A complex reverberation suppression method based on bayesian matrix factorization

Abstract

The application discloses a complex reverberation suppression method based on Bayesian matrix decomposition, and relates to the technical field of sonar detection. The method comprises the following steps: vectorizing each three-dimensional sonar image in an original three-dimensional sonar image sequence in time sequence, and then splicing the three-dimensional sonar image sequence into a two-dimensional data matrix; using a Bayesian matrix decomposition method to extract an indicator matrix of low-rank components and non-low-rank components from the two-dimensional data matrix; calculating each non-low-rank component through the low-rank components and the indicator matrix, and selecting a non-low-rank component with the maximum contrast as a sparse target component; reversely vectorizing the two-dimensional sparse target component into a three-dimensional sparse target image sequence; using nonlinear superposition on the three-dimensional sparse target image sequence to suppress fluctuation reverberation and enhance the target, and obtaining a moving target trajectory. Compared with other reverberation suppression algorithms, the application avoids the selection of an optimization algorithm regularization parameter, and can stably implement complex reverberation suppression to detect a moving small target.

Description

Technical Field

[0001] This application relates to the field of sonar detection technology, and in particular to a complex reverberation suppression method based on Bayesian matrix decomposition. Background Technology

[0002] Active sonar-based moving target detection is an effective means of underwater safety monitoring. However, complex reverberation from the surface, bottom, and water body severely affects the detection performance of active sonar and increases the false alarm rate. Therefore, complex reverberation suppression is an essential technical component of active detection. Robust principal component analysis, which utilizes the low-rank steady-state reverberation and sparsity of moving targets in multi-frame sonar images, is currently the mainstream method for reverberation suppression.

[0003] The relevant literature (FXGe, Y.Chen, and W.Li, “Target detection and tracking via structured convex optimization,” in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing-Proceedings, 2017.) proposes the Accelerated Proximal Gradient (APG) method to separate reverberation and the target. Based on this, numerous low-rank sparse reverberation suppression methods have been proposed. Patent CN120742282A proposes a multi-frame reverberation suppression method combining the alternating direction multiplier method with low-complexity high-order nonlinear cumulants, and patent CN120972152A proposes a reverberation suppression method and apparatus based on tensor robust principal component analysis.

[0004] The low-rank sparse algorithms mentioned above are all based on optimization algorithms. However, the regularization parameter that balances the sparse target and low-rank reverberation needs to be manually selected, which greatly limits the performance and practicality of these algorithms. Especially for complex shallow water environments with low signal-to-mixing ratios, an excessively large regularization parameter can cause low-energy targets to be decomposed into low-rank matrices. Furthermore, due to the complexity of underwater noise, a single Gaussian noise is difficult to represent the complex underwater environment. Summary of the Invention

[0005] In view of this, embodiments of this application provide a complex reverberation suppression method and related equipment based on Bayesian matrix decomposition to accurately separate sparse targets in sonar images.

[0006] One aspect of this application provides a complex reverberation suppression method based on Bayesian matrix decomposition, the method comprising the following steps:

[0007] Each frame of the original three-dimensional sonar image sequence is vectorized in time order and then stitched together into a two-dimensional data matrix;

[0008] The indicator variable matrix of low-rank and non-low-rank components is extracted from the two-dimensional data matrix using the Bayesian matrix decomposition method;

[0009] Each non-low-rank component is calculated using the low-rank component and the indicator variable matrix, and the non-low-rank component with the highest contrast is selected as the sparse target component.

[0010] The two-dimensional sparse target components are inversely vectorized into a three-dimensional sparse target image sequence;

[0011] Nonlinear superposition is used on the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, resulting in the trajectory of the moving target.

[0012] In some embodiments, the step of vectorizing each frame of the original three-dimensional sonar image sequence in temporal order and then stitching them together into a two-dimensional data matrix includes the following steps:

[0013] For the original three-dimensional sonar image sequence Each frame of the 3D sonar image is vectorized in chronological order and then stitched together to form the 2D data matrix. ;in, , , , For matrix vectorization operators, For the number of image frames, This refers to the total number of pixels in each frame of the three-dimensional sonar image.

