Sea ice prediction methods, devices and electronic equipment

Abstract

This disclosure relates to the field of sea ice detection technology, and discloses a sea ice prediction method, apparatus, and electronic device. The method includes: acquiring an initial sea ice prediction image, which is generated based on a sea ice numerical prediction model or a sea ice assimilation numerical prediction model; acquiring a sea ice tag image based on a historical ocean remote sensing inversion model; training a neural network based on the initial sea ice prediction image and the sea ice tag image to obtain a sea ice prediction model, wherein the sea ice prediction model includes a gradient sensing module used to distinguish different sea ice regions; and inputting a new initial sea ice prediction image into the sea ice prediction model to obtain sea ice data. This disclosure improves the efficiency of sea ice data prediction by training a neural network based on the characteristics of the sea ice data to obtain accurate predictions of sea ice data.

Description

Technical Field

[0001] This disclosure relates to the field of sea ice detection technology, specifically to sea ice prediction methods, apparatus, and electronic equipment. Background Technology

[0002] Arctic sea ice has a significant impact on human activities such as navigation in the Arctic. While the Arctic Ocean is covered by sea ice for most of the year, its extent varies seasonally and plays a crucial role in global climate regulation. Forecasting Arctic sea ice is extremely important, especially synoptic-scale forecasts, which are critical to navigation decisions. The main methods for Arctic sea ice forecasting are numerical weather prediction and statistical forecasting. Numerical weather prediction uses computers to solve mathematical equations based on physical laws to predict sea ice, while statistical forecasting uses large amounts of data and statistical methods to uncover the relationships between sea ice and other environmental factors. The parameters in numerical weather prediction methods require approximation based on human experience, affecting the accuracy of the forecast. Statistical forecasting is also limited by data and algorithmic constraints.

[0003] In view of this, the present invention is hereby proposed. Summary of the Invention

[0004] To address or at least partially address the aforementioned technical problems, embodiments of this disclosure provide a sea ice prediction method, apparatus, and electronic device that trains a neural network based on the characteristics of sea ice data to obtain accurate predictions of sea ice data and improve the efficiency of sea ice data prediction.

[0005] In a first aspect, embodiments of this disclosure provide a sea ice prediction method, the method comprising:

[0006] Acquire initial sea ice prediction images, which are generated based on sea ice numerical weather prediction models or sea ice assimilation numerical weather prediction models; Acquire sea ice labeled images based on historical ocean remote sensing inversion models; A neural network is trained based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model. The sea ice prediction model includes a gradient perception module, which is used to distinguish different sea ice regions. The new initial sea ice prediction image is input into the sea ice prediction model for prediction, and sea ice data is obtained.

[0007] Secondly, embodiments of this disclosure also provide a sea ice prediction device, the device comprising: An initial image acquisition module is used to acquire an initial predicted image of sea ice, which is generated based on a sea ice numerical weather prediction model or a sea ice assimilation numerical weather prediction model. The tag image acquisition module is used to acquire sea ice tag images based on historical ocean remote sensing inversion models; The model acquisition module is used to train a neural network based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model. The sea ice prediction model includes a gradient perception module, which is used to distinguish different sea ice regions. The sea ice prediction module is used to input new initial sea ice prediction images into the sea ice prediction model for prediction, and obtain sea ice data.

[0008] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising: one or more processors; a storage device for storing one or more programs; the one or more programs being executed by the one or more processors to implement the sea ice prediction method as described above.

[0009] This disclosure provides sea ice prediction methods, apparatus, and electronic devices. The method includes: acquiring an initial sea ice prediction image, which is generated based on a sea ice numerical weather prediction model or a sea ice assimilation numerical weather prediction model; acquiring a sea ice tag image based on a historical ocean remote sensing inversion model; training a neural network based on the initial sea ice prediction image and the sea ice tag image to obtain a sea ice prediction model, wherein the sea ice prediction model includes a gradient sensing module for distinguishing different sea ice regions; and inputting a new initial sea ice prediction image into the sea ice prediction model for prediction to obtain sea ice data. This disclosure improves the efficiency of detecting sea ice concentration by training a sea ice prediction model. Furthermore, by introducing a gradient sensing module into the sea ice prediction model, which distinguishes different sea ice regions and divides them into pure water or pure ice regions and edge ice regions, the accuracy of sea ice prediction is further improved. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0011] Figure 1 This is a flowchart of a sea ice prediction method according to an embodiment of the present disclosure.

