Active sonar super-resolution imaging method and device based on compressed sensing beam forming

Abstract

The invention provides an active sonar super-resolution imaging method and device based on compressed sensing beam forming. The method comprises the following steps: collecting echo signals which are received by an active sonar array element and are reflected by a scattering element in an observation area; generating a sensing matrix according to waveform prior information of the echo signal; according to an echo signal and the sensing matrix, constructing a target optimization function of compressed sensing beam forming; and solving the target optimization function to obtain a target echo signal corresponding to each scattering element in the observation area. According to the method, waveform prior information of an echo signal is added to a sensing matrix, and a target optimization function is constructed by using the echo signal and the sensing matrix, so that the internal sparsity of the target echo signal in an observation area is reduced, and super-resolution which breaks through the sampling rate limitation of the echo signal in distance is realized; and reverberation interference of the active sonar is effectively suppressed, so that the quality of echo signals received by the active sonar is greatly improved, and the accuracy of underwater target detection and recognition is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater acoustic signal processing, and particularly relates to an active sonar super-resolution imaging method and device based on compressed sensing beamforming. BACKGROUND

[0002] One of the most important means of underwater target detection and perception is active sonar, which has a wide range of applications in underwater target detection and identification, underwater monitoring, seabed topography mapping, and marine resource exploration.

[0003] Active sonar periodically transmits sound pulses to detect underwater areas. The echo received by the active sonar array after transmitting a sound pulse is called a ping. While receiving the scattering echo of underwater targets, the active sonar array also receives a large number of scattering echoes generated by random scatterers (such as bubbles, marine organisms, and seabed silt). These random scattering echoes are called reverberation, which can cause the quality of target echo signals to drop sharply.

[0004] Traditional active sonar generally uses conventional beamforming to form directivity in the direction and uses matched filtering to pulse compress the transmitted signal in the distance. However, the results obtained by conventional beamforming and matched filtering both have the problem of wide main lobe and high side lobe, making it difficult to accurately locate the direction and distance of the target signal, resulting in low resolution of the traditional active sonar in angle and distance, and thus leading to poor quality of the echo signal received by the active sonar.

[0005] For underwater target detection and identification tasks, high-resolution and high-quality active sonar echo signals are very important. The wide main lobe and high side lobe of conventional beamforming can easily cause weak targets to be submerged in reverberation and side lobe interference. Therefore, achieving a substantial improvement in the resolution of underwater active sonar and effectively suppressing reverberation can greatly improve the quality of the echo signal received by the active sonar, thereby improving the accuracy of underwater target detection and identification.

[0006] In recent years, the academic and industrial communities have made some explorations in improving the resolution of active sonar, mainly including adaptive beamforming, deconvolution beamforming, compressed sensing-based beamforming, and deep learning-based beamforming.

[0007] Among them, the adaptive beamformer is a classical beamformer based on statistical optimization, which can adaptively change the weights of beamforming according to the received data to achieve the effect of adaptive enhancement of signals and suppression of interference. This method obtains higher angular resolution than conventional beamforming, but the overall side lobe level does not decrease significantly, and the robustness is poor and easily affected by signal model errors.

[0008] Then, in 2017, Professor T.C. Yang from Zhejiang University proposed a deconvolution beamforming method for sonar. This method is a robust beamforming method that improves the angle resolution. It regards the output of the conventional beamforming as a convolution of a point spread function and the spatial distribution of the incident signal, and reconstructs the spatial distribution of the signal through deconvolution. This method greatly reduces the main lobe width and side lobe level of conventional beamforming, achieves an angle resolution beyond adaptive beamforming, and ensures the robustness consistent with conventional beamforming methods.

[0009] In 2020, the team from Zhejiang University further proposed an angle-distance two-dimensional deconvolution method. This method achieves super-resolution in angle (26 times higher than conventional beamforming) after angle deconvolution, and achieves a narrower compressed pulse in distance, with a width of only 1 / 10 of the conventional pulse compression width. However, the distance resolution is still limited by the sampling rate of the echo signal in the distance direction. In addition, the deconvolution method also has edge effects, and when the target signal deviates from the normal direction of the array at a large angle, the correctness of the deconvolution result will be affected.

[0010] Compressive sensing-based beamforming is a method that uses the intrinsic sparsity of target signals to achieve super-resolution. In 2019, the U.S. Ocean Research and Experiment Center proposed a frequency-domain compressive sensing beamforming method for synthetic aperture sonar based on array elements. This method overcomes the edge effect problem of deconvolution and achieves super-resolution in the navigation direction and distance direction. The essence of achieving super-resolution in the navigation direction is to benefit from achieving super-resolution in angle. However, the distance resolution of this method is still limited by the sampling rate of the echo signal, and the interference effect of reverberation is not considered.

[0011] In addition, this frequency-domain compressive sensing beamforming method needs to combine multiple ping data to make the optimization problem a well-posed problem. When this frequency-domain method is used for active sonar single ping data, the least squares problem with L1 norm constraint cannot be solved due to the high underdetermination of the sensing matrix in compressive sensing. This problem needs to introduce other constraints to make the target echo more sparse to solve it.

[0012] In recent years, beamforming methods based on supervised deep learning have also become a research direction in the field of acoustics. In 2018, a team from the University of Illinois proposed a deep neural network for beamforming of a speech microphone array, achieving better results than traditional beamforming methods. In 2023, a team from Tsinghua University proposed a multi-scale residual U-net network architecture that can be used for medical ultrasound array beamforming, achieving super-resolution imaging results. However, the beamforming method based on supervised deep learning requires a large amount of data for training, but the active sonar data in the field of underwater acoustics is scarce, and this supervised deep learning method cannot achieve good beamforming results in complex marine environments with only a small amount of data.

[0013] In summary, the current active sonar beamforming method based on supervised deep learning has good theoretical results, but due to the scarcity of underwater data sets, it is difficult to implement in actual underwater environments.

