A method and system for fusion analysis of marine meteorological data

Abstract

The present invention discloses a method and system for fusion analysis of marine meteorological data, the method comprising: step S1: establishing a marine meteorological data set; step S2: preprocessing the marine meteorological image data and the marine meteorological time series data respectively; step S3: constructing a marine meteorological data fusion analysis model; step S4: training the marine meteorological data fusion analysis model; step S5: acquiring the marine meteorological image data and the marine meteorological time series data of the target marine area; step S6: preprocessing the image data and the time series data of the target marine area, and inputting them into the marine meteorological data fusion analysis model; step S7: visually displaying the analysis results. The present invention makes full use of the complementarity of image data and time series data, realizes the effective combination and deep analysis of marine meteorological image data and time series data, and can comprehensively and accurately analyze marine meteorology.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological data fusion processing, and in particular relates to a method and system for fusion analysis of marine meteorological data. Background Art

[0002] With the development of science and technology and the in-depth exploration of the field of marine meteorology by human beings, the means of collecting marine meteorological data are becoming increasingly rich, and the amount of data is also showing explosive growth. These data not only include traditional time series data, such as temperature, humidity, wind speed and other measurement values, but also include a large amount of satellite and radar image data. These data provide us with valuable marine and meteorological information, which is of great value to the fields of weather forecasting, marine environment monitoring, climate research, etc.

[0003] However, how to effectively fuse and analyze these diverse data has become an urgent problem to be solved. Traditional data fusion methods often focus on a single type of data analysis, such as processing only time series data or image data, while ignoring the complementarity between the two types of data. In fact, time series data and image data each have unique information: time series data provides continuous and quantitative observations, while image data provides intuitive information on spatial distribution and morphological changes. At present, although some methods have attempted to combine the two types of data for analysis, they often fail to fully explore the intrinsic connection and complementarity between the two types of data, resulting in unsatisfactory analysis results. Especially in the complex and changeable marine meteorological environment, accurate data analysis is particularly important. Summary of the invention

[0004] In view of the defects existing in the above-mentioned prior art, the present invention provides a method for fusion analysis of marine meteorological data, the method comprising:

[0005] Step S1: establishing a marine meteorological data set, wherein the marine meteorological data set includes marine meteorological image data and marine meteorological time series data;

[0006] Step S2: preprocessing the marine meteorological image data and the marine meteorological time series data respectively;

[0007] Step S3: constructing a marine meteorological data fusion analysis model, wherein the marine meteorological data fusion analysis model includes a CNN module, an LSTM module, a hybrid domain attention mechanism module, and a feature fusion module;

[0008] The hybrid domain attention mechanism module includes a spatial attention module and a channel attention module; the output end of the CNN module is connected to the spatial attention module, the output end of the LSTM module is connected to the channel attention module, and the output end of the spatial attention module and the output end of the LSTM module are connected to the feature fusion module;

[0009] Step S4: training the marine meteorological data fusion analysis model based on the marine meteorological data set, and iterating until the loss function converges;

[0010] The pre-processed marine meteorological image data is input into the CNN module of the marine meteorological data fusion analysis model, and the pre-processed marine meteorological time series data is input into the LSTM module of the marine meteorological data fusion analysis model;

[0011] Step S5: Acquire marine meteorological image data and marine meteorological time series data of the target ocean area;

[0012] Step S6: pre-processing the marine meteorological image data and the marine meteorological time series data of the target ocean area, and inputting them into the trained marine meteorological data fusion analysis model to obtain the marine meteorological data fusion analysis results;

[0013] Step S7: Visually display the marine meteorological data fusion analysis results.

[0014] Preferably, in step S2, the preprocessing of the marine meteorological image data and the marine meteorological time series data respectively specifically includes:

[0015] S21: flipping and scaling the marine meteorological image data to achieve enhanced processing of the marine meteorological image data;

[0016] S22: performing noise removal and normalization processing on the marine meteorological time series data, and supplementing the missing values ​​in the marine meteorological time series data by using an interpolation filling method;

[0017] S23: aligning the timestamps of the marine meteorological image data and the marine meteorological time series data to ensure that each marine meteorological time series data point and its corresponding marine meteorological image data have the same timestamp.

