Discrete profile driven three-dimensional sound speed field reconstruction and ocean structure identification method and system

Abstract

The present application relates to the field of underwater acoustic environment detection and marine tomography, in particular to a method and system for discrete profile driven three-dimensional sound velocity field reconstruction and marine structure identification; the method comprises: performing quality control on discrete station vertical profile and unifying depth layer to calculate sound velocity value, constructing three-dimensional regular grid and generating low-rank three-dimensional background sound velocity field by using empirical orthogonal function decomposition, adopting optimal interpolation to fuse discrete observation and background field to generate continuous three-dimensional analysis sound velocity field, synchronously diagnosing thermocline fluctuation surface, joint front candidate area, vortex anomaly boundary candidate area and internal wave disturbance candidate area, outputting structured information and confidence for each type of identified structure, and outputting standardized data interface after reserving profile intersection test partition statistical error; the present application realizes high-precision reconstruction of three-dimensional sound velocity field and multi-type marine structure collaborative identification under sparse discrete profile driving.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic environment detection and ocean tomography technology, specifically to a method and system for discrete profile-driven three-dimensional sound velocity field reconstruction and ocean structure identification. Background Technology

[0002] The three-dimensional sound velocity field related structures such as the ocean thermocline, fronts, vortices, and internal waves play a key role in regulating underwater sound propagation paths and energy distribution. Accurately reconstructing the sound velocity field and simultaneously identifying the above structures is an important direction of continuous attention in the fields of underwater acoustic environment protection and marine monitoring.

[0003] Currently, ocean observation mainly relies on platforms such as mobile CTD (conductivity, temperature, depth) profilers, discardable CTD profilers, Argo buoys, and moored underwater moorings to obtain discrete vertical profiles. This type of data has high vertical resolution and accurate measurements, but the station intervals are large, resulting in a sparse and discrete spatial distribution.

[0004] To fill observational gaps, existing methods have incorporated historical climatological mean fields or ocean model reanalysis data as supplementary information. These methods generate regularly gridded three-dimensional sound velocity fields through objective analysis or data assimilation, while simultaneously utilizing single variables such as temperature gradients or sound velocity gradients to diagnose features like thermoclines and fronts. This forms the basic technical framework for current ocean environmental field reconstruction and structure identification. However, within this framework, sound velocity field reconstruction and identification of various ocean structures typically fall under different processing flows. Structure identification often relies on single-variable, single-depth-layer analysis, and there is still room for improvement in utilizing multi-variable collaborative constraints and three-dimensional spatial correlation information. Furthermore, the connection between the output format and the acoustic model requires further optimization. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method and system.

[0006] The technical solution of this invention: a method for reconstructing a three-dimensional sound velocity field and identifying marine structures driven by discrete profiles, comprising the following implementation steps: S1. Perform quality control on the vertical profiles of multiple discrete stations and unify the depth layer. Calculate the sound velocity value based on temperature, salinity and depth to form a standardized set of discrete vertical sound velocity observations. S2. Construct a three-dimensional regular grid covering the target sea area, collect historical ocean fields or model background fields as a sample set, calculate the mean field and extract several principal modes by empirical orthogonal function decomposition, estimate the coefficients of each mode using discrete observation profiles, and generate a low-rank three-dimensional background sound velocity field by weighted superposition of the mean field and principal modes. S3. Using the background sound velocity field as the prior field and the discrete profile sound velocity observation as the observation, the optimal interpolation method is used to fuse the data and generate a continuous three-dimensional analytical sound velocity field. S4. Based on the three-dimensional analysis of the sound velocity field and the corresponding temperature field, salinity field and density field, the maximum position of the vertical temperature gradient is extracted to form the undulating surface of the thermocline. The horizontal gradients of sound velocity, temperature, salinity and density are calculated and normalized to construct a joint front index and mark the front candidate area. The sound velocity field is smoothed and the background is separated. The candidate area of ​​the vortex boundary is identified by combining the anomaly intensity and the edge intensity. The internal wave disturbance proxy index is constructed by combining the thermocline displacement residual and the short-scale sound velocity residual and the candidate area of ​​the internal wave disturbance is marked. S5. Output structured information for each type of identified environmental structure, and calculate the confidence level of each type of structure by comprehensively considering strength, continuity, multivariate consistency and morphological features. S6. Extract a portion from all profiles as a validation set, reconstruct the three-dimensional sound velocity field using the remaining profiles, interpolate the predicted values ​​at the validation profile locations and compare them with the actual observations to calculate the error, and statistically analyze the error by region. Finally, output the three-dimensional sound velocity field and structural marker layer and provide a data interface.

