Submarine topography inversion method based on full-connection deep neural network

Abstract

The invention discloses a submarine topography inversion method based on a full-connection deep neural network. The submarine topography inversion method comprises the following steps: acquiring ship-measured water depth data for data preprocessing; a gravity anomaly data set is acquired, a gravity anomaly field is decomposed into long wave and short wave components, and residual long wave and residual short wave gravity anomalies are obtained through fitting difference; constructing a sea depth inversion model, and taking the gravity anomaly data and the latitude and longitude data as input; and performing neural network training on the sea depth inversion model through the ship survey water depth data, outputting a sea depth inversion result, and performing precision evaluation optimization on the sea depth inversion model. According to the method, ship-borne sounding data, long-wave gravity anomaly data, short-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data are combined, the precision of submarine topography inversion is improved, a full-connection deep neural network model is used, the inversion capability of the model is further improved, the model can capture topographic changes more accurately, and the accuracy of submarine topography inversion is improved. And the water depth inversion precision is better.

Description

Technical Field

[0001] The present invention relates to the field of marine surveying and mapping technology, and in particular to a seabed topography inversion method based on a fully connected deep neural network. Background Art

[0002] Seabed topography measurement is a basic marine mapping work. Traditional methods mainly rely on ships as measurement tools, usually combined with ship positioning technology and sonar technology to detect ocean depth and seabed obstacles. However, due to the influence of ship movement speed, navigation route and natural conditions, the depth measurement results are uneven. There are still many data blank areas in the vast ocean, which makes our understanding of seabed topography relatively limited. These blank areas limit the progress of scientific research, resource development and ecological protection.

[0003] Satellite altimetry technology has brought innovative technical paths to seabed topography mapping with its all-weather, uninterrupted operation capabilities, wide global coverage and rapid information acquisition. Gravity anomalies and seabed topography have a significant correlation within specific bands. Therefore, the ocean gravity anomaly data obtained through altimetry technology can be used to invert seabed topography and provide data support for Earth system science research.

[0004] In the study of seafloor topography inversion using gravity anomaly data, commonly used methods include gravity geological method and its improved GGM method, admittance function method, Smith & Sandwell method (S&S) and least squares method. These methods provide reliable models to a certain extent, but they still have several limitations. For example, although the gravity geological method and its improved method have high accuracy, they require dense control point data and ignore the nonlinear relationship between gravity anomaly and water depth, and the inversion accuracy in shallow sea areas is also poor; the admittance function law uses the correlation between water depth and topography to perform model conversion in the frequency domain, which is suitable for sea areas lacking ship survey data, but its spatial resolution is low. For this reason, some researchers began to combine gradient anomalies with gravity anomalies to obtain better inversion results. The nonlinear iterative least squares method still has shortcomings in computational efficiency and processing time compared with other modeling methods. Therefore, it is urgent to further explore and develop more efficient algorithms to improve modeling efficiency and inversion accuracy. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides a seabed topography inversion method based on a fully connected deep neural network to solve the above problems.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for seafloor topography inversion based on a fully connected deep neural network, comprising: obtaining ship-measured water depth data and performing data preprocessing; Based on the ship-measured water depth data, the gravity anomaly data is decomposed into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data, wherein after the gravity anomaly field is decomposed into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the gravity anomaly field with the ship-measured water depth data respectively; A sea depth inversion model is constructed, the gravity anomaly data set and the longitude and latitude data are used as input, a neural network inversion is performed on the sea depth prediction model, the neural network is trained on the sea depth inversion model through ship-measured water depth data, the sea depth inversion result is output, and the accuracy of the sea depth inversion model is evaluated and optimized according to the ship-measured water depth data.

[0008] As a preferred solution of the method for seabed topography inversion based on a fully connected deep neural network described in the present invention, the step of obtaining ship-measured water depth data and performing data preprocessing includes: Performing error screening on the ship-measured water depth data, interpolating the water depth of the world ocean seabed topography map to the ship-measured water depth point, and calculating the difference with the ship-measured water depth value; Based on the difference, the standard deviation of the two is calculated, and the sounding points with a difference exceeding three times the standard deviation are eliminated; According to the ratio of the number of control points to the number of check points of 4:1, the control points and the check points are obtained randomly.