[0014] In some embodiments, extracting the indicator variable matrix of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix factorization method includes the following steps:

[0015] The two-dimensional data matrix is ​​modeled as including low-rank components and non-low-rank components;

[0016] The two-dimensional data matrix is ​​decomposed into the low-rank component and the non-low-rank component using the Bayesian matrix decomposition method.

[0017] The low-rank components are updated and reconstructed until the reconstruction error of the low-rank components is less than a set threshold, and then the indicator variable matrix of the current low-rank components and non-low-rank components is output.

[0018] In some embodiments, modeling the two-dimensional data matrix as including low-rank and non-low-rank components includes the following steps:

[0019] The two-dimensional data matrix The model is as follows: ;in, For the low-rank component, Including steady-state reverberation; For the non-low-rank component, Including wave reverberation, moving targets, motion interference, and additive noise. It was modeled as a Gaussian mixture model;

[0020] The step of decomposing the two-dimensional data matrix into low-rank components and non-low-rank components using the Bayesian matrix decomposition method includes the following steps:

[0021] The two-dimensional data matrix is ​​processed using the Bayesian matrix decomposition method described above. Matrix decomposition is performed to decompose steady-state reverberation and moving targets; wherein the input of the Bayesian matrix decomposition method includes the two-dimensional data matrix. The degree of freedom parameters of the Wishart distribution Scale matrix parameters The number of Gaussian components in the Gaussian mixture model ;

[0022] Initialize low-rank components , low-rank components precision matrix Indicator variable matrix of non-low-rank components Weight parameters for each Gaussian component Mean of each Gaussian component Precision of each Gaussian component and the parameters of the Gaussian mixture model .

[0023] In some embodiments, the step of updating and reconstructing the low-rank components until the reconstruction error of the low-rank components is less than a set threshold, and then outputting the indicator variable matrix of the current low-rank components and non-low-rank components, includes the following steps:

[0024] Update low-rank components Each column and the variance of each column :

[0025] ;

[0026] ;

[0027] in, This represents the diagonalization operator for column vectors; As the first process variable, its first... item ; As the second process variable, its first... item , For data matrix The Line number List;

[0028] Update precision matrix and weight parameters ;

[0029] Update the mean of each Gaussian component and its precision :

[0030] ;

[0031] ;

[0032] in, Low-rank components The Line number List; ; ; ; variance The Line number List;

[0033] Update the indicator variable matrix of non-low-rank components ;

[0034] in, ;

[0035] Represents the digamma function;

[0036] Return to the updated low-rank component Each column and the variance of each column The steps continue until the low-rank components are reconstructed. The error is less than the stopping iteration error. Then output the current low-rank component. Indicator variable matrix of non-low-rank components .

[0037] In some embodiments, calculating each non-low-rank component using the low-rank component and the indicator variable matrix includes the following steps:

[0038] Indicator variable matrix through non-low-rank components Calculate each non-low-rank component ,Right now ;

[0039] Selecting the non-low-rank component with the highest contrast as the sparse target component includes the following steps:

[0040] In calculating each non-low-rank component Contrast ,Right now ;

[0041] Select the highest contrast value index Corresponding non-low-rank components As the sparse target component .

[0042] In some embodiments, the nonlinear superposition of the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, resulting in a moving target trajectory, includes the following steps:

[0043] For the sparse target image sequence in three dimensions Nonlinear superposition is performed to suppress wave reverberation and enhance the target, resulting in the trajectory of the moving target; wherein the expression for the nonlinear superposition is as follows:

[0044] ;

[0045] in, express lie in Pixel value at; The mean of the sparse target image sequence; The order of the nonlinear accumulation of the sparse target image sequence;

[0046] Output the sparse target image sequence and the corresponding nonlinear superposition results .

[0047] Another aspect of this application embodiment provides a complex reverberation suppression device based on Bayesian matrix decomposition, the device comprising:

[0048] The image vectorization unit is used to vectorize each frame of the original three-dimensional sonar image sequence in time order, and then stitch them together into a two-dimensional data matrix.

