A method and system for correcting the azimuth of shallow water acoustic and seabed ground acoustic homologous peaks based on deep learning

Abstract

The present application provides a kind of shallow water sound and seabed ground sound homologous peak azimuth correction method and system based on deep learning.The method obtains hydrophone array signal, seabed seismograph signal and array geometry and azimuth reference in the same observation period, extracts water sound time delay candidate peak and ground sound azimuth spectrum candidate peak, forms candidate azimuth correction peak pair and calculates candidate absolute azimuth correction amount, generates azimuth correction angle set under different sub-band, hydrophone element pair and seismograph component combination, obtains direction concentration degree using Rayleigh test, and embeds the direction concentration degree into the Dirichlet distribution parameter generation process of Evidential Deep Learning model, limits homologous evidence, determines homologous peak pair according to evidence proportion and uncertainty, determines absolute azimuth correction amount and corrects water sound positioning azimuth;The method can be used for shallow target positioning, and reduces the influence of non-homologous peak misadoption on positioning azimuth.

Description

Technical Field

[0001] This invention relates to the field of shallow sea acoustic positioning and seabed acoustic signal processing technology, and in particular to a method and system for correcting the orientation of shallow sea acoustic and seabed acoustic sources based on deep learning. Background Technology

[0002] Target localization in shallow waters typically relies on underwater buoy hydrophone arrays to receive radiated or scattered sound from a target. The target's azimuth is obtained through the time difference of arrival between array elements, cross-correlation peaks, and array geometry. In shallow water environments, sound velocity profiles, seabed reflections, seabed sediment coupling, and platform installation azimuth errors all collectively affect underwater acoustic azimuth estimation. Seabed seismometers can receive the ground acoustic response formed by the coupling of the underwater sound field to the seabed medium. Multi-component seismometer signals can yield ground acoustic direction information through component projection, polarization analysis, or azimuth spectrum analysis. When hydrophone arrays and seabed seismometers respond to the same target during the same observation period, a common source relationship exists between the underwater acoustic time delay peak and the ground acoustic azimuth spectrum peak, which can be used for azimuth correction. Reasonably utilizing the correspondence between underwater acoustic observations and ground acoustic observations helps to establish a unified azimuth correction quantity under a unified azimuth benchmark during shallow water buoy localization.

[0003] In existing technologies, underwater acoustic positioning typically involves first performing cross-correlation calculations on the received signals from different hydrophone array element pairs to extract time delay peaks that satisfy the conditions of peak amplitude, peak width, and sidelobe ratio. Then, the underwater acoustic azimuth is calculated by combining array aperture, element spacing, installation direction, and sound velocity parameters. Seafloor acoustic azimuth estimation generally utilizes three-component or multi-component seismograph signals, projecting or accumulating energy according to component azimuth references to form an azimuth spectrum and extracting the azimuth of spectral peaks. For scenarios involving joint underwater and ground acoustic positioning, a common approach is to compare the angle difference between candidate underwater acoustic azimuths and ground acoustic spectral peak azimuths, or to filter candidate peaks by combining peak intensity, spectral peak width, sidelobe ratio, artificial rules, and training models, and then select a peak pair as the basis for azimuth correction.

[0004] Existing methods for selecting underwater and ground acoustic peak pairs often rely on single frequency bands, single array element pairs, single component combinations, or local peak shape characteristics, making it difficult to fully reflect whether the same candidate peak pair maintains a consistent azimuth relationship under various observation conditions. In shallow seas, multipath propagation, reflection peaks, ground acoustic pseudo-spectral peaks, and local strong noise peaks can create seemingly reasonable angular differences, leading to the inclusion of non-homogeneous peak pairs in the azimuth correction process. If existing training models primarily output adoption results based on peak characteristics, lacking centralized constraints on circumferential angular direction and evidence uncertainty rejection mechanisms, erroneous peak pairs may be mistakenly adopted in complex shallow sea scenarios.

[0005] Therefore, there is an urgent need in the existing technology for a method to correct the orientation of the same source peak of shallow sea acoustics and seabed acoustics that can overcome the shortcomings of the existing technology. Summary of the Invention

[0006] This invention proposes a method and system for azimuth correction of homologous peaks in shallow water acoustic and seabed ground acoustic spectrum based on deep learning. The core technical problem to be solved by this application is: in scenarios where shallow water moorings simultaneously obtain signals from underwater acoustic arrays and seabed seismometers, how to determine peak pairs with an acceptable homologous relationship from multiple candidate peaks of underwater acoustic time delay and multiple candidate peaks of ground acoustic azimuth spectrum, and how to reject ground acoustic azimuth correction when the homologous relationship is insufficient, thereby avoiding non-homogeneous peaks from participating in underwater acoustic positioning azimuth correction.

[0007] This invention is achieved through the following technical solution: This invention proposes a method for correcting the azimuth of co-source peaks of shallow sea acoustics and seabed acoustics based on deep learning, the method comprising: S1. Obtain the hydrophone array signal of the shallow sea mooring, the seabed seismograph signal, and the array geometry and orientation reference. Obtain the underwater acoustic time delay candidate peak based on the hydrophone array signal, and obtain the ground acoustic orientation spectrum candidate peak based on the seabed seismograph signal. S2. Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair with an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. S3. For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction amount under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. S4. Perform Rayleigh test for circumferential angle data on the azimuth correction angle set to obtain the directional concentration of candidate azimuth correction peak pairs. Embed the directional concentration into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. The homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. S5. Based on the Dirichlet distribution and evidence threshold used for homology determination, homology determination is performed on candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or to reject azimuth correction peak pairs. S6. Determine the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, use the absolute azimuth correction amount to correct the underwater acoustic positioning azimuth, and output the shallow sea target positioning result in combination with array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be azimuth correction peak pairs not adopted, output the shallow sea target positioning result without adopting ground acoustic azimuth correction.

[0008] Furthermore, S1 specifically refers to: Acquire hydrophone array signals and seabed seismograph signals of shallow sea moorings within the same observation period, and obtain the array geometry and azimuth reference corresponding to the hydrophone array signals and the seabed seismograph signals; Based on the array geometry and orientation reference, cross-correlation calculations are performed on the received signals of different hydrophone array element pairs in the hydrophone array signal to form underwater acoustic time delay peak information; Based on the underwater acoustic delay peak information and the preset peak value conditions, candidate underwater acoustic delay peaks are determined; Based on the array geometry and azimuth reference, the azimuth spectrum of the received signals of different seismograph component combinations in the submarine seismograph signal is calculated to form ground acoustic azimuth spectrum peak information, and ground acoustic azimuth spectrum candidate peaks are determined according to the ground acoustic azimuth spectrum peak information and preset peak conditions.

[0009] Furthermore, S2 specifically refers to: Based on the array geometry and orientation reference, the candidate peaks of underwater acoustic time delay are converted into underwater acoustic positioning orientations, and the peak positions of the candidate peaks of the ground acoustic orientation spectrum are determined as the orientations of the ground acoustic spectrum peaks. Based on the underwater acoustic positioning azimuth and the ground acoustic spectrum peak azimuth, a pairing relationship is established between an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak to form a candidate azimuth correction peak pair. Based on the underwater acoustic positioning azimuth and the geosonic spectrum peak azimuth in the candidate azimuth correction peak pair, the angular difference between the geosonic spectrum peak azimuth and the underwater acoustic positioning azimuth is calculated to form the candidate absolute azimuth correction amount. The candidate absolute azimuth correction values ​​are associated with the corresponding candidate azimuth correction peak pairs, so that the candidate azimuth correction peak pairs are used as inputs to generate the set of azimuth correction angles.

[0010] Furthermore, S3 specifically refers to: Based on the candidate azimuth correction peak pairs, extract the underwater acoustic time delay candidate peaks and ground acoustic azimuth spectrum candidate peaks that make up the candidate azimuth correction peak pairs, and read the candidate absolute azimuth correction values ​​corresponding to the candidate azimuth correction peak pairs to form peak-to-azimuth relationships. Based on the peak orientation relationship, the hydrophone array signal is decomposed into different sub-frequency bands, and different hydrophone array element pairs participating in the underwater acoustic time delay candidate peak localization are selected in each different sub-frequency band to form an underwater acoustic observation signal. Based on the underwater acoustic observation signal and the array geometry and orientation reference, calculate the underwater acoustic observation orientation corresponding to the underwater acoustic time delay candidate peak for different sub-frequency bands and different hydrophone array element pairs; Based on the peak orientation relationship, the submarine seismograph signal is combined into components according to different seismograph component combinations, and the geosound observation orientation corresponding to the candidate peak of the geosound orientation spectrum is calculated under each different seismograph component combination. A different sub-frequency band, a different pair of hydrophone array elements, and a different combination of seismograph components are determined as an observation condition. Based on the underwater acoustic observation azimuth, the ground acoustic observation azimuth, and the candidate absolute azimuth correction amount under the same observation condition, the azimuth correction angle corresponding to the candidate absolute azimuth correction amount under each observation condition is generated. Based on the candidate absolute azimuth correction, the azimuth correction angle under each observation condition is aligned with the circumferential angle, and the azimuth correction angle under each observation condition is updated. For the same candidate azimuth correction peak pair, the azimuth correction angles under all observation conditions after the update are collected by circular angle aggregation to form an azimuth correction angle set, so that the azimuth correction angle set represents the degree of directional concentration of the candidate azimuth correction peak pair under multiple observation conditions.

[0011] Furthermore, S4 specifically refers to: Based on the candidate azimuth correction peak pairs, read the azimuth correction angle set corresponding to the candidate azimuth correction peak pairs, and convert the azimuth correction angles under each observation condition in the azimuth correction angle set into circular angle data for Rayleigh test. Perform a Rayleigh test on the circumferential angle data to calculate the composite length of the azimuth correction angle set in the circumferential direction and the significance of the test. Based on the directional synthesis length and the test significance quantity, the directional concentration of the candidate azimuth correction peak pair is determined, and the directional concentration is associated with the candidate azimuth correction peak pair that generates the azimuth correction angle set. The candidate underwater acoustic delay candidate peak, ground acoustic azimuth spectrum candidate peak, array geometry and azimuth reference, and candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair are input into the Evidential Deep Learning model to obtain the initial homology evidence value and non-adoption evidence value of the candidate azimuth correction peak pair. In the process of generating the Dirichlet distribution parameters corresponding to the homology evidence in the Evidential Deep Learning model, a constraint relationship is established on the initial homology evidence value based on the directional concentration, and the initial homology evidence value is updated according to the constraint relationship to form the homology evidence value; Based on the homology evidence value and the non-adoption evidence value, a Dirichlet distribution for homology determination of candidate orientation correction peak pairs is generated, such that the Dirichlet distribution for homology determination corresponds to the candidate orientation correction peak pairs.

[0012] Furthermore, S5 specifically refers to: Receive the Dirichlet distribution for homology determination and the evidence determination threshold corresponding to each candidate orientation correction peak pair, and determine the homology evidence value formed after directional concentration restriction and the corresponding non-adoption evidence value from the Dirichlet distribution for homology determination, thus forming the evidence value relationship of the candidate orientation correction peak pair. Based on the aforementioned evidence value relationship, calculate the proportion of source evidence values ​​in the Dirichlet distribution for source determination, calculate the proportion of non-adopted evidence values ​​in the Dirichlet distribution for source determination, and calculate the evidence uncertainty based on the Dirichlet distribution for source determination. Based on the aforementioned evidence judgment threshold, the same-source evidence threshold, the non-adoption evidence threshold, and the uncertainty threshold are determined to form the evidence judgment conditions for candidate orientation correction peak pairs. The proportion of evidence from the same source is compared with the threshold for evidence from the same source, the proportion of evidence not accepted is compared with the threshold for evidence not accepted, and the uncertainty of the evidence is compared with the threshold for uncertainty, thus forming the evidence comparison result of the candidate orientation correction peak pair; Based on the evidence comparison results, if the proportion of evidence from the same source is greater than or equal to the threshold for evidence from the same source, the proportion of evidence not adopted is less than the threshold for evidence not adopted, and the amount of uncertainty in the evidence is less than or equal to the threshold for uncertainty, the candidate orientation correction peak pair will be determined as an orientation correction peak pair from the same source. Based on the evidence comparison results, if the proportion of evidence from the same source is less than the threshold for evidence from the same source, the proportion of evidence not adopted is greater than or equal to the threshold for evidence not adopted, or the uncertainty of the evidence is greater than the threshold for uncertainty, the candidate orientation correction peak pair will be determined as the orientation correction peak pair not adopted. The homology determination is performed on all candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or azimuth correction peak pairs that are not adopted. Homologous azimuth correction peak pairs are used as input to determine the absolute azimuth correction amount, and azimuth correction peak pairs that are not adopted are used as input to determine whether all candidate azimuth correction peak pairs are azimuth correction peak pairs that are not adopted.

[0013] Furthermore, S6 specifically refers to: Based on the aforementioned homologous azimuth correction peak pair, read the candidate absolute azimuth correction amount and the Dirichlet distribution for homology determination corresponding to the homologous azimuth correction peak pair, and determine the homology evidence value of the homologous azimuth correction peak pair based on the Dirichlet distribution for homology determination. Based on the aforementioned evidence values ​​of the same origin, the absolute azimuth correction amount is determined from the candidate absolute azimuth correction amounts corresponding to the same azimuth correction peak pairs. Based on the absolute azimuth correction amount, the underwater acoustic positioning azimuth is angularly corrected to form the corrected underwater acoustic positioning azimuth, and the shallow sea target positioning result is output based on the corrected underwater acoustic positioning azimuth, array geometry and azimuth reference. If all candidate azimuth correction peak pairs are determined to be unacceptable, the underwater acoustic positioning azimuth is not corrected using absolute azimuth correction, and the shallow sea target positioning result without ground acoustic azimuth correction is output based on the underwater acoustic positioning azimuth, array geometry and azimuth reference.

