SAR (Synthetic Aperture Radar) water body extraction method based on scSE-U-Net model

Abstract

The invention provides an SAR (Synthetic Aperture Radar) water body extraction method based on an scSE-U-Net model, and relates to the field of radar remote sensing and ecological environment remote sensing, and the method comprises the following steps: obtaining dual-polarization SAR data and Yellow River source SAR data, and making and synthesizing a false color data set in combination with a principal component analysis method; based on the U-Net model and an scSE attention mechanism, an scSE-U-Net model is constructed; training an scSE-U-Net model by synthesizing a false color data set; performing semantic segmentation on the SAR data of the Yellow River source through an scSE-U-Net model to obtain a water body range of the Yellow River source region; and obtaining a long-time sequence lake extraction monitoring result in the water body range of the source region of the Yellow River by using DEM data and an artificial visual inspection method, and completing the extraction of the lake water body of the source region of the Yellow River. Compared with a traditional monitoring extraction method, the technical scheme of the application can obtain a lake water body extraction result with higher precision.

Description

Technical Field

[0001] The present application relates to the fields of radar remote sensing and ecological environment remote sensing, and in particular to a SAR water body extraction method based on the scSE-U-Net model. Background Art

[0002] At present, the main methods of lake monitoring are hydrological station monitoring and remote sensing monitoring. Traditional hydrological station monitoring can only detect changes in lake surface and lakeshore lines at a local spatiotemporal scale. However, in the source area of ​​the Yellow River in the northeast of the Qinghai-Tibet Plateau, the existence of thousands of lakes, including numerous thermal melt lakes, poses a challenge to data collection, and the harsh natural environment limits the availability of hydrological data. The development of remote sensing technology provides sufficient spatiotemporal resolution, global coverage, and relatively low cost, providing a potential solution for observing lakes at a larger spatiotemporal scale. Therefore, the use of remote sensing technology to monitor lakes in the source area of ​​the Yellow River is not only of great significance for the study and utilization of water resources in the region, but also provides scientific support for the sustainable utilization of water resources in the entire Yellow River Basin.

[0003] However, since optical sensors are easily affected by clouds and rain, it is difficult to provide large-scale high-quality cloud-free images. SAR plays an increasingly important role in water body monitoring due to its ability to work all day and its insensitivity to weather conditions such as clouds and rain. However, the presence of speckle noise and uneven grayscale distribution in SAR images limits the effectiveness of traditional water body extraction methods (such as threshold segmentation and target-based methods) in large-scale extraction tasks. Therefore, combining SAR data with intelligent water body extraction algorithms has become a research hotspot. Summary of the invention

[0004] The purpose of the present invention is to provide a SAR water body extraction method based on the scSE-U-Net model in order to solve the problems of low water body extraction accuracy and difficulty in extracting small water bodies in synthetic aperture radar using traditional methods.

[0005] The above-mentioned purpose of the present application is achieved through the following technical solutions: S1: Acquire dual-polarization SAR data and Yellow River source SAR data and perform preprocessing; S2: Analyze the preprocessed dual-polarization SAR data using principal component analysis to produce a synthetic false color data set; S3: Based on the U-Net semantic segmentation network and the scSE attention mechanism, a scSE-U-Net model is constructed; the scSE-U-Net model is trained by synthesizing a false color dataset to obtain the first scSE-U-Net model and the second scSE-U-Net model; S4: The first scSE-U-Net model and the second scSE-U-Net model are combined with the principal component analysis method to perform semantic segmentation on the Yellow River source SAR data and obtain the water body range in the Yellow River source area; S5: Using DEM data and manual visual inspection methods, we obtained the long-term lake extraction monitoring results within the water body range of the Yellow River source area and completed the extraction of lake water bodies in the Yellow River source area.

[0006] Optionally, step S1 includes: S11: Download dual-polarization SAR data of the Qinghai-Tibet Plateau. The dual-polarization SAR images contain water information. Downloading and processing the Yellow River source SAR data of the Yellow River source area in a preset year interval, including: the Yellow River source SAR data in a first preset year interval and the Yellow River source SAR data in a second preset year interval; the Yellow River source SAR data in the first preset year interval is single-polarization data in the VV band, and the Yellow River source SAR data in the second preset year interval is dual-polarization data in the VV / VH band; S12: Preprocess the dual-polarization SAR data and SAR data using SNAP software; Preprocessing includes: orbit correction, GRD boundary noise removal, thermal noise removal, radiation calibration, terrain correction, and backscatter coefficient decibel conversion.