[0014] Among the methods based on traditional models, the beamforming method based on frequency domain compressive sensing has the best results in practical applications compared to other methods, but it still has three defects: 1) the super-resolution in the distance direction cannot break through the limitation of the sampling rate; 2) when applied to active sonar single ping data, there is a problem of high underdetermination of the sensing matrix in compressive sensing; 3) the influence of severe underwater reverberation is not considered.

[0015] Therefore, how to solve the problem of low resolution in angle and distance and severe reverberation interference of the active sonar in the prior art is an important topic in the field of underwater acoustic signal processing that needs to be solved. SUMMARY

[0016] The present application provides a kind of active sonar super-resolution imaging method and device based on compressive sensing beamforming, to overcome the defects of low resolution in angle and distance and severe reverberation interference of the active sonar in the prior art, realize the substantial improvement of underwater active sonar resolution, and effectively suppress reverberation interference, thereby substantially improving the quality of active sonar received echo signal, to improve the accuracy of underwater target detection and identification.

[0017] In one aspect, the application provides an active sonar super-resolution imaging method based on compressed sensing beamforming, comprising: collecting echo signals reflected by scatterers in an observed area after being received by an active sonar array element; generating a sensing matrix according to waveform prior information of the echo signals, the sensing matrix being a mapping matrix of the scatterers in the observed area to the active sonar array element; constructing a target optimization function of compressed sensing beamforming according to the echo signals and the sensing matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem; solving the target optimization function to obtain target echo signals corresponding to each scatterer in the observed area; wherein the target echo signals corresponding to each scatterer form a super-resolution active sonar image, and the active sonar image contains a detection target.

[0018] Further, the received echo signals are preprocessed, including: in the case of a bandpass signal, sequentially performing a bandpass filtering operation, a quadrature demodulation operation and a matched filtering operation on the echo signals; or in the case of a single frequency signal, sequentially performing a quadrature demodulation operation and an envelope taking operation on the echo signals; accordingly, the constructing a target optimization function according to the echo signals and the sensing matrix comprises: constructing a target optimization function according to the preprocessed echo signals and the sensing matrix.

[0019] Further, the generating a sensing matrix according to waveform prior information of the echo signals comprises: obtaining waveform prior information corresponding to the echo signals according to a single acoustic pulse signal transmitted by the active sonar array element; adding the waveform prior information to an initial mapping matrix to obtain the sensing matrix; wherein the single acoustic pulse signal corresponds to the echo signal, the waveform prior information includes a delay, a phase and a waveform of each scatterer in the observed area to the active sonar array element, and matrix elements of the initial mapping matrix are all zero.

[0020] Further, the target optimization function is expressed as follows:

[0021] wherein, represents the solved target echo signal / target scatter amplitude matrix, S represents a scatter amplitude matrix corresponding to each scatterer in the observed area, W represents a weighted coefficient diagonal matrix, D represents the echo signal received by the active sonar array element, A represents the sensing matrix, and ∈ represents a hyperparameter of a constraint fitting error.

[0022] Further, the solving the target optimization function comprises: using a multiple-round weighted iteration interior point method to solve the target optimization function.

[0023] Further, the inner point method with multiple rounds of weighted iteration is used to solve the target optimization function, including: setting the weighted coefficient diagonal matrix as an identity matrix, and using the inner point method to solve the target optimization function to obtain an initial target echo signal; setting diagonal elements of the weighted coefficient diagonal matrix as inverses of amplitudes of the initial target echo signal, and using the inner point method again to solve the target optimization function to obtain a candidate target echo signal; iteratively updating the diagonal elements of the weighted coefficient diagonal matrix and solving the target optimization function to obtain the target echo signals corresponding to each scattering element in the observation region.

[0024] Further, the hyperparameter of the constraint fitting error is determined by the following steps: estimating reverberation energy levels according to correlations between echo signals received by different active sonar array elements; and determining the hyperparameter of the constraint fitting error according to the estimated reverberation energy levels.

[0025] In a second aspect, the present application further provides an active sonar super-resolution imaging device based on compressed sensing beamforming, including: an echo signal acquisition module, configured to acquire echo signals reflected by scattering elements in an observation region and received by active sonar array elements; a sensing matrix generation module, configured to generate a sensing matrix according to waveform prior information of the echo signals, the sensing matrix being a mapping matrix from the scattering elements in the observation region to the active sonar array elements; a target optimization function construction module, configured to construct a target optimization function of compressed sensing beamforming according to the echo signals and the sensing matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem; and a target optimization function solving module, configured to solve the target optimization function to obtain target echo signals corresponding to each scattering element in the observation region; wherein the target echo signals corresponding to each scattering element form an active sonar image with super-resolution, and the active sonar image contains a detection target.

[0026] In a third aspect, the present application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the active sonar super-resolution imaging method based on compressed sensing beamforming according to any one of the above aspects when executing the program.

[0027] In a fourth aspect, the present application further provides a non-transitory computer readable storage medium, having a computer program stored thereon, and the computer program is executable on a processor to implement the active sonar super-resolution imaging method based on compressed sensing beamforming according to any one of the above aspects.

[0028] The application provides an active sonar super-resolution imaging method based on compressed sensing beam forming. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0030] Figure 1 FIG. 1 is a flowchart of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiments of the application.

[0031] Figure 2 FIG. 2 is a schematic diagram of the generation of the sensing matrix provided by the embodiments of the application.

[0032] Figure 3 FIG. 3 is a schematic diagram of the resolution improvement effect in angle and distance of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiments of the application.

[0033] Figure 4 FIG. 4 is a schematic diagram of the resolution improvement and reverberation suppression effect of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiments of the application on the side-scan sonar simulation image.

[0034] Figure 5 FIG. 5 is a schematic diagram of the distance super-resolution effect of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiments of the application on the forward-looking sonar.