[0007] Preferably, in step S1, the specific process of performing quality control and unifying the depth layer on the acquired vertical profiles of multiple discrete stations, and calculating the sound velocity value based on temperature, salinity, and depth includes: Abnormal records with temperature or salinity exceeding the normal range for climatological conditions are removed; points that are not monotonically increasing in the depth sequence are removed; spike noise with vertical gradients exceeding a preset threshold is removed; and median filtering is applied to the profile. A uniform depth layer is set for the target sea area, and the temperature and salinity data of each profile are interpolated to the uniform depth layer using linear interpolation or cubic spline interpolation. At each uniform depth layer, the speed of sound is calculated using an empirical formula based on temperature, salinity, and depth.

[0008] Preferably, in step S2, the specific process of collecting historical ocean fields or model background fields as a sample set, calculating the mean field, and performing empirical orthogonal function decomposition to extract several principal modes includes: The mean field is obtained by averaging the sample set point by point, and the perturbation field is obtained by subtracting the mean field from each sample. The three-dimensional perturbation field is expanded into a one-dimensional vector according to the spatial grid to construct a sample matrix. The sample matrix is ​​then subjected to singular value decomposition to obtain the left singular vector matrix, the diagonal singular value matrix, and the right singular vector matrix. Each column of the left singular vector is taken as an empirical orthogonal function mode, and the number of main modes to be retained is determined based on the cumulative variance contribution rate.

[0009] Preferably, in step S3, the specific process of using the background sound velocity field as the prior field and employing the optimal interpolation method for data fusion includes: The background error covariance matrix is ​​set as a spatial correlation function, and the correlation scale of the spatial correlation function is determined based on the sea area scale, front direction, vortex radius, topographic constraints, or empirical orthogonal function mode energy. The observation error covariance matrix is ​​set as a diagonal matrix, and the diagonal elements of the diagonal matrix are determined based on the instrument type, profile quality control results, and vertical interpolation error. Gain weights are calculated using the background error covariance matrix and the observation error covariance matrix. The background field is then corrected using these gain weights to generate a continuous three-dimensional analytical sound velocity field.

[0010] Preferably, in step S4, the specific process of extracting the location of the maximum vertical temperature gradient that constitutes the undulating surface of the thermocline includes: At each horizontal grid point, the vertical gradient of temperature with depth is calculated based on the three-dimensional temperature field. The location with the strongest negative gradient within a preset depth range is found as the representative depth of the thermocline at that horizontal location. The thermocline undulating surface is formed by the representative depths of the thermoclines at all horizontal grid points.

[0011] Preferably, in step S4, the specific process of calculating the horizontal gradients of sound speed, temperature, salinity, and density, constructing a joint frontal index using normalized weights, and labeling frontal candidate regions includes: Calculate the horizontal gradients of sound velocity, temperature, salinity, and density proxy at a given depth layer. The gradients of each variable are normalized, and the joint front index is obtained by multiplying the normalized gradients of each variable by their corresponding weight coefficients and then summing them. Continuous regions where the joint front index exceeds an adaptive threshold are marked as frontal candidate regions; the weighting coefficients are either equally weighted or adaptively determined based on observation quality.

[0012] Preferably, in step S4, the specific process of smoothing the background of the sound velocity field and identifying vortex boundary candidate regions by combining anomaly intensity and edge intensity includes: Spatial smoothing of the sound velocity field at a given depth layer yields a smoothed background field, and subtraction of the smoothed background field from the original sound velocity field yields a sound velocity anomaly field. The absolute value of the sound velocity anomaly field is calculated as the anomaly intensity, and the horizontal gradient magnitude of the sound velocity anomaly field is calculated as the edge intensity. Regions that simultaneously satisfy the conditions of anomaly intensity exceeding the first threshold, edge intensity exceeding the second threshold, and spatial morphology continuity are marked as candidate regions for vortex anomaly boundaries.

