A method and system for obtaining target incoming wave direction for depth-finding side-scan sonar

Abstract

The present invention relates to the field of array signal processing technology, and in particular to a method and system for obtaining the direction of arrival of a target wave for a depth-measuring side-scan sonar. The method comprises: step 1: estimating the covariance matrix of the data matrix received by the full-receiving array; step 2: performing eigendecomposition on the covariance matrix to obtain a signal subspace and a noise subspace; step 3: obtaining a rough value of the target's estimated angle based on the signal subspace through the ESPRIT algorithm; step 4: substituting the rough value of the estimated angle into the signal steering vector for a first-order Taylor expansion, and utilizing the orthogonality of the noise subspace and the signal steering vector to obtain an estimated angle deviation value; and taking the sum of the rough value of the estimated angle and the estimated angle deviation value as the angle estimate of the target's incoming wave direction. The present invention uses an improved ESPRIT method combined with noise subspace compensation to ensure that its estimation result is substantially consistent with the accuracy of the MUSIC algorithm and that the amount of calculation is much less than that of the Root‑MUSIC algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of array signal processing, and in particular to a method and system for obtaining the direction of incoming waves of a target for use in a depth-scanning side-scan sonar. Background Art

[0002] Direction of Arrival (DOA) estimation is a core technology in array signal processing, used to identify the direction of arrival of a signal. Schmidt R et al. proposed the Multiple Signal Classification (MUSIC) algorithm for uniform linear arrays. This algorithm solves the covariance matrix of the array's received signal, performs eigenvalue decomposition on it, and obtains the signal subspace and noise subspace. The orthogonality between the signal steering vector and the noise subspace is then used to establish a spatial spectrum function, which is then subjected to a small-step spectral peak search to estimate the signal parameters. This algorithm achieves high estimation accuracy, but cannot directly obtain an analytical solution for the direction of arrival. It requires assistance from algorithms such as spectral peak search, resulting in high algorithmic complexity and long computational time. Building on this, Barabell AJ et al. proposed the Root-MUSIC algorithm, which uses a polynomial root-finding method instead of a spectral peak search, reducing the computational complexity to a certain extent. Roy R et al. proposed the Estimating Signal Parameter via Rotational Invariance Techniques (Esprit) algorithm. This algorithm divides the receiving array into two identical subarrays. The incident angles of the received signal on these two subarrays differ by a rotation factor, which contains the DOA estimate. By solving the generalized eigenvalue equation, the magnitude of this rotation factor can be determined, and the target's direction of arrival can be determined. Compared to the MUSIC algorithm, this algorithm is less complex and faster, and can directly obtain analytical solutions for the parameters, but its estimation accuracy is somewhat lacking. Summary of the Invention

[0003] The purpose of the present invention is to compensate for the lower accuracy of the ESPRIT algorithm compared to the MUSIC algorithm in bathymetric side-scan sonar, while avoiding a significant increase in computational complexity. The present invention proposes a method and system for acquiring the direction of incoming target waves in bathymetric side-scan sonar. By combining an improved ESPRIT method with noise subspace compensation, the method's estimation results are essentially consistent with the accuracy of the MUSIC algorithm, while ensuring that the computational complexity is far less than that of the Root-MUSIC algorithm.

[0004] To solve the above technical problems, the technical solution of the present invention provides a method for obtaining the direction of target wave for a depth-scanning side-scan sonar, comprising:

[0005] Step 1: Estimate the covariance matrix R of the data matrix X received by the full receiving array X ;

[0006] Step 2: Covariance matrix R X Perform eigendecomposition to obtain the signal subspace E S and noise subspace E N ;

[0007] Step 3: Using the ESPRIT algorithm, based on the signal subspace E S , get the estimated angle rough value θ of the target gl ;

[0008] Step 4: Estimate the rough angle θ gl Substitute the signal-directed vector for first-order Taylor expansion and use the noise subspace E N The orthogonality of the signal steering vector is used to obtain the estimated angle deviation value δ l ; Estimate the rough angle value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is taken as the angle estimation value θ l .

