Method and device for predicting ocean surface temperature, electronic equipment and storage medium

Abstract

The application provides a marine surface temperature prediction method and device, electronic equipment and storage medium, and belongs to the technical field of marine observation. The method comprises the following steps: based on first satellite image data, determining first observation data of each pixel in an image of a target sea area at a target spatial resolution, and based on second satellite image data, determining second observation data of each pixel in the image; performing fusion analysis on the first observation data and the second observation data of each pixel to determine a sea surface temperature prediction result of the target sea area at the target spatial resolution and a target time resolution; and the target time resolution is determined based on the time resolution of the second satellite image data. The application can effectively obtain a high-precision, high-spatial-temporal resolution and high-coverage SST fusion product, well meets the rapid development demand of the field of marine resource development, and greatly improves the application value of the SST product.

Description

Technical Field

[0001] This invention relates to the field of marine observation technology, and in particular to a method, apparatus, electronic device, and storage medium for predicting ocean surface temperature. Background Technology

[0002] Sea surface temperature (SST) is a crucial physical factor in ocean circulation, atmospheric circulation, and climate. Early SST observations primarily relied on ships and buoys, but this approach could not meet the demands for large-scale, real-time monitoring. Satellite remote sensing technology, with its advantages of wide coverage, high resolution, and long-term repeatable observations, has been widely applied to global SST observation. For example, the Fengyun-3E polar-orbiting meteorological satellite, equipped with the MERSI-II detector and using all infrared channels, can acquire SST data at 1 km / 1 day, meeting the high spatial resolution monitoring requirements for sea surface temperature. The Fengyun-4A geostationary satellite can acquire SST data at 4 km / 1 hour, enabling continuous observation of sea surface temperature changes.

[0003] In recent years, with the rapid development of fields such as climate change monitoring and marine resource development, the demand for dynamic high spatial resolution sea surface temperature product data has been increasing. However, due to the limitations of sensor performance, SST products retrieved from the aforementioned single satellite thermal infrared remote sensing data cannot simultaneously achieve both temporal and spatial resolution, thus failing to meet development needs. The application value of SST products needs to be improved. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for predicting ocean surface temperature, which addresses the shortcomings of existing technologies where SST products derived from single satellite thermal infrared remote sensing data cannot simultaneously achieve both temporal and spatial resolution, thus failing to meet development needs and limiting the application value of SST products.

[0005] This invention provides a method for predicting ocean surface temperature, comprising:

[0006] Based on the first satellite image data, the first observation data of each pixel in the image at the target spatial resolution of the target sea area is determined, and based on the second satellite image data, the second observation data of each pixel in the image is determined.

[0007] Wherein, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data;

[0008] The first and second observation data of each pixel are fused and analyzed to determine the sea surface temperature prediction result of the target sea area under the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0009] According to a method for predicting ocean surface temperature provided by the present invention, the step of fusing and analyzing the first and second observation data of each pixel to determine the predicted ocean surface temperature of the target sea area at the target spatial resolution and the target temporal resolution includes:

[0010] The first sea surface temperature of each pixel at the predicted time is determined by fusing the first sea surface temperature at the existing time and the second sea surface temperature at the existing time, as well as the second sea surface temperature at the predicted time. The second observation data includes the second sea surface temperature at the existing time and the second sea surface temperature at the predicted time. The first observation data includes the first sea surface temperature at the existing time.

[0011] Based on the first sea surface temperature of each pixel at the prediction time and the weight factor corresponding to each pixel, a neighborhood convolution operation is performed to determine the sea surface temperature of the center pixel in the neighborhood window at the prediction time; the weight factor is used to characterize the contribution of the pixels adjacent to the center pixel in the neighborhood window to the sea surface temperature of the center pixel.

[0012] After the neighborhood window has moved and traversed every pixel in the image, the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution is obtained.

[0013] According to the ocean surface temperature prediction method provided by the present invention, the weighting factor is determined based on the spectral fluctuation factor, the temporal fluctuation factor, and the spatial distance factor.

[0014] The spectral fluctuation factor is determined based on the difference between the first sea surface temperature and the second sea surface temperature at an existing time for each pixel; the temporal fluctuation factor is determined based on the average sea surface temperature at all existing times extracted from the second observation data of each pixel; and the spatial distance factor is determined based on the distance between each pixel and its corresponding center pixel.

[0015] According to a method for predicting ocean surface temperature provided by the present invention, after determining the second observation data for each pixel in the image, the method further includes:

[0016] The image is divided into multiple sub-image blocks according to a preset pixel size;

[0017] The first and second observation data of each pixel in each sub-image block are fused and analyzed to obtain the sea surface temperature prediction result of each sub-image block at the target time resolution.

[0018] The sea surface temperature prediction results of each sub-image block at the target temporal resolution are merged to obtain the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution.

[0019] According to a method for predicting ocean surface temperature provided by the present invention, the first satellite image data includes a first satellite image and corresponding observation data; the step of determining the first observation data of each pixel in the image of the target sea area at the target spatial resolution based on the first satellite image data includes:

[0020] The first satellite image is resampled to determine the image of the target sea area at the target spatial resolution;

[0021] The observation data corresponding to the first satellite image is registered to the pixel matrix of the image to obtain the first observation data of each pixel in the image.

[0022] According to a method for predicting ocean surface temperature provided by the present invention, the second satellite image data includes the temporal change of sea surface temperature, the second satellite image, and the corresponding observation data; the step of determining the second observation data for each pixel in the image based on the second satellite image data includes:

[0023] The second satellite image is resampled to determine the image of the target sea area at the target spatial resolution;

[0024] Register the observation data corresponding to the second satellite image into the pixel matrix of the image, and determine the existing observation data corresponding to each pixel in the pixel matrix;

[0025] Based on the sea surface temperature change over time and the existing observation data corresponding to each pixel in the pixel matrix, data fitting is performed to determine the observation data corresponding to the predicted time for each pixel in the pixel matrix.

