A radar echo data compression method based on target imaging prior

Abstract

The application discloses a radar echo data compression method based on target imaging prior, comprising the following steps: step one, obtaining the original echo signal of a radar, and establishing a target dimension reduction observation model based on an SF signal, the original echo signal, a reference signal and an observation matrix; step two, establishing an observation matrix and imaging performance joint optimization model according to the target dimension reduction observation model and the observation matrix; step three, iteratively solving the observation matrix and the target high-resolution range image according to the observation matrix and imaging performance joint optimization model to obtain an optimal observation matrix and compressed data. The application improves the data compression efficiency and obtains the target high-resolution imaging result, and improves the radar data compression performance.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of signal and information processing, and particularly relates to a radar echo data compression method based on target imaging prior. BACKGROUND

[0002] With the continuous development of information technology and the increasing demand of people for radar imaging performance, the radar signal bandwidth, pulse repetition frequency, polarization mode and observation mode are increasing, which leads to a sharp increase in imaging radar data volume, bringing great challenges to data acquisition, storage, transmission and processing of radar systems. Therefore, it is of great significance to study efficient compression methods for imaging radar echo data.

[0003] Generally, the original data of imaging radar has the characteristics of high entropy value and small power change, and the commonly used lossless compression method has unsatisfactory effect. At present, the coding compression methods represented by block adaptive quantization, block adaptive vector quantization and block adaptive tree vector quantization are one of the commonly used radar data compression methods. Although this kind of method can compress the data volume to a certain extent, it is realized under the Nyquist sampling condition, and the signal bandwidth will directly affect the data volume. The compressed sensing (CS) method provides a new idea for radar data compression. The compressed sensing method can break through the limitation of Nyquist sampling theorem. As long as the signal has compressibility or sparsity in a certain transform domain, a non-correlation observation method can be used to project the high-dimensional signal into a low-dimensional space through an observation matrix, and then the original signal can be accurately reconstructed in a probabilistic sense according to a small number of projection values by solving an optimization problem. Therefore, the radar data compression method based on compressed sensing can obtain better performance. However, the existing imaging radar data compression method based on compressed sensing is not flexible enough and does not have the adaptive ability to radar data (essentially to the observed target). In fact, when using the compressed sensing method to compress radar data, the structure of the observation matrix essentially reflects the compression method of the data, and the dimension of the observation matrix directly determines the amount of compressed data obtained finally. Obviously, in order to obtain better data compression ratio, the smaller the dimension of the observation matrix is, the better. At the same time, in order to avoid the storage and transmission cost of the observation matrix, the observation matrix should be a 0-1 structure with band constraint, that is, only the data is retained or deleted. Therefore, the optimization of the observation matrix is one of the keys to improve the performance of compressed sensing radar data compression.

[0004] Currently, some research results have been achieved in the optimization of observation matrix under the CS theory framework. Optimized projections for compressed sensing (IEEE Transactions on Signal Processing, 2007, 55(12): 5695-5702) reduces the correlation between the observation matrix and the sparse transform matrix by defining the Gram matrix and threshold shrinkage function, and realizes the optimization of the observation matrix. Sensing matrix optimization based on equiangular tight frames with consideration of sparse representation error (IEEE transactions on Multimedia. 2016, 18(10): 2040-2053) proposes a robust observation matrix design method based on equiangular tight frame, which improves the signal reconstruction accuracy. 3D scattering image reconstruction based on measurement optimization of a radar network (Remote Sensing Letters, 2020, 11(7): 697-706) proposes a three-dimensional imaging method based on observation optimization of a radar network, which optimizes the network observation mode based on the minimum mutual coherence of the sensing matrix in the worst case, and improves the target imaging performance. However, the existing observation matrix optimization methods have the following shortcomings: (1) Lack of adaptability to the target. Existing researches mostly take the irrelevance of the sensing matrix as the optimization target, and do not consider the influence of target characteristics on the optimization of the observation matrix; (2) Lack of adaptive adjustment ability of the dimension of the observation matrix. Existing researches usually optimize the element values in the observation matrix under the given dimension condition, and do not consider the optimization problem of the dimension of the observation matrix; (3) Difficult to meet the 0-1 structure constraint condition of the observation matrix. Existing researches mostly take random Gaussian matrix as the initial matrix for iterative optimization, and the element values of the observation matrix are not constrained in the optimization process. In order to reconstruct the signal, the observation matrix needs to be stored and transmitted, which cannot realize the real data compression.Although, Measurement matrix optimization for ISAR sparse imaging based on genetic algorithm (IEEE Geoscience and Remote Sensing Letters, 2016, 13(12): 1875-1879) studies the observation matrix optimization method under the radar target sparse imaging framework, but the method does not fully utilize the characteristics of the target, and it is difficult to optimize the imaging quality and the observation matrix at the same time, in addition, the method may fall into a local optimal solution.

