Sea ice extraction method based on rotating fan scatterometer

Abstract

The embodiment of the application provides a sea ice extraction method based on a rotating sector scatterometer, comprising: acquiring polar backscatter observation data collected by the rotating sector scatterometer, wherein the polar backscatter observation data comprises a plurality of groups of backscatter observation data obtained by performing multiple observations on a polar region by the rotating sector scatterometer; performing principal component analysis on the plurality of groups of backscatter observation data to obtain principal component rotation feature data of the backscatter observation data; and performing sea ice feature extraction based on the principal component rotation feature data to obtain distribution information of sea ice in the polar region.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of microwave remote sensing technology, and in particular to a sea ice extraction method, computing device and computer storage medium based on a rotating fan scatterometer. Background Technology

[0002] Sea ice acts as an amplifier of global climate change, and changes in polar sea ice have a significant impact on global climate change. Therefore, accurately obtaining information on the distribution of sea ice in the polar regions is of great importance.

[0003] In related technologies, fixed sector antenna scatterometers or rotating pen antenna scatterometers are typically used to collect polar backscattering observation data, and the distribution of sea ice in polar regions is analyzed through polar backscattering observation data.

[0004] In the process of realizing the concept of this invention, the inventors discovered that the sensor antenna of the fixed fan-shaped antenna scatterometer cannot perform rotational scanning and can only receive backscattered signals from a fixed azimuth direction, while the sensor antenna of the rotating pen-shaped scatterometer can only acquire backscattered signals from a fixed incident angle direction. Therefore, the extraction efficiency of sea ice distribution is low. Summary of the Invention

[0005] This invention provides a method and apparatus for sea ice extraction based on a rotating fan-shaped scatterometer.

[0006] In a first aspect, embodiments of the present invention provide a method for sea ice extraction using a rotating fan-shaped scatterometer, comprising:

[0007] The polar backscattering observation data collected by the rotating sector scatterer includes multiple sets of backscattering observation data obtained by the rotating sector scatterer from multiple observations of the polar region based on different azimuth angles and incident angles.

[0008] Principal component analysis was performed on the multiple sets of backscattering observation data to obtain the principal component rotation feature data of the backscattering observation data;

[0009] Sea ice features are extracted based on the principal component rotation feature data to obtain sea ice distribution information in the polar regions.

[0010] Secondly, embodiments of the present invention provide a sea ice extraction device for a rotating fan-shaped scatterometer, comprising:

[0011] The first acquisition module is used to acquire polar backscattering observation data collected by the rotating fan-shaped scattering meter. The polar backscattering observation data includes multiple sets of backscattering observation data obtained by the rotating fan-shaped scattering meter from multiple observations of the polar region based on different azimuth angles and incident angles.

[0012] The analysis module is used to perform principal component analysis on the multiple sets of backscattering observation data to obtain principal component rotation feature data of the backscattering observation data;

[0013] The extraction module is used to extract sea ice features based on the principal component rotation feature data to obtain the distribution information of sea ice in the polar regions.

[0014] These or other aspects of the invention will become more apparent from the following description of the embodiments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 The flowchart illustrates a sea ice extraction method based on a rotating sector scatterometer according to an embodiment of the present invention.

[0017] Figure 2 The illustration shows a schematic diagram of a sea ice extraction method based on a rotating sector scatterometer according to an embodiment of the present invention;

[0018] Figure 3 A schematic diagram illustrating the generation of prediction results is shown.

[0019] Figure 4 The diagram schematically illustrates a block diagram of a sea ice extraction device based on a rotating fan-shaped scatterometer according to an embodiment of the present invention.

[0020] Figure 5 The diagram illustrates a block diagram of a computing device according to one embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0022] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0023] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this invention are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0024] Sea ice acts as an amplifier of global climate change, and changes in polar sea ice have a significant impact on global climate change. Therefore, accurately obtaining information on the distribution of sea ice in the polar regions is of great importance.

[0025] Due to their remote location and harsh environment, direct observation of sea ice in polar regions is not only inefficient and costly, but also impossible to conduct large-scale observations. The rise and development of satellite remote sensing has provided an effective means of monitoring sea ice. Compared with optical remote sensing, microwave remote sensing is not affected by light and clouds and rain, and can conduct all-day, all-weather observations.

[0026] A satellite scatterometer is an active, non-imaging radar microwave remote sensing system that detects information about targets by transmitting microwave pulse signals to the Earth's surface and receiving the backscattered echo signals.

[0027] Satellite scatterometers primarily operate at frequencies of 5.3 GHz (C-band) and 13.5 GHz (Ku-band). The C-band, with its longer wavelength, offers stronger atmospheric penetration and is less affected by factors such as clouds and rain; while the Ku-band, with its higher frequency, provides a more pronounced contrast between ice and water. Satellite scatterometers can be categorized by their scanning method into fixed-antenna scatterometers and rotating-scan scatterometers. Fixed-antenna scatterometers, due to the fan-shaped ground footprint of their antenna beam, are also known as fixed-fan-beam scatterometers. Representative scatterometers include SASS, AMI-SCAT, NSCAT, and ASCAT. Fixed-fan-beam scatterometers can only provide backscattering coefficient observations at a few specific azimuth angles and cannot obtain continuous observations of the azimuth backscattering coefficient. Therefore, using fixed-fan-beam scatterometers for polar sea ice observations presents a technical challenge of low sea ice extraction efficiency.