[0018] Preferably, the CNN module performs feature extraction on the marine meteorological image data and obtains an image feature map, the spatial attention module calculates the score of each spatial position in the image feature map based on the marine meteorological time series data, and uses the sigmoid activation function to convert the score into a weight value between 0 and 1 to obtain a spatial attention weight map, and the spatial attention weight map and the image feature map are multiplied element by element to obtain the image features of the marine meteorological image data after adding spatial attention.

[0019] Preferably, the spatial attention module calculates the score of each spatial position in the image feature map based on the marine meteorological time series data, specifically including:

[0020] Represent the marine meteorological time series data as a vector of length 5: i , cosθ i , sinθ i ,s i , t i );where v i represents wind speed, cosθ i and sinθ i Indicates wind direction, s i Indicates air pressure, t i represents temperature, i represents the i-th time series data;

[0021] Input the vector into a fully connected layer, perform linear transformation through a weight matrix and a bias term, and perform nonlinear transformation through an activation function ReLU to obtain an output vector whose dimension matches the image feature map;

[0022] The output vector is multiplied by the image feature map, and normalized, and the normalized result is used as the score of each spatial position in the image feature map.

[0023] Preferably, the LSTM module performs feature extraction on the marine meteorological time series data and obtains time series features, and the channel attention module calculates channel attention weights based on information of different channels in the time series features, and multiplies the channel attention weights by the time series features to obtain the time series features of the marine meteorological time series data after adding channel attention.

[0024] Preferably, the feature fusion module is used to fuse the image features output by the spatial attention module and the temporal features output by the channel attention module, specifically including:

[0025] The size of the image feature is H×W×C, where H is the height, W is the width, and C is the number of channels. For each temporal feature, it is converted into a vector of H×W×1 through a fully connected layer, and then replicated C times to obtain the number of channels matching the image feature, thereby obtaining a temporal embedding feature;

[0026] The temporal embedding feature is concatenated with the image feature to obtain a feature fusion result.

[0027] Preferably, the loss function L loss Including the loss function L corresponding to the marine meteorological image data loss1 The loss function L corresponding to the marine meteorological time series dataloss2 , specifically:

[0028] L loss =ω1L loss1 +ω2L loss2 ;

[0029]

[0030]

[0031] Among them, ω1 and ω2 are weight coefficients, p(x i ) is the true probability distribution of image features in marine meteorological image data, q(x i ) is the predicted probability distribution of image features in marine meteorological image data, i is a pixel sample; n is the number of pixel samples; y is the true value of the marine meteorological time series data sample, f(x) is the predicted value of the marine meteorological time series data sample, and σ is a hyperparameter.

[0032] The present invention also discloses a marine meteorological data fusion analysis system, the system comprising:

[0033] A data acquisition module, wherein the data acquisition module is used to acquire the marine meteorological image data and the marine meteorological time series data;

[0034] A fusion analysis module, wherein the fusion analysis module is used to perform fusion analysis on the marine meteorological image data and the marine meteorological time series data;

[0035] A visualization platform, which is used to visualize the analysis results of the fusion analysis module.

[0036] The present invention also discloses a computer-readable storage medium, wherein the storage medium stores computer instructions, and when the computer instructions are executed by a processor, the marine meteorological data fusion analysis method is implemented.

[0037] The present invention also discloses a marine meteorological data fusion analysis device, comprising a memory and a processor, wherein the memory stores computer program code, and the processor is used to execute the computer program code to implement the marine meteorological data fusion analysis method.

[0038] The present invention uses CNN convolutional neural network to process marine meteorological image data, LSTM long short-term memory recursive neural network to process marine meteorological time series data, and performs feature fusion. The analysis results combine the spatial information of marine meteorological image data and the temporal dynamics of marine meteorological time series data, providing more comprehensive and accurate marine meteorological analysis results. The present invention introduces spatial attention mechanism and channel attention mechanism respectively, wherein the spatial attention mechanism enables the model to focus on the key areas of marine meteorological image data, improves the recognition ability of important image features, and the channel attention mechanism assigns larger weights to channels containing significant information through feature weighting, thereby improving the effectiveness of the temporal features of marine meteorological time series data. The present invention integrates marine meteorological time series information when constructing the spatial attention weight map, and further integrates image features and temporal features during feature fusion. The complementarity of image data and time series data is fully utilized through a two-step fusion method, and the effective combination and deep analysis of image data and time series data are realized, and the model's understanding and prediction capabilities of complex marine meteorological phenomena are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0040] Figure 1 is a flow chart showing a method for fusion analysis of marine meteorological data according to an embodiment of the present invention.

