An acoustic imaging method based on three-dimensional feature enhancement of underwater targets

Abstract

This invention provides an acoustic imaging method based on the enhancement of three-dimensional features of underwater targets, comprising: establishing an underwater three-dimensional acoustic imaging model; dividing the imaging area into multiple two-dimensional spherical slices centered at the sound source; discretizing each spherical slice into a set of grid points; constructing a signal receiving model; constructing a three-dimensional L1-TV norm optimization problem based on the underwater three-dimensional acoustic imaging model to enhance the three-dimensional features of the underwater target; decomposing the three-dimensional L1-TV norm optimization problem into several sub-problems using the alternating multiplier method; solving each sub-problem based on the Chambolle projection method and the near-end mapping algorithm to obtain its analytical or numerical solution, thereby achieving three-dimensional imaging of the underwater target. This invention effectively utilizes the three-dimensional features of the underwater imaging area and the contour information of the underwater target, resulting in fewer artifacts in the imaging results, clearer target contours, and improved imaging quality.

Description

Technical Field

[0001] This invention relates to the field of underwater three-dimensional imaging, and more specifically to an acoustic imaging method based on the enhancement of three-dimensional features of underwater targets. Background Technology

[0002] Beamforming can reconstruct 3D images from scattered signals recorded by a 2D array, and it is one of the most widely used techniques for acoustic array imaging. Due to the rapid development of digital signal processing, the use of 2D arrays and beamforming technology has become one of the mainstream solutions for 3D imaging systems. Therefore, researching beamforming technology to better apply it to 3D imaging systems is of great significance.

[0003] The bottleneck in developing underwater 3D imaging systems lies in the high hardware complexity of using large 2D arrays and the high software complexity of using digital beamforming. Therefore, existing underwater 3D imaging research mainly focuses on designing sparser 2D arrays and reducing the computational cost of beamforming to ensure real-time imaging. Due to hardware cost considerations, sparse 2D arrays are the preferred choice for underwater 3D systems. Unfortunately, due to the limited aperture and high sidelobe level of sparse 2D arrays, the imaging quality of underwater 3D systems using conventional convenient beamforming (CBF) methods is poor. Because of the Rayleigh criterion, the aperture of the 2D array used in an underwater 3D system determines the main lobe width of the beam pattern, which limits the azimuth resolution of the CBF method. Sparsity is achieved at the cost of high sidelobe levels, which leads to background noise and artifacts in the 3D acoustic image. These artifacts can overwhelm weak objects and cause blurring across different imaging ranges. Given the hardware limitations of sparse two-dimensional arrays, using high-performance beamforming methods to replace traditional methods has become a promising software approach to improve the quality of underwater 3D imaging. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and propose an acoustic imaging method based on the enhancement of three-dimensional features of underwater targets. This method effectively utilizes the three-dimensional features of the underwater imaging area and the contour information of the underwater target, resulting in fewer artifacts and clearer target contours in the imaging results, thereby improving the imaging quality.

[0005] In view of this, the present invention provides an acoustic imaging method based on three-dimensional feature enhancement of underwater targets, comprising: Step 1: Establish an underwater three-dimensional acoustic imaging model, divide the imaging area into multiple two-dimensional spherical slices with the sound source as the center, and discretize each spherical slice into a set of grid points to construct a signal receiving model; Step 2: Based on the underwater 3D acoustic imaging model, construct a 3D L1-TV norm optimization problem to enhance the 3D features of the underwater target; Step 3: Decompose the three-dimensional L1-TV norm optimization problem into several subproblems using the alternating multiplier method; Step 4: Based on the Chambolle projection method and the near-end mapping algorithm, solve each sub-problem to obtain its analytical or numerical solution, thereby realizing the three-dimensional imaging of underwater targets.

[0006] As an improvement to the above method, the specific process of establishing the underwater three-dimensional acoustic imaging model in step 1 includes: The imaging region is divided into circles centered on the sound source and with a radius of [missing information]. Multiple two-dimensional spherical slices, among which, k Indicates the first k A two-dimensional spherical slice , R The total number of slices is given, and each spherical slice is discretized into... The nth densely distributed grid point set is used to calculate the nth d ... Line 1 Direction vector of column grid points Furthermore, it is set that all scatterers within the imaging area are composed of dense point scatterers, and that all point scatterers are located on grid points; Perform a Fourier transform on the array received signal to obtain the received matrix. The expression is:

[0007] in, M Indicates the number of array element sensors. For array manifold matrix, The response of the imaging region to the transmitted signal. It is additive noise.