[0026] Based on the existing observation data corresponding to each pixel and the predicted observation data corresponding to each pixel, the second observation data of each pixel in the image is obtained.

[0027] According to the ocean surface temperature prediction method provided by the present invention, after performing fusion analysis on the first and second observation data for each pixel to determine the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution, the method further includes:

[0028] According to the preset sea surface temperature grading rendering standard, based on the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution, a sea surface temperature prediction map of the target sea area at the target spatiotemporal resolution is generated; the spatiotemporal resolution is determined based on the target spatial resolution and the target temporal resolution.

[0029] The present invention also provides a marine surface temperature prediction device, comprising:

[0030] The processing module is used to determine the first observation data of each pixel in the image of the target sea area at the target spatial resolution based on the first satellite image data, and to determine the second observation data of each pixel in the image based on the second satellite image data.

[0031] Wherein, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data;

[0032] The prediction module is used to perform fusion analysis on the first and second observation data of each pixel to determine the predicted sea surface temperature of the target sea area under the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0033] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the ocean surface temperature prediction method as described above.

[0034] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the ocean surface temperature prediction method as described above.

[0035] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the ocean surface temperature prediction method as described above.

[0036] The ocean surface temperature (SST) prediction method, apparatus, electronic device, and storage medium provided by this invention acquire two types of satellite observation data of a target sea area at the same target spatial resolution by utilizing high spatial resolution first satellite image data and high temporal resolution second satellite image data. Then, by performing data fusion analysis on the two types of satellite observation data of the image at the same target spatial resolution, the predicted sea surface temperature of the target sea area at the target spatiotemporal resolution is obtained. This achieves high spatiotemporal resolution data fusion at the time scale of high spatial resolution satellite image data and high temporal resolution satellite image data, which can effectively obtain high-precision, high spatiotemporal resolution, and high-coverage SST fusion products, well meeting the rapid development needs of fields such as marine resource development, and greatly improving the application value of SST products. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0038] Figure 1 This is one of the flowcharts of the ocean surface temperature prediction method provided by the present invention;

[0039] Figure 2 This is a schematic diagram comparing the accuracy of sea surface temperature reconstruction before and after in the ocean surface temperature prediction method provided by this invention;

[0040] Figure 3 This is the second schematic diagram of the ocean surface temperature prediction method provided by the present invention;

[0041] Figure 4 This is a schematic diagram illustrating the accuracy verification results of the ocean surface temperature prediction method provided by this invention;

[0042] Figure 5 This is a schematic diagram of the structure of the ocean surface temperature prediction device provided by the present invention;

[0043] Figure 6 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0045] The following is combined with Figures 1-6 The present invention describes a method, apparatus, electronic device, and storage medium for predicting ocean surface temperature.

[0046] Figure 1 This is one of the flowcharts illustrating the ocean surface temperature prediction method provided by the present invention, such as... Figure 1 As shown, it includes steps 110 and 120.

[0047] Step 110: Based on the first satellite image data, determine the first observation data of each pixel in the image of the target sea area at the target spatial resolution, and based on the second satellite image data, determine the second observation data of each pixel in the image.

[0048] Among them, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data.

[0049] Specifically, the first satellite image data described in this embodiment of the invention may be image data of sea surface temperature products segmented by 1km / 1 day collected by the Moderate Resolution Spectroscopic Imager (MERSI) carried by the Fengyun-3E polar-orbiting meteorological satellite, which has a multi-region high spatial resolution of 1km.

[0050] The second satellite image data described in this embodiment of the invention can be image data of sea surface temperature products at 4km / 1 hour collected by the multi-channel scanning imaging radiometer (AGRI) carried by the Fengyun-4A (FY-4A) geostationary satellite, which has a high temporal resolution of 1 hour and multiple time points.

[0051] The target sea area described in this embodiment of the invention refers to the sea area coverage area used for SST product monitoring.

[0052] In embodiments of the present invention, since FY-4A satellite imagery data is required, and FY-4A is a geostationary satellite with a fixed monitored sea area, the target sea area can be determined based on the monitored sea area of ​​FY-4A. Therefore, in embodiments of the present invention, the target sea area can be 40°E to 170°E, 60°S to 50°N, i.e., 40°E to 170°E longitude and 60°S to 50°N latitude.

[0053] In some embodiments, the target sea area may also be determined based on the monitoring sea area range of other different geostationary satellites.

[0054] It is understandable that the spatial resolution of the first satellite imagery data is higher than that of the second satellite imagery data, and the temporal resolution of the second satellite imagery data is higher than that of the first satellite imagery data. For example, the 1km spatial resolution of FY-3E satellite imagery data is higher than the 4km spatial resolution of FY-4A satellite imagery data; the 1-hour temporal resolution of FY-4A satellite imagery data is higher than the 1-day temporal resolution of FY-3E satellite imagery data.

[0055] The target spatial resolution described in the embodiments of the present invention can be determined based on the spatial resolution of the first satellite image data, specifically it can be a spatial resolution of 1 km.

[0056] The first observation data described in this embodiment of the invention can be the observation data corresponding to the image at the target spatial resolution obtained by observing the target sea area through the FY-3E satellite.

[0057] The second observation data described in this embodiment of the invention can be the observation data corresponding to the image at the target spatial resolution obtained by observing the target sea area through the FY-4A satellite.

[0058] It should be noted that, for ease of explanation, ocean surface temperature will be referred to as sea surface temperature for the following reasons.

[0059] In the embodiments of the present invention, it is first necessary to preprocess the satellite image data to solve the problem of different spatial resolutions and latitude and longitude information of sea surface temperature products from different sources.

