A method and system for ground penetrating radar time delay estimation based on coprime sampling
Through mutual quality sampling and off-grid sparse Bayesian learning algorithm, the problems of ground penetrating radar in signal coherence and computational complexity are solved, and efficient time delay estimation and time resolution improvement are achieved.
Patent Information
- Application Number
- CN202310458767.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-26
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2043-04-26
AI Technical Summary
Existing ground penetrating radars face signal coherence problems and increased computational complexity when improving resolution performance, especially under multipath effect and high sampling rates, which are difficult for traditional methods to effectively conduct delay estimation.
Using frequency domain sparse sampling and off-grid sparse Bayesian learning algorithms based on mutually-grain sparse Bayesian learning algorithms, through frequency domain mutual sampling, non-circular expansion and first-order Taylor expansion, the delay estimation of coherent echoes is realized and the sampling rate is reduced. and improve time resolution.
It effectively reduces the computing complexity and hardware burden of ground penetrating radar, improves time resolution, avoids system errors caused by over-complete matrix division, and achieves unbiased delay estimation.
Smart Images

Figure CN116559863B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of radar parameter estimation, and in particular relates to a ground penetrating radar time delay estimation method and system based on coprime sampling. Background Art
[0002] Time Delay Estimation (TDE) is an important problem in the research of Ground Penetrating Radar (GPR) system. Assuming that the layered material is determined and the dielectric constant of the material is known, the estimation of other structural parameters can be achieved based on the time delay estimation.
[0003] With the continuous development of materials and process technology, most materials tend to be thinner and lighter, which poses a challenge to the resolution performance of ground penetrating radar detection methods. To improve the resolution performance of ground penetrating radar, there are currently two ways to achieve it: one is to expand the radar bandwidth, and the other is to use signal processing methods to improve the temporal resolution of the radar. In order to ensure the normal use of existing conventional ground penetrating radars, ground penetrating radars require the resolution of the algorithm to be improved when the radar bandwidth is limited. Traditional parameter estimation methods based on Fast Fourier Transform (FFT) usually require a large number of samples and their resolution is directly limited by the bandwidth of the radar system; while subspace-based high-resolution algorithms represented by the MUSIC algorithm can achieve a resolution much higher than that of FFT.
[0004] Another key difficulty in ground penetrating radar signal processing is the coherence problem caused by the multipath effect, which makes the subspace method ineffective. Therefore, before estimating the delay of radar coherent echoes, the signal needs to be dephased, such as spatial smoothing technology, to meet the full rank requirement of the signal covariance matrix of subspace algorithms. However, dephasing restores the rank of the covariance matrix at the cost of losing effective bandwidth, which reduces the time resolution and also reduces the degree of freedom.
[0005] At the same time, the decorrelation method is strictly limited to the uniform sampling method. In order to ensure that no estimation ambiguity occurs, the high sampling rate brought by uniform sampling imposes a burden on the computational complexity of the algorithm. Sparse sampling brings new ideas to solve this problem: one is to use a sparse sampling structure, and the other is to adopt a sparse signal model. Sparse sampling structures such as coprime sampling usually combine multiple uniform sparse samplings with large spacing to achieve non-uniform spacing sparse sampling, but non-uniform sampling structures are not suitable for decorrelation interference processing. In recent years, convex optimization methods and compressed sensing methods have been research hotspots in parameter estimation and image processing. These two types of algorithms are not affected by signal coherence. Based on the signal model, sparse signals can be reconstructed and restored, but the computational complexity is high and there is a basis mismatch problem. In addition, most algorithms cannot achieve unbiased estimation of off-grid parameters of coherent signals in many cases due to the division of their search grids or overcomplete matrices. Summary of the invention
[0006] In view of the deficiencies in the prior art, the present invention provides a ground penetrating radar time delay estimation method and system based on coprime sampling.
[0007] In a first aspect, the present invention provides a ground penetrating radar time delay estimation method based on coprime sampling, comprising:
[0008] Perform frequency domain sparse sampling on multiple coherent echoes to obtain frequency domain mutually prime sampling received signals;
[0009] According to the non-circular characteristics of the coherent echo of the ground penetrating radar, the frequency domain coprime sampling received signal is non-circularly expanded to obtain the equivalent non-circular frequency domain received signal.
