A method and system for processing an ocean current meter signal by multi-scale time-frequency analysis

Abstract

This invention discloses a multi-scale time-frequency analysis method and system for ocean current meter signal processing, relating to the field of marine information processing. The method includes: S1: acquiring raw data and performing physical benchmark correction to obtain three-dimensional current velocity data and basic mass labels; S2: extracting first and second features from the three-dimensional current velocity data through spatiotemporal decoupling; S3: fusing and separating the three-dimensional current velocity data based on the first and second features to obtain an optimized set of flow field parameters, and generating the uncertainty of each parameter by combining the basic mass labels and error source information; S4: performing comprehensive quality assessment and grading based on the first and second features and the uncertainty, outputting a data product with mass labels and uncertainty information. By extracting spatiotemporal dual-dimensional features and using them to guide signal fusion and quality assessment, highly robust processing of ocean current data and end-to-end quality traceability are achieved, significantly improving the reliability of the data product.

Description

Technical Field

[0001] This invention relates to the field of marine information processing, specifically to a method and system for processing ocean current meter signals using multi-scale time-frequency analysis. Background Technology

[0002] Acoustic Doppler current profilers are the main equipment for acquiring the three-dimensional flow field structure of the ocean. The routine data processing flow of this equipment usually includes: coordinate and sound velocity correction of the raw beam data, threshold-based quality control based on diagnostic parameters, and time-frequency analysis of the flow velocity sequence to separate different physical components.

[0003] However, this process has significant limitations in complex observation scenarios with low to medium current velocities or multiple sources of disturbance such as strong waves and internal waves. Fixed quality control thresholds are difficult to accurately adapt to dynamically changing signal quality; traditional signal separation methods have limited effectiveness in processing low signal-to-noise ratio signals with multiple aliasing processes; and the final data products often lack systematic and quantitative reliability assessments. This restricts the ability to acquire high-confidence flow field data in complex marine environments.

[0004] Therefore, more robust and adaptive current meter signal processing techniques still need to be developed in this field. Summary of the Invention

[0005] Based on the shortcomings of the prior art described above, the purpose of this invention is to provide a method and system for processing ocean current meter signals using multi-scale time-frequency analysis, so as to solve the above-mentioned technical problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for processing ocean current meter signals using multi-scale time-frequency analysis, comprising:

[0007] S1: Collect raw observation data from the current meter, perform physical benchmark correction, and obtain three-dimensional current velocity data and basic mass labels;

[0008] S2: Spatiotemporal decoupling feature extraction of three-dimensional flow velocity data: The first feature is generated in the spatial dimension based on the characteristics of the profile structure, and the second feature is generated in the temporal dimension based on the characteristics of the signal multi-scale decomposition.

[0009] S3: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information.

[0010] S4: Perform comprehensive quality assessment and classification based on the first feature, the second feature, and the uncertainty, and output a data product with quality markings and uncertainty information.

[0011] The present invention is further configured such that S1 includes: a data correction step and a quality inspection step;

[0012] The data correction steps include:

[0013] Acquire raw observation data from the current meter, including raw beam velocity, raw diagnostic parameters, and synchronous observation data from auxiliary sensors. The raw diagnostic parameters include pulse effectiveness, velocity estimation correlation, and inter-beam closure error velocity. The synchronous observation data includes carrier attitude data and environmental parameters.

[0014] Based on the carrier attitude data, the original beam velocity is transformed from the instrument coordinate system to the geographic coordinate system through coordinate rotation;

[0015] Based on the inversion of sound velocity profiles using environmental parameters, the sound wave propagation path and depth positioning are corrected using a sound ray tracing algorithm based on the sound velocity profiles, the corrected beam incidence angle is obtained, and the three-dimensional flow velocity is recalculated.

[0016] For the data after coordinate transformation and sound velocity correction, pulse-level outlier detection based on statistical distribution, profile rationality check based on fluid kinematic continuity, and cross-validation based on multi-beam vector closure are performed sequentially to identify and remove outlier observations, thus obtaining preprocessed three-dimensional flow velocity data.

[0017] The present invention is further configured such that the quality inspection step includes:

[0018] Based on the original diagnostic parameters, the data reliability of each observation unit is evaluated, and corresponding basic quality labels are generated.

[0019] The present invention is further configured such that S2 includes: a spatial structure feature extraction step and a temporal dynamic feature extraction step;

[0020] The spatial structure feature extraction steps include:

[0021] Based on three-dimensional velocity data, the vertical coherence of velocity sequences at adjacent depth layers is calculated through cross-correlation analysis;

[0022] The vertical gradient of the three-dimensional flow velocity data is calculated using differential operations, and the stability of the vertical gradient in the time dimension is evaluated.

[0023] By combining vertical coherence, vertical gradient stability, and basic quality labels, a first feature is generated to characterize the continuity and reliability of the cross-sectional structure.

[0024] The present invention is further configured such that the time-dynamic feature extraction step includes:

[0025] Adaptive multi-scale decomposition was performed on the velocity time series of each depth layer in the three-dimensional velocity data to obtain multiple intrinsic mode components.

[0026] The instantaneous frequency characteristics of each intrinsic mode component are obtained by Hilbert transform analysis, the corresponding energy distribution characteristics are obtained by variance calculation or power spectrum analysis, and the corresponding vertical propagation characteristics are obtained by calculating the cross-correlation coefficients of the same mode components in adjacent depth layers.