[0012] Figure 2 This is a schematic diagram of the structure of a sea ice prediction device according to an embodiment of the present disclosure.

[0013] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0016] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0017] To address the aforementioned issues, embodiments of this disclosure provide a sea ice prediction method, apparatus, and electronic device. A sea ice prediction model is obtained by training a neural network based on the characteristics of sea ice data. This model is then used to predict sea ice data, thereby improving the efficiency and accuracy of sea ice data prediction.

[0018] Figure 1 This is a flowchart illustrating a sea ice prediction method according to an embodiment of this disclosure. The method can be executed by a sea ice prediction device, which can be implemented in software and / or hardware and can be configured in an electronic device. Figure 1 As shown, the method may specifically include the following steps: S110: Obtain an initial sea ice prediction image, which is generated based on a sea ice numerical weather prediction model or a sea ice assimilation numerical weather prediction model.

[0019] Specifically, initial sea ice prediction images can be generated using either a sea ice numerical weather prediction (NMR) model or a sea ice assimilation numerical weather prediction (STM) model. The NMR model predicts sea ice based on physical principles, including sea ice dynamics, thermodynamics, and thickness distribution, and possesses a certain degree of accuracy. The STM model, on the other hand, uses nudging and integrated Kalman filtering methods to combine short-term historical observation data with other factors such as sea ice concentration, thickness, sea surface temperature, and atmospheric ensemble to predict sea ice data.

[0020] S120: Obtain sea ice labeled images based on historical ocean remote sensing inversion models.

[0021] Among them, the historical ocean remote sensing inversion model is based on signals (electromagnetic waves) received by satellite sensors, such as various ocean surface elements (temperature, salinity, chlorophyll concentration, waves, etc.), and then based on mathematical models, the specific values ​​of the ocean elements that generated these signals are deduced from these signals to form sea ice images. Therefore, the sea ice label images obtained based on the historical ocean remote sensing inversion model are the true values ​​of sea ice data in the past and can be used as true labels for training data.

[0022] S130: A neural network is trained based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model, wherein the sea ice prediction model includes a gradient perception module, which is used to distinguish different sea ice regions.

[0023] In one specific embodiment, a neural network is trained based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model, including: The initial sea ice prediction image is input into the neural network to obtain the first sea ice prediction image; The root mean square error of pixels is determined based on the first sea ice prediction image and the sea ice label image; Using the root mean square error of pixels as the loss function, the sea ice prediction model is obtained when the loss function reaches its minimum or tends to stabilize.

[0024] For example, due to the spatiotemporal correlation of sea ice data, initial sea ice prediction images need to be acquired continuously in a certain time sequence. When the time of these initial sea ice prediction images coincides with that of the initial sea ice prediction images, sea ice label images are obtained based on historical ocean remote sensing inversion models. These initial sea ice prediction images are then input into a neural network to obtain the first sea ice prediction image. The sea ice label image is then used as the ground truth label to continuously optimize the deviation between the first sea ice prediction image and the sea ice label image. This deviation can be represented by the root mean square error (RMSE) between the two images. Using the RMSE as a loss function, the sea ice prediction model is obtained when the loss function is minimized or stabilized. At this point, the sea ice prediction model is a pre-trained model.

[0025] Furthermore, the sea ice prediction model includes a gradient sensing module, which is used to distinguish different sea ice regions.

[0026] It should be noted that the sea ice prediction model includes, but is not limited to, the following components. For example, the sea ice prediction model includes an input layer, a gradient sensing module, a key feature layer, and an output layer. Sea ice regions include land areas, pure water or pure ice areas, and marginal ice areas.