[0010] The first-order Taylor expansion is performed on the equivalent non-circular frequency domain received signal, and the time delays of multiple coherent echoes are determined using an off-grid sparse Bayesian learning algorithm.
[0011] Furthermore, the performing frequency domain sparse sampling on the multiple coherent echoes to obtain the frequency domain mutually prime sampling received signal includes:
[0012] Construct the expression of the received signal of frequency domain coprime sampling:
[0013] y C =Λ C A C s+n C ;
[0014] Among them, y C =[y(f1),y(f2),…,y(f M+N-1 )] T, is a (M+N-1)×1-dimensional discrete frequency domain received signal; M and N are the total number of sampling points of two sparse uniform samplings of mutually prime sampling, and M and N are mutually prime positive integers; the two sparse uniform samplings of mutually prime sampling share a starting sampling point as the reference frequency point, with a total of M+N-1 frequency domain sampling points; y(f i ) is the i-th frequency domain sampling point f i The received signal; T is the transposition operator; is the radar pulse frequency response matrix of the ground penetrating radar; diag(·) is the diagonal matrix operator; is the i-th frequency domain sampling point f i The frequency amplitude response of the radar pulse; A C =[a C (t1),a C (t2),…,a C (t K )], is the pattern matrix of coprime sampling in frequency domain; is the mode vector of the kth coherent echo; K is the total number of coherent echoes; e is a natural constant; j is an imaginary unit; π is the circumference constant; t k is the time delay of the kth coherent echo; s=[s1,s2,…,s K ] T , is the echo amplitude vector of the backscattered coherent echo signal of the ground penetrating radar; s k is the amplitude of the kth coherent echo; n C is a (M+N-1)×1 dimensional noise vector.
[0015] Furthermore, the non-circular expansion of the frequency domain mutually prime sampling received signal is performed according to the non-circular characteristics of the ground penetrating radar coherent echo to obtain an equivalent non-circular frequency domain received signal, including:
[0016] Construct an equivalent expression for the non-circular frequency domain received signal:
[0017]
[0018] Where J is the (M+N-1)-dimensional unit anti-diagonal matrix; (·) * conjugate operator; Λ is the extended radar pulse frequency response matrix; A=[a(t1),a(t2),…,a(t K )], is the expanded non-circular pattern matrix; is the pattern vector a C (t k ) is the non-circular pattern vector after expansion.
[0019] Furthermore, performing a first-order Taylor expansion on the equivalent non-circular frequency domain received signal and using an off-grid sparse Bayesian learning algorithm to determine the time delays of multiple coherent echoes includes:
[0020] Construct an overcomplete dictionary matrix covering the ground penetrating radar working range, using a uniformly divided grid point set For N S grid points, the number of grid points is N S >>K, grid step size Δτ=τ2-τ1, and each delay t k None of them fall on the grid points. is the distance t k The nearest grid point;
[0021] Build mode vectora C (t k )The non-circular pattern vector a(t k ) expression:
[0022]
[0023] in, is the grid point The non-circular pattern vector of is the non-circular pattern vector at the grid point The first order differential of ; is the off-grid error of the kth coherent echo;
[0024] Constructing a grid dictionary A S and the deviation dictionary B S The expression is:
[0025]
[0026] in, and The Nth S Grid points The non-circular mode vector on and the first-order differential of the non-circular mode vector;
[0027] The expression for constructing the off-grid overcomplete dictionary matrix Ξ is:
[0028] Ξ=A S +B S Ψ;
[0029] Where Ψ = diag{ψ}, is the off-grid error parameter vector The diagonal matrix formed by For N S The grid error of the grid points; the elements of ψ satisfy:
[0030]
[0031] Among them, the prior probability distribution of the off-grid error parameter vector ψ follows a uniform distribution Represents the error between the real time delay and the discrete grid points;
