Three-dimensional sea temperature space downscaling method and device and electronic equipment

Abstract

The invention discloses a three-dimensional sea temperature space downscaling method and device and electronic equipment, and relates to the field of image processing, and the method comprises the steps: obtaining a target data set which comprises wind field data, ocean current data, first sea temperature data and second sea temperature data, and inputting the wind field data, the ocean current data, the first sea temperature data and the second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing, and outputting three-dimensional sea temperature data. A target data set is obtained, wind field data, ocean current data and second sea temperature data are further introduced in the target data set except first sea temperature data, and the wind field data and the ocean current data contain information capable of representing the sea temperature data; and the first sea temperature data, the wind field data, the ocean current data and the second sea temperature data are sent to the three-dimensional sea temperature generation model for downscaling processing, so that the finally output sea temperature data is more accurate.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a method, device and electronic equipment for three-dimensional sea temperature spatial downscaling. Background Art

[0002] In recent years, deep learning (CNN, GAN, Transformer) has demonstrated excellent performance in 2D SST super-resolution. However, 3D super-resolution still requires independent layer-by-layer training, and the number of parameters and the workload for parameter tuning multiply with depth. While many existing models exist, most of them use low-resolution SST data as input, failing to incorporate key dynamic factors such as sea surface wind stress, Ekman suction, and geostrophic current transport. This results in insufficient physical consistency in the reconstruction results and inaccurate SST data. Therefore, a method for accurately acquiring SST data is urgently needed. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a method, apparatus, and electronic device for three-dimensional sea temperature spatial downscaling to solve the problem of how to obtain more accurate sea temperature data.

[0004] According to a first aspect, an embodiment of the present invention provides a method for spatial downscaling of three-dimensional sea temperature, comprising: Acquire a target data set, the target data set comprising: wind field data, ocean current data, first sea temperature data, and second sea temperature data, wherein the resolutions of the wind field data, ocean current data, and first sea temperature data are all smaller than the resolution of the second sea temperature data; The wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set are input into a three-dimensional sea temperature generation model for spatial downscaling processing, and three-dimensional sea temperature data is output, wherein the resolution of the three-dimensional sea temperature data is equal to the resolution of the second sea temperature data in the target data set.

[0005] The three-dimensional sea temperature spatial downscaling method provided in the present application obtains a target data set, removes the first sea temperature data from the target data set, and also introduces wind field data, ocean current data, and second sea temperature data. Since the wind field data and ocean current data also contain information that can characterize the sea temperature data, the first sea temperature data, wind field data, ocean current data, and second sea temperature data are sent to a three-dimensional sea temperature generation model for downscaling processing, which makes the sea temperature data finally output more accurate.

[0006] According to a second aspect, an embodiment of the present invention provides a device for three-dimensional sea temperature spatial downscaling, the device comprising: an acquisition module, configured to acquire a target data set, the target data set comprising: wind field data, ocean current data, first sea temperature data, and second sea temperature data, wherein the resolutions of the wind field data, ocean current data, and first sea temperature data are all smaller than the resolution of the second sea temperature data; A processing module is used to input the wind field data, ocean current data, first sea temperature data and second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing, and output three-dimensional sea temperature data, wherein the resolution of the three-dimensional sea temperature data is equal to the resolution of the second sea temperature data in the target data set.

[0007] The apparatus for three-dimensional sea temperature spatial downscaling provided in an embodiment of the present application acquires a target data set through an acquisition module, sends the target data set to a processing module, and performs spatial downscaling processing using a three-dimensional sea temperature generation model in the processing module. Diversity data is used to characterize more sea temperature details, and spatial downscaling is used to convert low-resolution data into high-resolution data, thereby highlighting more sea temperature details and making the final output sea temperature data more accurate.

[0008] According to a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the three-dimensional sea temperature spatial downscaling method described in the first aspect or any one embodiment of the first aspect by executing the computer instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings: Figure 1 A schematic flow chart of a method for three-dimensional sea temperature spatial downscaling provided in an embodiment of the present application.