[0005] Therefore, the compression efficiency of the existing imaging radar data compression method based on compressive sensing is low, and the radar data compression performance is poor. SUMMARY

[0006] In order to solve the above problems existing in the prior art, the present application provides a radar echo data compression method based on target imaging priori. The technical problem to be solved by the present application is solved by the following technical scheme:

[0007] A radar echo data compression method based on target imaging priori, comprising:

[0008] Step one, obtaining the original echo signal of the radar, and establishing a target dimension reduction observation model based on the SF signal, the original echo signal, the reference signal and the observation matrix;

[0009] Step two, establishing an observation matrix and imaging performance joint optimization model according to the target dimension reduction observation model and the observation matrix;

[0010] Step three, iteratively solving the observation matrix and the target high-resolution range profile according to the observation matrix and imaging performance joint optimization model to obtain the optimal observation matrix and compressed data.

[0011] In an embodiment of the present application, the SF signal is composed of N carrier frequencies with sub-pulses stepped by Δf, and each sub-pulse is in the form of a single frequency signal;

[0012] The expression of the i-th sub-pulse in each cluster of pulse trains is:

[0013]

[0014] Wherein, i=0, 1, …, N-1, T1 is the sub-pulse width, T r is the sub-pulse repetition period, f i =f c +iΔf is the carrier frequency of the i-th sub-pulse, f c is the starting carrier frequency of the pulse train, and t is the fast time;

[0015] The expression of the reference signal is:

[0016]

[0017] Where i=0,1,…,N-1, R ref is the distance from the reference point to the radar, T ref is the pulse width of the reference signal.

[0018] In an embodiment of the present application, the step one comprises:

[0019] Step 11, obtaining the original echo signal of the radar, and mixing the original echo signal and the reference signal to obtain the i-th sub-pulse at the slow time τ m ;

[0020] The expression of the i-th sub-pulse at the slow time τ m ;

[0021]

[0022] Where i=0,1,…,N-1, the target is composed of P scattering points, σ p is the scattering coefficient of the p-th scattering point, R p (τ m ) is the distance from the p-th scattering point to the radar at the time τ m , R ref is the distance from the reference point to the radar, R Δp (τ m ) = R p (τ m ) - R ref represents the distance from the p-th scattering point to the reference point at the slow time τ m ;

[0023] Step 12, performing Fourier transform on the i-th sub-pulse at the slow time τ m with respect to t'=t-iT r -2R ref / c and sampling at the frequency peak to obtain the target echo signal;

[0024] The expression of the target echo signal is:

[0025]

[0026] Step 13, performing Fourier transform on the target echo signal with respect to i to obtain the target high-resolution range image at the slow time τ m ;

[0027] The expression of the target high-resolution range image is:

[0028]

[0029] Step 14, dimension reduction observation is performed on the target echo signal in vector form using an M*N (M<N) dimensional observation matrix Φ, and a target dimension reduction observation model is established;

[0030] The target dimension reduction observation model is:

[0031] S c ′=ΦS c =ΦΨS H ;

[0032] Wherein, Ψ is a Fourier transform matrix, S H is a target high-resolution range profile in vector form at a slow time τ m , S c is a target echo signal in vector form, S c =[S c (0,τ m ),...,S c (i,τ m ),...,S c (N-1,τ m )] T , and T represents a transposition operation.