[0032] By combining the approximate error and noise, the off-grid received signal model expression is constructed:
[0033] y O =ΛΞx+n;
[0034] in, is a sparse source signal vector; For N S The source signal amplitude of grid points; n is the corresponding noise vector, and the noise obeys zero mean and variance is σ 2 The Gaussian distribution of the source signal vector x is sparse:
[0035]
[0036] The conditional likelihood function of the coherent echo signal is:
[0037]
[0038] in, N S dimensional identity matrix; det(·) is the determinant operator; (·) H is the conjugate transpose operator;
[0039] Each snapshot of the received signal is independent of each other. T The received signal matrix Y is:
[0040] Y = ΛΞX + N;
[0041] in, The source N T A matrix of snapshots; is the sparse source signal vector at the rth snapshot time, r=1,2,…,N T ; is the noise N T A matrix of snapshots, is the noise vector at the rth snapshot moment; the likelihood function of Y is:
[0042]
[0043] Among them, Π(·) is the cumulative multiplication operator; x m Obeying a Gaussian distribution with zero mean and unknown variance, the source covariance matrix Σ x =diag{γ x}, Determine the sparsity of the source, x j is the row vector form of the sparse source signal at the rth snapshot moment; y j is the received signal at the rth snapshot moment; γ m is the power of the sparse source vector:
[0044]
[0045] The prior probability of X is expressed as:
[0046]
[0047] in, N S ×1-dimensional zero vector; assume covariance γ x The prior for follows a gamma distribution:
[0048]
[0049] Among them, ρ is a positive gamma distribution hyperparameter. For a gamma distribution with hyperparameters a and b, Ground penetrating radar electromagnetic waves propagate in low-loss media, and the coherent echo amplitude s k Related to the reflection coefficient, and independent of each snapshot, we get:
[0050]
[0051] The posterior probability of X is:
[0052]
[0053] Here, the variance of the noise is estimated as γ0 follows a gamma distribution with hyperparameters c and d, p(γ0)=Γ(γ0|c,d); N T The mean U of the sparse source vectors of snapshots is:
[0054]
[0055] Among them, μ r is the mean of the sparse source vector at the rth snapshot moment; the source covariance matrix Σ x for:
[0056] Σ x =(γ0(ΛΞ) H ΛΞ+Σ -1 ) -1 ;
[0057] Update γ in each iteration m The values of γ0 and ψ are updated in each round as and ψnew :
[0058]
[0059] Until the algorithm converges, the delay estimation result is obtained; where [·] m,m is the element at the mth row and mth column of the matrix; d is the hyperparameter of the gamma distribution corresponding to the noise variance; ||·||2 is the L-2 norm of the matrix; E{·} is the expectation of the random process; tr{·} is the trace of the matrix.
[0060] In a second aspect, the present invention provides a ground penetrating radar time delay estimation system based on coprime sampling, comprising:
[0061] A frequency domain sparse sampling module is used to perform frequency domain sparse sampling on multiple coherent echoes to obtain a frequency domain mutually prime sampling received signal;
[0062] A non-circular expansion module is used to perform non-circular expansion on the frequency domain coprime sampling received signal according to the non-circular characteristics of the ground penetrating radar coherent echo, so as to obtain an equivalent non-circular frequency domain received signal;
[0063] The Taylor expansion module is used to perform first-order Taylor expansion on the equivalent non-circular frequency domain received signal and determine the time delay of multiple coherent echoes using an off-grid sparse Bayesian learning algorithm.
[0064] In a third aspect, the present invention provides a computer device comprising a processor and a memory; wherein, when the processor executes a computer program stored in the memory, the steps of the ground penetrating radar time delay estimation method based on coprime sampling described in the first aspect are implemented.
[0065] In a fourth aspect, the present invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, the steps of the ground penetrating radar time delay estimation method based on coprime sampling described in the first aspect are implemented.
