Flood monitoring method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a flood monitoring method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining digital elevation data, multispectral image data and time sequence Sentinel-1 data of a flood region; determining topographic features according to the digital elevation data; selecting a plurality of water body sample areas and non-water body sample areas from the multispectral image data as training samples; for each piece of Sentinel-1 data, performing image segmentation on the piece of Sentinel-1 data by using a multi-resolution segmentation algorithm to obtain an image segmentation result; extracting a polarization wave band feature, a texture feature, a water body index feature and a time SAR feature of the Sentinel-1 data, taking each feature as an input of an initial random forest model, and training the initial random forest model by using a training sample to obtain a training result; therefore, the trained random forest model has the capability of distinguishing the water body area and the non-water body area in the Sentinel-1 data. Through the method, the flood monitoring precision is improved.

Description

Technical Field

[0001] This application relates to the technical field of flood monitoring, and in particular, to a flood monitoring method, device, electronic device, and computer-readable storage medium. Background Art

[0002] Floods are natural disasters caused by heavy rain, snowmelt, dam breaks, or overflows of lakes or rivers. They cause huge damage to the natural environment and casualties worldwide every year. With the increase in extreme precipitation events due to climate change, it is expected that the frequency of future flood events will further increase, thus affecting people's livelihoods. Therefore, effective flood disaster monitoring is crucial for minimizing losses of life, property, and infrastructure, as well as industrial, agricultural, and other losses. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a flood monitoring method, device, electronic device, and computer-readable storage medium to improve the monitoring accuracy of floods.

[0004] In a first aspect, an embodiment of this application provides a flood monitoring method, including:

[0005] Obtaining digital elevation data, multispectral image data, and time series Sentinel-1 data within a monitoring time period of the flood-occurring area; wherein, the monitoring time period is the time period from before the flood disaster occurs to after the flood disaster occurs in the flood-occurring area;

[0006] Determining the topographic features of the flood-occurring area according to the digital elevation data; and selecting multiple water body sample areas and multiple non-water body sample areas from the multispectral image data as training samples;

[0007] For each Sentinel-1 data in the time series Sentinel-1 data, using a multi-resolution segmentation algorithm to perform image segmentation on this Sentinel-1 data to obtain a corresponding image segmentation result; wherein, the image segmentation result contains the segmentation images of each ground object in the Sentinel-1 data;

[0008] Extracting the polarization band features, texture features, water body index features, and time SAR features of this Sentinel-1 data;

[0009] Taking the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water index features, and the temporal SAR features of this Sentinel-1 data as the input of the initial random forest model corresponding to this Sentinel-1 data, and using the training samples to train the initial random forest model, so that the trained random forest model corresponding to this Sentinel-1 data has the ability to distinguish the water area and the non-water area in this Sentinel-1 data.

[0010] Combined with the first aspect, the embodiments of the present application provide a first possible implementation manner of the first aspect. Before using the multi-resolution segmentation algorithm to perform image segmentation on this Sentinel-1 data to obtain the corresponding image segmentation result, it further includes:

[0011] Preprocessing this Sentinel-1 data to obtain preprocessed Sentinel-1 data; where the preprocessing includes: applying orbit files, GRD boundary noise removal, thermal noise removal, radiometric calibration, filtering, terrain correction, and converting the backscattering intensity to decibels.

[0012] The using the multi-resolution segmentation algorithm to perform image segmentation on this Sentinel-1 data to obtain the corresponding image segmentation result includes:

[0013] Invoking the optimal segmentation scale estimation tool to determine the best segmentation scale parameter, the optimal shape factor value, and the compactness parameter of the preprocessed Sentinel-data;

[0014] Inputting the determined best segmentation scale parameter, the optimal shape factor value, and the compactness parameter into the multi-resolution segmentation algorithm, and using the multi-resolution segmentation algorithm to perform image segmentation on the preprocessed Sentinel-1 data to obtain the image segmentation result.

[0015] Combined with the first aspect, the embodiments of the present application provide a second possible implementation manner of the first aspect. The extracting the polarization band features, texture features, water index features, and temporal SAR features of this Sentinel-1 data includes:

[0016] Based on the VV polarization and VH polarization of this Sentinel-1 data, calculating the polarization ratio, polarization difference, polarization sum, and polarization product to obtain the polarization band features of this Sentinel-1 data;

[0017] SAR dual-polarization texture features are extracted from the VV polarization and VH polarization of the Sentinel-1 data respectively based on the gray-level co-occurrence matrix to obtain the texture features of the Sentinel-1 data; wherein, the texture features are used to reflect the spatial differences of ground objects.

[0018] Based on the VV polarization and VH polarization of the Sentinel-1 data, a dual-polarization water index is calculated, and the dual-polarization water index is used as the water index feature of the Sentinel-1 data; the water index feature is used to highlight water bodies to distinguish water bodies from other ground objects.

[0019] According to the backscattering data of the Sentinel-1 data and the time-series Sentinel-1 data, the time SAR feature of the Sentinel-1 data is calculated; wherein, the time SAR feature includes backscattering anomaly and Z-score; the backscattering anomaly is used to characterize the difference between the backscattering during flood disasters and the average backscattering during non-flood periods; the Z-score is used to characterize the difference between the backscattering during flood disasters and the average backscattering during the monitoring period.

[0020] Combined with the first aspect, the embodiment of the present application provides a third possible implementation manner of the first aspect, wherein the terrain features include elevation, slope, aspect, and mountain shadow.

[0021] Combined with the first aspect, the embodiment of the present application provides a fourth possible implementation manner of the first aspect, wherein the step of using the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water index features, and the time SAR features as the input of the initial random forest model corresponding to the Sentinel-1 data and training the initial random forest model using the training samples includes:

[0022] Input the terrain features, the polarization band features, the texture features, the water index features, and the time SAR features of the Sentinel-1 data into the initial random forest model corresponding to the Sentinel-1 data, and determine the importance ranking of each input feature through the initial random forest model.