[0009] As an improvement of the above method, step 1 specifically includes: estimating the covariance matrix R according to the data matrix received by the full receiving array X :

[0010]

[0011] Where K is the number of snapshots, X is the data matrix received by the entire array, (·) H Represents the conjugate transpose of a matrix.

[0012] As an improvement to the above method, step 2 specifically includes: X Perform eigendecomposition, arrange the eigenvalues ​​from large to small, and the eigenvectors related to the first L large eigenvalues ​​form the signal subspace E S , the eigenvectors associated with the remaining ML small eigenvalues ​​constitute the noise subspace E N , where M is the number of elements in the full-receiving array, and L is the number of far-field narrowband signals incident on the full-receiving array;

[0013]

[0014] in,(·) H represents the conjugate transpose of the matrix, ∑ S is the diagonal matrix composed of signal eigenvalues, ∑ N is a diagonal matrix composed of noise eigenvalues.

[0015] As an improvement to the above method, step 3 specifically includes:

[0016] Step 3-1: Divide the full receiving array into a first sub-array and a second sub-array that are identical and overlap each other; divide the signal subspace E S Split into the first subspace E S1 and the second subspace E S2 ; Among them, the first subspace E S1 is the signal subspace corresponding to the first subarray; the second subspace E S2 is the signal subspace corresponding to the second subarray:

[0017] E S1 =[I M-1 |0 M-1 ]E S

[0018] E S2 =[0 M-1 |I M-1 ]E S

[0019] Among them, I M-1 represents the (M-1)×(M-1) identity matrix, 0 M-1 represents the (M-1)×1 zero vector;

[0020] Step 3-2: Using the first subspace E S1 and the second subspace E S2 , obtain the first process matrix Ψ:

[0021] Ψ=E S1 + E S2

[0022] in,(·) + It means seeking pseudo-inverse;

[0023] Step 3-3: Perform eigenvalue decomposition on the first process matrix Ψ to obtain the eigenvalue vector z corresponding to the L largest eigenvalues l ;

[0024] Step 3-4: Based on the eigenvalue vector z corresponding to the L largest eigenvalues l , use the ESPRIT algorithm to solve and get the rough value of the estimated angle θ gl :

[0025]

[0026] Where λ is the wavelength of the far-field narrowband signal.

[0027] As an improvement to the above method, step 4 specifically includes:

[0028] Step 4-1: The rough value θ of the estimated angle obtained by the ESPRIT algorithm gl Substitute the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g :

[0029]

[0030] Among them, a gl is the steering vector of the lth estimated angle coarse value;

[0031] Step 4-2: The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm gl Substitute the derivative of the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g The derivative vector B g :

[0032]

[0033] Among them, b gl is the steering vector a of the lth estimated angle rough value gl The derivative value of

[0034] Step 4-3: Vector A formed by estimating the angle coarse value steering vector g and its derivative vector B g and noise subspace E N , get the estimated angle deviation value δ l :

[0035]

[0036] Re{·} is the real part of the complex number;

[0037] Step 4-4: Estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is taken as the angle estimation value θ l .

[0038] To achieve another object of the present invention, the present invention further provides a system for obtaining the direction of target wave for a depth-sounding side-scan sonar, comprising:

[0039] Covariance module, used to estimate the covariance matrix R of the data matrix X received by the full receiving array X ;

[0040] Decomposition module for the covariance matrix R X Perform eigendecomposition to obtain the signal subspace ES and noise subspace E N ;

[0041] ESPRIT module, through the ESPRIT algorithm, based on the signal subspace E S , get the estimated angle rough value θ of the target gl ;and

[0042] Estimation module, used to estimate the rough angle value θ gl Substitute the signal-directed vector into the first-order Taylor expansion to use the noise subspace E N The orthogonality of the signal steering vector is used to obtain the estimated angle deviation value δ l ; Used to obtain the estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is used as the angle estimate θ l .