[0066] The present invention provides a method and system for estimating time delay of a ground penetrating radar based on coprime sampling. Coprime sampling can greatly reduce the sampling rate of the ground penetrating radar, thereby reducing the hardware and computing burden of the ground penetrating radar to meet the real-time requirements of the ground penetrating radar system. At the same time, the method of the present invention utilizes the non-circular characteristics of the coherent echo of the ground penetrating radar to expand the signal in the frequency domain, effectively improving the time resolution of the ground penetrating radar. In addition, the off-grid sparse Bayesian learning method proposed by the method of the present invention can realize off-grid time delay estimation, effectively avoiding the system error problem caused by overcomplete matrix partitioning. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0068] Figure 1 A flow chart of a ground penetrating radar time delay estimation method based on coprime sampling provided by an embodiment of the present invention;
[0069] Figure 2 A schematic diagram of a ground penetrating radar backscatter coherent echo signal model provided by an embodiment of the present invention;
[0070] Figure 3 A schematic diagram of a frequency domain sampling distribution model provided by an embodiment of the present invention;
[0071] Figure 4 A schematic diagram of the result of implementing a single estimation with other coherent TDE algorithms provided by an embodiment of the present invention;
[0072] Figure 5 A schematic diagram of the root mean square error performance of the embodiment of the present invention and other TDE algorithms under different signal-to-noise ratios;
[0073] Figure 6 A schematic diagram of the root mean square error performance of the embodiment of the present invention and other TDE algorithms under different time-bandwidth products;
[0074] Figure 7 A structural diagram of a ground penetrating radar time delay estimation system based on coprime sampling provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0075] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0076] In one embodiment, if Figure 1 As shown, an embodiment of the present invention provides a ground penetrating radar delay estimation method based on coprime sampling, comprising:
[0077] Step 101 , performing frequency domain sparse sampling on multiple coherent echoes to obtain frequency domain mutually prime sampling received signals.
[0078] For example, the structure is as follows Figure 2 The ground penetrating radar receiving signal model shown in Figure 3 The frequency domain mutually prime sampling structure model shown performs frequency domain sparse sampling on multiple overlapping coherent ground penetrating radar echo signals to obtain frequency domain mutually prime sampling received signals.
[0079] like Figure 3 As shown in the figure, coprime sampling is composed of two sparse uniform samplings in the frequency domain: let the conventional uniform sampling interval be Δf, then the two sparse uniform sampling intervals of coprime sampling are NΔf and MΔf, corresponding to M and N sampling points respectively, where M and N are coprime positive integers, and share a starting sampling point as the reference frequency point, with a total of M+N-1 frequency domain sampling points.
[0080] Assume that there are K coherent echoes with time delays t1, t2, …, t k ,…,t K , construct the expression of the received signal of frequency domain coprime sampling:
[0081] y C =Λ C A C s+n C .
[0082] Among them, y C =[y(f1),y(f2),…,y(f M+N-1 )] T , is a (M+N-1)×1-dimensional discrete frequency domain received signal; M and N are the total number of sampling points of two sparse uniform samplings of mutually prime sampling, and M and N are mutually prime positive integers; the two sparse uniform samplings of mutually prime sampling share a starting sampling point as the reference frequency point, with a total of M+N-1 frequency domain sampling points; y(f i ) is the i-th frequency domain sampling point f i The received signal; T is the transposition operator; is the radar pulse frequency response matrix of the ground penetrating radar; diag(·) is the diagonal matrix operator; is the i-th frequency domain sampling point f i The frequency amplitude response of the radar pulse; A C =[a C (t1),a C (t2),…,a C (t K )], is the pattern matrix of coprime sampling in frequency domain; is the mode vector of the kth coherent echo; K is the total number of coherent echoes; e is a natural constant; j is an imaginary unit; π is the circumference constant; t k is the time delay of the kth coherent echo; s=[s1,s2,…,s K ] T , is the echo amplitude vector of the backscattered coherent echo signal of the ground penetrating radar; sk is the amplitude of the kth coherent echo; n C is a (M+N-1)×1 dimensional noise vector.
[0083] Step 102, according to the non-circular characteristics of the coherent echo of the ground penetrating radar, the frequency domain mutually prime sampling received signal is non-circularly expanded to obtain an equivalent non-circular frequency domain received signal.
[0084] Exemplarily, an equivalent expression of the non-circular frequency domain received signal is constructed:
[0085]
[0086] Where J is the (M+N-1)-dimensional unit anti-diagonal matrix; (·) * conjugate operator; Λ is the extended radar pulse frequency response matrix; A=[a(t1),a(t2),…,a(t K )], is the expanded non-circular pattern matrix; is the pattern vector a C (t k ) is the non-circular pattern vector after expansion.
[0087] Step 103 , performing a first-order Taylor expansion on the equivalent non-circular frequency domain received signal, and using an off-grid sparse Bayesian learning algorithm to determine the time delays of multiple coherent echoes.
