Sound speed profile reconstruction method and apparatus, device
By combining a nonlinear regression model and an autoregressive moving average model for sound velocity profile prediction, the failure of sound velocity profile reconstruction in existing technologies when crossing acoustic lattices is solved, and more accurate sound velocity profile prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING ZHONGAN INTELLIGENT INFORMATION TECH CO LTD
- Filing Date
- 2022-11-03
- Publication Date
- 2026-07-03
AI Technical Summary
Existing sound velocity profile reconstruction methods fail when crossing acoustic lattices and do not adequately consider time-varying factors, resulting in low accuracy of sound velocity profile results.
A sound velocity profile prediction method based on nonlinear regression and autoregressive moving average models is adopted, which combines the average state information and wave dynamic information of the data for prediction, and the accuracy is improved through a correction step.
It improves the accuracy of sound velocity profile prediction results, can more comprehensively consider time-varying factors, and enhances the accuracy of prediction.
Smart Images
Figure CN115828733B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of marine data analysis and processing technology, and in particular to a method, apparatus, and equipment for reconstructing sound velocity profiles. Background Technology
[0002] With the development and scaling up of technology, precise ocean acoustic field analysis is widely used in civilian and military fields. The sound velocity profile is one of the important waveguide elements for ocean acoustic field analysis and calculation. Therefore, constructing an accurate, large-scale, full-ocean-depth sound velocity profile has important theoretical and practical significance.
[0003] Currently, commonly used methods for sound velocity profile reconstruction mainly include mathematical extrapolation and the empirical orthogonal function (EOF) method. Extrapolation works well for regions with small sound velocity gradient changes and shallow extension depths; however, it cannot absorb effective information about acoustic strata and often fails when applied across acoustic strata. The empirical orthogonal function method, by combining multi-order data, can reflect the nonlinear changes in the sound velocity profile and has a significant advantage for reconstructing sound velocity profiles across strata. However, this method is constrained by the location of the maximum measured sound velocity depth, requiring additional methods to construct a full-depth sound velocity profile, and it does not fully consider the influence of time variations, resulting in insufficient accuracy of the obtained sound velocity profile results. Summary of the Invention
[0004] In view of this, this disclosure proposes a sound velocity profile reconstruction method, which can effectively improve the accuracy of sound velocity profile results.
[0005] According to one aspect of this disclosure, a method for reconstructing sound velocity profiles is provided, comprising:
[0006] Obtain the geographic location, topographic information, and time information of the target sea area currently being collected;
[0007] The geographic location information, the terrain information, and the time information are used as input data and input into a pre-built sound velocity profile prediction model. The sound velocity profile prediction model predicts the sound velocity profile based on the input data to obtain the corresponding predicted sound velocity profile.
[0008] The sound velocity profile prediction model is constructed based on a nonlinear regression model and an autoregressive moving average model.
[0009] In one possible implementation, the geographic location information includes the longitude and latitude of the target sea area, and the topographic information includes the elevation of the target sea area.
[0010] In one possible implementation, when the sound velocity profile prediction model obtains the predicted sound velocity profile based on the input data, it includes:
[0011] The nonlinear regression model is invoked to calculate the spatial mean estimate based on the input data, and the autoregressive moving average model is invoked to calculate the time series fluctuation estimate based on the input data.
[0012] The spatial mean state estimate and the time series fluctuation estimate are superimposed to obtain the predicted sound velocity profile.
[0013] In one possible implementation, after the predicted sound velocity profile is obtained by the sound velocity profile prediction model based on the input data, the step of correcting the predicted sound velocity profile is further included.
[0014] In one possible implementation, the predicted sound velocity profile is corrected based on the obtained measured sound velocity profile.
[0015] In one possible implementation, correcting the predicted sound velocity profile based on the obtained measured sound velocity profile includes:
[0016] Determine the overlapping region between the measured sound velocity profile and the predicted sound velocity profile, and update the data in the overlapping region of the predicted sound velocity profile to the measured sound velocity profile in the overlapping region.