[0041] Figure 2 A marine meteorological data fusion analysis model according to an embodiment of the present invention is shown.

[0042] Figure 3 The invention discloses a marine meteorological data fusion analysis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings, and "multiple" generally includes at least two.

[0045] It should be understood that although the terms first, second, third, etc. may be used to describe ... in the embodiments of the present invention, these ... should not be limited to these terms. These terms are only used to distinguish .... For example, without departing from the scope of the embodiments of the present invention, the first ... may also be referred to as the second ..., and similarly, the second ... may also be referred to as the first ....

[0046] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0047] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0048] It should also be noted that the term "includes", "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, so that a commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprising a ..." do not exclude the existence of other identical elements in the commodity or device including the elements.

[0049] like Figure 1 As shown, the present invention discloses a method for fusion analysis of marine meteorological data, the method comprising:

[0050] Step S1: Establishing a marine meteorological data set, wherein the marine meteorological data set includes marine meteorological image data and marine meteorological time series data.

[0051] Marine meteorological data mainly comes from a variety of observation methods, including meteorological station observation, radar observation, satellite remote sensing observation, and drone observation, etc. In the embodiment of the present invention, the marine meteorological image data mainly includes satellite remote sensing images of ocean temperature, clouds, etc. obtained by satellite remote sensing observation. Marine meteorological time series data mainly includes meteorological station observation data, which measures and records changes in meteorological elements such as temperature, humidity, wind speed, wind direction, air pressure, and precipitation by arranging meteorological observation stations on the ground.

[0052] Step S2: pre-processing the marine meteorological image data and the marine meteorological time series data respectively.

[0053] Since it involves the fusion of two different types of data, marine meteorological image data and marine meteorological time series data, data synchronization must be performed before the data is fused, that is, to ensure that meteorological data (such as wind speed, wind direction, air pressure, etc.) and corresponding image data (such as satellite cloud images, etc.) are correctly matched in time and space. In addition, in order to better fuse the two types of data, the marine meteorological time series data needs to be normalized according to the specific range and distribution, and the marine meteorological image data needs to be adjusted to a uniform size. In addition, when the meteorological station observes the marine meteorological time series data, there will be missing observation data, so it needs to be filled.

[0054] In order to achieve data synchronization and facilitate subsequent model analysis, in step S2, the preprocessing of the marine meteorological image data and the marine meteorological time series data respectively specifically includes:

[0055] S21: flipping and scaling the marine meteorological image data to achieve enhanced processing of the marine meteorological image data;

[0056] S22: performing noise removal and normalization processing on the marine meteorological time series data, and supplementing the missing values ​​in the marine meteorological time series data by using an interpolation filling method;

[0057] S23: aligning the timestamps of the marine meteorological image data and the marine meteorological time series data to ensure that each marine meteorological time series data point and its corresponding marine meteorological image data have the same timestamp.

[0058] Step S3: constructing a marine meteorological data fusion analysis model, wherein the marine meteorological data fusion analysis model includes a CNN module, an LSTM module, a hybrid domain attention mechanism module, and a feature fusion module;

[0059] Among them, the mixed domain attention mechanism module includes a spatial attention module and a channel attention module; the output end of the CNN module is connected to the spatial attention module, the output end of the LSTM module is connected to the channel attention module, and the output end of the spatial attention module and the output end of the LSTM module are connected to the feature fusion module.

[0060] In step S3, the CNN convolutional neural network module is used to extract features from the marine meteorological image data, and the LSTM module is used to extract features from the marine meteorological time series data.