[0088] For example, an overcomplete dictionary matrix covering the working range of the ground penetrating radar is constructed using a uniformly divided grid point set For N S grid points, the number of grid points is N S >>K, grid step size Δτ=τ2-τ1, and each delay t k None of them fall on the grid points. is the distance t k The nearest grid point; then the pattern vector after non-circular expansion is approximately the form of a first-order Taylor expansion, that is, the off-grid pattern vector:
[0089]
[0090] Build mode vectora C (t k )The non-circular pattern vector a(t k ) expression:
[0091]
[0092] in, is the grid point The non-circular pattern vector of is the non-circular pattern vector at the grid point The first order differential of ; is the off-grid error of the kth coherent echo.
[0093] Constructing a grid dictionary A S and the deviation dictionary B S The expression is:
[0094]
[0095] in, and The Nth S Grid points The non-circular mode vector on and the first-order differential of the non-circular mode vector;
[0096] The expression for constructing the off-grid overcomplete dictionary matrix Ξ is:
[0097] Ξ=A S +B S Ψ.
[0098] Where Ψ = diag{ψ}, is the off-grid error parameter vector The diagonal matrix formed by For N S The grid error of the grid points; the elements of ψ satisfy:
[0099]
[0100] Among them, the prior probability distribution of the off-grid error parameter vector ψ follows a uniform distribution Represents the error between the actual delay and the discrete grid points.
[0101] By combining the approximate error and noise, the off-grid received signal model expression is constructed:
[0102] y O =ΛΞx+n.
[0103] in, is a sparse source signal vector; For N S The source signal amplitude of grid points; n is the corresponding noise vector, and the noise obeys zero mean and variance is σ 2 The Gaussian distribution of the source signal vector x is sparse:
[0104]
[0105] The conditional likelihood function of the coherent echo signal is:
[0106]
[0107] in, N S dimensional identity matrix; det(·) is the determinant operator; (·) H is the conjugate transpose operator.
[0108] Each snapshot of the received signal is independent of each other. T The received signal matrix Y is:
[0109] Y=ΛΞX+N.
[0110] in, The source N T A matrix of snapshots; is the sparse source signal vector at the rth snapshot time, r=1,2,…,N T ; is the noise N T A matrix of snapshots, is the noise vector at the rth snapshot moment; the likelihood function of Y is:
[0111]
[0112] Among them, Π(·) is the cumulative multiplication operator; x m Obeying a Gaussian distribution with zero mean and unknown variance, the source covariance matrix Σ x =diag{γ x}, Determine the sparsity of the source, x r is the row vector form of the sparse source signal at the rth snapshot moment; y r is the received signal at the rth snapshot moment; γ m is the power of the sparse source vector:
[0113]
[0114] The prior probability of X is expressed as:
[0115]
[0116] in, N S ×1-dimensional zero vector; assume covariance γ x The prior for follows a gamma distribution:
[0117]
[0118] Among them, ρ is a positive gamma distribution hyperparameter. For a gamma distribution with hyperparameters a and b, Ground penetrating radar electromagnetic waves propagate in low-loss media, and the coherent echo amplitude s k Related to the reflection coefficient, and independent of each snapshot, we get:
[0119]
[0120] The posterior probability of X is:
[0121]
[0122] Here, the variance of the noise is estimated as γ0 follows a gamma distribution with hyperparameters c and d, p(γ0)=Γ(γ0|c,d); N T The mean U of the sparse source vectors of snapshots is:
[0123]
[0124] Among them, μ r is the mean of the sparse source vector at the rth snapshot moment; the source covariance matrix Σ x for:
[0125] Σ x =(γ0(ΛΞ) H ΛΞ+Σ -1 ) -1 .
[0126] Update γ in each iteration m The values of γ0 and ψ are updated in each round as and ψ new :
[0127]
[0128] Until the algorithm converges, the delay estimation result is obtained; where [·] m,m is the element at the mth row and mth column of the matrix; d is the hyperparameter of the gamma distribution corresponding to the noise variance; ||·||2 is the L-2 norm of the matrix; E{·} is the expectation of the random process; tr{·} is the trace of the matrix.
[0129] The effects of the embodiments of the present invention are further described below in conjunction with a simulation example.
[0130] The ground penetrating radar scene with a frequency band of [0.5, 2.5] GHz is simulated. Assuming that the number of echoes is 2, the true value of the delay of the first echo is t1 = 1.00ns, t2 = 1.71ns. The comparison algorithms (MUSIC algorithm, OMP algorithm, OGSBL algorithm) all use 21-point uniform sampling with full bandwidth, and the algorithm of the present invention (co-prime-OGSBL algorithm) uses 8-point co-prime sampling. For fair comparison, the number of grids of the MUSIC algorithm, the overcomplete dictionary matrix of the OMP algorithm and the OGSBL algorithm, and the overcomplete dictionary matrix of the proposed algorithm are all set to a grid division with a step size of 0.01ns.