[0027] Based on frequency characteristics, energy distribution, and vertical propagation properties, the physical process categories corresponding to each intrinsic mode component are identified;

[0028] Evaluate the frequency stability, energy concentration, and modal orthogonality of modal components belonging to the same physical process category;

[0029] The second feature is generated by combining frequency stability, energy concentration, modal orthogonality, and preset physical process energy weights.

[0030] The present invention is further configured such that the flow field parameter set includes: velocity profile, tidal current component and turbulence parameters, wherein the turbulence parameters include: turbulence intensity, Reynolds stress and turbulent kinetic energy dissipation rate.

[0031] The present invention is further configured such that S3 includes: a flow field parameter optimization step and an uncertainty quantification step;

[0032] The flow field parameter optimization steps include:

[0033] Using the first feature as the weight, the three-dimensional velocity data is vertically weighted and smoothed to reconstruct the optimized velocity profile.

[0034] Based on the temporal dynamic clarity indicated by the second feature, multi-scale filtering is performed on the time series of each depth layer in the three-dimensional flow velocity data to separate the turbulent components.

[0035] Based on the turbulent components, the turbulence intensity is obtained by calculating the variance of the corresponding fluctuating velocity, and the Reynolds stress is obtained by calculating the covariance of the fluctuating velocity in different directions.

[0036] Based on the turbulent components, the power spectrum is estimated to obtain the energy spectrum, and the slope is fitted in the inertial subregion based on the energy spectrum to obtain the turbulent kinetic energy dissipation rate.

[0037] Turbulence parameters are constructed by combining turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate.

[0038] The present invention is further configured such that the uncertainty quantification step includes:

[0039] Error source information includes: carrier attitude error introduced by the inertial measurement unit during the acquisition of carrier attitude data, and sound velocity profile error introduced by sensor accuracy or environmental disturbance during the inversion of sound velocity profile.

[0040] Using basic quality labels, error source information, and preset instrument inherent errors as inputs, the uncertainties of velocity profiles, tidal current components, and turbulence parameters are calculated respectively through an error propagation model.

[0041] The present invention is further configured such that S4 includes:

[0042] A comprehensive score is obtained by weighted fusion of the first feature, the second feature, and the uncertainty.

[0043] Based on a preset threshold range, the comprehensive score is mapped to discrete quality levels to generate the final quality label;

[0044] The final quality label and uncertainty are correlated with the optimized flow field parameter set and packaged into a standardized data product.

[0045] The present invention also provides a multi-scale time-frequency analysis current meter signal processing system, the system comprising:

[0046] Data calibration and quality control module: Collects raw observation data from the current meter, performs physical benchmark calibration, and obtains three-dimensional current velocity data and basic mass labels;

[0047] Spatiotemporal feature profiling module: performs spatiotemporal decoupling feature extraction on 3D flow velocity data: generates the first feature in the spatial dimension based on the characteristics of the cross-sectional structure, and generates the second feature in the temporal dimension based on the characteristics of the multi-scale decomposition of the signal;

[0048] Intelligent fusion and credibility generation module: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information.

[0049] Product Packaging and Quality Reporting Module: Performs comprehensive quality assessment and classification based on the first feature, the second feature, and uncertainty, and outputs data products with quality markings and uncertainty information.

[0050] This invention provides a multi-scale time-frequency analysis method and system for ocean current meter signal processing. The method comprises: S1: acquiring raw observation data from the ocean current meter and performing physical benchmark correction to obtain three-dimensional current velocity data and a basic mass label; S2: extracting spatiotemporal decoupling features from the three-dimensional current velocity data: generating a first feature based on profile structure characteristics in the spatial dimension, and generating a second feature based on the multi-scale decomposition characteristics of the signal in the temporal dimension; S3: fusing and separating the three-dimensional current velocity data based on the first and second features to obtain an optimized set of flow field parameters, and generating the uncertainty of each parameter by combining the basic mass label and error source information; S4: performing comprehensive quality assessment and grading based on the first feature, the second feature, and the uncertainty, and outputting a data product with mass labels and uncertainty information. The beneficial effects include:

[0051] A spatiotemporal decoupling dual-dimensional feature extraction method is proposed: a first feature representing structural integrity in the spatial dimension and a second feature representing dynamic clarity in the temporal dimension are used to quantify the vertical continuity of the velocity profile and the separability of physical components in the time series, respectively, thereby achieving a more refined and physically meaningful feature characterization of complex ocean current signals.

[0052] Establish a feature-driven intelligent fusion and credibility assessment mechanism: use the adaptive guidance of two-dimensional features to reconstruct the velocity profile and separate the physical components, and dynamically generate uncertain quantitative results that are directly related to data quality based on the reconstruction and separation results, which significantly improves the robustness of data processing and the reliability of output results in complex environments.

[0053] A complete data quality traceability and product output system has been established: by using multi-dimensional scoring, the first and second characteristics and uncertainty are comprehensively transformed into intuitive quality level marks, and associated with the final data product packaging, realizing full-process quality traceability from raw signals to standard data products, which greatly enhances the practical value of data and user trust.

[0054] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it according to the contents of the specification, and to make the above and other objectives, features and advantages of this application more obvious and understandable, a spatiotemporal decoupling dual-dimensional feature extraction method is proposed: by representing the first feature of structural integrity in the spatial dimension and the second feature of dynamic clarity in the time dimension, the vertical continuity of the velocity profile and the separability of physical components in the time series are quantified respectively, thereby realizing a more refined and physically meaningful feature characterization of complex ocean current signals.