[0017] Calculate the data error between the measured sound velocity profile and the predicted sound velocity profile in the non-overlapping region, and correct the predicted sound velocity profile in the non-overlapping region based on the data error.
[0018] In one possible implementation, when correcting the predicted sound velocity profile in the non-overlapping region based on the data error, the average error between the measured sound velocity profile and the predicted sound velocity profile in the overlapping region is used as the initial correction value, and the correction value is decreased sequentially according to the increasing region depth.
[0019] According to another aspect of this disclosure, a sound velocity profile reconstruction device is also provided, including an information acquisition module and a sound velocity profile prediction module.
[0020] The information acquisition module is configured to acquire the geographic location information, topographic information, and data acquisition time of the currently collected target sea area;
[0021] The sound velocity profile prediction module is configured to take the geographical location information, the terrain information and the data acquisition time as input data and input them into a pre-built sound velocity profile prediction model. The sound velocity profile prediction model then performs sound velocity profile prediction based on the input data to obtain the corresponding predicted sound velocity profile.
[0022] The sound velocity profile prediction model is constructed based on a nonlinear regression model and an autoregressive moving average model.
[0023] One possible implementation also includes a data correction module;
[0024] The data correction module is configured to correct the predicted sound velocity profile after the sound velocity profile prediction model obtains the predicted sound velocity profile based on the input data.
[0025] According to one aspect of this disclosure, a sound velocity profile reconstruction device is also provided, comprising:
[0026] processor;
[0027] Memory used to store processor-executable instructions;
[0028] The processor is configured to implement any of the preceding methods when executing the executable instructions.
[0029] By constructing a sound velocity profile prediction model using the NLR and ARMA models, and reconstructing the sound velocity profile of the target sea area based on the constructed sound velocity profile prediction model, the obtained sound velocity profile model can simultaneously combine the average state information and wave dynamic information of the data to predict the sound velocity profile. This means that the method of this application fully considers the time variation factor when predicting the sound velocity profile, thereby making the factors considered when predicting the sound velocity profile of the target sea area more comprehensive, which effectively improves the accuracy of the sound velocity profile prediction results.
[0030] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description
[0031] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.
[0032] Figure 1 A flowchart of a sound velocity profile reconstruction method according to an embodiment of this application is shown;
[0033] Figure 2 This diagram illustrates the process of calculating the predicted sound velocity profile based on the constructed sound velocity profile prediction model in a sound velocity profile reconstruction method according to an embodiment of this application.
[0034] Figure 3 This document illustrates a flowchart of the sound velocity profile reconstruction method according to an embodiment of this application, during the construction of a sound velocity profile prediction model.
[0035] Figure 4 A flowchart illustrating another embodiment of the sound velocity profile reconstruction method of this application is shown;
[0036] Figure 5 This diagram shows a structural block diagram of a sound velocity profile reconstruction device according to an embodiment of this application;
[0037] Figure 6 A structural block diagram of a sound velocity profile reconstruction device according to an embodiment of this application is shown. Detailed Implementation
[0038] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.
[0039] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.
[0040] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.
[0041] First, it should be noted that the sound velocity profile reconstruction method provided in this application is mainly used for data analysis in the marine field, especially for the analysis of the marine sound field. After the sound velocity profile of the target sea area is obtained by predicting the marine sound velocity profile using the method in this application, the obtained sound velocity profile results are then applied to the marine sound field analysis.
[0042] Example 1
[0043] Figure 1 A flowchart illustrating a sound velocity profile reconstruction method according to an embodiment of the present disclosure is shown. Figure 1 As shown, the method includes: Step S100, acquiring the geographical location information, topographic information, and time information of the currently collected target sea area. Here, it should be explained that the acquired geographical location information of the target sea area refers to the geographical location of the point in the sea area where the sound velocity profile prediction is currently being performed, which can be characterized by the longitude and latitude of that point. The acquired topographic information of the target sea area can be characterized by the elevation of the point in the target sea area. The time information refers to the time node information of the current sound velocity profile prediction of the target sea area.