[0014] This invention also proposes a deep learning-based system for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics, the system comprising: Acquisition module: Acquires hydrophone array signals, seabed seismograph signals, and array geometry and azimuth references of shallow sea moorings; obtains underwater acoustic time delay candidate peaks based on hydrophone array signals; and obtains ground acoustic azimuth spectrum candidate peaks based on seabed seismograph signals. Determination module: Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair by combining an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. Generation module: For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction quantities under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. The verification module performs a Rayleigh test on the azimuth correction angle set for circumferential angle data to obtain the directional concentration of candidate azimuth correction peak pairs. The directional concentration is then embedded into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. Finally, the homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. Decision module: Based on the Dirichlet distribution and evidence decision threshold used for homology determination, the candidate azimuth correction peak pairs are determined to be homology, and homology azimuth correction peak pairs are obtained or azimuth correction peak pairs are not adopted. Output module: Determines the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, corrects the underwater acoustic positioning azimuth using the absolute azimuth correction amount, and outputs the shallow sea target positioning result by combining array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be unacceptable, the shallow sea target positioning result without ground acoustic azimuth correction is output.

[0015] The present invention also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the deep learning-based method for correcting the orientation of shallow sea acoustic and seabed acoustic sources.

[0016] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the deep learning-based method for correcting the orientation of shallow sea acoustic and seabed acoustic sources.

[0017] The beneficial effects of this invention are: 1. This invention proposes an improved method for azimuth correction of source peaks in shallow sea acoustics and seabed ground acoustics. The key improvement lies in incorporating the circumferential centrality of the azimuth correction angle set into the evidence generation process of the Evidential Deep Learning model. Instead of directly outputting the adoption result after inputting the peak amplitude, peak width, sidelobe ratio of the underwater acoustics and the peak amplitude and width of the ground acoustic spectrum, this invention first performs a Rayleigh test on the azimuth correction angles formed by the same candidate azimuth correction peak pair under various observation conditions to obtain the directional centrality. Then, during the Dirichlet distribution parameter generation process, the directional centrality is used to constrain the initial source evidence value, while ensuring that the unadopted evidence value is not amplified by the directional centrality. Through this evidence constraint method, only when the azimuth correction angles exhibit circumferential centrality around the candidate absolute azimuth correction amount can source evidence enter the subsequent judgment process, reducing the possibility of shallow sea reflection peaks, ground acoustic pseudo-spectral peaks, or local strong noise peaks being mistakenly identified as source peak pairs.

[0018] 2. This invention proposes a method for constructing azimuth correction angle sets under multiple observation conditions. It binds a candidate azimuth correction peak pair to a candidate absolute azimuth correction value, and calculates the underwater acoustic observation azimuth and ground acoustic observation azimuth under different sub-frequency bands, different hydrophone array element pairs, and different seismograph component combinations. Then, it generates multiple azimuth correction angles using the candidate absolute azimuth correction value as the circumferential angle alignment reference. Existing processing methods typically rely on a single frequency band, a single array element pair, a single seismograph component combination, or a single local peak shape feature, making it difficult to determine whether candidate peak pairs maintain a consistent azimuth relationship after changes in observation conditions. This invention, through circumferential angle alignment and aggregation, makes the homology determination of candidate peak pairs not only depend on peak strength but also further constrained by multi-condition directional consistency, thus making it more suitable for positioning scenarios where shallow sea multipath propagation and seabed coupling responses coexist.

[0019] 3. This invention proposes a mechanism for accepting and rejecting homologous peak pairs in the shallow-sea mooring positioning process. Candidate peaks for underwater acoustic time delay and candidate peaks for ground acoustic azimuth spectrum are first combined to form candidate azimuth correction peak pairs. Then, based on homology determination, the proportion of homologous evidence, the proportion of rejected evidence, and the amount of evidence uncertainty are calculated using Dirichlet distribution, and an acceptance decision is made in conjunction with a threshold. For multiple candidate peak pairs sharing the same underwater acoustic time delay or the same ground acoustic azimuth spectrum candidate peak, this invention resolves conflicts based on the amount of evidence uncertainty and the proportion of homologous evidence, avoiding multiple mutually exclusive peak pairs from participating in azimuth correction simultaneously. When all candidate peak pairs fail to meet the acceptance criteria, this invention outputs a shallow-sea target positioning result that does not accept ground acoustic azimuth correction, avoiding the forced introduction of ground acoustic azimuth correction when homology is insufficient. Through the above process, this invention addresses the problem of reliably confirming homologous peak pairs in shallow-sea scenarios, forming a closed-loop processing method from candidate peak extraction, peak pair construction, direction concentration constraints, evidence determination to positioning output. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a flowchart of a method for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics based on deep learning, as proposed in this invention. Figure 2 This is a flowchart illustrating the determination of candidate peaks for underwater acoustic time delay and candidate peaks for ground acoustic azimuth spectrum in a method for correcting the azimuth of shallow sea acoustic and seabed acoustic sources based on deep learning proposed in this invention. Figure 3This is a flowchart for determining candidate azimuth correction peak pairs and candidate absolute azimuth correction values ​​in a deep learning-based method for azimuth correction of shallow sea acoustics and seabed acoustics. Figure 4 This is a flowchart illustrating the formation of the azimuth correction angle set in a deep learning-based method for azimuth correction of the common source peaks of shallow sea acoustics and seabed acoustics proposed in this invention. Figure 5 The flowchart for determining the direction concentration and homology of a method for correcting the azimuth of shallow sea acoustics and seabed acoustics based on deep learning proposed in this invention is generated using Dirichlet distribution. Figure 6 This is a flowchart of the homology determination process for a method for correcting the orientation of shallow sea acoustic and seabed acoustic sources based on deep learning, as proposed in this invention. Figure 7 This is a flowchart illustrating the determination of the absolute orientation correction amount and the output of shallow sea target positioning results for a method for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics based on deep learning proposed in this invention. Figure 8 This is a superimposed comparison of shallow sea target positioning results before and after applying the proposed deep learning-based method for correcting the azimuth of the same source peaks of shallow sea acoustics and seabed acoustics. The figure illustrates the spatial difference in shallow sea target positioning results before and after applying the proposed method. The horizontal axis represents distance, ranging from 1 to 5 km, and the vertical axis represents lateral deviation in meters, all represented by a unified horizontal coordinate system. The black solid line and black dots represent the reference target position or actual track, the red dashed line and red hollow dots represent the positioning results without applying the proposed method, and the blue solid line and blue hollow dots represent the positioning results after correction using the proposed method. As can be seen from the figure, without applying the proposed method, there is a significant shift in the positioning azimuth, with an azimuth error of approximately 8° to 25° and a horizontal error of approximately 100 to 500 m. After applying the proposed method, the positioning point significantly converges to the reference track, with the azimuth error reduced to approximately 1° to 4° and the horizontal error reduced to approximately 20 to 80 m. This figure visually demonstrates that the proposed method can improve the consistency between the shallow sea target positioning results and the reference target position through the determination of the same source peak and absolute azimuth correction. Figure 9This figure shows the source determination results of Rayleigh directional concentration constrained EDL evidence for a deep learning-based method for correcting the azimuth of shallow sea acoustic and seabed acoustic sources proposed in this invention. The figure is used to illustrate the influence of Rayleigh directional concentration on Evidential Deep Sound (EDL) evidence. The constraint effect of homology evidence on the Learning model is shown on the left, which is the circular distribution of 24 azimuth correction angles. The azimuth correction angles of true homology peak pairs are concentrated near the candidate absolute azimuth correction values, with a directional synthesis length R of approximately 0.82, a significance level p of less than 0.01, and a directional concentration of approximately 0.85. The azimuth correction angles of erroneous peak pairs are scattered, with an R of approximately 0.22, a p of greater than 0.05, and a directional concentration of approximately 0.05. The right side shows the proportion of homology evidence and the uncertainty of evidence. The proportion of homology evidence for true homology peak pairs is approximately 0.78%, higher than the 0.65 threshold, and the uncertainty of evidence is approximately 0.18%, lower than the 0.35 threshold, therefore, homology is accepted. The proportion of homology evidence for erroneous peak pairs is approximately 0.28%, and the uncertainty of evidence is approximately 0.62, therefore, homology is not accepted. This figure illustrates the inhibitory effect of directional concentration on erroneous peak pairs. Figure 10 This is a schematic diagram of the model input composition for a method for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics based on deep learning proposed in this invention. Figure 11 This is a schematic diagram of the direction concentration constraint evidence generation structure for a method for correcting the orientation of shallow water acoustic and seabed acoustic sources based on deep learning proposed in this invention. Figure 12 This is a schematic diagram of the homology determination and shallow sea target localization result output structure of a method for correcting the azimuth of shallow sea acoustic and seabed acoustic homology peaks based on deep learning proposed in this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Specifically, in combination Figures 1-12 This invention proposes a method for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics based on deep learning. The method includes: S1. Obtain the hydrophone array signal of the shallow sea mooring, the seabed seismograph signal, and the array geometry and orientation reference. Obtain the underwater acoustic time delay candidate peak based on the hydrophone array signal, and obtain the ground acoustic orientation spectrum candidate peak based on the seabed seismograph signal. S2. Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair with an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. S3. For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction amount under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. S4. Perform Rayleigh test for circumferential angle data on the azimuth correction angle set to obtain the directional concentration of candidate azimuth correction peak pairs. Embed the directional concentration into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. The homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. S5. Based on the Dirichlet distribution and evidence threshold used for homology determination, homology determination is performed on candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or to reject azimuth correction peak pairs. S6. Determine the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, use the absolute azimuth correction amount to correct the underwater acoustic positioning azimuth, and output the shallow sea target positioning result in combination with array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be azimuth correction peak pairs not adopted, output the shallow sea target positioning result without adopting ground acoustic azimuth correction.

[0024] Optionally, S1 is as follows: Acquire hydrophone array signals and seabed seismograph signals of shallow sea moorings within the same observation period, and obtain the array geometry and azimuth reference corresponding to the hydrophone array signals and the seabed seismograph signals; Based on the array geometry and orientation reference, cross-correlation calculations are performed on the received signals of different hydrophone array element pairs in the hydrophone array signal to form underwater acoustic time delay peak information; Based on the underwater acoustic delay peak information and the preset peak value conditions, candidate underwater acoustic delay peaks are determined; Based on the array geometry and azimuth reference, the azimuth spectrum of the received signals of different seismograph component combinations in the submarine seismograph signal is calculated to form ground acoustic azimuth spectrum peak information, and ground acoustic azimuth spectrum candidate peaks are determined according to the ground acoustic azimuth spectrum peak information and preset peak conditions.

[0025] Terminology Explanation: The array geometry and orientation reference refers to the spatial position of each hydrophone element in the hydrophone array, the spacing between elements, the array aperture, the array installation direction, and the orientation reference of each component of the seafloor seismograph relative to a unified horizontal coordinate system. The preset peak conditions refer to the conditions used to determine candidate peaks of underwater acoustic delay from the underwater acoustic delay peak information, including peak amplitude conditions, peak width conditions, and peak sidelobe ratio conditions. The preset spectral peak conditions refer to the conditions used to determine candidate peaks of the ground acoustic azimuth spectrum from the ground acoustic azimuth spectral peak information, including spectral peak amplitude conditions, spectral peak width conditions, and spectral peak sidelobe ratio conditions. The underwater acoustic time delay peak information refers to the time delay peak position, peak amplitude, peak width, and peak-side lobe ratio obtained by cross-correlation calculation of the received signals of different pairs of hydrophone array elements in the hydrophone array signal. The underwater acoustic time delay candidate peak refers to the time delay peak in the underwater acoustic time delay peak information that meets the preset peak value condition and is retained to form a candidate azimuth correction peak pair. The geosonic azimuth spectrum peak information refers to the azimuth peak azimuth, peak amplitude, peak width, and peak sidelobe ratio obtained by calculating the azimuth spectrum of the received signals of different seismograph component combinations in the submarine seismograph signal. The candidate peaks of the ground acoustic azimuth spectrum refer to the spectral peaks in the ground acoustic azimuth spectrum peak information that meet the preset spectral peak conditions and are retained to form candidate azimuth correction peak pairs.

[0026] Optionally, S2 is as follows: Based on the array geometry and orientation reference, the candidate peaks of underwater acoustic time delay are converted into underwater acoustic positioning orientations, and the peak positions of the candidate peaks of the ground acoustic orientation spectrum are determined as the orientations of the ground acoustic spectrum peaks. Based on the underwater acoustic positioning azimuth and the ground acoustic spectrum peak azimuth, a pairing relationship is established between an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak to form a candidate azimuth correction peak pair. Based on the underwater acoustic positioning azimuth and the geosonic spectrum peak azimuth in the candidate azimuth correction peak pair, the angular difference between the geosonic spectrum peak azimuth and the underwater acoustic positioning azimuth is calculated to form the candidate absolute azimuth correction amount. The candidate absolute azimuth correction values ​​are associated with the corresponding candidate azimuth correction peak pairs, so that the candidate azimuth correction peak pairs are used as inputs to generate the set of azimuth correction angles.

[0027] Terminology Explanation: The underwater acoustic positioning orientation refers to the estimated value of the underwater acoustic direction determined based on the underwater acoustic time delay candidate peak and the array geometry and orientation reference. The azimuth of the ground acoustic spectrum peak refers to the estimated ground acoustic direction corresponding to the peak position of the candidate peak of the ground acoustic azimuth spectrum under the array geometry and azimuth reference. The candidate azimuth correction peak pair refers to a homology determination object consisting of an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and corresponding to a candidate absolute azimuth correction amount. The candidate absolute azimuth correction amount refers to the candidate azimuth correction angle obtained by converting the angle difference between the azimuth of the ground acoustic spectrum peak and the azimuth of the underwater acoustic positioning to a unified azimuth reference based on the array geometry and azimuth reference.