[0010] Figure 2 A comparison of the root mean square error (RMS) at depth between the three-dimensional sea temperature spatial downscaling method provided in this embodiment and other methods.

[0011] Figure 3 A comparison of the peak signal-to-noise ratio at depth between the three-dimensional sea temperature spatial downscaling method provided in an embodiment of the present application and other methods.

[0012] Figure 4 A comparison of the mean absolute error in depth between the three-dimensional sea temperature spatial downscaling method provided in an embodiment of the present application and other methods.

[0013] Figure 5 Schematic diagram of the structure of the device for three-dimensional sea temperature spatial downscaling provided in an embodiment of the present application.

[0014] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0015] Reference numerals

[0016] 10 - acquisition module; 11 - processing module; 21 - processor; 20 - memory. DETAILED DESCRIPTION

[0017] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] The present invention provides a method, device and electronic device for spatial downscaling of three-dimensional sea temperature. Figure 1 The figure shows a flow chart of a method for three-dimensional sea temperature spatial downscaling provided by an embodiment of the present application. The method for three-dimensional sea temperature spatial downscaling provided by an embodiment of the present application is applied to a smart terminal or control system. This method uses wind field, ocean current, and sea temperature data as model inputs, and can effectively extract information about atmospheric elements (wind field) and ocean elements (ocean current) to guide the three-dimensional sea temperature spatial downscaling. Compared with downscaling models that only use low-resolution sea temperature as input, the obtained high-resolution sea temperature is more accurate.

[0019] Continue reading Figure 1 The three-dimensional sea temperature spatial downscaling method provided in the embodiment of the present application includes the following steps: S10, obtaining a target data set, the target data set including: wind field data, ocean current data, first sea temperature data and second sea temperature data, the resolution of the wind field data, ocean current data and first sea temperature data are all smaller than the resolution of the second sea temperature data.

[0020] S11, inputting the wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing, and outputting three-dimensional sea temperature data, the resolution of which is equal to the resolution of the second sea temperature data in the target data set.

[0021] In this embodiment, since the wind field data contains local or remote large-scale or mesoscale signals, and the ocean current data contains smaller-scale signals, such as mesoscale eddies or sub-mesoscale processes, these data can highlight the details of the sea temperature data, which is conducive to forming accurate three-dimensional sea temperature data that is close to the real environment.

[0022] Since the wind field data, ocean current data, and the first sea temperature data belong to data of different dimensions, in order to facilitate data processing, the second sea temperature data is introduced as a training label. The second sea temperature data is used as a reference to ensure the consistency of the data format when processing the wind field data, ocean current data, and the first sea temperature data, thereby facilitating or facilitating the data processing capabilities in subsequent steps. The first sea temperature data is low-resolution sea temperature data; the resolution of the second sea temperature data is high-resolution sea temperature data or super-resolution sea temperature data, and the output three-dimensional sea temperature data can be high-resolution sea temperature data or super-resolution sea temperature data.

[0023] In addition, spatial downscaling can convert low-resolution data into high-resolution data, which can obtain more accurate sea temperature data.

[0024] The three-dimensional sea temperature spatial downscaling method provided in an embodiment of the present application obtains a target data set, removes the first sea temperature data from the target data set, and introduces wind field data, ocean current data, and second sea temperature data. Since the wind field data and ocean current data also contain information that can characterize the sea temperature data, the first sea temperature data, wind field data, ocean current data, and second sea temperature data are fed into a three-dimensional sea temperature generation model for downscaling processing, which makes the sea temperature data ultimately output more accurate.

[0025] The three-dimensional sea temperature spatial downscaling method provided in this application, before obtaining the target dataset, further includes: S01, obtaining wind field original data, ocean current original data, first sea temperature original data, and second sea temperature original data in a target sea area.