[0007] Optionally, step S2 includes: S21: performing principal component analysis on the dual-polarization SAR data to determine the principal components of the dual-polarization SAR data; The principal components are mean filtered to obtain the PCA average texture features of dual-polarization SAR data; S22: linearly stretching the VV band of the dual-polarization SAR data, the VH band of the dual-polarization SAR data and the PCA average texture features to a grayscale range of 0-255; S23: according to the combination of the VV band corresponding to the R band, the VH band corresponding to the G band, and the PCA average texture feature corresponding to the B band, the three data are synthesized into an RGB image to obtain a first RGB image; S24: Use Labelme software to label water samples in the first RGB image to obtain a synthetic false color data set.

[0008] Optionally, step S3 includes: S31: Adopt the VGG16 structure and modify each group of convolutional layers and maximum pooling layers in the backbone feature extraction part of the U-Net semantic segmentation network; S32: Improve U-Net semantic segmentation network using scSE attention mechanism; S33: Train the scSE-U-Net model by synthesizing the false color data set to obtain the first scSE-U-Net model; S34: extract VV band data from the synthetic false color data set and construct a VV single polarization sample set; The scSE-U-Net model is trained by using the VV single-polarization sample set to obtain a second scSE-U-Net model for VV single-polarization data.

[0009] Optionally, step S4 includes: S41: The dual-polarization data of the Yellow River source SAR data are processed using the principal component analysis method to obtain a synthetic false color image of the Yellow River source; S42: Use the first scSE-U-Net model to extract water body information from the synthetic false color Yellow River source image and obtain the water body range in the Yellow River source area from 2017 to 2023; S43: stretch the single polarization data of the Yellow River source SAR data from 2014 to 2016 to the grayscale range of 0-255; The R band, the G band, and the B band are all assigned the single polarization data of the VV band of the Yellow River source SAR data, and synthesized into an RGB image to obtain a second RGB image; S44: Use the second scSE-U-Net model to extract water body information from the second RGB image and obtain the water body range in the source area of ​​the Yellow River from 2014 to 2016.

[0010] Optionally, step S5 includes: S51: Obtain DEM data of the Yellow River source area; use ArcMap 10.8 software to calculate the slope of the Yellow River source area; S52: setting a slope threshold; according to the slope conditions in the source area of ​​the Yellow River, within the water body range of the source area of ​​the Yellow River, removing areas with a slope greater than the slope threshold within the water body range of the source area of ​​the Yellow River; S53: Use manual visual inspection methods to verify the scope of the water bodies in the source area of ​​the Yellow River after elimination, and determine the long-term series lake extraction monitoring results within the scope of the water bodies in the source area of ​​the Yellow River.

[0011] An electronic device comprises a processor, a memory, a user interface and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes a SAR water body extraction method based on a scSE-U-Net model.

[0012] A computer-readable storage medium stores instructions. When the instructions are executed, a SAR water body extraction method based on a scSE-U-Net model is executed.

[0013] The beneficial effects of the technical solution provided by this application are: Based on the traditional U-Net architecture, the scSE-U-Net semantic segmentation neural network was constructed in combination with the scSE attention mechanism to extract lakes from synthetic false color images of dual-polarization SAR data. In order to train the scSE-U-Net semantic segmentation neural network, a synthetic false color dataset of VV+VH+PCA of different water body sizes in the Qinghai-Tibet Plateau was created. Based on the synthetic false color dataset, the performance of the scSE-U-Net water body extraction network and its ability to extract large lakes and small thermal melting lakes were verified. In the application stage, based on the synthetic false color dataset, the scSE-U-Net model was used to extract and analyze the water body information of lakes in the source area of ​​the Yellow River from 2014 to 2023. Through the scSE attention mechanism, the spatial detail information in the SAR image is fully retained while obtaining high-quality semantic context information, thereby improving the accuracy and effectiveness of extracting and segmenting large-scale water bodies from SAR images. Compared with traditional monitoring extraction methods, higher-precision lake water body extraction results can be obtained, so as to more comprehensively and effectively analyze the expansion and contraction changes of lakes in the source area of ​​the Yellow River over the years. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present application will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a step diagram in an embodiment of the present application; Figure 2 is the Landsat-8 optical image in the embodiment of the present application; Figure 3 This is a diagram showing the results of water extraction in an embodiment of the present application; Figure 4 It is a schematic diagram of the structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0016] An embodiment of the present application provides a SAR water body extraction method based on the scSE-U-Net model.