[0033] In an embodiment of the present application, the observation matrix and imaging performance joint optimization model is:

[0034]

[0035] Wherein, ||·||2 and ||·||1 represent L2 norm and L1 norm of a vector or a matrix respectively, is an element in the mth row and nth column of the observation matrix Φ, S D is an expected target high-resolution range profile, λ1 and λ2 represent weight parameters, and the observation matrix Φ is a 0-1 structure with constraints.

[0036] In an embodiment of the present application, the step three: iteratively solving the observation matrix and the target high-resolution range profile according to the observation matrix and imaging performance joint optimization model comprises:

[0037] Step 31, determining an expression of a reconstructed target high-resolution range profile according to the observation matrix and imaging performance joint optimization model;

[0038] Step 32, performing sparse reconstruction iterative calculation on the expression of the reconstructed target high-resolution range profile based on a fixed observation matrix, to obtain a reconstructed target high-resolution range profile and a reconstructed radar echo signal;

[0039] Step 33, solving the measurement matrix based on the joint optimization model of the measurement matrix and imaging performance and the reconstructed target high-resolution range image to obtain an optimized measurement matrix;

[0040] Step 34, repeat steps 32 and 33 to obtain the optimal observation matrix and compressed data.

[0041] In one embodiment of the present invention, the expression for reconstructing the high-resolution range image of the target is:

[0042]

[0043] Among them, y g represents the observed signal, y g =Φ g S c , A g represents the perception matrix, A g =Φ g Ψ,g represents the g-th iteration in the update iteration of the observation matrix and the target high-resolution range image, and I represents the identity matrix.

[0044] In one embodiment of the present invention, step 32 includes:

[0045] Step 321: Based on the perception matrix A g calculate Get the column with the largest matrix norm, where k represents the kth iteration of reconstructing the target high-resolution range image, A g,q Represents the perception matrix A g The qth column of the matrix, the column with the largest norm is located at q k ;

[0046] Step 322: Update the column support set Ώ of the perception matrix (k) =[Π (k-1) ,u k ] and the maximum column position record vector q (k) =q (k-1) ∪{q k}, and u k From A g Delete in;

[0047] Step 323: Based on the observed signal y g And the expression of reconstructing the target high-resolution range image updates the vector to be reconstructed X (k) =(Π (k)T Π (k) +λ2I) -1 (Π (k)T y g +λ2S D );

[0048] Step 324, updating the residual r (k) = y g - Π (k) X (k) ;

[0049] Step 325, setting k = k + 1, K max is the maximum iteration number, if k ≤ K max and return to step 321; otherwise, stop iteration, and obtain the reconstructed target high-resolution range profile S H,g and the reconstructed radar echo signal S c,g :

[0050] S H,g = {δ l}, wherein represents the lth element in X (k) , S c,g = ΨS H,g , and e represents a residual signal energy threshold.

[0051] In an embodiment of the present application, the specific steps of the step 33 include:

[0052] Step 331, determining an observation matrix solving model according to the observation matrix and the imaging performance joint optimization model and the reconstructed target high-resolution range profile;

[0053] wherein, the observation matrix solving model is:

[0054]

[0055] Step 332, solving the observation matrix solving model by using a branch and bound method, and obtaining an optimized observation matrix Φ g+1 .

[0056] In an embodiment of the present application, the specific steps of the step 34 are: setting the observation matrix and the target high-resolution range profile updating iteration number g = g + 1, repeating the step 32-step 33, and when the g reaches a maximum iteration number G, the iteration stops, and the optimal observation matrix and the compressed data are obtained.

[0057] The present application has the following beneficial effects:

[0058] The application is directed to a SF signal commonly used in radar target imaging, a target dimension reduction observation model based on CS is established, and data compression mode optimization is converted into an observation matrix optimization problem. On this basis, the internal consistency between the observation matrix structure and the data compression mode is analyzed, the target imaging prior information is fully utilized, the minimum observation matrix dimension and the maximum imaging similarity are taken as the optimization objectives, the 0-1 constraint of the data compression process on the observation matrix structure is considered, a joint optimization model of the observation matrix and the imaging performance based on the target imaging prior is established, and a corresponding observation matrix and target high-resolution range profile alternating iterative updating solution algorithm and a target high-resolution range profile sparse reconstruction algorithm based on the imaging prior are proposed, so that the optimal observation matrix and high-resolution imaging result are obtained, the data compression efficiency is improved, and the target high-resolution imaging result is obtained, and the radar data compression performance is improved.