[0026] S02, performing maximum and minimum normalization on the wind field raw data, the ocean current raw data, the first sea temperature raw data, and the second sea temperature raw data so that the wind field raw data, the ocean current raw data, the first sea temperature raw data, and the second sea temperature raw data tend to the target range, and outputting the wind field data, the ocean current data, the first sea temperature data, and the second sea temperature data.

[0027] In this embodiment, the wind field raw data, ocean current raw data, first sea temperature raw data and second sea temperature raw data can be images or digital data obtained by corresponding sensors for characterizing wind field, ocean current and sea temperature information. For example, the image data can be obtained by a shooting device.

[0028] In this embodiment, the wind field raw data, ocean current raw data, first sea temperature raw data and second sea temperature raw data are subjected to maximum and minimum normalization to reduce the scale difference between different category features, so as to facilitate faster data convergence during subsequent model iterative processing.

[0029] Preferably, the target range of the maximum and minimum normalization can be normalized to between [0, 1].

[0030] In this embodiment, the maximum and minimum normalization is performed, which may be specifically: First, determine the parameters or obtain the target dataset, including: wind field Wind, ocean current SC, low-resolution sea temperature dataset High-resolution sea temperature dataset .

[0031] Each parameter contains n data, and the data format can be: sea temperature and current data are in three-dimensional format, that is, the data format can be [depth level, height, width]; wind field is in two-dimensional format, that is, the data format can be [1, height, width]). .

[0032] in, is the low-resolution sea temperature data (first sea temperature data); is high-resolution sea temperature data (second sea temperature data); is the wind field data; For ocean current data.

[0033] Specifically, let the maximum value in the current data be , the minimum value is , for data The result after normalization Expressed as: .

[0034] The three-dimensional sea temperature spatial downscaling method provided in the present application further requires, before executing step S11, to splice and align the wind field data and the ocean current data, and output the spliced ​​and aligned spliced ​​data.

[0035] In this embodiment, when the depth levels of wind field data and ocean current data are different, it is necessary to unify the wind field data and ocean current data. For example, the ocean current and wind field formats can be (B, C, D, H, W), where B represents Batchsize, C represents the number of channels, D represents the number of depth levels, H represents the vertical resolution, and W represents the horizontal resolution. Since the wind field data has only one layer, D=1, both the wind field and the ocean current have C=1, and after data splicing and alignment, the format of the ocean current and wind field is (B, C, D+1, H, W). It can be understood that in this application, the wind field data and the ocean current data are spliced ​​and aligned to ensure the consistency of subsequent data processing and reduce the complexity of data analysis.

[0036] The three-dimensional sea temperature spatial downscaling method provided by this application, when executing step S11, may be: S111: Obtain a preset noise value.

[0037] In this embodiment, the preset noise value can be obtained from the normal distribution Sampling a noise , can also be obtained based on past experience, such as user-specified random noise.

[0038] S112, sending the wind field data, ocean current data, first sea temperature data and second sea temperature data to a feature extraction network for iterative processing, and outputting a target condition vector.

[0039] Optionally, in step S112, the spliced ​​data, the first sea temperature data, and the second sea temperature data are fed into a feature extraction network for iterative processing, and a target condition vector is output.

[0040] In this embodiment, the feature extraction network can be a cascaded ResBlock, HybridTransformerblock, ResBlock, HybridTransformerblock, and Fusion Layer architecture. Specifically, the feature extraction network can be 4 consecutive ResBlocks, 4 consecutive HybridTransformerblocks, 4 consecutive ResBlocks, 4 consecutive HybridTransformerblocks, and a Fusion Layer layer, and finally the output is a target condition vector of (B, 4, 1024).

[0041] S113 , performing nearest neighbor interpolation processing on the first sea temperature data, and outputting the processed first sea temperature data.

[0042] S114 , sending the preset noise value, the target condition vector, and the processed first sea temperature data into a denoising network for iterative calculation, and outputting three-dimensional sea temperature data.