[0061] Step S4: Training the marine meteorological data fusion analysis model based on the marine meteorological data set, iterating until the loss function converges

[0062] In step S4, selecting a suitable loss function can improve the accuracy of the marine meteorological data fusion analysis model. Since the embodiments of the present invention respectively use CNN convolutional neural network to process marine meteorological image data, and LSTM long short-term memory neural network to process marine meteorological time series data, a loss function is constructed correspondingly for the above processing process.

[0063] Step S5: Acquire the marine meteorological image data and marine meteorological time series data of the target ocean area.

[0064] In step S5, according to the target ocean area to be analyzed, the corresponding satellite remote sensing images and the temperature, wind direction, wind speed, air pressure and other data captured by the meteorological observation station are respectively obtained for fusion analysis.

[0065] Step S6: pre-processing the marine meteorological image data and the marine meteorological time series data of the target ocean area, and inputting them into the trained marine meteorological data fusion analysis model to obtain the marine meteorological data fusion analysis results;

[0066] Step S7: Visually display the marine meteorological data fusion analysis results.

[0067] This embodiment uses CNN convolutional neural network to process marine meteorological image data and LSTM long short-term memory recursive neural network to process marine meteorological time series data, and performs feature fusion. The analysis results combine the spatial information of marine meteorological image data and the temporal dynamics of marine meteorological time series data, providing more comprehensive and accurate marine meteorological analysis results.

[0068] like Figure 2As shown, in this embodiment, a CNN convolutional neural network is used to process marine meteorological image data. Marine meteorological image data usually contains a large amount of spatial information, such as cloud distribution, precipitation pattern, etc. CNN can automatically extract spatial features in these images through its convolution layer and pooling layer. These features are crucial for predicting weather patterns, storm paths, etc. In addition, through multi-layer convolution and pooling operations, CNN can gradually abstract higher-level features, which are very useful for subsequent feature fusion and prediction tasks. In this embodiment, the CNN convolutional neural network includes an input layer, a convolution layer, a pooling layer, and a fully connected layer.

[0069] The size of the input layer is determined according to the size of the input marine meteorological image data. In this embodiment, the size of the input marine meteorological image data after preprocessing is 256×256, so the size of the output layer is also set to 256×256.

[0070] Since marine meteorological image data is very complex and contains many different cloud types, precipitation types, etc., more convolutional layers may be needed to extract these complex features. In this embodiment, the number of convolutional layers may be 5. Each convolutional layer includes multiple layers of convolution kernels, specifically 32 3×3 convolution kernels and 32 7×7 convolution kernels. Among them, the 7×7 convolution kernel can capture global features in a larger range, such as the overall cloud type, precipitation area, etc. The 3×3 convolution kernel can capture more local and finer features, such as edges, textures, etc. In marine meteorological images, these features may include details of the clouds, subtle changes in precipitation patterns, etc.

[0071] Pooling layers are added between convolutional layers to reduce the number of parameters and the amount of computation while retaining key features in the image. Commonly used pooling operations include maximum pooling and average pooling. In this embodiment, a combination of maximum pooling and average pooling is used to perform pooling operations.

[0072] After the convolution layer and the pooling layer, a fully connected layer is used to flatten the feature map into a one-dimensional vector and output the final feature representation. The number of nodes in the fully connected layer can be set specifically according to the task requirements. After experimental verification, the optimal number of nodes in this embodiment is set to 256.

[0073] After the CNN convolutional neural network, the image feature map corresponding to the marine meteorological image data can be obtained.

[0074] like Figure 2As shown, in this embodiment, LSTM long short-term memory network is used to process marine meteorological time series data. Marine meteorological time series data contains information about changes in meteorological indicators over time. LSTM can capture the long-term dependencies and dynamic changes in these time series data through its unique memory unit and gating mechanism (including input gate, forget gate and output gate), and analyze meteorological data over a period of time, including changes in indicators such as temperature, humidity, wind speed and air pressure. The pre-processed marine meteorological time series data can be input into the LSTM long short-term memory network to obtain the corresponding time series features.