[0049] A matrix decomposition unit is used to extract indicator variable matrices of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix decomposition method.

[0050] The component selection unit is used to calculate each non-low-rank component using the low-rank component and the indicator variable matrix, and select the non-low-rank component with the highest contrast as the sparse target component.

[0051] A matrix inverse vectorization unit is used to inverse vectorize the two-dimensional sparse target components into a three-dimensional sparse target image sequence.

[0052] The target trajectory extraction unit is used to apply nonlinear superposition to the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, thereby obtaining the trajectory of the moving target.

[0053] Another aspect of this application embodiment provides an electronic device, including a processor and a memory;

[0054] The memory is used to store programs;

[0055] The processor executes the program to implement any of the methods described above.

[0056] Another aspect of this application provides a computer-readable storage medium storing a program that is executed by a processor to implement the method described in any of the above embodiments.

[0057] This application includes at least the following beneficial effects:

[0058] This application can vectorize each frame of a three-dimensional sonar image in the original three-dimensional sonar image sequence in temporal order, and then stitch them together into a two-dimensional data matrix. Using Bayesian matrix factorization, indicator variable matrices of low-rank and non-low-rank components are extracted from the two-dimensional data matrix. Each non-low-rank component is calculated using the low-rank component and indicator variable matrices, and the non-low-rank component with the highest contrast is selected as the sparse target component. The two-dimensional sparse target component is inversely vectorized into a three-dimensional sparse target image sequence. Nonlinear superposition is applied to the three-dimensional sparse target image sequence to suppress wave reverberation and enhance the target, obtaining the trajectory of the moving target. Compared to other reverberation suppression algorithms, this application avoids the need to optimize the selection of regularization parameters, and can robustly implement complex reverberation suppression to detect small moving targets. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 A flowchart illustrating a complex reverberation suppression method based on Bayesian matrix decomposition provided in this application embodiment;

[0061] Figure 2 An example flowchart of a complex reverberation suppression method based on Bayesian matrix decomposition provided for embodiments of this application;

[0062] Figure 3 This is a frame image result from the original sonar image sequence provided in the embodiments of this application;

[0063] Figure 4 The embodiments provided in this application provide for the application of the following: Figure 3 The result of ADMM reverberation suppression on the image frames;

[0064] Figure 5 The embodiments provided in this application provide for the application of the following: Figure 3 The result of APG reverberation suppression on the image frames;

[0065] Figure 6 The embodiments provided in this application provide for the application of the following: Figure 3 The result of VBRPCA reverberation suppression on the image frame;

[0066] Figure 7 The embodiments provided in this application provide for the application of the following: Figure 3 The result of BMD reverberation suppression on the image frame;

[0067] Figure 8 The target trajectory map is obtained by nonlinearly superimposing the sparse target image sequence after BMD reverberation suppression provided in the embodiments of this application.

[0068] Figure 9 This is a structural block diagram of a complex reverberation suppression device based on Bayesian matrix decomposition, provided for an embodiment of this application. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0070] Reference Figure 1 This application provides a complex reverberation suppression method based on Bayesian matrix decomposition, specifically including the following steps S100~S140:

[0071] S100: Vectorize each frame of the original three-dimensional sonar image sequence in time order, and then stitch them together into a two-dimensional data matrix;

[0072] S110: Extract the indicator variable matrix of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix decomposition method;

[0073] S120: Calculate each non-low-rank component using the low-rank component and the indicator variable matrix, and select the non-low-rank component with the largest contrast as the sparse target component;

[0074] S130: Inverse vectorize the two-dimensional sparse target components into a three-dimensional sparse target image sequence;

[0075] S140: Nonlinear superposition is used on the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, thereby obtaining the trajectory of the moving target.

[0076] The following section will provide a detailed introduction and explanation of the solutions in the embodiments of this application, using specific application examples.