[0044] After obtaining the above information, step S200 involves inputting the obtained geographic location information, terrain information, and time information as input data into a pre-built sound velocity profile prediction model. The sound velocity profile prediction model then predicts the sound velocity profile based on the input data, yielding the corresponding predicted sound velocity profile. It should be noted that in the method of this embodiment, the pre-built sound velocity profile prediction model is constructed based on a nonlinear regression model (NLR model) and an autoregressive moving average model (ARMA model).
[0045] The NLR model is trained using time-series data as the training dataset to extract wave dynamics information, while the ARMA model is trained using average continuous time-series data with a preset time period as the training dataset to extract average state information. Therefore, after constructing the sound velocity profile prediction model using both the NLR and ARMA models, the resulting sound velocity profile model can simultaneously combine the average state information and wave dynamics information of the data to predict the sound velocity profile. This allows the method in this embodiment to fully consider time-varying factors when predicting the sound velocity profile, resulting in a more comprehensive consideration of factors when predicting the sound velocity profile at target sea areas, thus effectively improving the accuracy of the sound velocity profile prediction results.
[0046] In one possible implementation, when obtaining the predicted sound velocity profile based on the sound velocity profile prediction model built on the nonlinear regression model and the autoregressive moving average model, it can be achieved in the following way.
[0047] In other words, firstly, a linear regression model is invoked to calculate the spatial mean estimate based on the input data, and then an autoregressive moving average model is invoked to calculate the time series fluctuation estimate based on the input data. Finally, the calculated spatial mean estimate and the time series fluctuation estimate are superimposed to obtain the predicted sound velocity profile.
[0048] For details, see Figure 2As shown, when constructing sound velocity profile prediction models using the NLR model and the ARMA model respectively, after obtaining the geographical location information (including longitude and latitude), topographic information (including elevation), and time information (including month) of the target sea area, the above information is input into the sound velocity profile prediction model. The sound velocity profile prediction model calls the NLR model and the ARMA model respectively. The NLR model processes the above information to obtain the corresponding spatial mean estimate, and the ARMA model processes the above information to obtain the time series fluctuation estimate. Then, the sound velocity profile prediction model superimposes the obtained spatial mean estimate and time series fluctuation estimate, and the superimposed result is the predicted sound velocity profile result.
[0049] Specifically, after the user inputs information such as longitude, latitude, time, and elevation, the sound velocity profile prediction model calls the NLR model and the ARMA model respectively to calculate the spatial mean state estimate and the time series fluctuation estimate. For the NLR model, when calculating the spatial mean state estimate based on the input information, it can select a matching sub-model from 12*n sub-models to provide the spatial mean state estimate of the sound velocity profile. For the ARMA model, when calculating the time series fluctuation estimate based on the input information, it selects a matching sub-model from 2*n sub-models to provide the time series fluctuation estimate of the sound velocity profile.
[0050] More specifically, when the NLR model selects a matching sub-model from 12*n sub-models to calculate the spatial mean estimate, each of the 12*n sub-models corresponds to a different month and a corresponding depth layer. That is, the sub-models included in the NLR model are: sub-NLR(n, i), where n represents the depth layer corresponding to the sub-model, and i represents the time (e.g., month) corresponding to the sub-model. It should also be noted that each month corresponds to multiple sub-models with different depth layers.
[0051] When determining the sub-models that match the input information, matching sub-models can be selected based on the input elevation and time information. In one possible implementation, sub-models corresponding to the time period can be selected first, and then all sub-models with a depth layer less than or equal to the elevation information can be selected from the sub-models with the corresponding time period based on the elevation information. Alternatively, sub-models with a depth layer less than or equal to the elevation information can be selected first, and then all sub-models with a depth layer less than or equal to the elevation information can be selected from the sub-models with a depth layer less than or equal to the elevation information based on the time information to match the input time information.
