Submarine topography deformation prediction method and system, electronic equipment and storage medium

Abstract

The invention provides a submarine topography deformation prediction method and system, electronic equipment and a storage medium, and belongs to the technical field of topography monitoring. The method comprises the following steps: acquiring submarine topography data acquired at a current moment and a plurality of historical moments to form a current multi-scale time sequence; decomposing the current multi-scale time sequence to obtain a current trend sequence, a current seasonal sequence and a current residual sequence; performing gradient descent optimization on the current residual error sequence to obtain a current optimal residual error sequence; generating a current optimal multi-scale time sequence according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence; performing multi-scale division and coding on the current optimal multi-scale time sequence to obtain a plurality of groups of subsequences; and inputting the plurality of groups of subsequences into a seabed deformation prediction model based on a self-adaptive multi-head attention mechanism for analysis to obtain submarine topography prediction data at a future moment. The method can improve the prediction accuracy of submarine topography deformation.

Description

Technical Field

[0001] The present application relates to the field of terrain monitoring technology, and in particular to a method, system, electronic device and storage medium for predicting seabed terrain deformation. Background Art

[0002] As a common geological phenomenon, the change of seabed topography has become an important research topic in the current field of geographic information science and engineering to accurately monitor and predict the change of seabed topography. With the advancement of sensor technology, the acquisition of seabed topography data has become increasingly convenient. How to effectively extract and fuse these seabed topography data has become the key to improving the accuracy of seabed topography deformation prediction. In the existing technology, the fast Fourier transform algorithm is mainly used to extract features from these seabed topography data, and then the extracted feature data is convoluted and folded. However, this processing method is usually only applicable to stationary sequences, and it is difficult to capture the dynamic changes and complex patterns of the seabed topography, resulting in inaccurate predictions of subsequent seabed topography deformation. Summary of the invention

[0003] The main purpose of this application is to propose a method, system, electronic device and storage medium for predicting seabed terrain deformation, which can more comprehensively capture the changing characteristics of seabed terrain and thus improve the prediction accuracy of seabed terrain deformation.

[0004] To achieve the above objectives, one aspect of the present application provides a method for predicting seabed terrain deformation, the method comprising:

[0005] Acquire seabed topographic data collected at the current moment and multiple historical moments to form a current multi-scale time series; wherein the seabed topographic data includes height change values ​​of multiple different positions in the seabed area to be measured relative to the sea level;

[0006] Decomposing the current multi-scale time series to obtain a current trend series, a current seasonal series, and a current residual series related to changes in seafloor topography;

[0007] Performing gradient descent optimization on the current residual sequence to obtain a current optimal residual sequence;

[0008] Generate a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence;

[0009] Performing multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences;

[0010] The multiple groups of subsequences are input into a seabed deformation prediction model based on an adaptive multi-head attention mechanism for analysis to obtain seabed topography prediction data at future moments.

[0011] Furthermore, performing gradient descent optimization on the current residual sequence to obtain a current optimal residual sequence includes:

[0012] Using the current residual sequence as the current residual sequence to be optimized;

[0013] Calculating a gradient value of a loss function on the current residual sequence to be optimized, wherein the loss function is used to measure the difference between the current multiscale time series and the optimized current multiscale time series, wherein the optimized current multiscale time series is obtained by optimizing the current residual sequence contained in the current multiscale time series;

[0014] Optimizing the current residual sequence to be optimized by using the gradient value to obtain an optimized current residual sequence;

[0015] Determine whether the preset iteration end condition is met; if so, use the optimized current residual sequence as the current optimal residual sequence; if not, use the optimized current residual sequence as the current residual sequence to be optimized, and then return to the step of calculating the gradient value generated by the loss function for the current residual sequence to be optimized.

[0016] Further, generating a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence includes:

[0017] Obtain a historical trend sequence at a first historical moment and a historical seasonal sequence at a second historical moment, wherein the first historical moment is the moment before the current moment, and the time interval between the second historical moment and the current moment is a seasonal cycle;

[0018] According to the difference between the current trend sequence and the historical trend sequence, the current trend sequence is optimized to obtain a current optimal trend sequence;

[0019] According to the difference between the current seasonal sequence and the historical seasonal sequence, the current seasonal sequence is optimized to obtain a current optimal seasonal sequence;

[0020] The current optimal trend sequence, the current optimal seasonal sequence and the current optimal residual sequence are fused to obtain the current optimal multi-scale time series.

[0021] Furthermore, the multi-scale division and encoding of the current optimal multi-scale time series to obtain multiple groups of subsequences includes:

[0022] According to the preset scale range, a plurality of different scale factors are determined;

[0023] For each of the scale factors, dividing the current optimal multi-scale time series according to the scale factor to obtain all initial subsequences;

[0024] According to the current optimal multi-scale time series, performing sliding average processing on the boundaries of all the initial subsequences to obtain all the optimal subsequences;

[0025] After position encoding is performed on all the optimal subsequences respectively, a group of subsequences is formed.

[0026] Furthermore, the seabed deformation prediction model based on the adaptive multi-head attention mechanism includes an encoder network, an adaptive multi-head attention network and a feedforward neural network; the multiple groups of subsequences include several optimal subsequences, and the multiple optimal subsequences correspond to carrying several position encoding information; the multiple groups of subsequences are input into the seabed deformation prediction model based on the adaptive multi-head attention mechanism for analysis, and the seabed terrain prediction data at the future moment is obtained, including:

[0027] Using the encoder network to extract features from the plurality of optimal subsequences respectively, to obtain a plurality of initial potential feature data related to the seabed topography change;

[0028] Generating a plurality of final latent feature data according to the plurality of initial latent feature data and the plurality of position coding information;

[0029] Using the adaptive multi-head attention network to extract features from the plurality of final potential feature data, respectively, to obtain a plurality of attention feature data related to changes in seabed topography;

[0030] The feedforward neural network is used to perform fusion analysis on the plurality of attention feature data to obtain the seabed topography prediction data at the future moment.