[0008] As an improvement to the above method, the three-dimensional L1-TV norm optimization problem constructed in step 2 is as follows:

[0009] Among them, the definition space for , operation This indicates rearranging the elements of a matrix. ; Operation Representation of operations The reverse, Indicates to The result of the rearrangement and These are characteristic parameters. It is a noise parameter. Representation space Euclidean norm in Representation space The TV norm in the context.

[0010] As an improvement to the above method, step 3 is decomposed into the following sub-problems: (1): (2): (3): (4): in, This is the result of the (q+1)th imaging. The superscript T indicates transpose. For Lagrange operators, , , and These are the Lagrange operators for the q-th iteration. and As an intermediate variable, , , , These are the results of the (q+1)th iteration. For data fitting parameters, For set The characteristic function of .

[0011] As an improvement to the above method, step 4 involves solving each sub-problem, specifically including: Repeat the following steps until the error between two consecutive iterations is small enough to obtain the imaging result: Step 4-1: Solve the subproblem (1): The solution is obtained by setting the first derivative of the objective function to 0: ; Among them, superscript H Indicates conjugate transpose; Step 4-2: Solve subproblem (2): By using vectors Orthogonal projection onto the feasible region of the hypersphere To solve this, we get:

[0012] in, ; Step 4-3: Solve subproblem (3): Using the soft threshold shrinkage operator Solution: in, , Representing a three-dimensional matrix The i, j, k-th elements , Representing the angle of complex variables; Step 4-4: Solving subproblem (4): Obtaining analytical solution based on Chambolle projection method:

[0013] Steps 4-5: Update the Lagrange operator using the alternating multiplier method: .

[0014] As an improvement to the above method, in subproblem (4), ,in, , Solving the projection using an iterative method Numerical solution: Defining space , initialize any ,along with , converges to ,in, The iterative formula is: in, and The four-dimensional matrix for the (n+1)th iteration is respectively and the four-dimensional matrix of the nth iteration The vector at positions i, j, k For learning rate, Represents finding space The gradient of the three-dimensional matrix in the image. This indicates the calculation of discrete divergence.

[0015] Compared with the prior art, the advantages of the present invention are: 1. An iterative solution method for the 3D TV optimization problem based on the Chambolle algorithm is proposed, and the algorithm is optimized by utilizing prior knowledge of contour features, which provides conditions for subsequent imaging algorithm solutions.

[0016] 2. By considering the complete three-dimensional features of the target and enhancing the three-dimensional contour features of the target, the present invention can image the entire three-dimensional region at once and improve the imaging quality.

[0017] 3. This invention decomposes the imaging algorithm into multiple sub-problems, including 3D TV norm regularization, and uses the alternating direction multiplier method to solve them iteratively, which significantly reduces the algorithm complexity and computational resource consumption, making it easier to deploy in practical applications. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall process of an acoustic imaging method based on three-dimensional feature enhancement of underwater targets according to the present invention. Figure 2 Specific features of the region that needs to be imaged; Figure 3 The true element distribution of a two-dimensional planar array; Figure 4 The imaging results are from a traditional beamforming algorithm; Figure 5 The imaging results are from the compressed beamforming algorithm; Figure 6 The imaging result is from the deconvolution algorithm; Figure 7 The imaging results are those of this invention. Detailed Implementation

[0019] This invention provides an acoustic imaging method based on the enhancement of three-dimensional features of underwater targets. The method includes: (1) establishing an underwater three-dimensional acoustic imaging model; (2) proposing an iterative solution to minimize the three-dimensional TV norm using the Chambolle projection method; (3) based on the aforementioned underwater three-dimensional acoustic imaging model, using the three-dimensional L1-TV norm to enhance the three-dimensional features of underwater targets, and constructing a three-dimensional L1-TV norm optimization problem; (4) using the alternating multiplier method to decompose the optimization problem into four easily solvable sub-problems; (5) based on the aforementioned Chambolle projection method and the near-end mapping solution algorithm, solving the sub-problems respectively to obtain their analytical or numerical solutions, thereby realizing the three-dimensional imaging of underwater targets.

[0020] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific examples.

[0021] Example 1 like Figure 1 As shown, Embodiment 1 of the present invention proposes an acoustic imaging method based on the enhancement of three-dimensional features of underwater targets, the implementation of which includes the following steps: Step 1: Establish an underwater three-dimensional acoustic imaging model, the specific implementation is as follows: Step 1-1) Divide the imaging area into circles centered on the sound source and with a radius of... A series of two-dimensional spherical slices, k Indicates the first k A two-dimensional spherical slice RThe total number of slices; and each spherical slice is discretized as The nth densely distributed grid point set is used to calculate the nth d ... Line 1 Direction vector of column grid points Assume that all scatterers in the imaging region consist of dense point scatterers, and that these point scatterers are located exactly at grid points.

