Sea ice classification and density inversion method based on deep learning

Abstract

The invention discloses a sea ice classification and intensity inversion method based on deep learning, relates to the technical field of remote sensing image processing, and aims to solve the problems in the prior art, and the method comprises the steps: obtaining original dual-polarization synthetic aperture radar data, and carrying out the preprocessing of the original dual-polarization synthetic aperture radar data; performing superpixel segmentation on the preprocessed data to obtain a plurality of superpixels; calculating the posterior probability of the superpixel by using a conditional random field model, and determining an uncertain superpixel unit; taking the uncertain super-pixel unit as input, outputting sea ice and seawater boundary lines in the uncertain super-pixel unit based on an Ice-WaterNet network model, and combining to obtain a sea ice classification result graph; and performing sea ice concentration inversion on the sea ice classification result map to obtain a sea ice concentration inversion result. According to the method, the inversion precision of sea ice classification and density is improved, and the method can play a more important role in the aspects of global climate change monitoring, marine environment protection, polar region resource development and utilization and the like.

Description

Technical Field

[0001] The present invention relates to a sea ice classification and density inversion method based on deep learning, and belongs to the technical field of remote sensing image processing. Background Art

[0002] Sea ice density is not only a key parameter for marine environmental monitoring and climate research, but also an important reference data for ensuring the safety of marine activities and promoting the development and utilization of polar resources. With the significant changes in global climate change and sea ice coverage, sea ice has had an important impact on the global climate system, water balance, and heat transfer, and has become an important factor in monitoring global climate change. At the same time, sea ice also poses a severe challenge to marine activities such as ship navigation, seabed mining, and polar ocean exploration, and may even cause catastrophic events. Therefore, accurate and efficient sea ice classification is particularly important.

[0003] Synthetic aperture radar (SAR) technology has unique advantages in sea ice monitoring due to its all-day and all-weather observation capabilities, strong penetration, and no cloud cover. The increasing amount of SAR image data and the increasing temporal and spatial resolution have made SAR the main data source for sea ice classification.

[0004] Sea ice classification research based on SAR can not only help scientists understand the physical characteristics and distribution patterns of sea ice and its interaction with climate change, but also provide a scientific basis for marine environmental forecasts and sea ice disaster warnings. Specifically: On the one hand, sea ice classification helps reveal the intrinsic connection between sea ice and the climate system. Through detailed sea ice classification, key information such as sea ice type, range, thickness, and age can be obtained, and then the impact of sea ice changes on the global climate can be analyzed, such as affecting ocean heat exchange and regulating atmospheric temperature. This information is of great significance for predicting global climate change trends and formulating response strategies.

[0005] On the other hand, sea ice classification information is an important reference for ensuring the safety of ship navigation. By real-time monitoring of the distribution and type of sea ice, accurate route planning and ice avoidance guidance can be provided for ships, reducing the risk of accidents such as ship stranding and collision caused by sea ice. At the same time, sea ice classification can also provide strong support for exploration and resource development activities in polar waters, ensuring the smooth progress of activities and the safety of personnel.

[0006] In addition, sea ice classification helps to assess the impact of sea ice changes on marine ecosystems. Sea ice is an important part of polar ecosystems, and its changes directly affect the habitat and food chain structure of polar organisms. Using sea ice classification information, we can monitor the growth and decline of sea ice, assess its impact on marine ecosystems, and provide a basis for decision-making in formulating protection measures and promoting sustainable development.

[0007] With the rapid development of artificial intelligence and deep learning technologies, the accuracy and efficiency of sea ice classification have been significantly improved. Traditional sea ice classification methods mainly rely on manual visual interpretation and remote sensing image processing technology, which have problems such as strong subjectivity and low efficiency. The SAR sea ice classification method based on deep learning can automatically extract sea ice features from SAR images and achieve efficient and accurate classification. This not only greatly improves the efficiency and accuracy of sea ice classification, but also provides the possibility for intelligent and automated sea ice monitoring and research.