[0009] As a preferred solution of the seabed topography inversion method based on a fully connected deep neural network described in the present invention, the gravity anomaly data set obtained includes: Decomposing the gravity anomaly field into a long-wave component and a short-wave component, calculating the short-wave gravity anomaly of the control point, and subtracting the short-wave gravity anomaly of the control point from the gravity anomaly of the control point to obtain the long-wave gravity anomaly of the control point; Gridding the long-wave gravity anomaly of the control point to construct a long-wave gravity anomaly field, obtaining the long-wave gravity anomaly of the unknown point by interpolation, and solving the short-wave gravity anomaly of the unknown point; The linear relationship between the long-wave and short-wave components and the ship-measured water depth values ​​is determined through a linear regression model. The ship-measured water depth is multiplied by the corresponding fitting slope to obtain the reference values ​​of the long-wave and short-wave gravity anomalies. The long-wave and short-wave gravity anomalies are subtracted from the reference values ​​to obtain the residual long-wave gravity anomalies and residual short-wave gravity anomalies.

[0010] As a preferred solution of the method for seabed topography inversion based on a fully connected deep neural network described in the present invention, constructing a sea depth inversion model includes: An input layer, wherein the input data of the input layer are longitude, latitude and gravity anomaly data, and the gravity anomaly data include short-wave gravity anomaly, long-wave gravity anomaly, residual short-wave gravity anomaly and residual long-wave gravity anomaly; Four hidden layers, the first hidden layer has 16 neurons, the second hidden layer has 32 neurons, the third hidden layer has 256 neurons, and the fourth hidden layer has 512 neurons. The neurons in each layer are connected to each other through weights and biases; The output layer outputs the ocean depth measurement result via the hidden layer.

[0011] As a preferred solution of the method for seabed topography inversion based on a fully connected deep neural network described in the present invention, the sea depth inversion model is expressed as: ; in, represents the input gravity field signal, y represents the output of the neuron, , Represent the input weight and neuron bias respectively. represents the activation function; Among them, the weights and biases are given initial values ​​in the model initialization stage, and the weights of each neuron are adjusted according to the inversion differences during training. The weight iteration formula is expressed as: ; in, represents the adjusted weight, represents the learning rate, represents the initial weight, loss represents a loss function that depends on the mean squared error.

[0012] As a preferred solution of the method for seabed topography inversion based on a fully connected deep neural network described in the present invention, the neural network inversion includes: The location information of the ship's sounding control points and the gravity anomaly data are divided into a training set and a test set; The training data set and the test data set are normalized to train the ocean depth inversion model; Initialize the ocean depth inversion model and randomly generate initial weights and deviations; Using the loss function based on the root mean square error of the inverted depth and the ship-borne depth, the network was trained for 40 epochs with an initial learning rate of 0.001, and the parameters of the model were adjusted using the Adam optimization algorithm; The gravity signal obtained by altimetry is fed into the trained ocean depth inversion model to obtain the inverted water depth of the measurement area.

[0013] As a preferred solution of the seabed topography inversion method based on a fully connected deep neural network described in the present invention, the accuracy assessment optimization includes: Through bicubic interpolation method, multiple comparison models are interpolated to the check points and compared with the ship-measured water depth data at the check points; The standard deviation, root mean square error and mean absolute percentage error are used as the accuracy evaluation indicators of the ocean depth inversion model and other comparison models.

[0014] In a second aspect, the present invention provides a seabed topography inversion system based on a fully connected deep neural network, comprising: Data acquisition module, used to acquire ship depth data and perform data preprocessing; A data processing module is used to decompose the gravity anomaly data into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data based on the ship-measured water depth data, wherein after the gravity anomaly field is decomposed into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the gravity anomaly field with the ship-measured water depth data respectively; The model building module is used to build a sea depth inversion model, take the gravity anomaly data set and longitude and latitude data as input, perform neural network inversion on the sea depth prediction model, train the neural network of the sea depth inversion model through ship-measured water depth data, output the sea depth inversion result, and perform accuracy assessment and optimization on the sea depth inversion model based on the ship-measured water depth data.