[0022] Steps 1-2) Performing a Fourier transform on the array received signal yields the receiver matrix. for:

[0023] in This represents the spectrum of the pulse signal emitted from an ideal source. Indicates in Randomly distributed on a plane The array element sensor is numbered as follows The position of the array element sensor, Indicates the time point at which the pulse echo signal is received. This represents the speed of sound in water. Indicates the first Individual elements The result of performing a short-time Fourier transform on the signal received at time t, where j represents the imaginary part.

[0024] Steps 1-3) Calculate the array manifold matrix :

[0025] Steps 1-4) From this, the signal model can be obtained as follows: in, This represents the additive noise generated during the sampling data of the array element sensor and the propagation of the acoustic signal. Indicates the imaging area's response to the emitted pulse signal The response.

[0026] Step 2: Based on the aforementioned underwater 3D acoustic imaging model, the 3D features of the underwater target are enhanced using the 3D L1-TV norm. The 3D L1-TV norm optimization problem is constructed as follows: Among them, the definition space for , operation This represents rearranging the elements of a matrix; for any... All ; Operation Representation of operations The reverse. , and These are characteristic parameters. It is a noise parameter.

[0027] Step 3: Decompose the optimization problem into four easily solvable subproblems using the alternating multiplier method: in, This is the result of the (q+1)th imaging. , For Lagrange operators, , , and These are the Lagrange operators for the q-th iteration. For data fitting parameters, , , , These are the results of the (q+1)th iteration. gather The characteristic function. After a certain number of iterations, the imaging result can be obtained. .

[0028] Step 4: Based on the Chambolle projection method and the algorithm for solving near-end mapping, solve the subproblems respectively to obtain their analytical or numerical solutions.

[0029] Step 4-1) Subproblem 1 can be solved by setting the first derivative of the objective function to 0: . Among them, superscript H Indicates conjugate transpose; Step 4-2) Subproblem 2 can be solved by changing the vector Orthogonal projection onto the feasible region of the hypersphere To solve this, we get: Among them, let Step 4-3) Subproblem 3 can be solved using the soft threshold shrinkage operator. To solve this, we need to find the answer: in, in, , Representing a three-dimensional matrix The i, j, k-th elements , Represents the perspective of complex variables.

[0030] Step 4-4) Subproblem 4 can be solved analytically using the Chambolle projection method:

[0031] in, express: in , The projection can be solved using an iterative method. Numerical solution: Defining space , initialize any ,along with , converges to ,in, The iterative formula is: in, and The four-dimensional matrix for the (n+1)th iteration is respectively and the four-dimensional matrix of the nth iteration The vector at positions i, j, k For learning rate, Represents finding space The gradient of the three-dimensional matrix in the equation. Discrete divergence Defined as:

[0032] Steps 4-5) Update the Lagrange operator using the alternating multiplier method:

[0033] Steps 4-6) Repeat steps 4-1) to 4-6) continuously, iteratively calculating. Finally, the imaging results can be obtained.

[0034] Simulation example: The technical effects of the present invention will be further explained below with reference to simulation: To verify the performance of the proposed method, the imaged region needs to have specific features, such as... Figure 2The values ​​of the cubes follow a Gaussian distribution with a mean of 0.3 and a variance of 0.1, and the values ​​of two sets of mutually perpendicular parallel lines are 1. To demonstrate that the present invention provides better imaging results for regions with the above-mentioned features, conventional beamforming algorithms, compressed beamforming algorithms, and deconvolution algorithms are selected for comparison.

[0035] Assuming 512 array element sensors are randomly distributed in In a uniformly densely arranged grid of points, the spacing between the points is equal to the wavelength of the selected sound wave, such as... Figure 3 The imaging system transmits a signal at a frequency of 600 kHz. The sinusoidal domain is uniformly divided into... A beam angle, forming A uniform field of view, and correspondingly divide the three-dimensional imaging region into uniform parts. One grid, that is In this invention, , and .set up To emphasize contour information in the depth direction, Gaussian white noise with a signal-to-noise ratio of -15dB is added to the imaging region, and the parameters are set... Slightly larger than the variance of noise ,Right now .

