Methods and apparatus for inverting the extent and type of Arctic sea ice

Abstract

This application provides a method and apparatus for inverting the extent and type of Arctic sea ice. The method includes: acquiring a first identification model; the first identification model is a model established based on first observation data and the corresponding ice-water classification results; the first observation data is observation data collected by a satellite the previous day; acquiring second observation data; the second observation data is observation data collected by the satellite on the current day; and determining the ice-water classification results corresponding to the second observation data based on the first identification model and the second observation data. This method aims to improve the accuracy, timeliness, and efficiency of sea ice identification.

Description

Technical Field

[0001] This application relates to the field of marine monitoring technology, and more specifically, to a method and apparatus for inverting the extent and type of Arctic sea ice. Background Technology

[0002] In existing technologies, data monitored by ocean satellites can be used to identify sea ice, that is, to identify sea ice and seawater in the ocean.

[0003] Traditional methods for sea ice identification require external data (such as prior data) as input, followed by the development of an identification model using maximum likelihood estimation or Bayesian methods. This model is suitable for identifying sea ice. However, this approach demands high accuracy from the external data, and the input of external data reduces timeliness and efficiency. Furthermore, the identification model does not consider the seasonal variations in the observed data, resulting in inconsistent accuracy across different seasons and months. For example, if the identification model is built based on winter data, its accuracy will be poor when used to identify summer sea ice.

[0004] Therefore, the existing sea ice identification methods are poor in terms of accuracy, timeliness, and efficiency. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for inverting the extent and type of Arctic sea ice, so as to improve the accuracy, timeliness and efficiency of sea ice identification.

[0006] In a first aspect, embodiments of this application provide a method for inverting the extent and type of Arctic sea ice, comprising: obtaining a first identification model; the first identification model is a model established based on first observation data and the ice-water classification result corresponding to the first observation data; the first observation data is observation data collected by a satellite the previous day; obtaining second observation data; the second observation data is observation data collected by the satellite on the current day; and determining the ice-water classification result corresponding to the second observation data according to the first identification model and the second observation data.

[0007] In this embodiment, compared with the prior art, on the one hand, the basis for constructing the recognition model is the daily observation data collected by the satellite, which does not require a large amount of external data input, thus improving the timeliness and efficiency of recognition. On the other hand, for the ice and water classification results of the day, the corresponding recognition model is a model constructed based on the observation data and ice and water classification results of the previous day, which fully considers the impact of seasonal changes on the accuracy of the recognition model. When the season changes, the recognition model will also change accordingly, avoiding the problem that the recognition model cannot accurately identify due to the seasonal changes in the observation data, thereby improving the accuracy of the recognition model.

[0008] As one possible implementation, obtaining the first recognition model includes: acquiring first observation data and the ice-water classification result corresponding to the first observation data; preprocessing the first observation data to obtain preprocessed first observation data; determining the classification label corresponding to the preprocessed first observation data based on the ice-water classification result corresponding to the first observation data; the classification label includes sea ice and seawater; and establishing the first recognition model based on the classification label corresponding to the preprocessed first observation data and the preprocessed first observation data.

[0009] In this embodiment of the application, the first observation data is first preprocessed, then the classification label corresponding to the preprocessed first observation data is determined, and a first recognition model is established based on the preprocessed first observation data and its corresponding classification label, thereby realizing the effective construction of the first recognition model.

[0010] As one possible implementation, determining the ice-water classification result corresponding to the second observation data based on the first recognition model and the second observation data includes: preprocessing the second observation data to obtain preprocessed second observation data; and inputting the preprocessed second observation data into the first recognition model to obtain the ice-water classification result corresponding to the second observation data.

[0011] In this embodiment of the application, since the data used to establish the first identification model is preprocessed observation data, when applying the first identification model, the second observation data is preprocessed before being input into the first identification model to improve the accuracy of the ice-water classification results.

[0012] As one possible implementation, the preprocessing of the second observation data to obtain preprocessed second observation data includes: removing land observation values ​​from the second observation data, and removing invalid observation values ​​according to a preset observation data range to obtain preprocessed second observation data.

[0013] In the embodiments of this application, by removing land observations, the influence of land observations on ice and water classification results can be avoided; and by removing invalid observations, the influence of invalid observations on ice and water classification results can be avoided; thereby improving the accuracy of ice and water classification results.

[0014] As one possible implementation, the preprocessing of the first observation data to obtain preprocessed first observation data includes: removing land observation values ​​from the first observation data, and removing invalid observation values ​​according to a preset observation data range to obtain preprocessed first observation data.

[0015] In this embodiment of the application, by removing land observations, the influence of land observations on the establishment of the first identification model can be avoided; and by removing invalid observations, the influence of invalid observations on the establishment of the first identification model can be avoided; thereby improving the accuracy of the first identification model.