[0028] Optionally, S3 specifically refers to: Based on the candidate azimuth correction peak pairs, extract the underwater acoustic time delay candidate peaks and ground acoustic azimuth spectrum candidate peaks that make up the candidate azimuth correction peak pairs, and read the candidate absolute azimuth correction values ​​corresponding to the candidate azimuth correction peak pairs to form peak-to-azimuth relationships. Based on the peak orientation relationship, the hydrophone array signal is decomposed into different sub-frequency bands, and different hydrophone array element pairs participating in the underwater acoustic time delay candidate peak localization are selected in each different sub-frequency band to form an underwater acoustic observation signal. Based on the underwater acoustic observation signal and the array geometry and orientation reference, calculate the underwater acoustic observation orientation corresponding to the underwater acoustic time delay candidate peak for different sub-frequency bands and different hydrophone array element pairs; Based on the peak orientation relationship, the submarine seismograph signal is combined into components according to different seismograph component combinations, and the geosound observation orientation corresponding to the candidate peak of the geosound orientation spectrum is calculated under each different seismograph component combination. A different sub-frequency band, a different pair of hydrophone array elements, and a different combination of seismograph components are determined as an observation condition. Based on the underwater acoustic observation azimuth, the ground acoustic observation azimuth, and the candidate absolute azimuth correction amount under the same observation condition, the azimuth correction angle corresponding to the candidate absolute azimuth correction amount under each observation condition is generated. Based on the candidate absolute azimuth correction, the azimuth correction angle under each observation condition is aligned with the circumferential angle, and the azimuth correction angle under each observation condition is updated. For the same candidate azimuth correction peak pair, the azimuth correction angles under all observation conditions after the update are collected by circular angle aggregation to form an azimuth correction angle set, so that the azimuth correction angle set represents the degree of directional concentration of the candidate azimuth correction peak pair under multiple observation conditions.

[0029] Terminology Explanation: The peak-to-directional relationship refers to the correspondence between the candidate peaks of underwater acoustic time delay, candidate peaks of ground acoustic directional spectrum, and candidate absolute directional correction in the same candidate directional correction peak pair. The underwater acoustic observation signal refers to the signal formed by the received signals of a different pair of hydrophone array elements within a different sub-frequency band, and used to calculate the azimuth of the underwater acoustic observation. The observation conditions refer to the azimuth correction angle generation conditions determined by a combination of a different sub-frequency band, a different pair of hydrophone array elements, and a different combination of seismograph components. The circumferential angle alignment refers to the process of converting the azimuth correction angle under each observation condition to the same continuous circumferential angle interval, using the candidate absolute azimuth correction amount as the alignment reference. The circular angle aggregation refers to the process of grouping all updated azimuth correction angles under all observation conditions corresponding to the same candidate azimuth correction peak pair into the same azimuth correction angle set. The underwater acoustic observation azimuth refers to the underwater acoustic direction estimate obtained based on the underwater acoustic observation signal and the array geometry and azimuth reference under a different sub-frequency band and a different pair of hydrophone array elements; The geosonic observation azimuth refers to the geosonic direction estimate obtained based on the submarine seismograph signal and the array geometry and azimuth reference under a different combination of seismograph components. The azimuth correction angle under each observation condition refers to the angle formed by aligning the azimuth of the ground acoustic observation relative to the azimuth of the underwater acoustic observation under the same observation condition with the candidate absolute azimuth correction amount as the reference. The set of azimuth correction angles refers to the set of azimuth correction angles under all observation conditions formed by the same candidate azimuth correction peak pair and aligned by circumferential angles. The degree of directional concentration refers to the degree to which the azimuth correction angles under each observation condition in the set of azimuth correction angles are clustered around the candidate absolute azimuth correction in the circumferential direction.

[0030] Optionally, S4 specifically refers to: Based on the candidate azimuth correction peak pairs, read the azimuth correction angle set corresponding to the candidate azimuth correction peak pairs, and convert the azimuth correction angles under each observation condition in the azimuth correction angle set into circular angle data for Rayleigh test. Perform a Rayleigh test on the circumferential angle data to calculate the composite length of the azimuth correction angle set in the circumferential direction and the significance of the test. Based on the directional synthesis length and the test significance quantity, the directional concentration of the candidate azimuth correction peak pair is determined, and the directional concentration is associated with the candidate azimuth correction peak pair that generates the azimuth correction angle set. The candidate underwater acoustic delay candidate peak, ground acoustic azimuth spectrum candidate peak, array geometry and azimuth reference, and candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair are input into the Evidential Deep Learning model to obtain the initial homology evidence value and non-adoption evidence value of the candidate azimuth correction peak pair. In the process of generating the Dirichlet distribution parameters corresponding to the homology evidence in the Evidential Deep Learning model, a constraint relationship is established on the initial homology evidence value based on the directional concentration, and the initial homology evidence value is updated according to the constraint relationship to form the homology evidence value; Based on the homology evidence value and the non-adoption evidence value, a Dirichlet distribution for homology determination of candidate orientation correction peak pairs is generated, such that the Dirichlet distribution for homology determination corresponds to the candidate orientation correction peak pairs.

[0031] Terminology Explanation: The circular angle data refers to the Rayleigh test input data formed by representing the azimuth correction angles under each observation condition in the azimuth correction angle set according to the angular periodicity. The direction synthesis length refers to the length obtained by synthesizing the circumferential directions corresponding to the circumferential angle data, and is used to characterize the degree of concentration of the azimuth correction angle set in the circumferential direction. The significance measure mentioned above refers to the statistic obtained by the Rayleigh test based on the composite length of the directions and the number of circumferential angle data, which is used to characterize whether the set of azimuth correction angles shows a concentration of circumferential directions; The Rayleigh test for circumferential angle data refers to testing whether the set of azimuth correction angles is concentrated in the circumferential direction, and outputting the composite length of the direction and the significance of the test. The directional concentration refers to a value that is not less than 0 and not greater than 1, determined based on the directional synthesis length and the test significance quantity, and is used to limit the formation of original evidence values ​​from initial homologous evidence values. The Evidential Deep Learning model refers to an evidence learning model that uses candidate orientation correction peak pairs as evidence generation objects and outputs initial homologous evidence values ​​and non-adopted evidence values. The initial homologous evidence value refers to the homologous evidence generated by the Evidential Deep Learning model based on the candidate azimuth correction peaks, corresponding underwater acoustic time delay candidate peaks, ground acoustic azimuth spectrum candidate peaks, array geometry and azimuth reference, and candidate absolute azimuth correction values, and which has not yet been subject to directional concentration constraints. The evidence value that is not adopted refers to the evidence generated by the Evidential Deep Learning model for candidate orientation correction peak pairs, indicating that orientation correction is not performed using the candidate orientation correction peak pairs; The constraint relationship refers to the relationship in which the directional concentration limits the initial homologous evidence value to form homologous evidence value during the generation process of the Dirichlet distribution parameters corresponding to homologous evidence in the Evidential Deep Learning model, so that the homologous evidence value does not exceed the initial homologous evidence value. The homology evidence value refers to the initial homology evidence value formed after being constrained by the directional concentration, and is used to generate evidence using the Dirichlet distribution for homology determination; The process of generating the Dirichlet distribution parameters corresponding to the homologous evidence refers to the internal calculation process of the Evidential DeepLearning model to convert the homologous evidence values ​​and the non-adopted evidence values ​​into the Dirichlet distribution parameters for homologation determination of candidate orientation correction peak pairs. The Dirichlet distribution used for homology determination refers to the Dirichlet distribution generated from the homology evidence values ​​and the non-adoption evidence values, and used to determine the homology of candidate orientation correction peak pairs.

[0032] Optional, S5 specifically includes: Receive the Dirichlet distribution for homology determination and the evidence determination threshold corresponding to each candidate orientation correction peak pair, and determine the homology evidence value formed after directional concentration restriction and the corresponding non-adoption evidence value from the Dirichlet distribution for homology determination, thus forming the evidence value relationship of the candidate orientation correction peak pair. Based on the aforementioned evidence value relationship, calculate the proportion of source evidence values ​​in the Dirichlet distribution for source determination, calculate the proportion of non-adopted evidence values ​​in the Dirichlet distribution for source determination, and calculate the evidence uncertainty based on the Dirichlet distribution for source determination. Based on the aforementioned evidence judgment threshold, the same-source evidence threshold, the non-adoption evidence threshold, and the uncertainty threshold are determined to form the evidence judgment conditions for candidate orientation correction peak pairs. The proportion of evidence from the same source is compared with the threshold for evidence from the same source, the proportion of evidence not accepted is compared with the threshold for evidence not accepted, and the uncertainty of the evidence is compared with the threshold for uncertainty, thus forming the evidence comparison result of the candidate orientation correction peak pair; Based on the evidence comparison results, if the proportion of evidence from the same source is greater than or equal to the threshold for evidence from the same source, the proportion of evidence not adopted is less than the threshold for evidence not adopted, and the amount of uncertainty in the evidence is less than or equal to the threshold for uncertainty, the candidate orientation correction peak pair will be determined as an orientation correction peak pair from the same source. Based on the evidence comparison results, if the proportion of evidence from the same source is less than the threshold for evidence from the same source, the proportion of evidence not adopted is greater than or equal to the threshold for evidence not adopted, or the uncertainty of the evidence is greater than the threshold for uncertainty, the candidate orientation correction peak pair will be determined as the orientation correction peak pair not adopted. The homology determination is performed on all candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or azimuth correction peak pairs that are not adopted. Homologous azimuth correction peak pairs are used as input to determine the absolute azimuth correction amount, and azimuth correction peak pairs that are not adopted are used as input to determine whether all candidate azimuth correction peak pairs are azimuth correction peak pairs that are not adopted.

[0033] Terminology Explanation: The evidence determination threshold refers to the set of thresholds used to compare and determine the homology determination using the Dirichlet distribution, including the homology evidence threshold, the evidence not accepted threshold, and the uncertainty threshold; The evidence value relationship refers to the correspondence between the homology evidence value, the non-adoption evidence value, and the candidate orientation correction peak pair in the Dirichlet distribution used for homology determination; The proportion of common-source evidence refers to the percentage of the value of common-source evidence in the total amount of evidence formed by the value of common-source evidence and the value of evidence not accepted. The percentage of evidence not accepted refers to the proportion of the value of evidence not accepted in the total amount of evidence formed by the value of evidence from the same source and the value of evidence not accepted. The evidence determination criteria refer to the comparison criteria jointly determined by the same source evidence threshold, the non-adoption evidence threshold, and the uncertainty threshold. The evidence comparison result refers to the result formed by comparing the proportion of evidence from the same source, the proportion of evidence not accepted, and the amount of evidence uncertainty with the evidence judgment conditions respectively. The uncertainty of evidence refers to the quantity determined by the total amount of evidence formed by the common evidence value and the non-adoption evidence value in the Dirichlet distribution used for the homology determination, and is used to indicate the degree of insufficiency of the homology determination evidence for the candidate orientation correction peak pair. The homologous azimuth correction peak pair refers to the candidate azimuth correction peak pair that meets the evidence judgment conditions and is used to determine the absolute azimuth correction amount. The term "not accepting azimuth correction peak pairs" refers to candidate azimuth correction peak pairs that do not meet the aforementioned evidence judgment conditions and are not used to determine the absolute azimuth correction amount.

[0034] Optional, S6 specifically includes: Based on the aforementioned homologous azimuth correction peak pair, read the candidate absolute azimuth correction amount and the Dirichlet distribution for homology determination corresponding to the homologous azimuth correction peak pair, and determine the homology evidence value of the homologous azimuth correction peak pair based on the Dirichlet distribution for homology determination. Based on the aforementioned evidence values ​​of the same origin, the absolute azimuth correction amount is determined from the candidate absolute azimuth correction amounts corresponding to the same azimuth correction peak pairs. Based on the absolute azimuth correction amount, the underwater acoustic positioning azimuth is angularly corrected to form the corrected underwater acoustic positioning azimuth, and the shallow sea target positioning result is output based on the corrected underwater acoustic positioning azimuth, array geometry and azimuth reference. If all candidate azimuth correction peak pairs are determined to be unacceptable, the underwater acoustic positioning azimuth is not corrected using absolute azimuth correction, and the shallow sea target positioning result without ground acoustic azimuth correction is output based on the underwater acoustic positioning azimuth, array geometry and azimuth reference.

[0035] Terminology Explanation: The absolute azimuth correction amount refers to the azimuth angle correction amount determined from the candidate absolute azimuth correction amounts corresponding to the same source azimuth correction peak pairs and used to correct the underwater acoustic positioning azimuth. The corrected underwater acoustic positioning azimuth refers to the azimuth obtained by angularly correcting the underwater acoustic positioning azimuth using the absolute azimuth correction amount. The shallow sea target positioning result refers to the target position result obtained by combining the corrected underwater acoustic positioning azimuth, array geometry and azimuth reference. The shallow sea target positioning result that does not adopt ground acoustic azimuth correction refers to the target position result obtained based on the underwater acoustic positioning azimuth, array geometry and azimuth reference without absolute azimuth correction when all candidate azimuth correction peak pairs are determined to be azimuth correction peak pairs that do not adopt azimuth correction.

[0036] Optionally, when generating the azimuth correction angles under each observation condition corresponding to the candidate absolute azimuth correction amount, under the same observation condition, the angle difference between the ground acoustic observation azimuth and the underwater acoustic observation azimuth is calculated. Using the candidate absolute azimuth correction amount as the circumferential angle alignment reference, the angle difference is converted to the same continuous circumferential angle interval as the candidate absolute azimuth correction amount to obtain the azimuth correction angles under the observation condition. The azimuth correction angles obtained under all observation conditions for the same candidate azimuth correction peak pair are aggregated to form a set of azimuth correction angles corresponding to the candidate azimuth correction peak pair.

