Anti-bottom echo interference target detection method and system suitable for multi-beam sounding sonar

Abstract

The invention belongs to the technical field of underwater sound, and particularly relates to an anti-bottom echo interference target detection method and system suitable for a multi-beam sounding sonar, and the method comprises the steps: constructing a binary sparse hypothesis test problem based on to-be-detected data and a constructed beam domain sparse signal model, the to-be-detected data is beam data corresponding to a certain sampling moment in the current frame; solving the background noise power, and solving the sparse vector based on the block sparse characteristic of the target and Bayesian learning; and judging whether a target exists in the to-be-detected data or not by using a generalized likelihood ratio detector. In the design process, priori information such as a beam data model and block sparse characteristics of a target is considered and utilized, so that the priori knowledge of the target is fully mined, and the sensing capability of the sonar to the environment is improved; interference suppression and target detection can be realized only based on the beam data of the current frame without using auxiliary data of front and back frames; parameters needing manual adjustment do not exist, and the method can be directly applied to an unmanned platform.

Description

Technical Field

[0001] The present invention belongs to the field of underwater acoustic technology, and in particular relates to a bottom echo interference resistant target detection method and system suitable for multi-beam bathymetric sonar. Background Art

[0002] With the widespread application of multi-beam bathymetric sonar in underwater engineering surveying, underwater search and rescue, and submarine pipeline monitoring, the demand for automatic target detection using beam data has been increasing in recent years. Beam data is obtained by beamforming the array element data at multiple preset angles. It can intuitively reflect the two-dimensional scattered sound field in the sector covered by the multi-beam bathymetric sonar transmission beam and save the position and shape information of the target. However, the beam data contains not only target echoes, but also seabed scattered echoes. The target echo close to the seabed will be in the background of seabed scattered echoes, which will interfere with target detection. Specifically, the sidelobes of the seabed scattered echoes are high, which is easy to cause false alarms, and at the same time, it shields weak targets and causes missed detections. How to improve the detection performance of targets under the interference of bottom echoes has become an urgent problem to be solved.

[0003] In order to realize automatic target detection in beam data, researchers have introduced the Constant False Alarm Rate (CFAR) technology into multi-beam bathymetric sonar target detection. Among them, the more classic ones are Cell-Averaged CFAR (CA-CFAR), Cell-Averaging Greatest-ofselection CFRA (CAGO-CFAR), Cell-Averaging Smallest-ofselection CFRA (CASO-CFAR) and Order Statistic CFAR (OS-CFAR). These classic methods directly use auxiliary data to estimate the background noise power, and then judge a single data point through an adaptive threshold. None of them considers the impact of bottom echo and its sidelobes on target detection. In response to the above problems, relevant scholars proposed an adaptive soft threshold denoising algorithm (ASTDA) to achieve target detection under bottom echo interference. The algorithm performs threshold segmentation through the maximum inter-class variance, divides the beam data into two parts: the target noise mixed area and the background area, then determines the sampling point number interval where the target is located based on the heading map, separates the background area from the target noise mixed area in the beam array map, and eliminates the background area data. Finally, the sidelobe noise position is determined through an adaptive soft threshold to eliminate the sidelobe noise to achieve target detection.

[0004] In practical applications, ASTDA has three obvious shortcomings: First, when determining the sampling point number interval where the target is located, a large amount of beam data before the frame is needed for statistical analysis. However, the underwater environment is very complex. If there is interference in the previous frame data, such as swimming fish and terrain changes, it will lead to statistical errors in the target interval; second, sidelobe suppression only depends on threshold selection, and targets with echo energy lower than the bottom echo sidelobe will be missed; third, there are parameters in the threshold setting that need to be manually adjusted, which will increase the difficulty of system operation and is not suitable for unmanned platforms. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and propose a target detection method and system for multi-beam bathymetric sonar with anti-bottom echo interference.

[0006] On the one hand, the present invention provides a method for detecting a target with anti-bottom echo interference applicable to a multi-beam bathymetric sonar, comprising:

[0007] Step 1: Based on the data to be detected and the constructed beam domain sparse signal model, a binary sparse hypothesis test problem is constructed, wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on the two-step generalized likelihood ratio criterion;

[0008] Step 2: solving the background noise power, and solving the sparse vector based on the block sparse characteristics of the target and Bayesian learning;

[0009] Step 3: Based on the solved background noise power and the sparse vector, a generalized likelihood ratio detector is used to determine whether there is a target in the data to be detected.