[0015] In a third aspect, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the seabed topography inversion method based on a fully connected deep neural network are implemented.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the seabed topography inversion method based on a fully connected deep neural network.

[0017] Compared with the prior art, the present invention has the following beneficial effects: by combining shipborne bathymetric data, long-wave gravity anomaly data and short-wave gravity anomaly data, the present invention can make full use of multi-source information and improve the accuracy of seabed topography inversion; by using a fully connected deep neural network model, it can capture complex nonlinear relationships, improve modeling efficiency, and further enhance the inversion capability of the model; the model can more accurately capture terrain changes, and the inverted water depth accuracy is better; by decomposing the gravity anomaly data into long-wave and short-wave components, the seabed topography features of different scales can be better analyzed; the residual gravity anomaly further refines the topography features and improves the resolution of the local topography; the trained model can complete a wide range of seabed topography inversion in a short time, and is suitable for large-scale marine surveying and mapping tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0019] Figure 1 A schematic diagram of the overall process of a method for seabed topography inversion based on a fully connected deep neural network according to an embodiment of the present invention; Figure 2 A schematic diagram of a seabed topography inversion process of a seabed topography inversion method based on a fully connected deep neural network according to an embodiment of the present invention; Figure 3 A schematic diagram of the distribution of the research area and shipborne bathymetric trajectory of a seafloor topography inversion method based on a fully connected deep neural network according to an embodiment of the present invention; Figure 4 A schematic diagram of a sea depth inversion model of a seabed topography inversion method based on a fully connected deep neural network according to an embodiment of the present invention; Figure 5 The difference histogram and scatter plot of FCD_Depth_SCS, GGM _Depth, GEBCO_2024, SIOv25.1 and DTU18 models compared with the ship-measured water depth at the check point of the seafloor topography inversion method based on a fully connected deep neural network according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the distribution of different areas A, B, C, and D of the seabed topography inversion method based on a fully connected deep neural network described in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0023] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0024] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0025] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0026] Example:

[0027] Reference Figure 1-Figure 5Tables 1 to 3 are an embodiment of the present invention, which provides a method for seafloor topography inversion based on a fully connected deep neural network, such as Figure 1 , 2 As shown, including: S101, obtaining ship-measured water depth data and performing data preprocessing; S102, based on the ship-measured water depth data, decomposing the gravity anomaly data into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data, wherein after decomposing the gravity anomaly field into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the gravity anomaly field with the ship-measured water depth data respectively; S103, constructing a sea depth inversion model, taking the gravity anomaly data set and the longitude and latitude data as input, performing neural network inversion on the sea depth prediction model, training the sea depth inversion model with the ship-measured water depth data, outputting the sea depth inversion result, and performing accuracy assessment and optimization on the sea depth inversion model based on the ship-measured water depth data.

[0028] It should be noted that due to the influence of the ship's moving speed, sailing route and natural conditions, the water depth measurement results are uneven. There are still many data blank areas in the vast ocean, which makes our understanding of the seabed topography relatively limited. Gravity anomalies and seabed topography have a significant correlation in specific bands. The ocean gravity anomaly data obtained by altimetry technology can be used to invert the seabed topography.

[0029] Preferably, Figure 3 As shown, obtaining ship-measured water depth data and performing data preprocessing include: Error screening is performed on the ship-measured water depth data, the water depth of GEBCO is interpolated to the ship-measured water depth point, and the difference with the ship-measured water depth value is calculated; Based on the difference, the standard deviation of the two is calculated, and the sounding points with a difference exceeding three times the standard deviation are eliminated; According to the ratio of the number of control points to the number of check points of 4:1, the control points and the check points are obtained randomly.