[0037] Optionally, when determining the directional concentration of candidate orientation correction peak pairs, the directional synthesis length is compared with a preset synthesis length condition, and the test significance quantity is compared with a preset significance condition. Based on the comparison result, a directional concentration less than or equal to 1 is determined. During the generation process of the Dirichlet distribution parameters corresponding to the homology evidence in the Evidential Deep Learning model, the directional concentration is multiplied by the initial homology evidence value to form the homology evidence value. The non-adoption evidence value without directional concentration amplification and the homology evidence value are used together to generate the Dirichlet distribution for homology determination of candidate orientation correction peak pairs.

[0038] Optionally, when performing the homology determination on all candidate azimuth correction peak pairs, for multiple candidate azimuth correction peak pairs that contain the same underwater acoustic time delay candidate peak or the same ground acoustic azimuth spectrum candidate peak, the proportion of homology evidence and the amount of evidence uncertainty corresponding to the multiple candidate azimuth correction peak pairs are compared. For candidate azimuth correction peak pairs that jointly satisfy the following conditions: the proportion of homology evidence is greater than or equal to the homology evidence threshold, the proportion of non-adopted evidence is less than the non-adopted evidence threshold, and the amount of evidence uncertainty is less than or equal to the uncertainty threshold, the candidate azimuth correction peak pair with the smallest amount of evidence uncertainty is selected as the homology azimuth correction peak pair. In the case of the same amount of evidence uncertainty, the candidate azimuth correction peak pair with the largest proportion of homology evidence is selected as the homology azimuth correction peak pair. The candidate azimuth correction peak pairs that are not selected are determined as non-adopted azimuth correction peak pairs.

[0039] This invention also proposes a deep learning-based system for correcting the orientation of the same source peaks of shallow sea acoustics and seabed acoustics, the system comprising: Acquisition module: Acquires hydrophone array signals, seabed seismograph signals, and array geometry and azimuth references of shallow sea moorings; obtains underwater acoustic time delay candidate peaks based on hydrophone array signals; and obtains ground acoustic azimuth spectrum candidate peaks based on seabed seismograph signals. Determination module: Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair by combining an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. Generation module: For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction quantities under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. The verification module performs a Rayleigh test on the azimuth correction angle set for circumferential angle data to obtain the directional concentration of candidate azimuth correction peak pairs. The directional concentration is then embedded into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. Finally, the homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. Decision module: Based on the Dirichlet distribution and evidence decision threshold used for homology determination, the candidate azimuth correction peak pairs are determined to be homology, and homology azimuth correction peak pairs are obtained or azimuth correction peak pairs are not adopted. Output module: Determines the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, corrects the underwater acoustic positioning azimuth using the absolute azimuth correction amount, and outputs the shallow sea target positioning result by combining array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be unacceptable, the shallow sea target positioning result without ground acoustic azimuth correction is output.

[0040] Example refer to Figures 1 to 12 This invention proposes a method for correcting the azimuth of the same source peaks of shallow sea acoustics and seabed acoustics based on deep learning, comprising: S1. Obtain the hydrophone array signal of the shallow sea mooring, the seabed seismograph signal, and the array geometry and orientation reference. Obtain the underwater acoustic time delay candidate peak based on the hydrophone array signal, and obtain the ground acoustic orientation spectrum candidate peak based on the seabed seismograph signal. S2. Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair with an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. S3. For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction amount under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. S4. Perform Rayleigh test for circumferential angle data on the azimuth correction angle set to obtain the directional concentration of candidate azimuth correction peak pairs. Embed the directional concentration into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. The homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. S5. Based on the Dirichlet distribution and evidence threshold used for homology determination, homology determination is performed on candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or to reject azimuth correction peak pairs. S6. Determine the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, use the absolute azimuth correction amount to correct the underwater acoustic positioning azimuth, and output the shallow sea target positioning result in combination with array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be azimuth correction peak pairs not adopted, output the shallow sea target positioning result without adopting ground acoustic azimuth correction.

[0041] In this embodiment, step S1 specifically includes: Shallow-water moorings collect signals from a hydrophone array and a seabed seismograph during the same observation period. Assume the hydrophone array contains... One hydrophone array element, This indicates the number of hydrophone array elements; the hydrophone array signal is represented as... ,in Indicates the sampling sequence number. This indicates the number of sampling points for the hydrophone array signal. Indicates the element number of the hydrophone array. This indicates the sampling time of the hydrophone array signal. Indicates the first Each hydrophone element during sampling time The received value, assuming the seabed seismograph signal contains One portion, This indicates the number of components of the seabed seismometer; the seabed seismometer signal is represented as... ,in This indicates the number of sampling points for the seabed seismograph signal. Indicates the component number of the seabed seismograph. This indicates the sampling time of the seabed seismograph signal. Indicates the first Each component at sampling time The received values ​​are based on the start time of the same observation period. and target sampling interval Constructing a unified timeline ,in , This represents the number of sampling points on a unified time axis, mapping the hydrophone array signal and the seafloor seismograph signal to a unified time axis to form a discrete sequence. and When the sampling time does not fall on the same time axis, the corresponding discrete value is obtained by linear interpolation of adjacent sampling times. When the boundary of the same time axis exceeds the original sampling time range, the boundary sampling value is used to maintain the corresponding discrete value. The array geometry and azimuth reference are obtained in correspondence with the hydrophone array signals and the seabed seismograph signals. The array geometry and azimuth reference includes the spatial position of each hydrophone element in the hydrophone array, the element spacing, the array aperture, the array installation direction, and the azimuth reference of each component of the seabed seismograph relative to a unified horizontal coordinate system. The spatial position of each hydrophone array element is represented as follows: ,in , , They represent the first The three-dimensional coordinates of each hydrophone element in a unified coordinate system, and the set of different hydrophone element pairs is represented as follows: ,in Indicates by the first The hydrophone array element and the first Hydrophone array element pairs, consisting of individual hydrophone array elements, are represented as sets of different seismograph component combinations. , Each element in the equation is a combination of at least two seafloor seismograph components with azimuth references. The orientation reference of each component is represented as follows: , Indicates the first The installation orientation of each component in a unified horizontal coordinate system; Based on the array geometry and orientation reference, cross-correlation calculations are performed on the received signals of different hydrophone element pairs in the hydrophone array signal. For each hydrophone element pair... ,use and As input, according to the The hydrophone array element and the first The element spacing between each hydrophone array element and the preset underwater acoustic propagation speed range determine the preset time delay search range. ,in This represents the set of discrete time delay values ​​that can be searched. The preset range of underwater acoustic propagation speed is given by the sound speed profile of the shallow sea observation area or the configuration parameters of the shallow sea moorings. For each discrete time delay... , This represents the sample offset on a unified time axis, and the first... The discrete sequence of the first hydrophone array element and the first Each hydrophone array element The cross-correlation curve is obtained by normalizing the offset discrete sequences. ,in Indicates the hydrophone array element pair In discrete time delay The cross-correlation value under the following conditions; Cross-correlation curves Local peak detection is performed to generate underwater acoustic time delay peak information. Local peak detection uses discrete time delay as an index. First, the peak position where the cross-correlation value is greater than the cross-correlation value of adjacent discrete time delays is determined. Then, the time delay peak position and peak amplitude corresponding to the peak position are read. The peak width is determined by the discrete time delay span between the peak amplitude and a preset proportion position. The peak-sidelobe ratio is determined by the ratio of the peak amplitude to the maximum sidelobe amplitude in a preset neighborhood. The underwater acoustic time delay peak information includes the time delay peak position, peak amplitude, peak width, and peak-sidelobe ratio, and retains the hydrophone array element pairs that generated the underwater acoustic time delay peak information. Each delay peak in the underwater acoustic delay peak information is represented as ,in Indicates the sequence number of the underwater acoustic delay peak. This includes the location of the delay peak, the peak amplitude, the peak width, the peak sidelobe ratio, and the identifiers of the hydrophone array elements; Candidate underwater acoustic time delay peaks are determined based on underwater acoustic time delay peak information and preset peak conditions. The preset peak conditions include peak amplitude, peak width, and peak-sidelobe ratio conditions. The peak amplitude condition requires the peak amplitude to be no less than a preset underwater acoustic peak amplitude threshold. The peak width condition requires the peak width to fall within a preset underwater acoustic peak width range. The peak-sidelobe ratio condition requires the peak-sidelobe ratio to be no less than a preset underwater acoustic peak-sidelobe ratio threshold. For each time delay peak in the underwater acoustic time delay peak information... The peak amplitude, peak width, and peak sidelobe ratio are compared with the peak amplitude condition, peak width condition, and peak sidelobe ratio condition, respectively. When the peak amplitude condition, peak width condition, and peak sidelobe ratio condition are all satisfied, the corresponding time delay peak is determined as the underwater acoustic time delay candidate peak. The underwater acoustic time delay candidate peak retains the time delay peak position, peak amplitude, peak width, peak sidelobe ratio, and hydrophone array element pair identifier, so that the underwater acoustic time delay candidate peak can be converted into underwater acoustic positioning orientation based on array geometry and orientation reference. Based on the array geometry and azimuth reference, the azimuth spectrum of the received signals from different combinations of seismograph components in the seafloor seismograph signal is calculated. For each combination of seismograph components... , This represents a combination of seafloor seismograph components involved in azimuth spectrum calculation, constructing an azimuth search set. ,in Indicates the search sequence number. This indicates the number of location searches. This represents the candidate azimuth angles in a unified horizontal coordinate system. For each candidate azimuth angle... Combining seismograph components Discrete sequence of each component According to the direction reference Projected onto candidate azimuth angle The projected component sequences are then accumulated along a unified time axis to obtain candidate azimuth angles. The projection sequence under the following conditions, for the projection sequence in Energy is accumulated within the range to obtain candidate azimuth angles. The corresponding azimuth spectral values, according to The azimuth values ​​of the candidate azimuths are arranged in order to form a ground acoustic azimuth spectrum; Local peak detection is performed on the ground acoustic azimuth spectrum to form ground acoustic azimuth peak information. Local peak detection uses candidate azimuth angles as indices. First, the azimuth of the peak whose azimuth value is greater than that of the adjacent candidate azimuth angle is determined. Then, the amplitude of the peak corresponding to the azimuth is read. The azimuth width is determined by the azimuth span between the peak amplitude and a preset proportion. The sidelobe ratio is determined by the ratio of the peak amplitude to the maximum sidelobe amplitude in the preset azimuth neighborhood. The ground acoustic azimuth peak information includes peak azimuth, peak amplitude, peak width, and sidelobe ratio, and retains the seismograph component combination that generated the ground acoustic azimuth peak information. Each spectral peak in the ground acoustic azimuth spectrum information is represented as ,in Indicates the spectral peak number of the ground sound direction. This includes spectral peak orientation, spectral peak amplitude, spectral peak width, spectral peak sidelobe ratio, and seismograph component combination identifiers; Candidate peaks in the ground acoustic azimuth spectrum are determined based on the ground acoustic azimuth spectral peak information and preset peak conditions. The preset peak conditions include peak amplitude, peak width, and sidelobe ratio. The peak amplitude condition requires the peak amplitude to be no less than a preset ground acoustic peak amplitude threshold. The peak width condition requires the peak width to fall within a preset ground acoustic peak width range. The sidelobe ratio condition requires the sidelobe ratio to be no less than a preset ground acoustic peak sidelobe ratio threshold. For each peak in the ground acoustic azimuth spectral peak information... The peak amplitude, peak width, and peak sidelobe ratio are compared with the peak amplitude condition, peak width condition, and peak sidelobe ratio condition, respectively. When the peak amplitude condition, peak width condition, and peak sidelobe ratio condition are all satisfied, the corresponding peak is determined as a candidate peak of the ground acoustic azimuth spectrum. The candidate peak of the ground acoustic azimuth spectrum retains the peak azimuth, peak amplitude, peak width, peak sidelobe ratio, and seismograph component combination identifier, so that the candidate peak of the ground acoustic azimuth spectrum can form a candidate azimuth correction peak pair with the candidate peak of the underwater acoustic time delay.