[0027] For the floating-point feature of sea ice concentration (SIC) which is strictly constrained within the range of [0, 1] and has a clear physical meaning, when optimizing a neural network model using quantization methods, traditional uniform quantization (such as simple Min-Max linear quantization to INT8) often erases extremely important physical gradient information. In ocean physics and sea ice dynamics, large areas of open water (SIC ≈ 0) and dense ice regions (SIC ≈ 1) are relatively stable; while in the marginal ice zone (MIZ) between the two, that is, the transition zone where 0 < SIC < 1, there are strong thermodynamic exchanges and complex dynamic deformations, which are extremely sensitive to quantization errors. Therefore, it is possible to first filter out the land areas that affect the prediction efficiency of sea ice prediction results.

[0028] In a specific embodiment, the sea ice prediction model further includes a land mask module provided before the gradient perception module, and the land mask module is used to filter out the sea ice land areas based on a mask.

[0029] Specifically, on the basis of the above embodiment, the sea ice prediction model may include an input layer, a land mask module, a gradient perception module, a key feature layer, and an output layer. Since sea ice images are different from traditional images, the range of the three RGB pixel channels of common images is [0 - 255], while sea ice images mainly describe sea ice concentration, and the range of sea ice concentration is [0 - 1], and sea ice images involve open water areas, dense sea ice areas, marginal sea ice areas between open water areas and dense sea ice areas, and land areas. Among them, land areas do not belong to sea ice, so land areas are noise, and the land values of land areas can be set to invalid values, and the land areas need to be removed by means of a mask. Therefore, a land mask module is provided before the gradient perception module to filter out the sea ice land areas using a mask.

[0030] In the embodiment of the present application, the irrelevant land areas are filtered out through the land mask module, reducing the amount of data processing and improving the efficiency of data processing.

[0031] After passing through the land mask module, the sea ice data will enter the gradient perception module, and the gradient perception module will process the sea ice data to obtain pure water or pure ice areas and marginal ice areas.

[0032] In one specific embodiment, the gradient sensing module is used to distinguish different sea ice regions, including: dividing the sea ice region into grids to obtain multiple grid regions; determining the sea ice concentration difference between a grid region and its adjacent grid regions; determining the regional gradient based on the sea ice concentration difference; and determining different sea ice regions based on the regional gradient and a set threshold, wherein the different sea ice regions include pure water or pure ice regions and edge ice regions.

[0033] Further, a first sea ice concentration of a grid region is obtained; a second sea ice concentration of adjacent grid regions located to the left or right of the first grid region in the horizontal direction is obtained; a third sea ice concentration of adjacent grid regions located below or above the first grid region in the vertical direction is obtained; a first concentration difference is determined based on the absolute value of the difference between the first and second sea ice concentrations; a second concentration difference is determined based on the absolute value of the difference between the first and third sea ice concentrations. The regional gradient is determined based on the absolute sum of the first and second concentration differences.

[0034] For example, see Table 1: Table 1

[0035] The sea ice concentration difference includes a first concentration difference and a second concentration difference. The obtained sea ice area is divided into multiple grid areas, and a grid area is determined. First sea ice concentration Based on the first sea ice concentration The grid region is defined as a grid region located in the horizontal direction. Left side or right side The second sea ice concentration in adjacent grid areas or and its location in a grid area in the vertical direction lower side or upper side The third sea ice concentration in adjacent grid areas or First sea ice concentration With the second sea ice concentration or The absolute value of the difference is determined as the first density difference. First sea ice concentration With the concentration of the third sea ice or The absolute value of the difference is determined as the second density difference. The first density difference Difference between second density The sum is determined as the regional gradient. Please refer to the following formulas (1) and (2): or , or (1); (2).