[0030] Specifically, a data set of 108°E to 121°E, 6°N to 23°N in the South China Sea was obtained. The data set is ship-measured water depth data, which can be obtained through the National Center for Environmental Information of the United States. The data set was preprocessed, and the error of the ship-measured water depth data was screened using the "3σ" principle based on the GEBCO_2024 model. The specific method is to interpolate the water depth of the GEBCO_2024 model to the ship-measured water depth point and calculate the difference with the ship-measured water depth value; based on these differences, the standard deviation of the two is further calculated, and the sounding points with a difference of more than three times the standard deviation are eliminated. Finally, a total of 8653 data points were eliminated, and 802,379 ship-borne sounding points were retained, with an elimination rate of approximately 1.07%.

[0031] According to the principle that the ratio of control points to check points is 4:1, 641903 ship-surveyed sounding points were randomly selected as control points, and the remaining 160476 ship-surveyed sounding points were used as check points. The control points are only used to train the neural network model, while the check points are specifically used to evaluate the accuracy of the sea depth inversion model (FCD_Depth_SCS).

[0032] Preferably, in step S102, obtaining the gravity anomaly data set includes: Decompose the gravity anomaly field into long-wave components and short-wave components, calculate the short-wave gravity anomaly of the control point, and obtain the long-wave gravity anomaly of the control point by subtracting the short-wave gravity anomaly of the control point from the gravity anomaly of the control point. The long-wave gravity anomaly of the control points is processed into a grid, and the long-wave gravity anomaly field is constructed. The long-wave gravity anomaly of the unknown points is obtained by interpolation, and the short-wave gravity anomaly of the unknown points is solved. The linear relationship between the long-wave and short-wave components and the ship-measured water depth values ​​is determined through a linear regression model. The ship-measured water depth is multiplied by the corresponding fitting slope to obtain the reference values ​​of the long-wave and short-wave gravity anomalies. The long-wave and short-wave gravity anomalies are subtracted from the reference values ​​to obtain the residual long-wave gravity anomalies and residual short-wave gravity anomalies.

[0033] Specifically, after obtaining the long-wave and short-wave gravity anomalies, the linear relationship between the long-wave and short-wave gravity anomalies and the ship-measured water depth values ​​is determined through a linear regression model, the ship-measured water depth values ​​are used as input variables, and the long-wave and short-wave gravity anomaly data are used as target variables, and the parameters of the model, namely the slope and intercept of the linear equation, are calculated to minimize the difference between the inversion of the model on the training data and the actual value; the ship-borne water depth data is multiplied by the slope obtained by fitting to obtain the reference long-wave and short-wave gravity anomaly data, and the reference long-wave and short-wave gravity anomalies are subtracted from the long-wave and short-wave gravity anomalies to obtain the residual long-wave and short-wave gravity anomalies; therefore, the gravity anomaly data feature data set consists of short-wave gravity anomalies, long-wave gravity anomalies, residual short-wave gravity anomalies and residual long-wave gravity anomaly data.

[0034] Since the relationship between GA and bathymetric data is not linear, the GA field is decomposed into long-wave and short-wave components. The short-wave gravity anomaly mainly originates from subtle changes in local bedrock topography, while the long-wave gravity anomaly originates from changes in mass distribution at deeper levels within the crust.

[0035] In an optional implementation, the shortwave gravity anomaly of the calculated control point is expressed as: ; in, Indicates The short-wave gravity anomaly of the point, G is the gravitational constant, is the density difference constant, For control points The sea depth is 5174 m, and D is the reference sea depth, which is set to the maximum water depth covered by the study area.

[0036] Preferably, in step S103, Figure 4 As shown in Figure 2, constructing the ocean depth inversion model includes: Input layer: the input data of the input layer are longitude, latitude and gravity anomaly data. The gravity anomaly data include short-wave gravity anomaly, long-wave gravity anomaly, residual short-wave gravity anomaly and residual long-wave gravity anomaly. Four hidden layers, the first hidden layer has 16 neurons, the second hidden layer has 32 neurons, the third hidden layer has 256 neurons, and the fourth hidden layer has 512 neurons. The neurons in each layer are connected to each other through weights and biases; The output layer outputs the depth measurement results through the hidden layer.