[0010] Preferably, the beam-domain sparse signal model is constructed in the following manner:

[0011] Establishing a response function of a far-field target to a multi-beam bathymetric sonar receiving array, wherein the multi-beam bathymetric sonar receiving array is a multi-element equidistant linear array;

[0012] Discretize the echo angle into multiple angle units and construct a complete beam-domain steering vector dictionary;

[0013] According to the constructed complete beam-domain steering vector dictionary and the established response function, a beam-domain sparse signal model is constructed.

[0014] The optimal beam-domain sparse signal model is:

[0015] x=Ψw+n

[0016] Among them, x is the data to be detected, Ψ is the observation matrix, w is the sparse vector, n is the noise vector, I is the identity matrix, is the background noise power spectral density.

[0017] Preferably, the binary sparse hypothesis testing problem constructed in step 1 is:

[0018]

[0019] Among them, H0 and H1 represent the hypothesis of no target and the hypothesis of target existence, respectively, x is the data to be detected, n is the noise vector, z k is auxiliary data, n k is the noise vector in the auxiliary data, assuming n k They satisfy the independent and identically distributed conditions, k is the auxiliary data index, ranging from 1 to K, Ψ is the observation matrix, and w is the sparse vector.

[0020] Preferably, the generalized likelihood ratio detector designed in step 3 is:

[0021]

[0022] in, and are the probability density functions of the data x to be detected under the hypothesis of no target H0 and the hypothesis of target existence H1, respectively. is the background noise power and w is the sparse vector, η is the detection threshold.

[0023] Preferably, the step 2 solves the sparse vector based on the block sparse characteristics of the target and the Bayesian formula, including:

[0024] The beam domain sparse signal model is expanded by a sliding window method to obtain an expanded beam domain sparse signal model, wherein the expanded beam domain sparse signal model includes a new block sparse vector, and the new block sparse vector has a known block structure;

[0025] Solve for a new block sparse vector;

[0026] According to the matrix operation, the sparse vector is restored using the solved new block sparse vector.

[0027] Better, solve the new block sparse vector, including:

[0028] Assuming that the new block sparse vector obeys Gaussian distribution, and there is correlation between elements in each block, and different blocks are independent of each other, the posterior distribution function of the new block sparse vector is established according to the Bayesian formula, wherein the established posterior distribution function contains hyperparameters;

[0029] estimating the hyperparameters using a second-kind maximum likelihood algorithm;

[0030] Substitute the estimated hyperparameter into the posterior mean of the established sparse vector to obtain a new block sparse vector.

[0031] Preferably, the posterior distribution function of the new block sparse vector is established as:

[0032]

[0033] in,

[0034]

[0035] Λ=diag{ξ1B1,ξ2B2,…,ξ P-h+1 B P-h+1}

[0036] Among them, w E is a new block sparse vector, x is the data to be detected, ξ i and B i are the i-th first and second hyperparameters respectively, i is the index of the observation matrix block, P is the number of angle units discretized into echo angles, h is the sliding window length, μ is the mean vector of the new block sparse vector, ∑ is the covariance matrix of the new block sparse vector, Λ is the diagonal matrix, Ψ E To expand the observation matrix, is the background noise power spectral density, I is the unit matrix, H represents the conjugate transpose, and -1 represents the inverse matrix.

[0037] Preferably, the hyperparameters are estimated using a second-class maximum likelihood algorithm, including:

[0038] Establish the following cost function:

[0039]

[0040] in, represents a hyperparameter;

[0041] The above cost function is optimized and solved using the expectation maximization algorithm to obtain the hyperparameters.

[0042] On the other hand, a system based on the above-mentioned method for detecting a target with anti-bottom echo interference suitable for multi-beam bathymetric sonar is provided, comprising: a problem construction unit, a solution unit and a detection unit; wherein:

[0043] The problem construction unit is used to construct a binary sparse hypothesis test problem based on the data to be detected and the constructed beam domain sparse signal model, wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on a two-step generalized likelihood ratio criterion;

[0044] The solving unit is used to solve the background noise power and solve the sparse vector based on the block sparse characteristics of the target and Bayesian learning;

[0045] The detection unit is used to use a generalized likelihood ratio detector to determine whether a target exists in the data to be detected based on the solved background noise power and sparse vector, wherein the generalized likelihood ratio detector is designed and implemented based on a two-step generalized likelihood ratio criterion.