[0013] Preferably, in step S4, the specific process of constructing the internal wave disturbance surrogate index by combining the thermocline displacement residual and the short-scale sound velocity residual, and marking the internal wave disturbance candidate region, includes: Calculate the low-frequency background of the undulating surface of the thermocline, and subtract the low-frequency background from the undulating surface of the thermocline to obtain the displacement residual of the thermocline. Calculate the short-scale sound velocity residuals of the sound velocity field at a given depth layer; An internal wave disturbance proxy index is constructed using the thermocline displacement residual, short-scale sound velocity residual, and scale conversion coefficient. Regions where the internal wave disturbance proxy index exceeds the adaptive threshold are marked as internal wave disturbance candidate regions.

[0014] Preferably, in step S5, the specific process of outputting structured information for each identified environmental structure and calculating the confidence level of each structure by integrating multiple factors includes: For each candidate structure, output the structure type, 3D spatial location, diagnostic index value, adaptive threshold, binary mask or boundary line, recommended visualization color and transparency, and associated index pointing to the sound velocity field; The confidence level is determined as follows: for frontal candidate regions, the confidence level is positively correlated with the degree to which the joint frontal index exceeds the threshold and the spatial consistency between the gradients of each variable; for vortex anomaly boundary candidate regions, the confidence level is positively correlated with the product of anomaly intensity and edge intensity, boundary closure, and multi-depth consistency; for internal wave disturbance candidate regions, the confidence level is positively correlated with the amplitude of the internal wave disturbance surrogate index, the regularity of wave train morphology, and the consistency of distribution along the strata. The confidence level was then normalized.

[0015] The technical solution of the present invention is a discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method as described in any one of claims 1 to 9.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs a discrete profile-driven three-dimensional sound velocity field reconstruction and ocean structure identification method and system. In sound velocity field reconstruction, a low-rank background field is constructed using empirical orthogonal function decomposition, effectively preserving the main spatial features of the target sea area. Then, optimal interpolation is used to fuse discrete profile observations, ensuring the reconstructed field possesses both the physical consistency of the background field and the true constraints of the observation data, significantly improving the continuity and reliability of the three-dimensional sound velocity field under sparse observation conditions. In ocean structure identification, the system simultaneously diagnoses thermocline undulations, frontal candidate regions, vortex boundary candidate regions, and internal wave disturbance candidate regions. In particular, it constructs a joint frontal index through multivariate normalized weighting and constructs an internal wave disturbance proxy index by coupling thermocline displacement residuals and short-scale sound velocity residuals, realizing the identification of multiple types of ocean structures. The integrated collaborative identification method makes fuller use of information and more reliable identification results. It introduces a confidence leveling mechanism, comprehensively considering factors such as intensity, continuity, multivariate consistency, and morphological features, assigning high, medium, and low confidence levels to each type of identification result. Simultaneously, it conducts regional error statistics through reserved profile cross-validation, quantitatively evaluating the applicability of the reconstructed field in complex environmental structures, effectively improving the robustness and reliability of the system in data-sparse or structurally complex regions. The three-dimensional sound velocity field and structural marker layer output by this invention can directly serve acoustic models such as ray tracing, parabolic propagation, and matched-field localization, providing high-quality data support for underwater sound propagation prediction and sonar performance evaluation, combining environmental structural semantic information with uncertainty quantification indicators. Attached Figure Description

[0017] Figure 1 This is a flowchart of a discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method proposed in this invention; Figure 2 This is a schematic diagram of the spatial mapping from a discrete vertical profile to a three-dimensional regular mesh. Figure 3 A schematic diagram of how to generate a three-dimensional analytical sound velocity field by fusing the empirical orthogonal function background field with the observation profile; Figure 4 This is a schematic diagram illustrating the extraction of the undulating surface of the thermocline. Figure 5 A schematic diagram for identifying candidate regions of multivariable fronts; Figure 6 Schematic diagram for identifying candidate regions for vortex anomaly boundaries; Figure 7 Schematic diagram for identifying candidate regions of internal wave disturbance; Figure 8 A visualization diagram of the three-dimensional sound velocity field and environmental structure layers; Figure 9 A schematic diagram for reserving observation profiles for cross-validation. Detailed Implementation