[0023] According to the importance ranking of each feature, add one feature to the initial random forest model in descending order to determine the classification accuracy of the initial random forest model after each feature is added.

[0024] According to the change in the classification accuracy of the initial random forest model, when the highest classification accuracy is determined, each feature added to the initial random forest model is determined, and each feature added when the classification accuracy is the highest is determined as the target feature corresponding to this Sentinel-1 data;

[0025] Take the image segmentation result corresponding to this Sentinel-1 data and each of the target features as the input of the initial random forest model corresponding to this Sentinel-1 data, and use the training samples to train the initial random forest model.

[0026] Combined with the fourth possible implementation manner of the first aspect, the embodiments of the present application provide a fifth possible implementation manner of the first aspect, wherein the method further includes:

[0027] For each Sentinel-1 data, after obtaining the trained random forest model corresponding to this Sentinel-1 data, fuse the feature values of each of the target features corresponding to this Sentinel-1 data into this Sentinel-1 data to obtain the fused Sentinel-1 data;

[0028] Input the fused Sentinel-1 data and the image segmentation result corresponding to this Sentinel-1 data into the random forest model, and based on the random forest model, identify the water area and non-water area in the fused Sentinel-1 data.

[0029] Combined with the fifth possible implementation manner of the first aspect, the embodiments of the present application provide a sixth possible implementation manner of the first aspect, wherein the method further includes:

[0030] Based on the identified water area and non-water area in the fused Sentinel-1 data, generate a binary classification image for distinguishing the water area and non-water area;

[0031] After obtaining the binary classification image corresponding to each Sentinel-1 data, for each pixel in the binary classification image, calculate the frequency at which the pixel is identified as water; wherein the frequency is used to reflect the duration of water at this pixel during the monitoring period;

[0032] Select multiple pixels from all the pixels as sample pixels, and determine whether each sample pixel is a permanent water pixel or a flood pixel in reality;

[0033] For each preset frequency threshold, compare the frequency corresponding to each of the sample pixels with the preset frequency threshold, so as to infer whether each of the sample pixels is a permanent water body pixel or a flood pixel based on the preset frequency threshold;

[0034] According to whether each of the sample pixels inferred is a permanent water body pixel or a flood pixel, and whether each of the sample pixels in the actual situation determined is a permanent water body pixel or a flood pixel, calculate the accuracy of the preset frequency threshold for distinguishing between permanent water body pixels and flood pixels;

[0035] According to the accuracy corresponding to each preset frequency threshold, select the preset frequency threshold with the highest accuracy as the target preset frequency threshold;

[0036] For each pixel of the binary classification image, compare the frequency of the pixel with the target preset frequency threshold. If the frequency of the pixel is greater than the target preset frequency threshold, it indicates that the pixel is a permanent water body pixel; if the frequency of the pixel is less than or equal to the target preset frequency threshold, it indicates that the pixel is a flood pixel.

[0037] In a second aspect, an embodiment of the present application further provides a flood monitoring device, including:

[0038] An acquisition module, configured to acquire digital elevation data, multi-spectral image data, and time series Sentinel-1 data during a monitoring time period of a flood occurrence area; wherein, the monitoring time period is the time period from before the flood disaster occurs to after the flood disaster occurs in the flood occurrence area;

[0039] A determination module, configured to determine the topographic features of the flood occurrence area according to the digital elevation data; and select a plurality of water body sample areas and a plurality of non-water body sample areas from the multi-spectral image data as training samples;

[0040] A segmentation module, configured to perform image segmentation on each Sentinel-1 data in the time series Sentinel-1 data by using a multi-resolution segmentation algorithm to obtain a corresponding image segmentation result; wherein, the image segmentation result includes the segmentation images of each ground object in the Sentinel-1 data;

[0041] An extraction module, configured to extract the polarization band feature, texture feature, water body index feature, and time SAR feature of the Sentinel-1 data;

[0042] A training module, configured to use the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water body index features, and the temporal SAR features of this Sentinel-1 data as the input of the initial random forest model corresponding to this Sentinel-1 data, and use the training samples to perform model training on the initial random forest model, so that the trained random forest model corresponding to this Sentinel-1 data has the ability to distinguish the water body area and the non-water body area in this Sentinel-1 data.

[0043] In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, the steps in any possible implementation manner of the first aspect described above are executed.

[0044] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the steps in any possible implementation manner of the first aspect described above are executed.

[0045] The flood monitoring method, device, electronic device, and computer-readable storage medium provided by the embodiments of the present application, wherein when training the initial random forest model, the terrain features of the flood occurrence area, the polarization band SAR features, texture features, water body index features, temporal SAR features of the Sentinel-1 data, and the segmentation images of each ground object in the Sentinel-1 data are combined, which is beneficial to improving the accuracy of the trained random forest model in identifying the water body area and the non-water body area.

[0046] To make the above objects, features, and advantages of the present application more obvious and understandable, the following specifically enumerates preferred embodiments and, in conjunction with the accompanying drawings, makes the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 Shows a flowchart of a flood monitoring method provided by an embodiment of the present application;

[0049] Figure 2 It shows a schematic diagram of an image segmentation result provided by an embodiment of the present application;

[0050] Figure 3 It shows a schematic structural diagram of a flood monitoring device provided by an embodiment of the present application;

[0051] Figure 4 It shows a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are only a part rather than all of the embodiments of the present application. Usually, the components of the embodiments of the present application described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided in the accompanying drawings below is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without making creative efforts fall within the scope of protection of the present application.