[0031] Further, the generating of a plurality of final latent feature data according to the plurality of initial latent feature data and the plurality of position coding information comprises:

[0032] Correspondingly concatenating the plurality of initial latent feature data and the plurality of position coding information to obtain a plurality of intermediate latent feature data;

[0033] The plurality of intermediate latent feature data are respectively linearly transformed to obtain the plurality of final latent feature data.

[0034] Furthermore, a plurality of attention heads are set in the adaptive multi-head attention network; the adaptive multi-head attention network is used to extract features of the plurality of final potential feature data respectively, and the plurality of attention feature data related to the change of the seabed topography are obtained, including:

[0035] For each of the final latent feature data, use the multiple attention heads to process the final latent feature data respectively to obtain multiple attention outputs;

[0036] Performing weighted summation on the multiple attention outputs to obtain a final attention output;

[0037] The final attention output is residually connected and normalized with the final potential feature data to obtain attention feature data.

[0038] To achieve the above objectives, another aspect of the present application provides a seabed terrain deformation prediction system, the system comprising:

[0039] An acquisition module is used to acquire seafloor topographic data collected at the current moment and multiple historical moments to form a current multi-scale time series; wherein the seafloor topographic data includes height change values ​​of multiple different positions in the seafloor area to be measured relative to the sea level;

[0040] A decomposition module, used for decomposing the current multi-scale time series to obtain a current trend sequence, a current seasonal sequence and a current residual sequence related to the change of the seabed topography;

[0041] An optimization module, used for performing gradient descent optimization on the current residual sequence to obtain a current optimal residual sequence;

[0042] A generating module, used for generating a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence;

[0043] A division module, used for performing multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences;

[0044] The prediction module is used to input the multiple groups of subsequences into a seabed deformation prediction model based on an adaptive multi-head attention mechanism for analysis to obtain seabed topography prediction data at a future moment.

[0045] To achieve the above objective, another aspect of the present application provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above method when executing the computer program.

[0046] To achieve the above objective, another aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0047] The present application includes at least the following beneficial effects: by decomposing the current trend sequence, the current seasonal sequence and the current residual sequence from the current multi-scale time series formed by the seabed topography data collected at different times, and then performing gradient descent optimization on the current residual sequence covering the subtle features of the seabed topography changes, reintegrating it together with the current trend sequence and the current seasonal sequence to generate the current optimal multi-scale time series, and finally performing multi-scale division and encoding on the current optimal multi-scale time series, and completing the prediction analysis through the seabed deformation prediction model, and introducing an adaptive multi-head attention mechanism in the prediction analysis process to more comprehensively capture the characteristics of the seabed topography changes, the prediction accuracy of the seabed topography deformation can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of a flow chart of a method for predicting seabed terrain deformation provided in an embodiment of the present application;

[0049] Figure 2 It is a structural schematic diagram of a submarine terrain deformation prediction system provided by an embodiment of the present application;

[0050] Figure 3 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of systems and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0052] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0053] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0055] The change of seabed topography not only affects the health of the ecosystem, such as leading to intensified climate change, rising sea levels, and frequent extreme weather events, but also affects the effective development and utilization of marine resources. As a common geological phenomenon, the accurate monitoring and prediction of seabed topography changes has become an important research topic in the current field of geographic information science and engineering.

[0056] With the advancement of sensor technology, it is becoming increasingly convenient to obtain submarine topography data, which also greatly facilitates the prediction of submarine topography deformation. However, these submarine topography data often show complex trends and seasonal changes. How to effectively extract and integrate these submarine topography data has become the key to improving the accuracy of submarine topography deformation prediction. Therefore, multi-scale analysis is of great significance in submarine topography research. It can analyze topographic changes at multiple levels, such as the movement of sediments, the formation of submarine cracks, and the interaction between climate change and geological tectonic movement. By integrating data at different scales, the complex mechanisms behind these changes can be revealed, thereby providing a scientific basis for the prediction of submarine topography changes.

[0057] In the existing technologies, such as the MSGNet (Multi-Scale Graph Network) model and the TimesNet model, the fast Fourier transform algorithm is mainly used to extract features from these seabed topography data, and then the extracted feature data is convolved and folded. However, this processing method is usually only applicable to stationary sequences and it is difficult to capture the dynamic changes and complex patterns of the seabed topography, resulting in inaccurate subsequent predictions of seabed topography deformation.

[0058] In view of this, an embodiment of the present application provides a method, system, electronic device and storage medium for predicting seabed terrain deformation. The scheme decomposes a current trend sequence, a current seasonal sequence and a current residual sequence from a current multi-scale time series formed by seabed terrain data collected at different times, and then performs gradient descent optimization on the current residual sequence covering subtle features of seabed terrain changes, and then reintegrates it together with the current trend sequence and the current seasonal sequence to generate a current optimal multi-scale time series. Finally, the current optimal multi-scale time series is multi-scale divided and encoded, and then prediction analysis is completed through a seabed deformation prediction model. In the prediction analysis process, an adaptive multi-head attention mechanism is introduced to more comprehensively capture the characteristics of seabed terrain changes, thereby improving the prediction accuracy of seabed terrain deformation.

[0059] A method for predicting seabed terrain deformation provided in an embodiment of the present application relates to the field of terrain monitoring technology, and can be applied to a terminal, a server, or software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, and a vehicle-mounted terminal, etc., but is not limited thereto; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network; the software can be an application that implements the above-mentioned method for predicting seabed terrain deformation, etc., but is not limited to the above forms.