[0075] When analyzing marine meteorological images, different areas may have different effects on the analysis results. In order to enable the marine meteorological data fusion analysis model to focus on the key areas of marine meteorological image data, such as Figure 2 As shown, in this embodiment, a spatial attention mechanism module is introduced after the CNN convolutional neural network, and the image feature map output by the CNN convolutional neural network is input into the spatial attention mechanism module for processing. The spatial attention module can help the model automatically locate the most important area in the image and perform a more detailed analysis on it.

[0076] For example, when analyzing a storm path, the area at the center of the storm may be the most critical. The spatial attention module can help the model identify and focus on this area more accurately, thereby improving the accuracy of the analysis.

[0077] In the past, when analyzing marine meteorological image data, only spatial information was considered from the marine meteorological image data. However, marine meteorological data has the important characteristic of temporal dynamics, and the above method ignores the important influence of marine meteorological data time series information on the analysis of marine meteorological image data. In order to make full use of marine meteorological data time series information, in this embodiment, the influence of marine meteorological data time series information is introduced when constructing the spatial attention weight map, realizing the first step of fusion of marine meteorological image data information and marine meteorological data time series information.

[0078] Specifically, the spatial attention module calculates the score of each spatial position in the image feature map based on the marine meteorological time series data, and uses the sigmoid activation function to convert the score into a weight value between 0 and 1 to obtain a spatial attention weight map, and multiplies the spatial attention weight map and the image feature map element by element to obtain the image features of the marine meteorological image data after adding spatial attention.

[0079] When calculating the score of each spatial position in the image feature map based on the marine meteorological time series data, the marine meteorological time series data is first represented as a vector with a length of 5: i , cosθ i, sinθ i ,s i , t i );where v i represents wind speed, cosθ i and sinθ i Indicates wind direction, s i Indicates air pressure, t i represents temperature, i represents the i-th time series data;

[0080] Then, the vector is input into the fully connected layer, linearly transformed by the weight matrix and the bias term, and nonlinearly transformed by the activation function ReLU to obtain an output vector whose dimension matches the image feature map;

[0081] Finally, the output vector is multiplied by the image feature map, and normalized, and the normalized result is used as the score of each spatial position in the image feature map.

[0082] After obtaining the score of each spatial position in the image feature map, the score is converted into a weight value between 0 and 1 using the sigmoid activation function, thereby obtaining a spatial attention weight map.

[0083] Marine meteorological time series data include temperature, humidity, wind speed, wind direction, air pressure, precipitation, etc. Under different marine meteorological environments, the importance of each meteorological index to marine meteorological analysis varies. Therefore, in order to enable the marine meteorological data fusion analysis model to focus on more important feature channels, such as Figure 2 As shown in the figure, after the LSTM long short-term memory network, the channel attention mechanism module is introduced. The time series features obtained after the LSTM long short-term memory network processes the marine meteorological time series data are input into the channel attention mechanism module.

[0084] The channel attention module calculates the channel attention weight according to the information of different channels in the time series features, and multiplies the channel attention weight by the time series features to obtain the time series features of the marine meteorological time series data after adding the channel attention.

[0085] like Figure 2 As shown, in order to make full use of the spatial information and temporal fluidity information of marine meteorological data, after obtaining the image features corresponding to the marine meteorological image data and the temporal features corresponding to the marine meteorological time series data respectively, both are input into the feature fusion module to obtain the final fusion analysis results.

[0086] In this embodiment, when performing feature fusion, it is necessary to normalize the image features and time series features. The size of the image features is H×W×C, where H is the height, W is the width, and C is the number of channels. In order to fuse the image features and time series features, for each time series feature, it is converted into a vector of H×W×1 through a fully connected layer, and then it is replicated C times to obtain the number of channels that matches the image feature, thereby obtaining a time series embedding feature;

[0087] The time series embedding feature is spliced ​​with the image feature to obtain a feature fusion result. This step performs the second step fusion of the marine meteorological image data information and the marine meteorological data time series information. The above feature fusion result contains both the spatial information corresponding to the marine meteorological image data and the time series information corresponding to the marine meteorological time series data, realizing the comprehensive feature representation of the marine meteorological information.