[0055] Establish a feature-driven intelligent fusion and credibility assessment mechanism: use the adaptive guidance of two-dimensional features to reconstruct the velocity profile and separate the physical components, and dynamically generate uncertain quantitative results that are directly related to data quality based on the reconstruction and separation results, which significantly improves the robustness of data processing and the reliability of output results in complex environments.

[0056] A complete data quality traceability and product output system is established: by using multi-dimensional scoring, the first and second features, as well as uncertainty, are comprehensively transformed into intuitive quality level labels, and associated with the final data product packaging. This achieves full-process quality traceability from raw signals to standard data products, greatly enhancing the practical value of the data and user trust. A spatiotemporal decoupling dual-dimensional feature extraction method is proposed: by representing the structurally sound first feature through the spatial dimension and the dynamically clear second feature through the temporal dimension, the continuity of the current profile in the vertical direction and the separability of the physical components in the time series are quantified respectively, achieving a more refined and physically meaningful feature characterization of complex ocean current signals.

[0057] Establish a feature-driven intelligent fusion and credibility assessment mechanism: use the adaptive guidance of two-dimensional features to reconstruct the velocity profile and separate the physical components, and dynamically generate uncertain quantitative results that are directly related to data quality based on the reconstruction and separation results, which significantly improves the robustness of data processing and the reliability of output results in complex environments.

[0058] A complete data quality traceability and product output system is formed: through multi-dimensional scoring, the first feature, the second feature, and uncertainty are comprehensively transformed into an intuitive quality level mark, and associated with the final data product packaging, realizing full-process quality traceability from the original signal to the standard data product, which greatly enhances the practical value of the data and user trust. The following are specific implementation methods of this application. Attached Figure Description

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

[0060] Figure 1 A flowchart illustrating a multi-scale time-frequency analysis current meter signal processing method is shown as an exemplary embodiment of the present invention;

[0061] Figure 2 This is a schematic diagram of the structure of a current meter signal processing system for multi-scale time-frequency analysis, which is an exemplary embodiment of the present invention. Detailed Implementation

[0062] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0063] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0064] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0065] Example 1:

[0066] A multi-scale time-frequency analysis method for ocean current meter signal processing, such as Figure 1 As shown, it includes:

[0067] S1: Collect raw observation data from the current meter, perform physical benchmark correction, and obtain three-dimensional current velocity data and basic mass labels;

[0068] S2: Spatiotemporal decoupling feature extraction of three-dimensional flow velocity data: The first feature is generated in the spatial dimension based on the characteristics of the profile structure, and the second feature is generated in the temporal dimension based on the characteristics of the signal multi-scale decomposition.

[0069] S3: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information.

[0070] S4: Perform comprehensive quality assessment and classification based on the first feature, the second feature, and the uncertainty, and output a data product with quality markings and uncertainty information.

[0071] The present invention is further configured such that S1 includes: a data correction step and a quality inspection step;

[0072] The data correction steps include:

[0073] Acquire raw observation data from the current meter, including raw beam velocity, raw diagnostic parameters, and synchronous observation data from auxiliary sensors. The raw diagnostic parameters include pulse effectiveness, velocity estimation correlation, and inter-beam closure error velocity. The synchronous observation data includes carrier attitude data and environmental parameters.

[0074] Based on the carrier attitude data, the original beam velocity is transformed from the instrument coordinate system to the geographic coordinate system through coordinate rotation;

[0075] Based on the inversion of sound velocity profiles using environmental parameters, the sound wave propagation path and depth positioning are corrected using a sound ray tracing algorithm based on the sound velocity profiles, the corrected beam incidence angle is obtained, and the three-dimensional flow velocity is recalculated.

[0076] After coordinate transformation and sound velocity correction, the data undergoes sequential processing including pulse-level outlier detection based on statistical distribution, profile rationality verification based on fluid kinematics continuity, and cross-validation based on multi-beam vector closure. This process identifies and removes outlier observations, resulting in preprocessed three-dimensional current velocity data. Specifically, firstly, various raw and synchronous observation data from the current meter are acquired simultaneously. The raw observation data includes: radial velocity directly measured by the acoustic beam (raw beam velocity), and raw diagnostic parameters generated by the signal processor to evaluate the quality of a single measurement. These raw diagnostic parameters include: pulse effectiveness reflecting the completeness of data acquisition, velocity estimation correlation characterizing the statistical reliability of velocity estimation, and inter-beam closure error velocity used to assess the internal consistency of the measurement. Synchronous observation data comes from other sensors, including but not limited to: carrier attitude data from the inertial measurement unit to determine the instantaneous spatial pointing of the beam, such as roll, pitch, and true heading angles; and environmental parameters from the temperature, salinity, and depth (TDM) sensor to calculate sound velocity, such as seawater temperature, salinity, and pressure. The data preprocessing process then proceeds as follows: First, coordinate system one is established. Based on the carrier attitude data, a rotation matrix from the instrument coordinate system to the stable geographic coordinate system is calculated using a direction cosine matrix or quaternion rotation method. This rotation matrix is ​​then used to correct the direction vector of each beam and re-invert it, thus obtaining the three-dimensional velocity components in the geographic coordinate system. Second, acoustic path and positioning corrections are performed. Based on environmental parameters, the sound velocity profile is calculated using the internationally accepted seawater state equation. Subsequently, based on Snell's law, iterative ray tracing is performed on the propagation path of each beam. This corrects the actual geometric distance and sound wave incident angle of each depth unit. Finally, the corrected parameters are used to recalculate the sound velocity-corrected three-dimensional flow velocity. The third step involves multi-level anomaly identification and removal: This cascaded filtering process includes: 1. Pulse-level outlier detection: Robust outlier detection is performed on pulse-level velocities using the median absolute deviation method. Pulses whose velocity values ​​deviate from the median by more than a set threshold are identified as outliers, such as the default 3 times the median absolute deviation; 2. Profile rationality verification: Based on the physical principle that fluid motion is continuous in the vertical direction, the vertical gradient of the velocity profile is verified, and abnormal depth layers with non-physical velocity peaks are removed; 3. Multi-beam consistency cross-validation: Internal consistency cross-validation is performed using the characteristics of multi-beam measurement. By analyzing the temporal or spatial distribution of the closure error velocity between beams, inconsistent observation units with continuously exceeding closure error limits are removed; Finally, the data that has passed the verification are integrated to obtain the preprocessed three-dimensional velocity data.