[0036] Imaging results as follows Figures 4-7 As shown. Comparison Figure 6 and Figure 7 Both the deconvolution algorithm and this invention produce good imaging results for cubes. However, because the deconvolution algorithm requires the independence of the sound source, it cannot distinguish objects that are close together. Therefore, for two sets of parallel lines that are perpendicular to each other, the deconvolution algorithm cannot distinguish two parallel lines that are close together, while this invention can. Figure 4 The imaging results from the conventional beamforming algorithm shown are similar to those from the deconvolution algorithm, failing to distinguish between two sets of parallel lines that are perpendicular to each other. Figure 5 The compressed beamforming algorithm shown in the simulation exhibits significant background noise. In conclusion, the present invention demonstrates the best imaging performance in this simulation.

[0037] 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 it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An acoustic imaging method based on three-dimensional feature enhancement of underwater targets, comprising: Step 1: Establish an underwater three-dimensional acoustic imaging model, divide the imaging area into multiple two-dimensional spherical slices with the sound source as the center, and discretize each spherical slice into a set of grid points to construct a signal receiving model; Step 2: Based on the underwater 3D acoustic imaging model, construct a 3D L1-TV norm optimization problem to enhance the 3D features of the underwater target; Step 3: Decompose the three-dimensional L1-TV norm optimization problem into several subproblems using the alternating multiplier method; Step 4: Based on the Chambolle projection method and the near-end mapping algorithm, solve each sub-problem to obtain its analytical or numerical solution, thereby realizing the three-dimensional imaging of underwater targets.

2. The acoustic imaging method based on underwater target three-dimensional feature enhancement according to claim 1, characterized in that, The specific process of establishing the underwater three-dimensional acoustic imaging model in step 1 includes: The imaging region is divided into circles centered on the sound source and with a radius of [missing information]. Multiple two-dimensional spherical slices, among which, k Indicates the first k A two-dimensional spherical slice , R The total number of slices is given, and each spherical slice is discretized into... The nth densely distributed grid point set is used to calculate the nth d ... Line 1 Direction vector of column grid points Furthermore, it is set that all scatterers within the imaging area are composed of dense point scatterers, and that all point scatterers are located on grid points; Perform a Fourier transform on the array received signal to obtain the received matrix. The expression is: ； in, M Indicates the number of array element sensors. For array manifold matrix, The response of the imaging region to the transmitted signal. It is additive noise.

3. The acoustic imaging method based on underwater target three-dimensional feature enhancement according to claim 2, characterized in that, The three-dimensional L1-TV norm optimization problem constructed in step 2 is as follows: ； Among them, the definition space for , operation This indicates rearranging the elements of a matrix. ; Operation Representation of operations The reverse, Indicates to The result of the rearrangement and These are characteristic parameters. It is a noise parameter. Representation space Euclidean norm in Representation space The TV norm in the context.

4. The acoustic imaging method based on underwater target three-dimensional feature enhancement according to claim 3, characterized in that, Step 3 is broken down into the following sub-problems: （1）： ； （2）： ； （3）： ； （4）： ； in, This is the result of the (q+1)th imaging. The superscript T indicates transpose. For Lagrange operators, , , and These are the Lagrange operators for the q-th iteration. and As an intermediate variable, , , , These are the results of the (q+1)th iteration. For data fitting parameters, For set The characteristic function of .

5. The acoustic imaging method based on underwater target three-dimensional feature enhancement according to claim 4, characterized in that, Step 4 involves solving each sub-problem, specifically including: Repeat the following steps until the error between two consecutive iterations is small enough to obtain the imaging result: Step 4-1: Solve the subproblem (1): The solution is obtained by setting the first derivative of the objective function to 0: ； Among them, superscript H Indicates conjugate transpose; Step 4-2: Solve subproblem (2): By using vectors Orthogonal projection onto the feasible region of the hypersphere To solve this, we get: ； ； in, ; Step 4-3: Solve subproblem (3): Using the soft threshold shrinkage operator Solution: ； ； in, , Representing a three-dimensional matrix The i, j, k-th elements , Representing the angle of complex variables; Step 4-4: Solving subproblem (4): Obtaining analytical solution based on Chambolle projection method: ； Steps 4-5: Update the Lagrange operator using the alternating multiplier method: 。 6. The acoustic imaging method based on underwater target three-dimensional feature enhancement according to claim 5, characterized in that, In the subproblem (4), ,in, , Solving the projection using an iterative method Numerical solution: Defining space , initialize any ,along with , converges to ,in, The iterative formula is: ； in, and The four-dimensional matrix for the (n+1)th iteration is respectively and the four-dimensional matrix of the nth iteration exist i , j , k The vector of position, For learning rate, Represents finding space The gradient of the three-dimensional matrix in the image. This indicates the calculation of discrete divergence.

Images