[0046] Compared with the prior art, the advantages of the present invention are:

[0047] 1. In the design process, the beam data model and the block sparse characteristics of the target are considered to fully explore the prior knowledge of the target and improve the sonar's perception of the environment;

[0048] 2. The mathematical model of beam data is integrated into the design of the detector, and the sparse recovery algorithm is used to greatly improve the ability to suppress the interference of the bottom echo sidelobe;

[0049] 3. Interference suppression and target detection can be achieved based only on the beam data of the current frame, without the need to use auxiliary data from previous and subsequent frames;

[0050] 4. There are no parameters that need to be adjusted manually. While ensuring the target detection performance, it also has CFAR performance and can be directly applied to unmanned platforms. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of a target detection method for resisting bottom echo interference applicable to multi-beam bathymetric sonar according to an embodiment of the present invention;

[0052] Figure 2 It is a flow chart of another method for detecting a target with anti-bottom echo interference applicable to a multi-beam bathymetric sonar according to an embodiment of the present invention;

[0053] Figure 3 is a distance dimension detection performance curve diagram of an embodiment of the present invention;

[0054] Figure 4 is a graph showing the angle dimension detection performance of an embodiment of the present invention;

[0055] Figure 5 4 is a false alarm probability curve diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In the scheme of the embodiment of the present invention, a beam domain sparse signal model is first constructed, and then a sparse binary hypothesis test problem is designed. Finally, the hypothesis test problem is solved and judged using a two-step generalized likelihood ratio test (GLRT) criterion. In order to improve the target detection performance under bottom echo interference, the present invention considers the prior knowledge of the block sparsity of the target during the detector design process, integrates it into the test statistic, and realizes the effective use of the prior information. In fact, the introduction of block sparsity makes the sparse recovery of the target no longer based on a single beam angle, but is achieved within a continuous angle range, avoiding the performance loss caused by reduced sparsity, while ensuring that the recovered target has a relatively complete outline.

[0057] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0058] like Figure 1 As shown, it is a flow chart of a method for detecting a target with anti-bottom echo interference applicable to a multi-beam bathymetric sonar according to an embodiment of the present invention, comprising:

[0059] Step 1: Based on the data to be detected and the constructed beam domain sparse signal model, a binary sparse hypothesis test problem is constructed; wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on the two-step generalized likelihood ratio criterion;

[0060] Step 2: solving the background noise power, and solving the sparse vector based on the block sparse characteristics of the target and Bayesian learning;

[0061] Step 3: Based on the solved background noise power and the sparse vector, a generalized likelihood ratio detector is used to determine whether there is a target in the data to be detected.

[0062] The detailed design process of the present invention is as follows:

[0063] 1. Problem Description

[0064] 1.1 Sparse Signal Model

[0065] Consider the multi-beam bathymetric sonar receiving array as an N-element equally spaced linear array, with a preset beam angle of Θ M ={θ1,θ2,...,θ M}, assuming that there is a point at an angle θ in the far field of the array i The goal is to MThe response function on can be expressed as formula (1):

[0066] x=α i Ψ(u i )(1) Where u i = sinθ i ,

[0067]

[0068] Where M is the number of preset beam angles, i is the target angle index, x is the data to be detected, Ψ(u i ) is the angle θ i is the response vector of the target at the preset beam angle, m is the preset beam angle index, ψ m (.) is the angle θ i The target response at the mth preset beam angle is d, λ is the wavelength, and N is the number of array elements.

[0069] Discretize the echo angle into P angle units θ P ={θ1,θ2,...,θ P}, then we can construct an overcomplete beam-domain steering vector dictionary as shown in formula (3):

[0070]

[0071] According to the dictionary, the beam domain sparse signal model shown in formula (4) can be obtained:

[0072] x=Ψw+n (4)

[0073] Among them, Ψ is the observation matrix, w is the sparse vector, and n is the noise vector, which obeys the complex Gaussian distribution, that is, I is the identity matrix, is the background noise power spectral density.