[0077] The present invention is further configured such that the quality inspection step includes:

[0078] Based on the original diagnostic parameters, the data reliability of each observation unit is evaluated, and corresponding basic quality labels are generated. Specifically, the input original diagnostic parameters include: pulse effectiveness, velocity estimation correlation, and inter-beam closure error velocity. The implementation process includes: first, comparing each parameter with multiple quality thresholds pre-set based on instrument performance and observation experience. The quality thresholds are determined based on the instrument's nominal performance and experience in typical observation environments. For example, the excellent threshold for velocity estimation correlation can be set to 0.7, and the acceptable threshold can be set to 0.5; the excellent threshold for inter-beam closure error velocity can be set to 0.05 m / s, and the acceptable threshold can be set to 0.15 m / s; the acceptable threshold for pulse effectiveness can be set to 70%. Subsequently, a pre-defined multi-parameter joint discrimination logic rule, such as using a rule engine or decision tree method, is used to comprehensively judge each observation unit point by point. The judgment logic is as follows: when the velocity estimation correlation and closure error velocity both reach the high standard of excellent threshold and the impulse effectiveness is qualified, it is marked as excellent; when the correlation and closure error velocity meet the basic qualified threshold, it is marked as qualified; when any parameter slightly deviates from the qualified threshold, it is marked as questionable; when the correlation or impulse effectiveness is lower than the minimum tolerance, or the closure error exceeds the maximum tolerance limit, it is marked as invalid. Finally, the step assigns a discrete basic quality label of "excellent", "qualified", "questionable" or "invalid" to each observation unit as independent quality dimension data, which is output synchronously with the preprocessed three-dimensional velocity data.

[0079] The present invention is further configured such that S2 includes: a spatial structure feature extraction step and a temporal dynamic feature extraction step;

[0080] The spatial structure feature extraction steps include:

[0081] Based on three-dimensional velocity data, the vertical coherence of velocity sequences at adjacent depth layers is calculated through cross-correlation analysis;

[0082] The vertical gradient of the three-dimensional flow velocity data is calculated using differential operations, and the stability of the vertical gradient in the time dimension is evaluated.

[0083] By integrating vertical coherence, vertical gradient stability, and basic quality labels, a first feature is generated to characterize the continuity and reliability of the profile structure. Specifically, the three-dimensional velocity data sequence in geographic coordinate system and the corresponding discrete quality level labels are used as input. 1. First, vertical coherence is calculated. For each pair of adjacent depth layers in the velocity profile, the eastward and northward velocity component sequences of the same time period are extracted. The Pearson correlation coefficient of the two sets of sequences is calculated to measure the degree of linear correlation. The Pearson correlation coefficient is a current technology, obtained by calculating the product of the covariance of the two sets of sequences and their respective standard deviations, which will not be elaborated here. Then, the Pearson correlation coefficients in the two directions are arithmetically averaged to obtain the vertical coherence coefficient characterizing the flow consistency of the adjacent layers. Finally, the average of the vertical coherence coefficients of all valid layer pairs (i.e., layer pairs whose corresponding data are not invalid) in the entire profile is calculated as the overall vertical coherence. 2. Next, calculate the vertical gradient and its temporal stability. Use the central difference method to calculate the magnitude of the velocity gradient vector at each time point for each depth layer along the depth direction. Then, for each depth layer, calculate the standard deviation and average value of all gradient values ​​within a continuous time window (such as one tidal cycle). By dividing the standard deviation by the average value, calculate the coefficient of variation of the gradient of that depth layer. The smaller the coefficient of variation, the more stable the gradient fluctuation of that layer is over time. Then, calculate the average value of the coefficients of variation of all depth layers as the average gradient coefficient of variation of the profile. In order to obtain a positive index characterizing "stability", the average gradient coefficient of variation is converted into a gradient stability index. The specific conversion formula is: Gradient stability index = 1 - average gradient coefficient of variation. 3. Finally, feature fusion is performed to generate the first feature. Based on the basic quality labels, the data involved in the calculation are cleaned and weighted. For example, invalid data points are excluded, and data labeled "suspicious" are given a lower weight when averaging, for example, a weight of 0.5. Then, a preset fusion rule is used to combine the average vertical coherence, average gradient stability, and the effective proportion of the overall data (i.e., the proportion of data with non-"invalid" labels). The preset fusion rule is, for example, to use a weighted summation method to assign weights to vertical coherence, the reciprocal of gradient stability (or 1 - stability value), and the effective proportion. The weight allocation scheme is 0.4, 0.4, and 0.2. Then, a scalar index between 0 and 1 is generated as the first feature, i.e., the spatial structure soundness index, through linear normalization. The higher the value of the first feature, the more sound and reliable the spatial structure of the current velocity profile.