[0043] As an improvement of the above system, the covariance module is specifically used to obtain the covariance matrix R according to the data matrix received by the full receiving array. X :

[0044]

[0045] Where K is the number of snapshots, X is the data matrix received by the entire array, (·) H Represents the conjugate transpose of a matrix.

[0046] As an improvement of the above system, the decomposition module is specifically used to: X Perform eigendecomposition, arrange the eigenvalues ​​from large to small, and the eigenvectors related to the first L large eigenvalues ​​form the signal subspace E S , the eigenvectors associated with the remaining ML small eigenvalues ​​constitute the noise subspace E N , where M is the number of elements in the full-receiving array, and L is the number of far-field narrowband signals incident on the full-receiving array;

[0047]

[0048] in,(·) H represents the conjugate transpose of the matrix, Σ S is the diagonal matrix composed of signal eigenvalues, ∑ N is a diagonal matrix composed of noise eigenvalues.

[0049] As an improvement to the above system, the ESPRIT module is specifically used to:

[0050] The full receiving array is divided into a first sub-array and a second sub-array that are completely identical and overlap each other; the signal subspace E S Split into the first subspace E S1 and the second subspace E s2 ; Among them, the first subspace E s1 is the signal subspace corresponding to the first subarray; the second subspace E S2 is the signal subspace corresponding to the second subarray:

[0051] E S1 =[I M-1 |0 M-1 ]E S

[0052] E S2 =[0 M-1 |I M-1 ]E S

[0053] Among them, I M-1 represents the (M-1)×(M-1) identity matrix, 0 M-1 represents the (M-1)×1 zero vector;

[0054] Using the first subspace E S1 and the second subspace F S2 , obtain the first process matrix Ψ:

[0055] Ψ=E S1 + E S2

[0056] in,(·) + It means seeking pseudo-inverse;

[0057] Perform eigenvalue decomposition on the first process matrix Ψ to obtain the eigenvalue vector z corresponding to the L largest eigenvalues l ;

[0058] Based on the eigenvalue vector z corresponding to the L large eigenvalues l , use the ESPRIT algorithm to solve and get the rough value of the estimated angle θ gl :

[0059]

[0060] Where λ is the wavelength of the far-field narrowband signal.

[0061] As an improvement to the above system, the estimation module is specifically configured to:

[0062] The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm glSubstitute the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g :

[0063]

[0064] Among them, a gl is the steering vector of the lth estimated angle coarse value;

[0065] The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm gl Substitute the derivative of the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g The derivative vector B g :

[0066]

[0067] Among them, b gl is the steering vector a of the lth estimated angle rough value gl The derivative value of

[0068] Vector A formed by estimating the angle rough value steering vector g and its derivative vector B g and noise subspace E N , get the estimated angle deviation value δ l :

[0069]

[0070] Re{·} is the real part of the complex number;

[0071] Get the estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is used as the angle estimate θ l .

[0072] Compared to existing technologies, the present invention offers advantages in that, based on the ESPRIT algorithm, it performs a first-order Taylor expansion on the signal steering vector. By leveraging the orthogonality between the noise subspace and the steering vector, it can directly derive an analytical solution for estimating the coarse deviation without requiring iterative calculations or spectral peak searching. This algorithm offers low computational complexity and high accuracy. When applied to bathymetric side-scan sonars, it can improve bathymetric accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A flow chart of a method for obtaining the target incoming wave direction for a depth-scanning sonar provided by the present invention;

[0074] Figure 2 Schematic diagram of the receiving array;

[0075] Figure 3 Schematic diagram showing the performance comparison between the method provided by the present invention and the traditional method under different signal-to-noise ratios;

[0076] Figure 4 Schematic diagram showing the performance comparison between the method provided by the present invention and the traditional method under different snapshot numbers;

[0077] Figure 5 This is a schematic diagram comparing the calculation time of the method provided by the present invention and the traditional method under different snapshot numbers. DETAILED DESCRIPTION

[0078] The technical solution provided by the present invention is further illustrated below with reference to embodiments.