[0016] As one possible implementation, after inputting the preprocessed second observation data into the first recognition model to obtain the ice-water classification result corresponding to the second observation data, the method further includes: post-processing the ice-water classification result using a preset sea ice misjudgment pixel removal algorithm to obtain a post-processed ice-water classification result.

[0017] In this embodiment of the application, after obtaining the initial classification result, the sea ice misclassification pixel removal algorithm is used for post-processing, which can avoid the influence of sea ice misclassification pixels on the ice-water classification result, and the final ice-water classification result has higher accuracy.

[0018] As one possible implementation, after inputting the preprocessed second observation data into the first recognition model to obtain the ice-water classification result corresponding to the second observation data, the method further includes: determining third observation data based on the ice-water classification result corresponding to the second observation data; the third observation data being the observation data in the second observation data whose ice-water classification result is sea ice; obtaining a second recognition model; the second recognition model being a model established based on fourth observation data and the sea ice type classification result corresponding to the fourth observation data; the fourth observation data being the observation data in the first observation data whose ice-water classification result is sea ice; and determining the sea ice type classification result corresponding to the third observation data based on the second recognition model and the third observation data.

[0019] In this embodiment, the identification model for sea ice type is based on daily satellite observation data, eliminating the need for extensive external data input and improving timeliness and efficiency. Furthermore, the identification model for the current day's sea ice classification results is built upon the previous day's observation data and classification results, fully considering the impact of seasonal changes on model accuracy. The model adapts to seasonal changes, preventing inaccurate identification due to seasonal variations in observation data and thus enhancing model accuracy.

[0020] As one possible implementation, obtaining the second identification model includes: determining fourth observation data based on the ice-water classification result corresponding to the first observation data; obtaining the sea ice type classification result corresponding to the fourth observation data; determining the classification label corresponding to the fourth observation data based on the sea ice type classification result corresponding to the fourth observation data; the classification label includes one-year ice and multi-year ice; and establishing the second identification model based on the classification label corresponding to the fourth observation data and the fourth observation data.

[0021] In this embodiment of the application, the observation data of sea ice on the previous day is first determined, and then the classification label corresponding to the observation data of sea ice on the previous day is determined based on the observation data of sea ice on the previous day and its sea ice type classification result. Finally, the second recognition model is effectively established.

[0022] As one possible implementation, the satellite is equipped with a microwave scatterometer and a scanning microwave radiometer; before acquiring the first observation data, the method further includes: determining the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; the first observation data is the observation data corresponding to the sensitive parameters; the second observation data is the observation data corresponding to the sensitive parameters.

[0023] In this embodiment of the application, by determining the sensitive parameters corresponding to the payloads on the satellite, the observation data is the observation data corresponding to the sensitive parameters, so that the observation data can be better used for ice and water classification.

[0024] As one possible implementation, determining the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively includes: determining the inter-class spacing of the observation parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; and determining the sensitive parameters based on the inter-class spacing of the observation parameters and a preset inter-class spacing value.

[0025] In this embodiment of the application, the influence of the observation parameters on the ice-water classification results is evaluated by the inter-class spacing of the observation parameters, thereby achieving effective determination of sensitive parameters.

[0026] Secondly, embodiments of this application provide an Arctic sea ice extent and type inversion device, comprising: various functional modules for implementing the Arctic sea ice extent and type inversion method described in the first aspect and any possible implementation of the first aspect.

[0027] Thirdly, embodiments of this application provide a readable storage medium storing a computer program, which, when executed by a computer, performs the Arctic sea ice extent and type inversion method as described in the first aspect and any possible implementation thereof. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating the method for inverting the extent and type of Arctic sea ice provided in this application embodiment;

[0030] Figure 2 A schematic diagram of the structure of the Arctic sea ice extent and type inversion device provided in the embodiments of this application.

[0031] Icons: 200 - Arctic sea ice extent and type inversion device; 210 - Acquisition module; 220 - Processing module. Detailed Implementation

[0032] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0033] To facilitate understanding of the technical solutions provided in the embodiments of this application, the application scenarios and hardware operating environment will be introduced before introducing the technical solutions.

[0034] The technical solutions of this application can be applied to the identification of ice and water and the identification of sea ice types in the Arctic. Of course, if there are other ocean environments or ocean conditions that are the same or similar to those in the Arctic, the technical solutions of this application can also be used for corresponding identification.