[0060] In embodiments of the present invention, the first observation data and the second observation data of each pixel in the image of the target sea area at the target spatial resolution can be determined based on data processing such as missing values, latitude and longitude conversion, and resampling of the first satellite image data and the second satellite image data.

[0061] In one specific embodiment, firstly, missing data values ​​are handled by uniformly assigning -999 to land areas and outlier regions in both FY-3E and FY-4A products. Next, latitude and longitude conversion is performed. Since FY-4A satellite data is arranged by row and column numbers and lacks latitude and longitude data, it can be matched using the officially provided latitude and longitude lookup table to achieve the conversion from row and column numbers to latitude and longitude. Then, resampling is performed, uniformly resampling both FY-3E and FY-4A products to a 1km × 1km spatial resolution according to a 0.01° × 0.01° latitude and longitude standard. The low spatial resolution FY-4A imagery is resampled to the 1km spatial resolution of FY-3E, ensuring that the FY-3E and FY-4A product imagery data maintain the same pixel size and coverage of the target sea area.

[0062] Based on the above embodiments, as an optional embodiment, the first satellite image data includes a first satellite image and corresponding observation data; based on the first satellite image data, determining the first observation data for each pixel in the image of the target sea area at the target spatial resolution includes:

[0063] The first satellite image was resampled to determine the image of the target sea area at the target spatial resolution;

[0064] The observation data corresponding to the first satellite image is registered into the pixel matrix of the image to obtain the first observation data of each pixel in the image.

[0065] Specifically, the pixel matrix described in the embodiments of the present invention refers to the pixel matrix of the image of the target sea area at the target spatial resolution. The image is determined based on the first satellite image, and the pixel matrix can be constructed based on the target spatial resolution of 1km.

[0066] In an embodiment of the present invention, a pixel matrix M of size 11000*13000 can be created within the coverage area of ​​the target sea area, which corresponds to a spatial resolution of 1km.

[0067] The first satellite imagery data acquired by FY-3E is stored in blocks, with an average of 200 images per day. Each image contains information about the sea surface temperature and its corresponding regional distribution.

[0068] Most scholars employ a single-scene, single-time-time prediction model in their research. However, calculating single-scene images using traditional weighted model algorithms is time-consuming. Therefore, in this embodiment of the invention, all first-day satellite images acquired by the FY-3E satellite can first be resampled to a target spatial resolution of 1km. For each image acquired by FY-3E, the row and column numbers of the pixels in each image in the pixel matrix M are output using the following formula in the multi-core processor of a high-performance calculator:

[0069] R(x i y j )=(Lat(x i y j )-Lat min ) / dis;

[0070] L(x i y j )=(Lon(x i y j )-Lon min ) / dis;

[0071] Among them, (x i y j ) represents a pixel in a satellite image; Lon(x) i y j ) and Lat(x i y j ) represent the latitude and longitude of the pixel's location; Lon min The value is 40, Lat min The value is -60, representing the initial latitude and longitude values; the dis value can be 0.01, corresponding to a spatial resolution of 1 km; R(x i y j L(x) represents the row number of the pixel in the pixel matrix M; i y j ) represents the column number of the pixel matrix M corresponding to the pixel.

[0072] Therefore, based on the created pixel matrix M of size 11000*13000 pixels, each FY-3E image is matched to the corresponding computer core, all pixels of the first satellite image are traversed, and the row number and column number of each pixel in the pixel matrix M are output.

[0073] Furthermore, in an embodiment of the present invention, the observation data corresponding to the first satellite image is registered into the pixel matrix M. This involves merging the row and column numbers obtained from each kernel operation, the sea surface temperature value corresponding to each pixel, and the satellite transit time, and registering them into the pixel matrix M. When multiple FY-3E satellite images cover the same pixel, the most recent observation time is taken, replacing the original value and updating the corresponding observation time. This forms a sea surface temperature value matrix and an observation time matrix at a spatial resolution of 11000*13000, thereby obtaining the first observation data for each pixel at the target spatial resolution.

[0074] The method of this invention determines the image of the target sea area at the target spatial resolution by resampling the first satellite image to the target spatial resolution, and determines the first observation data of each pixel in the image, thus preparing for the subsequent resampling of the second satellite image to the same target spatial resolution and the fusion of the two types of satellite image data, providing an accurate and reliable data source.

[0075] Based on the above embodiments, as an optional embodiment, the second satellite image data includes the sea surface temperature change over time, the second satellite image, and the corresponding observation data; based on the second satellite image data, determining the second observation data for each pixel in the image includes:

[0076] The second satellite imagery is resampled to determine the target sea area at the target spatial resolution;

[0077] Register the observation data corresponding to the second satellite image into the pixel matrix of the image, and determine the existing observation data at each time corresponding to each pixel in the pixel matrix;

[0078] Data fitting is performed based on the sea surface temperature change over time and the existing observation data corresponding to each pixel in the pixel matrix to determine the observation data corresponding to the predicted time in the pixel matrix.

[0079] Based on the existing observation data for each pixel and the predicted observation data for each pixel, the second observation data for each pixel in the image is obtained.

[0080] Specifically, in the embodiments of the present invention, the second satellite image data acquired by the FY-4A satellite may include the sea surface temperature change over time, the second satellite image, and the corresponding observation data. The sea surface temperature change over time may be the hourly change of sea surface temperature acquired by the FY-4A satellite, which can be extracted from the sea surface temperature data at multiple times acquired by the FY-4A satellite. The observation data may include the observation time and the corresponding observed sea surface temperature.

[0081] It should be noted that the existing time refers to the observation time corresponding to the valid data collected by the satellite, which can include multiple times; the predicted time refers to the time when the satellite misses data due to environmental interference, as well as the time when it is unable to conduct observations, which can also include multiple times.