[0035] Figure 6 is a processing result schematic diagram of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiment of the present application on the actually collected side-scan sonar data.

[0036] Figure 7 is a structural schematic diagram of the active sonar super-resolution imaging device based on compressed sensing beam forming provided by the embodiment of the present application.

[0037] Figure 8 is a structural schematic diagram of the electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0038] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0039] It should be noted that the supervised deep learning method in the current active sonar beam forming method has good effects in theory, but due to the scarcity of underwater data sets, it is difficult to implement in the actual underwater environment.

[0040] Among the methods based on the conventional model, the beam forming method based on frequency domain compressed sensing has the best effect in practical application compared with other methods, but there are still the following defects: 1) the super-resolution in the distance direction cannot break through the limitation of the sampling rate; 2) when applied to the active sonar single ping data, there is a problem of high underdetermination of the sensing matrix in compressed sensing; 3) the influence of serious reverberation under water is not considered.

[0041] In view of this, the present application provides an active sonar super-resolution imaging method based on compressed sensing beam forming, specifically, Figure 1 The flowchart of the active sonar super-resolution imaging method based on compressed sensing beam forming provided by the embodiment of the present application is shown.

[0042] As Figure 1 shown, the method comprises steps S110-S140, which will be described in detail below.

[0043] S110, collecting the echo signal reflected by the scattering element in the observed area received by the active sonar array element.

[0044] It can be understood that the active sonar array periodically emits a sound pulse to the observation area, and a plurality of scattering elements in the observation area receive and reflect the sound pulse, and then the active sonar array receives the echo signals reflected by each scattering element.

[0045] In this step, the echo signal includes not only the target scattering echo signal reflected by the target in the observation area, but also the scattering echo signal generated by other random scatterers (such as bubbles, marine organisms, and seabed silt) other than the target. The scattering echo signal generated by the random scatterer is reverberation, which will be suppressed in the subsequent step.

[0046] It should be noted that the received echo signal in this embodiment refers to the echo signal corresponding to a single sound pulse signal (one ping / single ping) emitted by the active sonar array.

[0047] It should also be noted that the active sonar array receives the echo signal from near to far according to different time delays.

[0048] In addition, after receiving the echo signal, in order to suppress the out-of-band noise of the transmitted signal, the echo signal can be preferably preprocessed, such as bandpass filtering, quadrature demodulation, and matched filtering. The preprocessing operation can be selected according to the actual situation of the echo signal, which is not limited here.

[0049] S120, generating a perception matrix according to waveform prior information of the echo signal, the perception matrix being a mapping matrix from the scattering element in the observation area to the active sonar array.

[0050] It can be understood that on the basis of step S110 of collecting the echo signal received by the active sonar array after the scattering element in the observation area is reflected, step S120 further acquires waveform prior information of the echo signal, and then generates a mapping matrix from the scattering element in the observation area to the active sonar array, i.e. the perception matrix, according to the waveform prior information.

[0051] The waveform prior information includes but is not limited to the time delay, phase, and waveform of each scattering element in the observation area to the active sonar array.

[0052] It is worth mentioning that the generation of the perception matrix in this embodiment does not need to use the received echo signal, but only needs to know the position of the active sonar array, the position of the scattering element in the observation area, and the sound pulse signal emitted by the active sonar array. The purpose is to prepare for constructing a target optimization function in the next step.

[0053] It should be further noted that by adding the waveform prior information of the echo signal in the perception matrix, the inherent sparsity of the target scattering echo amplitude matrix (target echo signal) in the observation area can be effectively reduced, thereby realizing the super-resolution that breaks through the sampling rate limit of the echo signal in the distance. The super-resolution means that the reverberation superimposed in each sampling point has been greatly reduced, so the embodiment can also realize the reverberation suppression of the active sonar.

[0054] S130, constructing a target optimization function of compressed sensing beamforming according to the echo signal and the perception matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem.

[0055] It can be understood that on the basis of generating the perception matrix according to the waveform prior information of the echo signal in step S120, further, step S130 will construct the target optimization function according to the echo signal collected in S110 and the perception matrix generated in S120.

[0056] Specifically, the embodiment converts the active sonar imaging into a convex optimization problem with sparse constraints, and uses the echo signal of a single ping and the perception matrix to construct a weighted sparse time-domain compressed sensing beamforming problem, that is, the target optimization function.

[0057] In addition, in order to enable the target optimization function to be better solved, a suitable hyperparameter can also be set to constrain the error of fitting data. The hyperparameter can be selected by experience or determined by estimating the reverberation level, which is not specifically limited here.

[0058] S140, solving the target optimization function to obtain the target echo signal corresponding to each scattering element in the observation area; wherein the target echo signal corresponding to each scattering element forms a super-resolution active sonar image, and the active sonar image contains the detection target.

[0059] It can be understood that on the basis of constructing the target optimization function according to the echo signal and the perception matrix in step S130, further, step S140 will solve the target optimization function to obtain the target echo signal corresponding to each scattering element in the observation area, thereby obtaining the finally formed active sonar image.

[0060] Specifically, the convex optimization method can be used to iteratively solve the weighted sparse compressed sensing beamforming problem, so that the position of the target echo signal can be accurately obtained, and the imaging result with super-resolution in the direction and distance, that is, the active sonar image containing the detection target, can be obtained.

[0061] In a specific embodiment, the convex optimization method is a multiple-round weighted iterative interior point method, and this solving method can further improve the performance of the active sonar super-resolution imaging method.