[0008] However, affected by the rising temperature during the melting period, the sea ice begins to melt, and its form gradually changes from solid to liquid or semi-solid. This process is accompanied by a decrease in the thickness of the ice layer, the formation of meltwater on the ice surface, and the destruction of the ice structure. These changes make the sea ice diverse and complex in appearance and internal structure, making the sea ice characteristics observed by remote sensing complex and difficult to describe with unified characteristics and models. In addition, the dynamic changes of sea ice are also a major difficulty in classification. During the melting process, sea ice may be affected by various external forces such as wind, waves, and currents, causing drift, fragmentation, and accumulation. These dynamic changes not only change the distribution and form of sea ice, but also increase the difficulty of sea ice classification to a certain extent.

[0009] In summary, the SAR sea ice classification during the melt period is affected by many factors, including the complexity of the physical changes of sea ice, the interference of environmental factors, and the complexity of data processing and analysis. Summary of the invention

[0010] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a sea ice classification and density inversion method based on deep learning, so as to improve the inversion accuracy of sea ice classification and density, and play a more important role in global climate change monitoring, marine environmental protection, polar resource development and utilization, etc.

[0011] To achieve the above object, the present invention is implemented by adopting the following technical solutions: The present invention provides a sea ice classification and density inversion method based on deep learning, comprising: Get raw dual-polarization synthetic aperture radar data; Preprocessing the original dual-polarization synthetic aperture radar data to obtain preprocessed dual-polarization synthetic aperture radar data; Performing superpixel segmentation on the preprocessed dual-polarization synthetic aperture radar data to obtain multiple superpixels; The conditional random field model is used to calculate the posterior probability of superpixels, and the superpixels whose posterior probability is within the set interval are taken as uncertain superpixel units; Taking the uncertain superpixel unit as input, the sea ice and seawater boundary lines within the uncertain superpixel unit are obtained based on the output of the pre-trained Ice-WaterNet network model. The sea ice and seawater boundaries in all uncertain superpixel units are combined to obtain the sea ice classification result map; The sea ice density is inverted according to the sea ice classification result map to obtain the sea ice density inversion result; Among them, the Ice-WaterNet network model is constructed based on the U-Net network model. The construction method includes connecting each convolutional layer in the decoder of the U-Net network model to an uncertainty measurement module, and the uncertainty measurement module is used to divide the uncertain superpixel unit into an uncertain area and a certain area, and calculate the uncertainty of the features in the uncertain area to give high weights to features with large uncertainty.

[0012] Further, the preprocessing of the original dual-polarization synthetic aperture radar data to obtain preprocessed dual-polarization synthetic aperture radar data includes: Read noise parameters and radiation calibration parameters from raw dual-polarization synthetic aperture radar data; The remote sensing image pixel brightness value of the original dual-polarization synthetic aperture radar data is converted into the backscattering coefficient according to the noise parameter and the radiation calibration parameter. The expression is as follows: ; in, represents the backscattering coefficient, Represents the brightness value of the remote sensing image pixel. represents the noise parameter, represents the radiation calibration parameter; The backscatter coefficient is converted into the backscatter coefficient of the reference incident angle, which is the preprocessed dual-polarization synthetic aperture radar data. Its expression is as follows: ; in, represents the corrected backscatter coefficient, i.e., the preprocessed dual-polarization synthetic aperture radar data, represents the backscattering coefficient, represents the reference incident angle, represents the slope coefficient, represents the local angle of incidence.

[0013] Furthermore, the superpixel segmentation of the preprocessed dual-polarization synthetic aperture radar data adopts a meanshift segmentation method.

[0014] Furthermore, the expression of the conditional random field model is as follows: ; in, represents the posterior probability, represents the normalization coefficient, Indicates The unit potential energy function of superpixels is Indicates superpixels and The potential energy function of superpixels; Indicates The class of superpixels, Indicates a class of superpixels, the class being either sea ice or sea water; and All represent superpixel sets and do not contain each other; Indicates The backscatter coefficient of the superpixel, represents the model weight parameter, represents the model input features, , Indicates superpixels and The cross-correlation function of superpixels is represents the Dilick function, when hour, The value is 1, otherwise it is 0.