[0082] Furthermore, the FY-4A low spatial resolution image data is resampled to a 1km spatial resolution. Following the aforementioned method, the second satellite image of FY-4A is resampled, and the observation data corresponding to the high temporal resolution second satellite image at multiple times are registered to the pixel matrix M of the target sea area image at the target spatial resolution. This also forms a sea surface temperature numerical matrix and an observation time matrix at an 11000*13000 spatial resolution, from which the observation data corresponding to each pixel in the pixel matrix M at multiple existing times can be obtained.

[0083] It should be noted that the second satellite imagery data acquired by the FY-4A satellite is easily affected by cloud cover and other factors, resulting in data gaps and voids at different times, which makes it impossible to extract information on temporal changes. Sea surface temperature fluctuates strongly on a monthly scale, making it difficult to fit with an appropriate curve, but it maintains a linear trend within a small range (on a daily scale or within half a day). Therefore, linear fitting is used to complete the 24-hour high temporal resolution imagery.

[0084] Furthermore, in an embodiment of the present invention, linear fitting calculations are performed based on the sea surface temperature change over time and the observation data of multiple existing times corresponding to each pixel in the pixel matrix to obtain the observation data of multiple predicted times corresponding to each pixel in the pixel matrix M, thereby completing and reconstructing the sea surface temperature data observed by the FY-4A satellite.

[0085] Specifically, the FY-4A satellite can acquire hourly changes in sea surface temperature, i.e., the instantaneous changes in sea surface temperature. In the short term, a linear fitting model can be used for each pixel to simulate the changing characteristics. The expression for the linear fitting model is:

[0086] C(x i y j ,t0)=a(x i y j )×Δt+C(x i y j , t k );

[0087] In the formula, C(x) i y j ,t0) and C(x i y j , t k () represent the predicted time t0 and the existing time t, respectively. k The corresponding sea surface temperature data observed by the FY-4A satellite; (x i y j ) represents the cell corresponding to row number i and column number j, and Δt represents the distance between t0 and t1. k The time interval between two moments, a(x) iy j () indicates the change in sea surface temperature of that pixel over time.

[0088] In an embodiment of the present invention, a temperature linear reconstruction method is used to complete the missing image data of FY-4A satellite observation by establishing a linear fitting curve pixel by pixel, while retaining the original data, thereby obtaining FY-4A satellite SST products with higher coverage. The accuracy verification of the results before and after reconstruction shows that this method is effective.

[0089] In embodiments of the present invention, the evaluation metrics used mainly include root mean square error (RMSE), mean absolute error (MAE), and linear correlation (R0). 2 ), RMSE and MAE are used to characterize the absolute error of SST temperature products, R 2 This indicates the degree of correlation between the SST linear fit value and the buoy value. The formulas for calculating RMSE, MAE, and R² are as follows:

[0090]

[0091]

[0092]

[0093] In the formula, i represents the sequence number, N represents the number of SST product data and buoy data that match, and P i P represents the value of SST product data. i ′ indicates that it is related to P i The corresponding buoy data value.

[0094] Figure 2 This is a schematic diagram comparing the accuracy of sea surface temperature reconstruction before and after in the sea surface temperature prediction method provided by this invention, as shown in the figure. Figure 2 As shown, the reconstructed 15-day FY-4A satellite SST product was validated using Argo buoy measurement data. Figure 2 (a) in the image shows the accuracy results of the FY-4A satellite SST product before SST reconstruction. Figure 2 (b) in the figure shows the accuracy result after SST reconstruction of the FY-4A satellite SST product.

[0095] like Figure 2 In (a) and (b), the number of verifiable buoy points N before reconstruction was only 282, but the number of buoy points N after reconstruction increased to 471, and the root mean square error (RMSE) and mean absolute error (MAE) also increased by 0.122 and 0.57 respectively, and the coverage of the image data was greatly improved.

[0096] Furthermore, in the pixel matrix M, based on the existing observation data corresponding to each pixel and the observation data corresponding to the predicted time, a sea surface temperature numerical matrix and observation time matrix with higher coverage of the target sea area at a spatial resolution of 11000*13000 can be obtained, thereby obtaining the second observation data of each pixel in the pixel matrix M at the target spatial resolution.

[0097] The method of this invention determines the corresponding image of the target sea area at the target spatial resolution by resampling the second satellite image to the target spatial resolution, and improves the coverage of the image observation data by linear fitting curves. It determines the second observation data of each pixel in the pixel matrix at the target spatial resolution, ensuring that the first and second satellite images maintain the same pixel size and coverage of the target sea area after resampling, thus providing a reliable data source for the spatiotemporal fusion of the two types of satellite image data.

[0098] Step 120: Perform fusion analysis on the first and second observation data for each pixel to determine the sea surface temperature prediction results for the target sea area under the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0099] Specifically, the target time resolution described in the embodiments of the present invention can be determined based on the time resolution of the second satellite image data, which can specifically be a time resolution of 1 hour.

[0100] Furthermore, in an embodiment of the present invention, a neighborhood convolution operation method can be used to fuse the first observation data and the second observation data of each pixel in the pixel matrix M to determine the sea surface temperature prediction result of the target sea area under the target spatial resolution and the target temporal resolution.

[0101] Based on the above embodiments, as an optional embodiment, the first and second observation data of each pixel are fused and analyzed to determine the sea surface temperature prediction result of the image at the target temporal resolution, including:

[0102] The first sea surface temperature of each pixel at the current time and the second sea surface temperature at the current time are fused together to determine the first sea surface temperature of each pixel at the predicted time; the second observation data includes the second sea surface temperature at the current time and the second sea surface temperature at the predicted time; the first observation data includes the first sea surface temperature at the current time.

[0103] Based on the first sea surface temperature of each pixel at the prediction time and the weight factor corresponding to each pixel, a neighborhood convolution operation is performed to determine the sea surface temperature of the center pixel in the neighborhood window at the prediction time; the weight factor is used to characterize the contribution of the pixels adjacent to the center pixel in the neighborhood window to the sea surface temperature of the center pixel.