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Figure 1 The flowchart illustrating a sea ice extraction method based on a rotating sector scatterometer according to an embodiment of the present invention is shown in the figure. Figure 1 As shown, the sea ice extraction method based on a rotating sector scatterometer may specifically include the following steps:

[0030] 101. Acquire polar backscattering observation data collected by a rotating sector scatterometer. The polar backscattering observation data includes multiple sets of backscattering observation data obtained by the rotating sector scatterometer from multiple observations of the polar region based on different azimuth angles and incident angles.

[0031] 102. Principal component analysis was performed on multiple sets of backscattering observation data to obtain the principal component rotation feature data of the backscattering observation data;

[0032] 103. Based on principal component rotation feature data, sea ice feature extraction was performed to obtain the distribution information of sea ice in the polar region.

[0033] According to embodiments of the present invention, rotating scanning scatterometers mostly employ two fixed incident angle beams, namely an inner beam and an outer beam, provided by dual feed sources with VV or HH polarization. Rotating scanning scatterometers can be further classified into pencil beam scatterometers and fan beam scatterometers based on beam shape. Examples of pencil beam scatterometers include SeaWinds, OSCAT, and HSCAT, while examples of fan beam scatterometers include CSCAT and WindRad. Rotating pencil beam scatterometers can obtain continuous observations in the azimuth direction, but only use 1-2 point beams for discrete observations in the elevation direction, resulting in a smaller measurement footprint. Rotating fan beam scatterometers can simultaneously obtain continuous observations of the backscattering coefficients in both the elevation and azimuth directions. Their greatest advantage lies in shortening the observation time for multiple azimuth angle combinations of the same target, which is beneficial for highly correlated measurement data.

[0034] According to embodiments of the present invention, the polar region may include the Antarctic and / or the Arctic.

[0035] In a preferred embodiment of the present invention, the rotating sector scatterometer can be implemented as the Sino-French Oceanographic Satellite. The Sino-French Oceanographic Satellite operates in a sun-synchronous orbit, orbiting the Earth 14 to 16 times per day, covering the North and South Poles once per orbit. Its onboard satellite scatterometer adopts a single-frequency (Ku) dual-polarization (VV and HH) rotating sector scanning mechanism, generating 4 to 8 backscattering observation data per observation.

[0036] According to an embodiment of the present invention, by performing principal component analysis on multiple sets of backscattering observation data, principal component rotation feature data that can reflect sea ice characteristics can be extracted from multiple sets of backscattering observation data, thereby improving the accuracy of extracting sea ice distribution information based on principal component rotation feature data.

[0037] In embodiments of the present invention, a pre-trained machine learning model can be used to predict principal component rotation feature data to obtain sea ice distribution information in polar regions; however, it is not limited to this, the machine learning model can also be trained using principal component rotation feature data to obtain a trained sea ice extraction model, and then the sea ice extraction model can be used to obtain sea ice distribution information in polar regions.

[0038] According to embodiments of the present invention, the distribution information of sea ice in polar regions may include, for example, sea ice type information and sea ice presence probability information.

[0039] According to an embodiment of the present invention, each set of backscattering observation data includes 8 sets of vertical polarization data and 8 sets of horizontal polarization data.

[0040] According to an embodiment of the present invention, a set of backscattering observation data can be the data collected by a rotating sector scatterometer during one polar orbit. Each time the rotating sector scatterometer orbits the poles, it generates both vertical polarization data and horizontal polarization data. By collecting both vertical and horizontal polarization data, the rotating sector scatterometer can simultaneously acquire ice surface information and sea surface information.

[0041] According to an embodiment of the present invention, principal component analysis is performed on multiple sets of backscattering observation data to obtain principal component rotation feature data of the backscattering observation data, including:

[0042] For each set of backscattering observation data, the polarization ratio data is calculated based on both vertical and horizontal polarization data.

[0043] A feature matrix is ​​constructed based on multiple sets of vertical polarization data, horizontal polarization data, and polarization ratio data.

[0044] Principal component analysis is performed on the feature matrix to obtain principal component rotation feature data.

[0045] According to embodiments of the present invention, after acquiring multiple sets of backscattering observation data, each set of backscattering observation data can be preprocessed separately. Specifically, polarimetric ratio data can be calculated based on the horizontal and vertical polarization data of each set of backscattering observation data. The polarimetric ratio data can be obtained by calculating the ratio of vertical polarization data to horizontal polarization data. For sea ice-covered areas, the polarimetric ratio data can be used to extract sea ice structure information, such as sea ice type, thickness, and concentration.

[0046] According to an embodiment of the present invention, the construction of the feature matrix based on multiple sets of vertical polarization data, horizontal polarization data, and polarization ratio data can be specifically implemented as follows:

[0047] The vertical polarization data, horizontal polarization data, and polarization ratio data were obtained from each set of backscattering observation data.

[0048] The first characteristic matrix, the second characteristic matrix, and the third characteristic matrix are constructed based on the vertical polarization data, horizontal polarization data, and polarization ratio data in each set of backscattering observation data.