[0043] Optionally, a mean square error between a preset noise value and a noise value corresponding to the three-dimensional sea temperature data is calculated.

[0044] Specifically, taking the generation of a high-resolution three-dimensional sea temperature image as an example, the implementation process can be as follows: First assumption , is a pre-set sequence, and .make , For the trained feature extraction network and It is a 3DUnet network (noise reduction network).

[0045] Assume wind field Wind, ocean current SC, low-resolution sea temperature data The sea temperature and current data are in three-dimensional format, i.e. [depth level, height, width]; the wind field is in two-dimensional format, i.e. [1, height, width].

[0046] First, users can get the normal distribution Sampling a noise .

[0047] Secondly, according to For each value in the sequence , perform the following steps: (1) Start with the normal distribution Sample a noise with the same size as the second sea temperature data .

[0048] (2) Input the wind field Wind and ocean current SC data into the feature extraction network FEN to obtain the feature information vector .

[0049] (3) Using nearest neighbor interpolation to process low-resolution sea temperature data Improve it to the same resolution as the second sea temperature data to obtain the processed first sea temperature data .

[0050] (4) Calculation of three-dimensional sea temperature data , the specific calculation formula is as follows: .

[0051] Among them, obtaining three-dimensional sea temperature data It may be a diffusion process to remove noise.

[0052] In this embodiment, in order to reduce data interference, the first sea temperature data after the nearest neighbor interpolation processing is The 3DUnet network is mainly used to perform noise reduction processing on the generated diffusion model under the guidance of the characteristic information vector.

[0053] To further improve the feature extraction network Understanding the 3DUnet network, its training process can be: First set up the wind field Wind, ocean current SC, and the first sea temperature dataset With the second SST dataset Contains n data, the sea temperature and current data are in three-dimensional format, i.e. [depth level, height, width], and the wind field is in two-dimensional format, i.e. [1, height, width]), among which, .

[0054] set up , is a pre-set sequence, and .make .

[0055] For each and the corresponding , perform the following steps: (1) From Sample a value in .

[0056] (2) From the normal distribution Sampling a noise .

[0057] (3) Calculation , recorded as .

[0058] (4) Input the trained feature extraction network to obtain the feature information vector ; Use nearest neighbor interpolation Raise it to The same resolution is obtained to obtain the first sea temperature data after processing .

[0059] (5) Calculation and The mean square error of .

[0060] (6) Optimize the error and repeat the above process until convergence to obtain the trained feature extraction network With 3DUnet network .

[0061] Advantages of the three-dimensional sea temperature spatial downscaling method provided in the embodiments of the present application: The three-dimensional sea temperature spatial downscaling method proposed in this method is driven by wind and ocean currents. It can simultaneously take into account the relationship between atmospheric elements and ocean elements and sea temperature, and use this information to drive the three-dimensional sea temperature spatial downscaling to obtain more accurate high-resolution sea temperature data.

[0062] By introducing wind field, ocean current and sea temperature data as model input, it is possible to effectively extract information on atmospheric elements (wind field) and ocean elements (ocean current) to guide the three-dimensional sea temperature spatial downscaling. Compared with the three-dimensional sea temperature generation model that only uses low-resolution sea temperature as input, the use of wind field and ocean current data can obtain more data details, which is conducive to obtaining more accurate and high-resolution sea temperature data (images).

[0063] Compared with the two-dimensional sea temperature spatial downscaling model, the method provided in this application no longer requires a separate training model for the sea temperature at each depth level, which can reduce additional model training time and tedious parameter adjustment process.