[0062] In the embodiment, the echo signal reflected by the scatter elements in the observed area is collected by the active sonar array elements, and a perception matrix is generated according to the waveform prior information of the echo signal, the perception matrix being a mapping matrix of the scatter elements in the observed area to the active sonar array elements, and then a target optimization function is constructed according to the echo signal and the perception matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem, so that the target echo signal corresponding to each scatter element in the observed area is obtained by solving the target optimization function, and the target echo signal corresponding to each scatter element forms a super-resolution active sonar image, and the active sonar image contains the detection target. The method adds the waveform prior information of the echo signal to the perception matrix, and constructs the target optimization function by using the echo signal and the perception matrix, thereby reducing the inherent sparsity of the target echo signal in the observed area, realizing the super-resolution which breaks through the sampling rate limit of the echo signal in the distance, effectively suppressing the reverberation interference of the active sonar, and greatly improving the quality of the active sonar received echo signal, thereby improving the accuracy of underwater target detection and recognition.

[0063] On the basis of the above-mentioned embodiments, the preprocessing process of the echo signal will be described in detail as follows.

[0064] The received echo signal is preprocessed, including: in the case of a bandpass signal, sequentially performing a bandpass filtering operation, a quadrature demodulation operation and a matched filtering operation on the echo signal; or in the case of a single frequency signal, sequentially performing a quadrature demodulation operation and an envelope taking operation on the echo signal.

[0065] It can be understood that after receiving the echo signal reflected by the observed area, in order to suppress noise, a series of preprocessing operations will be performed on the echo signal, and the specific preprocessing operation adopted is related to the type of the sound pulse signal transmitted by the active sonar array element (the type of the sound pulse signal is consistent with the type of the echo signal).

[0066] Specifically, in a specific embodiment, the sound pulse signal transmitted by the active sonar array element is a bandpass signal, such as a linear frequency modulation signal, in which case, first, a digital bandpass filter is used to bandpass filter the echo signal received by each active sonar array element to limit the bandwidth of the echo signal, thereby suppressing out-of-band noise.

[0067] Then, the bandpass filtered echo signal is quadrature demodulated to convert the carrier signal into a baseband signal and obtain the main information in the echo signal.

[0068] Finally, the baseband signal obtained by the quadrature demodulation is matched filtered, specifically, the baseband signal obtained by the quadrature demodulation is convolved with the echo signal received by each active sonar array element, to enhance the part of the echo signal related to the transmitted signal (the acoustic pulse signal transmitted by the active sonar array element) and suppress the noise part of the echo signal unrelated to the transmitted signal, so as to realize pulse compression of the echo signal. Finally, the preprocessed echo signal is obtained.

[0069] In another specific embodiment, the acoustic pulse signal transmitted by the active sonar array element is a single frequency signal, such as a continuous sine wave signal, etc. In this case, first, the echo signal is quadrature demodulated to convert the carrier signal to a baseband signal and obtain the main information in the echo signal.

[0070] Then, the baseband signal obtained by the quadrature demodulation is envelope taken, and the signal envelope is the profile of the signal waveform. Common envelope taking operations include the Matlab envelope taking function, so as to obtain the preprocessed echo signal.

[0071] Correspondingly, according to the echo signal and the sensing matrix, a target optimization function is constructed, including: according to the preprocessed echo signal and the sensing matrix, a target optimization function is constructed.

[0072] In this embodiment, by sequentially performing the bandpass filtering operation, the quadrature demodulation operation and the matched filtering operation on the echo signal in the case of the echo signal being a bandpass signal, or by sequentially performing the quadrature demodulation operation and the envelope taking operation on the echo signal in the case of the echo signal being a single frequency signal, the preprocessed echo signal is obtained, and then according to the preprocessed echo signal and the sensing matrix, a target optimization function is constructed, which is a weighted sparse compressed sensing beamforming problem, so as to solve the target optimization function to obtain the target echo signal corresponding to each scattering element in the observation region, and the target echo signal corresponding to each scattering element forms a super-resolution active sonar image, and the active sonar image contains the detection target. This method adds the waveform prior information of the echo signal to the sensing matrix, and constructs a target optimization function using the echo signal and the sensing matrix, which reduces the intrinsic sparsity of the target echo signal in the observation region, realizes super-resolution that breaks through the sampling rate limit of the echo signal in distance, effectively suppresses the reverberation interference of the active sonar, and further greatly improves the quality of the active sonar received echo signal, thereby improving the accuracy of underwater target detection and recognition.

[0073] On the basis of the above embodiment, further, the generation process of the sensing matrix will be described in detail below.

[0074] According to the waveform prior information of the echo signal, a perception matrix is generated, including: obtaining waveform prior information corresponding to the echo signal according to a single acoustic pulse signal emitted by the active sonar array element; adding the waveform prior information to an initial mapping matrix to obtain the perception matrix; wherein the single acoustic pulse signal corresponds to the echo signal, the waveform prior information includes the delay, phase and waveform of each scattering element in the observation region to the active sonar array element, and the matrix elements of the initial mapping matrix are all zero.

[0075] It can be understood that the embodiment innovatively introduces the prior information of the echo waveform into the perception matrix to enhance the inherent sparsity of the target echo signal, and the generation of the perception matrix in the embodiment does not need to use the received echo signal, but only needs to know the position of the active sonar array element, the position of the scattering element in the observation region and the acoustic pulse signal emitted by the active sonar array element and the like.

[0076] Specifically, first, the waveform prior information corresponding to the echo signal is obtained by using a single acoustic pulse signal emitted by the active sonar array element.

[0077] In the case of a bandpass signal, the echo prior information is the autocorrelation function (such as a sinc function with different delays) of the single acoustic pulse signal; in the case of a single frequency signal, the echo prior information is the envelope of the single acoustic pulse signal.

[0078] Whether it is the autocorrelation function of the single acoustic pulse signal or the envelope of the single acoustic pulse signal, the echo prior information includes the delay, phase and waveform of each scattering element in the observation region to the active sonar array element.

[0079] Wherein, the delay can be calculated according to the distance between each scattering element in the observation region and the active sonar array element, the phase can be obtained according to the orthogonal demodulation of the echo signal as a preprocessing operation, and the waveform can be obtained according to the single acoustic pulse signal emitted by the active sonar array element.