[0042] In this embodiment, step S2 specifically includes: After determining the candidate peaks for underwater acoustic time delay and the candidate peaks for ground acoustic azimuth spectrum, the candidate peaks for underwater acoustic time delay are represented as follows: ,in This indicates the number of candidate peaks for underwater acoustic delay. Indicates the candidate peak number of the underwater acoustic time delay. Indicates the first The candidate peak for underwater acoustic time delay, the first Each candidate peak for underwater acoustic time delay includes the peak location, peak amplitude, peak width, peak sidelobe ratio, and hydrophone array element pair identifiers. The candidate peaks of the ground acoustic azimuth spectrum are represented as follows: ,in This indicates the number of candidate peaks in the ground acoustic azimuth spectrum. Indicates the candidate peak number of the ground acoustic azimuth spectrum. Indicates the first The candidate peaks of the ground acoustic azimuth spectrum, the first Each candidate peak in the ground acoustic azimuth spectrum includes the peak azimuth, peak amplitude, peak width, peak sidelobe ratio, and seismograph component combination identifier. The unified horizontal coordinate system is determined using the azimuth reference in the array geometry and azimuth datum. The true north direction of the unified horizontal coordinate system is taken as the zero-degree direction, and the clockwise direction is taken as the positive angular direction. The azimuth angle range is... ; Based on the array geometry and orientation reference, the candidate peaks of underwater acoustic time delay are converted into underwater acoustic positioning orientation. For the first... One candidate peak for underwater acoustic delay Read the position of the latency peak and hydrophone array element identification ,in Indicates the first The discrete time delay corresponding to each candidate peak of underwater acoustic time delay Indicates the generation of the first The hydrophone array element pairs with candidate peaks of underwater acoustic time delay are read based on the array geometry and orientation reference. The spatial position of each hydrophone array element, the first The spatial location of each hydrophone element, the array installation direction, and the array azimuth reference are used to construct an underwater acoustic azimuth search set. ,in This indicates the number of underwater acoustic azimuth searches. Indicates the search sequence number for underwater acoustic orientation. The candidate underwater acoustic azimuth angle is represented in a unified horizontal coordinate system. The interval between adjacent candidate underwater acoustic azimuth angles is determined by the preset underwater acoustic azimuth step size, which is determined by the azimuth resolution configuration of the shallow sea mooring. For each candidate underwater acoustic azimuth angle Based on the candidate underwater acoustic azimuth angle Corresponding horizontal unit direction, the first The spatial position of each hydrophone element and the first The spatial position of each hydrophone element determines the candidate underwater acoustic azimuth of the hydrophone element relative to the baseline. The projected length in the direction is calculated using the underwater acoustic propagation speed. Convert the projection length into prediction delay, where This represents the speed of sound propagation in water. Given the sound velocity profile or shallow sea mooring configuration parameters of the shallow sea observation area within the same observation period, the predicted time delay and the location of the time delay peak will be determined. By comparing the results, the candidate underwater acoustic azimuth angle with the smallest absolute time delay difference was determined as the first... The underwater acoustic location of the candidate peaks of the underwater acoustic time delay is denoted as follows: When multiple candidate underwater acoustic azimuth angles correspond to the same absolute time delay difference, the circumferential angle difference between the candidate underwater acoustic azimuth angles and the array installation direction is calculated based on the array installation direction. The candidate underwater acoustic azimuth angle with the smallest circumferential angle difference is determined as the underwater acoustic positioning azimuth. ; Based on the array geometry and azimuth reference, the peak position of the candidate peaks in the ground acoustic azimuth spectrum is determined as the azimuth of the ground acoustic spectrum peak. For the first... Candidate peaks of the ground acoustic azimuth spectrum Read the spectral peak azimuth and seismograph component combination identifiers. When the spectral peak azimuth of the candidate peaks of the geosound azimuth spectrum has been represented according to a unified horizontal coordinate system, the spectral peak azimuth is directly determined as the geosound spectral peak azimuth, denoted as . When the azimuth of the candidate peaks of the geosonic azimuth spectrum is represented according to the local coordinates of the seafloor seismograph, the azimuth of the spectral peaks is converted from the local coordinates of the seafloor seismograph to the unified horizontal coordinate system based on the azimuth reference of each component of the seafloor seismograph in the array geometry and azimuth reference. The converted azimuth angle is then determined as the azimuth of the geosonic spectral peaks. ; Based on the underwater acoustic positioning azimuth and the ground acoustic spectral peak azimuth, a pairing relationship is established between a candidate underwater acoustic time delay peak and a candidate ground acoustic azimuth spectral peak. and each , will the One candidate peak for underwater acoustic delay With the Candidate peaks of the ground acoustic azimuth spectrum Forming candidate azimuth correction peak pairs, denoted as Candidate azimuth correction peak Includes underwater acoustic delay candidate peaks Candidate peaks of ground acoustic azimuth spectrum underwater acoustic positioning and the orientation of the geosound spectrum peaks One candidate azimuth correction peak pair corresponds to only one underwater acoustic time delay candidate peak and one ground acoustic azimuth spectrum candidate peak, so that underwater acoustic reflection peaks, ground acoustic pseudo-spectral peaks and real homologous peaks all enter the same candidate azimuth correction peak pair generation rule. Based on the underwater acoustic positioning azimuth and the geosonic spectrum peak azimuth in the candidate azimuth correction peak pair, candidate absolute azimuth correction values ​​are formed. Calculate the azimuth of the ground acoustic spectrum peaks Relative to underwater acoustic positioning The angle difference is converted to a unified circumferential angle range. When the angle difference is greater than At that time, subtract the angle difference When the angle difference is less than or equal to At that time, add to the angle difference The angle difference after completing the circumferential angle conversion is used as the candidate azimuth correction peak pair. The corresponding candidate absolute azimuth correction is denoted as ; Candidate absolute azimuth correction Associated with candidate orientation correction peak pairs The correlated candidate orientation correction peak pairs Includes underwater acoustic delay candidate peaks Candidate peaks of ground acoustic azimuth spectrum underwater acoustic positioning , geoacoustic spectrum peak orientation and candidate absolute azimuth correction amount Candidate absolute azimuth correction amount Indicates candidate azimuth correction peak pair Given a single azimuth correction relation, when generating the set of azimuth correction angles, candidate azimuth correction peak pairs are... As the binding object, candidate absolute azimuth correction amount As a reference for circumferential angle alignment, the azimuth correction angles obtained under different observation conditions, different hydrophone element pairs, and different seismograph component combinations, all correspond to candidate azimuth correction peak pairs. .

[0043] In this embodiment, step S3 specifically includes: For candidate orientation correction peak pairs Read the candidate azimuth correction peak pairs underwater acoustic delay candidate peak Candidate peaks of ground acoustic azimuth spectrum underwater acoustic positioning , geoacoustic spectrum peak orientation and candidate absolute azimuth correction amount , Indicates the candidate peak number of the underwater acoustic time delay. Indicates the candidate peak number of the ground acoustic azimuth spectrum and the candidate peak of the underwater acoustic time delay spectrum. Including the position of the time delay peak, peak amplitude, peak width, peak sidelobe ratio, and hydrophone array element pair identifiers, as well as candidate peaks in the ground acoustic azimuth spectrum. Including spectral peak orientation, spectral peak amplitude, spectral peak width, spectral peak sidelobe ratio, and seismograph component combination identifiers, underwater acoustic time delay candidate peaks are identified. Candidate peaks of ground acoustic azimuth spectrum underwater acoustic positioning , geoacoustic spectrum peak orientation and candidate absolute azimuth correction amount Together, they form a peak-to-positional relationship, enabling the correction of peak pairs in the same candidate orientation. Always bind the same candidate absolute azimuth correction value under different observation conditions ; The hydrophone array signal was decomposed into different sub-bands, and the observation frequency band was divided into four different sub-bands, denoted as follows: , , and , Indicates the first Sub-band Using passband range and A consistent bandpass filter, for the first Discrete sequence of hydrophone array elements Perform filtering to obtain the first... Discrete sequence of hydrophone array elements in sub-band , Indicates the element number of the hydrophone array. Indicates the sampling sequence number on the unified time axis. This indicates the number of sampling points on the unified time axis. The passband range of the bandpass filter is consistent with the corresponding sub-band. The filtered discrete sequence retains the unified time axis index. Different hydrophone array element pairs are selected within different sub-frequency bands to participate in the localization of underwater acoustic time delay candidate peaks, targeting the underwater acoustic time delay candidate peaks. Read the identifiers of the hydrophone array elements and combine them with the underwater acoustic positioning location. And array geometry and orientation reference, in different sets of hydrophone element pairs Three sets of hydrophone array elements were selected and denoted as follows: , and , This represents the set of hydrophone element pairs, which consists of different hydrophone elements in a hydrophone array. This indicates the selection sequence number for the hydrophone array element pair. Indicates the first The selection rule for the hydrophone array element pairs is: priority is given to retaining candidate peaks that generate underwater acoustic delay. For the remaining hydrophone array pairs, based on the underwater acoustic positioning direction... The spatial position of the hydrophone array elements and the underwater acoustic propagation speed determine the corresponding prediction time delay, and within a preset underwater acoustic local time delay window. Internally search for cross-correlation local peaks and preset the local time delay window for underwater acoustics. From the prediction delay before and after Composed of discrete time delay points, This indicates the half-width of the preset local time delay window for underwater acoustics. The value is a positive integer and is given by the shallow sea mooring configuration parameters, which will satisfy the preset underwater acoustic local time delay window. The hydrophone array elements are sorted from high to low according to the peak-sidelobe ratio, and the hydrophone array elements with the highest ranking are selected until three sets of hydrophone array elements are obtained. For the Sub-band and the first The selected hydrophone array elements are paired ,Will The two hydrophone array elements are denoted as follows: and , composed of sub-band discrete sequences and Forming underwater acoustic observation signals underwater acoustic observation signals Including the same sampling number Next The sub-band received value of the first hydrophone element and the... The sub-band received value of the first hydrophone element is used to calculate the second... Sub-band and the first The selected hydrophone array elements are used to observe the underwater acoustic azimuth from below. Based on underwater acoustic observation signals Based on the array geometry and azimuth reference, calculate the underwater acoustic observation azimuth for the first... Sub-band and the first The selected hydrophone array pairs are within a preset local underwater acoustic time delay window. Inside, for the first The hydrophone array element and the first Cross-correlation peak search is performed on the sub-band received values ​​of each hydrophone array element, and the cross-correlation peak search is performed within a preset underwater acoustic local time delay window. Each discrete time delay within the range undergoes sequence offset, sample multiplication and accumulation, and energy normalization. The local peak position that satisfies the peak amplitude condition is selected as the sub-band observation time delay. When the preset underwater acoustic local time delay window If there are no local peaks that meet the peak amplitude condition, a preset underwater acoustic local time delay window will be used. The position with the largest internal cross-correlation value is used as the sub-band observation delay. , Indicates the first The candidate peak of underwater acoustic time delay is at the first Sub-band and the first The observation delay obtained by the selected hydrophone array elements is based on... The spatial positions of the two hydrophone array elements, the array mounting direction, the array azimuth reference, and the underwater acoustic propagation speed are used to construct an underwater acoustic positioning system. The search range for the local underwater acoustic azimuth is centered. Within this range, the predicted time delay for each candidate underwater acoustic azimuth angle is determined, and the predicted time delay is compared with the sub-band observation time delay. The candidate underwater acoustic azimuth with the smallest absolute difference is determined as the underwater acoustic observation azimuth. ; The signals from the seabed seismographs were combined into components according to different seismograph component combinations, and candidate peaks in the ground acoustic azimuth spectrum were selected. Read the component combination identifiers of the seismograph and combine them with the orientation of the ground acoustic spectrum peaks. And array geometry and azimuth reference, in different seismograph component combinations Two seismograph component combinations are selected and denoted as follows: and , This represents the set of different seismograph component combinations formed by the components of the seabed seismograph. This indicates the selection sequence number for the seismograph component combination. Indicates the first The selected seismograph component combinations are chosen according to the following rule: priority is given to retaining candidate peaks that generate the ground acoustic azimuth spectrum. For the remaining seismograph component combinations, within the preset local azimuth window of the ground acoustic system... Internal calculation of local peaks in the azimuth spectrum, preset local azimuth window for ground sound. By the orientation of the geosound spectrum peaks Centered on, the window half-width is , This represents the half-width of the preset local azimuth window for ground acoustic imaging, in degrees, and is given by the azimuth resolution configuration parameters of the seafloor seismograph. It will satisfy the preset local azimuth window for ground acoustic imaging. The seismograph component combinations are sorted from high to low according to the spectral peak sidelobe ratio. The seismograph component combination with the highest ranking is selected until two seismograph component combinations are obtained. For the Selected seismograph component combinations Combination of seismograph components Discrete sequences of each component Form a combination signal of geoacoustic components , Indicates the component number of the seabed seismograph and the combined signal of the ground acoustic components. Including the same sampling number Lower seismograph component combination The received values ​​of each component, for the first component Selected seismograph component combinations In the preset local azimuth window Within the local structural geosound azimuth search range, for each candidate geosound azimuth angle within the search range, the geosound component signals are combined based on the azimuth reference of each component of the seafloor seismograph relative to a unified horizontal coordinate system. Projecting onto candidate ground acoustic azimuth angles and performing time-axis energy accumulation on the projected component sequences yields the azimuth spectrum values ​​corresponding to the candidate ground acoustic azimuth angles. The candidate ground acoustic azimuth angle with the largest azimuth spectrum value is then converted to a unified horizontal coordinate system, and the converted azimuth angle is determined as the ground acoustic observation azimuth. ; A combination of a different sub-frequency band, a different pair of hydrophone elements, and a different seismograph component is defined as an observation condition for candidate azimuth correction peak pairs. Each observation condition is determined by , and jointly determined, , , Each observation condition corresponds to a specific underwater acoustic observation azimuth. A ground acoustic observation azimuth and the same candidate absolute azimuth correction quantity Under the same observation conditions, the azimuth correction angles corresponding to the candidate absolute azimuth corrections are generated for each observation condition as follows: ; in, Indicates candidate azimuth correction peak pair In the Sub-band, first The selected hydrophone array elements and the first The azimuth correction angle is generated by a selected combination of seismograph components, in degrees. Indicates the candidate peak number of the underwater acoustic time delay. Indicates the candidate peak number of the ground acoustic azimuth spectrum. Indicates the sub-band number. This indicates the selection sequence number for the hydrophone array element pair. This indicates the selection sequence number for the seismograph component combination. Indicates candidate azimuth correction peak pair The corresponding candidate absolute azimuth correction, in degrees. Indicates the first The candidate peaks of the ground acoustic azimuth spectrum are at the first The ground acoustic observation azimuth, in degrees, is obtained from a selected combination of seismograph components. Indicates the first The candidate peak of underwater acoustic time delay is at the first Sub-band and the first The selected hydrophone array elements were used to obtain the underwater acoustic observation azimuth, in degrees. Represents pi (π). This represents the half-circle angle constant in degrees. Represents the sine function. Represents the cosine function. This represents the two-parameter arctangent function, whose output range is... , Used to convert degrees to radians. Used to convert radians to degrees, the formula first determines the azimuth of the ground acoustic observation. Relative to underwater acoustic observation azimuth Angular difference minus candidate absolute azimuth correction amount Then, the difference is limited to a certain value using the sine function, cosine function, and two-parameter arctangent function. Within the corresponding circumferential angle range, the candidate absolute azimuth correction amount is finally added back. This allows the azimuth correction angle under various observation conditions to be converted to the equivalent of the candidate absolute azimuth correction. Within the same continuous interval of inscribed angles; Correction peak pairs for the same candidate orientation The azimuth correction angles under all observation conditions were updated and then collected into circular angles. The sequence numbers were selected according to the seismograph component combinations. For the fastest changing index, select the sequence number for the hydrophone array element pair. For intermediate change index, sub-band sequence number The azimuth correction angles under each observation condition are arranged in order of the slowest change index, forming a set of azimuth correction angles. Azimuth correction angle set for Dimension, the first item is The second item is The third item is And arranged sequentially up to the twenty-fourth item according to the same index order. Azimuth correction angle set The azimuth correction angle set consists of azimuth correction angles under various observation conditions, based on four different sub-frequency bands, three different hydrophone array pairs, and two different seismograph component combinations. Each azimuth correction angle in the table corresponds to a candidate azimuth correction peak pair. and candidate absolute azimuth correction amount Azimuth correction angle set Used to indicate candidate orientation correction peak pairs Around candidate absolute azimuth correction under various observation conditions The degree of directional concentration is used as input for the Rayleigh test on circumferential angle data.