[0084] The present invention is further configured such that the time-dynamic feature extraction step includes:

[0085] Adaptive multi-scale decomposition was performed on the velocity time series of each depth layer in the three-dimensional velocity data to obtain multiple intrinsic mode components.

[0086] The instantaneous frequency characteristics of each intrinsic mode component are obtained by Hilbert transform analysis, the corresponding energy distribution characteristics are obtained by variance calculation or power spectrum analysis, and the corresponding vertical propagation characteristics are obtained by calculating the cross-correlation coefficients of the same mode components in adjacent depth layers.

[0087] Based on frequency characteristics, energy distribution, and vertical propagation properties, the physical process categories corresponding to each intrinsic mode component are identified;

[0088] Evaluate the frequency stability, energy concentration, and modal orthogonality of modal components belonging to the same physical process category;

[0089] The second feature is generated by integrating frequency stability, energy concentration, modal orthogonality, and preset physical process energy weights. Specifically, the input consists of a 3D velocity data sequence in a geographic coordinate system and corresponding discrete quality level labels. 1. Adaptive multi-scale decomposition is performed. First, the velocity sequence of each depth layer is processed using an ensemble empirical mode decomposition method. Gaussian white noise with a specific standard deviation (e.g., 0.2 times the standard deviation of the 3D velocity data sequence) is added multiple times to the original 3D velocity data sequence. Empirical mode decomposition is performed on the sequence after each noise addition, and all decomposition results are ensemble averaged to suppress mode aliasing, ultimately obtaining a series of intrinsic mode components arranged from high to low frequency. 2. Modal feature extraction and physical process identification are performed. Three key features of each component are extracted: the instantaneous frequency feature is obtained by calculating its instantaneous phase using Hilbert transform and differentiating it with respect to time; the energy distribution feature is obtained by calculating the component variance; and the vertical propagation characteristics are obtained by calculating the Pearson correlation coefficient of the same index component in adjacent depth layers. Based on the three key features mentioned above, physical process identification is performed. The feature vector of each component is matched with a pre-established template of typical physical process features. The templates are defined according to oceanographic knowledge, such as high-frequency noise (frequency > 1 Hz, low energy, weak vertical correlation), surface waves (frequency 0.05-0.5 Hz, energy decays exponentially with depth, vertical correlation coefficient close to 1), and tidal currents (fixed frequency, such as M2 confluence 1.93 cpd, high energy, synchronous vertical changes across layers). Matching is achieved by calculating the Euclidean distance between the component features and the features of each template, and assigning the component to the physical process category corresponding to the template with the smallest distance. 3. Intra-class modal quality assessment: For all components within the same identified physical process category, the quality of the physical process category is assessed: the standard deviation of the instantaneous frequency sequence of each component is calculated; the smaller the standard deviation, the higher the frequency stability. The proportion of the energy of the largest (dominant) component in the physical process category to the total energy of the category is calculated; the higher the proportion, the higher the energy concentration. The absolute value of the cross-correlation coefficient between any two different component time series within the category is calculated; the lower the absolute value, the better the modal orthogonality. 4. Generate a second feature by calculating a sharpness sub-score for each identified physical process. For example, the evaluation results of three indicators—frequency stability, energy concentration, and modal orthogonality—of a physical process (normalized to the 0-1 range) are weighted and averaged. The weights are preset to default values ​​of 0.4, 0.3, and 0.3, respectively. The resulting average is the sharpness sub-score for the physical process. Then, weights are assigned to each physical process based on its energy proportion to the total energy of all identified processes (for example, if tidal energy accounts for 60% and wave energy accounts for 30%, the weights are 0.6 and 0.3, respectively). The sharpness sub-score of each process is multiplied by its energy weight and summed to obtain a preliminary comprehensive value.Furthermore, the initial composite value is corrected (e.g., by multiplication) by combining the overall effective proportion of data reflected by the basic quality labels (such as the proportion of "high-quality" and "qualified" data to the total data). Finally, the corrected value is mapped to a closed interval of 0 to 1 through linear normalization, and the final output value is used as the second feature, namely the temporal dynamic clarity index. The higher the second feature, the clearer and easier it is to separate different physical processes in the flow velocity signal.