[0077] To address the shortcomings of existing technologies, this embodiment provides a complex reverberation suppression method based on Bayesian matrix decomposition, referring to... Figure 2 First, the original 3D sonar image sequence is vectorized frame by frame in chronological order and then concatenated to obtain a 2D data matrix. Then, the provided Bayesian matrix decomposition method is used to extract indicator variable matrices for both low-rank and non-low-rank components. Each non-low-rank component is calculated using these indicator variable matrices, and the non-low-rank component with the highest contrast is selected as the sparse target component. Next, the 2D sparse target matrix is ​​inversely vectorized into a 3D sparse target image sequence. Finally, nonlinear superposition is applied to the 3D sparse target image sequence to further suppress wave reverberation and enhance the target, yielding the moving target trajectory. Compared to other reverberation suppression algorithms, this method avoids optimizing the selection of regularization parameters. By unifying wave reverberation, moving targets, random interference, and noise into non-low-rank components and modeling them as Gaussian mixture models, it better reflects the real complex underwater environment and can robustly implement complex reverberation suppression to detect small moving targets.

[0078] The technical solution of this embodiment includes the following steps:

[0079] Step 1: Process the original 3D sonar image sequence Processing yields a two-dimensional data matrix .in, , , , For matrix vectorization operators, For the number of image frames, This represents the total number of pixels in each frame of the image.

[0080] Step 2: Convert the two-dimensional data matrix The model is as follows: .in, It is a low-rank component, mainly containing steady-state reverberation; The components are non-low-rank elements, mainly including fluctuating reverberation, moving targets, motion interference, and additive noise, and are modeled as a Gaussian mixture model. A variational Bayesian (VB) matrix decomposition method is used to separate steady-state reverberation from moving targets. The overall reverberation suppression target detection algorithm takes a two-dimensional data matrix as input. The degrees of freedom parameters of the Wishart distribution and scale matrix parameters Number of Gaussian components in a Gaussian mixture model Initialize low-rank components. ;low-rank component precision matrix ; Indicator variable matrix of non-low-rank components Weight parameters for each Gaussian component Mean of each Gaussian component Precision of each Gaussian component Gaussian mixture model parameters .

[0081] Step 3: Update low-rank components Each column and the variance of each column :

[0082] , .

[0083] in, This represents the diagonalization operator for column vectors; As a process variable, its first... item ; It is also a process variable, its first... item , For data matrix The Line number List.

[0084] Step 4: Update the precision matrix and weight parameters .

[0085] Step 5: Update the mean of each Gaussian component. and its precision :

[0086] , .

[0087] in, Low-rank components The Line number List; ; ; ; variance The Line number List.

[0088] Step 6: Update the indicator variable matrix of non-low-rank components .in:

[0089] ;

[0090] in, This represents the digamma function.

[0091] Step 7: Repeat steps 3, 4, 5, and 6 until the reconstructed low-rank components are obtained. The error is less than the stopping iteration error. The low-rank component is then output. Indicator variable matrix of non-low-rank components .

[0092] Step 8: Because of the residual It contains wave reverberation, moving targets, motion interference, and additive noise, so it is necessary to extract the moving target component. Specifically, this is done first through the indicator variable matrix of non-low-rank components. Calculate each non-low-rank component ,Right now Then calculate each non-low-rank component. Contrast ,Right now Finally, select the highest contrast value. index Corresponding non-low-rank components As a component of the motion target Target component Inverse vectorization yields a 3D sparse target image sequence. .

[0093] Step 9: Process the 3D sparse target image sequence Nonlinear superposition is then used to further suppress wave reverberation and enhance the target's trajectory. The specific nonlinear superposition formula is as follows:

[0094] ;

[0095] in, Representing the image matrix lie in Pixel value at; The mean of the sparse target image sequence; Let be the order of the nonlinear accumulation of the sparse target image sequence. The reverberation suppression algorithm based on Bayesian matrix factorization outputs the sparse target image sequence. and its nonlinear superposition result .

[0096] This embodiment describes a method for complex reverberation suppression of sonar image sequences containing small underwater moving targets, which can robustly detect the targets and form their motion trajectories. To verify the effectiveness and robustness of the Bayesian matrix decomposition (BMD) method, the embodiment processes and analyzes real experimental data and compares it with existing reverberation suppression methods such as variational Bayesian robust principal component analysis (VBRPCA), accelerated proximal gradient (APG), and alternating direction method of multipliers (ADMM).