[0044] In this embodiment, step S4 specifically includes: The candidate orientation correction peak pair is denoted as... ,in Indicates the candidate peak number of the underwater acoustic time delay. This indicates the candidate peak number of the ground acoustic azimuth spectrum. For candidate azimuth correction peak pairs... Read the candidate orientation correction peak pair The corresponding set of azimuth correction angles Azimuth correction angle set for Dimensions are selected according to the component combinations of the seismograph. For the fastest changing index, select the sequence number for the hydrophone array element pair. For intermediate change index, sub-band sequence number To arrange the indices of the slowest change, the first... Each azimuth correction angle is denoted as ,in , , , Read candidate azimuth correction peak pairs Corresponding candidate absolute azimuth correction amount Candidate absolute azimuth correction amount To determine the anchor angle, the set of azimuth correction angles is... The azimuth correction angles under each observation condition are converted into inscribed angle data for the Rayleigh test. The inscribed angle data are used relative to the candidate absolute azimuth correction. The radian angle representation fixes the input objects of the Rayleigh test to the same candidate azimuth correction peak pair. Around candidate absolute azimuth correction under various observation conditions Angular offset; Candidate orientation correction peak pair The corresponding candidate peaks for underwater acoustic time delay, candidate peaks for ground acoustic azimuth spectrum, array geometry and azimuth reference, and candidate absolute azimuth correction are organized as follows: Dimensional candidate peaks for input features Candidate peaks for input features It consists of four categories of values ​​arranged in a fixed order, the first category being... The underwater acoustic time delay candidate peak characteristics include time delay location, peak amplitude, peak width, peak-sidelobe ratio, cross-hydrophone element pair coherence, and underwater acoustic localization orientation. The cross-hydrophone element pair coherence is calculated using the average of the normalized cross-correlation peak values ​​of the three hydrophone element pairs participating in the underwater acoustic time delay candidate peak localization within a preset underwater acoustic local time delay window. The normalized cross-correlation peak values ​​are read from the cross-correlation curves of the corresponding hydrophone element pairs. The second type is... The characteristics of candidate peaks in the azimuth spectrum include azimuth angle, peak amplitude, peak width, peak sidelobe ratio, coherence of different seismograph component combinations, and local contrast of the azimuth spectrum. The coherence of different seismograph component combinations is the normalized peak response average of the two seismograph component combinations involved in the location of candidate peaks in the azimuth spectrum within a preset local azimuth window. The local contrast of the azimuth spectrum is the ratio of the peak amplitude to the average background spectral value within the preset local azimuth window. The average background spectral value is determined by the azimuth spectral values ​​within the preset local azimuth window that do not fall within the peak width coverage area. The third category is... The array geometry and orientation reference characteristics include hydrophone element spacing, array aperture, array orientation reference, and seabed seismograph orientation reference. Hydrophone element spacing is the spatial distance between pairs of hydrophone elements that generate candidate peaks for underwater acoustic delay. The array aperture is the maximum spatial distance between any two hydrophone elements in the array. The array orientation reference is the azimuth angle of the array installation direction relative to the zero-degree direction of a unified horizontal coordinate system. The seabed seismograph orientation reference is the azimuth angle of the zero-degree direction of the seabed seismograph's local coordinates used in the calculation of the ground acoustic azimuth spectrum relative to the zero-degree direction of a unified horizontal coordinate system. The fourth category is... Candidate absolute azimuth correction value, using ; Will Dimensional candidate peaks for input features Input Evidential Deep Learning model ,in This represents all the model parameters of the Evidential Deep Learning model. The Evidential Deep Learning model includes a candidate peak pair relation encoding structure and an evidence state generation structure. The candidate peak pair relation encoding structure uses... Neuron fully connected layer receives Dimensional candidate peaks for input features ,get The candidate peak is compared with the basic state, and then... Neuron fully connected layer receives Given the candidate peak relative to the base state, we obtain The state of the candidate peak pair relationship. Each neuron in the fully connected layer receives Dimensional candidate peaks for input features All values, including Each weight, One bias and correction linear unit, Each neuron in the fully connected layer receives The candidate peak represents all values ​​of the basic state, including Each weight, Each biased and corrected linear unit sets the negative input to zero and retains the non-negative input, so that the candidate peak pair relation coding structure obtains a non-negative activation state. The evidence state generation structure adopts Neuron fully connected layer receives The candidate peak pair relationship state is obtained. Maintain the state of evidence, and then adopt Neuron evidence output layer receiving The state of evidence is generated respectively. Initial homology evidence value and Wei does not accept evidence value , Each neuron in the fully connected layer receives All values ​​of the candidate peak pair relation state, including Each weight, One bias and correction linear unit, Each neuron in the neuronal evidence output layer receives All values ​​of the evidence state, including Each weight, Each bias and Softplus nonlinear unit sequentially performs natural exponentiation, increment by one, and natural logarithm on the real values ​​of the evidence output layer to obtain nonnegative evidence values ​​and initial homologous evidence values. Evidence from the same source that has not yet undergone directional concentration constraints will not be accepted. This indicates that candidate orientation correction peak pairs are not used. Evidence for orientation correction; The Rayleigh test for inscribed angle data is performed, and an orientation concentration constraint structure is set within the generation process of the Dirichlet distribution parameters corresponding to the common source evidence of the Evidential Deep Learning model. The orientation concentration constraint structure is based on the set of azimuth correction angles. Candidate absolute azimuth correction amount Initial homologous evidence value And the value of not accepting evidence For direct input, the directional synthesis length, significance level, directional concentration, homology evidence value, and Dirichlet distribution parameters for homology determination are calculated according to the following joint rules: ; ; ; ; ; ; ; ; ; in, Indicates the first Each azimuth correction angle relative to the candidate absolute azimuth correction amount The inscribed angle data, in radians. Indicates the candidate peak number of the underwater acoustic time delay. Indicates the candidate peak number of the ground acoustic azimuth spectrum. Indicates the serial number of the inscribed angle data. Indicates candidate azimuth correction peak pair In the Sub-band, first The selected hydrophone array elements and the first The azimuth correction angle is generated by a selected combination of seismograph components, in degrees. Indicates the first The sub-band number corresponding to each azimuth correction angle. Indicates the first The sequence number is selected for each hydrophone array element corresponding to each azimuth correction angle. Indicates the first The sequence number is selected for the seismograph component combination corresponding to each azimuth correction angle. Indicates candidate azimuth correction peak pair The corresponding candidate absolute azimuth correction, in degrees. Represents pi (π). This represents the half-circle angle constant in degrees. This indicates the number of inscribed angles, with a value of twenty-four. Indicates according to the inscribed angle data sequence number Accumulate. This represents the average component of the inscribed angle data in the cosine direction. This represents the average component of the inscribed angle data along the sine direction. Represents the cosine function. Represents the sine function. To represent the square root operation, Indicates multiplication operation. Indicates the direction of the composite length, with a value range of 1. , This represents the Rayleigh test statistic. This represents the significance level of the test, and its value range is [value range missing]. , This represents the natural exponential function. This represents the directional concentration, with a value range of [value missing]. , This represents the synthesis length threshold in the preset synthesis length conditions, with a value range of [value missing]. , This represents the significance threshold in the preset significance criteria, with a value range of [value range missing]. , This indicates an operation that takes the smaller value. Indicates the initial value of evidence of common origin. This represents the value of common-origin evidence after directional concentration constraints. This indicates that the evidence value will not be accepted. Indicates candidate azimuth correction peak pair The two-dimensional Dirichlet distribution parameter vector, Indicates the same category identifier, This indicates that category identifiers are not adopted. This represents the unit reference constant added when converting evidence values ​​to Dirichlet distribution parameters; the formula contains... Convert degrees to radians for use in trigonometric function calculations; in the formula... A dimensionless ratio, directional concentration Compared with the initial homology evidence value Dimensional matching during multiplication; The preset synthesis length condition is the directional synthesis length. Reaching the synthesis length threshold The presupposed significance condition is the significance level for testing. Not exceeding the significance threshold In the direction of the composite length The preset synthesis length condition is met and the significance level is tested. Under the premise of satisfying the preset significance condition, the directional concentration Based on the direction of the composite length and significance test The obtained value is limited to a range not exceeding one, in terms of directional composite length. The preset synthesis length condition is not met, or the significance level is not tested. If the preset significance condition is not met, the directional concentration will be... Setting it to zero, this joint rule reduces the directional concentration. Correction peaks originating from the same candidate orientation The twenty-four azimuth correction angles, and the direction concentration Enter the process of generating Dirichlet distribution parameters corresponding to homologous evidence; Based on directional concentration Establish initial homology evidence values The restriction relationship will affect the initial homologous evidence value. Updated to same-origin evidence value The value of evidence will not be accepted. Without directional concentration amplification, based on the value of homologous evidence And the value of not accepting evidence Generate a two-dimensional Dirichlet distribution parameter vector And by the two-dimensional Dirichlet distribution parameter vector Generate candidate orientation correction peak pairs Homology was determined using the Dirichlet distribution, with candidate orientation correction peaks. directional concentration Same source evidence value 1. Not accepting evidence value The homology determination is maintained using the Dirichlet distribution.