[0104] After moving the neighborhood window to traverse every pixel in the image, the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution is obtained.

[0105] Specifically, in the embodiments of the present invention, the first sea surface temperature can be the sea surface temperature at the existing time and the predicted time corresponding to the FY-3E satellite, and the second sea surface temperature can be the sea surface temperature at the existing time and the predicted time corresponding to the FY-4A satellite.

[0106] In an embodiment of the present invention, after obtaining the first and second observation data of each pixel in the 11000*13000 pixel pixel matrix M, the observation time and the observed sea surface temperature value are recorded respectively, and the two types of satellite observation data are fused and analyzed.

[0107] For a single pixel, when the effects of sensor bias, changes in sea surface thermal properties, and observation conditions are ignored, i.e., assuming that the difference between the two sensor readings remains constant between any given time and the predicted time, the following formula holds true:

[0108] Y(x i y j ,t0)-Y(x i y j , t k )=F(x i y j ,t0)-F(x i y j , t k );

[0109] Among them, (x i y j ) represents a pixel, t0 represents the prediction time, t k Indicates an existing time; Y(x) i y j Y(x) represents the predicted sea surface temperature at the time corresponding to the FY-3E satellite, i.e., the first sea surface temperature at the time of prediction; i y j , t k F(x) represents the sea surface temperature value observed by the FY-3E satellite at a given time, i.e., the first sea surface temperature at that given time; similarly, F(x) represents the sea surface temperature observed by the FY-3E satellite at a given time. i y j , t k ) and F(xi y j , t0) represent the second sea surface temperature corresponding to the FY-4A satellite at the existing time and the predicted time, respectively.

[0110] Therefore, based on the above formula, the first sea surface temperature of each pixel at the predicted time can be calculated by fusing the first sea surface temperature and the second sea surface temperature at the existing time for each pixel, as well as the second sea surface temperature at the predicted time.

[0111] It should be noted that in reality, the differences between different locations and times are not negligible. Furthermore, due to sea surface heat conduction, the temperature of the central pixel will be affected by neighboring pixels. Therefore, a weight matrix is ​​introduced using neighboring pixels to weight the central pixel.

[0112] Furthermore, a neighborhood convolution operation can be performed on the first sea surface temperature of each pixel at the prediction time and the corresponding weight factor of each pixel. As the neighborhood window moves one step, the sea surface temperature of the center pixel within the neighborhood window at the prediction time is calculated. This process can be expressed by the following formula:

[0113]

[0114] Where w represents the width of the weight matrix, which is also the size of the neighborhood window; Y(x w / 2 y w / 2 ,t0) represents the value of the center pixel within the neighborhood window at the prediction time, C ij Let represent the weighting factor matrix, where the weighting factors are used to characterize the contribution of pixels adjacent to the center pixel within the neighborhood window to the sea surface temperature value estimated by the center pixel.

[0115] Furthermore, in an embodiment of the present invention, after the neighborhood window has traversed every pixel in the pixel matrix M, the sea surface temperature prediction result of the target sea area under the target spatial resolution and the target temporal resolution can be obtained.

[0116] In the embodiments of the present invention, by using neighborhood convolution operation to perform fusion analysis on two types of satellite observation data of images with the same target spatial resolution, it is possible to effectively achieve high spatiotemporal resolution data fusion of high spatial resolution satellite image data and high temporal resolution satellite image data at the time scale, thereby improving the spatiotemporal resolution and prediction accuracy of SST fusion products.

[0117] The ocean surface temperature (SST) prediction method of this invention utilizes high spatial resolution first satellite imagery data and high temporal resolution second satellite imagery data to acquire two types of satellite observation data of the target sea area at the same target spatial resolution. Then, by performing data fusion analysis on the two types of satellite observation data of the image at the same target spatial resolution, the predicted sea surface temperature of the target sea area at the target spatiotemporal resolution is obtained. This achieves high spatiotemporal resolution data fusion at a time scale between high spatial resolution satellite imagery data and high temporal resolution satellite imagery data, effectively obtaining high-precision, high spatiotemporal resolution, and high-coverage SST fusion products. This well meets the rapid development needs of fields such as marine resource development and greatly improves the application value of SST products.

[0118] Based on the above embodiments, as an optional embodiment, the weighting factor is determined based on the spectral volatility factor, the temporal volatility factor, and the spatial distance factor;

[0119] Among them, the spectral fluctuation factor is determined based on the difference between the first sea surface temperature and the second sea surface temperature at the existing time for each pixel; the temporal fluctuation factor is determined based on the average sea surface temperature at all existing times extracted from the second observation data of each pixel; and the spatial distance factor is determined based on the distance between each pixel and its corresponding center pixel.

[0120] Specifically, in the embodiments of the present invention, considering the effects of spectral differences, temperature fluctuations, and spatial distances of ground features, a weighting factor C is introduced. ij Specifically, it can be based on the spectral volatility factor S ij Time volatility factor T ij and spatial distance factor D ij To determine.