[0017] Please refer to Figure 1 , Figure 1 : is a step diagram of a SAR water body extraction method based on the scSE-U-Net model in an embodiment of the present application, including: S1: Acquire dual-polarization SAR data and Yellow River source SAR data and perform preprocessing; S2: Analyze the preprocessed dual-polarization SAR data using principal component analysis to produce a synthetic false color data set; S3: Based on the U-Net semantic segmentation network and the scSE attention mechanism, a scSE-U-Net model is constructed; the scSE-U-Net model is trained by synthesizing a false color dataset to obtain the first scSE-U-Net model and the second scSE-U-Net model; S4: The first scSE-U-Net model and the second scSE-U-Net model are combined with the principal component analysis method to perform semantic segmentation on the Yellow River source SAR data and obtain the water body range in the Yellow River source area; S5: Using DEM data and manual visual inspection methods, we obtained the long-term lake extraction monitoring results within the water body range of the Yellow River source area and completed the extraction of lake water bodies in the Yellow River source area.

[0018] Step S1 includes: S11: Download dual-polarization SAR data of the Qinghai-Tibet Plateau. The dual-polarization SAR images contain water information. Download and process the Yellow River source SAR data in the Yellow River source area from 2014 to 2023. The Yellow River source SAR data from 2014 to 2016 are single-polarization data in the VV band, and the Yellow River source SAR data from 2017 to 2023 are dual-polarization data in the VV / VH band; S12: Preprocess the dual-polarization SAR data and SAR data using SNAP software; Preprocessing includes: orbit correction, GRD boundary noise removal, thermal noise removal, radiation calibration, terrain correction, and backscatter coefficient decibel conversion.

[0019] Specifically, dual-polarization SAR data of the Qinghai-Tibet Plateau were downloaded from the official website of the European Space Agency; both dual-polarization SAR data and SAR data were collected by the Sentinel-1 satellite. SNAP (SeNtinel Application Platform) software was provided by the European Space Agency.

[0020] Step S2 includes: S21: performing principal component analysis on the dual-polarization SAR data to determine the principal components of the dual-polarization SAR data; The principal components are mean filtered to obtain the PCA average texture features of dual-polarization SAR data; Specifically, SAR image noise can be reduced by averaging texture features.

[0021] S22: linearly stretching the VV band of the dual-polarization SAR data, the VH band of the dual-polarization SAR data and the PCA average texture features to a grayscale range of 0-255; Specifically, since the VV band of the dual-polarization SAR data, the VH band of the dual-polarization SAR data and the PCA average texture features are at different intensity levels, they are firstly linearly stretched to a grayscale range of 0-255 before synthesis.

[0022] S23: according to the combination of the VV band corresponding to the R band, the VH band corresponding to the G band, and the PCA average texture feature corresponding to the B band, the three data are synthesized into an RGB image to obtain a first RGB image; S24: Use Labelme software to label water samples in the first RGB image to obtain a synthetic false color data set.

[0023] Specifically, the synthetic false color dataset includes 1680 experimental samples from Sentinel-1 dual-polarization SAR images from the Qinghai-Tibet Plateau from January 1, 2017 to December 31, 2023, with a resolution of 256×256. The study area is the Sanjiangyuan Yellow River National Park. Please refer to Figure 2 . Taking into account that each channel image of polarimetric SAR only intuitively displays the gray image information of the observed target, and the single grayscale image limits the image presentation and the available target recognition feature mining, a three-band RGB image is constructed using a data synthesis method. Single polarization refers to (HH) or (VV), which is horizontal transmission and horizontal reception or vertical transmission and vertical reception. Dual polarization refers to adding another polarization mode to one polarization mode, such as HH / HV (HH: horizontal transmission and horizontal reception, HV: horizontal transmission and vertical reception). Full polarization requires the simultaneous transmission of H and V, that is, four polarization modes: HH / HV / VV / VH.

