated value as the PRI value retrieval range, and selects the pulse sequence pulse_list from the pulse sequence pulse_list. ii Remove the corresponding pulse description word; for each pulse sequence pulse_list in the radar signal preliminary clustering result pre_res ii , perform the above operations, separate the pulse sequence of each radar, distinguish different radar signals, and complete radar signal sorting.

[0147] In one embodiment of the present invention, ψ may be set to 0.2.

[0148] In the vine moss algorithm, according to the position X of the vine moss u The specific method for calculating the fitness fit is: first execute steps 4.1 to 4.9, and then calculate as follows:

[0149]

[0150] Where s is the total number of horizontal coordinates of each hump peak value obtained; c qis the abscissa value corresponding to the position of the q-th hump peak; a q is the abscissa value corresponding to the intersection of curve G x (t) and curve A x (t) on the left side of the q-th hump; b q is the abscissa value corresponding to the intersection of curve G x (t) and curve A x (t) on the right side of the q-th hump.

[0151] The present invention reconstructs the TOA difference distribution through an overlapping PRI bin structure, combines the idea of dynamic time origin calibration, and improves the adaptability of the algorithm. At the same time, a vine moss algorithm is designed, the Chebyshev mapping is used to construct the population initialization position distribution, the Lévy flight strategy is introduced to change the spore reproduction search method, and the environmental factor adaptation search mechanism is incorporated to avoid falling into local optimal solutions, improve the global search ability, and increase the parameter space coverage rate, enabling the proposed method to achieve PRI jitter radar signal sorting in an unknown environment.

[0152] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A PRI jitter radar signal sorting method based on the vine moss algorithm, characterized in that: The following steps are involved: Step 1: Receive the radar signal within the sampling time and convert it into an ordered pulse sequence according to the order of pulse arrival time; Step 2: Pre-sort the pulse sequence to obtain the preliminary clustering result of the radar signal; Step 3: For each pulse sequence in the preliminary clustering result of the radar signal, determine the minimum value of the PRI value range, the maximum value of the PRI value range and the number of PRI boxes; The vine moss algorithm is used, and the overlap rate of the PRI overlap box in the pulse main sorting, the adjustable coefficient of observation time, the adjustable coefficient of eliminating subharmonics and the adjustable coefficient of eliminating noise are taken as the optimization targets. The Chebyshev mapping is used to construct the population initialization position distribution, the Levy flight strategy is introduced to change the spore reproduction search method, and the environmental factor adaptation search mechanism is integrated to enhance the global search ability and obtain the optimal parameter combination. Step 4: According to the optimal parameter combination corresponding to each pulse sequence in the preliminary clustering results of radar signals, pulse primary sorting is performed on the preliminary clustering results of radar signals, the PRI value retrieval range of each pulse sequence in the preliminary clustering results of radar signals is determined, the corresponding pulse description words are removed from the preliminary clustering results of radar signals, the pulse sequence of each radar is separated, and radar signal sorting is completed.

2. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 1 is characterized in that: In step 1, the radar signal is converted into an ordered pulse sequence pulse_list: pulse_list={PDW1,PDW2,...,PDW i ,...,PDW n } Among them, PDW i is the ith pulse description word of the pulse sequence pulse_list, i=1,2,...,n, n is the total number of pulses in the pulse sequence; PDW i ={TOA i ,PW i ,BW i ,CF i ,PA i }, TOA i is the pulse arrival time of the ith pulse description word, PW i is the pulse width of the ith pulse description word, BW i is the bandwidth of the ith pulse description word, CF i is the carrier frequency of the ith pulse description word, PA i is the pulse amplitude of the i-th pulse description word; The pre-sorting process of the pulse sequence in step 2 is specifically as follows: Step 2.1: Calculate the PW of all pulse descriptors i , BW i and CF i The mean of Step 2.2: Calculate the PW of all pulse descriptors i , BW i and CF i The standard deviation of Step 2.3: Calculate the PW of each pulse descriptor word i , BW i and CF i Standard score of Step 2.4: Calculate the PW of each pulse descriptor word i , BW i and CF i The normalized result is: Step 2.5: Cluster the normalized results PW_uni, BW_uni and CF_uni, and then cluster the pulse description words PDW in the pulse sequence pulse_list according to the clustering results. i Clustering, get the radar signal preliminary clustering result pre_res: pre_res={pulse_list1,pulse_list2,...,pulse_list ii ,...,pulse_list f } Among them, pulse_list ii is the ii-th pulse sequence in the preliminary clustering result of the radar signal, ii=1,...,f, f is the total number of pulse sequences in the preliminary clustering result of the radar signal; the receiving time set of each pulse sequence in the preliminary clustering result of the radar signal is T, T={T1,T2,...,T ii ,...,T f }, T ii is the receiving time of the ii-th pulse sequence in the preliminary clustering result of the radar signal.

3. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 2 is characterized in that: In step 3, the Chebyshev mapping is used to construct the population initialization position distribution as follows: Among them, X u (1) is the initial position of the u-th vine moss, u = 1, ..., NUM, NUM is the number of vine mosses in the population; the position of each vine moss represents a parameter combination of the overlap rate e of a set of PRI overlap boxes, the adjustable coefficient of observation time α, the adjustable coefficient of subharmonic elimination β and the adjustable coefficient of noise elimination γ; R1 is a random number [0,1]; l_x is the lower limit of the vine moss group position parameter range, u_x is the upper limit of the vine moss group position parameter range, and h is a constant parameter.

4. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 3 is characterized in that: In step 3, the Levy flight strategy is introduced to change the spore reproduction search mode, specifically: Vine mosses search for spores by dispersing their spores with the wind. Due to the different wind speeds, the search mechanism of vine mosses is different: in, is the position of the new vine moss after the u-th vine moss spreads through spores in the Ite-th iteration; K is a constant parameter; R2 is a random number in [0,1], R3 is a random number in [0,1]; WF is the wind strength, MI is the maximum number of iterations; L is the coefficient of spore propagation distance under wind conditions; B is the expansion parameter of the Levy distribution step; z is a standard normal distribution random variable; μ is the parameter of the Levy distribution; X best is the position of the global optimal moss individual.

5. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 4 is characterized in that: The specific steps in step 3 are: Step 3.1: Set the parameters of the vine moss algorithm, including the number of vine mosses in the population num and the maximum number of iterations MI; initialize the current number of iterations Ite = 1, and use Chebyshev mapping to construct the population initialization position distribution X u (1); According to the initial position X of each vine moss u (1) Calculate the fitness of each vine moss and select the position of the vine moss with the smallest fitness value as the global optimal position X best ; Step 3.2: From X best Randomly take the vth dimension value X best,v As a threshold, compare it with all vine moss positions X u (Ite) the vth dimension value X u,v (Ite) comparison, if X u,v (Ite)>X best,v (Ite), then X u (Ite) is added to the set area_m(Ite); if X u,v (Ite)≤X best,v (Ite), then X u (Ite) joins the set area_l(Ite); Step 3.3: Compare the number of vine mosses in area_m(Ite) and area_l(Ite), and select the set with more vine mosses as area(Ite); Step 3.4: Calculate wind direction WD; Among them, sum(area(Ite)) is the number of all vine mosses in the set area(Ite); Step 3.5: Introduce the Levy flight strategy to change the spore reproduction search method and obtain the position of the new vine moss after each vine moss spreads through spores The spore dispersal distance coefficient L under gentle wind conditions is: Step 3.6: Incorporate the environmental factor adaptive search mechanism to enhance the global search capability, determine whether the environmental factor EV(Ite) meets the search conditions for sexual reproduction of vine mosses, and if the condition EV(Ite) ≥ 0.5 is met, sexual reproduction search is performed, otherwise vegetative reproduction search is performed to obtain the updated position X of each vine moss u (Ite+1); Among them, the environmental factor EV(Ite) is a random number between [0,1]; rand and N are both 1×4 random vectors, each element in rand is a random number between [0,1], and each element in N is a random number with standard Gaussian distribution; ⊙ is the operator for finding the product of two vector elements; R4 is a random number between [0,1]; Q is the propagation factor, H is a constant parameter; Step 3.7: Update the position X of each vine moss u (Ite+1) Calculate the fitness of each vine moss. If there is a vine moss updated position X u The fitness of (Ite+1) is less than the global optimal position X best The fitness of the global optimal position X is updated best ; Step 3.8: If Ite < MI, set Ite = Ite + 1 and return to step 3.2; otherwise, output the global optimal position X best The corresponding parameter combination.

6. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 5 is characterized in that: In step 3, the iith pulse sequence pulse_list in the preliminary clustering result of the radar signal is ii , determine the minimum value τ of the PRI value range min (ii) Maximum value of PRI value range τ max (ii) and the number of PRI bins F(ii), and τ min (ii) Not greater than pulse_list ii The starting point corresponds to the moment, τ max (ii) not less than pulse_list ii The time corresponding to the end point; then the box width wid of the PRI box is: The center coordinates of the x-th PRI box τ x is: x =(x-0.5)·wid+τ min , the width of the overlapped portion of the xth PRI box ov_wid x For: ov_wid x =e·τ x ; The range of the xth PRI box [C 1x ,C 2x ]for: When obtaining the optimal parameter combination through the vine moss algorithm, according to the position X of the vine moss u The specific method for calculating fitness is: Step 4.1: Position the moss according to the vine u ={e,α,β,γ}, obtain the overlap rate e, the observation time adjustable coefficient α, the sub-harmonic elimination adjustable coefficient β and the noise elimination adjustable coefficient γ of the PRI overlap box; Step 4.2: Initialize the cumulative spectrum values ​​of all PRI boxes G = {G1, G2, ..., G x ,...,G F }={0,0,...,0,...,0}, the time starting point O of all PRI boxes={O1,O2,...,O x ,...,O F }={0,0,...,0,...,0}, the number of times all PRI boxes are used flag={flag1,flag2,...,flag x ,...,flag F }={0,0,...,0,...,0};initialize j=2,i=1; Step 4.3: Calculate the arrival time difference ε between the i-th pulse description word and the j-th pulse description word; ε=TOA j -TOA i Step 4.4: If ε ≥ τ max , then let j = j + 1 and execute step 4.5; if ε ≤ τ min , then let i = i-1 and execute step 4.6; if τ min <ε<τ max , then execute step 4.7; Step 4.5: If j>num(pulse_list ii ), then execute step 4.8; if j≤num(pulse_list ii ), then let i = j-1, and return to step 4.3; num(pulse_list ii ) is the number of pulses in the ii-th pulse sequence in the pre-sorting result; Step 4.6: If i>1, return to step 4.3; if i≤1, set j=j+1 and return to step 4.5; Step 4.7: Determine the PRI box range that ε falls into [C 1x ,C 2x ], get the corresponding PRI box index x; since [C 1x ,C 2x ] includes the overlapped parts of other PRI boxes, for [C 1x ,C 2x ] Update the cumulative spectrum value G of all PRI boxes within the range x Then, let i=i-1 and return to step 4.6; Update the cumulative spectrum value G x The method is: Step 4.7.1: If [C 1x ,C 2x ]The number of times a PRI box is used within the range flag x =0, then set flag x =flag x +1, O x =TOA j ; Step 4.7.2: For [C 1x ,C 2x ] for all PRI boxes within the range, and calculate their initial phase η0, decomposition parameters U and W; Step 4.7.3: If [C 1x ,C 2x ] a PRI box within the range satisfies U = 1 and O x =TOA i Or U ≥ 2 and |W| ≤ ξ, then update its starting point to O x =TOA j ; ξ is the constant coefficient that determines the starting point of time; Step 4.7.4: Update [C 1x ,C 2x The cumulative spectrum value G of all PRI boxes within the range x ; G x =G x +D x in, Step 4.8: Calculate the pulse current density ρ ii ; Step 4.9: For each PRI bin, calculate the detection threshold A x ; Draw G x —τ x Discrete points and connect them to form a curve G x (t); draw A x —τ x Discrete points and connect them to form curve A x (t); if the curve G x The hump peak G in (t) x (τ x )>A x (τ x ), then obtain the horizontal coordinate of the hump peak value, that is, the hump peak value G x (τ x ) corresponding to τ x ; Step 4.10: Calculate the position X of the vine moss u The corresponding fitness fit; Where s is the total number of horizontal coordinates of each hump peak value obtained; c q is the horizontal coordinate value corresponding to the position of the qth hump peak; a q The left side of the qth hump, curve G x (t) and curve A x (t) The horizontal coordinate value corresponding to the intersection point; b q The right side of the qth hump, curve G x (t) and curve A x (t) The horizontal coordinate value corresponding to the intersection point.

7. The PRI jitter radar signal sorting method based on the vine moss algorithm according to claim 6 is characterized in that: In step 4, each pulse sequence pulse_list in the radar signal preliminary clustering result pre_res is ii The corresponding optimal parameter combination, according to step 4.1 to step 4.9, takes the horizontal axis coordinate of the detected hump peak as the PRI estimated value, sets the upper and lower deviation range ψ of the PRI estimated value as the PRI value retrieval range, and selects the pulse sequence pulse_list from the pulse sequence pulse_list. ii Remove the corresponding pulse description word; for each pulse sequence pulse_list in the radar signal preliminary clustering result pre_res ii , perform the above operations, separate the pulse sequence of each radar, distinguish different radar signals, and complete radar signal sorting.

8. A computer device / equipment / system comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images