[0037] It should be noted that the ocean terrain information is complex, and the inversion model needs to combine multi-source gravity field elements. Due to the complex characteristics of the input data, it is necessary to perform tasks such as classification and regression. Therefore, the sea depth inversion model is constructed as a fully connected neural network (FC-DNN). The sea depth inversion model contains an input layer, multiple hidden layers and an output layer. The hidden layer can assist in extracting relevant data features, but it may also lead to extended training time and overfitting. In order to achieve the best balance, four hidden layers are set. Each neuron in the sea depth inversion model uses specific operations to extract data features, and the neurons in each layer are connected to each other through weights and biases.

[0038] Preferably, the ocean depth inversion model is expressed as: ; in, represents the input gravity field signal, y represents the output of the neuron, , Represent the input weight and neuron bias respectively. represents the activation function; Among them, the weights and biases are given initial values ​​in the model initialization stage, and the weights of each neuron are adjusted according to the inversion differences during training. The weight iteration formula is expressed as: ; in, represents the adjusted weight, represents the learning rate, represents the initial weight, loss represents a loss function that depends on the mean squared error.

[0039] During model training, a reverse calculation is performed for each forward calculation, and the optimal value is obtained after multiple adjustments to the weights, so that the difference between the inverted water depth and the ship's depth is minimized.

[0040] It should be noted that the activation function used in this model is the rectified linear unit (ReLU) activation function. The data set needs to be normalized before the activation function to ensure that the data set falls within the range of 0 to 1. The sea depth inversion model also has the function of back-propagation optimization network. This function is mainly used to adjust the weights of the sea depth inversion model. In the model initialization stage, the weights and biases of the neural network are assigned initial values. During training, the weights of each neuron will be adjusted according to the differences in inversion.

[0041] Preferably, in step S103, performing neural network inversion includes: The location information of the ship's sounding control points and the gravity anomaly data are divided into a training set and a test set to train the sea depth inversion model, wherein the high-quality ship depth is used as training data and the remaining inspection data is used for testing; Perform maximum and minimum normalization on the training data and gravity field data to ensure that the data falls between 0 and 1, greatly reducing the difference in input data; Initialize the ocean depth inversion model and randomly generate initial weights and deviations; The network was trained for 40 epochs with an initial learning rate of 0.001 using a loss function based on the root mean square error of the inverted depth and the shipborne depth. In order to update the weights according to the learning rate and loss function, the parameters of the model were adjusted by the Adam optimization algorithm to optimize the performance. The gravity signal obtained by altimetry is fed into the trained ocean depth inversion model to obtain the inverted water depth of the measurement area.

[0042] Preferably, the accuracy assessment optimization includes: Through bicubic interpolation method, multiple comparison models are interpolated to the check points and compared with the ship-measured water depth data at the check points; The standard deviation, root mean square error and mean absolute percentage error are used as the accuracy evaluation indicators of the ocean depth inversion model and other comparison models.

[0043] Specifically, in order to evaluate the accuracy of the established FCD_Depth_SCS model, this study used the gravity geological method (GGM) to invert the submarine topography model of the South China Sea, named the GGM_Depth model. In addition, the GEBCO_2024, SIOv25.1 and DTU18 models were introduced. The FCD_Depth_SCS, GGM_Depth, GEBCO_2024, SIOv25.1 and DTU18 models were interpolated to the checkpoints by the bicubic interpolation method and compared with the ship-measured water depth data at the checkpoints. The standard deviation (STD), root mean square error (RMS) and mean absolute percentage error (MPAE / %) were used as evaluation indicators.

[0044] The mean absolute percentage error (MAPE / %) is an indicator widely used in inversion accuracy assessment. It is mainly used to measure the difference between the inverted value and the actual value. The calculation formula is as follows: ; in, Indicates The depth inversion value, For the The measured water depth value, The number of check points.

[0045] like Figure 5 As shown in the figure, the difference histograms and scatter plots of the FCD_Depth_SCS, GGM _Depth, GEBCO_2024, SIOv25.1 and DTU18 models compared with the ship-measured water depth at the check point: the first column is the FCD_Depth_SCS model, the second column is the GGM _Depth model, the third column is the GEBCO_2024 model, the fourth column is the SIOv25.1 model, and the fifth column is the DTU18 model.