[0053] Machine learning classification techniques have been gradually applied to flood monitoring based on SAR images and achieved high extraction accuracy. Machine learning methods can efficiently process high-dimensional data, make full use of image feature information, and the trained models can be applied to multiple images of the same type. Therefore, some researchers are committed to the research on flood monitoring using machine learning methods and multi-temporal SAR images. The use of multi-temporal SAR provides a means to better understand the seasonal behavior of different land cover classes. At the same time, time series analysis can be used to obtain multi-temporal features, enabling more detailed extraction of flood-related classes. From the perspective of the classification object, although pixel-based flood information extraction is the most commonly used method, it only considers the information of the pixels themselves and ignores the spatial arrangement, texture, and morphological features of the objects. The fragmented and redundant classification results may not meet the accuracy requirements. While the object-oriented method segments the remote sensing image into image objects as the basic units for classification. After segmentation, the inherent features of each object (such as spectral, texture, and shape features) and the features describing the relationships between objects (such as connectivity and proximity) are used for object classification. The resulting results have relatively low redundancy and fragmentation, significantly improved land cover classification accuracy, and the noise problem is well improved. Therefore, the object-oriented machine learning method is becoming a very effective image classification method and is increasingly widely used in flood monitoring research. Among them, the random forest algorithm has high accuracy and anti-noise interference ability. At the same time, the importance of features can be estimated to achieve dimensionality reduction in the high-dimensional feature space. Compared with the current popular similar algorithms, it has a faster learning speed, higher accuracy, and stability.

[0054] Although many achievements have been made in flood identification using SAR images, it is still challenging to classify water types in flood disaster events from post-disaster remote sensing images because the backscattering signal of temporary water bodies is affected by the characteristics of both the water bodies and the flooded ground objects. From the perspective of the classification object, although pixel-based flood information extraction is the most commonly used method, it only considers the information of the pixels themselves and ignores the spatial arrangement, texture, and morphological features of the objects. The fragmented and redundant classification results may not meet the accuracy requirements.

[0055] Considering the above problems, based on this, the embodiments of the present application provide a flood monitoring method, device, electronic device, and computer-readable storage medium, which combine polarization band features, texture features, water body index features, time SAR features, and terrain features to train the initial random forest model, thereby improving the accuracy of the trained random forest model in identifying water body areas. The following will be described through embodiments.

[0056] To facilitate the understanding of this embodiment, first, a flood monitoring method disclosed in the embodiments of the present application will be introduced in detail. As Figure 1As shown, the following steps S101-S105 are included:

[0057] S101: Acquire digital elevation data, multispectral image data, and time series Sentinel-1 data within a monitoring period of the flooded area; wherein the monitoring period is the period from before the flood disaster occurs to after the flood disaster occurs in the flooded area.

[0058] In this embodiment, the digital elevation data (DEM data) of the flooded area is used to describe the spatial distribution of the surface terrain elevation in the flooded area.

[0059] The multispectral image data is the HJ-2 multispectral verification data, which is a data set generated by the multispectral payload of the Environmental Disaster Reduction No. 2 (HJ-2A / B) satellite and has undergone standardized preprocessing. It is mainly used to verify the radiation accuracy, geometric positioning accuracy and reliability of ground object classification of remote sensing images. The multispectral image data is generated and acquired within a preset time after the flood occurs (for example, within 3 days, including 3 days). In other words, the time interval between the generation time of the multispectral image data of the flooded area and the day of the flood is less than the preset time (for example, 3 days, including 3 days). Exemplarily, assuming that the day the flood disaster occurs is July 8, 2021, then the generation and acquisition time of the multispectral image data can be July 10, 2021.

[0060] In this embodiment, the time series Sentinel-1 data of the sample area within the monitoring period is obtained from the GEE platform. In this embodiment, the obtained time series Sentinel-1 data can be time series Sentinel-1 BIW mode GRD product data.

[0061] The time series Sentinel-1 data contains multiple Sentinel-1 data with a time sequence relationship. Sentinel-1 data can specifically be Sentinel-1B IW mode GRD product data. Among them, Sentinel-1B IW mode GRD product data (Sentinel-1B IW mode first-level ground detection distance (GRD) product data) belongs to the first-level product (Level-1) in the Sentinel-1 data system. It is a radar intensity image generated by obtaining the surface backscatter signal through synthetic aperture radar (SAR) and multi-view processing.

[0062] Sentinel-1 data contains radar backscatter intensity, geospatial information, and metadata. Among them, the radar backscatter intensity is used to reflect the electromagnetic wave reflection characteristics of the earth's surface and can be used to invert surface parameters such as soil moisture and water body extent. The geospatial information includes longitude and latitude coordinates, spatial resolution, and coverage. The metadata records satellite orbit parameters, imaging time, polarization mode (such as VV+VH dual polarization), and processing level information.

[0063] The monitoring time period is a period from before to after the flood disaster in the flood-affected area. Exemplarily, assuming that the flood disaster occurred on July 8, 2021, then the monitoring time period can be from May 1, 2021, to September 1, 2021 (a total of 124 days). Then, the time series Sentinel-1 data may contain 10 Sentinel-1 data with a time series relationship (i.e., a chronological relationship).

[0064] In this embodiment, the flood-affected area can be one or more. When there are multiple flood-affected areas, each flood-affected area corresponds to its own digital elevation data, multispectral image data, and time series Sentinel-1 data. At this time, each step in this embodiment is sequentially executed for each flood-affected area.

[0065] S102: Determine the topographic features of the flood-affected area based on the digital elevation data; and select multiple water body sample areas and multiple non-water body sample areas from the multispectral image data as training samples.