[0059] The application will be further described in detail below with reference to the drawings and embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A flowchart of a radar echo data compression method based on a target imaging prior provided by the embodiment of the application is shown in the figure.

[0061] Figure 2 A target high-resolution range profile (target HRRP) simulation result under a full data condition is shown in the figure.

[0062] Figure 3 A target high-resolution range profile (target HRRP) reconstructed according to compressed data provided by the embodiment of the application is shown in the figure.

[0063] Figure 4 A comparison between an original echo signal and the original echo signal reconstructed according to compressed data provided by the embodiment of the application is shown in the figure.

[0064] Figure 5 A target two-dimensional image reconstructed according to compressed data provided by the embodiment of the application is shown in the figure.

[0065] Figure 6 A target high-resolution range profile (target HRRP) reconstructed according to compressed data of a random observation matrix under the same data compression rate is shown in the figure.

[0066] Figure 7 A target high-resolution range profile (target HRRP) reconstructed according to compressed data of a Gram matrix optimization method under the same data compression rate is shown in the figure.

[0067] Figure 8 An echo signal reconstructed according to compressed data of a random observation matrix under the same data compression rate is shown in the figure.

[0068] Figure 9 An echo signal reconstructed according to compressed data of a Gram matrix optimization method under the same data compression rate is shown in the figure. DETAILED DESCRIPTION

[0069] The application will be described in further detail below with reference to specific embodiments, but the embodiments of the application are not limited thereto.

[0070] As shown in the figure, a radar echo data compression method based on target imaging prior includes: Figure 1

[0071] Step one, obtain the original echo signal of the radar, and establish a target dimension reduction observation model based on the SF signal, the original echo signal, the reference signal and the observation matrix. For the step-frequency (SF) signal commonly used in radar target imaging, a target dimension reduction observation model based on compressed sensing (CS) is established, and the data compression mode optimization is converted into an observation matrix optimization problem.

[0072] Wherein, the SF signal is composed of N carrier frequencies with Δf stepped sub-pulses, and each sub-pulse is in the form of a single frequency signal;

[0073] The expression of the i-th sub-pulse in each cluster of pulse trains is:

[0074]

[0075] Wherein, i=0,1,…,N-1, T1 is the sub-pulse width, T r is the sub-pulse repetition period, f i =f c +iΔf is the carrier frequency of the i-th sub-pulse, f c is the starting carrier frequency of the pulse train, t is the fast time, i.e. the time within the pulse train;

[0076] The expression of the reference signal is:

[0077]

[0078] Wherein, i=0,1,…,N-1, R ref is the distance from the reference point to the radar, T ref is the pulse width of the reference signal.

[0079] Specifically, step one includes steps 11-14:

[0080] Step 11, obtain the original echo signal of the radar, and perform mixing processing on the original echo signal and the reference signal based on the “walk-stop” model to obtain the i-th sub-pulse at the slow time τ m ;

[0081] The expression of the i-th sub-pulse at the slow time τ m ;​

[0082]

[0083] Where i = 0, 1, ..., N-1, the target consists of P scattering points, σ p is the scattering coefficient of the pth scattering point, R p (τ m ) is τ m The distance from the pth scattering point to the radar at the moment, R ref is the distance from the reference point to the radar, R Δp (τ m )=R p (τ m )-R ref represents the slow time τ m The distance from the pth scattering point to the reference point at time t′ = t-iT r -2R ref / c,s c (t,i,τ m ) can be expressed as:

[0084]

[0085] Step 12, for s c (t′,i,τ m ) and perform Fourier transform on t′, we can get:

[0086]

[0087] Wherein, i=0,1,…,N-1, and f is the frequency domain representation corresponding to t′.