[0036] Optionally, in another specific embodiment, a first grid region and a second grid region adjacent to each other in the horizontal direction of a grid region are obtained; a first sea ice concentration of the first grid region and a second sea ice concentration of the second grid region are determined; a third and a fourth sea ice concentration of the first grid region and a fifth and a sixth sea ice concentration of the second grid region adjacent to each other in the vertical direction are determined; a third and a fourth grid region adjacent to each other in the vertical direction of a grid region are obtained; a seventh sea ice concentration of the third grid region and an eighth sea ice concentration of the fourth grid region are determined. The concentration is determined by: 1) determining the ninth and tenth sea ice concentrations of the third grid region in the horizontal direction, and the eleventh and twelfth sea ice concentrations of the fourth grid region in the horizontal direction; 2) determining a first concentration difference based on the absolute sum of the first, second, third, fourth, fifth, and sixth sea ice concentrations; and 3) determining a second concentration difference based on the absolute sum of the seventh, eighth, ninth, tenth, eleventh, and twelfth sea ice concentrations. Furthermore, determining the regional gradient based on the sea ice concentration difference includes: determining the regional gradient based on the absolute sum of the first concentration difference and the second concentration difference.

[0037] For example, Figure 2 This is a structural schematic diagram of sea ice concentration in different grid regions according to an embodiment of the present disclosure, such as... Figure 2 As shown, the acquired sea ice area is divided into multiple grid regions, and one grid region is determined. sea ​​ice concentration Obtain a grid region The first grid region adjacent in the horizontal direction and the second grid area Determine the first grid region. First sea ice concentration and the second grid area Second sea ice concentration Determine the first grid region. Third sea ice concentration in adjacent grid areas in the vertical direction and the fourth sea ice concentration and the second grid area The fifth sea ice concentration in adjacent grid areas in the vertical direction and the sixth sea ice concentration .

[0038] Get a grid region The third grid region adjacent in the vertical direction and the fourth grid area Determine the third grid region The seventh sea ice concentration and the fourth grid area The eighth sea ice concentration Determine the third grid region The ninth sea ice concentration in adjacent grid areas in the horizontal direction and the tenth sea ice concentration and the fourth grid area The eleventh sea ice concentration in adjacent grid areas in the horizontal direction and the twelfth sea ice concentration The sea ice concentration difference includes a first concentration difference and a second concentration difference, based on the first sea ice concentration. Second sea ice concentration Third sea ice concentration Fourth sea ice concentration Fifth sea ice concentration and the sixth sea ice concentration The absolute value of the sum determines the first density difference. Based on the seventh sea ice concentration Eighth Sea Ice Concentration Ninth Sea Ice Concentration 10th Sea Ice Concentration Eleventh Sea Ice Concentration and the twelfth sea ice concentration The absolute value of the sum determines the second density difference. ; the first density difference Difference between second density The absolute sum is determined as the regional gradient. For reference, see formulas (3), (4) and (5) below: or or (3); or (4); (5).

[0039] It should be noted that the horizontal and vertical directions of the grid area are perpendicular to each other and lie in the same plane, and the plane is parallel to the ground or ocean plane.

[0040] In one specific embodiment, when the regional gradient is greater than or equal to a set threshold, it is determined to be an edge ice region; when the regional gradient is less than the set threshold, it is determined to be a pure water or pure ice region.

[0041] Based on the above embodiments, after obtaining the regional gradient... Then, the regional gradient can be... The data is compared with a set threshold T to identify different sea ice regions. Specifically, the set threshold T can be 0.05, which is used when the regional gradient... When the gradient is greater than or equal to 0.05, it is identified as an edge ice region; when the regional gradient... When the value is less than 0.05, it is identified as a pure water or pure ice region. In subsequent sea ice data processing, pure water or pure ice regions can be defined to trigger an INT8 fixed-point quantization branch, while edge ice regions can trigger an FP16 half-precision floating-point branch.

[0042] This embodiment distinguishes between pure water or pure ice regions and edge ice regions by using regional gradients and preset thresholds. It can process sea ice data in parallel, filter out model noise, and quantize nonlinear asymmetric functions. Pure water or pure ice regions trigger INT8 fixed-point quantization branches, while edge ice regions trigger FP16 half-precision floating-point branches, improving the efficiency and accuracy of data processing. Furthermore, by using different model branches, the accuracy of model predictions is improved.