[0015] Furthermore, the Ice-WaterNet network model is an encoder-decoder structure, wherein the number of convolutional layers in the encoder and decoder are both 5, the encoder is used to calculate channel and spatial features, the decoder is used to extract different types of floating ice using multi-scale context information, and each convolutional layer in the decoder is provided with an attention module, which is used to distinguish sea ice from sea water using the difference between HH and HV channels to determine the buffer area.

[0016] Furthermore, the data processing process of the uncertainty measurement module includes: For each uncertain superpixel unit, the buffer area corresponding to the maximum output probability during model deconvolution is recorded as the uncertain area; Using a rectangular model as a filter window, continuously changing the perceptual field of view of the filter window, obtaining the sea ice sea boundary line in the uncertain area to redivide the uncertain area into a certain area and an uncertain area, until the size of the filter window reaches or exceeds the size of the minimum circumscribed rectangle of the uncertain superpixel unit, wherein the aspect ratio of the rectangular model is the aspect ratio of the minimum circumscribed rectangle of the uncertain superpixel unit; Calculate the uncertainty of the features in the uncertainty region of the uncertain superpixel unit, and obtain the uncertainty values ​​of the features in the uncertainty region of all uncertain superpixel units; The features in the uncertainty region of all uncertain superpixel units are sorted according to their uncertainty values, and features with large uncertainty values ​​are given high weights.

[0017] Furthermore, the uncertainty measurement module calculates the uncertainty of the feature in the uncertainty region in the uncertain superpixel unit using the following expression: ; in, represents the sea ice forecast uncertainty, represents the uncertainty of seawater prediction, represents the uncertainty characteristics of sea ice, represents the uncertainty characteristics of seawater, Indicates the offset.

[0018] Furthermore, the pre-training method of the Ice-WaterNet network model includes: Obtaining an ice condition map data set, wherein the ice condition map data set is annotated with a reference true value; The Ice-WaterNet network model is trained with the ice map dataset as input and the sea ice and sea water boundary lines as output. During the training process, the loss function is used to adjust the model parameters of the Ice-WaterNet network model. When the loss function value reaches the convergence condition, the pre-trained Ice-WaterNet network model is obtained.

[0019] Furthermore, the loss function adopts a binary cross entropy function, which is expressed as: ; in, represents the binary cross entropy function, Represents the first The output of the convolutional layer, Indicates the number of convolutional layers of the Ice-WaterNet network model, Represents the true value of the benchmark.

[0020] Furthermore, the sea ice density inversion is performed according to the sea ice classification result map to obtain the sea ice density inversion result, including: Downsampling the sea ice classification result according to a preset resolution to obtain a downsampled sea ice classification result; The downsampled sea ice classification results are projected according to the projection grid to obtain the percentage of sea ice pixels in each sea ice classification result projection grid, that is, the sea ice density of each sea ice classification result projection grid; The sea ice density of each sea ice classification result projection grid is combined to obtain the sea ice density inversion result.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention uses a conditional random field model to extract uncertain sea ice objects at the superpixel level, and then constructs an Ice-WaterNet deep learning model and uses an iterative strategy to optimize the segmentation edges of superpixels, thereby improving the inversion accuracy of sea ice classification and density. This can play a more important role in global climate change monitoring, marine environmental protection, and polar resource development and utilization. The present invention can effectively improve the inversion accuracy of sea ice classification and density during the melting period through uncertainty superpixel extraction and uncertainty measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of a flow chart of a method for sea ice classification and density inversion based on deep learning in an embodiment of the present invention; Figure 2 A schematic diagram of a processing flow of an Ice-WaterNet model of a sea ice classification and density inversion method based on deep learning in an embodiment of the present invention; Figure 3 A schematic diagram of comparison of the intersection-over-union ratio of CRF, U-net, and Ice-WaterNet of a sea ice classification and density inversion method based on deep learning in an embodiment of the present invention; Figure 4 It is a schematic diagram of comparing the accuracy of CRF, U-net, and Ice-WaterNet of a sea ice classification and density inversion method based on deep learning in an embodiment of the present invention; Figure 5 It is a schematic diagram of the comparison of the recall rates of CRF, U-net, and Ice-WaterNet of the sea ice classification and density inversion method based on deep learning in an embodiment of the present invention; Figure 6 A schematic diagram of the comparison of F1 scores of CRF, U-net, and Ice-WaterNet of a sea ice classification and density inversion method based on deep learning in an embodiment of the present invention; Figure 7 A schematic diagram of sea ice density inversion results of Ice-WaterNet based on a deep learning-based sea ice classification and density inversion method in an embodiment of the present invention; Figure 8Schematic diagram of ASI's sea ice density inversion results. DETAILED DESCRIPTION