[0080] After obtaining the waveform prior information corresponding to the echo signal, the waveform prior information is added to the initial mapping matrix, and the desired perception matrix can be generated.

[0081] Wherein, the initial mapping matrix is a matrix with all zero matrix elements.

[0082] Figure 2 A generation schematic diagram of the perception matrix provided by the embodiment of the application is shown. Figure 2 As shown, Figure 2 The left side is the region where the active sonar array element is located, Figure 2 and the right side is the observation region.

[0083] If there is an independent scattering element in the observation area, the scattering echo of the scattering element will reach each active sonar array element with different time delays but the same waveform, and there is a many-to-many correspondence between each scattering element in the observation area and each active sonar array element.

[0084] The generation process of the perception matrix is to calculate the waveform prior information of the echo signal and put it into the initial mapping matrix.

[0085] Suppose that the number of active sonar array elements is M, the number of sampling points in each array element is L, the number of distance networks obtained by dividing the observation area is K, the number of angle networks is N, and the active sonar array element receives data (echo signal) as The scattering amplitude matrix of the scattering element in the observation area is The reverberation in the received data is The perception matrix is The relationship between the perception matrix, the scattering amplitude matrix of the scattering element, the echo signal, and the reverberation is represented by the following formula (1).

[0086] D = AS + E (1).

[0087] According to the above, the perception matrix can be generated.

[0088] In this embodiment, the waveform prior information corresponding to the echo signal is obtained according to a single sound pulse signal emitted by the active sonar array element, and the waveform prior information is added to the initial mapping matrix to obtain the perception matrix. Then, a target optimization function is constructed according to the echo signal and the perception matrix, and the target optimization function is a weighted sparse compressed sensing beamforming problem. Thus, the target optimization function is solved to obtain the target echo signal corresponding to each scattering element in the observation area, and the target echo signal corresponding to each scattering element forms a super-resolution active sonar image, and the active sonar image contains a detection target. This method adds the waveform prior information of the echo signal to the perception matrix, and constructs a target optimization function using the echo signal and the perception matrix, which reduces the intrinsic sparsity of the target echo signal in the observation area, realizes super-resolution that breaks through the sampling rate limit of the echo signal in the distance, effectively suppresses the reverberation interference of the active sonar, and greatly improves the quality of the received echo signal of the active sonar, thereby improving the accuracy of underwater target detection and recognition.

[0089] On the basis of the above embodiment, further, the process of constructing and solving the target optimization function will be described in detail.

[0090] According to the echo signal and the sensing matrix, a target optimization function is constructed. Specifically, in order to more accurately solve the target scattering amplitude matrix (target echo signal) of the observation region, a super-resolution active sonar image is obtained, and the optimization problem of time-domain compressed sensing beamforming is constructed by using the element receiving data (echo signal) of a single ping and the sensing matrix, that is, the target optimization function. Specifically, the constructed target optimization function is as follows.

[0091]

[0092] In formula (2), represents the solved target echo signal / target scattering amplitude matrix, S represents the scattering amplitude matrix corresponding to each scattering element in the observation region, W represents a weighted coefficient diagonal matrix, D represents the echo signal received by the active sonar element, A represents the sensing matrix, and ∈ represents the hyperparameter of the constraint fitting error.

[0093] The significance of the above target optimization function is that, under the premise that the fitting error is not higher than ∈, the L1 norm of WS is minimized, that is, the sparse target echo signal amplitude and spatial distribution estimation can be obtained.

[0094] It should be noted that, in the target optimization function, the weighted coefficient diagonal matrix W is a diagonal matrix, the diagonal initial value of which is the result of the conventional beamforming of the observation region, and the diagonal element after that is designed as the reciprocal of the target scattering echo amplitude S after each iteration. In this way, the coefficient with large amplitude can be retained in the iteration process, and the coefficient with small amplitude can be converged to 0 as soon as possible, so as to enhance the sparsity of the solved .

[0095] The hyperparameter ∈ of the constraint fitting error can be artificially set reasonably according to experience, or can be set reasonably by estimating the reverberation level.

[0096] In a specific embodiment, the reverberation energy level E is estimated according to the correlation between the echo signals received by different active sonar elements, and then the hyperparameter ∈ of the constraint fitting error is determined according to the estimated reverberation energy level. The value of ∈ can be slightly less than ||E||2, and better sound field reconstruction effect can be obtained.

[0097] It is worth mentioning that, by adding the echo prior information of the echo signal in the sensing matrix, the internal sparsity of the target scattering echo amplitude matrix S of the observation region is reduced, and thus the super-resolution which breaks through the sampling rate limit of the echo signal in the distance is realized. The super-resolution means that the reverberation superimposed in each sampling point has been greatly reduced, and therefore the reverberation suppression of the active sonar can also be realized.

[0098] As to super-resolution, it is actually achieved by increasing the number of angle grid N and distance grid K under the premise of utilizing the inherent sparsity of target echo signals (see Figure 2 ). Because the target optimization function utilizes all the received data within a single ping to solve the target scattering amplitude matrix of the observation region In addition, the matrix s is made more sparse by introducing the prior information of active sonar target echo (i.e. the to-be-solved echo amplitude parameters in S are much smaller than N×K), so the number of equations of the target optimization function is much larger than the number of unknowns, which is a solvable well-posed or overdetermined problem. Even if the number of grids of the to-be-solved region (observation region) is increased (i.e. N and K are increased), the target optimization function can still be solved. In actual implementation, the dimension size of A and S in the target optimization function can be appropriately increased (i.e. N and K are increased) to obtain better angle super-resolution and distance super-resolution.

[0099] On the basis of constructing the target optimization function as in formula (2), further, a convex optimization method is used to iteratively solve the target optimization function to obtain the target echo signals corresponding to each scattering element in the observation region, and obtain the super-resolution active sonar image in direction and distance.