[0121] In an embodiment of the present invention, the spectral fluctuation factor S ij It is determined based on the difference between the first sea surface temperature and the second sea surface temperature at a given time for each pixel. Specifically, due to the influence of sensor scanning attitude, cloud cover, and aerosol absorption at different spatial locations, there is an inversion bias between FY-3E and FY-4A satellite data. This bias can be calculated by the absolute value of the difference detected by different sensors at the same pixel, i.e., by calculating the absolute value of the difference between the first and second sea surface temperatures at a given time for each pixel. This can be obtained using the following formula:

[0122] S ij ′=|Y(x i y j , t k )-F(xi y j , t k )|;

[0123] At the same moment, S ij The larger the value of ′, the greater the difference between high spatial resolution images and low spatial resolution images, and the lower the weight should be assigned. Therefore, the above factors need to be transformed to determine the spectral fluctuation factor S. ij ,Right now:

[0124]

[0125] In an embodiment of the present invention, the time volatility factor T ij This is determined based on the average sea surface temperature at all available times, extracted from the second observation data of each pixel. Specifically, the sea surface temperature fluctuation within the weighted range is extracted from the high temporal resolution FY-4A imagery of the entire 24-hour period at the prediction time. The coverage area varies at different times due to cloud cover. To integrate more information, the average of all valid temperature times is extracted from the 24-hour data, i.e., the average sea surface temperature at all available times is extracted from the second observation data of each pixel, resulting in:

[0126]

[0127] Where m represents the total number of valid data points per pixel in 24 hours, and T represents the total number of valid data points per day. ij The larger the value of ′, the greater the fluctuation of sea surface temperature in the region, and the lower the weight should be assigned. Therefore, the above factors need to be transformed to determine the time fluctuation factor T. ij ,Right now:

[0128]

[0129] In an embodiment of the present invention, the spatial distance factor D ij It is determined based on the distance between each pixel and its corresponding central pixel. Specifically, according to geographical laws, features that are geographically closer have similar properties, and sea surface temperatures are also closer. However, spatial isolation leads to increased differences, hence the introduction of geographic distance:

[0130]

[0131] Pixels farther from the center pixel are assigned lower weights, while pixels closer to the center pixel have higher spatial similarity and should be assigned higher weights in the calculation. Therefore, the above factors need to be transformed to determine the spatial distance factor D. ij ,Right now:

[0132]

[0133] Furthermore, in this embodiment of the invention, by adjusting the spectral fluctuation factor S... ij Time volatility factor T ij and spatial distance factor D ij Perform a product operation and then normalize to obtain the weighting factor C. ij Its specific expression is as follows:

[0134]

[0135] Among them, C ij ′=S ij *T ij *D ij .

[0136] The method of this invention introduces a weighting factor by considering the effects of sensor differences, temperature fluctuations, and spatial distances to ground objects. This factor is used as the convolution kernel in the neighborhood convolution operation, which facilitates the subsequent fusion calculation of two types of satellite image data under the same target spatial resolution and can further improve the accuracy of the acquired SST fusion product.

[0137] Based on the above embodiments, as an optional embodiment, after determining the second observation data for each pixel in the image, the method further includes:

[0138] The image is divided into multiple sub-image blocks according to the preset pixel size;

[0139] The first and second observation data of each pixel in each sub-image block are fused and analyzed to obtain the sea surface temperature prediction result of each sub-image block at the target temporal resolution.

[0140] The sea surface temperature prediction results of each sub-image block at the target temporal resolution are merged to obtain the sea surface temperature prediction results of the target sea area at both the target spatial resolution and the target temporal resolution.

[0141] Specifically, the preset pixel size described in the embodiments of the present invention can be 1000*1000 pixels or 500*500 pixels, which can be set according to actual calculation needs.

[0142] In an embodiment of the present invention, in order to accelerate the efficiency of fusion calculation of high spatial resolution satellite image data and high temporal resolution satellite image data, the image of the target sea area at the target spatial resolution can be divided, which is essentially the division of the pixel matrix M of the image.

[0143] In an embodiment of the present invention, the 110000*13000 pixel matrix M of the image can be divided into blocks according to a preset pixel size, such as dividing the image into 11*13 sub-image blocks according to a pixel size of 1000*1000.

[0144] Furthermore, by processing each sub-image block's data within a single kernel and employing neighborhood convolution, the first and second observation data for each pixel within each sub-image block are fused and analyzed to obtain the sea surface temperature prediction result for each sub-image block at the target temporal resolution. During this process, a shared memory strategy can be used to share the weight factor matrix during computation, thereby reducing memory consumption.

[0145] Finally, the sea surface temperature prediction results of each sub-image block at the target temporal resolution are merged, and the sea surface temperature prediction results are restored to the pixel size of the 110000*13000 pixel matrix M. Thus, the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution can be obtained.

[0146] In embodiments of the present invention, high-performance computing power, including a 24-core i9 processor and 64G of storage memory, can be used to perform data fusion calculations on one day of satellite imagery data collected by FY-3E and 24 hours of imagery data collected by FY-4A, thereby completing the production of fused data.

[0147] The method of this invention, by dividing the target sea area into images at the target spatial resolution and performing block data fusion analysis, can effectively accelerate the efficiency of high spatial resolution and high temporal resolution satellite image data fusion calculation and improve the performance of the acquired SST fusion products.

[0148] Based on the above embodiments, as an optional embodiment, after fusing and analyzing the first and second observation data for each pixel to determine the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution, the method further includes:

[0149] According to the preset sea surface temperature grading rendering standard, based on the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution, a sea surface temperature prediction map of the target sea area at the target spatiotemporal resolution is generated; the spatiotemporal resolution is determined based on the target spatial resolution and the target temporal resolution.

[0150] Specifically, the preset sea surface temperature grading rendering standard described in the embodiments of the present invention refers to a color rendering standard determined based on the sea surface temperature grading standard.

[0151] The sea surface temperature prediction map in this embodiment refers to a dynamic map used to display the predicted SST distribution of a target sea area, based on the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution, and obtained through a classification and grading rendering method.

[0152] Furthermore, in an embodiment of the present invention, according to a preset sea surface temperature grading rendering standard, based on the sea surface temperature prediction results of each pixel in the target sea area at the target spatial resolution and the target temporal resolution, color rendering is performed pixel by pixel to generate a sea surface temperature prediction map of the target sea area at the target spatiotemporal resolution, and the map is displayed.

[0153] The method of this invention can effectively improve the spatiotemporal mapping accuracy of SST fusion products by performing hierarchical rendering and mapping of the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution.