[0131] The performance estimation standard of the present invention in the simulation is the relative root mean square error (RRMSE) defined as:
[0132]
[0133] Where Q is the number of Monte Carlo experiments, represents the delay estimation result of the kth backscattered echo in the qth experiment, t k represents the true value of the time delay of the kth backscattered echo, where t K is the maximum echo delay. In the following simulation, the number of Monte Carlo experiments is 200. The signal-to-noise ratio (SNR) is defined as the ratio of the energy of the first echo to the energy of the noise, in dB.
[0134] Simulation 1: Figure 4 The figure shows the result of one-time estimation by the method (Co-prime-OGSBL) of the present invention and the MUSIC algorithm, the OMP algorithm and the OGSBL algorithm, with a signal-to-noise ratio of 10 dB. In the working range of 0 to 10 ns, the other three comparison algorithms have periodic pseudo peaks, which affect the detection results; the algorithm proposed in the present invention uses a lower sampling rate and fewer sampling points, and no pseudo peaks appear in the search range, achieving effective and unambiguous coherent signal delay estimation.
[0135] Simulation 2: Figure 5 The figure shows the curves of the performance of the method of the present invention, the MUSIC algorithm, the OMP algorithm and the OGSBL algorithm as the signal-to-noise ratio changes. To ensure the fairness of the performance comparison, the working range of the ground penetrating radar is shortened to 0ns to 2.5ns for RRMSE performance comparison. Figure 5 It can be seen from the figure that the delay estimation performance of the algorithm described in the present invention remains better than that of the other three comparison algorithms.
[0136] Simulation 3: Figure 6The figure shows the performance curves of the algorithm of the present invention, the MUSIC algorithm, the OMP algorithm and the OGSBL algorithm under different bandwidth-time products. In this simulation, other parameters are fixed, and the delay t2 of the second echo varies between 0.2ns and 0.85ns, that is, BΔτ=B(t2-t1) varies between 0.2 and 1.5. Figure 6 It can be seen that the time resolution of the algorithm proposed in this invention is better than that of other comparison algorithms.
[0137] In summary, from the analysis of the simulation effect diagram and the running time, it can be seen that the ground penetrating radar delay estimation method proposed in the present invention can greatly reduce the computational complexity and meet the requirements of ground penetrating radar parameter estimation for resolution and real-time performance. At the same time, the algorithm can significantly improve the estimation accuracy and have better performance in harsh environments, when the signal-to-noise ratio is low or when the echoes are highly overlapped.
[0138] Based on the same inventive concept, an embodiment of the present invention also provides a ground penetrating radar time delay estimation system based on coprime sampling. Since the principle of solving the problem by this system is similar to that of the aforementioned ground penetrating radar time delay estimation method based on coprime sampling, the implementation of this system can refer to the implementation of the ground penetrating radar time delay estimation method based on coprime sampling, and the repeated parts will not be repeated.
[0139] In another embodiment, the ground penetrating radar delay estimation system based on coprime sampling provided by the embodiment of the present invention is as follows: Figure 7 As shown, including:
[0140] The frequency domain sparse sampling module 10 is used to perform frequency domain sparse sampling on multiple coherent echoes to obtain frequency domain mutually prime sampling received signals.
[0141] The non-circular expansion module 20 is used to perform non-circular expansion on the frequency domain mutually prime sampling received signal according to the non-circular characteristics of the ground penetrating radar coherent echo, so as to obtain an equivalent non-circular frequency domain received signal.
[0142] The Taylor expansion module 30 is used to perform a first-order Taylor expansion on the equivalent non-circular frequency domain received signal and determine the time delays of multiple coherent echoes using an off-grid sparse Bayesian learning algorithm.
[0143] For more specific working processes of the above modules, please refer to the corresponding contents disclosed in the above embodiments, which will not be repeated here.
[0144] In another embodiment, the present invention provides a computer device, including a processor and a memory; wherein, when the processor executes a computer program stored in the memory, the steps of the above-mentioned ground penetrating radar time delay estimation method based on coprime sampling are implemented.