[0100] Specifically, first, the constructed target optimization function is weighted to speed up its convergence process and obtain better sound field reconstruction effect; then, the interior point method in the convex optimization method is used to iteratively solve the weighted target optimization function, and the weight is updated once at the end of each iteration.

[0101] The interior point method can be implemented in the CVX toolbox of Matlab.

[0102] The interior point method with multiple rounds of weighting iteration is used to solve the target optimization function. Specifically, first, the diagonal elements of the weighting coefficient diagonal matrix W are set to the results of the conventional beamforming of the observation region, and the interior point method is used to solve the target optimization function to obtain the initial target echo signal.

[0103] Then, the diagonal elements of the weighting coefficient diagonal matrix are set to the inverse of the initial target echo signal, that is, a smaller amplitude coefficient in the target echo signal is given a larger weight, and the interior point method is used again to solve the target optimization function, which can make the small amplitude coefficient quickly converge to 0 to obtain the candidate target echo signal.

[0104] The process of repeatedly (generally three times) updating the diagonal elements of the weighting coefficient diagonal matrix and solving the target optimization function is repeated, and the target scattering amplitude matrix of the observation region, that is, the target echo signals corresponding to each scattering element in the observation region, is obtained, so that the super-resolution active sonar image is obtained.

[0105] In some other embodiments, the target optimization function is represented as follows in equation (3).

[0106]

[0107] In equation (3), μ represents a hyperparameter that adjusts the proportion of the regular term and the quadratic term.

[0108] It should be noted that equation (3) is transformed by the Lagrange multiplier method from equation (2), and both are equivalent. Similarly, the hyperparameter μ can also be appropriately set by estimating the reverberation energy level. For specific setting process, refer to the hyperparameter ε.

[0109] In the case of equation (3) as the target optimization function, the solving method will also change accordingly.

[0110] Specifically, for the target optimization function (3), a gradient descent method of multiple rounds of weighted iteration in convex optimization method can be used to solve it. After each gradient descent ends, the diagonal elements of the weighted coefficient diagonal matrix are set to the reciprocals of the target scattering amplitude matrix, that is, the coefficients with smaller amplitudes in the target echo signal are given greater weights. After such a weighted gradient descent method is iterated multiple times (generally three times), a more sparse target scattering amplitude matrix is obtained, that is, the target echo signal corresponding to each scatterer in the observation area, so as to obtain a super-resolution active sonar image.

[0111] In this embodiment, the target optimization function is solved by using an interior point method or a gradient descent method of multiple rounds of weighted iteration, to obtain the target echo signal corresponding to each scatterer in the observation area. The target echo signal corresponding to each scatterer forms a super-resolution active sonar image, and the active sonar image contains the detection target. This method adds the waveform prior information of the echo signal to the perception matrix, and constructs a target optimization function using the echo signal and the perception matrix, thereby reducing the inherent sparsity of the target echo signal in the observation area, achieving super-resolution that breaks through the sampling rate limit of the echo signal in distance, effectively suppressing the reverberation interference of the active sonar, and further greatly improving the quality of the received echo signal of the active sonar, thereby improving the accuracy of underwater target detection and recognition.

[0112] As described in the above embodiments, the active sonar super-resolution imaging method based on compressed sensing beamforming proposed in the embodiments of the present application can achieve super-resolution in angle and distance, and the super-resolution in distance can break through the sampling rate limit of the echo signal. In addition, the embodiments of the present application also achieve effective reverberation suppression.

[0113] To be specific, in one embodiment, Figure 3The schematic diagram of the resolution improvement effect in angle and distance of the active sonar super-resolution imaging method based on compressed sensing beamforming provided by the embodiment of the application is shown.

[0114] In Figure 3 , the experiment simulates a forward-looking sonar array with dozens of array elements, and the array element spacing is half a wavelength. The transmitted signal is a linear frequency modulation signal with a carrier frequency of 100 kHz and a bandwidth of 20 kHz. Three point targets are set in the experiment, and the intensity of the point targets is 0 dB. The angles of the point targets are θ = [20°, 20°, 23°], and the distances of the point targets are x = [51, 51 + 1.75Δx, 51] m (Δx = 0.0187 m).

[0115] Figure 3 (a) in (d) is the result obtained by the conventional beamforming and the matched filtering, Figure 3 (b) and (c) in (d) are respectively Figure 3 The cross-sectional view of (a) in (d) at θ = 20° and x = 51 m, from which it can be seen that the angle resolution and the distance resolution of the forward-looking sonar image obtained by the conventional beamforming and the matched filtering method are very low, and the three point targets cannot be distinguished at all.

[0116] Figure 3 (d) in (d) is the result obtained by the frequency domain compressed sensing beamforming method, Figure 3 (e) and (f) in (d) are respectively Figure 3 The cross-sectional view of (d) in (d) at θ = 20° and x = 51 m, from which it can be seen that the angle resolution of the frequency domain compressed sensing beamforming method is greatly improved compared with the conventional beamforming, and the point targets at two different angles can be distinguished, but the two point targets with a distance of less than 0.05 m cannot be distinguished.

[0117] Figure 3 (g) in (d) is the result obtained by the active sonar super-resolution imaging method based on compressed sensing beamforming proposed in the application, Figure 3 (h) and (i) in (d) are respectively Figure 4 The cross-sectional view of (g) in (d) at θ = 20° and x = 51 m, from which it can be seen that the method proposed in the application can accurately solve the three point targets, and high resolution is obtained in the angle direction and the distance direction, which exceeds the conventional beamforming and the frequency domain compressed sensing beamforming method.

[0118] In another embodiment, Figure 4 The schematic diagram of the resolution improvement and reverberation suppression effect of the active sonar super-resolution imaging method based on compressed sensing beamforming provided by the embodiment of the application on the side-scan sonar simulation image is shown.