[0035] Ice-water identification can be understood as dividing the Arctic sea ice and seawater regions, where the sea ice region is mostly ice and the seawater region is mostly water. Sea ice type identification can be understood as identifying the type of ice within the sea ice region, including annual ice, multi-year ice, etc. In the technical solution provided in this application embodiment, ice-water identification is performed first. After obtaining the ice-water identification result, sea ice type identification is then performed for the sea ice region; if the ice-water identification result does not include a sea ice region, then sea ice type identification is unnecessary.

[0036] The data basis for the Arctic sea ice extent and type inversion method in this application embodiment is the data monitored daily by satellite. The data processing result is the corresponding ice-water classification result or sea ice type classification result. Therefore, the hardware operating environment of this Arctic sea ice extent and type inversion method can be a satellite. After obtaining the monitoring data, the satellite determines the ice-water classification result or sea ice type classification result and then synchronizes it to the monitoring end, such as a monitoring end in the Arctic. The hardware operating environment of this Arctic sea ice extent and type inversion method can also be a monitoring end. After obtaining the monitoring data, the satellite synchronizes it to the monitoring end, and the monitoring end determines the ice-water classification result or sea ice type classification result based on the monitoring data. Of course, the hardware operating environment is not limited to these two implementation methods. In practical applications, the hardware operating environment of this method can be reasonably changed according to different application scenarios, and is not limited in this application embodiment.

[0037] The satellite involved in this application embodiment may be HY-2B (Haiyang-2B satellite), which is a marine dynamic environment monitoring satellite capable of high-frequency, large-area, and large-scale monitoring of the marine dynamic environment. The HY-2B satellite carries main payloads such as a microwave scatterometer, a radar altimeter, and a scanning microwave radiometer. Data collected by the microwave scatterometer and scanning microwave radiometer payloads can serve as the data basis for this application embodiment.

[0038] Based on the above description of application scenarios and hardware operating environments, please refer to... Figure 1 Here is a flowchart of the Arctic sea ice extent and type inversion method provided in this application embodiment, the method including:

[0039] Step 110: Obtain the first identification model. The first identification model is a model built based on the first observation data and the corresponding ice and water classification results; the first observation data is the observation data collected by the satellite the previous day.

[0040] Step 120: Obtain the second observation data. The second observation data is the observation data collected by the satellite on that day.

[0041] Step 130: Determine the ice-water classification result corresponding to the second observation data based on the first identification model and the second observation data.

[0042] In this embodiment, compared with the prior art, on the one hand, the basis for constructing the recognition model is the daily observation data collected by the satellite, which does not require a large amount of external data input, thus improving the timeliness and efficiency of recognition. On the other hand, for the ice and water classification results of the day, the corresponding recognition model is a model constructed based on the observation data and ice and water classification results of the previous day, which fully considers the impact of seasonal changes on the accuracy of the recognition model. When the season changes, the recognition model will also change accordingly, avoiding the problem that the recognition model cannot accurately identify due to the seasonal changes in the observation data, thereby improving the accuracy of the recognition model.

[0043] In the embodiments of this application, steps 110-130 can be understood as the process of ice water identification. The detailed implementation of each step will be described below.

[0044] In step 110, the first identification model is a model built based on the observation data collected by the satellite the previous day and the corresponding ice and water classification results. This first identification model can be built either after the previous day's observation data collection is completed or when ice and water identification is required on the current day. In other words, the first identification model can be built in real-time or pre-built.

[0045] In this embodiment, all observation data are data after data projection and preprocessing. Mature data projection and preprocessing techniques exist for studying Arctic sea ice. This technique proposes a projection plane grid with minimal deformation within the high-latitude sea ice coverage area. This grid is projected using the polar ellipsoidal projection method (normal axis conformal azimuth projection), with no deformation across 70° north and south latitude, and a deformation rate of 6% at higher latitudes. The projected Arctic plane grid size is 448×304, with a spatial resolution of 25km. Using this projection method, daily satellite-monitored orbital data can be projected and the arithmetic mean can be used to obtain daily Arctic observation data with different parameters. In practical applications, this observation data can be represented in the form of images, with accompanying information including the various observation parameters of the payloads carried on the satellite.

[0046] As an optional implementation, the process of establishing the first identification model includes: acquiring first observation data and the ice-water classification results corresponding to the first observation data; preprocessing the first observation data to obtain preprocessed first observation data; determining the classification label corresponding to the preprocessed first observation data based on the ice-water classification results corresponding to the first observation data; the classification label includes sea ice and seawater; and establishing the first identification model based on the classification label corresponding to the preprocessed first observation data and the preprocessed first observation data.