[0060] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0061] Figure 1 is an optional flow chart of a method for predicting seabed terrain deformation provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S106:

[0062] Step S101, acquiring seafloor topography data collected at the current moment and at multiple historical moments to form a current multi-scale time series;

[0063] Step S102, decomposing the current multi-scale time series to obtain a current trend series, a current seasonal series, and a current residual series related to changes in seafloor topography;

[0064] Step S103, performing gradient descent optimization on the current residual sequence to obtain the current optimal residual sequence;

[0065] Step S104, generating a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence;

[0066] Step S105, performing multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences;

[0067] Step S106: input multiple groups of subsequences into the seabed deformation prediction model based on the adaptive multi-head attention mechanism for analysis to obtain the seabed topography prediction data at the future moment.

[0068] Steps S101 to S106 shown in the embodiment of the present application can more comprehensively capture the changing characteristics of the seabed topography, thereby improving the prediction accuracy of the seabed topography deformation.

[0069] In step S101 of some embodiments, multiple historical moments are all before the current moment, and the multiple historical moments and the current moment are continuous moments, and the seabed topography data collected at each moment includes the height change values ​​of multiple different positions in the seabed area to be measured relative to the sea level at that moment, which can be obtained by deploying displacement sensors at multiple different positions in the seabed area to be measured; in addition, the current multi-scale time series can be understood as being formed by arranging all seabed topography data in order from the earliest to the latest according to the collection time.

[0070] In step S102 of some embodiments, the STL (Seasonal and Trend decomposition using Loess) algorithm is used to decompose the current multi-scale time series into a current seasonal series, a current trend series, and a current residual series related to the seabed topography change, which can be represented by the following expression:

[0071] X t =S t +T t +R t ;

[0072] In the formula, X t is the current multi-scale time series, S t is the current seasonal series, which is used to characterize the seasonal fluctuation of seafloor subsidence. t is the current trend sequence, which is used to characterize the long-term trend of seafloor subsidence. t is the current residual sequence, which is used to characterize the random fluctuation of seafloor subsidence.

[0073] Specifically, the STL algorithm is a general and robust time series decomposition method, and the corresponding implementation process includes: after selecting a suitable seasonal cycle, firstly, the current multi-scale time series is analyzed by a local weighted regression smoothing method to fit a trend line therefrom, thereby decomposing the current trend sequence from the current multi-scale time series to remove long-term changes. In this process, it is possible to focus on instantaneous effects related to changes in seabed topography, such as sediment movement or ocean currents on the seabed; then, the current multi-scale time series after the initial decomposition is smoothed and analyzed by a local weighted regression smoothing method to identify short-term seasonal fluctuations therefrom, thereby continuing to decompose the current seasonal sequence from the current multi-scale time series after the initial decomposition, in this process, it is possible to focus on the regular changes in the shape and structure of the seabed area to be measured caused by water level changes in some cases; finally, the current multi-scale time series after the second decomposition is output as the current residual sequence, which represents the random fluctuations in the seabed topography data, and helps to identify abnormal phenomena or short-term subtle changes that have not been captured.

[0074] It should be noted that the number of all trend component values ​​contained in the current trend sequence, the number of all seasonal component values ​​contained in the current seasonal sequence, and the number of all residual component values ​​contained in the current residual sequence are the same as the number of all seabed topography data contained in the current multi-scale time series, and there is a time correspondence between the data contained in these four sequences.

[0075] In the above step S102, the current multi-scale time series is a non-stationary series. By introducing the STL algorithm to perform feature decomposition on the current multi-scale time series, the spatiotemporal data related to seabed topography deformation (such as tides, ocean currents and other phenomena) can be better processed.

[0076] As an optional implementation, before decomposing the current multi-scale time series, the current multi-scale time series may be preprocessed to ensure the integrity and accuracy of the collected data.

[0077] In the present application, the Z-score algorithm is preferably used to perform preprocessing such as cleaning and filling on the current multi-scale time series. The corresponding implementation process includes: for all height change values ​​collected by the same displacement sensor at different times contained in the current multi-scale time series, the average value and standard deviation between all height change values ​​are calculated, and then each height change value is subtracted from the average value, and the subtraction result is divided by the standard deviation, and it is determined whether the absolute value of the division result is greater than a preset value as an empirical rule. The preset value is preferably set to 3. If so, the height change value is retained (that is, the height change value is judged to be a normal value). If not, the height change value is replaced with the average value (that is, the height change value is judged to be an abnormal value). This mean filling strategy can avoid data deviation.

[0078] In some embodiments, the above step S103 may include but is not limited to steps S201 to S204:

[0079] Step S201: taking the decomposed current residual sequence as the current residual sequence to be optimized.

[0080] Step S202, calculating the gradient value generated by the loss function for the current residual sequence to be optimized; wherein the loss function is mainly used to measure the difference between the current multi-scale time series and the optimized current multi-scale time series, and the optimized current multi-scale time series is obtained by optimizing the current residual sequence contained in the current multi-scale time series.

[0081] In this step, the mean square error (MSE) function is preferably used as the loss function, and the corresponding mathematical expression is as follows:

[0082]

[0083] Where: X i =S i +T i +R i ,

[0084] In the formula, L(R t ) is the current residual sequence R t The loss function set by the optimization problem, N is the current multi-scale time series X t The number of all seafloor topography data contained in is the optimized current multi-scale time series, R t ′ is the current residual sequence after optimization, X i is the current multi-scale time series X t The i-th seafloor topography data contained in is the optimized seafloor topography data contained in the optimized current multi-scale time series, S i is the current seasonal series S t The i-th seasonal component value contained in T i is the current trend sequence T t The i-th trend component value contained in R i is the current residual sequence R t The i-th residual component value contained in R i ′ is the optimized current residual sequence R t ′ contains the i-th optimized residual component value, and X i , S i 、T i , R i and R i There is a time correspondence between ′, that is, the time corresponding to these data is the same.

[0085] In this step, the gradient value of the loss function for the current residual sequence to be optimized can be determined using the following expression:

[0086]

[0087] In the formula, is the loss function L(R t ) The current residual sequence R to be optimized t The resulting gradient value.