[0024] Step S3 includes: S31: Adopt the VGG16 structure and modify each group of convolutional layers and maximum pooling layers in the backbone feature extraction part of the U-Net semantic segmentation network; Specifically, based on the U-Net semantic segmentation network, in the backbone extraction part, inspired by the original U-Net model, the U-Net backbone feature extraction part is similar to VGG, consisting of stacked convolution and maximum pooling layers, which modifies the backbone feature extraction part of the U-Net network.

[0025] S32: Improve U-Net semantic segmentation network using scSE attention mechanism; Specifically, an sSE (spatial dimension) module and a cSE (channel dimension) module are introduced, and two feature maps obtained by passing through the two modules respectively are added along the channel dimension to obtain a feature-enhanced feature map.

[0026] S33: Train the scSE-U-Net model by synthesizing the false color data set to obtain the first scSE-U-Net model; Specifically, the synthetic false color dataset is divided into a training set and a test set at a ratio of 80% and 20%, respectively. The training set is used to train the semantic segmentation model, and the test set is used to verify the extraction accuracy of the water body extraction model. The constructed scSE-U-Net is trained using the divided training set, and the training loss and MIoU are used as the accuracy indicators of the semantic segmentation model to evaluate the water body extraction accuracy of the model under the current training results. The hyperparameters are adjusted and retrained based on the evaluation results until the hyperparameters of the model with the best water body extraction accuracy are obtained. The calculation formula of MIoU is as follows:

[0027] Among them, n is the number of prediction categories of the model; TP indicates pixels predicted by the model as water bodies, but are actually water bodies; FP indicates pixels predicted by the model as water bodies, but are actually background pixels; FN indicates pixels predicted by the model as background, but are actually water bodies; TN indicates pixels predicted by the model as background, but are actually background pixels.

[0028] S34: extract VV band data from the synthetic false color data set and construct a VV single polarization sample set; The scSE-U-Net model is trained by using the VV single-polarization sample set to obtain a second scSE-U-Net model for VV single-polarization data.

[0029] Step S4 includes: S41: The dual-polarization data of the Yellow River source SAR data are processed using the principal component analysis method to obtain a synthetic false color image of the Yellow River source; S42: Use the first scSE-U-Net model to extract water body information from the synthetic false color Yellow River source image and obtain the water body range in the Yellow River source area from 2017 to 2023; S43: stretch the single polarization data of the Yellow River source SAR data from 2014 to 2016 to the grayscale range of 0-255; The R band, the G band, and the B band are all assigned the single polarization data of the VV band of the Yellow River source SAR data, and synthesized into an RGB image to obtain a second RGB image; S44: Use the second scSE-U-Net model to extract water body information from the second RGB image and obtain the water body range in the source area of ​​the Yellow River from 2014 to 2016.

[0030] Step S5 includes: S51: Obtain DEM data of the Yellow River source area; use ArcMap 10.8 software to calculate the slope of the Yellow River source area; S52: setting a slope threshold; according to the slope conditions in the source area of ​​the Yellow River, within the water body range of the source area of ​​the Yellow River, removing areas with a slope greater than the slope threshold within the water body range of the source area of ​​the Yellow River; Specifically, the areas with slopes greater than the slope threshold are mistakenly identified as water bodies, and the errors caused by mountain shadows are minimized through screening.

[0031] S53: Use manual visual inspection methods to verify the scope of the water bodies in the source area of ​​the Yellow River after elimination, and determine the long-term series lake extraction monitoring results within the scope of the water bodies in the source area of ​​the Yellow River.

[0032] Specifically, a method combining DEM data and manual work is used to minimize the impact of SAR image noise, mountain shadows, existing river channels and other factors, thereby obtaining long-term series lake extraction monitoring results in the Yellow River source area.