[0046] Table 1 shows the statistical results of the comparison between the FCD_Depth_SCS, GGM_Depth, GEBCO_2024, SIOv25.1 and DTU18 models and the ship-measured water depth values ​​at the checkpoint.

[0047] Table 1: Comparison of different models and ship-measured water depth values ​​(unit: m) Model MAX MIN MEAN STD RMS MAPE / % FCD_Depth_SCS 1245.725 -857.709 0.187 44.755 44.756 2.903 GGM_Depth 1133.773 -1524.209 0.321 58.874 58.875 3.663 GEBCO_2024 842.874 -842.579 2.181 82.234 82.263 3.733 SIOv25.1 2137.575 -1964.993 6.786 108.241 108.454 5.149 DTU18 2130.733 -2762.660 31.904 186.967 189.670 9.095 By comparing the difference between the water depth inverted by the FCD_Depth_SCS model and the actual measured depth at the check point, the standard deviation (STD) is 44.755 m, and the mean absolute percentage error (MAPE) of the difference is 2.903%, which are better than the STD (58.874 m, 82.234 m, 108.241 m, 189.670 m) and MAPE (3.663%, 3.733%, 5.149%, 9.095%) of the difference between the GGM_Depth model, GEBCO_2024 model, SIOv25.1 model and DTU18 model and the actual measured water depth at the check point, showing a smaller inversion discreteness and higher accuracy, which indicates that the FCD_Depth_SCS model obtained by inverting gravity anomaly data through the FCDNN neural network is effective.

[0048] Table 2 shows the accuracy performance of the five models, FCD_Depth_SCS, GGM_Depth, GEBCO_2024, SIOv25.1 and DTU18, at different distances from the coastline.

[0049] Table 2: Accuracy results of different models at different distances from the coastline (unit: m) Table 3 shows the statistical results of the FCD_Depth_SCS, GGM_Depth, GEBCO_2024, SIOv25.1 and DTU18 models compared with the ship-measured water depth data in the range of water depth greater than 500 m and less than 500 m, respectively.

[0050] Table 3: Comparison results of different models with ship-measured water depth data in different water depth ranges (unit: m) Compared with the measured sea depth values ​​at check points in different water depth ranges and at different distances from the shore, the results show that the FCD_Depth_SCS model is stable in deep-water inversion, while other models such as SIOv25.1 and DTU18 have obvious error fluctuations at larger water depths; and its advantage is greater in the open sea area. As the distance from the coastline decreases, the accuracy of the model gradually decreases, indicating that the FCD_Depth_SCS model is superior to other models in inversion accuracy and stability.

[0051] It should be noted that the present invention can make full use of multi-source information and improve the accuracy of seabed topography inversion by combining shipborne bathymetric data, long-wave gravity anomaly data and short-wave gravity anomaly data. The fully connected deep neural network model can capture complex nonlinear relationships and further enhance the inversion capability of the model. The model can more accurately capture terrain changes and invert water depth with better accuracy. By decomposing the gravity anomaly data into long-wave and short-wave components, the seabed topography features of different scales can be better analyzed. The residual gravity anomaly further refines the topographic features and improves the resolution of local terrain. The trained model can complete a large range of seabed topography inversion in a short time and is suitable for large-scale marine surveying and mapping tasks.

[0052] The above is a schematic scheme of a method for seabed topography inversion based on a fully connected deep neural network in this embodiment. It should be noted that the technical scheme of the seabed topography inversion system based on a fully connected deep neural network and the technical scheme of the seabed topography inversion method based on a fully connected deep neural network belong to the same concept. The details not described in detail in the technical scheme of the seabed topography inversion system based on a fully connected deep neural network in this embodiment can be found in the description of the technical scheme of the seabed topography inversion method based on a fully connected deep neural network.

[0053] The seabed topography inversion system based on a fully connected deep neural network in this embodiment includes: Data acquisition module, used to acquire ship depth data and perform data preprocessing; A data processing module is used to decompose the gravity anomaly data into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data based on the ship-measured water depth data, wherein after the gravity anomaly field is decomposed into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the ship-measured water depth data respectively; The model building module is used to build a sea depth inversion model. It uses the gravity anomaly data set and longitude and latitude data as input, performs neural network inversion on the sea depth prediction model, trains the neural network of the sea depth inversion model through ship-measured water depth data, outputs the sea depth inversion results, and performs accuracy assessment and optimization on the sea depth inversion model based on the ship-measured water depth data.