[0079] The following describes a method for obtaining the target wave direction for a depth-scanning side-scan sonar provided by the present invention in conjunction with the accompanying drawings and specific examples. The method proposes an improved ESPRIT algorithm combined with noise subspace compensation. Figure 1 The flowchart of the method of the present invention is given. Figure 2 As shown in FIG, a uniform linear array with M elements is divided into two identical sub-arrays, each of which includes M-1 elements. Specifically, the following steps are included:

[0080] There is a uniform linear array composed of M array elements, and the spacing between the array elements is d. L far-field narrowband signals are sent at angles θ1, θ2, ...θ l ,…,θ L The wavelength of the signal incident on this linear array is λ, and the data matrix received by the entire array is X.

[0081] X=AS+N

[0082] Where A is the array manifold matrix composed of steering vectors, A=[a1,a2,…,a l ,…,a L ], where a l is the lth steering vector, l is an ordinal number, l = 1, 2, ... L, j is an imaginary number, S is a matrix of signal amplitudes, and N is additive white Gaussian noise.

[0083] The receiving array is divided into two identical and overlapping sub-arrays, each with M-1 elements. Assume that the data received by the first sub-array is the first data matrix X1, and the data received by the second sub-array is the second data matrix X2. For the same signal, the phase difference between the signals received by the first sub-array and the second sub-array is only φ l ,in The first data matrix X1 is composed of the first M-1 columns of the data matrix X received by the entire array, and the second data matrix X2 is composed of the last M-1 columns of the data matrix X received by the entire array, that is:

[0084] X1=[I M-1 |0 M-1 ]X

[0085] X2=[0 M-1 |I M-1 ]X

[0086] Among them, I M-1 represents the (M-1)×(M-1) identity matrix, O M-1 Represents the (M-1)×1 zero vector.

[0087] 2. Estimate the covariance matrix R based on the data matrix received by the entire array X

[0088]

[0089] Where K is the number of snapshots, X is the data matrix received by the entire array, (·) H Represents the conjugate transpose of a matrix.

[0090] 3. Covariance matrix R X Perform eigendecomposition to obtain the signal subspace E S and noise subspace E N

[0091]

[0092] The covariance matrix R of the actual received data X Perform eigenvalue decomposition and arrange the eigenvalues ​​from large to small. The eigenvectors related to the first L large eigenvalues ​​constitute the signal subspace E S , the eigenvectors associated with the remaining ML small eigenvalues ​​constitute the noise subspace E N ,∑ S and ∑ N are diagonal matrices consisting of signal eigenvalues ​​and noise eigenvalues ​​respectively.

[0093] 4. ESPRIT algorithm solves the rough value θ of the estimated angle gl

[0094]

[0095] z l is the eigenvalue vector corresponding to the L large eigenvalues, z l Specifically obtained from the following:

[0096] The signal subspace E S Split into two subspaces, the first subspace E S1 is the signal subspace corresponding to the first subarray; the second subspace ES2 is the signal subspace corresponding to the second subarray, that is:

[0097] E S1 =[I M-1 |0 M-1 ]E S

[0098] E S2 =[0 M-1 |I M-1 ]E S

[0099] Ψ=E S1 + E S2

[0100] I M-1 represents the (M-1)×(M-1) identity matrix, 0 M-1 represents the (M-1)×1 zero vector;

[0101] (·) + Indicates the pseudo-inverse, performs eigenvalue decomposition on the first process matrix Ψ, and obtains the eigenvalue vector z corresponding to the L largest eigenvalues l , l=1, 2,…L.

[0102] 5. The rough value θ of the estimated angle obtained by the ESPRIT algorithm gl Substitute the signal steering vector and its derivative to obtain the vector A consisting of the estimated angle rough value steering vector g and the estimated angle rough value steering vector A g The derivative vector B g

[0103]

[0104] where a gl is the steering vector for estimating the rough angle value, l=1,2,…L,b gl for a gl The derivative value, b gl =a' gl ,θ gl Get a rough value of the estimated angle for the ESPRIT algorithm.