[0023] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0024] like Figure 1 As shown, an embodiment of the present invention provides a sea ice classification and density inversion method based on deep learning, comprising the following steps: The original dual-polarization synthetic aperture radar data is obtained. In this embodiment, the original dual-polarization synthetic aperture radar data is original dual-polarization Setninel-1 SAR data.

[0025] Preprocess the original dual-polarization synthetic aperture radar data, including radiation calibration and incident angle correction. Specifically: Radiation calibration: read the noise parameters and radiation calibration parameters from the annotation document of the original dual-polarization Setninel-1 SAR data, and convert the remote sensing image pixel brightness value of the original dual-polarization Setninel-1 SAR data into the backscattering coefficient. The expression is as follows: ; in, represents the backscattering coefficient, Represents the brightness value of the remote sensing image pixel. represents the noise parameter, Represents the radiation calibration parameters.

[0026] Incident angle correction: The backscatter coefficient is converted into the backscatter coefficient of the reference incident angle, which is the preprocessed dual-polarization synthetic aperture radar data. The expression is as follows: ; in, represents the corrected backscatter coefficient, i.e., the preprocessed dual-polarization synthetic aperture radar data, represents the backscattering coefficient, represents the reference incident angle, represents the slope coefficient, represents the local angle of incidence.

[0027] In this embodiment, The value is 2. The value is 26°.

[0028] The pre-processed dual-polarization synthetic aperture radar data (i.e., the corrected backscatter coefficient) is segmented into superpixels. Specifically: Firstly, the preprocessed dual-polarization synthetic aperture radar data are divided into multiple superpixels using the meanshift segmentation method.

[0029] Then, the superpixel-level conditional random field (CRF) model is used to calculate the posterior probability of the superpixel. The expression of the conditional random field model is as follows: ; in, represents the posterior probability, represents the normalization coefficient, Indicates The unit potential energy function of superpixels is Indicates superpixels and The potential energy function of superpixels; Indicates The class of superpixels, Indicates a class of superpixels, the class being either sea ice or sea water; and All represent superpixel sets and do not contain each other; Indicates The backscatter coefficient of the superpixel, represents the model weight parameter, represents the model input features, , Indicates superpixels and The cross-correlation function of superpixels is represents the Dilick function, when hour, The value is 1, otherwise it is 0.

[0030] The superpixel segmentation process uses the minimum energy function method and outputs the final posterior probability , the superpixels whose posterior probabilities are within the set interval are marked as uncertain superpixel units, which effectively reduces the amount of calculation while ensuring the accuracy of sea ice water classification. In this embodiment, the set interval is 0.15~0.85.

[0031] Build the Ice-WaterNet network model: The Ice-WaterNet network model is built based on the U-Net network model. The Ice-WaterNet network model is an encoder-decoder structure, in which the number of convolutional layers in the encoder and decoder are 5. The encoder is used to calculate channel and spatial features, and the decoder is used to use multi-scale context information to extract different types of floating ice. Each convolutional layer in the decoder is also equipped with an attention module, which describes long-distance global information in an explicit way. It can use the difference between HH and HV channels to distinguish sea ice and seawater to determine the buffer area, and solves the problem of ice and water classification ambiguity caused by frost flowers on open waters and ice channels caused by wind and waves, thereby improving the accuracy of ice and water classification.

[0032] The improvement of the Ice-WaterNet network model over the U-Net network model is that each convolutional layer in the decoder of the U-Net network model is connected to an uncertainty measurement module, which is used to divide the uncertain superpixel units into uncertain areas and certain areas, and calculate the uncertainty values ​​of the features in the uncertain areas to give high weights to features with large uncertainties.