[0018] Example 1: This invention proposes a discrete profile-driven three-dimensional sound velocity field reconstruction and ocean structure identification method, such as... Figure 1 As shown, the specific implementation steps include the following: S1. Perform quality control on the multiple discrete station vertical profiles obtained, remove outliers and depth reversal points, unify each profile to the same depth layer through interpolation, and then calculate the sound velocity value of each depth layer based on temperature, salinity and depth using empirical sound velocity formulas to form a standardized discrete vertical sound velocity observation set. S2. Construct a three-dimensional regular grid covering the target sea area, collect historical ocean fields or model background fields as a sample set, calculate the mean field and perform empirical orthogonal function decomposition on the mean-removed samples, extract several main modes, estimate the coefficients of each mode using discrete observation profiles and take the average, and superimpose the mean field with the main modes to generate a low-rank three-dimensional background sound velocity field that preserves the main spatial features of the area. S3. Using the empirical orthogonal function background field as the prior field, the discrete profile sound velocity observation is used as the observation. The optimal interpolation method is used for data fusion. The gain weight is calculated through the background error covariance matrix (based on the spatial correlation function to set the correlation scale) and the observation error covariance matrix (determined according to the instrument type and quality control results) to perform observation correction on the background field and generate a continuous three-dimensional analytical sound velocity field. S4. Based on the three-dimensional analysis field and the corresponding temperature, salinity, and density fields, diagnose the environmental structure: extract the location of the maximum vertical temperature gradient to form the undulating surface of the thermocline; calculate the horizontal gradients of sound velocity, temperature, salinity, and density and construct a joint front index by normalizing and weighting, marking high-value continuous areas as front candidate areas; perform smooth background separation on the sound velocity field, and identify vortex boundary candidate areas by combining anomaly intensity and edge intensity; construct an internal wave disturbance surrogate index by combining thermocline displacement residuals and short-scale sound velocity residuals, and mark high-value areas as internal wave disturbance candidate areas. S5. Output structured information for each identified environmental structure, including structure type, three-dimensional spatial location, diagnostic index value, adaptive threshold, binary mask or boundary line, recommended visualization color and transparency, and association index pointing to the sound velocity field; at the same time, calculate the confidence level of each structure by integrating multiple factors (such as intensity, continuity, multivariate consistency, morphological features, etc.). S6. Extract a portion from all profiles as a validation set, reconstruct the three-dimensional sound velocity field using only the remaining profiles, interpolate the predicted values ​​at the validation profile locations and compare them with the actual observations, calculate the root mean square error, mean absolute error, bias, and maximum absolute error, and separately calculate the regional errors of the frontal region, vortex region, internal wave region, and background region; finally, output the three-dimensional sound velocity field and structure marker layer as a standardized grid data file, and provide data interfaces for acoustic models such as ray tracing, parabolic equation propagation, and matched field localization.

[0019] In one specific embodiment, step S1 aims to convert the acquired multiple discrete vertical profiles into a set of sound velocity observations at a unified depth layer, as detailed below: Please see Figure 2 This illustrates a schematic diagram of the mapping of discrete profile observation points to a three-dimensional regular mesh. S11. Deploy or select several observation stations in the target sea area, and each station acquires temperature, salinity and depth profiles from the sea surface to a preset depth; the observation platform can be a CTD, XCTD, Argo buoy, underway profiler, underwater mobile platform or moored mooring, etc. Assume a total of N profiles are obtained, and the original data of the i-th profile contains a depth sequence. Temperature sequence and salinity sequence Its planar position is ; Quality control operations include, but are not limited to: Remove abnormal records where the temperature or salinity exceeds the normal range for climatological conditions; Remove points that are not monotonically increasing in the depth sequence (depth reversal); Remove spike noise with vertical gradients exceeding a preset threshold; Median filtering (window size of 3-5 sampling points) is applied to the profile to smooth instrument noise; S12. Set the uniform depth layer of the target sea area as follows: The depth range is from the sea surface to the maximum observation depth (e.g., 0–1000 m), and the number of layers M can be determined according to the resolution requirements; For each profile, linear interpolation or cubic spline interpolation is used to interpolate the original temperature and salinity data onto a uniform depth layer, resulting in a standardized temperature profile. and salinity profile ; Where m = 1, 2, ..., M; For missing layers that appear near the sea surface or are deeper than the deepest observation depth of the profile after interpolation, if the depth of the missing layer is shallower than the shallowest effective depth of the profile, the shallowest effective value is extrapolated; if the depth of the missing layer is deeper than the deepest effective depth of the profile, the deepest effective value is extended downward, or it is marked as missing and excluded in subsequent fusion. S13. At each uniform depth layer, the speed of sound is calculated using the empirical formula based on temperature T (°C), salinity S (PSU), and depth z (m). This embodiment uses the Mackenzie formula: ; It should be noted that the empirical sound speed formula can also adopt the Chen-Millero formula, the Del Grosso formula, or other equivalent seawater sound speed models. This yields the sound velocity value of the i-th profile at a uniform depth layer. ; The sound velocity observations from all profiles constitute the discrete observation set Y: .