[0049] According to an embodiment of the present invention, principal component analysis is performed on the feature matrix to obtain the principal component rotation feature data of the backscattering observation data, which can be specifically implemented as follows:

[0050] Principal component analysis was performed on the first feature matrix, the second feature matrix, and the third feature matrix respectively to obtain the first feature data, the second feature data, and the third feature data corresponding to the first feature matrix, the second feature matrix, and the third feature matrix, respectively.

[0051] The first feature data, the second feature data, and the third feature data are concatenated to obtain the principal component rotation feature data.

[0052] According to an embodiment of the present invention, by retaining the vertical polarization data and horizontal polarization data in each set of backscattering observation data, and calculating the polarization ratio data based on the vertical polarization data and horizontal polarization data, and performing principal component analysis on the matrix composed of the vertical polarization data, horizontal polarization data, and polarization ratio data, the characteristics of sea ice observed by the rotating satellite scatterometer can be preserved, thereby improving the accuracy of extracting principal component rotation feature data from the characteristics of sea ice through principal component analysis.

[0053] According to an embodiment of the present invention, principal component analysis is performed on the feature matrix to obtain principal component rotation feature data of the backscattering observation data, including:

[0054] The feature matrix is ​​standardized to obtain the standardized matrix;

[0055] Transform the standardized matrix into a covariance matrix;

[0056] Calculate the eigenvalues ​​and eigenvectors of the covariance matrix;

[0057] The feature vectors corresponding to the feature values ​​that meet the preset conditions are used as the target feature vectors.

[0058] Principal component rotation feature data are obtained based on the target feature vector and the normalized matrix.

[0059] According to an embodiment of the present invention, when performing principal component analysis, principal component analysis can be performed on the obtained first feature matrix, second feature matrix and third feature matrix respectively. Therefore, in the specific implementation process of principal component analysis, the first feature matrix, second feature matrix and third feature matrix can be processed in the same way. The following description focuses on performing principal component analysis on the first feature matrix.

[0060] According to an embodiment of the present invention, in order to eliminate the weight imbalance problem caused by differences in dimensions and numerical ranges among different variables, and to ensure that the importance of each feature contributes equally to the extraction of sea ice features, the first feature matrix can first be standardized to obtain a standardized matrix. The standardized matrix obtained after standardization can retain the important information of the original data while removing unnecessary noise and irrelevant information, thus better reflecting the relationship between the features.

[0061] According to embodiments of the present invention, the first feature matrix can be standardized by, for example, eigenvalue decomposition, Z-score normalization, or range scaling.

[0062] According to an embodiment of the present invention, the standardized matrix can be transformed into a covariance matrix by the following formulas (1) and (2).

[0063] X′=X-μ; (1)

[0064]

[0065] Where X can represent the normalization matrix, μ can represent the mean of each dimension in the normalization matrix X, S can represent the covariance matrix, and m can represent the dimension of the normalization matrix X.

[0066] According to an embodiment of the present invention, the preset condition can be the top m largest feature values ​​among all feature values, where m can be a positive integer.

[0067] According to an embodiment of the present invention, after obtaining the target feature vector, the transpose of the target feature vector can be calculated first to obtain the transposed feature vector, and then the transposed feature vector is multiplied by the normalized matrix to obtain the principal component rotation feature data.

[0068] According to an embodiment of the present invention, the principal component rotation feature data includes historical principal component rotation feature data and current day principal component rotation feature data.

[0069] According to an embodiment of the present invention, in order to increase the amount of data, when acquiring polar backscattering observation data, the backscattering observation data of the rotating sector scatterometer for the previous 7 days (including the current day; if observations are missing, data can be acquired further back, but not exceeding 20 days) can be acquired as historical backscattering observation data. The observation data acquired by the rotating sector scatterometer on the current date can be the backscattering observation data for that day. Furthermore, after performing principal component analysis on the backscattering observation data, the data corresponding to the historical backscattering observation data in the principal component rotation of the backscattering observation data can be considered historical principal component rotation feature data, and the data corresponding to the current day's backscattering observation data can be considered the current day's principal component rotation feature data.

[0070] According to an embodiment of the present invention, sea ice feature extraction based on principal component rotation feature data yields sea ice distribution information in the polar regions, including:

[0071] A sea ice extraction model is generated by training historical principal component rotation feature data. The sea ice extraction model includes multiple sub-models.

[0072] Input the principal component rotation feature data of the day into the sea ice extraction model and output the distribution information.

[0073] According to embodiments of the present invention, the sub-model can be implemented as, for example, a multivariate logistic regression classification model, a K-nearest neighbor classification model, a decision tree model, a BP neural network model, a support vector machine model, a Gaussian Naive Bayes model, a random forest model, etc.

[0074] According to embodiments of the present invention, by constructing a sea ice feature extraction model using multiple sub-models, the results of sea ice feature extraction can be avoided due to classification errors in one model, thereby improving the accuracy of sea ice feature extraction.

[0075] According to an embodiment of the present invention, training a sea ice extraction model using historical principal component rotation feature data includes:

[0076] Obtain a pre-defined polar sampling area and reference information on the sea ice distribution corresponding to the sampling area;

[0077] Determine the observation data corresponding to the sampling area from the historical principal component rotation feature data;

[0078] Determine the reference distribution information of sea ice corresponding to the observation data from the sea ice distribution reference information;

[0079] A sample dataset is generated based on principal component rotation feature data and sea ice extraction reference distribution information.