[0088] Step S203: Utilize the gradient value to optimize the current residual sequence to be optimized, and obtain the optimized current residual sequence, which can be determined by the following expression:

[0089]

[0090] Where α is the learning rate used to control the update step size.

[0091] Step S204: determine whether the preset iteration end condition is met; if so, the optimized current residual sequence R′ is t Output as the current optimal residual sequence; if it does not meet the requirement, the optimized current residual sequence R′ t As the current residual sequence R to be optimized t , and returns to execute the above step S202.

[0092] Among them, the iteration end condition can be set to that the change of the loss function is less than the preset change threshold. At this time, it is necessary to calculate the current loss value according to the loss function using the optimized current residual sequence obtained in the current optimization process, and use the optimized current residual sequence obtained in the previous optimization process to calculate the historical loss value, and then calculate the absolute value of the difference between the current loss value and the historical loss value and use it as the change of the loss function; the iteration end condition can also be set to that the current loss value converges to less than the preset loss threshold, and can also be set to that the current number of optimization executions reaches the preset maximum number of iterative optimization executions. This application does not make specific limitations on this.

[0093] In steps S201 to S204 shown in the embodiment of the present application, considering that the current residual sequence decomposed by the STL algorithm is often ignored, that is, the subtle features of the seabed topography changes it covers are easily directly regarded as noise, the present application adopts the gradient descent method to optimize the current residual sequence, which can avoid information loss when the subsequent fusion generates the current optimal multi-scale time series, and also enable the subsequent model to better capture the potential patterns and structures in the residuals, thereby improving the interpretability of the prediction.

[0094] In some embodiments, the above step S104 may include but is not limited to steps S301 to S304:

[0095] Step S301, obtaining a historical trend sequence at a first historical moment, the first historical moment being the moment before the current moment, the historical trend sequence being obtained by decomposing a multi-scale time series collected at the first historical moment using an STL algorithm; and obtaining a historical seasonal sequence at a second historical moment, the time interval between the second historical moment and the current moment being a seasonal cycle, the historical seasonal sequence being also obtained by decomposing a multi-scale time series collected at the second historical moment using an STL algorithm.

[0096] Step S302: According to the difference between the current trend sequence and the historical trend sequence, the current trend sequence is optimized to obtain the current optimal trend sequence, which can be determined by the following expression:

[0097] ΔT t =T t -T t-1 , T′ t =T t +ΔT t ;

[0098] In the formula, T′ t is the current optimal trend sequence, T t-1 is the historical trend sequence at the first historical moment, Δt t For the current trend sequence T t And the historical trend sequence T t-1 The first-order difference results between .

[0099] Step S303: According to the difference between the current seasonal sequence and the historical seasonal sequence, the current seasonal sequence is optimized to obtain the current optimal seasonal sequence, which can be determined by the following expression:

[0100] ΔS t =S t -S t-s , S′ t =S t +ΔS t ;

[0101] In the formula, S′ t is the current optimal seasonal sequence, S t-s is the historical seasonal series at the second historical moment, s is the seasonal period, ΔS t is the current seasonal series S t and the historical seasonal series S t-s The first-order difference results between .

[0102] Step S304: fuse the current optimal trend sequence, the current optimal seasonal sequence and the current optimal residual sequence to obtain the current optimal multi-scale time series, which can be determined by the following expression:

[0103] X′ t =S′ t +T′ t +R′ t ;

[0104] In the formula, X′ t It is the current optimal multi-scale time series.

[0105] Steps S301 to S304 shown in the embodiment of the present application optimize the current trend sequence and the current seasonal sequence decomposed by the STL algorithm by using a first-order difference processing method, which can highlight the dynamic characteristics of the seabed topography changes, avoid static influences, and facilitate subsequent models to better capture the time series characteristics.

[0106] In some embodiments, the above step S105 may include but is not limited to steps S401 to S402:

[0107] Step S401: Determine a plurality of different scale factors according to a preset scale range.

[0108] In this step, the scale range is preferably set to [1,10], and the number of all seabed topography data contained in the current optimal multi-scale time series should be much larger than 10, the number of all seabed topography data contained in the current optimal multi-scale time series is the same as the number of all seabed topography data contained in the current multi-scale time series, and different multiple scale factors are selected from the scale range in a random manner. The value of each scale factor is an integer, and the scale factor mainly represents the amount of data required to be included in each initial subsequence subsequently divided.

[0109] Step S402: For each scale factor, firstly divide the current optimal multi-scale time series according to the scale factor to obtain all initial subsequences; secondly, perform sliding average processing on the boundaries of all initial subsequences according to the current optimal multi-scale time series to obtain all optimal subsequences; finally, position-code all optimal subsequences to form a group of subsequences. According to this embodiment, the current optimal multi-scale time series is divided and encoded using multiple scale factors to obtain multiple groups of corresponding subsequences.

[0110] In this step, the number of all initial subsequences divided from the current optimal multi-scale time series can be determined using the following expression:

[0111]

[0112] Where M is the number of all initial subsequences, N is the number of all seafloor topography data contained in the current optimal multi-scale time series, s is the scale factor, The floor symbol.

[0113] Exemplarily, starting from the first seabed topography data contained in the current optimal multi-scale time series, s seabed topography data are extracted from the current optimal multi-scale time series to form a first initial sub-sequence; starting from the s+1th seabed topography data contained in the current optimal multi-scale time series, s seabed topography data are extracted from the current optimal multi-scale time series to form a second initial sub-sequence; starting from the 2s+1th seabed topography data contained in the current optimal multi-scale time series, s seabed topography data are extracted from the current optimal multi-scale time series to form a third initial sub-sequence; and so on, until the Mth initial sub-sequence is formed.

