A high-resolution target bright spot extraction method under reverberation background

Abstract

The application provides a high-resolution target highlight extraction method in a reverberation background, belongs to the field of underwater acoustic signal processing, and particularly relates to the field of active sonar recognition research. In order to improve the performance of highlight extraction, the difference between the time-frequency domain reverberation and target echo correlation is used to model the two as a low-rank matrix and a non-low-rank matrix. In the Bayesian framework, the low-rank matrix is modeled, and the Laplace and generalized inverse Gaussian mixture distribution prior is given to the non-low-rank part to enhance the robustness of matrix recovery. After the Bayesian inference is completed through MCMC sampling, the time-frequency matrix formed by the target echo after reverberation suppression is subjected to RLSTFRFT processing to improve the time-frequency resolution, and the performance of highlight extraction is comprehensively improved. The simulation and experimental results verify the effectiveness of the method, and the performance is obviously improved compared with the existing method. The achievement realizes the performance improvement of highlight extraction, and provides a technical foundation for underwater target recognition.

Description

Technical Field

[0001] This invention belongs to the field of underwater acoustic signal processing, and more precisely, to the category of underwater active sonar identification. It is mainly a method for extracting high-resolution target bright spots in a reverberant background. Background Technology

[0002] Underwater target identification is one of the main technologies studied in sonar technology. Accurate identification of underwater targets relies on effective feature extraction. The target bright spot model is a practical target model based on the echo generation mechanism. The target bright spot structure is also a widely used underwater target feature; therefore, target echo bright spots and their extraction are crucial for underwater target identification. Current research on target echo bright spots can be mainly divided into three categories: theoretical and predictive research on underwater target echo bright spots, research on target echo bright spots using measured data, and simulation and identification research of target echoes based on bright spot models. Target echo bright spots contain target information; the better the bright spot extraction, the more beneficial it is for target identification. Research on target echo bright spot extraction can be divided into one-dimensional target echo (i.e., instantaneous signal) bright spot extraction and two-dimensional target echo (i.e., beam domain signal) bright spot extraction. One-dimensional target echo bright spot extraction is represented by methods such as matched filtering and time-frequency analysis. Two-dimensional target echo bright spot extraction can be effectively accomplished using existing image sonar combined with array processing technology. Low-rank matrix restoration theory has developed rapidly over the past decade and is now widely used in image processing, and in recent years it has also been gradually applied to underwater acoustic engineering. In target highlight extraction, Zhu Guangping et al. have achieved separation of target echo and reverberation in the time-frequency domain based on the different correlations in the energy distribution of target echo and reverberation. This work provides a new approach for underwater target highlight extraction. However, this work still has shortcomings such as insufficient robustness, performance highly dependent on parameter adjustments, failure to consider highlight clustering, and failure to improve highlight performance from a resolution perspective. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a high-resolution target bright spot extraction method in a reverberant background, further improving the bright spot extraction performance and providing a technical foundation for active sonar recognition.

[0004] The objective of this invention is achieved through the following technical solution: A method for extracting high-resolution target bright spots against a reverberant background, comprising the following steps:

[0005] Step 1: Perform time-frequency analysis on the active sonar received signal to obtain its time-frequency distribution;

[0006] Step 2: Perform Bayesian modeling on the low-rank matrix L representing the reverberation time-frequency matrix;

[0007] Step 3: Perform Bayesian modeling on the received signal time-frequency matrix Y, and assign a Laplace and generalized inverse Gaussian mixture prior distribution to the non-low-rank part representing the target echo;

[0008] Step 4: Place a first-order Markov random field on the non-low-rank part of the time-frequency matrix to ensure the clustering effect of the target echo bright spots in the time-frequency domain;

[0009] Step 5: Perform Bayesian inference on the model through MCMC sampling, and output the low-rank part representing the reverberation and the non-low-rank part representing the target echo;

[0010] Step 6: After improving the time-frequency resolution of the non-low-rank part of the output by deconvolution short-time fractional Fourier transform, the target echo bright spot extraction result is output.