[0074] 1.2 Hypothesis Testing Problem

[0075] According to the above sparse signal model, the following binary sparse hypothesis test problem can be constructed:

[0076]

[0077] Among them, H0 and H1 represent the hypothesis of no target and the hypothesis of target existence, respectively, x is the data to be detected, n is the noise vector, z k is auxiliary data, n k is the noise vector in the auxiliary data, assuming n kThey satisfy the independent and identically distributed conditions, k is the auxiliary data index, ranging from 1 to K, K is the number of auxiliary data, Ψ is the observation matrix, and w is the sparse vector.

[0078] 2. Detector design

[0079] For problem (5), a two-step GLRT criterion is used to design an adaptive detection scheme. The criterion consists of two parts: 1) Assuming the noise power and sparse vector w are known, design a GLRT detector, 2) use the auxiliary data and the data to be detected to estimate the noise power and sparse vector w respectively, and use the estimated values ​​to replace the detector and w.

[0080] The probability density function of x under the two assumptions can be expressed as shown in formula (6) and formula (7) respectively:

[0081]

[0082] The GLRT detector can be expressed as shown in formula (8):

[0083]

[0084] Wherein, η is the detection threshold.

[0085] The noise power can be estimated by the least squares method shown in formula (9):

[0086]

[0087] Where M is the number of preset beam angles, K is the number of auxiliary data, k is the auxiliary data index, ‖z k ‖2 represents z k The second-order norm of .

[0088] Next, w is estimated based on the block sparseness of the target. Considering that the block structure in w is unknown, the observation model is first expanded by the sliding window method, as shown in formula (10):

[0089] x=Ψ E w E +n(10)

[0090] in To expand the observation matrix, is the i-th observation matrix block, is the new block sparse vector, is the i-th block vector. E It has two advantages. One is E There is a known block structure that can be used for block sparse recovery.E The dimensions of each block vector are the same, which is beneficial for subsequent operations. According to matrix operations, w can be restored by the following formula (11):

[0091]

[0092] in is a dimension reduction matrix.

[0093] In order to w E Perform sparse recovery, assuming w E It obeys Gaussian distribution, and there is correlation between elements in each block, and different blocks are independent of each other, that is, in:

[0094] Λ=diag{ξ1B1,ξ2B2,…,ξ P-h+1 B P-h+1}(12)

[0095] where ξ i and B i are the first and second hyperparameters of the ith hyperparameter, both unknown. According to the Bayesian formula, w E The posterior distribution of

[0096]

[0097] in

[0098]

[0099] ξ i and B i It can be estimated by the second type of maximum likelihood algorithm. This estimation is equivalent to minimizing the following cost function

[0100]

[0101] in represents the unknown hyperparameter. Next, the expectation maximization (EM) algorithm is used to optimize and solve equation (16). E As a hidden variable in the EM algorithm, the (r+1)th parameter update can be expressed as shown in the following formula (17):

[0102]

[0103] Among them, f(w E |x,Θ (r) ) is w E The posterior distribution of f(x,w E |Θ) represents x and w E The joint distribution ofi and B i are coupled within this distribution. It is relatively complicated to directly use formula (17) to optimize Θ. In order to simplify the optimization process, the Bayesian formula can be used to rewrite the objective function as shown in formula (18):

[0104]

[0105] Since the first term in formula (18) is independent of Θ, the optimization of Θ can be expressed by formula (19):

[0106]

[0107] By solving formula (20) and formula (21), we can get ξ i and B i The update process:

[0108]

[0109] It should be noted that i and B i The solution needs to be obtained through an iterative process. After the algorithm converges, μ is w E The estimated value of w can be obtained by using formula (11). In the subsequent simulation, the detector is referred to as BSBL-GLRT method. The above process is to substitute the estimated hyperparameters into the posterior mean of the established sparse vector to obtain a new block sparse vector.

[0110] The specific execution process of the above method can be as follows Figure 2 As shown, the following steps are included:

[0111] (a) Input discretized angle set Θ P . By Θ P An over-complete beam-domain steering vector dictionary Ψ is constructed, and the block-sparse property of the target is utilized to expand the sparse recovery model using formula (10).

[0112] (b) Using auxiliary data z k The background noise power is estimated by formula (9): Initialize the hyperparameters ξ0 and B0, and calculate the mean vector μ and covariance matrix Σ using formulas (14) and (15).

[0113] (c) Construct the optimization problem shown in formula (16), and use formula (20) and formula (21) to calculate ξ using the EM algorithm, with ξ0 and B0 as the initial values. i and B i to update.

[0114] (d) If Go to step (e); otherwise, update μ and ∑ using formula (14) and formula (15), and go to step (c).