[0054] This embodiment also provides an electronic device, which is applicable to the case of seabed topography inversion based on a fully connected deep neural network, including: A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the seabed topography inversion method based on a fully connected deep neural network as proposed in the above embodiment.

[0055] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the seabed topography inversion method based on a fully connected deep neural network as proposed in the above embodiment.

[0056] The storage medium proposed in this embodiment and the method for implementing seabed topography inversion based on a fully connected deep neural network proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0057] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0058] Example:

[0059] Reference Figure 6 And Table 4, which is an embodiment of the present invention, provides a seabed topography inversion method based on a fully connected deep neural network. In order to verify its beneficial effects, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0060] In order to further verify the model effect, compare the inversion results of different sea areas, further analyze the correlation between the accuracy of the seabed topography model based on neural network inversion and the size of the sea area, and provide an implementation case for the subsequent inversion of the global seabed topography model, four sea areas of different sizes were selected in the South China Sea for experiments.

[0061] Table 4 shows the comparative statistical results of the inversion results of the four sea areas A, B, C, and D using the FCDNN neural network and the ship-measured water depth data. It also provides the comparison results of the FCD_Depth_SCS model with the ship-measured water depth in each sea area.

[0062] Table 4: Comparison results between the sea depth inversion model and ship-measured water depth in various sea areas This paper selects four sea areas of different sizes (A, B, C, D) in the South China Sea to analyze the correlation between model accuracy and sea area size and topography. Figure 6 The results show that the non-regional model has significant advantages in areas A and C where the terrain is flat or has little change, and its STD and MAPE are lower than those of the partitioned model. In area B, although the terrain is relatively complex, the seabed is flat, and the non-regional model still shows a slightly better performance than the partitioned model. In areas with complex terrain (such as area D), the STD of the difference between the FCD_Depth_SCS_D of the partitioned model and the ship-measured water depth at the check point is 98.529 m, while the STD of the non-regional model is 163.879 m, and the MAPE (2.008%) is also better, indicating that the partitioned model can capture terrain changes more accurately and has better accuracy in inverting water depth.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for seafloor topography inversion based on a fully connected deep neural network, characterized in that: include: Obtain ship-measured water depth data and gravity anomaly data, and perform data preprocessing; Based on the ship-measured water depth data, the gravity anomaly data is decomposed into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data, wherein after the gravity anomaly field is decomposed into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the gravity anomaly field with the ship-measured water depth data respectively; A sea depth inversion model is constructed, the gravity anomaly data set and the longitude and latitude data are used as input, a neural network inversion is performed on the sea depth prediction model, the neural network is trained on the sea depth inversion model through ship-measured water depth data, the sea depth inversion result is output, and the accuracy of the sea depth inversion model is evaluated and optimized according to the ship-measured water depth data.

2. The method for seabed topography inversion based on a fully connected deep neural network according to claim 1, characterized in that: The obtaining of ship-measured water depth data and performing data preprocessing includes: Performing error screening on the ship-measured water depth data, interpolating the water depth of the world ocean seabed topographic map model to the ship-measured water depth point, and calculating the difference with the ship-measured water depth value; Based on the difference, the standard deviation of the two is calculated, and the sounding points with a difference exceeding three times the standard deviation are eliminated; According to the ratio of the number of control points to the number of check points of 4:1, the control points and the check points are obtained randomly.