[0088] exist S c (f,i,τ m ) is sampled, and the target echo signal S c (i,τ m ):

[0089]

[0090] Step 13: Target echo signal S c (i,τ m ) Perform Fourier transform on i and get slow time τ m Target high-resolution range profile (target HRRP) at the moment;

[0091] The expression of the target high-resolution range image is:

[0092]

[0093] Step 14, at a given slow time instant τ m , the target echo signal S c (i,τ m ) is recorded as the form of N-dimensional observation vector S c =[S c (0,τ m ),...,S c (i,τ m ),...,S c (N-1,τ m )] T , where the superscript T represents the transposition operation. Under the framework of compressed sensing theory, the target echo signal S in the form of a vector is measured using the M×N (M<N) dimensional measurement matrix Φ. c Conduct dimensionality reduction observation and establish target dimensionality reduction observation model;

[0094] The target dimension reduction observation model is established as:

[0095] S c ′=ΦS c =ΦΨS H ;

[0096] Where Ψ is the Fourier transform matrix, S H is the slow time τ m The target high-resolution range image S at the moment H (F f ,τ m ) in vector form, usually, S H With sparsity, the target HRRP and the original signal S can be achieved by solving an optimization problem c On this basis, the target two-dimensional high-resolution imaging result can be obtained through azimuth Fourier transform processing.

[0097] Clearly, in the aforementioned signal dimensionality reduction observation process, the dimension of the measurement matrix determines the dimensionality of the reduced observation data, and therefore the data compression ratio. Furthermore, the structure of the measurement matrix reflects the data compression method, directly influencing both the target imaging result and the quality of the original signal reconstruction. Therefore, optimizing the measurement matrix design can optimize the data compression method, improve the imaging radar data compression performance, and achieve satisfactory target imaging results.

[0098] Step two, according to the target dimensionality reduction observation model and the internal consistency of the observation matrix structure and the data compression method, the observation matrix and the imaging performance joint optimization model based on the target imaging prior is established. On the basis of transforming the data compression method optimization into the observation matrix optimization, in order to realize the real sense of data compression, the observation matrix is designed as a 0-1 structure with constraints, that is, there is only one element with a value of 1 in each row of the observation matrix, and at most one element with a value of 1 in each column, 0 represents the corresponding data deletion, and 1 represents the corresponding data retention. This process is easy to implement in actual processing, and can avoid the increase of new data caused by observation matrix storage and transmission. In addition, making full use of the target feature information can effectively improve the data compression ratio and the signal reconstruction performance. Therefore, based on the target imaging prior information, the observation matrix and the imaging performance joint optimization model is established:

[0099]

[0100] Wherein, ||·||2 and ||·||1 represent the L2 norm and the L1 norm of the vector or matrix respectively, is the element in the mth row and the nth column of the observation matrix Φ, S D is the expected target high-resolution range image, that is, the target high-resolution range image obtained under the full data condition, the first term is the data fitting term, which reflects the fitting degree between the recovered signal and the original signal, the second term ||Φ||1 reflects the dimension of the observation matrix, that is, the data amount after compression, and the third term reflects the target imaging quality, that is, the similarity between the true imaging result and the expected result, and also reflects the signal reconstruction performance. The weight parameters λ1 and λ2 reflect the trade-off degree between the data fitting degree, the data compression ratio and the target imaging quality. The constraint condition reflects the 0-1 structure constraint of the observation matrix.

[0101] Step three, the optimal observation matrix and the compressed data are obtained by iteratively solving the observation matrix and the target high-resolution range image according to the observation matrix and the imaging performance joint optimization model. For the above observation matrix and the imaging performance joint optimization model, the observation matrix and the target high-resolution range image are iteratively optimized and solved. Initialize the maximum iteration number G, let the iteration number g = 1, first, randomly generate an initial observation matrix Φ1 with M × N dimension which satisfies the constraint condition requirement, wherein N is the original signal length, M is the initial given observation dimension, that is, the signal length after compression. According to the demand of the observation data amount in the compression sensing theory, its value can be set as M = c1K·ln(N) (wherein K is the signal sparsity, and c1 is a small constant). On this basis, the observation matrix and the target high-resolution range image are iteratively updated, and the specific process is as follows:

[0102] Step 31, determining the expression of the reconstruction target high-resolution range profile based on the imaging prior according to the observation matrix and the imaging performance joint optimization model, specifically:

[0103] For the gth iteration, the observation matrix Φ g is solved under the condition of fixed reconstruction target high-resolution range profile S H,g .