[0080] A sea ice extraction model was trained using a sample dataset.

[0081] According to embodiments of the present invention, sea ice distribution reference information may include the sea ice type corresponding to each polar sampling area. The sea ice type may include, for example, open water, open sea ice, and closed sea ice. Specifically, sampling areas with a sea ice percentage of 0-30% may be defined as open water, sampling areas with a sea ice percentage of 30-70% may be defined as open sea ice, and sampling areas with a sea ice percentage of 70-100% may be defined as closed sea ice.

[0082] According to an embodiment of the present invention, by spatiotemporally matching historical principal component rotation feature data with polar sampling areas and sea ice distribution reference information, multiple sampling areas in the historical principal component rotation feature data can be labeled with sea ice distribution, thereby constructing sample data for training the sea ice extraction model and realizing supervised training of the sea ice extraction model.

[0083] According to an embodiment of the present invention, training a sea ice extraction model using sample data includes:

[0084] The sample dataset is divided into a training dataset and a test dataset;

[0085] Multiple sub-models were trained using the training dataset to obtain the sea ice extraction model.

[0086] According to an embodiment of the present invention, the sample dataset can be divided into a training dataset and a test dataset in a 7:3 ratio.

[0087] According to an embodiment of the present invention, since the sea ice feature extraction model is composed of multiple sub-models, each sub-model can be trained separately.

[0088] According to an embodiment of the present invention, the principal component rotation feature data of the day is input into the sea ice extraction model, and the output distribution information includes:

[0089] The test dataset is input into multiple sub-models, and the first prediction result corresponding to each of the multiple sub-models is output.

[0090] The confusion matrix is ​​obtained based on the first prediction result and the reference distribution information extracted from sea ice.

[0091] The weight factors corresponding to each sub-model are obtained based on the confusion matrix;

[0092] The principal component rotation feature data of the day are input into multiple sub-models respectively, and the second prediction results corresponding to the multiple sub-models are output respectively.

[0093] Based on multiple weighting factors and the second prediction results corresponding to each weighting factor, a third prediction result corresponding to each sub-model is obtained;

[0094] Distribution information is determined from multiple third-party prediction results.

[0095] According to an embodiment of the present invention, firstly, the test dataset can be input into multiple sub-models respectively to obtain the first prediction result Y of the multiple sub-models. Pred .

[0096] According to embodiments of the present invention, the performance of a model can be evaluated using a confusion matrix by comparing the true labels with the model's predictions to statistically determine the classification between different categories.

[0097] According to an embodiment of the present invention, obtaining the confusion matrix based on the first prediction result and sea ice extraction reference distribution information can be specifically achieved as follows:

[0098]

[0099] In this context, True Positive = TP represents the number of true values ​​that are positive and the model output is positive; False Negative = FN represents the number of true values ​​that are positive and the model output is negative; False Negative = FN represents the number of true values ​​that are positive and the model output is negative; and True Negative = TN represents the number of true values ​​that are negative and the model output is negative.

[0100] According to an embodiment of the present invention, after generating the confusion matrix, the weight factors corresponding to each sub-model can be obtained based on the confusion matrix.

[0101] According to an embodiment of the present invention, the weight factors corresponding to each sub-model based on the confusion matrix can be specifically implemented as follows: formulas (3) to (5):

[0102] P = TP / (TP + FP); (3)

[0103] R = TP / (TP + FN) (4)

[0104] F1=2PR / (P+R) (5)

[0105] Where P can represent the precision of the model's secondary metrics, R can represent the recall, and F1 can represent the weighting factor.

[0106] According to an embodiment of the present invention, the principal component rotation feature data of the day can be input into each sub-model to obtain the second prediction result output by each sub-model.

[0107] According to an embodiment of the present invention, the largest third prediction result among multiple third prediction results can be used as distribution information.

[0108] According to an embodiment of the present invention, acquiring polar backscattering observation data collected by a rotating sector scatterometer includes:

[0109] Acquire global orbital observation data from the rotating sector scatterometer;

[0110] Polar backscattering observation data were determined from global orbital observation data.

[0111] According to an embodiment of the present invention, global orbital observation data may include all data collected by a rotating sector scatterometer while it is orbiting the Earth.

[0112] According to an embodiment of the present invention, after obtaining global orbital observation data, the global orbital observation data can be filtered to obtain polar backscattering observation data of the polar region from the global orbital observation data.

[0113] According to an embodiment of the present invention, determining polar backscattering observation data from global orbital observation data includes:

[0114] Determine the latitude and longitude information of global orbital observation data;

[0115] Based on latitude and longitude information, determine the first observation data in global orbital observation data that corresponds to the polar latitude and longitude information of the polar region;

[0116] The first observation data is projected onto the polar grid to generate polar backscattering observation data.

[0117] According to embodiments of the present invention, latitude and longitude information can be used to filter global orbital observation data to obtain polar backscattering observation data in polar regions.