[0052] For example, if the elevation information is 100m and the time is July, then when calculating the spatial mean value, the sub-models selected by the NLR model should include all sub-models with the month being July and the depth layer being less than or equal to 100m.
[0053] Correspondingly, when calculating the time series volatility estimate, it should be noted that the sub-models included in the ARMA model are sub-ARMA(n, j). Here, n also represents the depth layer corresponding to the sub-model, and j corresponds to the time period series.
[0054] When determining the ARMA sub-model that matches the input information, a matching sub-model can also be selected based on the input elevation and time information. It should be noted that when selecting the matching ARMA sub-model based on the input elevation and time information, it is necessary to first calculate the monthly average value of the data for different time periods, and then determine the sub-model based on the calculation results.
[0055] For example, if the elevation information is 100m and the time is July, then when calculating the estimated value of the fluctuation of this time series, the sub-models selected by the ARMA model should include all sub-models (random noise and periodic variations) with a depth layer less than or equal to 100m and a time period of July.
[0056] Furthermore, in the method of this application embodiment, when building the sound velocity profile prediction model based on the NLR model and the ARMA model, the NLR model and the ARMA model are first trained separately, and then the trained NLR model and the ARMA model are fused to build the sound velocity profile prediction model.
[0057] Among them, see Figure 3 As shown, the dataset used for training the NLR and ARMA models can be historical HYCOM temperature-salinity assimilation data. That is, historical HYCOM temperature-salinity assimilation data is used as the original dataset source, and this original dataset source is split into two parts: one part is used as the test dataset for subsequent model performance evaluation, and the other part is used as the training dataset for model training.
[0058] In one possible implementation, the original dataset source can be split into 20% and 80% portions. 20% of the data is used as the test dataset, and 80% is used as the training dataset. After splitting the original dataset source, the NLR and ARMA models can be trained using the determined training dataset.
[0059] For details, see Figure 3 Since the training dataset used is based on daily data and contains considerable noise, it needs to be preprocessed to remove noise before training the NLR and ARMA models using the segmented training dataset. One possible approach is to use smoothing to remove noise. Specifically, the monthly average of the three-dimensional temperature and salinity is first calculated, and then the data in the training dataset is smoothed based on the calculated monthly average.
[0060] Next, the mean state information and wave dynamics information are extracted and calculated from the training dataset. It should be noted that the mean state information refers to the monthly average value of the three-dimensional temperature and salinity data in the training dataset. That is, there are m monthly average continuous time series three-dimensional temperature and salinity data as the training set A, used for training the ARMA model.
[0061] Wave dynamics information refers to the fluctuation of the three-dimensional temperature and salinity data in the training dataset over time based on the monthly average. That is, the three-dimensional temperature and salinity data in the training dataset is divided into blocks according to 1-12 months, and then the average value of the three-dimensional temperature and salinity data for each month is calculated. The data obtained by dividing the time series data into blocks by month and then averaging is used as the training set B to train the NLR model.
[0062] Furthermore, when training the NLR and ARMA models, it is necessary to spatially and temporally discretize the previously calculated and extracted three-dimensional monthly average data, so that the discretized data can be used as the final sample data for model training.
[0063] Specifically, when discretizing the training dataset, the depth space of the data is first discretized to obtain multiple depth layers. Then, the time series of the data is decomposed according to the depth layers to extract the period and fluctuation of the time series.
[0064] In one possible implementation, when discretizing the training dataset in depth space, the depth layers can be segmented according to the "dense at the top, sparse at the bottom" principle based on the temperature and salinity fluctuation characteristics. For example, by segmenting the training dataset into depth layers, n depth layers can be obtained, namely depth layer 1, depth layer 2, depth layer 3, ..., depth layer n. In one possible implementation, when segmenting the aggregated training dataset into depth layers, the depth difference between any two adjacent depth layers can be set to [1 meter, 50 meters]. Preferably, it can be set to 50 meters, that is, every 50 meters constitutes one depth layer. Then, the time series of data in each depth layer is decomposed to extract the period and fluctuation of the time series of each depth layer. Finally, the extracted period and fluctuation are used for training and learning.