[0104] At this time, Φ g can be regarded as a constant, and only S H,g is optimized and solved, and the optimization objective function in the observation matrix and the imaging performance joint optimization model is recorded as:

[0105]

[0106] The above formula is derived with respect to S H,g :

[0107]

[0108] For simplicity, the observation signal y g = Φ g S c , the perception matrix A g = Φ g Ψ, and dF(S H,g ) can be expressed as:

[0109]

[0110] Let the above formula equal to 0, and the expression of the reconstruction target high-resolution range profile based on the imaging prior is:

[0111]

[0112] Step 32, based on the fixed observation matrix, the expression of the reconstruction target high-resolution range profile based on the imaging prior is sparsely reconstructed and iteratively calculated to obtain the reconstruction target high-resolution range profile and the reconstruction radar echo signal, and the specific process is as follows:

[0113] Step 320, input the observation signal y g , the perception matrix A g and the residual signal energy threshold e; and initialize: the signal residual r (0) =y g , the column support set of the perception matrix the maximum column position record vector the maximum iteration number K max of sparse reconstruction, the current iteration number k = 1;

[0114] Step 321, calculating Get the column with the largest matrix norm, where k represents the kth iteration of reconstructing the target high-resolution range image, A g,q Represents the perception matrix A g The qth column of the maximum column is recorded as q k ;

[0115] Step 322, update the column support set of the perception matrix: (k) =[Π (k-1) ,u k ], update the maximum column position record vector: q (k) =q (k-1) ∪{q k}, will u k From A g Delete in;

[0116] Step 323: Update the vector to be reconstructed according to the expression of reconstructing the target high-resolution range image: (k) =(∏ (k)T ∏ (k) +λ2I) -1 (∏ (k)T y g +λ2S D )

[0117] Step 324, update the residual r (k) =y g -∏ (k) X (k) ;

[0118] Step 325, let k = k + 1, if k ≤ K max and Return to step 321; otherwise, stop the iteration and obtain the reconstructed target high-resolution range image S H,g and reconstruct the radar echo signal S c,g :

[0119] S H,g ={δ l},in Represents X (k) The lth element in S c,g =ΨS H,g , K max is the maximum number of iterations.

[0120] Step 326: Output the reconstructed target high-resolution range image S H,g and reconstruct the radar echo signal S c,g .

[0121] Step 33: Solve the observation matrix based on the joint optimization model of the observation matrix and imaging performance and reconstruct the target high-resolution range image to obtain the optimized observation matrix Φg+1 Step 33 includes step 331 and step 332:

[0122] Step 331, determining an observation matrix solving model according to the observation matrix and the imaging performance joint optimization model and the reconstructed target high-resolution range profile, specifically:

[0123] For the gth iteration, determining the reconstructed target high-resolution range profile S H,g under the condition that the observation matrix is fixed H,g g+1 : At this time, only the observation matrix is optimized and solved. It should be noted that the observation matrix dimension is one of the important parameters to be optimized. Generally, the initial given observation dimension M is greater than the optimal observation dimension, so in the observation matrix optimization process, the case where a certain row element takes a value of 0 is allowed, and the all-0 row is deleted after optimization. At this time, the observation matrix solving model can be represented as:

[0124]

[0125] Step 332, solving the observation matrix solving model by using a branch and bound method to obtain an optimized observation matrix g+1 . The observation matrix solving model is a typical 0-1 programming problem, and the classical branch and bound method is used to solve the observation matrix, and then the all-0 row in the obtained observation matrix is deleted, that is, the optimized observation matrix g+1 .

[0126] Step 34, repeating steps 32-33 to obtain the optimal observation matrix and compressed data. Specifically:

[0127] Let g = g + 1, repeat steps 32-33 to obtain the optimal observation matrix H and and further obtain the data compression result

[0128] After obtaining the data compression result, the target high-resolution range profile and the original echo signal can be accurately reconstructed from the compressed data by using the above imaging-prior-based target high-resolution range profile sparse reconstruction algorithm, that is, the calculation process of steps 320-326.