3. The method for seabed topography inversion based on a fully connected deep neural network according to claim 2, characterized in that: The obtaining of the gravity anomaly data set comprises: Decomposing the gravity anomaly field into a long-wave component and a short-wave component, calculating the short-wave gravity anomaly of the control point, and subtracting the short-wave gravity anomaly of the control point from the gravity anomaly of the control point to obtain the long-wave gravity anomaly of the control point; Gridding the long-wave gravity anomaly of the control point to construct a long-wave gravity anomaly field, obtaining the long-wave gravity anomaly of the unknown point by interpolation, and solving the short-wave gravity anomaly of the unknown point; The linear relationship between the long-wave and short-wave components and the ship-measured water depth values ​​is determined through a linear regression model. The ship-measured water depth is multiplied by the corresponding fitting slope to obtain the reference values ​​of the long-wave and short-wave gravity anomalies. The long-wave and short-wave gravity anomalies are subtracted from the reference values ​​to obtain the residual long-wave gravity anomalies and residual short-wave gravity anomalies.

4. The method for seabed topography inversion based on a fully connected deep neural network according to claim 1 or 3, characterized in that: Building a sea depth inversion model includes: An input layer, wherein the input data of the input layer are longitude, latitude and gravity anomaly data, and the gravity anomaly data include short-wave gravity anomaly, long-wave gravity anomaly, residual short-wave gravity anomaly and residual long-wave gravity anomaly; Four hidden layers, the first hidden layer has 16 neurons, the second hidden layer has 32 neurons, the third hidden layer has 256 neurons, and the fourth hidden layer has 512 neurons. The neurons in each layer are connected to each other through weights and biases; The output layer outputs the ocean depth inversion result via the hidden layer.

5. The method for seabed topography inversion based on a fully connected deep neural network according to claim 4, characterized in that: The ocean depth inversion model is expressed as: ； in, represents the input gravity field signal, y represents the output of the neuron, , Represent the input weight and neuron bias respectively. represents the activation function; Among them, the weights and biases are given initial values ​​in the model initialization stage, and the weights of each neuron are adjusted according to the inversion differences during training. The weight iteration formula is expressed as: ； in, represents the adjusted weight, represents the learning rate, represents the initial weight, loss represents a loss function that depends on the mean squared error.

6. The method for seabed topography inversion based on a fully connected deep neural network according to claim 1, characterized in that: The neural network inversion comprises: The location information of the ship's sounding control points and the gravity anomaly data are divided into a training set and a test set; The training data set and the test data set are normalized to train the ocean depth inversion model; Initialize the ocean depth inversion model and randomly generate initial weights and deviations; Using the loss function based on the root mean square error of the inverted depth and the ship-borne depth, the network was trained for 40 epochs with an initial learning rate of 0.001, and the parameters of the model were adjusted using the Adam optimization algorithm; The gravity signal obtained by altimetry is fed into the trained ocean depth inversion model to obtain the inverted water depth of the measurement area.

7. The method for seabed topography inversion based on a fully connected deep neural network according to claim 6, characterized in that: The accuracy assessment optimization includes: Through bicubic interpolation method, multiple comparison models are interpolated to the check points and compared with the ship-measured water depth data at the check points; The standard deviation, root mean square error and mean absolute percentage error are used as the accuracy evaluation indicators of the ocean depth inversion model and other comparison models.

8. A seabed topography inversion system based on a fully connected deep neural network, used in the seabed topography inversion method based on a fully connected deep neural network as described in any of claims 1 to 7, characterized in that: include, Data acquisition module, used to acquire ship depth data and perform data preprocessing; A data processing module is used to decompose the gravity anomaly data into short-wave gravity anomaly data, long-wave gravity anomaly data, residual short-wave gravity anomaly data and residual long-wave gravity anomaly data based on the ship-measured water depth data, wherein after the gravity anomaly field is decomposed into long-wave components and short-wave components, the residual long-wave gravity anomaly and the residual short-wave gravity anomaly are obtained by fitting and subtracting the gravity anomaly field with the ship-measured water depth data respectively; The model building module is used to build a sea depth inversion model, take the gravity anomaly data set and longitude and latitude data as input, perform neural network inversion on the sea depth prediction model, train the neural network of the sea depth inversion model through ship-measured water depth data, output the sea depth inversion result, and perform accuracy assessment and optimization on the sea depth inversion model based on the ship-measured water depth data.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the seabed topography inversion method based on a fully connected deep neural network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the seabed topography inversion method based on a fully connected deep neural network as described in any one of claims 1 to 7.

Images