[0105] 6. Estimated angle rough value guide vector A g and its derivative vector B g , noise subspace E N Substitute into the formula to obtain the deviation value δ of the estimated angle l

[0106]

[0107] Re{·} is the real part of the complex number.

[0108] The specific derivation is as follows:

[0109] Taylor expansion is performed on the signal steering vector at the estimated angle coarse value position, and the total steering vector is expressed as (A e +B e Δ), A e =[a g1 ,a g2 ,…,a gl ,…,a gL ], A e is the matrix of steering vectors in the coarse value direction, B e =[a g1 ′,a g2 ′…,a gl ′,…,a gL ′],B e A e The derivative matrix of , let Δ be a diagonal matrix, is the deviation of the estimated angle, Δ=diag[δ1,δ2…,…,δ l ,…,δ L ], δ l =θ l -θ gl ,θ l is the angle of the lth far-field narrowband signal, θ gl Solve the ESPRIT algorithm to obtain a rough value of the estimated angle;

[0110] E N is the noise subspace. According to the orthogonality of the signal-noise subspace and the steering vector, the deviation Δ of the estimated angle can be obtained as:

[0111]

[0112] Tr(·) is the trace of the matrix, let D be a block diagonal matrix, D = diag[b g1 ,b g2 ,…,b gl ,…,b gL ], this formula is equivalent to:

[0113]

[0114] in, represents the Kronecker product, I L is an L×L unit matrix, δ is a vector consisting of L deviation values, δ=[δ1,δ2…,δ l ,…,δ L ] T, δ l is the deviation value of the estimated angle.

[0115] Let the first process value The optimal solution can be found as:

[0116] δ=-(Re{D H UD}) -1 ·(Re{D H UA g})

[0117] Re{·} is the real part of the complex number. After simplifying the formula, the deviation value of the estimated angle is:

[0118]

[0119] 7. Obtain the angle estimate θ of the ESPRIT correction algorithm l =θ gl +δ l ,θ gl The rough value θ of the estimated angle obtained by the ESPRIT algorithm gl , δ l is the deviation value of the estimated angle.

[0120] Assume that the rough angle value estimated by the ESPRIT algorithm is θ gl , then the angle estimation value θ obtained by the present invention is l is the rough value of the estimated angle θ gl Deviation δ from the estimated angle l Applying it to the depth-scanning sonar, we get the estimated angle θ l Then estimate the water depth.

[0121] Figure 3 The figure compares the direction-finding performance of the method of the present invention and the traditional method under different signal-to-noise ratios. The simulation parameter conditions are: the number of uniform linear array elements is 10, two coherent signal sources are incident from 20° and 50° respectively, the number of snapshots is 50, and the element spacing is half a wavelength. The signal-to-noise ratio range is -5dB to 20dB, with a step size of 5dB. Based on the above experimental conditions, 1000 Monte Carlo simulation experiments were carried out to estimate the incident angle and calculate the mean square error. It can be seen from the figure that under different signal-to-noise ratios, the error of the method of the present invention is almost consistent with that of the Root-MUSIC algorithm and does not require complex peak search. The direction-finding accuracy is better than that of the ESPRIT algorithm.

[0122] Figure 4The direction-finding performance of the method of the present invention is compared with that of the traditional method under different snapshot numbers. The simulation conditions are: the number of uniform linear array elements is 10, two independent signal sources are incident from 20° and 50° respectively, the signal-to-noise ratio is 10dB, and the element spacing is half a wavelength. The number of snapshots varies from 25 to 150, with a step size of 25. Based on the above experimental conditions, 1000 Monte Carlo simulation experiments were performed to estimate the incident angle and calculate the mean square error. As can be seen from the figure, under different snapshot numbers, the error of the method of the present invention is almost consistent with that of the Root-MUSIC algorithm and does not require complex peak search. The direction-finding accuracy is better than that of the ESPRIT algorithm.