[0045] In this embodiment, step S5 specifically includes: The candidate orientation correction peak pair is denoted as... ,in Indicates the candidate peak number of the underwater acoustic time delay. This indicates the candidate peak number of the ground acoustic azimuth spectrum. For each candidate azimuth correction peak pair... Read the candidate orientation correction peak pair The corresponding homology determination uses a Dirichlet distribution, which is composed of a two-dimensional Dirichlet distribution parameter vector. express, ,in This represents the Dirichlet distribution parameters corresponding to the same categories. This indicates that the Dirichlet distribution parameters corresponding to the category are not adopted. Indicates the same category identifier, This indicates that the category identifier is not adopted. According to the Dirichlet distribution parameter generation process of homology evidence in the Evidential Deep Learning model, the Dirichlet distribution parameter is obtained by adding a unit baseline constant to the evidence value. Therefore, from... The value of the evidence of common origin is obtained by subtracting the unit reference constant. ,from The value of evidence not to be accepted is obtained by subtracting the unit reference constant. Homologous evidence value 1. Not accepting evidence value and candidate orientation correction peak pair The correspondence between them forms the evidence value relationship; Same source evidence value in evidence value relationship The directional concentration constraint has been applied, and the value of the evidence from the same source is... Derived from the product of the initial homologous evidence value and the directional concentration, the evidence value is not adopted. Without amplifying the directional concentration, the evidence value relationship is used to determine the source of the evidence, so that the initial source evidence generated by the local peak shape is subject to the circumferential directional concentration constraint of the azimuth correction angle set before entering the determination. Based on the relationship between the evidence values, candidate azimuth correction peak pairs are determined. Total amount of evidence The proportion of evidence from the same source Percentage of evidence not accepted and the uncertainty of evidence , Indicates the value of evidence from the same source Value of evidence not accepted The total amount of evidence formed together, and the number of two-dimensional judgment categories, are denoted as follows: and take Calculated as follows: ; ; ; ; in, Indicates candidate azimuth correction peak pair The total amount of evidence, This represents the value of common-origin evidence after directional concentration constraints. This indicates that the evidence value will not be accepted. Indicates the proportion of evidence from the same source. The percentage of evidence not accepted. Indicates the amount of uncertainty in the evidence. This indicates the number of two-dimensional judgment categories and the total amount of evidence. When it is zero, the proportion of homologous evidence will be... and the percentage of evidence not accepted All are set to zero, and determined by the uncertainty of the evidence. Indicates the degree of insufficiency of evidence; The evidence determination threshold is expressed as ,in, Indicates the threshold for evidence determination. Indicates the threshold for evidence of common origin. This indicates a threshold for not accepting evidence. This represents the threshold of uncertainty, and the threshold is determined based on evidence. The criteria for determining evidence include: the proportion of evidence from the same source. Greater than or equal to the threshold of common source evidence The percentage of evidence not accepted Less than the threshold for not accepting evidence Uncertainty of evidence Less than or equal to the uncertainty threshold Threshold for homologous evidence Threshold for not accepting evidence and uncertainty threshold All use closed intervals The values ​​within the range are given by the verification samples of the shallow-sea mooring or the mission configuration parameters; Candidate azimuth correction peak pair Comparison of enforcement evidence, including the proportion of evidence from the same source. Threshold for Homologous Evidence Comparison, the percentage of evidence not accepted Threshold for not accepting evidence Comparison, adjusting the uncertainty of evidence With uncertainty threshold Comparison to form candidate azimuth correction peak pairs The results of the evidence comparison show the proportion of evidence from the same source. Greater than or equal to the threshold of common source evidence Percentage of evidence not accepted Less than the threshold for not accepting evidence And the uncertainty of evidence Less than or equal to the uncertainty threshold In this case, the candidate azimuth correction peak will be paired Assuming a preliminary homologous orientation correction peak pair, in terms of the proportion of homologous evidence... Less than the threshold of homologous evidence Percentage of evidence not accepted Greater than or equal to the threshold for not accepting evidence or uncertainty of evidence Greater than the uncertainty threshold In this case, the candidate azimuth correction peak will be paired The orientation correction peak pair was deemed unacceptable. When performing homology determination on all candidate azimuth correction peak pairs, preliminary homology azimuth correction peak pairs and non-adopted azimuth correction peak pairs are first obtained according to the evidence determination conditions. Then, conflict resolution is performed on preliminary homology azimuth correction peak pairs that contain the same underwater acoustic time delay candidate peak or the same ground acoustic azimuth spectrum candidate peak. For the same underwater acoustic time delay candidate peak... Constructing a shared candidate set of underwater acoustic peaks for preliminary source orientation correction. ,in Indicates the first One candidate peak for underwater acoustic delay. This indicates that the same underwater acoustic time delay candidate peak is included. The preliminary set of homologous azimuth correction peak pairs, for the same candidate peak in the geosonic azimuth spectrum Constructing a candidate set of ground acoustic shared peaks for preliminary source orientation correction. ,in Indicates the first Candidate peaks of the ground acoustic azimuth spectrum. This indicates that the candidate peaks of the same ground acoustic azimuth spectrum are included. Preliminary set of homologous orientation correction peak pairs; When underwater acoustics share candidate set When multiple preliminary homologous orientation correction peak pairs are included, compare the evidentiary uncertainties corresponding to the multiple preliminary homologous orientation correction peak pairs. Preserving the uncertainty of evidence The smallest preliminary homologous orientation correction peak pair, with evidence uncertainty. Under the same circumstances, the proportion of evidence retained from the same source The largest preliminary homology correction peak pair, with varying degrees of uncertainty in the evidence. and the proportion of homologous evidence If all are the same, then according to the candidate peak number of underwater acoustic time delay. Candidate peak numbers of ground acoustic azimuth spectrum The lexicographically ordered preliminary homologous orientation correction peak pairs are retained, and the local acoustic shared candidate set is used. When there are multiple preliminary azimuth correction peak pairs, the same comparison order is used to retain one preliminary azimuth correction peak pair, and the unretained preliminary azimuth correction peak pairs are updated to not adopted azimuth correction peak pairs. After conflict resolution, the retained preliminary homologous azimuth correction peak pairs are determined to be homologous azimuth correction peak pairs. The homologous azimuth correction peak pairs retain candidate absolute azimuth correction values, and the homologousness determination uses a Dirichlet distribution and homologous evidence values. 1. Not accepting evidence value The proportion of evidence from the same source Percentage of evidence not accepted and the uncertainty of evidence If a azimuth correction peak pair is not adopted, the non-adoption status mark and the corresponding evidence comparison result are retained. If all candidate azimuth correction peak pairs are determined to be non-adopted, then the result is that all candidate azimuth correction peak pairs are non-adopted.

[0046] In this embodiment, step S6 specifically includes: Let the set of homologous orientation correction peak pairs be denoted as... It was determined to be a pair of azimuth correction peaks from the same source. , This indicates that all source azimuth correction peak pairs are retained for determining the absolute azimuth correction amount. Indicates by the first The candidate peak of underwater acoustic time delay and the first Candidate azimuth correction peak pairs are composed of candidate peaks in the ground acoustic azimuth spectrum. Indicates the candidate peak number of the underwater acoustic time delay. This indicates the candidate peak number in the ground acoustic azimuth spectrum, for the set of azimuth correction peak pairs from the same source. Each homologous orientation correction peak pair Read the same orientation correction peak pair Corresponding candidate absolute azimuth correction amount Homology determination uses Dirichlet distribution, candidate absolute orientation correction. Indicates the same-source orientation correction peak pair The given candidate azimuth correction values ​​are in degrees. Homology determination uses the Dirichlet distribution, derived from the parameter vector of the two-dimensional Dirichlet distribution. express, , Indicates the same-source orientation correction peak pair The two-dimensional Dirichlet distribution parameter vector, This represents the Dirichlet distribution parameters corresponding to the same categories. This indicates that the Dirichlet distribution parameters corresponding to the category are not adopted. Indicates the same category identifier, This indicates that the category identifier is not adopted. Based on the Dirichlet distribution parameter generation process of homology evidence in the Evidential Deep Learning model, the Dirichlet distribution parameters are obtained by adding a unit baseline constant to the evidence value. Subtracting unit reference constant The same orientation correction peak pair was obtained. Homogeneous evidence value ,from Subtracting unit reference constant The same orientation correction peak pair was obtained. The value of not accepting evidence Homologous evidence value The evidence values ​​are those obtained after directional concentration constraints, which are derived from the azimuth correction angle set using the Rayleigh test for circumferential angle data. These evidence values ​​are not accepted. Amplification of concentration without directional bias; When the same orientation correction peak pair set When there is only one pair of azimuth correction peaks from the same source, the candidate absolute azimuth correction value corresponding to the one pair of azimuth correction peaks is determined as the absolute azimuth correction value, denoted as . , This represents the azimuth correction amount used to correct the underwater acoustic positioning azimuth, in degrees. When the azimuth correction peaks of the same source are set... When multiple azimuth correction peak pairs are included, the values ​​should be based on the source evidence value. Multiple pairs of homologous orientation correction peaks are sorted from largest to smallest, and homologous evidence values ​​are selected. The largest homologous orientation correction peak pair, if multiple homologous orientation correction peak pairs have the same source evidence value If they are the same, then the uncertainty in the evidence related to the homology orientation correction peak pair has already been read when determining homology. Selecting the uncertainty of evidence The smallest homologous orientation correction peak pair, The Dirichlet distribution represents the degree of inadequacy of evidence for determining homology. This indicates the uncertainty in the evidence for multiple homology orientation correction peak pairs. If they are the same, then the proportion of homology evidence already associated with the homology orientation correction peak pair when determining homology is determined. The proportion of homologous evidence selected The largest homologous orientation correction peak pair This indicates the proportion of homologous evidence values ​​in the total amount of evidence. If multiple homologous orientation correction peak pairs represent homologous evidence, then... If they are the same, then they are ranked according to the candidate peak number of the underwater acoustic time delay. Candidate peak numbers of ground acoustic azimuth spectrum The numerical order is used to select the first source orientation correction peak pair, and the selected source orientation correction peak pair is denoted as... , This indicates the selected candidate peak number for underwater acoustic time delay. This indicates the selected candidate peak number of the ground acoustic azimuth spectrum. Corresponding candidate absolute azimuth correction amount Determined as absolute azimuth correction amount ; Read the candidate peak number of underwater acoustic delay Corresponding underwater acoustic positioning direction , Indicates according to the first The candidate peaks of underwater acoustic time delay and the estimated underwater acoustic direction determined by the array geometry and azimuth reference are given in degrees, along with the absolute azimuth correction. Add to underwater acoustic positioning The summed angles are then converted to a unified horizontal coordinate system. Within the azimuth range, if the summed angle is greater than or equal to Then, continuously subtract the summed angles. until the angle falls into If the summed angle is less than Then, continuously add to the summed angles. until the angle falls into The azimuth angle after angle conversion is recorded as the corrected underwater acoustic positioning azimuth. , Indicates the use of absolute azimuth correction amount underwater acoustic positioning The orientation obtained after angle correction; Based on the corrected underwater acoustic positioning direction The array geometry and orientation reference form the shallow sea target positioning result, and the array reference point is denoted as... , This indicates the reference position of the hydrophone array in a unified horizontal coordinate system. Indicates the horizontal x-coordinate of the array reference point. Representing the horizontal ordinate of the array reference point, the first... The underwater acoustic localization distance corresponding to each candidate peak of underwater acoustic time delay is denoted as . , Indicates the array reference point Distance to the target horizontal position, underwater acoustic positioning distance The process of obtaining the corrected underwater acoustic positioning azimuth is as follows: For the azimuth direction, the effective detection range of the shallow sea mooring is discretized into multiple candidate distances according to a preset distance step size. For each candidate distance, the range is determined from the array reference point. Along the corrected underwater acoustic positioning direction Candidate horizontal positions are generated. Based on the candidate horizontal positions, the spatial positions of each hydrophone element, and the underwater acoustic propagation speed, the predicted propagation delay difference for different hydrophone element pairs is obtained. The predicted propagation delay difference is then compared with the... The positions of the candidate underwater acoustic time delay peaks and the observation delays of different hydrophone array element pairs participating in the underwater acoustic time delay candidate peak localization are compared. The candidate distance with the smallest total time delay residual is determined as the underwater acoustic localization distance. The underwater acoustic propagation speed is given by the sound speed profile of the shallow sea observation area or the configuration parameters of the shallow sea moorings within the same observation period; In a unified horizontal coordinate system, zero degrees is true north, and the positive angle direction is clockwise. For shallow-sea mooring operation mode that outputs horizontal coordinates, the target horizontal coordinate is... From the horizontal coordinate of the array reference point Add underwater acoustic positioning distance The horizontal coordinate of the target is obtained by projection onto the horizontal axis of a unified horizontal coordinate system. The horizontal ordinate of the array reference point Add underwater acoustic positioning distance The projection onto the vertical axis of a unified horizontal coordinate system yields the corrected underwater acoustic positioning orientation during projection. First, convert from angle to radian. Use the sine component for the horizontal axis projection and the cosine component for the vertical axis projection. The shallow sea target location result is represented as follows: ,in Represents the horizontal coordinate of the target. Represents the horizontal ordinate of the target. This indicates the corrected underwater acoustic positioning azimuth. For the shallow sea mooring mode that only outputs the azimuth line, the shallow sea target positioning result is represented as follows: ,in and Given the starting point of the azimuth line, Provide the direction of the azimuth line; When the same orientation correction peak pair set When empty, it indicates that all candidate azimuth correction peak pairs are determined to be unacceptable. In the case that all candidate azimuth correction peak pairs are determined to be unacceptable, no absolute azimuth correction is generated. Without using ground acoustic azimuth correction for underwater acoustic positioning, the candidate peak number of the underwater acoustic time delay used to output the underwater acoustic positioning results is denoted as... , This indicates the sequence number of the underwater acoustic time delay candidate peaks selected without ground acoustic azimuth correction, determined by sorting the candidate peaks from highest to lowest according to their peak-sidelobe ratio. If multiple candidate underwater acoustic time delay peaks have the same peak-side lobe ratio, they are determined according to the peak amplitude from high to low. Read underwater acoustic positioning location And according to the distance from the underwater acoustic positioning The same candidate distance search process determines the underwater acoustic positioning distance. For shallow-sea mooring operation mode that outputs horizontal coordinates, the target horizontal coordinate is... From the horizontal coordinate of the array reference point Add underwater acoustic positioning distance underwater acoustic positioning The target's horizontal ordinate is obtained by projecting it onto the horizontal axis. The horizontal ordinate of the array reference point Add underwater acoustic positioning distance underwater acoustic positioning The shallow sea target localization result, obtained by projecting along the corresponding vertical axis without adopting ground acoustic azimuth correction, is expressed as follows: For the shallow-sea mooring operation mode that only outputs azimuth lines, the shallow-sea target positioning results without adopting ground acoustic azimuth correction are expressed as follows: .

[0047] This invention directly addresses the aforementioned technical problem through a closed-loop technical path involving candidate peak extraction, peak pair construction, consistency verification under multiple observation conditions, evidence learning and judgment, and correction or rejection of output. In target positioning scenarios where shallow-sea moorings simultaneously acquire hydrophone array signals and seafloor seismograph signals, shallow-sea multipath propagation, seafloor medium coupling, array installation azimuth deviation, and local strong noise can cause uncertainty in the homology between underwater acoustic time delay peaks and ground acoustic azimuth spectrum peaks. This can lead to non-homogeneous peak pairs being used for underwater acoustic positioning azimuth correction. This invention addresses this issue by relying solely on hydrophone array signals within the same observation period... Under the conditions of underwater acoustic cross-correlation curves and ground acoustic azimuth spectrum, candidate peaks are first extracted from underwater acoustic time delay candidate peaks and ground acoustic azimuth spectrum candidate peaks to form candidate azimuth correction peak pairs, and candidate absolute azimuth correction amounts are calculated. Then, around the same candidate peak pair, azimuth correction angle sets are generated under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations. The angle differences under various observation conditions are classified into the same continuous interval by circumferential angle alignment, so that the directional consistency of candidate peak pairs under changes in microscopic observation conditions becomes a verifiable object. Compared to methods that rely solely on peak intensity, angle difference thresholds, or ordinary classification models to select peak pairs, this invention combines the Rayleigh test and an Evidential Deep Learning model into the process of determining originating peak pairs. The Rayleigh test addresses the concentration of circumferential angle directions within the azimuth correction angle set, adapting to azimuth periodicity and avoiding the discontinuities caused by ordinary linear difference statistics near zero degrees. The Evidential Deep Learning model… The initial evidence of common origin output by the learning model is not directly used for final adoption. Instead, it is constrained by the directional concentration during the generation of Dirichlet distribution parameters and, together with the evidence of non-adoption, forms a Dirichlet distribution for determining common origin. By jointly determining the proportion of common origin evidence, the proportion of non-adoption evidence, and the uncertainty of evidence, this invention can reject the use of ground acoustic azimuth correction when the common origin relationship is insufficient or the directions of multiple observation conditions are inconsistent. Furthermore, it handles mutually exclusive peak pairs that share the same underwater acoustic candidate peak or the same ground acoustic candidate peak through a conflict resolution mechanism. Thus, this invention enables the key intermediate quantities in shallow sea mooring positioning, namely the common origin azimuth correction peak pairs and their corresponding absolute azimuth correction quantities, to be determined or rejected under limited observation data and complex shallow sea environments, thereby forming a closed loop for shallow sea target positioning processing that is acceptable, rejectable, and deliverable.