[0097] Figure 3 This is a frame from the original sonar image sequence; the target is circled in red. Figures 4 to 7 To Figure 3 The results obtained by applying four reverberation suppression methods to the data frames show that although the four methods can separate the target, they have different degrees of suppression of background reverberation. Among them, the BMD reverberation suppression method in this embodiment has the best performance, which can basically suppress all complex reverberation and achieve the best performance in detecting small moving targets. Figure 8 The target trajectory map is obtained by nonlinearly superimposing the sparse target image sequence extracted after complex BMD reverberation suppression in this embodiment, demonstrating the robustness of the method in this embodiment for moving target detection.

[0098] Beneficial effects:

[0099] 1. The Bayesian matrix factorization algorithm designed in this embodiment can achieve complex reverberation suppression and can be better applied to the field of active detection of small moving underwater targets. It avoids the selection of regularization parameters in optimization methods and uses Gaussian mixture models to model non-low-rank components, which is more in line with the real complex underwater environment.

[0100] 2. Compared with existing reverberation suppression methods, the method in this embodiment suppresses more reverberation energy, adapts to various complex reverberation scenarios, and improves the robustness of the algorithm and the target detection performance.

[0101] Reference Figure 9 This application provides a complex reverberation suppression device based on Bayesian matrix decomposition, comprising:

[0102] The image vectorization unit is used to vectorize each frame of the original three-dimensional sonar image sequence in time order, and then stitch them together into a two-dimensional data matrix.

[0103] A matrix decomposition unit is used to extract indicator variable matrices of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix decomposition method.

[0104] The component selection unit is used to calculate each non-low-rank component using the low-rank component and the indicator variable matrix, and select the non-low-rank component with the highest contrast as the sparse target component.

[0105] A matrix inverse vectorization unit is used to inverse vectorize the two-dimensional sparse target components into a three-dimensional sparse target image sequence.

[0106] The target trajectory extraction unit is used to apply nonlinear superposition to the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, thereby obtaining the trajectory of the moving target.

[0107] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0108] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0109] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0110] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0112] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0113] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0114] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0115] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0116] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A complex reverberation suppression method based on Bayesian matrix decomposition, characterized in that, The method includes the following steps: Each frame of the original three-dimensional sonar image sequence is vectorized in time order and then stitched together into a two-dimensional data matrix; The indicator variable matrix of low-rank and non-low-rank components is extracted from the two-dimensional data matrix using the Bayesian matrix decomposition method; Each non-low-rank component is calculated using the low-rank component and the indicator variable matrix, and the non-low-rank component with the highest contrast is selected as the sparse target component. The two-dimensional sparse target components are inversely vectorized into a three-dimensional sparse target image sequence; Nonlinear superposition is used on the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, resulting in the trajectory of the moving target.

2. The complex reverberation suppression method based on Bayesian matrix decomposition according to claim 1, characterized in that, The process of vectorizing each frame of the original three-dimensional sonar image sequence in chronological order and then stitching them together into a two-dimensional data matrix includes the following steps: For the original three-dimensional sonar image sequence Each frame of the 3D sonar image is vectorized in chronological order and then stitched together to form the 2D data matrix. ;in, , , , For matrix vectorization operators, For the number of image frames, This refers to the total number of pixels in each frame of the three-dimensional sonar image.

3. The complex reverberation suppression method based on Bayesian matrix decomposition according to claim 1, characterized in that, The step of extracting the indicator variable matrix of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix factorization method includes the following steps: The two-dimensional data matrix is ​​modeled as including low-rank components and non-low-rank components; The two-dimensional data matrix is ​​decomposed into the low-rank component and the non-low-rank component using the Bayesian matrix decomposition method. The low-rank components are updated and reconstructed until the reconstruction error of the low-rank components is less than a set threshold, and then the indicator variable matrix of the current low-rank components and non-low-rank components is output.