[0123] Figure 5 The following figure compares the computation time of the proposed method and the traditional method under different snapshot numbers. The simulation conditions are: a uniform linear array with 10 elements, two independent signal sources incident at angles of 20° and 50°, a signal-to-noise ratio of 10 dB, an element spacing of half a wavelength, and 50 snapshots. Based on these experimental conditions, 1000 experiments were performed and the average computation time was calculated. It can be seen that the computation time of the proposed method is slightly longer than that of the ESPRIT algorithm, but much shorter than that of the Root-MUSIC algorithm.

[0124] Finally, it should be noted that the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention and are intended to be encompassed by the claims of the present invention.

Claims

1. A method for obtaining a target wave direction for a depth-scanning side-scan sonar, comprising: Step 1: Estimate the covariance matrix R of the data matrix X received by the full receiving array X ; Step 2: Covariance matrix R X Perform eigendecomposition to obtain the signal subspace E S and noise subspace E N ; Step 3: Using the ESPRIT algorithm, based on the signal subspace E S , get the estimated angle rough value θ of the target gl ; Step 4: Estimate the rough angle θ gl Substitute the signal-directed vector for first-order Taylor expansion and use the noise subspace E N The orthogonality of the signal steering vector is used to obtain the estimated angle deviation value δ l ; Estimate the rough angle value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is taken as the angle estimation value θ l ; The step 2 specifically includes: X Perform eigendecomposition, arrange the eigenvalues ​​from large to small, and the eigenvectors related to the first L large eigenvalues ​​form the signal subspace E S , the eigenvectors associated with the remaining ML small eigenvalues ​​constitute the noise subspace E N , where M is the number of elements in the full-receiving array, and L is the number of far-field narrowband signals incident on the full-receiving array; in,(·) H represents the conjugate transpose of the matrix, Σ S is the diagonal matrix composed of signal eigenvalues, Σ N is a diagonal matrix composed of noise eigenvalues; The step 3 specifically includes: Step 3-1: Divide the full receiving array into a first sub-array and a second sub-array that are identical and overlap each other; divide the signal subspace E S Split into the first subspace E S1 and the second subspace E S2 ; Among them, the first subspace E S1 is the signal subspace corresponding to the first subarray; the second subspace E S2 is the signal subspace corresponding to the second subarray: E S1 =[I M-1 |0 M-1 ]E S E S2 =[0 M-1 |I M-1 ]E S Among them, I M-1 represents the (M-1)×(M-1) identity matrix, 0 M-1 represents the (M-1)×1 zero vector; Step 3-2: Using the first subspace E S1 and the second subspace E S2 , obtain the first process matrix Ψ: Ψ=E S1 + AND S2 in,(·) + It means seeking pseudo-inverse; Step 3-3: Perform eigenvalue decomposition on the first process matrix Ψ to obtain the eigenvalue vector z corresponding to the L largest eigenvalues l ; Step 3-4: Based on the eigenvalue vector z corresponding to the L largest eigenvalues l , use the ESPRIT algorithm to solve and get the rough value of the estimated angle θ gl : Where λ is the wavelength of the far-field narrowband signal.

2. The method for obtaining the target wave direction for a depth-scanning sonar according to claim 1, characterized in that: The step 1 specifically includes: estimating the covariance matrix R according to the data matrix received by the full receiving array X : Where K is the number of snapshots, X is the data matrix received by the entire array, (·) H Represents the conjugate transpose of a matrix.

3. The method for obtaining the target wave direction for a depth-scanning sonar according to claim 1, characterized in that: The step 4 specifically includes: Step 4-1: The rough value θ of the estimated angle obtained by the ESPRIT algorithm gl Substitute the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g : Among them, a gl is the steering vector of the lth estimated angle coarse value; Step 4-2: The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm gl Substitute the derivative of the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g The derivative vector B g : Among them, b gl is the steering vector a of the lth estimated angle rough value gl The derivative value of Step 4-3: Vector A formed by estimating the angle coarse value steering vector g and its derivative vector B g and noise subspace E N , get the estimated angle deviation value δ l : Re{·} is the real part of the complex number; Step 4-4: Estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is taken as the angle estimation value θ l .