[0065] The NLR model training involves constructing Nth-order nonlinear spatial features based on the depth, latitude, and longitude of the temperature-salinity dataset. Machine learning is then used to train the NLR model, which consists of 12*n sub-models. One possible implementation is to construct second-order nonlinear spatial features for training the NLR model. Models constructed using second-order nonlinear spatial features exhibit better robustness; higher-order features can lead to overfitting.
[0066] For training the ARMA model, the stationarity of time series periodic data and fluctuation data is tested to learn and construct the ARMA model. The constructed ARMA model consists of 2*n sub-models.
[0067] After training the NLR and ARMA models using any of the above methods, the trained NLR and ARMA models can be fused and combined with empirical formulas for sound speed to construct a sound speed prediction model. The inputs to the constructed sound speed prediction model are longitude, latitude, elevation, and time information, including year and month. The output is the predicted sound speed at full ocean depth and the corresponding depth.
[0068] Furthermore, in the method of this application embodiment, after constructing the aforementioned sound velocity prediction model, in order to improve the accuracy of the sound velocity prediction model, one possible implementation also includes a test correction operation on the sound velocity prediction model. That is, using the test set segmented from the original data source set as test data, the constructed sound velocity prediction model is tested for model performance, and the model with the best robustness is selected as the optimal sound velocity profile prediction model (i.e., the sound velocity profile prediction model ultimately used in sound velocity prediction).
[0069] Specifically, when using the test dataset to correct the constructed sound speed prediction module, the test data in the test dataset can be compared with the sound speed profile results output by the sound speed prediction model. The sound speed prediction model can be corrected based on the error results, and the model with the best robustness can be selected as the final sound speed profile prediction model.
[0070] In addition, in one possible implementation, see [link to relevant documentation]. Figure 4 To further improve the accuracy of the sound velocity profile results, the method in this embodiment of the application also includes an operation of correcting the predicted sound velocity profile obtained by the sound velocity profile prediction model based on the input data.
[0071] Specifically, when correcting the predicted sound velocity profile, correction can be performed based on the obtained measured sound velocity profile. In one possible implementation, before correcting the obtained predicted sound velocity profile, a determination is made as to whether correction is necessary. If correction is deemed appropriate, the correction step is executed. If no correction is deemed necessary, the result is directly output. It should be noted that the measured sound velocity profile used for correction refers to a non-full-ocean-depth sound velocity profile. Specifically, those skilled in the art will understand that the measured sound velocity profile refers to the actually detected ocean depth and corresponding sound velocity data.
[0072] The determination of whether the currently output predicted sound velocity profile can be corrected can be based on whether a measured sound velocity profile for correction exists. If no measured sound velocity profile for correction exists, the entire ocean depth and the predicted sound velocity profile are directly output. If a measured sound velocity profile for correction exists, correction of the predicted sound velocity profile based on the measured sound velocity profile is performed.
[0073] In one possible implementation, when correcting the predicted sound velocity profile based on the measured sound velocity profile, it is mainly based on the existence of an overlapping region between the measured and predicted sound velocity profiles. That is, firstly, the measured sound velocity profile is obtained, and then the overlapping region between the measured and predicted sound velocity profiles is determined.
[0074] After determining the overlapping region between the measured and predicted sound velocity profiles, the data can be directly updated within this region. That is, the predicted sound velocity profile result for the overlapping region is directly updated to the corresponding measured sound velocity profile result. For example, for a certain overlapping region A, if the measured sound velocity profile result is V1 and the predicted sound velocity profile result is V2, then the sound velocity profile result for overlapping region A can be directly updated from V2 to V1.