4. The complex reverberation suppression method based on Bayesian matrix decomposition according to claim 3, characterized in that, Modeling the two-dimensional data matrix into components including low-rank and non-low-rank components includes the following steps: The two-dimensional data matrix The model is as follows: ;in, For the low-rank component, Including steady-state reverberation; For the non-low-rank component, Including wave reverberation, moving targets, motion interference, and additive noise. It was modeled as a Gaussian mixture model; The step of decomposing the two-dimensional data matrix into low-rank components and non-low-rank components using the Bayesian matrix decomposition method includes the following steps: The two-dimensional data matrix is ​​processed using the Bayesian matrix decomposition method described above. Matrix decomposition is performed to decompose steady-state reverberation and moving targets; wherein the input of the Bayesian matrix decomposition method includes the two-dimensional data matrix. The degree of freedom parameters of the Wishart distribution Scale matrix parameters The number of Gaussian components in the Gaussian mixture model ; Initialize low-rank components , low-rank components precision matrix Indicator variable matrix of non-low-rank components Weight parameters for each Gaussian component Mean of each Gaussian component Precision of each Gaussian component and the parameters of the Gaussian mixture model .

5. A complex reverberation suppression method based on Bayesian matrix decomposition according to claim 4, characterized in that, The process of updating and reconstructing the low-rank components until the reconstruction error of the low-rank components is less than a set threshold, and then outputting the indicator variable matrix of the current low-rank and non-low-rank components, includes the following steps: Update low-rank components Each column and the variance of each column : ； ； in, This represents the diagonalization operator for column vectors; As the first process variable, its first... item ; As the second process variable, its first item , For data matrix The Line number List; Update precision matrix and weight parameters ; Update the mean of each Gaussian component and its precision : ； ； in, Low-rank components The Line number List; ; ; ; For variance The Line number List; Update the indicator variable matrix of non-low-rank components ; in, ; Represents the digamma function; Return to the updated low-rank component Each column and the variance of each column The steps continue until the low-rank components are reconstructed. The error is less than the stopping iteration error. Then output the current low-rank component. Indicator variable matrix of non-low-rank components .

6. The complex reverberation suppression method based on Bayesian matrix decomposition according to claim 1, characterized in that, The calculation of each non-low-rank component using the low-rank component and the indicator variable matrix includes the following steps: Indicator variable matrix through non-low-rank components Calculate each non-low-rank component ,Right now ; Selecting the non-low-rank component with the highest contrast as the sparse target component includes the following steps: In calculating each non-low-rank component Contrast ,Right now ; Select the highest contrast value index Corresponding non-low-rank components As the sparse target component .

7. The complex reverberation suppression method based on Bayesian matrix decomposition according to claim 1, characterized in that, The process of using nonlinear superposition on the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, thereby obtaining the trajectory of the moving target, includes the following steps: For the sparse target image sequence in three dimensions Nonlinear superposition is performed to suppress wave reverberation and enhance the target, resulting in the trajectory of the moving target; wherein the expression for the nonlinear superposition is as follows: ； in, express lie in Pixel value at; The mean of the sparse target image sequence; The order of the nonlinear accumulation of the sparse target image sequence; Output the sparse target image sequence and the corresponding nonlinear superposition results .

8. A complex reverberation suppression device based on Bayesian matrix decomposition, characterized in that, The device includes: The image vectorization unit is used to vectorize each frame of the original three-dimensional sonar image sequence in time order, and then stitch them together into a two-dimensional data matrix. A matrix decomposition unit is used to extract indicator variable matrices of low-rank and non-low-rank components from the two-dimensional data matrix using the Bayesian matrix decomposition method. The component selection unit is used to calculate each non-low-rank component using the low-rank component and the indicator variable matrix, and select the non-low-rank component with the highest contrast as the sparse target component. A matrix inverse vectorization unit is used to inverse vectorize the two-dimensional sparse target components into a three-dimensional sparse target image sequence. The target trajectory extraction unit is used to apply nonlinear superposition to the sparse target image sequence in three dimensions to suppress wave reverberation and enhance the target, thereby obtaining the trajectory of the moving target.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the method as described in any one of claims 1 to 7.

Images