[0047] In this process, both the first observation data and the corresponding ice-water classification results are known information and can be directly obtained. The first observation data can be multiple sets of observation data, such as observation data from different times of the previous day, or observation data with different observation parameters from the previous day. The ice-water classification results corresponding to the first observation data can be understood as the ice-water classification results corresponding to the observation data from the previous day. Based on the description of the observation data, these ice-water classification results can be directly marked on the image; for example, white areas on the image corresponding to the observation data represent seawater, black areas represent land or invalid observation data, and gray areas represent sea ice.

[0048] If a model is built directly based on the first observation data and its corresponding ice and water classification results, irrelevant data may affect the accuracy of the model. Therefore, the first observation data can be preprocessed. As an optional implementation method, the preprocessing process includes: removing land observation values ​​from the first observation data, and removing invalid observation values ​​according to a preset observation data range to obtain preprocessed first observation data.

[0049] Land observations can be removed using the land masking method; invalid observations can be removed using the observation data range, where all data within the preset observation data range is removed. For example, the backscattering coefficient ranges from -80 to 10, and the brightness temperature is less than 300K.

[0050] After obtaining the first preprocessed observation data, corresponding classification labels are set for the observation data based on the ice-water classification results. The classification labels include two types: sea ice and seawater. If the classification result of the observation data is sea ice, then its corresponding classification label is sea ice; if the classification result of the observation data is seawater, then its corresponding classification label is seawater.

[0051] After determining the classification labels, corresponding labels are assigned to the preprocessed first observation data. Then, machine learning is performed based on the preprocessed first observation data and the assigned labels to establish a first recognition model. The machine learning algorithm can be a support vector machine classification algorithm or other classification algorithms, which are not limited in this embodiment. The method for constructing a classification model using machine learning is a mature technology in this field and will not be described in detail in this embodiment.

[0052] In this embodiment, the first observation data is preprocessed, then the classification label corresponding to the preprocessed first observation data is determined, and a first recognition model is established based on the preprocessed first observation data and its corresponding classification label, thereby achieving effective construction of the first recognition model. By removing land observation values, the influence of land observation values ​​on the ice-water classification results can be avoided; and by removing invalid observation values, the influence of invalid observation values ​​on the ice-water classification results can be avoided, thereby improving the accuracy of the ice-water classification results.

[0053] After obtaining the first identification model in step 110, the second observation data is obtained in step 120. The second observation data is the observation data collected by the satellite on that day, i.e., the observation data to be identified (classified). In practical applications, the observation data collected by the satellite each day can be used as the observation data to be identified for ice and water classification; alternatively, the observation data to be identified can be acquired every cycle, and ice and water classification can be performed based on the observation data within the cycle; or other implementation methods may be used, which are not limited in the embodiments of this application.

[0054] In the embodiments of this application, the identification model for each day's observation data is established based on the observation data of the previous day and the corresponding classification results. There is a special case: the determination of the classification result for the observation data of the first day. This classification result can be determined using existing methods; the determination of the classification results for observation data after the first day can all adopt the technical solution of the embodiments of this application.

[0055] In step 130, the ice-water classification result corresponding to the second observation data is determined based on the first identification model and the second observation data. Since the data used to establish the first identification model is preprocessed observation data, the second observation data can also be preprocessed before being input into the first identification model. Therefore, as an optional implementation, step 130 includes: preprocessing the second observation data to obtain preprocessed second observation data; and inputting the preprocessed second observation data into the first identification model to obtain the ice-water classification result corresponding to the second observation data.

[0056] The preprocessing process for the second observation data is the same as that for the first observation data, and may include: removing land observation values ​​from the second observation data, and removing invalid observation values ​​according to a preset observation data range to obtain the preprocessed second observation data. The specific implementation method of this preprocessing process is the same as that for the first observation data, and will not be repeated here.

[0057] In step 130, the ice and water classification results corresponding to the observation data of the day can be obtained. In the ice and water classification results, in addition to dividing sea ice and seawater, invalid observation areas and land observation areas can also be divided. Specifically, the division can be achieved by using the observation values ​​removed during preprocessing. For example, the observation data corresponding to the land observation values ​​is the land observation area, and the area corresponding to the invalid observation values ​​is the invalid observation area.

[0058] In this embodiment, both the first and second observation data can be filtered data. It is understood that the satellite carries multiple payloads, including but not limited to microwave scatterometers and scanning microwave radiometers. Each payload corresponds to at least one parameter, and multiple payloads correspond to multiple parameters. In this embodiment, the identification capability of multiple parameters can be used to filter the data input to the identification model. The identification capability of a parameter can be understood as the degree of influence of the parameter on the final identification result. The greater the influence, the better the identification capability; the lower the influence, the worse the identification capability. This can be characterized by the inter-class distance.

[0059] Therefore, before step 110, the method further includes: determining the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; the first observation data is the observation data corresponding to the sensitive parameters; and the second observation data is the observation data corresponding to the sensitive parameters.