[0145] For more specific processes of the above method, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.
[0146] In another embodiment, the present invention provides a computer-readable storage medium for storing a computer program; when the computer program is executed by a processor, the steps of the above-mentioned ground penetrating radar time delay estimation method based on coprime sampling are implemented.
[0147] For more specific processes of the above method, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.
[0148] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system, device, and storage medium disclosed in the embodiment, since they correspond to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0149] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention or some parts of the embodiments.
[0150] The present invention has been described in detail above in conjunction with specific implementations and exemplary examples, but these descriptions cannot be understood as limiting the present invention. Those skilled in the art understand that, without departing from the spirit and scope of the present invention, a variety of equivalent substitutions, modifications or improvements may be made to the technical solution of the present invention and its implementation methods, all of which fall within the scope of the present invention. The scope of protection of the present invention shall be subject to the attached claims.
Claims
1. A ground penetrating radar time delay estimation method based on coprime sampling, characterized in that: include: Perform frequency domain sparse sampling on multiple coherent echoes to obtain frequency domain mutually prime sampling received signals; According to the non-circular characteristics of the coherent echo of the ground penetrating radar, the frequency domain coprime sampling received signal is non-circularly expanded to obtain the equivalent non-circular frequency domain received signal. The first-order Taylor expansion is performed on the equivalent non-circular frequency domain received signal, and the time delays of multiple coherent echoes are determined using an off-grid sparse Bayesian learning algorithm.
2. The method for estimating ground penetrating radar time delay according to claim 1, characterized in that: The performing frequency domain sparse sampling on the multiple coherent echoes to obtain a frequency domain mutually prime sampling received signal includes: Construct the expression of the received signal of frequency domain coprime sampling: y C =L C A C s+n C ; Among them, y C =[y(f1),y(f2),…,y(f M+N-1 )] T , is a (M+N-1)×1-dimensional discrete frequency domain received signal; M and N are the total number of sampling points of two sparse uniform samplings of mutually prime sampling, and M and N are mutually prime positive integers; the two sparse uniform samplings of mutually prime sampling share a starting sampling point as the reference frequency point, with a total of M+N-1 frequency domain sampling points; y(f i ) is the i-th frequency domain sampling point f i The received signal; T is the transposition operator; is the radar pulse frequency response matrix of the ground penetrating radar; diag(·) is the diagonal matrix operator; is the i-th frequency domain sampling point f i The frequency amplitude response of the radar pulse; A C =[a C (t1),a C (t2),…,a C (t K )], is the pattern matrix of coprime sampling in frequency domain; is the mode vector of the kth coherent echo; K is the total number of coherent echoes; e is a natural constant; j is an imaginary unit; π is the circumference constant; t k is the time delay of the kth coherent echo; s=[s1,s2,…,s K ] T , is the echo amplitude vector of the backscattered coherent echo signal of the ground penetrating radar; s k is the amplitude of the kth coherent echo; n C is a (M+N-1)×1 dimensional noise vector.
3. The method for estimating ground penetrating radar time delay according to claim 2, characterized in that: The method of performing non-circular expansion on the frequency domain mutually prime sampling received signal according to the non-circular characteristics of the ground penetrating radar coherent echo to obtain an equivalent non-circular frequency domain received signal includes: Construct an equivalent expression for the non-circular frequency domain received signal: Where J is the (M+N-1)-dimensional unit anti-diagonal matrix; (·) * conjugate operator; Λ is the extended radar pulse frequency response matrix; A=[a(t1),a(t2),…,a(t K )], is the expanded non-circular pattern matrix; is the pattern vector a C (t k ) is the non-circular pattern vector after expansion.