[0064] like Figures 2 to 4 As shown, Figure 2 This is a comparison of the root mean square error (RMS) at depth between the three-dimensional sea temperature spatial downscaling method provided in the embodiment of the present application and other methods. Figure 3 This is a comparison of the peak signal-to-noise ratio at depth of the three-dimensional sea temperature spatial downscaling method provided in the embodiment of the present application and other methods. Figure 4 This is a comparison of the mean absolute error of the three-dimensional sea temperature spatial downscaling method provided in this application embodiment and other methods in depth. Figures 2 to 4 Correspondingly, DSSR3 represents the root mean square error, peak signal-to-noise ratio and mean absolute error output at different depths using the model corresponding to the three-dimensional sea temperature spatial downscaling method provided by this application; while Bicubic, MGRO, VSRT, DIFFDS and EDSR3D represent the root mean square error, peak signal-to-noise ratio and mean absolute error output at different depths using the models corresponding to other methods. Figures 2 to 4 In the figure, the ordinate is depth, and the abscissa is root mean square error, peak signal-to-noise ratio, and mean absolute error. The smaller the root mean square error and mean absolute error, the better the output result; the larger the peak signal-to-noise ratio, the better the reconstruction result of the output image. Figures 2 to 4 The comparison results show that the root mean square error, peak signal-to-noise ratio, and mean absolute error of the method / model provided in this application are superior to other methods at the depth level (comparison methods: Bicubic, MGRO, VSRT, DIFFDS, EDSR3D).

[0065] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0066] Accordingly, please refer to Figure 5 , an embodiment of the present invention provides a device for three-dimensional sea temperature spatial downscaling, the device comprising: The acquisition module 10 is used to acquire a target data set, which includes: wind field data, ocean current data, first sea temperature data and second sea temperature data. The resolution of the wind field data, ocean current data and first sea temperature data is smaller than the resolution of the second sea temperature data. For details, please refer to step S10.

[0067] Processing module 11 is used to input the wind field data, ocean current data, first sea temperature data and second sea temperature data in the target data set into the three-dimensional sea temperature generation model for spatial downscaling processing, and output three-dimensional sea temperature data. The resolution of the three-dimensional sea temperature data is equal to the resolution of the second sea temperature data in the target data set. For details, refer to step S11.

[0068] The apparatus for three-dimensional sea temperature spatial downscaling provided in an embodiment of the present application acquires a target data set through an acquisition module, sends the target data set to a processing module, and performs spatial downscaling processing using a three-dimensional sea temperature generation model in the processing module. Diversity data is used to characterize more sea temperature details, and spatial downscaling is used to convert low-resolution data into high-resolution data, thereby highlighting more sea temperature details and making the final output sea temperature data more accurate.

[0069] The embodiment of the present invention further provides an electronic device, such as Figure 6 As shown, the electronic device may include a processor 21 and a memory 20, wherein the processor 21 and the memory 20 may be connected via a bus or other means. Figure 6 The bus connection is taken as an example.

[0070] The processor 21 may be a central processing unit (CPU). The processor 21 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of these chips.

[0071] The memory 20 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as the program instructions / modules corresponding to the three-dimensional sea temperature spatial downscaling method in the embodiment of the present invention (for example, Figure 5The processor 21 executes the non-transient software programs, instructions, and modules stored in the memory 20 to perform various functional applications and data processing of the processor 21, thereby implementing the three-dimensional sea temperature spatial downscaling method in the above method embodiment.

[0072] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created by the processor 21, etc. In addition, the memory 20 may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 21, and these remote memories may be connected to the processor 21 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.

[0073] The one or more modules are stored in the memory 20 and when executed by the processor 21, perform the following steps: Figure 1 The method for three-dimensional sea temperature spatial downscaling in the illustrated embodiment.

[0074] For details of the above electronic equipment, please refer to Figure 1 The corresponding descriptions and effects in the embodiments shown can be understood and will not be repeated here.

[0075] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented through hardware associated with computer program instructions. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD), or a solid-state drive (SSD). The storage medium can also include a combination of the above-mentioned types of memory.