[0043] S140: Input the new initial sea ice prediction image into the sea ice prediction model to make predictions and obtain sea ice data.

[0044] The sea ice data includes sea ice concentration, sea ice thickness, and sea surface temperature. Based on the above embodiments, a new initial sea ice prediction image is generated using a sea ice numerical prediction model or a sea ice assimilation numerical prediction model. The new initial sea ice prediction image is then input into the sea ice prediction model (which has been trained) for prediction to obtain sea ice data such as sea ice concentration and sea ice thickness.

[0045] In summary, embodiments of this disclosure provide a sea ice prediction method, apparatus, and electronic device. The method includes: acquiring an initial sea ice prediction image, the initial sea ice prediction image being generated based on a sea ice numerical prediction model or a sea ice assimilation numerical prediction model; acquiring a sea ice tag image obtained based on a historical ocean remote sensing inversion model; training a neural network based on the initial sea ice prediction image and the sea ice tag image to obtain a sea ice prediction model, wherein the sea ice prediction model includes a gradient sensing module, the gradient sensing module being used to distinguish different sea ice regions; and inputting a new initial sea ice prediction image into the sea ice prediction model for prediction to obtain sea ice data. This disclosure improves the efficiency of sea ice concentration detection by training a sea ice prediction model. Furthermore, by introducing a gradient sensing module into the sea ice prediction model, which distinguishes different sea ice regions and divides sea ice regions into pure water or pure ice regions and edge ice regions, it further improves the accuracy of sea ice prediction.

[0046] Figure 2 This is a schematic diagram of the sea ice prediction device in an embodiment of this disclosure. Figure 2 As shown, the device includes an initial image acquisition module 210, a label image acquisition module 220, a model acquisition module 230, and a sea ice prediction module 240.

[0047] The initial image acquisition module 210 is used to acquire an initial predicted sea ice image, which is generated based on a sea ice numerical prediction model or a sea ice assimilation numerical prediction model. The label image acquisition module 220 is used to acquire sea ice label images based on historical ocean remote sensing inversion models; The model acquisition module 230 is used to train a neural network based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model. The sea ice prediction model includes a gradient perception module, which is used to distinguish different regions of sea ice. The sea ice prediction module 240 is used to input new initial sea ice prediction images into the sea ice prediction model for prediction to obtain sea ice data.

[0048] In one specific embodiment, the model acquisition module 230 is further configured to divide the sea ice region into grids to obtain multiple grid regions; determine the sea ice concentration difference between a grid region and its adjacent grid regions; determine the regional gradient based on the sea ice concentration difference; and determine different sea ice regions based on the regional gradient and a set threshold, wherein the different sea ice regions include pure water or pure ice regions and edge ice regions.

[0049] In one specific embodiment, the model acquisition module 230 is further configured to acquire a first sea ice concentration of a grid region, a second sea ice concentration of a grid region adjacent to it in the horizontal direction, and a third sea ice concentration of a grid region adjacent to it in the vertical direction; and determine the sea ice concentration difference based on the first sea ice concentration, the second sea ice concentration, and the third sea ice concentration.

[0050] In one specific embodiment, the model acquisition module 230 is further configured to acquire a first grid region and a second grid region that are horizontally adjacent to a grid region; determine a first sea ice concentration of the first grid region and a second sea ice concentration of the second grid region; determine a third and a fourth sea ice concentration of the first grid region and a fifth and a sixth sea ice concentration of the second grid region that are vertically adjacent to the first grid region; acquire a third and a fourth grid region that are vertically adjacent to a grid region; and determine a seventh sea ice concentration of the third grid region. The concentration of sea ice in the fourth grid region and the eighth sea ice concentration in the fourth grid region are determined; the ninth and tenth sea ice concentrations in the horizontally adjacent grid regions of the third grid region and the eleventh and twelfth sea ice concentrations in the horizontally adjacent grid regions of the fourth grid region are determined; the sea ice concentration difference is determined based on the first, second, third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh and twelfth sea ice concentrations.