[0033] The present application provides an embodiment to extract monitoring results of long-term lake series, analyze the expansion and contraction change rules of the Yellow River source lake, and conduct a comprehensive analysis of the changes in lake expansion and contraction combined with climate factors such as annual precipitation and annual average surface temperature. The specific steps are as follows: S61: Based on the long-term lake extraction monitoring results, statistics are compiled on the changes in the total number of lakes, the number of large lakes (with an area greater than 1 square kilometer), the number of small lakes (with an area less than 1 square kilometer), and the total area of ​​lakes, the area of ​​large lakes (with an area greater than 1 square kilometer), and the area of ​​small lakes (with an area less than 1 square kilometer) in the Yellow River source area from 2014 to 2023, and the expansion and contraction patterns of the Yellow River source lakes are analyzed; S62: Download the reanalysis dataset ERA5 (ERA5-Landmonthly averaged data from 1950 to present) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) and calculate the annual average temperature data and annual total precipitation data in the source area of ​​the Yellow River from 2014 to 2023; S63: Using the correlation analysis method, the annual precipitation, annual average temperature, total lake area and total number of lakes in the region were correlated and the correlation coefficient R was calculated. 2 value.

[0034] Specifically, using data statistical analysis and correlation analysis methods, the article explains the expansion and contraction patterns of the Yellow River source lakes and the impact of climate factors on these patterns.

[0035] Specifically, taking into account the large number of lakes in the source area of ​​the Yellow River, the harsh natural environment and the relatively small scale of a large number of thermal melt lakes in the area, optical remote sensing is easily disturbed by cloudy and rainy weather, making it difficult to obtain high-quality images. The all-weather characteristics of synthetic aperture radar (SAR) images were used to perform principal component analysis (PCA) on the Sentinel-1 dual-polarization data, and the first principal component was mean filtered and synthesized with the VV and VH bands to construct a VV+VH+PCA false color synthetic dataset.

[0036] In one embodiment, in order to obtain high-quality semantic context information while fully retaining the spatial detail information in SAR images, thereby improving the accuracy and effectiveness of extracting and segmenting water bodies from SAR images, a scSE-U-Net semantic segmentation neural network was constructed based on the traditional U-Net architecture, which effectively extracted water body information in the source area of ​​the Yellow River from 2014 to 2023.

[0037] In another embodiment, a method combining DEM and manual detection was used to post-process the water body results, minimizing the impact of mountain shadows, river channels, SAR image noise, etc., and obtaining the results of surface lake extraction in the experimental area. Ultimately, compared with traditional monitoring and extraction methods, higher-precision lake water extraction results can be obtained, thereby more comprehensively and effectively analyzing the expansion and contraction changes of the Yellow River source lakes over the years.

[0038] Please refer to Figure 3, which is a display result of water body extraction of the extraction model of the present invention, wherein Figure 3(a) and Figure 3(d) are images of the VV band in the Sentinel-1 SAR image, Figure 3(b) and Figure 3(e) are real water bodies in the images annotated with Labelme software, and Figure 3(c) and Figure 3(f) are water body information in the images extracted using the scSE-U-Net model proposed in the present invention. It can be seen that the present invention obtains a higher-precision water body extraction result by utilizing the false color sample set synthesized by dual-polarization SAR data and the scSE-U-Net semantic segmentation neural network, which is more conducive to the subsequent analysis of long-term series lake expansion and contraction changes and major climate influencing factors.

[0039] The present application also discloses an electronic device. Figure 4 , Figure 4 The electronic device 500 may include: at least one processor 501 , at least one network interface 504 , a user interface 503 , a memory 505 , and at least one communication bus 502 .

[0040] The communication bus 502 is used to realize the connection and communication between these components.

[0041] The user interface 503 may include a display screen, and the optional user interface 503 may also include a standard wired interface or a wireless interface.

[0042] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0043] The present application also discloses a computer-readable storage medium storing a plurality of instructions suitable for loading by a processor to execute the above-mentioned SAR water body extraction method based on the scSE-U-Net model.

[0044] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0045] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A SAR water extraction method based on scSE-U-Net model, characterized in that: The method comprises the following steps: S1: Acquire dual-polarization SAR data and Yellow River source SAR data and perform preprocessing; S2: Analyze the preprocessed dual-polarization SAR data using principal component analysis to produce a synthetic false color data set; S3: Based on the U-Net semantic segmentation network and the scSE attention mechanism, a scSE-U-Net model is constructed; the scSE-U-Net model is trained by synthesizing a false color dataset to obtain the first scSE-U-Net model and the second scSE-U-Net model; S4: The first scSE-U-Net model and the second scSE-U-Net model are combined with the principal component analysis method to perform semantic segmentation on the Yellow River source SAR data and obtain the water body range in the Yellow River source area; S5: Using DEM data and manual visual inspection methods, we obtained the long-term lake extraction monitoring results within the water body range of the Yellow River source area and completed the extraction of lake water bodies in the Yellow River source area.