[0114] In this step, for each initial subsequence, according to the current optimal multi-scale time series, the boundary of the initial subsequence is processed by sliding average to obtain the optimal subsequence, which can be determined by the following expression:

[0115]

[0116] In the formula, x′ i is the boundary x about the initial subsequence i The sliding average processing result, x i It can be understood as the i-th seafloor topography data contained in the current optimal multi-scale time series, x i-m is the ith seafloor topography data contained in the current optimal multi-scale time series, and k is the preset sliding window size.

[0117] For example, assuming that the scale factor is 3, the current optimal multi-scale time series X′ t All the seafloor topography data contained in j ,...,x N-1}, j = 0, 1, ..., N-1, i = j + 1, then the scale factor is used to calculate the optimal multi-scale time series X′ t Divide and get initial subsequences, denoted as: {x0,x1,x2}, {x3,x4,x5}, {x6,x7,x8}, ..., {x N-3 ,x N-2 ,x N-1};

[0118] Assuming that the sliding window size is 2, considering the current optimal multi-scale time series X′ t It is impossible to find a seabed topography data that falls in front of the boundary x0 of the first initial subsequence {x0,x1,x2}. Here, it is assumed that there is no need to perform sliding average processing on the first initial subsequence {x0,x1,x2}, that is, the first initial subsequence {x0,x1,x2} is taken as the first optimal subsequence; for the second initial subsequence {x3,x4,x5}, its boundary x3 is updated to x′3=(x2+x3) / 2, and its boundary x5 is updated to x′5=(x4+x5) / 2, thereby obtaining the second optimal subsequence {x′3,x4,x′5}; for the third initial subsequence {x6,x7,x8}, its boundary x6 is updated to x′6=(x5+x6) / 2, and its boundary x8 is updated to x′8=(x7+x8) / 2, thereby obtaining the third optimal subsequence {x′6,x7,x′8}; and so on, until obtaining An optimal subsequence.

[0119] In this step, for each optimal subsequence, a sine function and a cosine function are preferably used to generate the position coding information corresponding to the optimal subsequence. The expression of the sine function is as follows:

[0120]

[0121] The expression of the cosine function is as follows:

[0122]

[0123] In the formula, PE(pos,2r) and PE(pos,2r+1) are the position encoding data calculated by different functions, pos is the position index, r is the dimension index, and d is the position index. model is the dimension of the position coding information, and this application preferably sets d model =2.

[0124] For example, for the first optimal subsequence {x0, x1, x2}, the value range of the dimension index r is [0, 1, d model -1], the value range of the position index pos is [0,1,2], which corresponds to the data subscript of the first optimal subsequence {x0,x1,x2}. It is first determined that only when xr=0 can the values ​​of 2r and 2r+1 fall within the value range of the dimension index r, and then the relevant position encoding data is calculated through the above sine function and the above cosine function, which is specifically expressed as follows:

[0125] When pos=0:

[0126]

[0127] When pos=1:

[0128]

[0129] When pos=2:

[0130]

[0131] Therefore, it can be determined that the position encoding information corresponding to the first optimal subsequence {x0,x1,x2} includes [0,1], [0.8415,0.5403] and [0.9093,-0.4161].

[0132] In steps S401 to S402 shown in the embodiment of the present application, the current optimal multi-scale time series is divided into multiple scales, so that the subsequent model can capture the features at different levels and time scales, so as to enhance the model's overall understanding and prediction capabilities of the changes in the seabed topography; by performing sliding average processing on the boundaries of the initial sub-sequences obtained by the division, it is possible to avoid feature mutations at the boundaries that may have a negative impact on the subsequent model predictions; by position encoding the optimal sub-sequence obtained by optimization, it is possible for the subsequent model to capture the sequential characteristics of the changes in the seabed topography.

[0133] In some embodiments, the above step S106 may include but is not limited to steps S501 to S505:

[0134] Step S501, input multiple groups of subsequences into a seabed deformation prediction model based on an adaptive multi-head attention mechanism, wherein the multiple groups of subsequences include several optimal subsequences corresponding to several position coding information. The seabed deformation prediction model based on the adaptive multi-head attention mechanism mainly includes an encoder network, an adaptive multi-head attention network and a feedforward neural network, and multiple attention heads are set in the adaptive multi-head attention network.

[0135] Step S502: extract features from a plurality of optimal subsequences respectively through the encoder network to obtain a plurality of initial potential feature data related to the change of the seabed topography.

[0136] In this step, for each optimal subsequence, the optimal subsequence is input into the encoder network for feature extraction to obtain initial potential feature data, which can be determined by the following expression:

[0137] P = Sigmoid(W1X′+b1);

[0138] Where P is the initial potential feature data, X′ is the optimal subsequence, W1 is the weight matrix of the encoder network, b1 is the bias of the encoder network, Sigmoid is the activation function, and ReLU function can also be used.

[0139] Step S503: Generate a plurality of final latent feature data according to a plurality of initial latent feature data and a plurality of position coding information.

[0140] In this step, a number of initial latent feature data and a number of position coding information are concatenated correspondingly to obtain a number of intermediate latent feature data; and then a number of intermediate latent feature data are linearly transformed respectively to obtain a number of final latent feature data.

[0141] Specifically, for each initial latent feature data, the initial latent feature data and its corresponding position coding information are concatenated to obtain intermediate latent feature data, which can be determined by the following expression:

[0142] P′=[P;PE];

[0143] Then the intermediate latent feature data is linearly transformed to obtain the final latent feature data, which can be determined by the following expression:

[0144] P″=W3P′+b3;

[0145] Wherein, P′ is the intermediate potential feature data, PE is the position encoding information corresponding to the initial potential feature data P, [;] is the concatenation symbol, P″ is the final potential feature data, W3 is the preset first weight matrix, and b3 is the preset first bias.

[0146] By concatenating and linearly transforming the initial potential feature data with its corresponding position encoding information, subsequent models can better learn and understand the diversity and complexity of seabed topography changes, which helps to improve the model's predictive ability and ensure data consistency in subsequent calculations.

