阅读说明：本技术 一种适用于侧扫声呐的自适应底混响抑制方法 (Self-adaptive bottom reverberation suppression method suitable for side-scan sonar ) 是由 许枫 马龙双 刘佳 于 2020-06-02 设计创作，主要内容包括：本发明公开了一种适用于侧扫声呐的自适应底混响抑制方法,所述方法包括：对接收到的每帧离散回波数据进行分段,分成多个数据段,其中,数据段长度为发射脉冲宽度的两倍,且相邻数据段1/2重叠；根据改进Burg算法估计每个数据段AR模型的阶数及系数；利用AR模型系数构建白化滤波器,利用该白化滤波器对数据段进行白化匹配滤波处理；对白化匹配滤波处理后的一帧离散回波数据进行相空间重构,构造二阶Hankel矩阵；对二阶Hankel矩阵进行多分辨二分奇异值分解,并通过所选奇异值进行回波重构完成目标回波与底混响的分离,实现底混响抑制。本发明的方法可以提高信混比,改善声图质量,对于实现侧扫声呐沉底静态小目标探测具有重要意义。(The invention discloses a self-adaptive bottom reverberation suppression method suitable for a side scan sonar, which comprises the following steps: segmenting each frame of received discrete echo data, dividing the discrete echo data into a plurality of data segments, wherein the length of each data segment is twice the width of a transmitting pulse, and adjacent data segments 1/2 are overlapped; estimating the order and the coefficient of the AR model of each data segment according to the improved Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter; performing phase space reconstruction on the discrete echo data of the frame after the whitening matched filtering processing to construct a second-order Hankel matrix; and performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation, thereby realizing the suppression of the bottom reverberation. The method can improve the signal-to-mixture ratio and the quality of the sonogram, and has important significance for realizing the detection of the side-scan sonar sinking-bottom static small target.)

1. An adaptive bottom reverberation suppression method suitable for a side scan sonar, the method comprising:

segmenting each frame of received discrete echo data, dividing the discrete echo data into a plurality of data segments, wherein the length of each data segment is twice the width of a transmitting pulse, and adjacent data segments 1/2 are overlapped;

estimating the order and the coefficient of the AR model of each data segment according to the improved Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter;

performing phase space reconstruction on the discrete echo data of the frame after the whitening matched filtering processing to construct a second-order Hankel matrix; and performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation, thereby realizing the suppression of the bottom reverberation.

2. The adaptive bottom reverberation suppression method suitable for the side scan sonar according to claim 1, characterized in that each frame of received discrete echo data is segmented into a plurality of data segments, wherein the length of each data segment is twice of the emission pulse width, and adjacent data segments 1/2 are overlapped; the method specifically comprises the following steps:

segmenting each frame of received discrete echo data, and dividing the discrete echo data into M data segments, wherein each data segment comprises 2s sampling points, namely the mth data segment is a sequence:

w_m(1),…w_m(s),…w_m(2s)

where t is 1 ≦ s, i.e., adjacent data segments 1/2 overlap.

3. The adaptive bottom reverberation suppression method suitable for the side scan sonar according to claim 2, wherein the order and the coefficient of each data segment AR model are estimated according to a modified Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter; the method specifically comprises the following steps:

when m is 1, for the 1 st data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 1 st data segment, k is more than or equal to 1 and less than or equal to p₁，p₁Order, u, of the AR model for the data segment₁(n) is a white noise sequence; n is more than or equal to 1 and less than or equal to 2 s;

construction of the 1 st whitening Filter with AR model coefficients, its transfer function H₁(z) is as follows:

wherein z is a variable;

whitening matched filtering processing is carried out on the 1 st data segment sequence through the 1 st whitening filter, and a one-dimensional whitening sequence w 'of the 1 st data segment is obtained'₁(n)：

Applying 1 st whitening filter to 3 rd data segment sequence w₃(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 3 rd data segment'₃(n)：

When m is 2, for the 2 nd data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 2 nd data segment, k is more than or equal to 1 and less than or equal to p₂，p₂Order, u, of the AR model for the data segment₂(n) is a white noise sequence;

constructing a second whitening filter with AR model coefficients, the transfer function H of which₂(z) is as follows:

for the 2 nd sequence of data segments w through the 2 nd whitening filter₂(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 2 nd data segment'₂(n)：