[0011] The aforementioned time-frequency analysis of the active sonar received signal involves performing a Short Time Fractional Fourier Transform (STFRFT) on the active sonar received signal to obtain its time-frequency matrix. The expression for STFRFT is as follows:

[0012]

[0013] Where s(t) is the transmitted signal, g(t) is the window function, α is the transformation angle, and K is the variable angle. α (t,u) represents the transform kernel, where u is the parameter of the fractional Fourier transform.

[0014] Furthermore, the obtained time-frequency matrix Y is represented as...

[0015] Y = L + S

[0016] Where L is a low-rank matrix representing the time-frequency matrix of reverberation, and S is a non-low-rank matrix representing the time-frequency matrix of target echo;

[0017] The time-frequency matrix of the received signal Perform Bayesian modeling for

[0018]

[0019] η ij ~GIG(p,a,b)

[0020] in The mean is a i b j T The scale is η ij The Laplace distribution is given by GIG, which represents the generalized inverse Gaussian distribution, and p, a, b are hyperparameters.

[0021] In step four, in order to achieve the clustering effect of non-low-rank matrix pixels (target echoes in the time-frequency matrix), in Place a first-order Markov random field in the image, and define a potential function for each neighboring pixel (i,j) and (p,q).

[0022]

[0023] ψ(Λ ij ,Λ pq )=exp{-α|logΛ ij -logΛ pq |}

[0024] The strength of the prior is controlled by the parameter α.

[0025] In step six, the output matrix S is subjected to deconvolution short-time fractional Fourier transform (RLSTFRFT) to improve its time-frequency resolution, and the target bright spot extraction result is output. The iterative update expression of RLSTFRFT is as follows:

[0026]

[0027] in This indicates the relevant operation, k is the number of iterations, and the initial value RLFRFT0 is set to FRFT.

[0028] The beneficial effects of this invention are as follows: First, the received signal from the active sonar is transformed into the time-frequency domain by performing a short-time fractional Fourier transform (STFRFT). When the transmitted signal is a linear frequency modulated signal, the echo signal exhibits sparsity in the time-frequency domain as a few diagonal lines, while the reverberation within the bandwidth exhibits correlation and low rank in the time-frequency domain. Bayesian modeling is performed on the low-rank matrix composed of reverberation under the Markov Bayesian Robust Matrix Factorization (MBRMF) model, and a Laplace and generalized inverse Gaussian mixture prior distribution is assigned to the sparse matrix composed of the target echo. Furthermore, to ensure the clustering of the time-frequency matrix elements of the target echo bright spot, a first-order Markov random field is placed between the sparse matrix elements. Then, MCMC sampling is used to infer the parameters in the Bayesian model, thereby achieving low-rank sparse decomposition, i.e., separation of reverberation and target bright spot. By performing deconvolution on the sparse matrix composed of target bright spots using short-time fractional Fourier transform (RLSTFRFT), the resolution of target bright spots is further improved, thus comprehensively enhancing the performance of the bright spots. Simulation and experimental results show that this method is effective for extracting target echo bright spots. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art or ordinary skills, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart of a method for extracting high-resolution target highlights against a reverberant background.

[0031] Figure 2 The target model used in the simulation is a rigid cylinder with one end being a frustum and the other end being a hemisphere.

[0032] Figure 3 This is a schematic diagram of the simulation results when the signal-to-mixing ratio is -5dB.

[0033] Figure 4 A schematic diagram showing the variation of the third-order Rayleigh entropy with the signal-to-mixing ratio for results obtained by different methods.