[0066] In this embodiment, to improve the reliability of random forest classification, topographic features derived from digital elevation data are introduced as auxiliary inputs. Among them, the topographic features include the elevation, slope, aspect, and mountain shadow of the flood-affected area.

[0067] In this embodiment, Sentinel-1 data is actually a grayscale image, and features such as water bodies and roads in the image are represented in black. Therefore, it is difficult to distinguish water bodies in Sentinel-1 data. Based on this, in this embodiment, the multispectral image data is used as training samples. The multispectral image data belongs to a multispectral image, and different colors are used for the water body area and the non-water body area in it. Therefore, the water body area can be marked based on the multispectral image data.

[0068] Specifically, assume that in this embodiment, there are multiple water body areas (such as 500 water body areas) and multiple non-water body areas (such as 1200 non-water body areas) in the flood-affected area. Then, 500 water body areas and 1200 non-water body areas will be marked in the multispectral image data.

[0069] Select multiple water body regions (e.g., 50 water body regions) from the multi-spectral image data as water body sample regions, and select multiple non-water body regions (e.g., 100 non-water body regions) as non-water body sample regions, and use the selected multiple water body sample regions (e.g., 50 water body sample regions) and multiple non-water body sample regions (e.g., 100 non-water body sample regions) as training samples.

[0070] The training samples may include: the positions of each water body sample region and the water body annotation labels corresponding to each water body sample region, and the positions of each non-water body sample region and the non-water body annotation labels corresponding to each non-water body sample region.

[0071] Among them, the number of selected water body sample regions is less than the total number of water body regions. Similarly, the number of selected non-water body sample regions is less than the total number of non-water body regions.

[0072] In this embodiment, when selecting multiple water body sample regions and multiple non-water body sample regions from the multi-spectral image data, specifically, multiple water body sample regions and multiple non-water body sample regions may be randomly selected from the multi-spectral image data.

[0073] S103: For each Sentinel-1 data in the time series Sentinel-1 data, use the multi-resolution segmentation algorithm to perform image segmentation on this Sentinel-1 data to obtain the corresponding image segmentation result; where the image segmentation result contains the segmented images of each ground object in the Sentinel-1 data.

[0074] In this embodiment, the ground objects include water bodies, vegetation, construction land, etc. The image segmentation result contains the segmented images corresponding to each ground object in the flood area. For example, the water body corresponds to a segmented image. Among them, the segmented image corresponding to each ground object refers to: the area where this ground object is located is segmented from the Sentinel-1 data, and the average value of all pixels in the area where this ground object is located is calculated to obtain the segmented image. As Figure 2 shown, a schematic diagram of an image segmentation result is shown.

[0075] In this embodiment, each Sentinel-1 data corresponds to its own image segmentation result.

[0076] In a possible implementation manner, before performing step S103, each Sentinel-1 data may also be preprocessed through the following steps:

[0077] For each Sentinel-1 data in the time series Sentinel-1 data, preprocess the Sentinel-1 data to obtain the preprocessed Sentinel-1 data; wherein, the preprocessing includes: applying an orbit file, removing GRD boundary noise, removing thermal noise, radiometric calibration, filtering, terrain correction, and converting the backscatter intensity to decibels;

[0078] Among them, the orbit file is used in the preprocessing of GRD product data to correct the position and velocity information of the satellite to eliminate geolocation errors caused by satellite orbit parameter deviations. Removing thermal noise is used to eliminate radar system noise (such as thermal noise and sensor inherent noise) and improve the signal-to-noise ratio (by default, the VH and VV polarization channels are retained). Radiometric calibration is used to convert the original digital value (DN) to the backscatter coefficient (σ0 or β0) to achieve radiometric consistency of data in different time phases.

[0079] Use SRTM for terrain correction. Specifically, use SRTM or higher-resolution DEM data to correct terrain distortion through the Range-Doppler model to generate a geocoded image. Convert the backscatter intensity to decibels (dB) through a logarithmic scale to enhance the contrast of low-reflection areas. Use a Refined Lee filter with a 3×3 pixel window size to remove speckle noise, suppress the speckle noise inherent in the Sentinel-1 data, and enhance the texture features of ground objects.

[0080] In a possible implementation manner, when performing step S103 to segment the Sentinel-1 data using a multi-resolution segmentation algorithm to obtain the corresponding image segmentation result, it can be specifically performed according to the following steps:

[0081] Call the optimal segmentation scale estimation tool (ESP2) to determine the best segmentation scale parameter (such as 80 pixels), the optimal shape factor value (such as 0.1), and the compactness parameter (such as 0.5) of the preprocessed Sentinel- data;

[0082] Input the determined best segmentation scale parameter, the optimal shape factor value, and the compactness parameter into the multi-resolution segmentation algorithm, and use the multi-resolution segmentation algorithm to segment the preprocessed Sentinel-1 data to obtain the image segmentation result.

[0083] S104: Extract the polarization band features, texture features, water body index features, and time SAR features of the Sentinel-1 data of this image.

[0084] In a possible implementation manner, when performing step S104, it can be specifically performed according to the following steps S1041-S1044:

[0085] S1041: Calculate the polarization ratio, polarization difference, polarization sum, and polarization product based on the VV polarization and VH polarization of the Sentinel-1 data, and obtain the polarization band characteristics of the Sentinel-1 data.

[0086] In this embodiment, each Sentinel-1 data corresponds to a VV (vertical transmit and vertical receive) polarization and a VH (vertical transmit and horizontal receive) polarization. For each Sentinel-1 data, calculate the ratio of the VV polarization to the VH polarization in the Sentinel-1 data (i.e., the polarization ratio), calculate the difference between the VV polarization and the VH polarization (i.e., the polarization difference), calculate the sum of the VV polarization and the VH polarization (i.e., the polarization sum), and calculate the product of the VV polarization and the VH polarization (i.e., the polarization product). Take the calculated polarization ratio, polarization difference, polarization sum, and polarization product as the polarization band characteristics of the Sentinel-1 data.