[0075] For the correction of the predicted sound velocity profile in non-overlapping regions, the data error between the measured and predicted sound velocity profiles in the non-overlapping regions is calculated, and the predicted sound velocity profile in the non-overlapping regions is corrected based on this data error. In one possible implementation, when correcting the predicted sound velocity profile in the non-overlapping regions based on the data error, the average error between the measured and predicted sound velocity profiles in the overlapping regions is used as the initial correction value, and the correction value is decreased sequentially according to the increasing region depth.
[0076] It should be noted here that the overlapping region refers to the depth range where the measured sound velocity and the predicted sound velocity overlap. Generally, there is usually only one overlapping region. It should also be pointed out that when updating the correction value by sequentially decreasing it according to the increasing depth of the region, the deceleration rate ranges from 5% to 15% of the initial correction value. Preferably, the deceleration rate can be 10% of the initial correction value.
[0077] For example, if the sea depth at a certain point is 4000m, an on-site measurement is conducted to obtain a measured sound velocity profile with a total depth of 300m. Based on the longitude, latitude, time, and elevation of the point, a predicted sound velocity profile with a total depth of 4000m is obtained using a sound velocity profile prediction model. The overlap region between the measured and predicted sound velocity profiles is 0–300m. The average value of the sound velocity difference in the overlap region is calculated as the initial correction value. Considering that the impact of surface disturbances on the deep ocean environment decreases with depth, the correction value is reduced by 10% for every 100m increase in depth. The measured sound velocity is used for the 0–300m range, while the predicted sound velocity is superimposed with the correction value for the 300m–4000m range. Finally, the full-depth sound velocity profile from 0 to 4000m at that point is reconstructed.
[0078] It should be noted that, although... Figures 1 to 4 The sound velocity profile reconstruction method described above has been presented as an example, but those skilled in the art will understand that this disclosure is not limited thereto. In fact, users can flexibly set the specific implementation of each step according to their personal preferences and / or actual application scenarios, as long as the NLR model and ARMA model can be integrated to reconstruct the sound velocity profile.
[0079] Accordingly, based on any of the aforementioned sound velocity profile reconstruction methods, this application also provides a sound velocity profile reconstruction device. Since the working principle of the sound velocity profile reconstruction device of this application is the same as or similar to the principle of the sound velocity profile reconstruction method provided in this application, the repeated details will not be elaborated further.
[0080] See Figure 5The sound velocity profile reconstruction device 100 provided in this application includes an information acquisition module 110 and a sound velocity profile prediction module 120. The information acquisition module 110 is configured to acquire the geographical location information, topographic information, and data acquisition time of the currently collected target sea area. The sound velocity profile prediction module 120 is configured to input the geographical location information, topographic information, and data acquisition time as input data into a pre-constructed sound velocity profile prediction model, which then predicts the sound velocity profile based on the input data to obtain the corresponding predicted sound velocity profile. The sound velocity profile prediction model is constructed based on a nonlinear regression model and an autoregressive moving average model.
[0081] In one possible implementation, a data correction module (not shown in the figure) is also included. The data correction module is configured to correct the predicted sound velocity profile after the sound velocity profile prediction model 120 obtains the predicted sound velocity profile based on the input data.
[0082] Furthermore, according to another aspect of this disclosure, a sound velocity profile reconstruction device 200 is also provided. See also... Figure 6 The sound velocity profile reconstruction apparatus 200 of this disclosure includes a processor 210 and a memory 220 for storing executable instructions of the processor 210. The processor 210 is configured to implement any of the aforementioned sound velocity profile reconstruction methods when executing the executable instructions.
[0083] It should be noted here that the number of processors 210 can be one or more. Furthermore, the sound velocity profile reconstruction device 200 of this embodiment may also include an input device 230 and an output device 240. The processors 210, memory 220, input device 230, and output device 240 can be connected via a bus or other means, without specific limitations here.
[0084] The memory 220, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and various modules, such as the program or module corresponding to the sound velocity profile reconstruction method of this disclosure embodiment. The processor 210 executes various functional applications and data processing of the sound velocity profile reconstruction device 200 by running the software program or module stored in the memory 220.