[0034] Figure 5 This is a schematic diagram of the experimental results for a real target. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0036] like Figure 1 As shown, this invention proposes a high-resolution target bright spot extraction method against a reverberant background. It achieves the separation of target echo and reverberation in the time-frequency domain within a Bayesian framework, and further enhances bright spot resolution using fractional-order deconvolution. The specific steps include:

[0037] 1. First, perform a short-time fractional Fourier transform (STFRFT) on the active sonar received signal to obtain its time-frequency matrix. The expression for STFRFT is:

[0038]

[0039] Where s(t) is the transmitted signal, g(t) is the window function, α is the transformation angle, and K is the variable angle. α (t,u) represents the transform kernel, where u is the parameter of the fractional Fourier transform.

[0040] 2. Represent the obtained time-frequency matrix Y as follows:

[0041] Y = L + S

[0042] Where L is a low-rank matrix representing the time-frequency matrix of reverberation, and S is a sparse (non-low-rank) matrix representing the time-frequency matrix of target echo.

[0043] 3. Perform Bayesian modeling on the low-rank matrix L as follows:

[0044] L = AB T

[0045]

[0046] in, r << m, n; a i and b j These represent the i-th row of A and the j-th row of B, respectively. Let be a normal distribution with mean μ and variance τ; μ0, v0, β0, and W0 are hyperparameters. Let W represent the Wishart distribution with scale matrix W and degrees of freedom v.

[0047] 4. For the time-frequency matrix Perform Bayesian modeling for

[0048]

[0049] η ij ~GIG(p,a,b)

[0050] in The mean is a i b j T The scale is η ij The Laplace distribution is given by GIG, which represents the generalized inverse Gaussian distribution, and p, a, b are hyperparameters.

[0051] 5. To achieve the clustering effect of non-low-rank matrix pixels (target echoes in the time-frequency matrix), in Place a first-order Markov random field in the image. Define a potential function for each neighboring pixel (i,j) and (p,q).

[0052]

[0053] ψ(Λ ij ,Λ pq )=exp{-α|logΛ ij -logΛ pq |}

[0054] The strength of the prior is controlled by the parameter α.

[0055] 6. The established Bayesian model can be approximated using MCMC sampling, thus achieving the recovery of the low-rank matrix L. The basic logic of MCMC sampling is to obtain a series of independent samples from the target posterior distribution, thereby approximating the target distribution. Due to space limitations, the inference results are not listed here.

[0056] 7. Output the target echo matrix S after suppressing reverberation via S=YL.

[0057] 8. The time-frequency resolution of the output matrix S is improved by performing a deconvolution short-time fractional Fourier transform (RLSTFRFT), resulting in the extracted target bright spots. The iterative update expression for RLSTFRFT is as follows:

[0058]

[0059] in This indicates the relevant operation, k is the iteration number, and the initial value RLFRFT0 is generally set to FRFT.

[0060] like Figure 3 The simulation results shown are for a signal-to-mixing ratio of -5dB. Figure 3 (a) is the result after STFRFT. Figure 3 (b) for in Figure 3 (a) The result after processing with existing principal component tracking (PCP) based reverberation suppression methods. Figure 3 (c) shows the result after reverberation suppression using this method. Figure 3 (d) is in Figure 3 (c) Results of improving resolution using fractional-order deconvolution. In the simulation, the distance between the receiver and the target was 100m. The timing of each bright spot on the target was obtained through image pulse theory, and the amplitude of each bright spot was calculated using physical acoustics. The transmitted signal was a linear frequency modulated signal with a frequency band of 30kHz-40kHz, a pulse width of 1ms, and a sampling rate of 200kHz. As can be seen from the results, compared with the existing PCP method, this method has better reverberation suppression capability, improves time-frequency resolution, and comprehensively improves the performance of the bright spots.

[0061] like Figure 4 The graphs show the variation of the third-order Rayleigh entropy with the signal-to-mixing ratio obtained by different methods. A smaller third-order Rayleigh entropy indicates a more concentrated energy in the time-frequency domain, resulting in better bright spot extraction. The results show that matrix factorization effectively improves bright spot extraction performance, and this method outperforms existing methods.