[0087] In this embodiment, each Sentinel-1 data corresponds to its own polarization band characteristics.

[0088] S1042: Based on the Gray-Level Co-occurrence Matrix (GLCM), extract the SAR dual-polarization texture characteristics of the VV polarization and VH polarization of the Sentinel-1 data respectively, and obtain the texture characteristics of the Sentinel-1 data; among them, the texture characteristics are used to reflect the spatial differences of ground objects.

[0089] In this embodiment, calculate the texture characteristics such as Angular Second Moment, Entropy, Mean, Variance, Contrast, Correlation, Homogeneity, and Dissimilarity of the VV polarization and VH band of the Sentinel-1 data respectively based on the Gray-Level Co-occurrence Matrix (GLCM).

[0090] S1043: Calculate the Sentinel-1 Dual-Polarized Water Index (SDWI) based on the VV polarization and VH polarization of the Sentinel-1 data, and take the dual-polarized water index as the water index characteristic of the Sentinel-1 data; the water index characteristic is used to highlight the water body to distinguish the water body from other ground objects except the water body.

[0091] In this embodiment, the dual-polarization water body index is calculated through the following operation formula:

[0092] K SDWI = ln(10·VV·VH) - 8

[0093] where VV represents VV polarization, VH represents VH polarization, and K SDWI represents the dual-polarization water body index.

[0094] S1044: Calculate the temporal SAR features of this Sentinel-1 data based on the backscattering data of this Sentinel-1 data and the time-series Sentinel-1 data; wherein, the temporal SAR features include backscattering anomaly and Z-score; the backscattering anomaly is used to characterize the difference between the backscattering during the flood disaster and the average backscattering during the non-flood period; the Z-score is used to characterize the difference between the backscattering during the flood disaster and the average backscattering of the first monitoring time period.

[0095] In this embodiment, the backscattering anomaly is calculated through the following formula:

[0096]

[0097] where Δσ(t) represents the backscattering anomaly; σ(t) is the backscattering data on the day of the flood disaster; represents the average backscattering data during the non-flood period (i.e., excluding the day of the flood disaster).

[0098] The Z-score is calculated through the following formula:

[0099]

[0100] where Z represents the Z-score; std(σ) represents the standard deviation of the backscattering data of all the first Sentinel-1B IW-mode GRD product data.

[0101] S105: Use the terrain features, the image segmentation result of this Sentinel-1 data, the polarization band features, the texture features, the water body index features, and the temporal SAR features as the input of the initial random forest model corresponding to this Sentinel-1 data, and use the training samples to train the initial random forest model so that the trained random forest model corresponding to this Sentinel-1 data has the ability to distinguish the water body area and the non-water body area in this Sentinel-1 data.

[0102] In this embodiment, each piece of Sentinel-1 data corresponds to an initial random forest model. Each piece of Sentinel-1 data is used to train its corresponding initial random forest model with its corresponding data pair to obtain the random forest model corresponding to each piece of Sentinel-1 data. Among them, the trained random forest model can identify each water area and non-water area in this piece of Sentinel-1 data.

[0103] In a possible implementation manner, when performing step S105 to train the initial random forest model corresponding to each piece of Sentinel-1 data, it can be specifically performed according to the following steps:

[0104] S1051: Input the terrain feature, as well as the polarization band feature, texture feature, water index feature, and temporal SAR feature of this piece of Sentinel-1 data into the initial random forest model corresponding to this piece of Sentinel-1 data, and determine the importance ranking of each input feature through this initial random forest model.

[0105] In this embodiment, the importance degree of the features (i.e., polarization band feature, texture feature, water index feature, temporal SAR feature, terrain feature) directly affects the classification accuracy of the initial random forest model. Calculating the importance ranking of each feature and making an optimal selection can improve the classification speed of the initial random forest model and reduce data redundancy.

[0106] In this embodiment, the importance ranking of each feature corresponding to each piece of Sentinel-1 data may be different.

[0107] S1042: According to the importance ranking of each feature, add one feature to the initial random forest model in turn from high to low to determine the classification accuracy of the initial random forest model after each feature is added.

[0108] Exemplarily, assume that the importance ranking of each feature corresponding to one piece of Sentinel-1 data from high to low is: polarization band feature, texture feature, water index feature, temporal SAR feature, terrain feature.

[0109] At this time, first fuse the polarization band feature corresponding to this piece of Sentinel-1 data into this piece of Sentinel-1 data, and input the fused Sentinel-1 data and its corresponding image segmentation result into the initial random forest model corresponding to this piece of Sentinel-1 data, and calculate the classification accuracy of this initial random forest model at this time based on the training samples.

[0110] Next, the polarization band features and texture features corresponding to this Sentinel-1 data are fused into this Sentinel-1 data, and the fused Sentinel-1 data and its corresponding image segmentation result are input into the initial random forest model corresponding to this Sentinel-1 data. Based on the training samples, the classification accuracy of the initial random forest model at this time is calculated.

[0111] And so on. Each time one more feature is fused, the classification accuracy of the initial random forest model corresponding to this Sentinel-1 data after adding each feature can be obtained.

[0112] S1053: According to the change in the classification accuracy of the initial random forest model, when the highest classification accuracy is determined, each feature added to this initial random forest model is determined, and each feature added when the classification accuracy is the highest is determined as the target feature corresponding to this Sentinel-1 data.

[0113] In this embodiment, through steps S1051 - S1053, the screening of multiple features corresponding to this Sentinel-1 data is completed, and the optimal feature combination corresponding to this Sentinel-1 data is found.