For the 4 th data segment sequence w through the 2 nd whitening filter₄(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 4 th data segment'₄(n)：

When M is more than or equal to 3 and less than or equal to M-2, for the mth data segment, estimating the order and the coefficient of the AR model of the current data segment according to an improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the mth data segment, k is more than or equal to 1 and less than or equal to p_m，p_mOrder, u, of the AR model for the data segment_m(n) is a white noise sequence;

constructing the mth whitening filter with the coefficients of the AR model, the transfer function H_m(z) is as follows:

for the m +2 data segments w through the m-th whitening filter_m+2(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the m +2 data segment'_m+2(n)：

And splicing the first s sampling points of each one-dimensional whitening sequence:

w′₁(1),…w′₁(s),w′₂(1),…w′₂(s),…w′_m-1(1),…w′_m-1(s),w′_m(1),…w′_m(s)

thus, one frame of discrete echo data w' (N) after whitening matched filtering processing is obtained, wherein N is 1, 2.

4. The adaptive bottom reverberation suppression method suitable for the side scan sonar according to claim 3, wherein a second-order Hankel matrix is constructed by performing phase space reconstruction on one frame of discrete echo data after whitening matched filtering processing; performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation so as to realize the suppression of the bottom reverberation; the method specifically comprises the following steps:

step S1) constructing a 2 × (N-1) dimensional Hankel matrix according to the discrete echo data of the frame after the whitening matched filtering process, where N is 1, 2.

Performing singular value decomposition on H:

wherein, U_2×2＝[u₁,u₂]Left singular matrix of H, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Right singular matrix of H, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s₀＝[diag(σ_a0,σ_d0)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_a0At large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A₁；σ_d0For small singular values, the reconstructed signal is an SVD detail signal, denoted as D₁；

Step S2) configuration A_j2 × (N-1) dimensional Hankel matrix H_j(ii) a The initial value of j is 1;

step S3) for the matrix H_jTo carry outSingular value decomposition:

wherein, U_2×2＝[u₁,u₂]Is A_jLeft singular matrix of, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Is A_jRight singular matrix, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s_j＝[diag(σ_aj,σ_dj)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_ajAt large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A_j+1；σ_djFor small singular values, the reconstructed signal is an SVD detail signal, denoted as D_j+1；

Step S4) calculating the large singular value σ_ajAnd small singular value σ_djRatio r of_j：

Step S5) calculating the difference e_j：

e_j＝r_j-r_j-1

When j is 1, r_j-1＝0；

Step S6) judgment e_jIf the judgment result is less than 1, the decomposition is finished and the step S7 is entered, otherwise, the step S2 is entered by adding 1 to j);

step S7) of outputting the target echo signal a separated from the bottom reverberation_j。

5. The adaptive bottom reverberation suppression method suitable for a side scan sonar according to claim 4, further comprising: and drawing a sound map of the target echo signal output by the multi-frame discrete echo data.

Technical Field

The invention relates to the field of side-scan sonars, in particular to a self-adaptive bottom reverberation suppression method suitable for the side-scan sonars.

Background

The side scan sonar is the underwater bottom static small target detection equipment which is widely applied at present. When the side-scan sonar underwater small target detection device works, two narrow beams are formed by the side-scan sonar in the direction perpendicular to the course direction, and when the side-scan sonar underwater small target detection device is used for detecting underwater small targets, bright spots generated by target echoes and sound shadows caused by shielding are main characteristics for target detection. However, the bottom reverberation, especially when the bottom is a substrate such as a sand bottom, a hard mud bottom, and the like, causes a decrease in image contrast, a decrease in image definition, and the like, and is a main factor affecting the working performance of the side scan sonar.

Commonly used methods for suppressing bottom reverberation can be divided into two main categories: waveform design and signal processing. Under the condition that the sonar working form is determined, signal processing is carried out at a receiving end to restrain bottom reverberation. The more common signal processing algorithms mainly include pre-whitening techniques and filtering techniques. The filtering technology, such as fractional Fourier transform, transforms the target echo into a fractional Fourier domain for filtering processing to achieve the purpose of suppressing reverberation, but the method is only suitable for linear frequency modulation signals. The pre-whitening technology assumes that the reverberation signal is locally stable, blocks the data and constructs an AR model, so that the received reverberation signal is pre-whitened, and finally reverberation suppression is realized through matched filtering. The algorithm has certain requirements on signal-to-noise ratio and is not strong in universality.