[0020] In one specific embodiment, step S2 constructs a background sample set using historical ocean fields, model background fields, or reanalysis fields, and generates a low-rank three-dimensional background sound velocity field through empirical orthogonal function (EOF) decomposition. The specific implementation method is as follows: S21. Establish a three-dimensional regular grid covering the target sea area. The horizontal grid resolution is set according to the data density and computing power, and the vertical depth layer is consistent with the unified depth layer in step S1. ; Set the number of horizontal grid points to The vertical number of layers is M, and the total number of grid points is... ; In the formula, and These represent the number of grid points in the longitude direction and the number of grid points in the latitude direction, respectively. S22. Collect historical field, model background field, or reanalysis field data of the target sea area to form a three-dimensional sound velocity sample set. ; In the formula, K is the total number of samples (e.g., 365 daily samples). It should be noted that the background data sources include WOA climatological mean field, HYCOM reanalysis data, GLORYS reanalysis data, or historical CTD / Argo data; if historical data is lacking, the current observation profile can be used to generate an expanded sample through spatial perturbation. S23. Calculate the mean field by averaging the sample set point by point: ; S24. Subtract the mean field from each sample to obtain the perturbation field: ; The three-dimensional perturbation field is expanded into a one-dimensional vector according to the spatial grid, and a sample matrix X (dimension L×K) is constructed. Singular value decomposition is then performed on X: ; In the formula, U is The left singular vector matrix; for Diagonal singular value matrix; for The right singular vector matrix; r is the number of retained principal modes; The columns of the left singular vectors are the empirical orthogonal function modes. ; It should be noted that the value of r is determined based on the cumulative variance contribution rate, and is usually selected as the minimum number of modes required for the cumulative variance contribution rate to reach a preset threshold (e.g., 95%). S25. Using the N discrete sound velocity profiles obtained in step S1 to observe Y, estimate the coefficients corresponding to each principal mode. The estimation method can be the least squares method, that is, at the observation location... At a point that minimizes the sum of squares of the difference between the reconstructed sound velocity and the observed sound velocity; Finally, the low-rank three-dimensional background sound velocity field was reconstructed. : .

[0021] In a specific embodiment, step S3 will generate the empirical orthogonal function background field (low-rank three-dimensional background sound velocity field) constructed in step S2. As a priori field, the discrete profile observation Y obtained in step S1 is used as the observation quantity and fused using the optimal interpolation method to generate a continuous three-dimensional analytical sound velocity field. ,like Figure 3 As shown, the specific implementation method is as follows: The core calculation formula for optimal interpolation is as follows: ; In the formula, B is the background error covariance matrix; The three-dimensional sound velocity field after analysis; y is the observation vector composed of all discrete profile observations in step S1; H is the observation operator that maps from the three-dimensional mesh to the observation position; R is the observation error covariance matrix; It should be noted that the background error covariance matrix B is set as a spatial correlation function, and its correlation scale is determined according to the sea area scale, front direction, vortex radius, topographic constraints or empirical orthogonal function mode energy; the observation error covariance matrix R is determined according to the instrument type, profile quality control results and vertical interpolation error, and is usually set as a diagonal matrix, with its diagonal elements being the error variance of each observation.