[0090] The present invention is further configured such that the flow field parameter set includes: velocity profile, tidal current component, and turbulence parameters, wherein the turbulence parameters include: turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate. Specifically, the velocity profile refers to the final output vertical distribution of water flow velocity in the geographic coordinate system after a series of processing steps such as physical benchmark correction, mass filtering, and feature-guided weighted fusion. It is a two-dimensional data array containing three velocity components: east, north, and vertical, which intuitively reflects the changes in ocean current velocity and direction with depth and time. The tidal current component refers to the deterministic periodic flow portion generated by astronomical tidal forces separated from the original mixed velocity time series. It is usually characterized in the form of the harmonic constant of the main astronomical tidal constituents, including: amplitude, phase, ellipticity, and the direction of the principal axis of the ellipse. It is used for high-precision tidal current forecasting to ensure navigation and operational safety, and helps to remove deterministic signals from observation data, thereby revealing non-tidal dynamic processes such as wind-driven currents more clearly. Turbulence parameters specifically refer to a set of physical quantities used to quantify random small-scale fluctuations in ocean currents, including: turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate. Turbulence intensity is the ratio of the standard deviation of turbulent fluctuation velocities to the time-averaged flow velocity, characterizing the degree of turbulence and mixing potential. Reynolds stress is the negative value of the covariance of fluctuation velocities in different directions, reflecting the momentum flux or shear stress caused by turbulence. Turbulent kinetic energy dissipation rate is the rate at which turbulent kinetic energy per unit mass of fluid is converted into heat energy due to viscosity. It is usually estimated by analyzing the characteristic slope of the turbulent energy spectrum in the inertial subregion and is a key physical quantity for directly quantifying ocean mixing intensity and energy dissipation. These three turbulence parameters collectively characterize the details of turbulent motion driving ocean mixing and energy cascades.

[0091] The present invention is further configured such that S3 includes: a flow field parameter optimization step and an uncertainty quantification step;

[0092] The flow field parameter optimization steps include:

[0093] Using the first feature as the weight, the three-dimensional velocity data is vertically weighted and smoothed to reconstruct the optimized velocity profile.

[0094] Based on the temporal dynamic clarity indicated by the second feature, multi-scale filtering is performed on the time series of each depth layer in the three-dimensional flow velocity data to separate the turbulent components.

[0095] Based on the turbulent components, the turbulence intensity is obtained by calculating the variance of the corresponding fluctuating velocity, and the Reynolds stress is obtained by calculating the covariance of the fluctuating velocity in different directions.

[0096] Based on the turbulent components, the power spectrum is estimated to obtain the energy spectrum, and the slope is fitted in the inertial subregion based on the energy spectrum to obtain the turbulent kinetic energy dissipation rate.

[0097] Turbulence parameters are constructed by combining turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate. Specifically, the flow field parameter optimization steps, guided by the first and second features, fuse and separate the three-dimensional velocity data to generate optimized velocity profiles, tidal current components, and turbulence parameters. First, the three-dimensional velocity data is reconstructed using vertical weighted smoothing with the first feature, namely the spatial structure soundness index, as weights. Specifically, the distribution of the first feature along the depth direction is normalized into weight coefficients, and the velocity values ​​of each depth layer in the original velocity profile are weighted and averaged. Depth layers with higher weights occupy a larger proportion in the reconstruction, thereby suppressing abnormal jumps in structural discontinuities, enhancing the vertical continuity of the profile, and ultimately outputting an optimized velocity profile with a more reasonable and smoother physical structure. Second, based on the signal separability indicated by the second feature, namely the temporal dynamic clarity index, adaptive multi-scale filtering is performed on the velocity time series of each depth layer. When the second eigenvalue is high, a finely divided filter group (such as a set of Butterworth bandpass filters with well-defined passband boundaries) is used to extract the tidal current frequency band and the turbulent high-frequency band signals respectively. When the second eigenvalue is low, a wider passband and stronger smoothing are used to ensure robust separation. After the above processing, the periodic tidal current component and the randomly fluctuating turbulent component are separated from the original sequence. Then, turbulence parameters are calculated based on the separated turbulent components. The turbulence intensity is specifically obtained by calculating the variance of the turbulent fluctuating velocity sequence, and the square root of the variance divided by the time-averaged velocity is the relative turbulence intensity. The Reynolds stress is obtained by calculating the covariance of the fluctuating velocity sequences in different directions (such as east and vertical, north and vertical), where the negative value is the corresponding Reynolds stress component. The turbulent kinetic energy dissipation rate is estimated by calculating the power spectrum of the turbulent components: First, the Welch method is used to estimate the power spectrum of the turbulent velocity sequence to obtain the energy spectrum. Then, frequency bands conforming to the inertial subregion theory are identified in double logarithmic coordinates. The energy spectrum curve of this frequency band is fitted with least-squares linearly, and the slope obtained from the fitting is substituted into the Kolmogorov formula to calculate the turbulent kinetic energy dissipation rate. Finally, the calculated turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate are integrated into a turbulence parameter set.

[0098] The present invention is further configured such that the uncertainty quantification step includes:

[0099] Error source information includes: carrier attitude error introduced by the inertial measurement unit during the acquisition of carrier attitude data, and sound velocity profile error introduced by sensor accuracy or environmental disturbance during the inversion of sound velocity profile.