[0088] After building the marine meteorological data fusion analysis model, it needs to be trained to obtain an accurate model. First, the established marine meteorological data set is divided into a training set, a data set, and a validation set. The training set is used to train the model, and the validation set and the test set are used to evaluate the performance of the model. During the evaluation process, attention can be paid to the model's ability to capture key meteorological phenomena, meteorological changes, and so on.

[0089] Secondly, during the training process, a loss function needs to be set to ensure model convergence. In this embodiment, since the model uses CNN convolutional neural network and LSTM long short-term memory network to process marine meteorological image data and marine meteorological time series data respectively, the loss function L corresponding to the model is set. loss Including the loss function L corresponding to the marine meteorological image data loss1 The loss function L corresponding to the marine meteorological time series data loss2 , specifically:

[0090] L loss =ω1L loss1 +ω2L loss2 ;

[0091]

[0092]

[0093] Among them, ω1 and ω2 are weight coefficients, p(x i ) is the true probability distribution of image features in marine meteorological image data, q(x i) is the predicted probability distribution of image features in marine meteorological image data, i is a pixel sample; n is the number of pixel samples; y is the true value of the marine meteorological time series data sample, f(x) is the predicted value of the marine meteorological time series data sample, and σ is a hyperparameter.

[0094] Figure 3 This is another embodiment of the present invention. Figure 3 As shown, this embodiment relates to a marine meteorological data fusion analysis system 300, which is mainly composed of a data acquisition module 310, a fusion analysis module 320 and a visualization platform 330. Among them:

[0095] The data acquisition module 310 is used to obtain the marine meteorological image data and the marine meteorological time series data, establish a marine meteorological data set, and obtain the marine meteorological data of the target ocean area to be analyzed;

[0096] The fusion analysis module 320 includes a marine meteorological data fusion analysis model, which is used to use the model to perform fusion analysis on the marine meteorological image data and the marine meteorological time series data of the target ocean area;

[0097] The visualization platform 330 is used to visualize the analysis results of the fusion analysis module so that users can intuitively understand the meteorological conditions of the target ocean area.

[0098] Among them, the visualization platform 330 uses known virtual reality technology, WebGIS technology, etc. to map the feature fusion results output by the fusion analysis module 320 into scatter plots, color spots, flow field maps, animations and other forms to help users quickly obtain key information and improve decision-making efficiency and accuracy.

[0099] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0100] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0101] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0102] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0103] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the name of a unit does not, in some cases, limit the unit itself.

[0104] The above introduces the preferred embodiments of the present invention, which is intended to make the spirit of the present invention clearer and easier to understand, but is not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection outlined by the claims attached to the present invention.

Claims

1. A method for fusion analysis of marine meteorological data, the method comprising: Step S1: establishing a marine meteorological data set, wherein the marine meteorological data set includes marine meteorological image data and marine meteorological time series data; Step S2: preprocessing the marine meteorological image data and the marine meteorological time series data respectively; Step S3: constructing a marine meteorological data fusion analysis model, wherein the marine meteorological data fusion analysis model includes a CNN module, an LSTM module, a hybrid domain attention mechanism module, and a feature fusion module; The hybrid domain attention mechanism module includes a spatial attention module and a channel attention module; the output end of the CNN module is connected to the spatial attention module, the output end of the LSTM module is connected to the channel attention module, and the output end of the spatial attention module and the output end of the channel attention module are connected to the feature fusion module; Step S4: training the marine meteorological data fusion analysis model based on the marine meteorological data set, and iterating until the loss function converges; The pre-processed marine meteorological image data is input into the CNN module of the marine meteorological data fusion analysis model, and the pre-processed marine meteorological time series data is input into the LSTM module of the marine meteorological data fusion analysis model; Step S5: Acquire marine meteorological image data and marine meteorological time series data of the target ocean area; Step S6: pre-processing the marine meteorological image data and the marine meteorological time series data of the target ocean area, and inputting them into the trained marine meteorological data fusion analysis model to obtain the marine meteorological data fusion analysis results; Step S7: Visually displaying the fusion analysis results of the marine meteorological data; The CNN module performs feature extraction on the marine meteorological image data and obtains an image feature map. The spatial attention module calculates the score of each spatial position in the image feature map based on the marine meteorological time series data, and uses the sigmoid activation function to convert the score into a weight value between 0 and 1 to obtain a spatial attention weight map. The spatial attention weight map and the image feature map are multiplied element by element to obtain the image features of the marine meteorological image data after adding spatial attention.