[0022] In a specific embodiment, step S4 calculates multivariable environmental structure diagnostic indicators based on the continuous three-dimensional analytical sound velocity field generated in step S3 and its corresponding temperature field, salinity field, and density surrogate field. According to these diagnostic indicators, thermocline undulation surfaces, frontal candidate regions, vortex anomaly boundary candidate regions, and internal wave disturbance candidate regions are generated, as detailed below: S41. Extract the undulating surface of the thermocline, such as Figure 4 As shown, at each horizontal grid point (x,y), based on the three-dimensional temperature field... Calculate the vertical temperature gradient with depth. ; Find the location with the strongest negative gradient within a preset depth range, and use it as the representative depth of the thermocline at that horizontal location: ; From all horizontal grid points The undulating surface of the thermocline is used to represent the three-dimensional spatial variation of the thermocline depth; S42. Identify candidate regions for multivariable fronts, such as Figure 5 As shown, at a given depth layer or multiple depth layers, the sound velocity c, temperature T, salinity S, and density proxy are calculated respectively. Horizontal gradient: , , , ; In the formula, For horizontal gradient operators; Normalize the gradients of each variable to construct the joint frontal index: ; In the formula, , , and These are the weighting coefficients for each variable; In one implementation, the weighting coefficients of each variable are equally weighted, i.e. ; In another implementation, the weighting coefficients are adaptively determined based on observation quality, sea area type, and mission requirements; Combined frontal index Continuous regions exceeding a preset threshold or a high quantile threshold are marked as frontal candidate regions. In this embodiment, the threshold is an adaptive threshold within the range of 85% to 95% quantile. S43. Identify candidate vortex anomaly boundaries, such as Figure 6 As shown, for a given single-depth layer or multiple depth layers, the formula for smoothing the background separation of the sound velocity field is as follows: ; In the formula, To represent a smoothing operator in two-dimensional or three-dimensional space, Gaussian filtering, median filtering, or moving average filtering can be used. Calculate the intensity of local sound velocity anomalies: ; Calculate the anomaly edge strength: ; Regions that simultaneously meet the conditions of anomaly intensity exceeding a set threshold, edge intensity exceeding a set threshold, and spatial morphological continuity are marked as vortex anomaly boundary candidate regions. In a preferred embodiment, candidate boundaries are further optimized and screened by considering factors such as closure, curvature, area range, scale range, and multi-depth consistency. S44, such as Figure 7 As shown, identification is based on the internal wave manifestations as thermocline displacement and short-scale sound velocity disturbances. First, the undulating surface of the thermocline is calculated. Low-frequency background: ; Calculate the residual displacement of the thermocline: ; Calculate the short-scale sound velocity residuals at a given depth layer: ; Constructing an internal wave disturbance proxy index: ; In the formula, The scale conversion factor from thermocline displacement to sound velocity disturbance; Will Regions exceeding the adaptive threshold are marked as candidate regions for internal wave disturbances; It should be noted that the candidate regions are further screened based on strip-like, wave-like, distribution along the lattice, multi-depth coherence, and time series consistency.

[0023] In a specific embodiment, step S5 outputs structured information, confidence assessment, and visualization parameters for each type of candidate structure, such as... Figure 8 As shown, the specific implementation method is as follows: For each candidate structure identified in step S4, at least the following information should be output: Structure type; Three-dimensional spatial position; Diagnostic indicator values; Adaptive threshold; Binary mask or boundary line; Confidence level; Visualize color and transparency; The association index with the three-dimensional sound velocity field; The confidence level is determined by a combination of the following factors: For frontal candidate regions, the confidence level is positively correlated with the degree to which the joint frontal index exceeds the threshold and the spatial consistency between the gradients of each variable. For candidate regions of vortex anomaly boundaries, the confidence level is positively correlated with the product of anomaly intensity and edge intensity, boundary closure, and multi-depth consistency. For candidate regions of internal wave disturbances, the confidence level is positively correlated with the amplitude, wave train morphology regularity, and consistency of distribution along the lattice of the internal wave disturbance proxy index. In one implementation, the confidence level is calculated by a weighted average of the aforementioned relevant indicators and normalized to the [0,1] interval.

[0024] In one specific embodiment, step S6 involves quantitatively evaluating the reconstruction reliability through reserved profile cross-validation and providing a standardized data interface, such as... Figure 9 As shown, the specific implementation method is as follows: S61. To evaluate the reliability of the reconstruction, a portion of the N vertical profiles from step S1 is extracted as a validation set, and the three-dimensional sound velocity field is reconstructed using only the remaining profiles. Predicted profiles are obtained by interpolation at the validation profile locations and compared with the actual observed profiles. Output the following metrics: Root mean square error (RMSE); Mean absolute error (MAE); Bias; Maximum absolute error; Depth error; Regional error; The errors in the frontal region, vortex candidate region, internal wave candidate region, and ordinary background region were statistically analyzed separately to further evaluate the applicability of this method in complex environmental structures. S62. Output a three-dimensional sound velocity field and environmental structure marker layers for use in three-dimensional visualization, sound ray tracing, parabolic propagation model, and sound source localization algorithm. The data interface outputs data in a general grid data format, including horizontal coordinates, vertical coordinates, sound velocity values, and various structure marker layers, which can be directly used as environmental input for subsequent acoustic models.