[0129] The effect of the present application is further illustrated by simulation experiments as follows:

[0130] Experimental parameters: radar transmits an SF signal: f c = 15 GHz, T r ​=39.063μs, T1 = 4.8828μs, Δf = 4.6875MHz, N = 64. The target consists of five scattering points, with coordinates relative to the reference point (target center) at (8, -7, 0), (-8, -2, 0), (0, 0, 0), (5, 5, 0), (-3, 8, 0), in meters. The target center coordinates are (200, 10000, 0) meters, and it flies along the x-axis at a speed of 300m / s. The remaining simulation parameters are set as follows: c1 = 1.2, G = 50, K max =10, e=0.1, λ1=1, λ2=0.15.

[0131] The target high resolution range image (target HRRP) under full data conditions is as follows Figure 2 As shown in Figure 1, the target is used as the desired imaging result. The measurement matrix is ​​optimized by the method of the present invention to achieve data compression and reconstruct the target high resolution range image (target HRRP) and the original signal. The dimension of the optimal measurement matrix is ​​11, and the reconstructed target HRRP is as follows: Figure 3 As shown, it is basically consistent with the expected imaging results. Figure 4 A comparison diagram of the reconstructed signal and the original signal is given, and the correlation coefficient between the two is 0.9814. The above results show that the data compression rate of the method of the present invention can reach 7.11, and at the same time, it can achieve accurate reconstruction of the target image and the original echo signal. On this basis, through the azimuth Fourier transform processing, the target two-dimensional imaging result can be obtained, as shown in the figure. Figure 5 As shown, it is consistent with the target scattering distribution.

[0132] Under the same measurement matrix dimension, that is, when the data compression rate is the same, the measurement matrix obtained by the random measurement matrix and the Gram matrix optimization method is used to image the target, and the target imaging prior information is not used in the imaging process. The traditional orthogonal matching pursuit algorithm is used to reconstruct the signal. The target HRRPs are as follows: Figure 6 and Figure 7 As shown, there is a big difference from the expected imaging result. The reconstructed signals obtained on this basis are as follows Figure 8 and Figure 9 As shown in FIG, there is a large error with the original signal. This experimental result proves the performance advantage of the method of the present invention.

[0133] The application firstly establishes a target dimension reduction observation model based on CS for the SF signal commonly used in radar target imaging, and converts the data compression mode optimization into an observation matrix optimization problem. On this basis, the internal consistency between the observation matrix structure and the data compression mode is analyzed, the target imaging prior information is fully utilized, the minimum observation matrix dimension and the maximum imaging similarity are taken as the optimization objectives, the 0-1 constraint of the data compression process on the observation matrix structure is considered, a joint optimization model of the observation matrix and the imaging performance based on the target imaging prior is established, and a corresponding observation matrix and target high-resolution range profile alternating iterative updating solving algorithm and a target high-resolution range profile sparse reconstruction algorithm based on the imaging prior are proposed, so that the optimal observation matrix and high-resolution imaging result are obtained, and the radar data compression performance is improved.

[0134] In addition, the terms "first", "second", "third", etc. are used only for descriptive purposes and are not to be construed as indicating or implying relative importance or an indicated number of technical features. Therefore, the features defined with "first", "second", etc. can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly specified.

[0135] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in the present application.

[0136] The above is a further detailed description of the present application in combination with specific preferred embodiments, and the specific implementation of the present application cannot be limited to these descriptions. For those skilled in the art, without departing from the concept of the present application, a number of simple deductions or replacements can be made, which should be regarded as falling within the protection scope of the present application.

Claims

1. A method for radar echo data compression based on target imaging priors, characterized in that, The application relates to a radar target high-resolution range image compression method, which comprises the following steps: Step one, obtaining a radar original echo signal, and establishing a target dimension-reduced observation model based on an SF signal, the original echo signal, a reference signal and an observation matrix; Step two, establishing an observation matrix and imaging performance joint optimization model according to the target dimension-reduced observation model and the observation matrix; wherein the observation matrix and imaging performance joint optimization model is: ； wherein, and denote the L2-norm and the L1-norm of a vector or matrix, respectively, is an observation matrix is the element in the m row n column of is the desired target high resolution range profile; and denote a weight parameter, the observation matrix is a 0-1 structure with constraints; is the target high resolution range profile at a slow time moment in vector form, is the target echo signal in vector form, T denotes the transposition operation; Step three, iteratively solving the observation matrix and the target high-resolution range image according to the observation matrix and imaging performance joint optimization model to obtain an optimal observation matrix and compressed data.