Disclosure of Invention

The invention provides a bottom reverberation suppression method combining sub-optimal detection with data self-driveability and multi-resolution binary singular value decomposition aiming at the problem that the detection and identification performance of a side-scan sonar on a sunk static small target is reduced under the influence of bottom reverberation.

In order to achieve the above object, the present invention provides an adaptive bottom reverberation suppression method suitable for a side scan sonar, including:

segmenting each frame of received discrete echo data, dividing the discrete echo data into a plurality of data segments, wherein the length of each data segment is twice the width of a transmitting pulse, and adjacent data segments 1/2 are overlapped;

estimating the order and the coefficient of the AR model of each data segment according to the improved Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter;

performing phase space reconstruction on the discrete echo data of the frame after the whitening matched filtering processing to construct a second-order Hankel matrix; and performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation, thereby realizing the suppression of the bottom reverberation.

As an improvement of the above method, each frame of received discrete echo data is segmented into a plurality of data segments, wherein the length of the data segment is twice the width of the transmitted pulse, and the adjacent data segments 1/2 are overlapped; the method specifically comprises the following steps:

segmenting each frame of received discrete echo data, and dividing the discrete echo data into M data segments, wherein each data segment comprises 2s sampling points, namely the mth data segment is a sequence:

w_m(1),…w_m(s),…w_m(2s)

where t is 1 ≦ s, i.e., adjacent data segments 1/2 overlap.

As an improvement of the above method, the estimating the order and coefficient of the AR model of each data segment according to the modified Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter; the method specifically comprises the following steps:

when m is 1, for the 1 st data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 1 st data segment, k is more than or equal to 1 and less than or equal to p₁，p₁Order, u, of the AR model for the data segment₁(n) is a whiteA noise sequence; n is more than or equal to 1 and less than or equal to 2 s;

construction of the 1 st whitening Filter with AR model coefficients, its transfer function H₁(z) is as follows:

wherein z is a variable;

whitening matched filtering processing is carried out on the 1 st data segment sequence through the 1 st whitening filter, and a one-dimensional whitening sequence w 'of the 1 st data segment is obtained'₁(n)：

Applying 1 st whitening filter to 3 rd data segment sequence w₃(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 3 rd data segment'₃(n)：

When m is 2, for the 2 nd data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 2 nd data segment, k is more than or equal to 1 and less than or equal to p₂，p₂Order, u, of the AR model for the data segment₂(n) is a white noise sequence;

constructing a second whitening filter with AR model coefficients, the transfer function H of which₂(z) is as follows:

for the 2 nd sequence of data segments w through the 2 nd whitening filter₂(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 2 nd data segment'₂(n)：

For the 4 th data segment sequence w through the 2 nd whitening filter₄(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 4 th data segment'₄(n)：

When M is more than or equal to 3 and less than or equal to M-2, for the mth data segment, estimating the order and the coefficient of the AR model of the current data segment according to an improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the mth data segment, k is more than or equal to 1 and less than or equal to p_m，p_mOrder, u, of the AR model for the data segment_m(n) is a white noise sequence;

constructing the mth whitening filter with the coefficients of the AR model, the transfer function H_m(z) is as follows:

for the m +2 data segments w through the m-th whitening filter_m+2(n) performing whitening matched filtering processing to obtain the m +2 dataOne-dimensional whitening sequence of segments w'_m+2(n)：

And splicing the first s sampling points of each one-dimensional whitening sequence:

w′₁(1),…w′₁(s),w′₂(1),…w′₂(s),…w′_m-1(1),…w′_m-1(s),w′_m(1),…w′_m(s)

thus, one frame of discrete echo data w' (N) after whitening matched filtering processing is obtained, wherein N is 1, 2.

As an improvement of the above method, the phase space reconstruction is performed on the discrete echo data of the frame after the whitening matched filtering processing, and a second-order Hankel matrix is constructed; performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation so as to realize the suppression of the bottom reverberation; the method specifically comprises the following steps:

step S1) constructing a 2 × (N-1) dimensional Hankel matrix according to the discrete echo data of the frame after the whitening matched filtering process, where N is 1, 2.