[0147] Step S504: extract features from a number of final potential feature data respectively through the adaptive multi-head attention network to obtain a number of attention feature data related to the change of the seabed topography.

[0148] In this step, for each final potential feature data, the final potential feature data is input into the adaptive multi-head attention network for feature extraction to obtain attention feature data. The corresponding implementation methods include the following:

[0149] First, the final potential feature data is processed by multiple attention heads to obtain multiple attention outputs, which can be determined by the following expression:

[0150]

[0151]

[0152] Secondly, multiple attention outputs are weighted and summed to obtain the final attention output, which can be determined using the following expression:

[0153]

[0154] Finally, the final attention output is residually connected and normalized with the final potential feature data to obtain the attention feature data, which can be determined by the following expression:

[0155] U′ end =LayerNorm(U end +P″);

[0156] In the formula, Q h is the query matrix corresponding to the h-th attention head, is the query weight matrix corresponding to the h-th attention head, K h is the key matrix corresponding to the h-th attention head, is the key weight matrix corresponding to the h-th attention head, V h is the value matrix corresponding to the h-th attention head, is the value weight matrix corresponding to the hth attention head, d k is the key matrix K h The dimension of the query matrix Q h The dimension and value matrix V h The dimension is also d k , K j is the key matrix K h The j-th dimension data contained in h The data in the jth column is contained in k , T is the transposition symbol, A hj is the query matrix Q h Key matrix K h The j-th dimension data K contained in j The attention score, α hj is the weight value, U h is the attention output obtained by the h-th attention head processing the final potential feature data P″, U end is the final attention output, H is the number of multiple attention heads, and β h is the weight coefficient corresponding to the h-th attention head, which represents the feature contribution corresponding to the h-th attention head, h = 1, 2, ..., H, U′ endis the attention feature data, and LayerNorm is the normalization function.

[0157] By introducing an adaptive multi-head attention mechanism, different attention heads are used to learn different feature representations, and weight coefficients are used to adjust the contribution of attention heads. This allows the model to focus on more important attention heads to learn more important feature representations, thereby more accurately capturing the dynamic characteristics of seabed topography changes and enhancing the model's ability to predict seabed topography deformation.

[0158] Step S505: A plurality of attention feature data are fused and analyzed through the feedforward neural network to obtain the seabed topography prediction data at a future moment.

[0159] In this step, the feedforward neural network includes an input layer, four hidden layers and an output layer, each hidden layer is used to map the input to a new space through a linear transformation, and the specific implementation method includes:

[0160] Through the input layer, several attention feature data are concatenated to obtain a feature data to be tested;

[0161] The feature data to be tested is passed to four hidden layers for processing in turn to obtain the final feature data, which can be determined by the following expression:

[0162] Hidden1=ReLU(W4·Input FFN +b4),

[0163] Hidden2=ReLU(W5·Hidden1+b5),

[0164] Hidden3=ReLU(W6·Hidden2+b6),

[0165] Hidden4=ReLU(W7·Hidden3+b7);

[0166] The final feature data is processed through the output layer to obtain the predicted seabed topography data at the future moment, which can be determined by the following expression:

[0167] Y=W8·Hidden4+b 8 / ;

[0168] In the formula, Input FFNis the feature data to be tested, W4 is the weight matrix of the first hidden layer, b4 is the bias of the first hidden layer, Hidden1 is the output of the first hidden layer, W5 is the weight matrix of the second hidden layer, b5 is the bias of the second hidden layer, Hidden2 is the output of the second hidden layer, W6 is the weight matrix of the third hidden layer, b6 is the bias of the third hidden layer, Hidden3 is the output of the third hidden layer, W7 is the weight matrix of the fourth hidden layer, b7 is the bias of the fourth hidden layer, Hidden4 is the output of the fourth hidden layer, that is, the final feature data, W8 is the weight matrix of the output layer, b8 is the bias of the output layer, and Y is the seabed topography prediction data.

[0169] Steps S501 to S505 shown in the embodiment of the present application can effectively improve the prediction accuracy of seabed topography changes by inputting multiple groups of subsequences into the encoder network, the adaptive multi-head attention network and the feedforward neural network for analysis in sequence.

[0170] In some embodiments, the training process for the seabed deformation prediction model is mainly divided into two parts. The first part is to further train the encoder network, and the second part is to use the trained encoder network, the adaptive multi-head attention network and the feedforward neural network to build an initial seabed deformation prediction model and then perform overall training. In this overall training process, only the adaptive multi-head attention network and the feedforward neural network are updated with parameters.

[0171] Specifically, the training process for the encoder network includes the following:

[0172] The encoder network and the additional decoder network are used to build the initial autoencoder model. The data processing process inside the autoencoder model can be expressed as follows:

[0173] Y m =Sigmoid(W1X in +b1),

[0174] Y out =Sigmoid(W2Y m +b2);

[0175] In the formula, X in is the input of the encoder network, that is, the input of the autoencoder model, Y m is the output of the encoder network, W2 is the weight matrix of the decoder network, b2 is the bias of the decoder network, and Y out is the output of the decoder network, that is, the output of the autoencoder model.

[0176] A first training data set is obtained, and the first training data set can be obtained in the following manner: determining a number of different historical time periods; for each historical time period, obtaining the historical seabed topography historical data collected for the same seabed area to be measured in the historical time period and forming a corresponding historical multi-scale time series; decomposing the historical multi-scale time series to obtain a historical trend sequence, a historical seasonal sequence, and a historical residual sequence related to the change of the seabed topography; performing gradient descent optimization on the historical residual sequence to obtain a historical optimal residual sequence; generating a historical optimal multi-scale time series according to the historical trend sequence, the historical seasonal sequence, and the historical optimal residual sequence; performing multi-scale division and encoding on the historical optimal multi-scale time series to obtain multiple groups of historical sub-sequences; taking the multiple groups of historical sub-sequences corresponding to each historical time period as a training sample, thereby forming the first training data set.