[0060] As an optional implementation, determining the sensitive parameters includes: determining the inter-class spacing of the observation parameters corresponding to the microwave scatterometer and the scanning microwave radiometer, respectively; and determining the sensitive parameters based on the inter-class spacing of the observation parameters and a preset inter-class spacing value.

[0061] For ice and water identification, the class distance can be defined as: in, These are the mean values ​​of sea ice and seawater observation parameters, respectively; D represents the variance of the observed parameters for sea ice and seawater, respectively. IW This represents the distance between the two types of samples: sea ice and seawater. The larger this value, the better the ice-water identification ability of this observation parameter and the better the sea ice range extraction effect; conversely, the smaller the value, the worse the ice-water identification information extraction ability of this observation parameter.

[0062] The preset inter-class spacing value can be set according to the required ice and water recognition capability, and is not limited in this embodiment.

[0063] In this embodiment of the application, the parameters involved in the evaluation may include, but are not limited to: the H-polarization backscattering coefficient obtained by the microwave scatterometer. V-polarization backscattering coefficient The standard deviation (Δσ) of the backscattering coefficient in the H-polarization mode H), standard deviation of backscattering coefficient in V polarization mode (Δσ) V ) and backscattering coefficient polarization ratio (σ V / σ H Five parameters; brightness temperature and polarization gradient ratio, and spectral gradient ratio, obtained from the scanning microwave radiometer in five bands: 18.7V, 18.7H, 23.8V, 37V, and 37H.

[0064] Among them, the polarization gradient ratio (PR(18.7)) and the spectral gradient ratio (GR(37 / 18.7)) are defined as follows: PR(18.7) = (T b,18.7V -T b,18.7H ) / (T b,18.7V +T b,18.7H ), GR(37 / 18.7)=(T b,37V -T b,18.7H ) / (T b,37V +T b,18.7H ).

[0065] Among them, T b,18.7V The brightness temperature is 18.7V, T b,18.7H The brightness temperature is in the 18.7H band, T b,37V The brightness temperature is 37V.

[0066] In this embodiment, ice-water classification and sea ice type classification can ultimately be performed using observation data from a microwave scatterometer and a scanning microwave radiometer. The microwave scatterometer has HH and VV polarization, operates at a frequency of 13.256 GHz, has a backscattering coefficient measurement accuracy of 0.5 dB, and a measurement range of -40 dB to 20 dB. The scanning microwave radiometer is a microwave radiometer with 9 channels at 5 frequencies, where the 5 frequencies are 6.6, 10.7, 18.7, 23.8, and 37 GHz. Except for 23.8 GHz, which is V polarized, the other frequencies are H and V dual polarized.

[0067] In this embodiment of the application, the final selected sensitive parameter can be: σ V / σ H , Δσ V T b,18.7V T b,18.7H , PR(18.7).

[0068] In this embodiment of the application, the influence of the observation parameters on the ice-water classification results is evaluated by the inter-class spacing of the observation parameters, thereby achieving effective determination of sensitive parameters.

[0069] Once the sensitive parameters are determined, when acquiring observation data, only the observation data corresponding to the sensitive parameters need to be acquired; other observation data unrelated to the sensitive parameters do not need to be acquired.

[0070] In this embodiment of the application, the ice-water classification result output by the first recognition model can be used as the initial classification result, and further processing can be performed based on the initial classification result. Therefore, as an optional implementation, after step 130, the method further includes: post-processing the ice-water classification result using a preset sea ice misclassification pixel removal algorithm to obtain a post-processed ice-water classification result.

[0071] Among them, the sea ice prediction pixel removal algorithm and its corresponding post-processing include, but are not limited to: morphological erosion expansion method to remove sea ice misjudgment pixels caused by noise; and climatological maximum sea ice range mask to remove sea ice misjudgment pixels caused by high wind speed and other phenomena in open sea areas.

[0072] Specifically, due to the roughness of the ocean surface caused by high wind speeds, backscattering signals are enhanced. The microwave signals observed in open seawater are similar to those of sea ice, which can easily lead to confusion between open seawater microwave signals and sea ice, resulting in misclassification of open seawater as sea ice during classification. At the same time, during data projection preprocessing, some grid cells will have no observations, becoming invalid parameter noise that cannot be distinguished as seawater or sea ice.

[0073] To address the two phenomena mentioned above, the morphological erosion expansion method is first used to remove sea ice misclassification pixels caused by noise. A rhombus structure with a radius of 2 pixels is used to erode and dilate the initial ice-water classification result, which can partially eliminate some sea ice misclassification areas with fewer pixels. Then, a climatological maximum sea ice range mask is used to remove larger sea ice misclassification pixels in open sea areas caused by high wind speeds and other phenomena. If the initial sea ice classification result is outside the climatological maximum sea ice range, it is classified as seawater. After these two post-processing steps, the final ice-water classification result for the day is obtained.