Performing singular value decomposition on H:

wherein, U_2×2＝[u₁,u₂]Left singular matrix of H, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Right singular matrix of H, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s₀＝[diag(σ_a0,σ_d0)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_a0At large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A₁；σ_d0For small singular values, the reconstructed signal is an SVD detail signal, denoted as D₁；

Step S2) configuration A_j2 × (N-1) dimensional Hankel matrix H_j(ii) a The initial value of j is 1;

step S3) for the matrix H_jSingular value decomposition is carried out:

wherein, U_2×2＝[u₁,u₂]Is A_jLeft singular matrix of, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Is A_jRight singular matrix, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s_j＝[diag(σ_aj,σ_dj)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_ajAt large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A_j+1；σ_djFor small singular values, the reconstructed signal is an SVD detail signal, denoted as D_j+1；

Step S4) calculating the large singular value σ_ajAnd small singular value σ_djRatio r of_j：

Step S5) calculating the difference e_j：

e_j＝r_j-r_j-1

When j is 1, r_j-1＝0；

Step S6) judgment e_jIf < 1, if yes, the decomposition is finished, and the process goes to step S7), otherwise, the process goes to step S7j plus 1, go to step S2);

step S7) of outputting the target echo signal a separated from the bottom reverberation_j。

As an improvement of the above method, the method further comprises: and drawing a sound map of the target echo signal output by the multi-frame discrete echo data.

The invention has the advantages that:

the method comprises the steps of firstly carrying out whitening matched filtering processing on a received signal, then adopting iterative binary singular value decomposition, and realizing the separation of target echo and bottom reverberation through subspace processing; the method can improve the signal-to-mixture ratio and improve the quality of the sonogram, and has important significance for realizing the detection of the side-scan sonar sinking-bottom static small target.

Drawings

FIG. 1 is a flow chart of the adaptive bottom reverberation suppression method suitable for side scan sonar according to the present invention;

FIG. 2 is a schematic diagram of a multi-resolution binary singular value decomposition path;

FIG. 3 is an original sonogram and its test results;

FIG. 4 is a sound diagram after processing and its detection result.

Detailed Description

The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

The invention provides a self-adaptive bottom reverberation suppression method suitable for a side scan sonar. Firstly, segmenting preprocessed echo data and estimating the order and the coefficient of an AR model of a data segment according to an improved Burg algorithm; then, a whitening filter is constructed by utilizing the AR model coefficient to perform whitening matched filtering processing on the data section; then, carrying out phase space reconstruction on the whitened data to construct a second-order Hankel matrix; and finally, performing multi-resolution binary singular value decomposition on the matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation so as to realize the suppression of the bottom reverberation. The method improves the signal-to-noise ratio and improves the quality of the sound image.

As shown in fig. 1, the adaptive bottom reverberation suppression method applied to a side scan sonar according to the present invention includes:

step 1) segmenting each frame of received discrete echo data, dividing the discrete echo data into a plurality of data segments, wherein the length of each data segment is twice of the width of a transmitting pulse, and adjacent data segments 1/2 are overlapped; the method specifically comprises the following steps:

segmenting each frame of received discrete echo data, and dividing the discrete echo data into M data segments, wherein each data segment comprises 2s sampling points, namely the mth data segment is a sequence:

w_m(1),…w_m(s),…w_m(2s)

where t is 1 ≦ s, i.e., adjacent data segments 1/2 overlap.

Step 2) estimating the order and the coefficient of the AR model of each data segment according to the improved Burg algorithm; constructing a whitening filter by utilizing the AR model coefficient, and performing whitening matched filtering processing on the data section by utilizing the whitening filter; the method specifically comprises the following steps:

when m is 1, for the 1 st data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 1 st data segment, k is more than or equal to 1 and less than or equal to p₁，p₁Order, u, of the AR model for the data segment₁(n) is a white noise sequence; n is more than or equal to 1 and less than or equal to 2 s;

construction of the 1 st whitening Filter with AR model coefficients, its transfer function H₁(z) is as follows:

wherein z is a variable;

carrying out whitening matched filtering processing on the 1 st data segment sequence through the 1 st whitening filter to obtain a one-dimensional whitening sequence w of the 1 st data segment′₁(n)：

Applying 1 st whitening filter to 3 rd data segment sequence w₃(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 3 rd data segment'₃(n)：