[0118] According to an embodiment of the present invention, projecting the first observation data onto a polar grid to generate polar backscattering observation data includes:

[0119] Read the latitude and longitude information of each observation orbit from the first observation data;

[0120] Calculate the kilometers corresponding to latitude and longitude information according to the rules of polar equatorial projection;

[0121] Using the kilometer number as the row and column number, the first observation data is projected onto the polar grid to generate polar backscattering observation data.

[0122] Figure 2 The illustration shows a schematic diagram of a sea ice extraction method based on a rotating sector scatterometer according to an embodiment of the present invention.

[0123] The following combination Figure 2 This illustration demonstrates the specific implementation of the sea ice extraction method based on a rotating sector scatterometer provided in the embodiments of the present invention.

[0124] (1) Polar gridding of observation data from rotating sector satellite scatterometer: Select data from the polar region from global orbital observations and project and interpolate them into the polar grid to generate a backscattered polar grid interpolation dataset, which is used to prepare data for principal component analysis to extract ice and water identification features.

[0125] Step 1: In order to extract sea ice from the data for the current date, obtain the observation data of the China-France Oceanographic Satellite scatterometer for the previous 7 days (including the current day; if observations are missing, obtain data further back, but not exceeding 20 days). There are about 16 observations per day, and each observation is a three-dimensional dataset, stored as 1700 scan lines * 42 wind vector unit grids * 16 observation eyes (8 horizontal polarizations and 8 vertical polarizations), as a 7-day * 16-track backscattering orbit observation dataset;

[0126] Step 2: Based on the backscatter orbit observation dataset, read the latitude and longitude of each orbit. According to the polar equatorial projection rules (443*314 grids in the Northern Hemisphere and 408*302 grids in the Southern Hemisphere), calculate the number of kilometers in the north and south grids for the latitude and longitude. Then convert the number of kilometers into row and column numbers and store the orbit observation dataset according to the row and column numbers to form a dataset with 7 days, 16 observations (8 horizontal and 8 vertical polarization layers), 16 orbits, and N backscatter observations, i.e., backscatter orbit projection dataset (7 days * 16 observations * 16 orbits * N observations);

[0127] Step 3: Based on the backscattering orbit projection dataset, calculate the average value of N backscattering observations for each observation number on a daily basis, and convert the three-dimensional list data into a backscattering grid dataset (7 days * 16 observation numbers * 16 orbits * [Northern Hemisphere 443 * 314, Southern Hemisphere 408 * 302]).

[0128] Step 4: Based on the backscattering grid dataset, using days as the unit, take the first orbital observation for both horizontal and vertical polarization as the reference, i.e., the 2nd to 8th observations reference the 1st observation, and the 9th to 16th observations reference the 8th observation. First, calculate the cumulative histogram of the 1st and 9th observations as the specified histogram. Then, calculate the cumulative histogram of the 2nd to 8th and 10th to 16th observations as the histogram to be matched. Calculate the absolute value of the difference between each gray level of the histogram to be matched and each gray level of the orbital histogram, and its corresponding minimum value. The gray level corresponding to the minimum value is the matched observation value. Generate the backscattering grid histogram matching dataset (7 days * 16 observations * 16 orbits * [Northern Hemisphere 443 * 314, Southern Hemisphere 408 * 302]).

[0129] Step 5: Based on the backscattering grid histogram matching dataset, the 16-track grid data is spliced ​​according to spatial range, using days and observation counts respectively, for both the Northern and Southern Hemispheres. Wavelet transform is performed on overlapping areas to achieve smoothing. First, the data is converted to floating-point type, and wavelet transform is performed using wavelet basis functions to decompose high-frequency and low-frequency components. Low-frequency components below a certain threshold and high-frequency components above a certain threshold are set to 0 to eliminate noise. Then, the processed low-frequency and high-frequency components are subjected to inverse wavelet transform to obtain the smoothed result, i.e., the backscattering grid mosaic dataset (7 days * 16 observation counts * [Northern Hemisphere 443 * 314, Southern Hemisphere 408 * 302]).

[0130] Step 6: Based on the backscattering grid mosaic dataset, perform bilinear spatial interpolation on each observation number in days to fill the gaps between orbits where there are no backscattering observations, and generate a backscattering polar grid interpolation dataset (7 days * 16 observation numbers * [Northern Hemisphere 443 * 314, Southern Hemisphere 408 * 302]).

[0131] (2) Principal component extraction of backscattered polar grid interpolation dataset: In order to solve the problem of classification feature extraction caused by multi-angle observation in the sea ice extraction process of the rotating fan mechanism satellite scatterometer.

[0132] Step 7: Based on the backscattering grid interpolation dataset, calculate the ratio of the first 8 observations (vertical polarization) and the last 8 observations (horizontal polarization) in the North and South Poles on a daily basis. This is the 8 polarization ratio observations. Generate a 7-day backscattering feature image dataset of 8 vertical polarizations, 8 horizontal polarizations, and 8 polarization ratios for both the Northern and Southern Hemispheres (7 days * [8 vertical polarizations, 8 horizontal polarizations, 8 polarization ratios] * [Northern Hemisphere 443 * 314, Southern Hemisphere 408 * 302]).