[0076] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A three-dimensional sea temperature spatial downscaling method, characterized in that: include: Acquire a target data set, the target data set comprising: wind field data, ocean current data, first sea temperature data, and second sea temperature data, wherein the resolutions of the wind field data, ocean current data, and first sea temperature data are all smaller than the resolution of the second sea temperature data; The wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set are input into a three-dimensional sea temperature generation model for spatial downscaling processing, and three-dimensional sea temperature data is output, wherein the resolution of the three-dimensional sea temperature data is equal to the resolution of the second sea temperature data in the target data set.

2. The three-dimensional sea temperature spatial downscaling method according to claim 1, characterized in that: Before obtaining the target data set, the following steps are included: Obtaining wind field raw data, ocean current raw data, first sea temperature raw data, and second sea temperature raw data in the target sea area; The wind field raw data, the ocean current raw data, the first ocean temperature raw data, and the second ocean temperature raw data are normalized to make them approach a target range, and the wind field data, the ocean current data, the first ocean temperature raw data, and the second ocean temperature raw data are output, where the target range is [0, 1].

3. The three-dimensional sea temperature spatial downscaling method according to claim 1, characterized in that: Before inputting the wind field data, ocean current data, first sea temperature data and second sea temperature data in the target data set into the three-dimensional sea temperature generation model for spatial downscaling processing, the method includes: splicing and aligning the wind field data and the ocean current data, and outputting the spliced ​​and aligned spliced ​​data.

4. The method for three-dimensional sea temperature spatial downscaling according to claim 3, characterized in that: The step of inputting the wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set into the three-dimensional sea temperature generation model for spatial downscaling processing and outputting the three-dimensional sea temperature data further includes: Sending the wind field data, ocean current data, first sea temperature data and second sea temperature data into a feature extraction network for iterative processing, and outputting a target condition vector; or, The spliced ​​data, the first sea temperature data, and the second sea temperature data are sent to a feature extraction network for iterative processing, and a target condition vector is output.

5. The three-dimensional sea temperature spatial downscaling method according to claim 4, characterized in that: The feature extraction network includes: a ResBlock, a HybridTransformerblock, a ResBlock, a HybridTransformerblock, and a Fusion Layer architecture cascaded in sequence.

6. The three-dimensional sea temperature spatial downscaling method according to claim 4, characterized in that: The step of inputting the wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing to output three-dimensional sea temperature data includes: Perform nearest neighbor interpolation processing on the first sea temperature data, and output the processed first sea temperature data.

7. The three-dimensional sea temperature spatial downscaling method according to claim 6, characterized in that: The step of inputting the wind field data, ocean current data, first sea temperature data, and second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing to output three-dimensional sea temperature data includes: Get the preset noise value; The preset noise value, the target condition vector and the processed first sea temperature data are sent to a noise reduction network for iterative calculation to output three-dimensional sea temperature data.

8. The three-dimensional sea temperature spatial downscaling method according to claim 7, characterized in that: The step of sending the preset noise, the target condition vector, and the processed first sea temperature data into a noise reduction network for iterative calculation to output three-dimensional sea temperature data further includes: Calculate the mean square error between the preset noise value and the noise value corresponding to the three-dimensional sea temperature data.

9. A device for three-dimensional sea temperature spatial downscaling, characterized in that: include: an acquisition module, configured to acquire a target data set, the target data set comprising: wind field data, ocean current data, first sea temperature data, and second sea temperature data, wherein the resolutions of the wind field data, ocean current data, and first sea temperature data are all smaller than the resolution of the second sea temperature data; A processing module is used to input the wind field data, ocean current data, first sea temperature data and second sea temperature data in the target data set into a three-dimensional sea temperature generation model for spatial downscaling processing, and output three-dimensional sea temperature data, wherein the resolution of the three-dimensional sea temperature data is equal to the resolution of the second sea temperature data in the target data set.

10. An electronic device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the three-dimensional sea temperature spatial downscaling method according to any one of claims 1 to 8 by executing the computer instructions.

Images