4. The method for estimating ground penetrating radar time delay according to claim 3, characterized in that: The first-order Taylor expansion is performed on the equivalent non-circular frequency domain received signal, and the off-grid sparse Bayesian learning algorithm is used to determine the time delays of multiple coherent echoes, including: Construct an overcomplete dictionary matrix covering the ground penetrating radar working range, using a uniformly divided grid point set For N S grid points, the number of grid points is N S >>K, grid step size Δτ=τ2-τ1, and each delay t k None of them fall on the grid points. is the distance t k The nearest grid point; Build mode vectora C (t k )The non-circular pattern vector a(t k ) expression: in, is the grid point The non-circular pattern vector of is the non-circular pattern vector at the grid point The first order differential of is the off-grid error of the kth coherent echo; Constructing a grid dictionary A S and the deviation dictionary B S The expression is: in, and The Nth S Grid points The non-circular mode vector on and the first-order differential of the non-circular mode vector; The expression for constructing the off-grid overcomplete dictionary matrix Ξ is: Ξ=A S +B S P; Where Ψ = diag{ψ}, is the off-grid error parameter vector The diagonal matrix formed by For N S The grid error of the grid points; the elements of ψ satisfy: Among them, the prior probability distribution of the off-grid error parameter vector ψ follows a uniform distribution Represents the error between the real time delay and the discrete grid points; By combining the approximate error and noise, the off-grid received signal model expression is constructed: y O =ΛΞx+n; in, is a sparse source signal vector; For N S The source signal amplitude of grid points; n is the corresponding noise vector, and the noise obeys zero mean and variance is σ 2 The Gaussian distribution of the source signal vector x is sparse: The conditional likelihood function of the coherent echo signal is: in, N S dimensional identity matrix; det(·) is the determinant operator; (·) H is the conjugate transpose operator; Each snapshot of the received signal is independent of each other. T The received signal matrix Y is: Y = ΛΞX + N; in, The source N T A matrix of snapshots; is the sparse source signal vector at the rth snapshot time, r=1,2,…,N T ; is the noise N T A matrix of snapshots, is the noise vector at the rth snapshot moment; the likelihood function of Y is: Among them, Π(·) is the cumulative multiplication operator; x m Obeying a Gaussian distribution with zero mean and unknown variance, the source covariance matrix Σ x =diag{γ x }, Determine the sparsity of the source, x r is the row vector form of the sparse source signal at the rth snapshot moment; y r is the received signal at the rth snapshot moment; γ m is the power of the sparse source vector: The prior probability of X is expressed as: in, N S ×1-dimensional zero vector; assume covariance γ x The prior for follows a gamma distribution: Among them, ρ is a positive gamma distribution hyperparameter. For a gamma distribution with hyperparameters a and b, Ground penetrating radar electromagnetic waves propagate in low-loss media, and the coherent echo amplitude s k Related to the reflection coefficient, and independent of each snapshot, we get: The posterior probability of X is: Here, the variance of the noise is estimated as γ0 follows a gamma distribution with hyperparameters c and d, p(γ0)=Γ(γ0|c,d); N T The mean U of the sparse source vectors of snapshots is: Among them, μ r is the mean of the sparse source vector at the rth snapshot moment; the source covariance matrix Σ x for: S x =(γ0(ΛΞ) H LX+S -1 ) -1 ; Update γ in each iteration m The values of γ0 and ψ are updated in each round as and ψ new : Until the algorithm converges, the delay estimation result is obtained; where [·] m,m is the element at the mth row and mth column of the matrix; d is the hyperparameter of the gamma distribution corresponding to the noise variance; ||·||2 is the L-2 norm of the matrix; E{·} is the expectation of the random process; tr{·} is the trace of the matrix.
5. A ground penetrating radar time delay estimation system based on coprime sampling, characterized in that: include: A frequency domain sparse sampling module is used to perform frequency domain sparse sampling on multiple coherent echoes to obtain a frequency domain mutually prime sampling received signal; A non-circular expansion module is used to perform non-circular expansion on the frequency domain coprime sampling received signal according to the non-circular characteristics of the ground penetrating radar coherent echo, so as to obtain an equivalent non-circular frequency domain received signal; The Taylor expansion module is used to perform first-order Taylor expansion on the equivalent non-circular frequency domain received signal and determine the time delay of multiple coherent echoes using an off-grid sparse Bayesian learning algorithm.
6. A computer device, characterized in that: It comprises a processor and a memory; wherein, when the processor executes the computer program stored in the memory, it implements the steps of the ground penetrating radar time delay estimation method based on coprime sampling as described in any one of claims 1 to 4.
7. A computer-readable storage medium, characterized in that: Used to store computer programs; when the computer programs are executed by the processor, the steps of the ground penetrating radar time delay estimation method based on coprime sampling described in any one of claims 1 to 4 are implemented.