[0074] In this embodiment of the application, after obtaining the initial classification result, the sea ice misclassification pixel removal algorithm is used for post-processing, which can avoid the influence of sea ice misclassification pixels on the ice-water classification result, and the final ice-water classification result has higher accuracy.

[0075] Through the above embodiments, relatively accurate ice-water classification results can be obtained. After obtaining the ice-water classification results, sea ice type identification can be performed based on the ice-water classification results. If there is no sea ice range in the ice-water classification results, there is no need to identify the sea ice type. If there is a sea ice range in the ice-water classification results, then sea ice type identification is required.

[0076] As an optional implementation, after step 130, the method further includes: determining third observation data based on the ice-water classification result corresponding to the second observation data; the third observation data being the observation data in the second observation data whose ice-water classification result is sea ice; obtaining a second identification model; the second identification model being a model established based on the fourth observation data and the sea ice type classification result corresponding to the fourth observation data; the fourth observation data being the observation data in the first observation data whose ice-water classification result is sea ice; and determining the sea ice type classification result corresponding to the third observation data based on the second identification model and the third observation data.

[0077] In this process, the observation data that needs to be classified into sea ice types on the same day is first determined based on the ice and water classification results corresponding to the observation data of that day, which is the third observation data. This can be determined based on the post-processed ice and water classification results.

[0078] The second identification model is the established identification model; similar to the first identification model, it can be established in real time or in advance. Its establishment process may include: determining the fourth observation data based on the ice-water classification result corresponding to the first observation data; obtaining the sea ice type classification result corresponding to the fourth observation data; determining the classification label corresponding to the fourth observation data based on the sea ice type classification result; the classification labels include one-year ice and multi-year ice; and establishing the second identification model based on the classification label corresponding to the fourth observation data and the fourth observation data.

[0079] First, the sea ice observation data for the previous day is determined based on the ice-water classification results corresponding to the previous day's observation data; this is the fourth observation data. Since the previous day's sea ice observation data also undergoes sea ice type classification, the sea ice type classification result corresponding to the fourth observation data is known and can be directly obtained. When determining the classification label corresponding to the fourth observation data, if the sea ice type classification result is one-year ice, the corresponding classification label is also one-year ice; if the sea ice type classification result is multi-year ice, the corresponding classification label is also multi-year ice. After determining this, corresponding classification labels are set for the observation data, and then machine learning algorithms are used to perform machine learning to establish the second recognition model. This part can refer to the implementation method of machine learning for the first recognition model and will not be repeated here.

[0080] Based on the second identification model, the third observation data is input into the second identification model, and the second identification model can output the classification result of sea ice type.

[0081] In the process of classifying sea ice types, no data preprocessing or post-processing is required. The sea ice type classification result output by the second identification model can be used as the final classification result.

[0082] In this embodiment, the third and fourth observation data can be observation data corresponding to observation parameters with good sea ice type identification capabilities. Since the third and fourth observation data are data from the first and second observation data, respectively, when determining sensitive parameters in the aforementioned embodiments, in addition to considering ice and water identification capabilities, sea ice type identification capabilities can also be considered.

[0083] That is, in addition to determining the inter-class spacing characterizing the ability to identify ice and water, the inter-class spacing characterizing the ability to identify sea ice types can also be determined. Then, when determining sensitive parameters, the inter-class spacing characterizing the ability to identify ice and water and the inter-class spacing characterizing the ability to identify sea ice types are combined. For example, observation parameters whose inter-class spacing characterizing both the ability to identify ice and water and the ability to identify sea ice types are both greater than the corresponding preset inter-class spacing are determined as sensitive parameters. The method for determining the inter-class spacing characterizing the ability to identify sea ice types can refer to the method for determining the inter-class spacing characterizing the ability to identify ice and water.

[0084] Once the daily ice and water classification results or sea ice type classification results are obtained, they can be stored and queried or used for other applications whenever needed.

[0085] The technical solution provided in this application uses data from the previous day for modeling. This approach not only eliminates reliance on external data and improves operational efficiency, but also considers the temporal variations in microwave radiation scattering characteristics of different land features, thereby enhancing the accuracy of sea ice extent (i.e., ice-water identification) and sea ice type information extraction. In addition to microwave scatterometer data, the algorithm input also incorporates scanning microwave radiometer data, fully utilizing both active and passive microwave remote sensing data. By considering the microwave radiation scattering characteristics of different land features, the accuracy of sea ice information extraction can be further improved.