[0062] like Figure 5 The results shown are experimental results for real targets. Figure 5 (a) The result of performing an STFRFT on the received data. Figure 5(b) shows the results after processing using this method. The target in the experiment was a hollow spherical crown-shaped cylinder with a steel shell, constructed based on a scaled-down model. The top was a hemisphere with a diameter of 6 cm, and the bottom was a cylinder with the same diameter as the hemisphere and a height of 18 cm. The transmitted signal used a linear frequency modulated signal with a frequency band of 340 kHz-440 kHz, a pulse width of 0.5 ms, a sampling rate of 20 MHz, and was transmitted vertically towards the target. The results show that this method remains effective in the experimental data.

[0063] In summary, to improve the performance of bright spot extraction, this invention models the time-frequency reverberation and target echo as low-rank and non-low-rank matrices respectively, based on the difference in their correlation. The low-rank matrix is ​​modeled within a Bayesian framework, while the non-low-rank part is given a Laplace and generalized inverse Gaussian mixture priors to enhance the robustness of matrix recovery. After completing Bayesian inference through MCMC sampling, the resulting time-frequency matrix composed of the reverberation-suppressed target echo is processed using RLSTFRFT to improve time-frequency resolution, thus comprehensively enhancing the performance of bright spot extraction. Simulation and experimental results verify the effectiveness of this method, showing a significant performance improvement compared to existing methods. This invention achieves improved bright spot extraction performance and provides a technical foundation for underwater target identification.

[0064] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for extracting a high-resolution target highlight point in a reverberation background, characterized in that: The steps comprise the following: Step one: time-frequency analysis is performed on the active sonar receiving signal to obtain its time-frequency distribution; Step two: Bayesian modeling is performed on the low-rank matrix L representing the reverberation time-frequency matrix; Step three: Bayesian modeling is performed on the receiving signal time-frequency matrix Y, and a Laplace and generalized inverse Gaussian mixture distribution prior is given to the non-low-rank part representing the target echo; Step four: a first-order Markov random field is placed on the non-low-rank part in the time-frequency matrix to ensure the clustering effect of the target echo bright spot in the time-frequency domain; Step five: Bayesian inference of the model is realized through MCMC sampling, and the low-rank part representing the reverberation and the non-low-rank part representing the target echo are outputted; Step six: after the non-low-rank part outputted is subjected to deconvolution short-time fractional Fourier transform to improve the time-frequency resolution, the target echo bright spot extraction result is outputted.

2. The method of claim 1, wherein: The time-frequency analysis performed on the active sonar receiving signal is short-time fractional Fourier transform STFRFT performed on the active sonar receiving signal to obtain its time-frequency matrix, and the expression of STFRFT is where s(t) is the transmitted signal, g(t) is the window function, a is the transform angle, K α (t,u) is the transform kernel, and u is the parameter of the fractional Fourier transform, 3. The method of claim 2, wherein: The obtained time-frequency matrix Y is expressed as Y=L+S Wherein, L is a low-rank matrix representing the time-frequency matrix of reverberation, and S is a non-low-rank matrix representing the time-frequency matrix of the target echo; The time-frequency matrix of the received signal Bayesian modeling is performed η ij ~ GIG(p, a, b) wherein denotes the mean a i b j T , scale η ij Laplace distribution, GIG denotes the generalized inverse Gaussian distribution, p, a, b are hyperparameters.

4. The method of claim 3, wherein: In step four, in A first-order Markov random field is placed in the middle, and a potential function is defined for adjacent pixel points (i, j) and (p, q) ψ(Λ ij ,Λ pq ) = exp{-α|logΛ ij -logΛ pq |} The strength of the prior is controlled by the parameter α.

5. The method of claim 4, wherein: In step six, the matrix S outputted is subjected to deconvolution short-time fractional Fourier transform RLSTFRFT to improve the time-frequency resolution, and the target bright spot extraction result is outputted, and the iterative update expression of RLSTFRFT is wherein represents a correlation operation, k is the iteration number, and the initial value RLFRFT0 is set to FRFT.

Images