[0114] S1054: The image segmentation result and each target feature of this Sentinel-1 data are used as the input of the initial random forest model corresponding to this Sentinel-1 data, and the initial random forest model is trained using the training samples.

[0115] In this embodiment, after finding the optimal feature combination (i.e., at least one target feature) corresponding to this Sentinel-1 data, next, the initial random forest model corresponding to this Sentinel-1 data is trained. Among them, when training the initial random forest model corresponding to this Sentinel-1 data, the specific process can be:

[0116] For each Sentinel-1 data, the feature values of each target feature corresponding to this Sentinel-1 data are fused in this Sentinel-1 data to obtain the fused Sentinel-1 data;

[0117] The fused Sentinel-1 data and the image segmentation result corresponding to this Sentinel-1 data are input into the initial random forest model to be trained, and the initial classification result of the water body sample area and the non-water body sample area in this Sentinel-1 data is output through the initial random forest model;

[0118] Based on the training samples, calculate the loss value between the training samples and the initial classification result, and use this loss value to train the initial random forest model; wherein, the training samples include the positions of each water body sample region and the corresponding water body annotation labels for each water body sample region, as well as the positions of each non-water body sample region and the corresponding non-water body annotation labels for each non-water body sample region.

[0119] When the preset model training cutoff condition is met, stop training the initial random model to obtain the trained random forest model corresponding to this Sentinel-1 data. The trained random forest model has the ability to distinguish water body regions and non-water body regions in this Sentinel-1 data.

[0120] In a possible implementation manner, for each Sentinel-1 data, after obtaining the trained random forest model corresponding to this Sentinel-1 data, the water body regions and non-water body regions in this Sentinel-1 data can be classified according to the following steps S201 - S202:

[0121] S201: Integrate the feature values of each of the target features corresponding to this Sentinel-1 data into this Sentinel-1 data to obtain the integrated Sentinel-1 data.

[0122] S202: Input the integrated Sentinel-1 data and the image segmentation result corresponding to this Sentinel-1 data into the random forest model, and based on the random forest model, identify the water body regions and non-water body regions in the integrated Sentinel-1 data.

[0123] In this embodiment, for each Sentinel-1 data, use some water body sample regions (such as 50 water body sample regions) and non-water body sample regions (such as 100 non-water body sample regions) in this Sentinel-1 data to train the initial random forest model corresponding to this Sentinel-1 data. After training is completed, the trained random forest model can distinguish all the water body regions (such as 500 water body sample regions) and non-water body regions (such as 1200 non-water body sample regions) in this Sentinel-1 data.

[0124] In a possible implementation manner, after executing step S202 and obtaining the recognition results corresponding to each Sentinel-1 data, the following steps S203 - S209 can also be executed:

[0125] S203: Based on the identified water area and non-water area in the fused Sentinel-1 data, generate a binary classification image for distinguishing the water area and the non-water area.

[0126] In this embodiment, the water area in the binary classification image is represented by one color, and the non-water area is represented by another color.

[0127] S204: After obtaining the binary classification image corresponding to each Sentinel-1 data, for each pixel in the binary classification image, calculate the frequency at which the pixel is identified as water; wherein, the frequency is used to reflect the duration of water at the pixel during the monitoring period.

[0128] In this embodiment, a pixel refers to the smallest regional unit in the binary classification image. When calculating the frequency at which each pixel is identified as water in each binary classification image, the following formula can be used for calculation:

[0129]

[0130] where N W represents the number of times the pixel W is identified as water; N a represents the total number of Sentinel-1 data; f W represents the frequency at which the pixel W is identified as water in each binary classification image.

[0131] In this embodiment, the greater the frequency corresponding to each pixel, the longer the duration of water at the pixel; conversely, the smaller the frequency, the shorter the duration of water at the pixel.

[0132] S205: Select multiple pixels from all pixels as sample pixels, and determine whether each sample pixel is a permanent water pixel or a flood pixel in the real situation.

[0133] In this embodiment, each pixel corresponds to a position within the flood occurrence area. In order to identify the permanent water area (i.e., the area with water before the flood occurred) and the flood area (i.e., the area without water before the flood occurred and with water after the flood occurred) within the flood occurrence area, in this embodiment, first select multiple pixels (e.g., hundreds of pixels) from all pixels (e.g., tens of thousands of pixels) as sample pixels. Since the number of sample pixels is limited, therefore, according to the position corresponding to each sample pixel, whether it is located in the permanent water area or the flood area in the real situation, to determine whether each sample pixel is a permanent water pixel or a flood pixel in the real situation.

[0134] Among them, that the sample pixel is a permanent water body pixel in the real situation means that the position corresponding to the sample pixel in the flood area is a permanent water body area. That the sample pixel is a flood pixel in the real situation means that the position corresponding to the sample pixel in the flood area is a flood area.

[0135] S206: For each preset frequency threshold, compare the frequency corresponding to each sample pixel with the preset frequency threshold to infer whether each sample pixel is a permanent water body pixel or a flood pixel based on the preset frequency threshold.

[0136] In this embodiment, the value range of the preset frequency threshold can be 0.1 - 1. If the step size is 0.1, then the values of the preset frequency threshold can be: 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.

[0137] Taking the preset frequency threshold of 0.2 as an example for illustration, as shown in Table 1 below, assuming there are 8 sample pixels (a - h), when inferring whether each sample pixel is a permanent water body pixel or a flood pixel based on the preset frequency threshold of 0.2:

[0138] The frequency of sample pixel a, 0.1, is less than or equal to the preset frequency threshold of 0.2. Therefore, sample pixel a is inferred as a flood pixel.

[0139] The frequency of sample pixel b, 0.2, is less than or equal to the preset frequency threshold of 0.2. Therefore, sample pixel b is inferred as a flood pixel.