2. The method of claim 1, wherein, The SF signal is composed of sub-pulses of one carrier frequency in steps, each sub-pulse being in the form of a single-frequency signal. The expression of the first sub-pulse in each cluster of pulse trains is: ​ ； wherein, , is a sub-pulse width, is a sub-pulse repetition period, is a carrier frequency of the th sub-pulse, is a pulse train start carrier frequency, is a fast time; The expression of the reference signal is: ； wherein, , is the distance of the reference point to the radar, is the pulse width of the reference signal.

3. The method of claim 2, wherein, The step one comprises: Step 11, obtaining a raw echo signal of the radar, mixing the raw echo signal and the reference signal to obtain a slow time first sub-pulse of the moment ​ The slow time The expression of the first sub-pulse at the moment ； wherein, the target consists of scattering points, is the scattering coefficient of the scattering point, is the distance of the scattering point to the radar at the moment t, is the distance of the scattering point to the radar, denotes the slow time is the distance of the scattering point to the reference point at the moment t. Step 12, slow down time The moment Sub-pulses do Fourier transform is performed and sampling is performed at the frequency peak to obtain the target echo signal; The expression of the target echo signal is: ， ； Step 13, making Fourier transform of the target echo signal with respect to i the slow time instance to obtain the target high resolution range profile. The expression of the target high-resolution range image is: ； Step 14, using M X N Observation matrix The target echo signal in the form of a vector is reduced in dimension, and a target dimension reduction observation model is established, wherein, M < N ; The target dimension-reduced observation model is: ； wherein is a Fourier transform matrix, is a slow time target high resolution range profile at time is a target echo signal in vector form, T denotes a transposition operation.

4. The method of claim 1, wherein, The step three: iteratively solving the observation matrix and the target high-resolution range image according to the observation matrix and imaging performance joint optimization model, comprises: Step 31, determining the expression of a reconstructed target high-resolution range image according to the observation matrix and imaging performance joint optimization model; Step 32, performing sparse reconstruction iteration calculation on the expression of the reconstructed target high-resolution range image based on the fixed observation matrix to obtain a reconstructed target high-resolution range image and a reconstructed radar echo signal; Step 33, solving the observation matrix based on the observation matrix and imaging performance joint optimization model and the reconstructed target high-resolution range image to obtain an optimized observation matrix; Step 34, repeating steps 32-33 to obtain an optimal observation matrix and compressed data.

5. The method of claim 4, wherein, The expression of the reconstructed target high-resolution range image is: ； in, represents the observed signal, , represents the perception matrix, , represents the first iteration in the update of the observation matrix and the target high-resolution range image iterations, I represents the identity matrix; Indicates the The observation matrix obtained by iteration.

6. The method of claim 5, wherein, The step 32 comprises: Step 321, according to the perception matrix Calculate , get the column with the maximum matrix norm, wherein, Indicates the first iteration of the reconstruction target high-resolution range image, Indicates the first column of the perception matrix q , and the position of the column with the maximum matrix norm is ; Step 322: Update the column support set of the perception matrix and maximum column position record vector , and from Delete in; Step 323, updating the vector to be reconstructed according to the observed signal and the expression of the reconstructed target high-resolution range profile ; Step 324, update the residual ; Step 325, let , be the maximum iteration number, if and , return to step 321; otherwise, stop iteration, and get the reconstructed target high-resolution range profile and the reconstructed radar echo signal : wherein , denotes the l element of the , residual signal energy threshold.

7. The method of claim 4, wherein, The specific steps of the step 33 comprise: Step 331, determining an observation matrix solving model according to the observation matrix and imaging performance joint optimization model and the reconstructed target high-resolution range image; Wherein the observation matrix solving model is: ； Step 332, a branch and bound method is used to solve the model of the observation matrix to obtain an optimized observation matrix .

8. The method of claim 4, wherein, The specific step of the step 34 is: let the observation matrix and the target high-resolution range image update the iteration times , repeat the step 32-step 33, when the maximum iteration times G are reached, the iteration stops, and the optimal observation matrix and the compressed data are obtained.

Images