[0033] The channel and spatial features of any superpixel unit in the encoder are calculated and then fused in the encoder through the cross attention mechanism. Through each iteration, the convolutional features of each convolutional layer are updated, and the uncertainty measurement module divides the uncertain superpixel unit into uncertain areas and certain areas, and calculates the uncertainty of the features in the uncertain area, giving greater weight to the features with greater uncertainty.

[0034] The data processing process of the uncertainty measurement module includes: For any superpixel unit, the edge of the superpixel in the initial state is recorded as T0, and the edge at the Kth corrosion is recorded as Tk. At this time, Tk and the edge in the T0 state are used as buffers. The rectangular model (the aspect ratio is the aspect ratio of the minimum bounding rectangle of this superpixel) is used in the corrosion process. Through the spatial attention mechanism of U-Net, the buffer corresponding to the maximum output probability during U-Net deconvolution is marked as the uncertainty area of ​​U-Net. The termination condition of the loop is: the filter window exceeds or reaches the size of the minimum bounding rectangle of the superpixel. That is, the U-Net here obtains the potential boundary of sea ice and seawater in the superpixel by changing the perception field of the window, thereby dividing the uncertain superpixel into a deterministic area and an uncertain area.

[0035] The uncertainty of the features in the uncertainty region of the uncertain superpixel unit is calculated, and the uncertainty values ​​of the features in the uncertainty region of all uncertain superpixel units are obtained.

[0036] The features in the uncertainty region of all uncertain superpixel units are sorted according to their uncertainty values, and features with large uncertainty values ​​are given high weights.

[0037] The uncertainty measurement module calculates the uncertainty of the features in the uncertainty region of the uncertain superpixel unit as follows: ; in, represents the sea ice forecast uncertainty, represents the uncertainty of seawater prediction, represents the uncertainty characteristics of sea ice, represents the uncertainty characteristics of seawater, Indicates the offset.

[0038] In this embodiment, the offset is 0.5. The uncertainty measurement module reduces the uncertainty of superpixels of different scales in the layer-by-layer processing of the convolution layer and prevents the loss of features, so that the sea ice classification result obtained by the final output is more accurate.

[0039] Pre-train the constructed Ice-WaterNet network model: Obtain an ice condition map dataset, in which the benchmark true values ​​are annotated.

[0040] The ice map dataset is used as input and the sea ice and sea water boundary lines are used as output to train the Ice-WaterNet network model. During the training process, the loss function is used to adjust the model parameters of the Ice-WaterNet network model. The loss function uses a binary cross entropy function, which is expressed as: ; in, represents the binary cross entropy function, Represents the first The output of the convolutional layer, Indicates the number of convolutional layers of the Ice-WaterNet network model, Represents the true value of the benchmark.

[0041] When the loss function value reaches the convergence condition, the pre-trained Ice-WaterNet network model is obtained.

[0042] The uncertain superpixel unit is taken as input, and the sea ice and sea water boundary lines within the uncertain superpixel unit are obtained based on the output of the pre-trained Ice-WaterNet network model. The sea ice and sea water boundary lines within all uncertain superpixel units are combined to obtain the sea ice classification result map.

[0043] Sea ice concentration inversion: The sea ice classification result is downsampled according to a preset resolution to obtain a downsampled sea ice classification result.

[0044] The downsampled sea ice classification results are projected according to the projection grid to obtain the percentage of sea ice pixels in each sea ice classification result projection grid, that is, the sea ice density of each sea ice classification result projection grid. The sea ice density of each sea ice classification result projection grid is combined to obtain the sea ice density inversion result.