[0115] (e) Restore the sparse vector w according to formula (11).

[0116] (f) Obtain the judgment result according to formula (8).

[0117] 3. Performance Analysis

[0118] The detection performance of the BSBL-GLRT method is analyzed by Monte Carlo simulation method. In the simulation, the receiving array is set as a half-wavelength uniform linear array, the number of array elements N = 64, the sonar operating frequency f0 = 300kHz, the preset beam angle number M = 128, uniformly distributed in the sinθ domain [-1, 1], and the corresponding angle domain range is [-90°, 90°]. Gaussian white noise is added to each channel, and the noise between different channels is independent of each other.

[0119] First, consider the detection of a certain time slice data. In the simulation, the false alarm probability P is set fa =10 -3 , the number of auxiliary data K = 10, through 100 / P fa The detection threshold was calculated by Monte Carlo simulation. 3 Monte Carlo simulation statistical detection probability P d The signal-to-noise ratio is defined as

[0120] Figure 3 The detection performance curves of BSBL-GLRT and ASTDA in the distance dimension are given. It can be seen from the figure that the detection probability of both algorithms increases with the increase of SNR. When SNR = -15dB, the detection probability of BSBL-GLRT method is about 23% higher than that of ASTDA. When SNR reaches -10dB, the detection probability of both methods can reach 100%. Figure 4 The detection performance curves of BSBL-GLRT and ASTDA in the angle dimension are given. It can be seen from the figure that the detection probability of BSBL-GLRT is higher than that of ASTDA, and it has better target detection performance. Figure 5The false alarm probability curves of the two algorithms in angle dimension detection are given. As can be seen from the figure, the false alarm probability of BSBL-GLRT does not change with SNR, and has CFAR performance, but the false alarm probability of ASTDA increases with the increase of SNR, and does not have CFAR performance. The reason why BSBL-GLRT has CFAR performance is that it can better suppress the bottom echo and target side lobe, reducing the influence of the side lobe on target detection. The reason why ASTDA does not have CFAR performance is that its ability to suppress side lobes is weak. As SNR increases, the side lobe value becomes larger, and the side lobe is also detected as a target, resulting in false alarms. The simulation results show that the method proposed by the present invention can effectively resist the interference of bottom echo and side lobe, and has good target detection performance.

[0121] When using beam data for target detection, the bottom echo and its side lobes will interfere with target detection. Targets in the interference area are easily submerged by the bottom echo side lobes, resulting in poor target detection performance. ASTDA uses multi-frame auxiliary data to remove bottom echo interference, but has high requirements on the quality of auxiliary data, has poor bottom echo side lobe suppression capability, and requires manual real-time parameter adjustment. The embodiment of the present invention proposes a constant false alarm target detection method based on block sparse Bayesian learning. During the design process, this method considers the use of prior information such as the beam data model and the block sparse characteristics of the target. It can suppress bottom echo interference without relying on other frame auxiliary data, does not require manual parameter adjustment, and has CFAR performance while ensuring target detection performance.

[0122] In addition, an embodiment of the present invention further provides a system based on the above-mentioned anti-bottom echo interference target detection method applicable to multi-beam bathymetric sonar, comprising: a problem construction unit, a solution unit and a detection unit, wherein:

[0123] The problem construction unit is used to construct a binary sparse hypothesis test problem based on the data to be detected and the constructed beam domain sparse signal model, wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on a two-step generalized likelihood ratio criterion;

[0124] The solving unit is used to solve the background noise power and solve the sparse vector based on the block sparse characteristics of the target and Bayesian learning;

[0125] The detection unit is used to determine whether there is a target in the data to be detected based on the solved background noise power and sparse vector by using a generalized likelihood ratio detector.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A method for detecting a target with anti-bottom echo interference suitable for a multi-beam bathymetric sonar, comprising: Step 1: Based on the data to be detected and the constructed beam domain sparse signal model, a binary sparse hypothesis test problem is constructed, wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on the two-step generalized likelihood ratio criterion; Step 2: solving the background noise power, and solving the sparse vector based on the block sparse characteristics of the target and Bayesian learning; Step 3: Based on the solved background noise power and the sparse vector, a generalized likelihood ratio detector is used to determine whether there is a target in the data to be detected.

2. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 1, characterized in that: The beam-domain sparse signal model is constructed in the following manner: Establishing a response function of a far-field target to a multi-beam bathymetric sonar receiving array, wherein the multi-beam bathymetric sonar receiving array is a multi-element equidistant linear array; Discretize the echo angle into multiple angle units and construct a complete beam-domain steering vector dictionary; According to the constructed complete beam-domain steering vector dictionary and the established response function, a beam-domain sparse signal model is constructed.

3. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 2, characterized in that: The constructed beam domain sparse signal model is: x=Ψw+n Among them, x is the data to be detected, Ψ is the observation matrix, w is the sparse vector, n is the noise vector, I is the identity matrix, is the background noise power spectral density.

4. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 1, characterized in that: The binary sparse hypothesis testing problem constructed in step 1 is: Among them, H0 and H1 represent the hypothesis of no target and the hypothesis of target existence, respectively, x is the data to be detected, n is the noise vector, z k is auxiliary data, n k is the noise vector in the auxiliary data, assuming n k They satisfy the independent and identically distributed conditions, k is the auxiliary data index, ranging from 1 to K, Ψ is the observation matrix, and w is the sparse vector.

5. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 1, characterized in that: The generalized likelihood ratio detector designed in step 3 is: in, and are the probability density functions of the data x to be detected under the hypothesis of no target H0 and the hypothesis of target existence H1, respectively. is the background noise power and w is the sparse vector, η is the detection threshold.

6. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 1, characterized in that: In step 2, solving the sparse vector based on the block sparse characteristics of the target and the Bayesian formula includes: The beam domain sparse signal model is expanded by a sliding window method to obtain an expanded beam domain sparse signal model, wherein the expanded beam domain sparse signal model includes a new block sparse vector, and the new block sparse vector has a known block structure; Solve for a new block sparse vector; According to the matrix operation, the sparse vector is restored using the solved new block sparse vector.

7. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 6, characterized in that: Solve for a new block sparse vector, including: Assuming that the new block sparse vector obeys Gaussian distribution, and there is correlation between elements in each block, and different blocks are independent of each other, the posterior distribution function of the new block sparse vector is established according to the Bayesian formula, wherein the established posterior distribution function contains hyperparameters; estimating the hyperparameters using a second-kind maximum likelihood algorithm; Substitute the estimated hyperparameter into the posterior mean of the established sparse vector to obtain a new block sparse vector.

8. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 7, characterized in that: The posterior distribution function of the new block sparse vector is established as: in, Λ=diag{ξ1B1,ξ2B2,…,ξ P-h+1 B P-h+1 } Among them, w E is a new block sparse vector, x is the data to be detected, ξ i and B i are the i-th first and second hyperparameters respectively, i is the index of the observation matrix block, P is the number of angle units discretized into echo angles, h is the sliding window length, μ is the mean vector of the new block sparse vector, ∑ is the covariance matrix of the new block sparse vector, Λ is the diagonal matrix, Ψ E To expand the observation matrix, is the background noise power spectral density, I is the unit matrix, H represents the conjugate transpose, and -1 represents the inverse matrix.

9. The method for detecting targets with anti-bottom echo interference suitable for multi-beam bathymetric sonar according to claim 8, characterized in that: The hyperparameters are estimated using a second-class maximum likelihood algorithm, including: Establish the following cost function: in, represents a hyperparameter; The above cost function is optimized and solved using the expectation maximization algorithm to obtain the hyperparameters.

10. A system based on the method for detecting targets with anti-bottom echo interference applicable to multi-beam bathymetric sonar according to claim 1, characterized in that: include: A problem construction unit, a solution unit and a detection unit; wherein: The problem construction unit is used to construct a binary sparse hypothesis test problem based on the data to be detected and the constructed beam domain sparse signal model, wherein the data to be detected is the beam data corresponding to a certain sampling time in the current frame, and the test statistics of the binary sparse hypothesis test problem are background noise power and sparse vector, which are derived from a generalized likelihood ratio detector designed and implemented based on a two-step generalized likelihood ratio criterion; The solving unit is used to solve the background noise power and solve the sparse vector based on the block sparse characteristics of the target and Bayesian learning; The detection unit is used to use a generalized likelihood ratio detector to determine whether a target exists in the data to be detected based on the solved background noise power and sparse vector, wherein the generalized likelihood ratio detector is designed and implemented based on a two-step generalized likelihood ratio criterion.

Images