2. The SAR water body extraction method based on the scSE-U-Net model as claimed in claim 1, characterized in that: Step S1 includes: S11: Download dual-polarization SAR data of the Qinghai-Tibet Plateau. The dual-polarization SAR images contain water information. Downloading and processing the Yellow River source SAR data of the Yellow River source area in a preset year interval, including: the Yellow River source SAR data in a first preset year interval and the Yellow River source SAR data in a second preset year interval; the Yellow River source SAR data in the first preset year interval is single-polarization data in the VV band, and the Yellow River source SAR data in the second preset year interval is dual-polarization data in the VV / VH band; S12: Preprocess the dual-polarization SAR data and SAR data using SNAP software; Preprocessing includes: orbit correction, GRD boundary noise removal, thermal noise removal, radiation calibration, terrain correction, and backscatter coefficient decibel conversion.

3. The SAR water body extraction method based on the scSE-U-Net model as claimed in claim 1, characterized in that: Step S2 includes: S21: performing principal component analysis on the dual-polarization SAR data to determine the principal components of the dual-polarization SAR data; The principal components are mean filtered to obtain the PCA average texture features of dual-polarization SAR data; S22: linearly stretching the VV band of the dual-polarization SAR data, the VH band of the dual-polarization SAR data and the PCA average texture features to a grayscale range of 0-255; S23: according to the combination of the VV band corresponding to the R band, the VH band corresponding to the G band, and the PCA average texture feature corresponding to the B band, the three data are synthesized into an RGB image to obtain a first RGB image; S24: Use Labelme software to label water samples in the first RGB image to obtain a synthetic false color data set.

4. The SAR water body extraction method based on the scSE-U-Net model as claimed in claim 1, characterized in that: Step S3 includes: S31: Adopt the VGG16 structure and modify each group of convolutional layers and maximum pooling layers in the backbone feature extraction part of the U-Net semantic segmentation network; S32: Improve U-Net semantic segmentation network using scSE attention mechanism; S33: Train the scSE-U-Net model by synthesizing the false color data set to obtain the first scSE-U-Net model; S34: extract VV band data from the synthetic false color data set and construct a VV single polarization sample set; The scSE-U-Net model is trained by using the VV single-polarization sample set to obtain a second scSE-U-Net model for VV single-polarization data.

5. The SAR water body extraction method based on the scSE-U-Net model as claimed in claim 2, characterized in that: Step S4 includes: S41: The dual-polarization data of the Yellow River source SAR data are processed using the principal component analysis method to obtain a synthetic false color image of the Yellow River source; S42: using the first scSE-U-Net model, extracting water body information from the synthetic false color Yellow River source image to obtain the water body range of the Yellow River source area in the second preset year interval; S43: stretching the single polarization data of the Yellow River source SAR data in the first preset year interval to a grayscale range of 0-255; The R band, the G band, and the B band are all assigned the single polarization data of the VV band of the Yellow River source SAR data, and synthesized into an RGB image to obtain a second RGB image; S44: Using the second scSE-U-Net model, extract water body information from the second RGB image to obtain the water body range of the Yellow River source area in the first preset year interval.

6. The SAR water body extraction method based on the scSE-U-Net model as claimed in claim 1, characterized in that: Step S5 includes: S51: Obtain DEM data of the Yellow River source area; use ArcMap 10.8 software to calculate the slope of the Yellow River source area; S52: setting a slope threshold; according to the slope conditions in the source area of ​​the Yellow River, within the water body range of the source area of ​​the Yellow River, removing areas with a slope greater than the slope threshold within the water body range of the source area of ​​the Yellow River; S53: Use manual visual inspection methods to verify the scope of the water bodies in the source area of ​​the Yellow River after elimination, and determine the long-term series lake extraction monitoring results within the scope of the water bodies in the source area of ​​the Yellow River.

7. An electronic device, characterized in that: The electronic device comprises a processor (501), a memory (505), a user interface (503) and a network interface (504), wherein the memory (505) is used to store instructions, the user interface (503) and the network interface (504) are used to communicate with other devices, and the processor (501) is used to execute the instructions stored in the memory (505) so that the electronic device executes the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 6 is executed.

Images