[0133] Step 8: Based on the backscattering feature image data, convert the 8 horizontal polarizations, 8 vertical polarizations, and 8 polarization ratios of the Northern and Southern Hemispheres into column vectors, on a daily basis, and arrange them into 3 new matrices, namely, vertical polarization VV m*n,8 =[VV1 m*n,1 VV2 m*n,1 ..., VV8 m*n,1 ], Horizontal polarization HH m*n,8 =[HH1 m*n,1 HH2 m*n,1 ..., HH8 m*n,1 With polarization ratio VH m*n,8 =[VH1 m*n,1 VH2 m*n,1 ..., VH8 m*n,1 ];

[0134] Step 9: Perform singular value decomposition on the new matrices VV, HH, and VH using eigenvalue decomposition. Standardize these three matrices, denoted as VVS, HHS, and VHS respectively. Calculate the covariances of VVS, HHS, and VHS, denoted as VVC, HHC, and VHC. Solve for the eigenvalues ​​and eigenvectors of the covariance matrices using eigenvalue decomposition. Arrange the eigenvectors column-wise to form the principal component transformation coefficient matrices VVE, HHE, and VHE. The new feature images corresponding to VV, HH, and VH are VV, HH, and VH respectively. New =VV E T VV S HH New =HH E T HH S VH New =VH E T VH S ;

[0135] Step 10: Based on the new feature image, select the top 2, 2, and 4 principal components of horizontal polarization, vertical polarization, and polarization ratio, respectively, i.e., VV PC1 VV PC2 HH PC1 HH PC2 VH PC1 VH PC2 VH PC3 VH PC4 , as the backscattered principal component images of the northern and southern hemisphere grids;

[0136] Step 11: Based on the 7-day backscattered principal component images of the Northern and Southern Hemispheres and the preset sampling areas of the Northern and Southern Hemispheres, perform spatiotemporal matching with prior sea ice concentration data, and classify the ice-water type according to the range of sea ice concentration: 0-30% sea ice concentration is open water (OW), 30-70% sea ice concentration is open sea ice (OI), and 70-100% sea ice concentration is closed sea ice (CI), denoted as D:

[0137] D = {X, Y} = {VV} PC1 VV PC2 HH PC1 HH PC2 VH PC1 VH PC2 VH PC3 VH PC4 ,Y};

[0138] Where Y represents the sea ice type determined based on sea ice concentration, D is used as the sampling dataset for polar modeling, and based on the polar modeling sampling dataset D, it is divided into training sample datasets D in a 7:3 ratio. Train and test sample dataset D Test :

[0139] D Train ={X Train ,Y Train};

[0140] D Test ={X Test ,Y Test};

[0141] (3) Single model classification: Training a single classification model to prepare for classification using a set of multiple models.

[0142] Step 12: Based on the North and South Pole modeling sampling datasets, train 70% of the data using multivariate logistic regression classification, K-nearest neighbor classification, decision tree, BP neural network, support vector machine, Gaussian Naive Bayes, and random forest respectively to train the sea ice extraction single model and generate single model sets MSigle={MLG,MKNN,MDC,MBP,MSVC,MGB} for the Northern and Southern Hemispheres respectively.

[0143] Step 13: Based on the single model sets for the Northern and Southern Hemispheres, first predict the 30% test sample dataset to generate the test sample prediction result YPred. Combined with YTest, calculate the confusion matrices for OW, OI, and CI, as shown in the table below:

[0144]

[0145] In this context, True Positive = TP represents the number of true values ​​that are positive and the model output is positive; False Negative = FN represents the number of true values ​​that are positive and the model output is negative; False Negative = FN represents the number of true values ​​that are positive and the model output is negative; and True Negative = TN represents the number of true values ​​that are negative and the model output is negative.

[0146] Based on the confusion matrices of OW, OI, and CI, we can calculate the secondary metrics for evaluating model accuracy, namely precision (P), recall (R), and the tertiary metric, F1 score.

[0147] P = TP / (TP + FP);

[0148] R = TP / (TP + FN);

[0149] F1 = 2PR / (P+R);

[0150] Based on the F1 score for each model and each category in both the Northern and Southern Hemispheres, F1 OW ={F1 OW LG F1 OW KNN F1 OW DCT F1 OW BP F1 OW SVC F1 OW GB F1 OW RFC}, F1 OI ={F1 OI LG F1 OI KNN F1 OI DCT F1 OI BP F1 OI SVC F1 OI GB F1 OI RFC}, F1 CI ={F1 CI LG F1 CI KNN F1 CI DCT F1 CI BP F1 CI SVC F1 CI GB F1 CI RFC Then, predictions are made on the backscattered principal component images of the Northern and Southern Hemispheres for the 7th day, generating single-model sea surface type probability prediction images for both hemispheres: Prob OW ={Prob OW LG Prob OW KNN Prob OW DCT Prob OW BP Prob OW SVC ProbOW GB Prob OW RFC}, Prob OI ={Prob OI LG Prob OI KNN Prob OI DCT Prob OI BP Prob OI SVC Prob OI GB Prob OI RFC}, Prob CI ={Prob CI LG Prob CI KNN Prob CI DCT Prob CI BP Prob CI SVC Prob CI GB Prob CI RFC}

[0151] (4) Set classification based on multiple models: The multi-model set method is used for sea ice detection to solve the problem of low accuracy of traditional sea ice extraction using a single model.