[0140] The frequency of sample pixel c, 0.3, is greater than the preset frequency threshold of 0.2. Therefore, sample pixel c is inferred as a permanent water body pixel.

[0141] The frequency of sample pixel d, 0.4, is greater than the preset frequency threshold of 0.2. Therefore, sample pixel d is inferred as a permanent water body pixel.

[0142] The frequency of sample pixel e, 0.5, is greater than the preset frequency threshold of 0.2. Therefore, sample pixel e is inferred as a permanent water body pixel.

[0143] The frequency of sample pixel f, 0.6, is greater than the preset frequency threshold of 0.2. Therefore, sample pixel f is inferred as a permanent water body pixel.

[0144] The frequency of sample pixel g, 0.2, is less than or equal to the preset frequency threshold of 0.2. Therefore, sample pixel g is inferred as a flood pixel.

[0145] The frequency of sample pixel h, 0.1, is less than or equal to the preset frequency threshold of 0.2. Therefore, sample pixel h is inferred as a flood pixel.

[0146]

[0147] Table 1

[0148] S207: Calculate the accuracy of the preset frequency threshold for distinguishing between permanent water body pixels and flood pixels based on the inferred permanent water body pixels or flood pixels of each sample pixel and the determined permanent water body pixels or flood pixels of each sample pixel under actual conditions.

[0149] Combined with the foregoing embodiments, then, the accuracy of using the preset frequency threshold to distinguish between permanent water body pixels and flood pixels can be calculated as: 7 / 8 = 85.7%.

[0150] S208: Screen out the preset frequency threshold with the highest accuracy as the target preset frequency threshold according to the accuracy corresponding to each preset frequency threshold.

[0151] In this embodiment, after calculating the accuracy corresponding to each preset frequency threshold, screen out the preset frequency threshold with the highest accuracy from each preset frequency threshold as the target preset frequency threshold. Exemplarily, the target preset frequency threshold can be 0.3.

[0152] S209: For each pixel of the binary classification image, compare the frequency of the pixel with the target preset frequency threshold. If the frequency of the pixel is greater than the target preset frequency threshold, it means that the pixel is a permanent water body pixel; if the frequency of the pixel is less than or equal to the target preset frequency threshold, it means that the pixel is a flood pixel.

[0153] In this embodiment, after screening out the target preset frequency threshold, the binary classification image can be classified according to the target preset frequency threshold for each pixel to determine whether each pixel is a permanent water body pixel or a flood pixel, that is, classify the position corresponding to each pixel in the flood occurrence area to determine whether each position is a permanent water body area or a flood area.

[0154] Specifically, when the frequency of the pixel is greater than the target preset frequency threshold, it means that the pixel is a permanent water body pixel, and it also means that the position corresponding to the pixel in the flood occurrence area is a permanent water body area. When the frequency of the pixel is less than or equal to the target preset frequency threshold, it means that the pixel is a flood pixel, and it also means that the position corresponding to the pixel in the flood occurrence area is a flood area.

[0155] Based on the same technical concept, the embodiment of the present application also provides a flood monitoring device, as Figure 3 shown, the device includes:

[0156] An acquisition module 301, configured to acquire digital elevation data, multi-spectral image data of a flood-occurring area, and time-series Sentinel-1 data during a monitoring time period; wherein, the monitoring time period is the time period from before the flood disaster occurs to after the flood disaster occurs in the flood-occurring area;

[0157] A determination module 302, configured to determine the terrain features of the flood-occurring area according to the digital elevation data; and select a plurality of water body sample areas and a plurality of non-water body sample areas from the multi-spectral image data as training samples;

[0158] A segmentation module 303, configured to perform image segmentation on each Sentinel-1 data in the time-series Sentinel-1 data by using a multi-resolution segmentation algorithm to obtain a corresponding image segmentation result; wherein, the image segmentation result includes the segmentation images of each ground object in the Sentinel-1 data;

[0159] An extraction module 304, configured to extract the polarization band features, texture features, water body index features, and time SAR features of the Sentinel-1 data;

[0160] A training module 305, configured to use the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water body index features, and the time SAR features as inputs of an initial random forest model corresponding to the Sentinel-1 data, and use the training samples to perform model training on the initial random forest model, so that the trained random forest model corresponding to the Sentinel-1 data has the ability to distinguish water body areas and non-water body areas in the Sentinel-1 data.

[0161] Optionally, the device further includes:

[0162] A preprocessing module, configured to preprocess the Sentinel-1 data to obtain preprocessed Sentinel-1 data before the segmentation module 303 performs image segmentation on the Sentinel-1 data by using a multi-resolution segmentation algorithm to obtain a corresponding image segmentation result; wherein, the preprocessing includes: applying an orbit file, removing GRD boundary noise, removing thermal noise, radiometric calibration, filtering processing, terrain correction, and converting the backscattering intensity to decibels;

[0163] When the segmentation module 303 is configured to perform image segmentation on the Sentinel-1 data by using a multi-resolution segmentation algorithm to obtain a corresponding image segmentation result, it is specifically configured to:

[0164] Call the optimal segmentation scale estimation tool to determine the optimal segmentation scale parameters, the optimal shape factor value, and the compactness parameters of the preprocessed Sentinel-1 data;

[0165] Input the determined optimal segmentation scale parameters, the optimal shape factor value, and the compactness parameters into the multi-resolution segmentation algorithm, and use the multi-resolution segmentation algorithm to perform image segmentation on the preprocessed Sentinel-1 data to obtain an image segmentation result.