[0085] Input device 230 can be used to receive input digital numbers or signals. These signals may include key signals related to user settings and function control of the device / terminal / server. Output device 240 may include a display device such as a screen.
[0086] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.
Claims
1. A method of sound speed profile reconstruction, characterized by, include: Obtain the geographic location, topographic information, and time information of the target sea area currently being collected; The geographic location information, the terrain information, and the time information are used as input data and input into a pre-built sound velocity profile prediction model. The sound velocity profile prediction model predicts the sound velocity profile based on the input data to obtain the corresponding predicted sound velocity profile. The sound velocity profile prediction model is constructed based on a nonlinear regression model and an autoregressive moving average model. The method further includes a step of correcting the predicted sound velocity profile based on the measured sound velocity profile obtained by the sound velocity profile prediction model after obtaining the predicted sound velocity profile based on the input data. When correcting the predicted sound velocity profile based on the obtained measured sound velocity profile, the following steps are included: Determine the overlapping region between the measured sound velocity profile and the predicted sound velocity profile, and update the data in the overlapping region of the predicted sound velocity profile to the measured sound velocity profile in the overlapping region. Calculate the data error between the measured sound velocity profile and the predicted sound velocity profile in the non-overlapping region, and correct the predicted sound velocity profile in the non-overlapping region based on the data error; When correcting the predicted sound velocity profile in the non-overlapping region based on the data error, the average error between the measured sound velocity profile and the predicted sound velocity profile in the overlapping region is used as the initial correction value, and the correction value is decreased sequentially according to the increasing region depth.
2. The method of claim 1, wherein, The geographic location information includes the longitude and latitude of the target sea area, and the topographic information includes the elevation of the target sea area.
3. The method of claim 1, wherein, When the sound velocity profile prediction model obtains the predicted sound velocity profile based on the input data, it includes: The nonlinear regression model is invoked to calculate the spatial mean estimate based on the input data, and the autoregressive moving average model is invoked to calculate the time series fluctuation estimate based on the input data. The spatial mean state estimate and the time series fluctuation estimate are superimposed to obtain the predicted sound velocity profile.
4. A sound speed profile reconstruction apparatus characterized by comprising: It includes an information acquisition module and a sound velocity profile prediction module; The information acquisition module is configured to acquire the geographic location information, topographic information, and data acquisition time of the currently collected target sea area; The sound velocity profile prediction module is configured to take the geographical location information, the terrain information and the data acquisition time as input data and input them into a pre-built sound velocity profile prediction model. The sound velocity profile prediction model then performs sound velocity profile prediction based on the input data to obtain the corresponding predicted sound velocity profile. The sound velocity profile prediction model is constructed based on a nonlinear regression model and an autoregressive moving average model. The method further includes a step of correcting the predicted sound velocity profile based on the measured sound velocity profile obtained by the sound velocity profile prediction model after obtaining the predicted sound velocity profile based on the input data. When correcting the predicted sound velocity profile based on the obtained measured sound velocity profile, the following steps are included: Determine the overlapping region between the measured sound velocity profile and the predicted sound velocity profile, and update the data in the overlapping region of the predicted sound velocity profile to the measured sound velocity profile in the overlapping region. Calculate the data error between the measured sound velocity profile and the predicted sound velocity profile in the non-overlapping region, and correct the predicted sound velocity profile in the non-overlapping region based on the data error; When correcting the predicted sound velocity profile in the non-overlapping region based on the data error, the average error between the measured sound velocity profile and the predicted sound velocity profile in the overlapping region is used as the initial correction value, and the correction value is decreased sequentially according to the increasing region depth.
5. The apparatus of claim 4, wherein, It also includes a data correction module; The data correction module is configured to correct the predicted sound velocity profile after the sound velocity profile prediction model obtains the predicted sound velocity profile based on the input data.
6. A sound velocity profile reconstruction device, characterized by, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the method of any one of claims 1 to 3 when executing the executable instructions.