[0025] Example 2: A discrete profile-driven three-dimensional sound velocity field reconstruction and ocean structure identification method proposed in this invention is used to execute the discrete profile-driven three-dimensional sound velocity field reconstruction and ocean structure identification method proposed in Example 1, and includes: Memory; processor; A computer program stored in the memory and capable of running on the processor; The processor executes a computer program to implement a discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method as described in Embodiment 1 above.

[0026] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for reconstructing a three-dimensional sound velocity field and identifying marine structures using discrete profile-driven methods, characterized in that, The implementation steps include the following: S1. Perform quality control on the vertical profiles of multiple discrete stations and unify the depth layer. Calculate the sound velocity value based on temperature, salinity and depth to form a standardized set of discrete vertical sound velocity observations. S2. Construct a three-dimensional regular grid covering the target sea area, collect historical ocean fields or model background fields as a sample set, calculate the mean field and extract several principal modes by empirical orthogonal function decomposition, estimate the coefficients of each mode using discrete observation profiles, and generate a low-rank three-dimensional background sound velocity field by weighted superposition of the mean field and principal modes. S3. Using the background sound velocity field as the prior field and the discrete profile sound velocity observation as the observation, the optimal interpolation method is used to fuse the data and generate a continuous three-dimensional analytical sound velocity field. S4. Based on the three-dimensional analysis of the sound velocity field and the corresponding temperature field, salinity field and density field, the maximum position of the vertical temperature gradient is extracted to form the undulating surface of the thermocline. The horizontal gradients of sound velocity, temperature, salinity and density are calculated and normalized to construct a joint front index and mark the front candidate area. The sound velocity field is smoothed and the background is separated. The candidate area of ​​the vortex boundary is identified by combining the anomaly intensity and the edge intensity. The internal wave disturbance proxy index is constructed by combining the thermocline displacement residual and the short-scale sound velocity residual and the candidate area of ​​the internal wave disturbance is marked. S5. Output structured information for each type of identified environmental structure, and calculate the confidence level of each type of structure by comprehensively considering strength, continuity, multivariate consistency and morphological features. S6. Extract a portion from all profiles as a validation set, reconstruct the three-dimensional sound velocity field using the remaining profiles, interpolate the predicted values ​​at the validation profile locations and compare them with the actual observations to calculate the error, and statistically analyze the error by region. Finally, output the three-dimensional sound velocity field and structural marker layer and provide a data interface.

2. The method for discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification according to claim 1, characterized in that, In step S1, the specific process of performing quality control and unifying the depth layer on the acquired vertical profiles of multiple discrete stations, and calculating the sound velocity value based on temperature, salinity, and depth includes: Abnormal records with temperature or salinity exceeding the normal range for climatological conditions are removed; points that are not monotonically increasing in the depth sequence are removed; spike noise with vertical gradients exceeding a preset threshold is removed; and median filtering is applied to the profile. A uniform depth layer is set for the target sea area, and the temperature and salinity data of each profile are interpolated to the uniform depth layer using linear interpolation or cubic spline interpolation. At each uniform depth layer, the speed of sound is calculated using an empirical formula based on temperature, salinity, and depth.

3. The method for discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification according to claim 2, characterized in that, In step S2, the specific process of collecting historical ocean fields or model background fields as a sample set, calculating the mean field, and performing empirical orthogonal function decomposition to extract several principal modes includes: The mean field is obtained by averaging the sample set point by point, and the perturbation field is obtained by subtracting the mean field from each sample. The three-dimensional perturbation field is expanded into a one-dimensional vector according to the spatial grid to construct a sample matrix. The sample matrix is ​​then subjected to singular value decomposition to obtain the left singular vector matrix, the diagonal singular value matrix, and the right singular vector matrix. Each column of the left singular vector is taken as an empirical orthogonal function mode, and the number of main modes to be retained is determined based on the cumulative variance contribution rate.