[0152] Figure 3 A schematic diagram illustrating the generation of prediction results is shown.

[0153] Step 14: As Figure 3 As shown, based on the F1 scores of the Northern and Southern Hemispheres and the single-model sea surface type probability prediction images of the Northern and Southern Hemispheres, the F1 score is used as a weighting coefficient to fuse the predicted probabilities of the three sea surface types:

[0154] Prob_OW=F1 OW LG *Prob OW LG +F1 OW KNN *Prob OW KNN +F1 OW DCT *

[0155] Prob OWDCT +F1 OW BP *Prob OW BP +F1 OW SVC *Prob OW SVC +F1 OW GB *Prob OW GB +F1 OW RFC *Prob OW RFC 。

[0156] Prob_OI=F1 OI LG *Prob OI LG +F1 OI KNN *Prob OI KNN +F1 OI DCT *Prob OI DCT +

[0157] F1 OI BP *Prob OI BP +F1 OI SVC *Prob OI SVC +F1 OI GB *Prob OI GB +F1 OI RFC *Prob OI RFC 。

[0158] Prob_CI=F1 CI LG *Prob CI LG +F1 CI KNN *Prob CI KNN +F1 CI DCT *Prob CI DCT +

[0159] F1 CI BP*Prob CI BP +F1 CI SVC *Prob CI SVC +F1 CI GB *Prob CI GB +F1 CI RFC *Prob CI RFC .

[0160] Step 15: Based on the predicted probabilities of the three sea surface types, take the value with the highest probability among Porb_OW, Prob_OI, and Prob_CI. The type corresponding to the maximum value is the type of that pixel, and generate the preliminary classification results of Arctic and Antarctic sea ice.

[0161] Step 16: Based on the preliminary classification results of the Arctic and Antarctic sea ice, perform post-classification processing. Use the mode operator to process the patch noise in the image. The mode gradient operator is defined as an 8*8 sliding window. For each pixel area of ​​the image after classification, the classification categories of the 64 pixels in this window are counted. The category with the largest number of categories is taken as the final category of the pixel, and the patch noise correction image of the preliminary classification of Arctic and Antarctic sea ice is generated.

[0162] Step 17: Based on the preliminary classification of polar sea ice patch noise correction images, Arctic: using the effective sea ice mask file for the month of the prediction date (available from the U.S. Snow and Ice Data Center), correct the pixels affected by land spillover and those with almost no sea ice in historical climate periods to extract Arctic sea ice; Antarctica: using the reanalyzed monthly average sea surface temperature data of the past 40 years, correct the pixels with sea surface temperatures above 4°C in historical climate periods to extract Antarctic sea ice.

[0163] Figure 4 The schematic diagram illustrates a block diagram of a sea ice extraction device based on a rotating fan-shaped scatterometer according to an embodiment of the present invention, as shown below. Figure 4 As shown, the sea ice extraction device based on a rotating fan-shaped scatterometer may specifically include:

[0164] The first acquisition module 401 is used to acquire polar backscattering observation data collected by the rotating sector scattering meter. The polar backscattering observation data includes multiple sets of observation data obtained by the rotating sector scattering meter from multiple observations of the polar region based on different azimuth angles and incident angles.

[0165] Analysis module 402 is used to perform principal component analysis on multiple sets of observation data to obtain principal component rotation feature data of backscattering observation data;

[0166] The extraction module 403 is used to extract sea ice features based on principal component rotation feature data to obtain sea ice distribution information in the polar regions.

[0167] Figure 4 The aforementioned sea ice extraction device based on a rotating fan-shaped scatterometer can perform... Figure 1 The implementation principle and technical effects of the sea ice extraction method based on a rotating sector scatterometer described in the illustrated embodiment will not be repeated here. The specific operation methods of each module and unit in the sea ice extraction device based on the rotating sector scatterometer in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0168] In one possible design, the sea ice extraction device based on a rotating fan-shaped scatterometer provided in this embodiment of the invention can be implemented as a computing device, such as... Figure 5 As shown, the computing device may include a storage component 501 and a processing component 502;

[0169] The storage component 501 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 502 to implement the sea ice extraction method based on a rotating fan-shaped scatterometer provided in the embodiments of the present invention.

[0170] Of course, computing devices may also include other components, such as input / output interfaces and communication components. Input / output interfaces provide an interface between processing components and peripheral interface modules, which can be output devices, input devices, etc. Communication components are configured to facilitate wired or wireless communication between the computing device and other devices.

[0171] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0172] When the computing device is a physical device, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device.

[0173] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a computer, can implement the sea ice extraction method based on a rotating sector scatterometer provided in this invention.

[0174] This invention also provides a computer program product, including a computer program that, when executed by a computer, can implement the sea ice extraction method based on a rotating sector scatterometer provided in this invention.

[0175] The processing component in the corresponding embodiments described above may include one or more processors to execute computer instructions to complete all or part of the steps in the method described above. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the method described above.

[0176] Storage components are configured to store various types of data to support operation within the device. Storage components can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0177] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0178] 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.

[0179] 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.