[0154] Figure 3 This is the second flowchart of the ocean surface temperature prediction method provided by the present invention, as shown below. Figure 3 As shown, the first satellite image is an FY-3E satellite image, and the second satellite image is an FY-4A satellite image. The specific steps of this method include:

[0155] First, multi-core parallel data integration was performed on FY-3E (1km / day) satellite imagery to complete multi-core parallel resampling of FY-3E data, resulting in an image of the observed sea area at a spatial resolution of 1km. Then, by performing data preprocessing on the FY-3E satellite imagery observation data, including missing value handling, the processed observation data was registered to the pixel matrix of the image at a spatial resolution of 1km, obtaining the first observation data for each pixel in the image.

[0156] Then, for the FY-4A (4km / hour) satellite imagery, data preprocessing was performed, including missing value handling and latitude / longitude extraction. The low spatial resolution FY-4A imagery was resampled to the 1km spatial resolution of FY-3E, ensuring that the FY-3E satellite imagery data maintained the same pixel size and sea area coverage as the FY-4A satellite imagery. Subsequently, the FY-4A satellite imagery observation data was registered to the pixel matrix of the 1km spatial resolution imagery, determining the existing observation data for each pixel in the pixel matrix.

[0157] Furthermore, based on the sea surface temperature change over time and the existing observation data corresponding to each pixel in the pixel matrix, multi-time data linear fitting is performed to estimate the observation data corresponding to the predicted time for each pixel in the pixel matrix. Thus, based on the existing observation data corresponding to each pixel and the observation data corresponding to the predicted time for each pixel, the second observation data for each pixel in the image is obtained.

[0158] Next, considering the effects of neighborhood pixel homogeneity, temperature temporal variation fluctuations, and geographic distance, the temporal fluctuation factor, spectral fluctuation factor, and spatial distance factor are calculated separately, and the weighting factor is integrated across these multiple factors. Based on the weighting factor, neighborhood convolution operations are performed on the first and second observation data for each pixel in the image, and the image can be divided into blocks. Using multi-kernel parallel fusion computation, the SST fusion result of the observed sea area at a target spatiotemporal resolution of 1 km / 1 hour is obtained.

[0159] Finally, the accuracy of the SST fusion results was verified and evaluated using the Argo buoy dataset.

[0160] Figure 4 This is a schematic diagram illustrating the accuracy verification results of the ocean surface temperature prediction method provided by this invention, as shown below. Figure 4 As shown, in this embodiment, to verify the data accuracy of the SST fusion results, 15 days of 24-hour daily 1km resolution data prediction were performed. Spatiotemporal matching verification was conducted using Argo buoy data. The root mean square error (RMSE) of the SST fusion results was 1.115, the mean absolute error (MAE) was less than 1%, and the correlation R... 2 Greater than 0.9, with stable accuracy.

[0161] The ocean surface temperature prediction device provided by the present invention is described below. The ocean surface temperature prediction device described below can be referred to in correspondence with the ocean surface temperature prediction method described above.

[0162] Figure 5 This is a schematic diagram of the ocean surface temperature prediction device provided by the present invention, as shown below. Figure 5 As shown, it includes:

[0163] The processing module 510 is used to process data to determine the first observation data of each pixel in the image of the target sea area at the target spatial resolution based on the first satellite image data, and to determine the second observation data of each pixel in the image based on the second satellite image data.

[0164] Wherein, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data;

[0165] The prediction module 520 is used to perform fusion analysis on the first observation data and the second observation data of each pixel to determine the sea surface temperature prediction result of the target sea area under the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0166] The ocean surface temperature prediction device described in this embodiment can be used to execute the above-described ocean surface temperature prediction method embodiment. Its principle and technical effects are similar, and will not be repeated here.

[0167] The ocean surface temperature (SST) prediction device of this invention acquires two types of satellite observation data of a target sea area at the same target spatial resolution by utilizing high spatial resolution first satellite image data and high temporal resolution second satellite image data. Then, by performing data fusion analysis on the two types of satellite observation data of the image at the same target spatial resolution, the prediction result of the sea surface temperature of the target sea area at the target spatiotemporal resolution is obtained. This realizes high spatiotemporal resolution data fusion of high spatial resolution satellite image data and high temporal resolution satellite image data at the time scale, which can effectively obtain high-precision, high spatiotemporal resolution and high coverage SST fusion products, which well meet the rapid development needs of marine resource development and other fields, and greatly improve the application value of SST products.

[0168] Figure 6 This is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as... Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute the ocean surface temperature prediction method provided by the above methods. The method includes: determining first observation data for each pixel in the image of the target sea area at a target spatial resolution based on first satellite image data, and determining second observation data for each pixel in the image based on second satellite image data; wherein the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data; performing fusion analysis on the first observation data and the second observation data of each pixel to determine the sea surface temperature prediction result of the target sea area at the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0169] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0170] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the ocean surface temperature prediction method provided by the above methods. The method includes: determining first observation data for each pixel in an image of a target sea area at a target spatial resolution based on first satellite image data, and determining second observation data for each pixel in the image based on second satellite image data; wherein the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data; performing fusion analysis on the first and second observation data for each pixel to determine the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0171] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the ocean surface temperature prediction method provided by the methods described above. This method includes: determining first observation data for each pixel in an image of a target sea area at a target spatial resolution based on first satellite image data, and determining second observation data for each pixel in the image based on second satellite image data; wherein the spatial resolution of the first satellite image data is higher than the spatial resolution of the second satellite image data, and the temporal resolution of the second satellite image data is higher than the temporal resolution of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data; and performing a fusion analysis on the first and second observation data for each pixel to determine a predicted sea surface temperature for the target sea area at the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data.