[0119] In Figure 4 , an experimental simulation of a side-scan sonar array with dozens of elements is performed, the transmitted signal is a linear frequency modulation signal with a carrier frequency of 100 kHz and a bandwidth of 20 kHz. A simulated cylindrical target is set in the experiment, the cylindrical target is 2.8 m long and 0.28 m in radius, and is placed vertically to the array direction. During the experiment, reverberation with a signal-to-reverberation ratio of 10 dB is added to the data.

[0120] Figure 4 (a) in (a) of Figure 4 (b) in (b) of Figure 4 is a cross-section graph of (a) in (a) of from which it can be seen that the sidelobe interference exists in the cylindrical target echo obtained by the conventional beamforming, and the reverberation with a signal-to-reverberation ratio of 10 dB affects the quality of the target echo signal.

[0121] Figure 4 (c) in (c) of Figure 5 (d) in (d) of Figure 5 is a cross-section graph of (c) in (c) of from which it can be seen that the method proposed in the present application can effectively suppress the sidelobe in the distance direction in the side-scan sonar image, improve the distance resolution, and effectively suppress the reverberation, thereby improving the signal-to-reverberation ratio to 43 dB.

[0122] In another embodiment, Figure 5 shows the distance super-resolution effect achieved by the active sonar super-resolution imaging method based on the compressive sensing beamforming provided by the embodiment of the present application on a forward-looking sonar.

[0123] In Figure 5 , three point targets are set, which are located at x = [51, 51 + 4.25Δx, 51-3.5Δx] m. Figure 5 The results in (a) of Figure 6 (a) in (a) of

[0124] Figure 6 (b) in (b) of Figure 6The compressed sensing beamforming method shown in (c) in the table 1 adopts 4 times super resolution, and the positions of three point targets are all accurately recovered, and the amplitudes are also basically correctly recovered.

[0125] Therefore, the super resolution capability of the method provided by the application can effectively improve the target resolution capability of the active sonar in the distance direction, and make the imaging of the active sonar more accurate.

[0126] In other embodiments, Figure 6 The processing result schematic diagram of the active sonar super resolution imaging method based on compressed sensing beamforming provided by the embodiment of the application is shown. Figure 7 The data shown in the table 2 are actual echoes obtained by scanning two seabed cylindrical targets.

[0127] Figure 7 The images (a) and (b) in the table 2 are obtained after the conventional beamforming and matched filtering processing of the side scan sonar data, and it can be seen that the resolution of the image is low, and the reverberation interference is serious, so that the characteristic echoes of the targets are partially covered, and the effect of underwater target detection and identification is affected.

[0128] The results after the processing of the method provided by the application are respectively shown in (c) and (d) in the table 2, and it can be seen that the resolution of the characteristic echoes of the cylindrical targets is significantly improved, and the reverberation is effectively suppressed. Figure 8

[0129] In summary, the active sonar super resolution imaging method based on compressed sensing beamforming provided by the application can realize the super resolution and reverberation suppression of the active sonar, and good effects are achieved in objective indicators and subjective quality.

[0130] Corresponding to the active sonar super resolution imaging method based on compressed sensing beamforming described in each of the above embodiments, the application further provides an active sonar super resolution imaging device based on compressed sensing beamforming.

[0131] Specifically, Figure 8 The structure schematic diagram of the active sonar super resolution imaging device based on compressed sensing beamforming provided by the embodiment of the application is shown.

[0132] As ​ ​As shown, the device comprises: an echo signal acquisition module 710, configured to acquire echo signals reflected by scatterers in an observed area after being reflected by active sonar array elements; a perception matrix generation module 720, configured to generate a perception matrix according to waveform prior information of the echo signals, the perception matrix being a mapping matrix of the scatterers in the observed area to the active sonar array elements; a target optimization function construction module 730, configured to construct a target optimization function according to the echo signals and the perception matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem; and a target optimization function solving module 740, configured to solve the target optimization function to obtain target echo signals corresponding to each of the scatterers in the observed area; wherein the target echo signals corresponding to each of the scatterers form a super-resolution active sonar image, and the active sonar image contains a detection target.

[0133] In this embodiment, the echo signal acquisition module 710 acquires echo signals reflected by scatterers in an observed area after being reflected by active sonar array elements, the perception matrix generation module 720 generates a perception matrix according to waveform prior information of the echo signals, the perception matrix being a mapping matrix of the scatterers in the observed area to the active sonar array elements, and then the target optimization function construction module 730 constructs a target optimization function of compressed sensing beamforming according to the echo signals and the perception matrix, the target optimization function being a weighted sparse compressed sensing beamforming problem. Thus, the target optimization function solving module 740 solves the target optimization function to obtain target echo signals corresponding to each of the scatterers in the observed area, and the target echo signals corresponding to each of the scatterers form a super-resolution active sonar image, and the active sonar image contains a detection target. The device adds waveform prior information of the echo signals to the perception matrix, and constructs a target optimization function using the echo signals and the perception matrix, thereby reducing the inherent sparsity of the target echo signals in the observed area, achieving super-resolution that breaks through the sampling rate limit of the echo signals in distance, effectively suppressing reverberation interference of the active sonar, and greatly improving the quality of the active sonar received echo signals, thereby improving the accuracy of underwater target detection and recognition.

[0134] It should be noted that the active sonar super-resolution imaging device based on compressed sensing beamforming provided in this embodiment can be mutually corresponding to the active sonar super-resolution imaging method based on compressed sensing beamforming described in the above embodiments, and will not be described here.