[0086] Based on the same inventive concept, please refer to Figure 2 This application also provides an Arctic sea ice extent and type inversion device, including: an acquisition module 210 and a processing module 220.

[0087] The acquisition module 210 is used to acquire a first identification model; the first identification model is a model established based on first observation data and the corresponding ice-water classification results of the first observation data; the first observation data is observation data collected by the satellite the previous day; and to acquire second observation data; the second observation data is observation data collected by the satellite on the current day. The processing module 220 is used to determine the ice-water classification results corresponding to the second observation data based on the first identification model and the second observation data.

[0088] In this embodiment, the acquisition module 210 is further configured to acquire the first observation data and the ice-water classification result corresponding to the first observation data; the processing module 220 is further configured to: preprocess the first observation data to obtain preprocessed first observation data; determine the classification label corresponding to the preprocessed first observation data according to the ice-water classification result corresponding to the first observation data; the classification label includes sea ice and seawater; and establish the first recognition model according to the classification label corresponding to the preprocessed first observation data and the preprocessed first observation data.

[0089] In this embodiment of the application, the processing module 220 is further configured to preprocess the second observation data to obtain preprocessed second observation data; and input the preprocessed second observation data into the first recognition model to obtain the ice-water classification result corresponding to the second observation data.

[0090] In this embodiment of the application, the processing module 220 is specifically used to: remove land observation values ​​from the second observation data, and remove invalid observation values ​​according to a preset observation data range to obtain preprocessed second observation data.

[0091] In this embodiment of the application, the processing module 220 is further configured to: remove land observation values ​​from the first observation data, and remove invalid observation values ​​according to a preset observation data range to obtain preprocessed first observation data.

[0092] In this embodiment of the application, the processing module 220 is further configured to: perform post-processing on the ice-water classification result using a preset sea ice misjudgment pixel removal algorithm to obtain the post-processed ice-water classification result.

[0093] In this embodiment, the processing module 220 is further configured to determine the third observation data based on the ice-water classification result corresponding to the second observation data; the third observation data is the observation data in the second observation data whose ice-water classification result is sea ice; the acquisition module 210 is further configured to acquire the second recognition model; the second recognition model is a model established based on the fourth observation data and the sea ice type classification result corresponding to the fourth observation data; the fourth observation data is the observation data in the first observation data whose ice-water classification result is sea ice; the processing module 220 is further configured to determine the sea ice type classification result corresponding to the third observation data based on the second recognition model and the third observation data.

[0094] In this embodiment, the processing module 220 is specifically used to: determine the fourth observation data based on the ice-water classification result corresponding to the first observation data; the acquisition module 210 is further used to acquire the sea ice type classification result corresponding to the fourth observation data; the processing module 220 is further used to: determine the classification label corresponding to the fourth observation data based on the sea ice type classification result corresponding to the fourth observation data; the classification label includes one-year ice and multi-year ice; and establish the second recognition model based on the classification label corresponding to the fourth observation data and the fourth observation data.

[0095] In this embodiment of the application, the processing module 220 is further configured to: determine the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; the first observation data is the observation data corresponding to the sensitive parameters; and the second observation data is the observation data corresponding to the sensitive parameters.

[0096] In this embodiment of the application, the processing module 220 is specifically used to: determine the inter-class spacing of the observation parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; and determine the sensitive parameter based on the inter-class spacing of the observation parameters and a preset inter-class spacing value.

[0097] The Arctic sea ice extent and type inversion device 200 corresponds to the Arctic sea ice extent and type inversion method in the aforementioned embodiments. Therefore, the implementation methods of each functional module refer to the implementation methods of each step of the Arctic sea ice extent and type inversion method in the aforementioned embodiments, and will not be repeated here.

[0098] Based on the same inventive concept, embodiments of this application also provide a readable storage medium storing a computer program, which, when run by a computer, executes the Arctic sea ice extent and type inversion method described in embodiments of this application.