[0172] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0173] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting ocean surface temperature, characterized in that, include: Based on the first satellite image data, the first observation data of each pixel in the image at the target spatial resolution of the target sea area is determined, and based on the second satellite image data, the second observation data of each pixel in the image is determined. Wherein, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data; The first and second observation data of each pixel are fused and analyzed to determine the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data. The step of fusing and analyzing the first and second observation data for each pixel to determine the predicted sea surface temperature for the target sea area at the target spatial resolution and target temporal resolution includes: The first sea surface temperature of each pixel at the predicted time is determined by fusing the first sea surface temperature at the existing time and the second sea surface temperature at the existing time, as well as the second sea surface temperature at the predicted time. The second observation data includes the second sea surface temperature at the existing time and the second sea surface temperature at the predicted time. The first observation data includes the first sea surface temperature at the existing time. Based on the first sea surface temperature of each pixel at the prediction time and the weight factor corresponding to each pixel, a neighborhood convolution operation is performed to determine the sea surface temperature of the center pixel in the neighborhood window at the prediction time; the weight factor is used to characterize the contribution of the pixels adjacent to the center pixel in the neighborhood window to the sea surface temperature of the center pixel. After the neighborhood window has moved and traversed every pixel in the image, the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution is obtained.

2. The ocean surface temperature prediction method according to claim 1, characterized in that, The weighting factor is determined based on the spectral volatility factor, the temporal volatility factor, and the spatial distance factor. The spectral fluctuation factor is determined based on the difference between the first sea surface temperature and the second sea surface temperature at an existing time for each pixel; the temporal fluctuation factor is determined based on the average sea surface temperature at all existing times extracted from the second observation data of each pixel; and the spatial distance factor is determined based on the distance between each pixel and its corresponding center pixel.

3. The ocean surface temperature prediction method according to claim 1, characterized in that, After determining the second observation data for each pixel in the image, the method further includes: The image is divided into multiple sub-image blocks according to a preset pixel size; The first and second observation data of each pixel in each sub-image block are fused and analyzed to obtain the sea surface temperature prediction result of each sub-image block at the target time resolution. The sea surface temperature prediction results of each of the sub-image blocks at the target temporal resolution are merged to obtain the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution.

4. The ocean surface temperature prediction method according to any one of claims 1-3, characterized in that, The first satellite image data includes a first satellite image and corresponding observation data; the first observation data, based on the first satellite image data, determining each pixel in the image at the target spatial resolution for the target sea area, includes: The first satellite image is resampled to determine the image of the target sea area at the target spatial resolution; The observation data corresponding to the first satellite image is registered to the pixel matrix of the image to obtain the first observation data of each pixel in the image.

5. The ocean surface temperature prediction method according to any one of claims 1-3, characterized in that, The second satellite image data includes the sea surface temperature change over time, the second satellite image, and the corresponding observation data; the step of determining the second observation data for each pixel in the image based on the second satellite image data includes: The second satellite image is resampled to determine the image of the target sea area at the target spatial resolution; Register the observation data corresponding to the second satellite image into the pixel matrix of the image, and determine the existing observation data corresponding to each pixel in the pixel matrix; Based on the sea surface temperature change over time and the existing observation data corresponding to each pixel in the pixel matrix, data fitting is performed to determine the observation data corresponding to the predicted time for each pixel in the pixel matrix. Based on the observation data of the existing time corresponding to each pixel and the observation data of the predicted time corresponding to each pixel, the second observation data of each pixel in the image is obtained.

6. The ocean surface temperature prediction method according to any one of claims 1-3, characterized in that, After fusing and analyzing the first and second observation data for each pixel to determine the predicted sea surface temperature of the target sea area at the target spatial resolution and target temporal resolution, the method further includes: According to the preset sea surface temperature grading rendering standard, based on the sea surface temperature prediction results of the target sea area at the target spatial resolution and the target temporal resolution, a sea surface temperature prediction map of the target sea area at the target spatiotemporal resolution is generated; the spatiotemporal resolution is determined based on the target spatial resolution and the target temporal resolution.

7. A device for predicting ocean surface temperature, characterized in that, include: The processing module is used to determine the first observation data of each pixel in the image of the target sea area at the target spatial resolution based on the first satellite image data, and to determine the second observation data of each pixel in the image based on the second satellite image data. Wherein, the spatial resolution of the first satellite image data is higher than that of the second satellite image data, and the temporal resolution of the second satellite image data is higher than that of the first satellite image data; the target spatial resolution is determined based on the spatial resolution of the first satellite image data; The prediction module is used to perform fusion analysis on the first and second observation data of each pixel to determine the predicted sea surface temperature of the target sea area under the target spatial resolution and the target temporal resolution; the target temporal resolution is determined based on the temporal resolution of the second satellite image data. The step of fusing and analyzing the first and second observation data for each pixel to determine the predicted sea surface temperature for the target sea area at the target spatial resolution and target temporal resolution includes: The first sea surface temperature of each pixel at the predicted time is determined by fusing the first sea surface temperature at the existing time and the second sea surface temperature at the existing time, as well as the second sea surface temperature at the predicted time. The second observation data includes the second sea surface temperature at the existing time and the second sea surface temperature at the predicted time. The first observation data includes the first sea surface temperature at the existing time. Based on the first sea surface temperature of each pixel at the prediction time and the weight factor corresponding to each pixel, a neighborhood convolution operation is performed to determine the sea surface temperature of the center pixel in the neighborhood window at the prediction time; the weight factor is used to characterize the contribution of the pixels adjacent to the center pixel in the neighborhood window to the sea surface temperature of the center pixel. After the neighborhood window has moved and traversed every pixel in the image, the predicted sea surface temperature of the target sea area at the target spatial resolution and the target temporal resolution is obtained.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the ocean surface temperature prediction method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the ocean surface temperature prediction method as described in any one of claims 1 to 6.

Images