4. A system for obtaining the direction of incoming target waves for a depth-finding side-scan sonar, comprising: Covariance module, used to estimate the covariance matrix R of the data matrix X received by the full receiving array X ; Decomposition module for the covariance matrix R X Perform eigendecomposition to obtain the signal subspace E S and noise subspace E N ; ESPRIT module, through the ESPRIT algorithm, based on the signal subspace E S , get the estimated angle rough value θ of the target gl ; and Estimation module, used to estimate the rough angle value θ gl Substitute the signal-directed vector into the first-order Taylor expansion to use the noise subspace E N The orthogonality of the signal steering vector is used to obtain the estimated angle deviation value δ l ; Used to obtain the estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is used as the angle estimate θ l ; The decomposition module is specifically used to: X Perform eigendecomposition, arrange the eigenvalues ​​from large to small, and the eigenvectors related to the first L large eigenvalues ​​form the signal subspace E S , the eigenvectors associated with the remaining ML small eigenvalues ​​constitute the noise subspace E N , where M is the number of elements in the full-receiving array, and L is the number of far-field narrowband signals incident on the full-receiving array; in,(·) H represents the conjugate transpose of the matrix, ∑ S is the diagonal matrix composed of signal eigenvalues, ∑ N is a diagonal matrix composed of noise eigenvalues; The ESPRIT module is specifically used to: The full receiving array is divided into a first sub-array and a second sub-array that are completely identical and overlap each other; the signal subspace E S Split into the first subspace E S1 and the second subspace E S2 ; Among them, the first subspace E S1 is the signal subspace corresponding to the first subarray; the second subspace E S2 is the signal subspace corresponding to the second subarray: E S1 =[I M-1 |0 M-1 ]E S E S2 =[0 M-1 |I M-1 ]E S Among them, I M-1 represents the (M-1)×(M-1) identity matrix, 0 M-1 represents the (M-1)×1 zero vector; Using the first subspace E S1 and the second subspace E S2 , obtain the first process matrix Ψ: Ψ=E S1 + AND S2 in,(·) + It means seeking pseudo-inverse; Perform eigenvalue decomposition on the first process matrix Ψ to obtain the eigenvalue vector z corresponding to the L largest eigenvalues l ; Based on the eigenvalue vector z corresponding to the L large eigenvalues l , use the ESPRIT algorithm to solve and get the rough value of the estimated angle θ gl : Where λ is the wavelength of the far-field narrowband signal.

5. The system for obtaining target wave direction for bathymetric side-scan sonar according to claim 4, characterized in that: The covariance module is specifically used to obtain the covariance matrix R according to the data matrix received by the full receiving array X : Where K is the number of snapshots, X is the data matrix received by the entire array, (·) H Represents the conjugate transpose of a matrix.

6. The system for obtaining target wave direction for bathymetric side-scan sonar according to claim 4, characterized in that: The estimation module is specifically used for: The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm gl Substitute the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g : Among them, a gl is the steering vector of the lth estimated angle coarse value; The rough value θ of the estimated angle obtained by solving the ESPRIT algorithm gl Substitute the derivative of the signal steering vector to obtain the vector A consisting of the estimated angle rough value steering vector g The derivative vector B g : Among them, b gl is the steering vector a of the lth estimated angle rough value gl The derivative value of Vector A formed by estimating the angle rough value steering vector g and its derivative vector B g and noise subspace E N , get the estimated angle deviation value δ l : Re{·} is the real part of the complex number; Get the estimated angle rough value θ gl Deviation from the estimated angle δ l The sum of the target wave direction is used as the angle estimate θ l .

Images