[0051] In one specific embodiment, the model acquisition module 230 is further configured to determine a first concentration difference based on the first, second, third, fourth, fifth, and sixth sea ice concentrations; determine a second concentration difference based on the seventh, eighth, ninth, tenth, eleventh, and twelfth sea ice concentrations; and determine a regional gradient based on the absolute sum of the first and second concentration differences.

[0052] In one specific embodiment, the model acquisition module 230 is further configured to determine the region as an edge ice region when the gradient of the region is greater than or equal to a set threshold, and to determine the region as a pure water or pure ice region when the gradient of the region is less than the set threshold.

[0053] In one specific embodiment, the model acquisition module 230 is used to set a land mask module before the gradient perception module, the land mask module being used to filter out sea ice land areas based on a mask.

[0054] In one specific embodiment, the model acquisition module 230 is used to input the initial sea ice prediction image into a neural network to obtain a first sea ice prediction image. Based on the first sea ice prediction image and the sea ice label image, the root mean square error of pixels is determined. The root mean square error of pixels is used as a loss function. When the loss function reaches its minimum or tends to stabilize, the sea ice prediction model is obtained.

[0055] The sea ice prediction apparatus provided in the embodiments of this disclosure can execute the steps in the sea ice prediction method provided in the method embodiments of this disclosure, and has the execution steps and beneficial effects, which will not be repeated here.

[0056] Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of this disclosure. See below for details. Figure 3 It shows a schematic diagram of the structure of an electronic device 500 suitable for implementing embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0057] like Figure 3 As shown, the electronic device 500 may include a processing unit 501, a ROM 502, a RAM 503, a bus 504, an input / output (I / O) interface 505, an input device 506, an output device 507, a storage device 508, and a communication device 509. The processing unit (e.g., a central processing unit, a graphics processor, etc.) 501 can perform various appropriate actions and processes according to a program stored in the read-only memory (ROM) 502 or a program loaded from the storage device 508 into the random access memory (RAM) 503 to implement the sea ice prediction method as described in the embodiments of this disclosure. The RAM 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, the ROM 502, and the RAM 503 are interconnected via the bus 504. The input / output (I / O) interface 505 is also connected to the bus 504.

[0058] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts, thereby implementing the sea ice prediction method as described above. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of embodiments of this disclosure.

[0059] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0060] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: acquire an initial sea ice prediction image, the initial sea ice prediction image being generated based on a sea ice numerical prediction model or a sea ice assimilation numerical prediction model; acquire a sea ice labeled image obtained based on a historical ocean remote sensing inversion model; train a neural network based on the initial sea ice prediction image and the sea ice labeled image to obtain a sea ice prediction model, wherein the sea ice prediction model includes a gradient sensing module used to distinguish different sea ice regions; and input a new initial sea ice prediction image into the sea ice prediction model for prediction to obtain sea ice data.

[0061] Optionally, when one or more of the above-described procedures are executed by the electronic device, the electronic device may also perform other steps described in the above embodiments.

[0062] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0063] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A sea ice prediction method, characterized in that, The method includes: Acquire initial sea ice prediction images, which are generated based on sea ice numerical weather prediction models or sea ice assimilation numerical weather prediction models; Acquire sea ice labeled images based on historical ocean remote sensing inversion models; A neural network is trained based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model. The sea ice prediction model includes a gradient perception module, which is used to distinguish different sea ice regions. The new initial sea ice prediction image is input into the sea ice prediction model for prediction, and sea ice data is obtained.

2. The sea ice prediction method according to claim 1, characterized in that, The gradient sensing module is used to distinguish different sea ice regions, including: The sea ice region is divided into grids to obtain multiple grid regions; Determine the difference in sea ice concentration between a grid region and its adjacent grid regions; Based on the sea ice concentration difference, the regional gradient is determined; Based on the regional gradient and a set threshold, different sea ice regions are determined, including pure water or pure ice regions and edge ice regions.