[0100] Using basic quality labels, error source information, and preset instrument inherent errors as inputs, the uncertainties of velocity profiles, tidal current components, and turbulence parameters are calculated through an error propagation model. Specifically, the uncertainty quantification step generates quantified confidence intervals for velocity profiles, tidal current components, and turbulence parameters by integrating data quality and various measurement errors. Input parameters include: basic quality labels generated by the quality inspection step, error source information characterizing system errors, and instrument inherent errors provided by the current meter manufacturer. Error source information specifically includes: carrier attitude errors obtained from the inertial measurement unit specifications, such as roll and pitch angle measurement errors of ±0.1 degrees; and sound velocity profile errors obtained from sensor accuracy and uncertainty assessment of the inversion model. The specific implementation process is as follows: First, discrete basic quality label levels are mapped to continuous quality correction coefficients, for example, 1.0 for excellent, 1.2 for acceptable, and 1.5 for questionable, and local or global quality correction factors are calculated accordingly for different output parameters. Subsequently, based on the root sum-of-squares synthesis rule of error propagation, the expanded uncertainties of the three types of outputs are calculated respectively. For the velocity profile, the inherent errors of the calculation instrument, the velocity errors introduced by attitude errors through the partial derivative relationship of coordinate rotation, and the velocity errors introduced by sound velocity errors through the ray-tracking model are synthesized and multiplied by the local mass correction factor for the corresponding depth layer to obtain the expanded uncertainty of each velocity component. For the tidal current component, the average uncertainty of the original time-series data used to extract the tidal current is mainly evaluated by combining the model residuals or result dispersion of the tidal current separation algorithm itself. The two are synthesized and multiplied by the global mass correction factor for the analysis period to obtain the uncertainty of the tidal current amplitude and phase. For turbulence parameters, the uncertainties of turbulence intensity and Reynolds stress are estimated based on the statistical sampling error theory. For example, the uncertainty of turbulence intensity can be approximated by dividing its estimated value by twice the square root of the number of independent samples of fluctuating velocity. The uncertainty of turbulent kinetic energy dissipation rate is evaluated by analyzing the residuals and standard errors of the linear fitting of the energy spectrum slope of the inertial subregion. These statistical and fitting errors are synthesized with the uncertainties of the original turbulence data and then multiplied by the global mass correction factor for the calculation period to finally obtain the expanded uncertainty of each turbulence parameter.

[0101] The present invention is further configured such that S4 includes:

[0102] A comprehensive score is obtained by weighted fusion of the first feature, the second feature, and the uncertainty.

[0103] Based on a preset threshold range, the comprehensive score is mapped to discrete quality levels to generate the final quality label;

[0104] The final quality label and uncertainty are associated with the optimized flow field parameter set and packaged into a standardized data product. Specifically, using the first feature (spatial structure soundness index), the second feature (temporal dynamic clarity index), and the uncertainty of the flow field parameters as inputs, a final quality label is generated through weighted fusion and level mapping, and then packaged into a standard data product. The implementation process is as follows: First, a weighted fusion calculation is performed, directly using the first and second features. At the same time, the overall uncertainty index, such as the average relative uncertainty of the velocity profile, is normalized to the range of 0 to 1. A linear weighted sum is then performed using preset weights. For example, a weight of 0.4 is assigned to the first feature, a weight of 0.4 is assigned to the second feature, and the overall uncertainty level is converted into a positive index and then assigned a weight of 0.2, resulting in a comprehensive score between 0 and 1. Subsequently, continuous scores are mapped to discrete quality levels based on preset threshold ranges. For example, a comprehensive score greater than or equal to 0.85 corresponds to Grade A (Excellent), between 0.70 and 0.85 to Grade B (Good), between 0.50 and 0.70 to Grade C (Medium), and below 0.50 to Grade D (Poor). Based on this, a final quality label is generated. Finally, correlation and standardization encapsulation are performed. The optimized velocity profile, tidal current components, and turbulence parameters are used as core data variables. The corresponding expanded uncertainty value associated with each variable is used as an attribute. The final quality label, along with the processing method version, processing timestamp, and various characteristic indices and comprehensive score values, are written into a NetCDF format file conforming to the CF convention. This outputs a self-describing standardized data product that integrates high-confidence flow field data, quantitative confidence assessment, and clear quality conclusions.

[0105] Example 2:

[0106] Please see Figure 2 This exemplary multi-scale time-frequency analysis current meter signal processing system includes:

[0107] Data calibration and quality control module: Collects raw observation data from the current meter, performs physical benchmark calibration, and obtains three-dimensional current velocity data and basic mass labels;

[0108] Spatiotemporal feature profiling module: performs spatiotemporal decoupling feature extraction on 3D flow velocity data: generates the first feature in the spatial dimension based on the characteristics of the cross-sectional structure, and generates the second feature in the temporal dimension based on the characteristics of the multi-scale decomposition of the signal;

[0109] Intelligent fusion and credibility generation module: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information.

[0110] Product Packaging and Quality Reporting Module: Performs comprehensive quality assessment and classification based on the first feature, the second feature, and uncertainty, and outputs data products with quality markings and uncertainty information.

[0111] It should be noted that the multi-scale time-frequency analysis current meter signal processing system provided in the above embodiments and the multi-scale time-frequency analysis current meter signal processing method provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the multi-scale time-frequency analysis current meter signal processing system provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for processing ocean current meter signals using multi-scale time-frequency analysis, characterized in that, include: S1: Collect raw observation data from the current meter, perform physical benchmark correction, and obtain three-dimensional current velocity data and basic mass labels; S2: Spatiotemporal decoupling feature extraction of 3D flow velocity data: A first feature is generated in the spatial dimension based on the profile structure characteristics, and a second feature is generated in the temporal dimension based on the signal multi-scale decomposition characteristics. S2 includes a spatial structure feature extraction step and a temporal dynamic feature extraction step. The spatial structure feature extraction step includes: calculating the vertical coherence of flow velocity sequences at adjacent depth layers based on the 3D flow velocity data through cross-correlation analysis; calculating the vertical gradient of the 3D flow velocity data through differential operations and evaluating the stability of the vertical gradient in the temporal dimension; and generating a first feature to characterize the continuity and reliability of the profile structure by integrating vertical coherence, vertical gradient stability, and basic quality labels. The temporal dynamic feature extraction step... The steps include: performing adaptive multi-scale decomposition on the velocity time series of each depth layer in the three-dimensional velocity data to obtain multiple intrinsic modal components; obtaining the instantaneous frequency characteristics of each intrinsic modal component through Hilbert transform analysis, obtaining the corresponding energy distribution characteristics through variance calculation or power spectrum analysis, and obtaining the corresponding vertical propagation characteristics by calculating the cross-correlation coefficients of the same modal components in adjacent depth layers; identifying the physical process category corresponding to each intrinsic modal component based on frequency characteristics, energy distribution, and vertical propagation characteristics; evaluating the frequency stability, energy concentration, and modal orthogonality of modal components belonging to the same physical process category; and generating a second feature by combining frequency stability, energy concentration, modal orthogonality, and preset physical process energy weights. S3: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information. S4: Perform comprehensive quality assessment and classification based on the first feature, the second feature, and the uncertainty, and output a data product with quality markings and uncertainty information.

2. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 1, characterized in that, S1 includes: a data correction step and a quality inspection step; The data correction steps include: Acquire raw observation data from the current meter, including raw beam velocity, raw diagnostic parameters, and synchronous observation data from auxiliary sensors. The raw diagnostic parameters include pulse effectiveness, velocity estimation correlation, and inter-beam closure error velocity. The synchronous observation data includes carrier attitude data and environmental parameters. Based on the carrier attitude data, the original beam velocity is transformed from the instrument coordinate system to the geographic coordinate system through coordinate rotation; Based on the inversion of sound velocity profiles using environmental parameters, the sound wave propagation path and depth positioning are corrected using a sound ray tracing algorithm based on the sound velocity profiles, the corrected beam incidence angle is obtained, and the three-dimensional flow velocity is recalculated. For the data after coordinate transformation and sound velocity correction, pulse-level outlier detection based on statistical distribution, profile rationality check based on fluid kinematic continuity, and cross-validation based on multi-beam vector closure are performed sequentially to identify and remove outlier observations, thus obtaining preprocessed three-dimensional flow velocity data.

3. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 2, characterized in that, The quality inspection steps include: Based on the original diagnostic parameters, the data reliability of each observation unit is evaluated, and corresponding basic quality labels are generated.

4. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 1, characterized in that, The flow field parameter set includes: velocity profile, tidal current component and turbulence parameters, wherein the turbulence parameters include: turbulence intensity, Reynolds stress and turbulent kinetic energy dissipation rate.

5. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 4, characterized in that, S3 includes: a flow field parameter optimization step and an uncertainty quantification step; The flow field parameter optimization steps include: Using the first feature as the weight, the three-dimensional velocity data is vertically weighted and smoothed to reconstruct the optimized velocity profile. Based on the temporal dynamic clarity indicated by the second feature, multi-scale filtering is performed on the time series of each depth layer in the three-dimensional flow velocity data to separate the turbulent components. Based on the turbulent components, the turbulence intensity is obtained by calculating the variance of the corresponding fluctuating velocity, and the Reynolds stress is obtained by calculating the covariance of the fluctuating velocity in different directions. Based on the turbulent components, the power spectrum is estimated to obtain the energy spectrum, and the slope is fitted in the inertial subregion based on the energy spectrum to obtain the turbulent kinetic energy dissipation rate. Turbulence parameters are constructed by combining turbulence intensity, Reynolds stress, and turbulent kinetic energy dissipation rate.

6. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 5, characterized in that, The uncertainty quantification step includes: Error source information includes: carrier attitude error introduced by the inertial measurement unit during the acquisition of carrier attitude data, and sound velocity profile error introduced by sensor accuracy or environmental disturbance during the inversion of sound velocity profile. Using basic quality labels, error source information, and preset instrument inherent errors as inputs, the uncertainties of velocity profiles, tidal current components, and turbulence parameters are calculated respectively through an error propagation model.

7. The method for processing ocean current meter signals using multi-scale time-frequency analysis according to claim 1, characterized in that, S4 includes: A comprehensive score is obtained by weighted fusion of the first feature, the second feature, and the uncertainty. Based on a preset threshold range, the comprehensive score is mapped to discrete quality levels to generate the final quality label; The final quality label and uncertainty are correlated with the optimized flow field parameter set and packaged into a standardized data product.

8. A multi-scale time-frequency analysis current meter signal processing system, used to implement the multi-scale time-frequency analysis current meter signal processing method according to any one of claims 1-7, characterized in that, include: Data calibration and quality control module: Collects raw observation data from the current meter, performs physical benchmark calibration, and obtains three-dimensional current velocity data and basic mass labels; Spatiotemporal feature profiling module: performs spatiotemporal decoupling feature extraction on 3D flow velocity data: generates the first feature in the spatial dimension based on the characteristics of the cross-sectional structure, and generates the second feature in the temporal dimension based on the characteristics of the multi-scale decomposition of the signal; Intelligent fusion and credibility generation module: Based on the first and second features, the three-dimensional flow velocity data is fused and separated to obtain an optimized set of flow field parameters. The uncertainty of each parameter is generated by combining the basic mass label and error source information. Product Packaging and Quality Reporting Module: Performs comprehensive quality assessment and classification based on the first feature, the second feature, and uncertainty, and outputs data products with quality markings and uncertainty information.

Images