4. The discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method according to claim 3, characterized in that, In step S3, the specific process of data fusion using the background sound velocity field as the prior field and the optimal interpolation method includes: The background error covariance matrix is ​​set as a spatial correlation function, and the correlation scale of the spatial correlation function is determined based on the sea area scale, front direction, vortex radius, topographic constraints, or empirical orthogonal function mode energy. The observation error covariance matrix is ​​set as a diagonal matrix, and the diagonal elements of the diagonal matrix are determined based on the instrument type, profile quality control results, and vertical interpolation error. Gain weights are calculated using the background error covariance matrix and the observation error covariance matrix. The background field is then corrected using these gain weights to generate a continuous three-dimensional analytical sound velocity field.

5. The discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method according to claim 4, characterized in that, In step S4, the specific process of extracting the location of the maximum vertical temperature gradient that constitutes the undulating surface of the thermocline includes: At each horizontal grid point, the vertical gradient of temperature with depth is calculated based on the three-dimensional temperature field. The location with the strongest negative gradient within a preset depth range is found as the representative depth of the thermocline at that horizontal location. The thermocline undulating surface is formed by the representative depths of the thermoclines at all horizontal grid points.

6. The discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method according to claim 5, characterized in that, In step S4, the specific process of calculating the horizontal gradients of sound speed, temperature, salinity, and density, constructing a joint frontal index using normalized weights, and labeling frontal candidate regions includes: Calculate the horizontal gradients of sound velocity, temperature, salinity, and density proxy at a given depth layer. The gradients of each variable are normalized, and the joint front index is obtained by multiplying the normalized gradients of each variable by their corresponding weight coefficients and then summing them. Continuous regions where the joint front index exceeds an adaptive threshold are marked as frontal candidate regions; the weighting coefficients are either equally weighted or adaptively determined based on observation quality.

7. The method for discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification according to claim 6, characterized in that, In step S4, the specific process of smoothing the background of the sound velocity field and identifying vortex boundary candidate regions by combining anomaly intensity and edge intensity includes: Spatial smoothing of the sound velocity field at a given depth layer yields a smoothed background field, and subtraction of the smoothed background field from the original sound velocity field yields a sound velocity anomaly field. The absolute value of the sound velocity anomaly field is calculated as the anomaly intensity, and the horizontal gradient magnitude of the sound velocity anomaly field is calculated as the edge intensity. Regions that simultaneously satisfy the conditions of anomaly intensity exceeding the first threshold, edge intensity exceeding the second threshold, and spatial morphology continuity are marked as candidate regions for vortex anomaly boundaries.

8. The method for discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification according to claim 7, characterized in that, In step S4, the specific process of constructing the internal wave disturbance surrogate index by combining the thermocline displacement residual and the short-scale sound velocity residual, and marking the internal wave disturbance candidate region, includes: Calculate the low-frequency background of the undulating surface of the thermocline, and subtract the low-frequency background from the undulating surface of the thermocline to obtain the displacement residual of the thermocline. Calculate the short-scale sound velocity residuals of the sound velocity field at a given depth layer; An internal wave disturbance proxy index is constructed using the thermocline displacement residual, the short-scale sound velocity residual, and the scale conversion coefficient. Regions where the internal wave disturbance proxy index exceeds the adaptive threshold are marked as internal wave disturbance candidate regions.

9. The method for discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification according to claim 8, characterized in that, In step S5, the specific process of outputting structured information for each type of identified environmental structure, and calculating the confidence level of each type of structure by integrating multiple factors, includes: For each candidate structure, output the structure type, 3D spatial location, diagnostic index value, adaptive threshold, binary mask or boundary line, recommended visualization color and transparency, and associated index pointing to the sound velocity field; The confidence level is determined as follows: for frontal candidate regions, the confidence level is positively correlated with the degree to which the joint frontal index exceeds the threshold and the spatial consistency between the gradients of each variable; for vortex anomaly boundary candidate regions, the confidence level is positively correlated with the product of anomaly intensity and edge intensity, boundary closure, and multi-depth consistency; for internal wave disturbance candidate regions, the confidence level is positively correlated with the amplitude of the internal wave disturbance surrogate index, the regularity of wave train morphology, and the consistency of distribution along the strata. The confidence level was then normalized.

10. A discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification system, characterized in that, include: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements a discrete profile-driven three-dimensional sound velocity field reconstruction and marine structure identification method as described in any one of claims 1 to 9.

Images