2. A method for fusion analysis of marine meteorological data according to claim 1, characterized in that: In the step S2, the preprocessing of the marine meteorological image data and the marine meteorological time series data respectively specifically includes: S21: flipping and scaling the marine meteorological image data to achieve enhanced processing of the marine meteorological image data; S22: performing noise removal and normalization processing on the marine meteorological time series data, and supplementing the missing values ​​in the marine meteorological time series data by using an interpolation filling method; S23: aligning the timestamps of the marine meteorological image data and the marine meteorological time series data to ensure that each marine meteorological time series data point and its corresponding marine meteorological image data have the same timestamp.

3. The method for fusion analysis of marine meteorological data according to claim 1, characterized in that: The spatial attention module calculates the score of each spatial position in the image feature map according to the marine meteorological time series data, specifically including: Represent the marine meteorological time series data as a vector of length 5: i , cosθ i , sinθ i ,s i , t i );where v i represents wind speed, cosθ i and sinθ i Indicates wind direction, s i Indicates air pressure, t i represents temperature, i represents the i-th time series data; Input the vector into a fully connected layer, perform linear transformation through a weight matrix and a bias term, and perform nonlinear transformation through an activation function ReLU to obtain an output vector whose dimension matches the image feature map; The output vector is multiplied by the image feature map, and normalized, and the normalized result is used as the score of each spatial position in the image feature map.

4. A method for fusion analysis of marine meteorological data as claimed in claim 2, characterized in that: The LSTM module extracts features from the marine meteorological time series data and obtains time series features. The channel attention module calculates channel attention weights based on information from different channels in the time series features, and multiplies the channel attention weights by the time series features to obtain the time series features of the marine meteorological time series data after adding channel attention.

5. The method for fusion analysis of marine meteorological data according to claim 1, characterized in that: The feature fusion module is used to fuse the image features output by the spatial attention module and the temporal features output by the channel attention module, specifically including: The size of the image feature is H×W×C, where H is the height, W is the width, and C is the number of channels. For each temporal feature, it is converted into a vector of H×W×1 through a fully connected layer, and then replicated C times to obtain the number of channels matching the image feature, thereby obtaining a temporal embedding feature; The temporal embedding feature is concatenated with the image feature to obtain a feature fusion result.

6. The method for fusion analysis of marine meteorological data according to claim 1, characterized in that: The loss function L loss Including the loss function L corresponding to the marine meteorological image data loss1 The loss function L corresponding to the marine meteorological time series data loss2 , specifically: L loss =ω1L loss1 +ω2L loss2 ; Among them, ω1 and ω2 are weight coefficients, p(x i ) is the true probability distribution of image features in marine meteorological image data, q(x i ) is the predicted probability distribution of image features in marine meteorological image data, i is a pixel sample; n is the number of pixel samples; y is the true value of the marine meteorological time series data sample, f(x) is the predicted value of the marine meteorological time series data sample, and σ is a hyperparameter.

7. A marine meteorological data fusion analysis system, the system is implemented based on the marine meteorological data fusion analysis method according to claim 1, characterized in that: The system comprises: A data acquisition module, wherein the data acquisition module is used to acquire the marine meteorological image data and the marine meteorological time series data; A fusion analysis module, wherein the fusion analysis module is used to perform fusion analysis on the marine meteorological image data and the marine meteorological time series data; A visualization platform, which is used to visualize the analysis results of the fusion analysis module.

8. A computer-readable storage medium, wherein the storage medium stores computer instructions, and when the computer instructions are executed by a processor, the marine meteorological data fusion analysis method according to any one of claims 1 to 6 is implemented.

9. A marine meteorological data fusion analysis device, comprising a memory and a processor, characterized in that: The memory stores computer program code, and the processor is used to execute the computer program code to implement a marine meteorological data fusion analysis method as described in any one of claims 1-6.

Images