[0045] like Figure 2 As shown in the figure, taking the Sentinel-1 SAR data of the Fram Strait region in the Arctic on August 1, 2023 as an example, for the SAR data that has been radiometrically calibrated and ensemble corrected, the initial segmentation result is obtained through CRF superpixel segmentation, and the superpixels with a posterior probability between 0.15 and 0.85 are marked as uncertain superpixels. For any uncertain superpixel unit, the uncertain superpixel is segmented based on the multi-layer convolution and iterative strategy of Ice-WaterNet. In the loop process, it can be found that as the filter window increases, more sea ice objects are observed, so we only need to focus on Figure 2 The sea ice object in area 1 is sufficient. After the sea ice object in area 1 is accurately extracted, the remaining pixels in area 1 and all pixels in area 2 will be marked as open waters. The segmentation edge corresponding to the minimum uncertainty is taken as the final uncertain superpixel segmentation result.

[0046] Figure 3~Figure 6 This is the accuracy verification result of the ice-water classification result based on Ice-WaterNet of the present invention. Among the four evaluation indicators, the Ice-WaterNet model of the present invention performs best, indicating that the algorithm of the present invention has the best performance in the segmentation accuracy of the ice-water boundary and the suppression of the misclassification of seawater targets; on the other hand, Ice-WaterNet has a smaller fluctuation range among the four evaluation indicators, indicating that the model has better stability. Specifically, the intersection-union ratio reflects the similarity between the classification result and the true value, among which the CRF and U-Net results are similar, while the classification result of Ice-WaterNet is closer to the true value. Although the accuracy of U-Net is higher than that of CRF, the recall rate is lower than that of the CRF result, indicating that there are more false alarms in the U-Net model, that is, the uncertainty of sea ice is greater. The CRF result in the F1 score is also better than the U-Net result, indicating that CRF can better achieve a balance between accuracy and recall. In addition, Ice-WaterNet performs best in the F1 score, which shows that the method of the present invention performs best in ice-water classification.

[0047] Figure 7 and Figure 8This is the sea ice density inversion result based on Ice-WaterNet and ASI density products. Overall, the two density results have good consistency, and the Ice-WaterNet result of the present invention has a higher resolution and can show more sea ice details. In addition, in low-density areas such as the edge of the sea ice and polynyas, the texture features of the method of the present invention are richer.

[0048] In summary, the deep learning-based sea ice classification and density inversion method provided by the present invention improves the inversion accuracy of sea ice classification and density, and can play a more important role in global climate change monitoring, marine environmental protection, polar resource development and utilization, etc.

[0049] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A sea ice classification and density inversion method based on deep learning, characterized in that: include: Get raw dual-polarization synthetic aperture radar data; Preprocessing the original dual-polarization synthetic aperture radar data to obtain preprocessed dual-polarization synthetic aperture radar data; Performing superpixel segmentation on the preprocessed dual-polarization synthetic aperture radar data to obtain multiple superpixels; The conditional random field model is used to calculate the posterior probability of superpixels, and the superpixels whose posterior probability is within the set interval are taken as uncertain superpixel units; Taking the uncertain superpixel unit as input, the sea ice and seawater boundary lines within the uncertain superpixel unit are obtained based on the output of the pre-trained Ice-WaterNet network model. The sea ice and seawater boundaries in all uncertain superpixel units are combined to obtain the sea ice classification result map; The sea ice density is inverted according to the sea ice classification result map to obtain the sea ice density inversion result; Among them, the Ice-WaterNet network model is constructed based on the U-Net network model. The construction method includes connecting each convolutional layer in the decoder of the U-Net network model to an uncertainty measurement module, and the uncertainty measurement module is used to divide the uncertain superpixel unit into an uncertain area and a certain area, and calculate the uncertainty of the features in the uncertain area to give high weights to features with large uncertainty.

2. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The preprocessing of the original dual-polarization synthetic aperture radar data to obtain preprocessed dual-polarization synthetic aperture radar data includes: Read noise parameters and radiation calibration parameters from raw dual-polarization synthetic aperture radar data; The remote sensing image pixel brightness value of the original dual-polarization synthetic aperture radar data is converted into the backscattering coefficient according to the noise parameter and the radiation calibration parameter. The expression is as follows: ； in, represents the backscattering coefficient, Represents the brightness value of the remote sensing image pixel. represents the noise parameter, represents the radiation calibration parameters; The backscatter coefficient is converted into the backscatter coefficient of the reference incident angle, which is the preprocessed dual-polarization synthetic aperture radar data. Its expression is as follows: ； in, represents the corrected backscatter coefficient, i.e., the preprocessed dual-polarization synthetic aperture radar data, represents the backscattering coefficient, represents the reference incident angle, represents the slope coefficient, represents the local angle of incidence.

3. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The superpixel segmentation of the preprocessed dual-polarization synthetic aperture radar data adopts the meanshift segmentation method.

4. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The expression of the conditional random field model is as follows: ； in, represents the posterior probability, represents the normalization coefficient, Indicates The unit potential energy function of superpixels is Indicates superpixels and The potential energy function of superpixels; Indicates The class of superpixels, Indicates a class of superpixels, the class being either sea ice or sea water; and All represent superpixel sets and do not contain each other; Indicates The backscatter coefficient of the superpixel, represents the model weight parameter, represents the model input features, , Indicates superpixels and The cross-correlation function of superpixels is represents the Dilick function, when hour, The value is 1, otherwise it is 0.

5. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The Ice-WaterNet network model is an encoder-decoder structure, wherein the number of convolutional layers in the encoder and decoder are both 5, the encoder is used to calculate channel and spatial features, the decoder is used to extract different types of floating ice using multi-scale context information, and each convolutional layer in the decoder is provided with an attention module, which is used to distinguish sea ice from sea water by using the difference between HH and HV channels to determine the buffer area.

6. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The data processing process of the uncertainty measurement module includes: For each uncertain superpixel unit, the buffer area corresponding to the maximum output probability during model deconvolution is recorded as the uncertain area; Using a rectangular model as a filter window, continuously changing the perceptual field of view of the filter window, obtaining the sea ice sea boundary line in the uncertain area to redivide the uncertain area into a certain area and an uncertain area, until the size of the filter window reaches or exceeds the size of the minimum circumscribed rectangle of the uncertain superpixel unit, wherein the aspect ratio of the rectangular model is the aspect ratio of the minimum circumscribed rectangle of the uncertain superpixel unit; Calculate the uncertainty of the features in the uncertainty region of the uncertain superpixel unit, and obtain the uncertainty values ​​of the features in the uncertainty region of all uncertain superpixel units; The features in the uncertainty region of all uncertain superpixel units are sorted according to their uncertainty values, and features with large uncertainty values ​​are given high weights.

7. The method for sea ice classification and density inversion based on deep learning according to claim 6, characterized in that: The uncertainty measurement module calculates the uncertainty of the feature in the uncertainty region of the uncertain superpixel unit as follows: ； in, represents the sea ice forecast uncertainty, represents the uncertainty of seawater prediction, represents the uncertainty characteristics of sea ice, represents the uncertainty characteristics of seawater, Indicates the offset.

8. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The pre-training method of the Ice-WaterNet network model includes: Obtaining an ice condition map data set, wherein the ice condition map data set is annotated with a reference true value; The Ice-WaterNet network model is trained with the ice map dataset as input and the sea ice and sea water boundary lines as output. During the training process, the loss function is used to adjust the model parameters of the Ice-WaterNet network model. When the loss function value reaches the convergence condition, the pre-trained Ice-WaterNet network model is obtained.

9. The method for sea ice classification and density inversion based on deep learning according to claim 8, characterized in that: The loss function adopts a binary cross entropy function, which is expressed as: ； in, represents the binary cross entropy function, Represents the first The output of the convolutional layer, Indicates the number of convolutional layers of the Ice-WaterNet network model, Represents the true value of the benchmark.

10. The method for sea ice classification and density inversion based on deep learning according to claim 1, characterized in that: The sea ice density inversion is performed according to the sea ice classification result map to obtain the sea ice density inversion result, including: Downsampling the sea ice classification result according to a preset resolution to obtain a downsampled sea ice classification result; The downsampled sea ice classification results are projected according to the projection grid to obtain the percentage of sea ice pixels in each sea ice classification result projection grid, that is, the sea ice density of each sea ice classification result projection grid; The sea ice density of each sea ice classification result projection grid is combined to obtain the sea ice density inversion result.

Images