[0180] 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 extracting sea ice based on a rotating sector scatterometer, characterized in that, include: The polar backscattering observation data collected by the rotating sector scatterer includes multiple sets of backscattering observation data obtained by the rotating sector scatterer from multiple observations of the polar region based on different azimuth angles and incident angles. Principal component analysis was performed on the multiple sets of backscattering observation data to obtain the principal component rotation feature data of the backscattering observation data; Sea ice features are extracted based on the principal component rotation feature data to obtain the distribution information of sea ice in the polar region. Each set of backscattering observation data includes vertical polarization data and horizontal polarization data; The principal component analysis of multiple sets of backscattering observation data yields the principal component rotation feature data of the backscattering observation data, including: For each set of backscattering observation data, the polarization ratio data is calculated based on the vertical polarization data and the horizontal polarization data, respectively. A feature matrix is ​​constructed based on multiple sets of vertical polarization data, horizontal polarization data, and polarization ratio data. Principal component analysis is performed on the feature matrix to obtain the principal component rotation feature data; The construction of the feature matrix is ​​specifically achieved by obtaining the vertical polarization data, horizontal polarization data, and polarization ratio data from each group of backscattering observation data. The first characteristic matrix, the second characteristic matrix, and the third characteristic matrix are constructed based on the vertical polarization data, horizontal polarization data, and polarization ratio data in each set of backscattering observation data, respectively. The principal component rotation feature data of the backscattering observation data are as follows: Principal component analysis was performed on the first feature matrix, the second feature matrix, and the third feature matrix respectively to obtain the first feature data, the second feature data, and the third feature data corresponding to the first feature matrix, the second feature matrix, and the third feature matrix, respectively. The first feature data, the second feature data, and the third feature data are concatenated to obtain the principal component rotation feature data.

2. The method according to claim 1, characterized in that, Principal component analysis of the characteristic matrix yields principal component rotation characteristic data of the backscattering observation data, which also includes: The feature matrix is ​​standardized to obtain the standardized matrix; Transform the standardized matrix into a covariance matrix; Calculate the eigenvalues ​​and eigenvectors of the covariance matrix; The feature vectors corresponding to the feature values ​​that meet the preset conditions are used as the target feature vectors. Principal component rotation feature data are obtained based on the target feature vector and the normalized matrix.

3. The method according to claim 1, characterized in that, The principal component rotation feature data includes historical principal component rotation feature data and principal component rotation feature data for the current day; Sea ice feature extraction was performed based on principal component rotation feature data, yielding information on sea ice distribution in the polar regions, including: A sea ice extraction model is trained using the historical principal component rotation feature data, and the sea ice extraction model includes multiple sub-models. The principal component rotation features of the day are input into the sea ice extraction model, and the distribution information is output.

4. The method according to claim 3, characterized in that, The sea ice extraction model is trained using historical principal component rotation feature data, including: Obtain a pre-defined polar sampling area and sea ice distribution reference information corresponding to the sampling area; From the historical principal component rotation feature data, determine the principal component rotation feature data corresponding to the sampling region; Determine the sea ice extraction reference distribution information corresponding to the principal component rotation feature data from the sea ice distribution reference information; A sample dataset is generated based on the principal component rotation feature data and the sea ice extraction reference distribution information. The sea ice extraction model was trained using the sample dataset.

5. The method according to claim 4, characterized in that, The sea ice extraction model is trained using a sample dataset, including: The sample dataset is divided into a training dataset and a test dataset; The multiple sub-models are trained using the training dataset to obtain the sea ice extraction model.

6. The method according to claim 5, characterized in that, The principal component rotation feature data for the day is input into the sea ice extraction model, and the output distribution information includes: The test dataset is input into the multiple sub-models respectively, and the first prediction result corresponding to each of the multiple sub-models is output; Based on the first prediction result and the sea ice extraction reference distribution information, a confusion matrix is ​​obtained; Based on the confusion matrix, the weight factors corresponding to each of the sub-models are obtained respectively; The principal component rotation feature data of the day are respectively input into the multiple sub-models, and the second prediction results corresponding to the multiple sub-models are output respectively; Based on the multiple weighting factors and the second prediction results corresponding to the weighting factors respectively, a third prediction result corresponding to each sub-model is obtained; The distribution information is determined from multiple of the third prediction results.

7. The method according to claim 1, characterized in that, The polar backscattering observation data acquired by the rotating sector scatterometer includes: Acquire global orbital observation data from the rotating sector scatterometer; The polar backscattering observation data is determined from the global orbital observation data.

8. The method according to claim 7, characterized in that, The polar backscattering observation data is determined from the global orbital observation data, including: Determine the latitude and longitude information of the global orbital observation data; Based on the latitude and longitude information, determine the first backscattering observation data in the global orbital observation data that corresponds to the polar latitude and longitude information of the polar region; The first backscattering observation data is projected onto the polar grid to generate the polar backscattering observation data.

9. The method according to claim 8, characterized in that, Projecting the first backscattering observation data onto a polar grid to generate the polar backscattering observation data includes: Read the latitude and longitude information of each observation orbit from the first backscattering observation data; Calculate the number of kilometers corresponding to the latitude and longitude information according to the rules of polar equatorial projection; Using the kilometer number as the row and column number, the first backscattering observation data is projected onto the polar grid to generate the polar backscattering observation data.

Images