[0099] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0100] Furthermore, 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0101] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0102] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0103] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for inverting the extent and type of Arctic sea ice, characterized in that, include: Obtain the first recognition model; The first identification model is a model built based on the first observation data and the ice-water classification results corresponding to the first observation data; The first observation data is the observation data collected by the satellite the previous day; Acquire second observation data; the second observation data is the observation data collected by the satellite on that day; Based on the first identification model and the second observation data, determine the ice-water classification result corresponding to the second observation data; The ice-water classification results are used to divide the sea ice extent and the seawater extent; The satellite is equipped with a microwave scatterometer and a scanning microwave radiometer. Before acquiring the first observation data, the method further includes: The sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer are determined respectively; the first observation data is the observation data corresponding to the sensitive parameters; the second observation data is the observation data corresponding to the sensitive parameters; the sensitive parameters corresponding to the microwave scatterometer include: backscattering coefficient in H polarization mode, backscattering coefficient polarization ratio, and backscattering coefficient standard deviation in V polarization mode; the sensitive parameters corresponding to the scanning microwave radiometer include: brightness temperature in the 18.7V band, brightness temperature in the 18.7H band, and polarization gradient ratio. The method further includes: The third observation data is determined based on the ice-water classification result corresponding to the second observation data; the third observation data is the observation data in the second observation data whose ice-water classification result is sea ice; Obtain a second identification model; the second identification model is a model established based on the fourth observation data and the sea ice type classification results corresponding to the fourth observation data; the fourth observation data is the observation data in the first observation data whose ice-water classification result is sea ice; The sea ice type classification result corresponding to the third observation data is determined based on the second identification model and the third observation data.

2. The method according to claim 1, characterized in that, The acquisition of the first recognition model includes: Obtain the first observation data and the corresponding ice-water classification results; The first observation data is preprocessed to obtain preprocessed first observation data; The classification label corresponding to the preprocessed first observation data is determined based on the ice-water classification result corresponding to the first observation data; the classification label includes sea ice and seawater. The first recognition model is established based on the classification label corresponding to the preprocessed first observation data and the preprocessed first observation data.

3. The method according to claim 2, characterized in that, The step of determining the ice-water classification result corresponding to the second observation data based on the first recognition model and the second observation data includes: The second observation data is preprocessed to obtain preprocessed second observation data; The preprocessed second observation data is input into the first recognition model to obtain the ice-water classification result corresponding to the second observation data.

4. The method according to claim 2, characterized in that, The step of preprocessing the first observation data to obtain preprocessed first observation data includes: Remove land observation values ​​from the first observation data and remove invalid observation values ​​according to a preset observation data range to obtain preprocessed first observation data.

5. The method according to claim 3, characterized in that, After inputting the preprocessed second observation data into the first recognition model to obtain the ice-water classification result corresponding to the second observation data, the method further includes: The ice-water classification results are post-processed using a preset sea ice misclassification pixel removal algorithm to obtain post-processed ice-water classification results.

6. The method according to claim 1, characterized in that, The acquisition of the second recognition model includes: The fourth observation data is determined based on the ice-water classification results corresponding to the first observation data. Obtain the sea ice type classification result corresponding to the fourth observation data; The classification label corresponding to the fourth observation data is determined based on the sea ice type classification result; the classification label includes one-year ice and multi-year ice. The second identification model is established based on the classification label corresponding to the fourth observation data and the fourth observation data.

7. The method according to claim 1, characterized in that, The determination of the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer includes: Determine the interclass spacing of the observation parameters corresponding to the microwave scatterometer and the scanning microwave radiometer, respectively; The sensitive parameters are determined based on the inter-class spacing of the observed parameters and the preset inter-class spacing value.

8. A device for inverting the extent and type of Arctic sea ice, characterized in that, include: The acquisition module is used to acquire the first recognition model; The first identification model is a model built based on the first observation data and the ice and water classification results corresponding to the first observation data; the first observation data is the observation data collected by the satellite the previous day; The acquisition module is also used to acquire second observation data; the second observation data is the observation data collected by the satellite on that day; The processing module is used to determine the ice-water classification result corresponding to the second observation data based on the first identification model and the second observation data; the ice-water classification result is used to divide the sea ice range and the seawater range. The satellite is equipped with a microwave scatterometer and a scanning microwave radiometer; the processing module is further configured to: determine the sensitive parameters corresponding to the microwave scatterometer and the scanning microwave radiometer respectively; the first observation data is the observation data corresponding to the sensitive parameters; The second observation data is the observation data corresponding to the sensitive parameters; the sensitive parameters corresponding to the microwave scatterometer include: backscattering coefficient in H polarization mode, backscattering coefficient polarization ratio, and backscattering coefficient standard deviation in V polarization mode; the sensitive parameters corresponding to the scanning microwave radiometer include: brightness temperature in the 18.7V band, brightness temperature in the 18.7H band, and polarization gradient ratio. The processing module is also used for: The third observation data is determined based on the ice-water classification result corresponding to the second observation data; the third observation data is the observation data in the second observation data whose ice-water classification result is sea ice; The acquisition module is further configured to: acquire a second identification model; the second identification model is a model established based on the fourth observation data and the sea ice type classification result corresponding to the fourth observation data; the fourth observation data is the observation data in the first observation data whose ice-water classification result is sea ice; The processing module is further configured to: determine the sea ice type classification result corresponding to the third observation data based on the second identification model and the third observation data.

Images