[0177] Determine the reconstruction loss function, which preferably adopts the mean square error loss function, which is mainly used to characterize the input X of the autoencoder model in With output Y out The difference between.

[0178] When the autoencoder model is forward trained using the first training data set, the reconstruction loss function is used to optimize the first network parameters contained in the autoencoder model, the first network parameters including two weight matrices (W1, W2) and two biases (b1, b2), and the first network parameters are updated through a back propagation algorithm until the reconstruction loss function reaches a convergence state, thereby obtaining a trained autoencoder model, and then obtaining a trained encoder network therefrom.

[0179] Specifically, the overall training process for the seabed deformation prediction model includes the following:

[0180] A second training data set is obtained, and the second training data set is updated based on the first training data set. Specifically, for each historical time period, historical seabed topography data collected for the same seabed area to be measured at the next moment of the historical time period is obtained, and the next moment of the historical time period is taken as the future moment, and the historical seabed topography data at the next moment of the historical time period is taken as the real seabed topography data at the future moment, and the real seabed topography data at the future moment is taken as the real label carried by the multiple groups of historical subsequences corresponding to the historical time period; that is, the second training data set can be obtained by assigning a corresponding real label to each training sample contained in the first training data set through the above implementation method.

[0181] A training loss function is determined. The training loss function may adopt a mean square error function or a mean absolute error function, which is mainly used to characterize the difference between the output of the seabed deformation prediction model and the true label.

[0182] When the seabed deformation prediction model is forward trained using the second training data set, the training loss function is used to optimize the second network parameters included in the seabed deformation prediction model, wherein the second network parameters include five weight matrices (W4, W5, W6, W7, W8), five biases (b4, b5, b6, b7, b8) and weight coefficients corresponding to multiple attention heads (β1, β2, ..., β H ), update the second network parameters through the back propagation algorithm or the Adam optimization algorithm until the training loss function reaches a convergence state, and then obtain a trained seabed deformation prediction model.

[0183] Exemplarily, the training loss function can be expressed by the following expression:

[0184]

[0185] Where, L final is the training loss function, S is the number of all training samples contained in the second training data set, T s is the true label corresponding to the s-th training sample contained in the second training data set, Y s is the prediction result obtained by analyzing the sth training sample by the seabed deformation prediction model;

[0186] The second network parameters are updated by the back propagation algorithm, which is specifically performed as follows:

[0187]

[0188]

[0189]

[0190] In the formula, is the loss gradient, W p ′ is the updated weight matrix, b′ p is the updated bias, p = 4, 5, 6, 7, 8, β′ h is the updated weight coefficient corresponding to the h-th attention head, U h is the output of the hth attention head, and η is the learning rate.

[0191] The method for predicting seabed topography deformation proposed in the embodiment of the present application decomposes the current trend sequence, the current seasonal sequence and the current residual sequence from the current multi-scale time series formed by the seabed topography data collected at different times, and then performs gradient descent optimization on the current residual sequence covering the subtle features of the seabed topography changes, and then reintegrates it together with the current trend sequence and the current seasonal sequence to generate the current optimal multi-scale time series. Finally, the current optimal multi-scale time series is multi-scale divided and encoded, and the prediction analysis is completed through the seabed deformation prediction model. In the prediction analysis process, an adaptive multi-head attention mechanism is introduced to more comprehensively capture the characteristics of seabed topography changes, thereby improving the prediction accuracy of seabed topography deformation.

[0192] See also Figure 2 The present application also provides a schematic diagram of a submarine terrain deformation prediction system, which can implement the above-mentioned submarine terrain deformation prediction method. The system includes:

[0193] The acquisition module 601 is used to acquire the seafloor topography data collected at the current moment and multiple historical moments to form a current multi-scale time series;

[0194] The multiple historical moments are all before the current moment, and the multiple historical moments and the current moment are continuous moments. The seabed topography data collected at each moment includes the height change values ​​of multiple different positions in the seabed area to be measured relative to the sea level at the moment, which can be collected by arranging displacement sensors at multiple different positions in the seabed area to be measured;

[0195] A decomposition module 602 is used to decompose the current multi-scale time series to obtain a current seasonal series, a current trend series and a current residual series related to the change of the seabed topography;

[0196] The optimization module 603 is used to perform gradient descent optimization on the current residual sequence to obtain the current optimal residual sequence;

[0197] A generating module 604 is used to generate a current optimal multi-scale time series according to the current seasonal sequence, the current trend sequence and the current optimal residual sequence;

[0198] A division module 605 is used to perform multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences;

[0199] The prediction module 606 is used to input multiple groups of subsequences into the seabed deformation prediction model based on the adaptive multi-head attention mechanism for analysis to obtain the seabed topography prediction data at the future moment.

[0200] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0201] The embodiment of the present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method for predicting seabed terrain deformation when executing the computer program. The electronic device may include any intelligent terminal such as a tablet computer and a car computer.

[0202] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0203] See also Figure 3 , Figure 3 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0204] The processor 701 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0205] The memory 702 may be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 may store an operating system and other application programs. When the technical solution provided in the embodiment of the present application is implemented by software or firmware, the relevant program code is stored in the memory 702, and the processor 701 calls and executes the technical solution provided in the embodiment of the present application;

[0206] Input / output interface 703, used to implement information input and output;

[0207] Communication interface 704, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0208] A bus 705 that transmits information between the various components of the device (e.g., the processor 701, the memory 702, the input / output interface 703, and the communication interface 704);

[0209] The processor 701 , the memory 702 , the input / output interface 703 and the communication interface 704 are connected to each other in communication within the device via the bus 705 .

[0210] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned seabed terrain deformation prediction method is implemented.

[0211] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0212] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0213] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0214] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0215] The system embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0216] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0217] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0218] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0219] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0220] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0221] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0222] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0223] The preferred embodiments of the present application are described above with reference to the accompanying drawings, but the scope of the rights of the present application is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present application should be within the scope of the rights of the present application.