[0135] ​ An example of an electronic device is shown in the physical structure diagram as ​As shown, the electronic device can include a processor 810, a communications interface 820, a memory 830, and a communications bus 840, wherein the processor 810, the communications interface 820, and the memory 830 complete mutual communication through the communications bus 840. The processor 810 can invoke a logic instruction in the memory 830 to execute a compressive sensing beamforming-based active sonar super-resolution imaging method, which includes: collecting echo signals reflected by scatterers in an observed area after being received by an active sonar array element; generating a sensing matrix according to waveform prior information of the echo signals, the sensing matrix being a mapping matrix of the scatterers in the observed area to the active sonar array element; constructing a target optimization function of compressive sensing beamforming according to the echo signals and the sensing matrix, the target optimization function being a weighted sparse compressive sensing beamforming problem; solving the target optimization function to obtain target echo signals corresponding to each scatterer in the observed area; wherein the target echo signals corresponding to each scatterer form a super-resolution active sonar image, and the active sonar image contains a detection target.

[0136] In addition, the logic instruction in the memory 830 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0137] In another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the method of compressive sensing beamforming-based active sonar super-resolution imaging provided by the above method, and the method comprises: collecting echo signals reflected by scatterers in an observed region and received by an active sonar array element; generating a sensing matrix according to waveform prior information of the echo signals, the sensing matrix being a mapping matrix of the scatterers in the observed region to the active sonar array element; constructing a target optimization function of compressive sensing beamforming according to the echo signals and the sensing matrix, the target optimization function being a weighted sparse compressive sensing beamforming problem; and solving the target optimization function to obtain target echo signals corresponding to each scatterer in the observed region; wherein the target echo signals corresponding to each scatterer form a super-resolution active sonar image, and the active sonar image contains a detection target.

[0138] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0139] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary universal hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0140] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An active sonar super-resolution imaging method based on compressed sensing beamforming, characterized in that, include: Collect echo signals received by the active sonar array elements after reflection by scattering elements in the observation area; Based on the prior waveform information of the echo signal, a sensing matrix is ​​generated, which is a mapping matrix from scattering elements to active sonar array elements in the observation area. Based on the echo signal and the sensing matrix, a target optimization function for compressed sensing beamforming is constructed, wherein the target optimization function is a weighted sparse compressed sensing beamforming problem. Solving the target optimization function yields the target echo signal corresponding to each scattering element in the observation region; The target echo signal corresponding to each scattering element forms a super-resolution active sonar image, and the active sonar image contains the detected target.

2. The active sonar super-resolution imaging method based on compressed sensing beamforming according to claim 1, characterized in that, The received echo signal is preprocessed, including: When the echo signal is a bandpass signal, bandpass filtering, quadrature demodulation, and matched filtering are performed on the echo signal in sequence. or, When the echo signal is a single-frequency signal, quadrature demodulation and envelope extraction operations are performed sequentially on the echo signal; correspondingly, constructing the target optimization function based on the echo signal and the sensing matrix includes: Based on the preprocessed echo signal and the sensing matrix, a target optimization function is constructed.

3. The active sonar super-resolution imaging method based on compressed sensing beamforming according to claim 1, characterized in that, The step of generating a sensing matrix based on the prior waveform information of the echo signal includes: Based on the single acoustic pulse signal emitted by the active sonar array element, obtain the waveform prior information corresponding to the echo signal; The waveform prior information is added to the initial mapping matrix to obtain the perception matrix; The single acoustic pulse signal corresponds to the echo signal, the waveform prior information includes the delay, phase and waveform of each scattering element to the active sonar array element in the observation area, and all matrix elements of the initial mapping matrix are zero.

4. The active sonar super-resolution imaging method based on compressed sensing beamforming according to any one of claims 1-3, characterized in that, The objective optimization function is expressed as follows: in, denoted as the solved target echo signal / target scattering amplitude matrix, S represents the scattering amplitude matrix corresponding to each scattering element in the observation area, W represents the weighted coefficient diagonal matrix, D represents the echo signal received by the active sonar array element, A represents the sensing matrix, and ∈ represents the hyperparameters constraining the fitting error.

5. The active sonar super-resolution imaging method based on compressed sensing beamforming according to claim 4, characterized in that, Solving the objective optimization function includes: The objective optimization function is solved using the interior-point method with multiple rounds of weighted iteration.

6. The active sonar super-resolution imaging method based on compressed sensing beamforming according to claim 5, characterized in that, The method of solving the objective optimization function using a multi-round weighted iterative interior-point method includes: The weighted coefficient diagonal matrix is ​​set as the identity matrix, and the interior point method is used to solve the target optimization function to obtain the initial target echo signal; The diagonal elements of the weighted coefficient diagonal matrix are set to the reciprocal of the amplitude of the initial target echo signal, and the interior point method is used again to solve the target optimization function to obtain the candidate target echo signal. The diagonal elements of the weighted coefficient diagonal matrix are repeatedly updated iteratively, and the target optimization function is solved to obtain the target echo signal corresponding to each scattering element in the observation area.

7. The active sonar super-resolution imaging method based on compressed sensing beamforming according to claim 6, characterized in that, The hyperparameters of the constraint fitting error are determined through the following steps: The reverberation energy level is estimated based on the correlation between the echo signals received by different active sonar array elements. The hyperparameters of the constraint fitting error are determined based on the estimated reverberation energy level.

8. An active sonar super-resolution imaging device based on compressed sensing beamforming, characterized in that, include: The echo signal acquisition module is used to acquire the echo signal received by the active sonar array element and reflected by the scattering element in the observation area; The sensing matrix generation module is used to generate a sensing matrix based on the waveform prior information of the echo signal. The sensing matrix is ​​a mapping matrix from scattering elements to active sonar array elements in the observation area. The objective optimization function construction module is used to construct an objective optimization function for compressed sensing beamforming based on the echo signal and the sensing matrix. The objective optimization function is a weighted sparse compressed sensing beamforming problem. The target optimization function solving module is used to solve the target optimization function to obtain the target echo signal corresponding to each scattering element in the observation area. The target echo signal corresponding to each scattering element forms a super-resolution active sonar image, and the active sonar image contains the detected target.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the active sonar super-resolution imaging method based on compressed sensing beamforming as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the active sonar super-resolution imaging method based on compressed sensing beamforming as described in any one of claims 1 to 7.

Images