When m is 2, for the 2 nd data segment, estimating the order and the coefficient of the AR model of the current data segment according to the improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the 2 nd data segment, k is more than or equal to 1 and less than or equal to p₂，p₂Order, u, of the AR model for the data segment₂(n) is a white noise sequence;

constructing a second whitening filter with AR model coefficients, the transfer function H of which₂(z) is as follows:

for the 2 nd sequence of data segments w through the 2 nd whitening filter₂(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 2 nd data segment'₂(n)：

Passing through a 2 nd whitening filterFor the 4 th data segment sequence w₄(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the 4 th data segment'₄(n)：

When M is more than or equal to 3 and less than or equal to M-2, for the mth data segment, estimating the order and the coefficient of the AR model of the current data segment according to an improved Burg algorithm:

wherein the content of the first and second substances,is the coefficient of the AR model of the mth data segment, k is more than or equal to 1 and less than or equal to p_m，p_mOrder, u, of the AR model for the data segment_m(n) is a white noise sequence;

constructing the mth whitening filter with the coefficients of the AR model, the transfer function H_m(z) is as follows:

for the m +2 data segments w through the m-th whitening filter_m+2(n) carrying out whitening matched filtering processing to obtain a one-dimensional whitening sequence w 'of the m +2 data segment'_m+2(n)：

And splicing the first s sampling points of each one-dimensional whitening sequence:

w′₁(1),…w′₁(s),w′₂(1),…w′₂(s),…w′_m-1(1),…w′_m-1(s),w′_m(1),…w′_m(s)

thus, one frame of discrete echo data w' (N) after whitening matched filtering processing is obtained, wherein N is 1, 2.

Step 3) carrying out phase space reconstruction on the frame of discrete echo data subjected to whitening matched filtering processing to construct a second-order Hankel matrix; performing multi-resolution binary singular value decomposition on the second-order Hankel matrix, and performing echo reconstruction through the selected singular value to complete the separation of the target echo and the bottom reverberation so as to realize the suppression of the bottom reverberation; the method specifically comprises the following steps:

step 3-1) constructing a 2 × (N-1) dimensional Hankel matrix according to the discrete echo data of one frame after whitening matched filtering processing, wherein N is 1, 2.

Performing singular value decomposition on H:

wherein, U_2×2＝[u₁,u₂]Left singular matrix of H, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Right singular matrix of H, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s₀＝[diag(σ_a0,σ_d0)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_a0At large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A₁；σ_d0For small singular values, the reconstructed signal is an SVD detail signal, denoted as D₁；

Step 3-2) construction A_j2 × (N-1) dimensional Hankel matrix H_j(ii) a The initial value of j is 1;

step 3-3) for matrix H_jSingular value decomposition is carried out:

wherein, U_2×2＝[u₁,u₂]Is A_jLeft singular matrix of, V_(N-1)×(N-1)＝[v₁,v₂,…v_N-1]Is A_jRight singular matrix, vector u_i、v_iAre respectively a matrix U_2×2And V_(N-1)×(N-1)The ith column vector of (1); s_j＝[diag(σ_aj,σ_dj)]Is a diagonal matrix, 0_2×(N-3)Is a 2 x (N-3) dimensional zero matrix; sigma_ajAt large singular value, the reconstructed signal is an SVD approximate signal, and is marked as A_j+1；σ_djFor small singular values, the reconstructed signal is an SVD detail signal, denoted as D_j+1(ii) a Such as the decomposition path shown in fig. 2.

Step 3-4) calculating a large singular value sigma_ajAnd small singular value σ_djRatio r of_j：

Step 3-5) calculating the difference e_j：

e_j＝r_j-r_j-1

When j is 1, r_j-1＝0；

Step 3-6) judgment of e_jIf the value is less than 1, finishing the decomposition and entering a step 3-7), otherwise, adding 1 to j and entering a step 3-2);

step 3-7) outputting a target echo signal A separated from the bottom reverberation_j。

And 4) carrying out data processing of the steps 1) to 3) on each frame of data and forming a graph to obtain a sound graph after the bottom reverberation is suppressed.

Simulation analysis:

the water depth is about 50 meters, the seabed is a sediment bottom, the side scan sonar and the arrangement of the simulation target are shown in figure 3, wherein the simulation target is an iron cylinder. The side scan sonar passes through the simulation target from two sides of the target at a speed of no more than 4 knots under water at 25 meters in different directions and different distances. The multi-frame data is combined into the graph, and the processing result is shown in fig. 4.

Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.