Claims

1. A method for predicting seabed topography deformation, characterized in that: The method comprises: Acquire seabed topographic data collected at the current moment and multiple historical moments to form a current multi-scale time series; wherein the seabed topographic data includes height change values ​​of multiple different positions in the seabed area to be measured relative to the sea level; Decomposing the current multi-scale time series to obtain a current trend series, a current seasonal series, and a current residual series related to changes in seafloor topography; Performing gradient descent optimization on the current residual sequence to obtain a current optimal residual sequence; Generate a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence; Performing multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences; The multiple groups of subsequences are input into a seabed deformation prediction model based on an adaptive multi-head attention mechanism for analysis to obtain seabed topography prediction data at future moments.

2. The method for predicting seabed topography deformation according to claim 1, characterized in that: The step of performing gradient descent optimization on the current residual sequence to obtain the current optimal residual sequence comprises: Using the current residual sequence as the current residual sequence to be optimized; Calculating a gradient value of a loss function on the current residual sequence to be optimized, wherein the loss function is used to measure the difference between the current multiscale time series and the optimized current multiscale time series, wherein the optimized current multiscale time series is obtained by optimizing the current residual sequence contained in the current multiscale time series; Optimizing the current residual sequence to be optimized by using the gradient value to obtain an optimized current residual sequence; Determine whether the preset iteration end condition is met; if so, use the optimized current residual sequence as the current optimal residual sequence; if not, use the optimized current residual sequence as the current residual sequence to be optimized, and then return to the step of calculating the gradient value generated by the loss function for the current residual sequence to be optimized.

3. The method for predicting seabed topography deformation according to claim 1, characterized in that: The generating the current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence comprises: Obtain a historical trend sequence at a first historical moment and a historical seasonal sequence at a second historical moment, wherein the first historical moment is the moment before the current moment, and the time interval between the second historical moment and the current moment is a seasonal cycle; According to the difference between the current trend sequence and the historical trend sequence, the current trend sequence is optimized to obtain a current optimal trend sequence; According to the difference between the current seasonal sequence and the historical seasonal sequence, the current seasonal sequence is optimized to obtain a current optimal seasonal sequence; The current optimal trend sequence, the current optimal seasonal sequence and the current optimal residual sequence are fused to obtain the current optimal multi-scale time series.

4. The method for predicting seabed topography deformation according to claim 1, characterized in that: The multi-scale division and encoding of the current optimal multi-scale time series to obtain multiple groups of subsequences includes: According to the preset scale range, a plurality of different scale factors are determined; For each of the scale factors, dividing the current optimal multi-scale time series according to the scale factor to obtain all initial subsequences; According to the current optimal multi-scale time series, performing sliding average processing on the boundaries of all the initial subsequences to obtain all the optimal subsequences; After position encoding is performed on all the optimal subsequences respectively, a group of subsequences is formed.

5. The method for predicting seabed topography deformation according to claim 4, characterized in that: The seabed deformation prediction model based on the adaptive multi-head attention mechanism includes an encoder network, an adaptive multi-head attention network and a feedforward neural network; the multiple groups of subsequences include several optimal subsequences, and the multiple optimal subsequences correspond to several position encoding information; the multiple groups of subsequences are input into the seabed deformation prediction model based on the adaptive multi-head attention mechanism for analysis, and the seabed terrain prediction data at the future moment is obtained, including: Using the encoder network to extract features from the plurality of optimal subsequences respectively, to obtain a plurality of initial potential feature data related to the seabed topography change; Generating a plurality of final latent feature data according to the plurality of initial latent feature data and the plurality of position coding information; Using the adaptive multi-head attention network to extract features from the plurality of final potential feature data, respectively, to obtain a plurality of attention feature data related to changes in seabed topography; The feedforward neural network is used to perform fusion analysis on the plurality of attention feature data to obtain the seabed topography prediction data at the future moment.

6. The method for predicting seabed topography deformation according to claim 5, characterized in that: The step of generating a plurality of final latent feature data according to the plurality of initial latent feature data and the plurality of position coding information comprises: Correspondingly concatenating the plurality of initial latent feature data and the plurality of position coding information to obtain a plurality of intermediate latent feature data; The plurality of intermediate latent feature data are respectively linearly transformed to obtain the plurality of final latent feature data.

7. The method for predicting seabed topography deformation according to claim 5, characterized in that: The adaptive multi-head attention network is provided with a plurality of attention heads; the adaptive multi-head attention network is used to extract features of the plurality of final potential feature data respectively, and the plurality of attention feature data related to the change of the seabed topography is obtained, including: For each of the final latent feature data, use the multiple attention heads to process the final latent feature data respectively to obtain multiple attention outputs; Performing weighted summation on the multiple attention outputs to obtain a final attention output; The final attention output is residually connected and normalized with the final potential feature data to obtain attention feature data.

8. A seabed terrain deformation prediction system, characterized in that: The system comprises: An acquisition module is used to acquire seafloor topographic data collected at the current moment and multiple historical moments to form a current multi-scale time series; wherein the seafloor topographic data includes height change values ​​of multiple different positions in the seafloor area to be measured relative to the sea level; A decomposition module, used for decomposing the current multi-scale time series to obtain a current trend sequence, a current seasonal sequence and a current residual sequence related to the change of the seabed topography; An optimization module, used for performing gradient descent optimization on the current residual sequence to obtain a current optimal residual sequence; A generating module, used for generating a current optimal multi-scale time series according to the current trend sequence, the current seasonal sequence and the current optimal residual sequence; A division module, used for performing multi-scale division and encoding on the current optimal multi-scale time series to obtain multiple groups of sub-sequences; The prediction module is used to input the multiple groups of subsequences into a seabed deformation prediction model based on an adaptive multi-head attention mechanism for analysis to obtain seabed topography prediction data at a future moment.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images