3. The sea ice prediction method according to claim 2, characterized in that, The sea ice concentration difference includes a first concentration difference and a second concentration difference. Determining the sea ice concentration difference between a grid region and its adjacent grid regions includes: Obtain the first sea ice concentration for a grid region; Obtain the second sea ice concentration of the adjacent grid regions located to the left or right of the first grid region in the horizontal direction; Obtain the third sea ice concentration of the adjacent grid region located below or above the first grid region in the vertical direction; The first concentration difference is determined based on the absolute value of the difference between the first sea ice concentration and the second sea ice concentration. The second concentration difference is determined based on the absolute value of the difference between the first sea ice concentration and the third sea ice concentration.

4. The sea ice prediction method according to claim 2, characterized in that, The sea ice concentration difference includes a first concentration difference and a second concentration difference. Determining the sea ice concentration difference between a grid region and its adjacent grid regions includes: Obtain the first and second grid regions that are horizontally adjacent to a given grid region; Determine the first sea ice concentration in the first grid region and the second sea ice concentration in the second grid region; Determine the third and fourth sea ice concentrations of the first grid region in the vertical direction adjacent to the grid region, and the fifth and sixth sea ice concentrations of the second grid region in the vertical direction adjacent to the grid region; Obtain the third and fourth adjacent grid regions in the vertical direction of a given grid region; Determine the seventh sea ice concentration in the third grid region and the eighth sea ice concentration in the fourth grid region; Determine the ninth and tenth sea ice concentrations of the horizontally adjacent grid regions in the third grid region, and the eleventh and twelfth sea ice concentrations of the horizontally adjacent grid regions in the fourth grid region; Based on the absolute value of the sum of the first sea ice concentration, the second sea ice concentration, the third sea ice concentration, the fourth sea ice concentration, the fifth sea ice concentration, and the sixth sea ice concentration, the first concentration difference is determined. The second concentration difference is determined based on the absolute value of the sum of the seventh, eighth, ninth, tenth, eleventh, and twelfth sea ice concentrations.

5. The sea ice prediction method according to claim 3 or 4, characterized in that, The determination of the regional gradient based on the sea ice concentration difference includes: The regional gradient is determined based on the absolute sum of the first density difference and the second density difference.

6. The sea ice prediction method according to claim 2, characterized in that, When the gradient of the region is greater than or equal to a set threshold, it is determined to be an edge ice region; When the gradient of the region is less than a set threshold, it is determined to be a pure water or pure ice region.

7. The method for obtaining the sea ice prediction model according to claim 1, characterized in that, The sea ice prediction model also includes a land mask module set before the gradient sensing module, which is used to filter out sea ice land areas based on a mask.

8. The sea ice prediction method according to claim 1, characterized in that, The process of training a neural network based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model includes: The initial sea ice prediction image is input into the neural network to obtain the first sea ice prediction image; Based on the first sea ice prediction image and sea ice label image, determine the root mean square error of pixels; Using the root mean square error of pixels as the loss function, the sea ice prediction model is obtained when the loss function reaches its minimum or tends to stabilize.

9. A sea ice prediction device, characterized in that, The device includes: An initial image acquisition module is used to acquire an initial predicted image of sea ice, which is generated based on a sea ice numerical weather prediction model or a sea ice assimilation numerical weather prediction model. The tag image acquisition module is used to acquire sea ice tag images based on historical ocean remote sensing inversion models; The model acquisition module is used to train a neural network based on the initial sea ice prediction image and the sea ice label image to obtain a sea ice prediction model. The sea ice prediction model includes a gradient perception module, which is used to distinguish different regions of sea ice. The sea ice prediction module is used to input new initial sea ice prediction images into the sea ice prediction model for prediction, and obtain sea ice data.

10. An electronic device, the electronic device comprising: One or more processors; Storage device, the storage device being used to store one or more programs, Wherein, when the one or more programs are executed by the one or more processors, they implement the method according to any one of claims 1-8.

Images