[0048] This invention proposes a method and system for azimuth correction of co-source peaks in shallow sea acoustics and seabed ground acoustics based on deep learning. The method acquires hydrophone array signals, seabed seismograph signals, and array geometry and azimuth references within the same observation period. It extracts candidate peaks for underwater acoustic time delay and candidate peaks for ground acoustic azimuth spectrum, forming candidate azimuth correction peak pairs and calculating candidate absolute azimuth correction values. Azimuth correction angle sets are generated under different sub-frequency bands, hydrophone array element pairs, and seismograph component combinations. The Rayleigh test is used to obtain directional concentration, which is then embedded into the Dirichlet distribution parameter generation process of the Evidential Deep Learning model to restrict co-source evidence. Co-source peak pairs are determined based on evidence proportion and uncertainty, and the absolute azimuth correction value is determined to correct the underwater acoustic positioning azimuth. The method described in this invention can be used for shallow sea target positioning, reducing the impact of misadoption of non-co-source peaks on the positioning azimuth.

[0049] The present invention also proposes an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the deep learning-based method for correcting the orientation of shallow sea acoustic and seabed acoustic sources.

[0050] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the deep learning-based method for correcting the orientation of shallow sea acoustic and seabed acoustic sources.

[0051] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0052] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0053] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0054] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as execution by a hardware decoding processor, or as a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0055] The foregoing has provided a detailed description of the method and system for correcting the azimuth of shallow sea acoustic and seabed acoustic sources based on deep learning, as proposed in this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for correcting the azimuth of common source peaks in shallow sea acoustics and seabed acoustics based on deep learning, characterized in that, The method includes: S1. Obtain the hydrophone array signal of the shallow sea mooring, the seabed seismograph signal, and the array geometry and orientation reference. Obtain the underwater acoustic time delay candidate peak based on the hydrophone array signal, and obtain the ground acoustic orientation spectrum candidate peak based on the seabed seismograph signal. S2. Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair with an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. S3. For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction amount under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. S4. Perform Rayleigh test for circumferential angle data on the azimuth correction angle set to obtain the directional concentration of candidate azimuth correction peak pairs. Embed the directional concentration into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. The homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. S5. Based on the Dirichlet distribution and evidence threshold used for homology determination, homology determination is performed on candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or to reject azimuth correction peak pairs. S6. Determine the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, use the absolute azimuth correction amount to correct the underwater acoustic positioning azimuth, and output the shallow sea target positioning result in combination with array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be azimuth correction peak pairs not adopted, output the shallow sea target positioning result without adopting ground acoustic azimuth correction.

2. The method according to claim 1, characterized in that, S1 specifically refers to: Acquire hydrophone array signals and seabed seismograph signals of shallow sea moorings within the same observation period, and obtain the array geometry and azimuth reference corresponding to the hydrophone array signals and the seabed seismograph signals; Based on the array geometry and orientation reference, cross-correlation calculations are performed on the received signals of different hydrophone array element pairs in the hydrophone array signal to form underwater acoustic time delay peak information; Based on the underwater acoustic delay peak information and the preset peak value conditions, candidate underwater acoustic delay peaks are determined; Based on the array geometry and azimuth reference, the azimuth spectrum of the received signals of different seismograph component combinations in the submarine seismograph signal is calculated to form ground acoustic azimuth spectrum peak information, and ground acoustic azimuth spectrum candidate peaks are determined according to the ground acoustic azimuth spectrum peak information and preset peak conditions.

3. The method according to claim 1, characterized in that, S2 specifically refers to: Based on the array geometry and orientation reference, the candidate peaks of underwater acoustic time delay are converted into underwater acoustic positioning orientations, and the peak positions of the candidate peaks of the ground acoustic orientation spectrum are determined as the orientations of the ground acoustic spectrum peaks. Based on the underwater acoustic positioning azimuth and the ground acoustic spectrum peak azimuth, a pairing relationship is established between an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak to form a candidate azimuth correction peak pair. Based on the underwater acoustic positioning azimuth and the geosonic spectrum peak azimuth in the candidate azimuth correction peak pair, the angular difference between the geosonic spectrum peak azimuth and the underwater acoustic positioning azimuth is calculated to form the candidate absolute azimuth correction amount. The candidate absolute azimuth correction values ​​are associated with the corresponding candidate azimuth correction peak pairs, so that the candidate azimuth correction peak pairs are used as inputs to generate the set of azimuth correction angles.

4. The method according to claim 1, characterized in that, S3 specifically refers to: Based on the candidate azimuth correction peak pairs, extract the underwater acoustic time delay candidate peaks and ground acoustic azimuth spectrum candidate peaks that make up the candidate azimuth correction peak pairs, and read the candidate absolute azimuth correction values ​​corresponding to the candidate azimuth correction peak pairs to form peak-to-azimuth relationships. Based on the peak orientation relationship, the hydrophone array signal is decomposed into different sub-frequency bands, and different hydrophone array element pairs participating in the underwater acoustic time delay candidate peak localization are selected in each different sub-frequency band to form an underwater acoustic observation signal. Based on the underwater acoustic observation signal and the array geometry and orientation reference, calculate the underwater acoustic observation orientation corresponding to the underwater acoustic time delay candidate peak for different sub-frequency bands and different hydrophone array element pairs; Based on the peak orientation relationship, the submarine seismograph signal is combined into components according to different seismograph component combinations, and the geosound observation orientation corresponding to the candidate peak of the geosound orientation spectrum is calculated under each different seismograph component combination. A different sub-frequency band, a different pair of hydrophone array elements, and a different combination of seismograph components are determined as an observation condition. Based on the underwater acoustic observation azimuth, the ground acoustic observation azimuth, and the candidate absolute azimuth correction amount under the same observation condition, the azimuth correction angle corresponding to the candidate absolute azimuth correction amount under each observation condition is generated. Based on the candidate absolute azimuth correction, the azimuth correction angle under each observation condition is aligned with the circumferential angle, and the azimuth correction angle under each observation condition is updated. For the same candidate azimuth correction peak pair, the azimuth correction angles under all observation conditions after the update are collected by circular angle aggregation to form an azimuth correction angle set, so that the azimuth correction angle set represents the degree of directional concentration of the candidate azimuth correction peak pair under multiple observation conditions.

5. The method according to claim 1, characterized in that, S4 specifically refers to: Based on the candidate azimuth correction peak pairs, read the azimuth correction angle set corresponding to the candidate azimuth correction peak pairs, and convert the azimuth correction angles under each observation condition in the azimuth correction angle set into circular angle data for Rayleigh test. Perform a Rayleigh test on the circumferential angle data to calculate the composite length of the azimuth correction angle set in the circumferential direction and the significance of the test. Based on the directional synthesis length and the test significance quantity, the directional concentration of the candidate azimuth correction peak pair is determined, and the directional concentration is associated with the candidate azimuth correction peak pair that generates the azimuth correction angle set. The candidate underwater acoustic delay candidate peak, ground acoustic azimuth spectrum candidate peak, array geometry and azimuth reference, and candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair are input into the Evidential Deep Learning model to obtain the initial homology evidence value and non-adoption evidence value of the candidate azimuth correction peak pair. In the process of generating the Dirichlet distribution parameters corresponding to the homology evidence in the Evidential Deep Learning model, a constraint relationship is established on the initial homology evidence value based on the directional concentration, and the initial homology evidence value is updated according to the constraint relationship to form the homology evidence value; Based on the homology evidence value and the non-adoption evidence value, a Dirichlet distribution for homology determination of candidate orientation correction peak pairs is generated, such that the Dirichlet distribution for homology determination corresponds to the candidate orientation correction peak pairs.

6. The method according to claim 1, characterized in that, S5 specifically refers to: Receive the Dirichlet distribution for homology determination and the evidence determination threshold corresponding to each candidate orientation correction peak pair, and determine the homology evidence value formed after directional concentration restriction and the corresponding non-adoption evidence value from the Dirichlet distribution for homology determination, thus forming the evidence value relationship of the candidate orientation correction peak pair. Based on the aforementioned evidence value relationship, calculate the proportion of source evidence values ​​in the Dirichlet distribution for source determination, calculate the proportion of non-adopted evidence values ​​in the Dirichlet distribution for source determination, and calculate the evidence uncertainty based on the Dirichlet distribution for source determination. Based on the aforementioned evidence judgment threshold, the same-source evidence threshold, the non-adoption evidence threshold, and the uncertainty threshold are determined to form the evidence judgment conditions for candidate orientation correction peak pairs. The proportion of evidence from the same source is compared with the threshold for evidence from the same source, the proportion of evidence not accepted is compared with the threshold for evidence not accepted, and the uncertainty of the evidence is compared with the threshold for uncertainty, thus forming the evidence comparison result of the candidate orientation correction peak pair; Based on the evidence comparison results, if the proportion of evidence from the same source is greater than or equal to the threshold for evidence from the same source, the proportion of evidence not adopted is less than the threshold for evidence not adopted, and the amount of uncertainty in the evidence is less than or equal to the threshold for uncertainty, the candidate orientation correction peak pair will be determined as an orientation correction peak pair from the same source. Based on the evidence comparison results, if the proportion of evidence from the same source is less than the threshold for evidence from the same source, the proportion of evidence not adopted is greater than or equal to the threshold for evidence not adopted, or the uncertainty of the evidence is greater than the threshold for uncertainty, the candidate orientation correction peak pair will be determined as the orientation correction peak pair not adopted. The homology determination is performed on all candidate azimuth correction peak pairs to obtain homologous azimuth correction peak pairs or azimuth correction peak pairs that are not adopted. Homologous azimuth correction peak pairs are used as input to determine the absolute azimuth correction amount, and azimuth correction peak pairs that are not adopted are used as input to determine whether all candidate azimuth correction peak pairs are azimuth correction peak pairs that are not adopted.

7. The method according to claim 1, characterized in that, S6 specifically refers to: Based on the aforementioned homologous azimuth correction peak pair, read the candidate absolute azimuth correction amount and the Dirichlet distribution for homology determination corresponding to the homologous azimuth correction peak pair, and determine the homology evidence value of the homologous azimuth correction peak pair based on the Dirichlet distribution for homology determination. Based on the aforementioned evidence values ​​of the same origin, the absolute azimuth correction amount is determined from the candidate absolute azimuth correction amounts corresponding to the same azimuth correction peak pairs. Based on the absolute azimuth correction amount, the underwater acoustic positioning azimuth is angularly corrected to form the corrected underwater acoustic positioning azimuth, and the shallow sea target positioning result is output based on the corrected underwater acoustic positioning azimuth, array geometry and azimuth reference. If all candidate azimuth correction peak pairs are determined to be unacceptable, the underwater acoustic positioning azimuth is not corrected using absolute azimuth correction, and the shallow sea target positioning result without ground acoustic azimuth correction is output based on the underwater acoustic positioning azimuth, array geometry and azimuth reference.

8. A deep learning-based system for correcting the orientation of shallow sea acoustic and seabed acoustic peaks, characterized in that, The system includes: Acquisition module: Acquires hydrophone array signals, seabed seismograph signals, and array geometry and azimuth references of shallow sea moorings; obtains underwater acoustic time delay candidate peaks based on hydrophone array signals; and obtains ground acoustic azimuth spectrum candidate peaks based on seabed seismograph signals. Determination module: Based on the underwater acoustic time delay candidate peak and the ground acoustic azimuth spectrum candidate peak, form a candidate azimuth correction peak pair by combining an underwater acoustic time delay candidate peak and a ground acoustic azimuth spectrum candidate peak, and determine the candidate absolute azimuth correction amount corresponding to the candidate azimuth correction peak pair according to the array geometry and azimuth reference. Generation module: For the candidate azimuth correction peak pair, under different sub-frequency bands, different hydrophone array element pairs and different seismograph component combinations, generate azimuth correction angles corresponding to the candidate absolute azimuth correction quantities under various observation conditions, forming an azimuth correction angle set. The azimuth correction angle set is used to represent the degree of directional concentration of the candidate azimuth correction peak pair under various observation conditions. The verification module performs a Rayleigh test on the azimuth correction angle set for circumferential angle data to obtain the directional concentration of candidate azimuth correction peak pairs. The directional concentration is then embedded into the Dirichlet distribution parameter generation process corresponding to the homology evidence of the Evidential Deep Learning model. This allows the initial homology evidence values ​​generated by the Evidential Deep Learning model to be constrained by the directional concentration to form homology evidence values. Finally, the homology evidence values ​​and the non-adopted evidence values ​​are used to generate the Dirichlet distribution for homology determination of candidate azimuth correction peak pairs. Decision module: Based on the Dirichlet distribution and evidence decision threshold used for homology determination, the candidate azimuth correction peak pairs are determined to be homology, and homology azimuth correction peak pairs are obtained or azimuth correction peak pairs are not adopted. Output module: Determines the absolute azimuth correction amount based on the candidate absolute azimuth correction amount corresponding to the same azimuth correction peak pair, corrects the underwater acoustic positioning azimuth using the absolute azimuth correction amount, and outputs the shallow sea target positioning result by combining array geometry and azimuth reference. When all candidate azimuth correction peak pairs are determined to be unacceptable, the shallow sea target positioning result without ground acoustic azimuth correction is output.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-7.

Images