[0166] Optionally, when the extraction module 304 is used to extract the polarization band features, texture features, water body index features, and temporal SAR features of this Sentinel-1 data, it specifically is used for:

[0167] Based on the VV polarization and VH polarization of this Sentinel-1 data, calculate the polarization ratio, polarization difference, polarization sum, and polarization product to obtain the polarization band features of this Sentinel-1 data;

[0168] Based on the gray-level co-occurrence matrix, perform SAR dual-polarization texture feature extraction on the VV polarization and VH polarization of this Sentinel-1 data respectively to obtain the texture features of this Sentinel-1 data; wherein, the texture features are used to reflect the spatial differences of ground objects;

[0169] Based on the VV polarization and VH polarization of this Sentinel-1 data, calculate the dual-polarization water body index, and use the dual-polarization water body index as the water body index feature of this Sentinel-1 data; the water body index feature is used to highlight water bodies to distinguish water bodies from other ground objects;

[0170] According to the backscattering data of this Sentinel-1 data and the time series Sentinel-1 data, calculate the temporal SAR features of this Sentinel-1 data; wherein, the temporal SAR features include backscattering anomalies and Z-scores; the backscattering anomalies are used to characterize the difference between the backscattering during flood disasters and the average backscattering during non-flood periods; the Z-scores are used to characterize the difference between the backscattering during flood disasters and the average backscattering during the monitoring period.

[0171] Optionally, the terrain features include elevation, slope, aspect, and mountain shadow.

[0172] Optionally, the training module 305 is configured to use the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water body index features, and the temporal SAR features of this Sentinel-1 data as the input of the initial random forest model corresponding to this Sentinel-1 data, and use the training samples to train the initial random forest model, so that the trained random forest model corresponding to this Sentinel-1 data has the ability to distinguish the water body area and the non-water body area in this Sentinel-1 data. Specifically, it is configured to:

[0173] Input the terrain features, the polarization band features, the texture features, the water body index features, and the temporal SAR features of this Sentinel-1 data into the initial random forest model corresponding to this Sentinel-1 data, and determine the importance ranking of each input feature through the initial random forest model;

[0174] According to the importance ranking of each feature, add one feature to the initial random forest model in descending order to determine the classification accuracy of the initial random forest model after each feature is added;

[0175] According to the change of the classification accuracy of the initial random forest model, determine the features added to the initial random forest model when the classification accuracy is the highest, and determine the features added when the classification accuracy is the highest as the target features corresponding to this Sentinel-1 data;

[0176] Use the image segmentation result corresponding to this Sentinel-1 data and each of the target features as the input of the initial random forest model corresponding to this Sentinel-1 data, and use the training samples to train the initial random forest model.

[0177] Optionally, the device further includes:

[0178] A fusion module, configured to, for each Sentinel-1 data, after obtaining the trained random forest model corresponding to this Sentinel-1 data, fuse the feature values of each of the target features corresponding to this Sentinel-1 data into this Sentinel-1 data to obtain the fused Sentinel-1 data;

[0179] An identification module, configured to input the fused Sentinel-1 data and the image segmentation result corresponding to this Sentinel-1 data into the random forest model, and based on the random forest model, identify the water body area and the non-water body area in the fused Sentinel-1 data.

[0180] Optionally, the device further includes:

[0181] A generation module, configured to generate a binary classification image for distinguishing a water area and a non-water area based on the identified water area and non-water area in the fused Sentinel-1 data of this image;

[0182] A first calculation module, configured to calculate the frequency of each pixel in the binary classification image being identified as water after obtaining the binary classification image corresponding to each Sentinel-1 data; wherein, the frequency is used to reflect the duration of water at this pixel during the monitoring period;

[0183] A selection module, configured to select a plurality of pixels from all the pixels as sample pixels, and determine whether each of the sample pixels is a permanent water pixel or a flood pixel under actual conditions;

[0184] A first comparison module, configured to compare the frequency corresponding to each of the sample pixels with the preset frequency threshold for each preset frequency threshold, so as to infer whether each of the sample pixels is a permanent water pixel or a flood pixel based on the preset frequency threshold;

[0185] A second calculation module, configured to calculate the accuracy of the preset frequency threshold for distinguishing between permanent water pixels and flood pixels according to the inferred permanent water pixels or flood pixels of each of the sample pixels and the determined permanent water pixels or flood pixels of each of the sample pixels under actual conditions;

[0186] A screening module, configured to screen out the preset frequency threshold with the highest accuracy as the target preset frequency threshold according to the accuracy corresponding to each preset frequency threshold;

[0187] A second comparison module, configured to compare the frequency of each pixel in the binary classification image with the target preset frequency threshold. If the frequency of the pixel is greater than the target preset frequency threshold, it indicates that the pixel is a permanent water pixel; if the frequency of the pixel is less than or equal to the target preset frequency threshold, it indicates that there is a flood pixel at this pixel.

[0188] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application, including: a processor 401, a memory 402, and a bus 403. The memory 402 stores machine-readable instructions executable by the processor 401. When the electronic device runs the above information processing method, the processor 401 communicates with the memory 402 through the bus 403, and the processor 401 executes the machine-readable instructions to execute the method steps in the first embodiment.

[0189] The embodiment of this application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the method steps in the first embodiment.

[0190] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described device, electronic device, and computer-readable storage medium can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0191] In several embodiments provided by this application, it should be understood that the disclosed methods, devices, electronic devices, and computer-readable storage media can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some communication interfaces. The indirect coupling or communication connection of devices or modules can be in an electrical, mechanical, or other form.

[0192] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0193] In addition, in each embodiment of this application, each functional unit can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0194] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0195] Finally, it should be noted that: the above-mentioned embodiments are only specific implementation manners of this application, used to illustrate the technical solutions of this application, rather than limiting it. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed in this application can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A flood monitoring method, characterized in that, Including: Obtaining digital elevation data, multispectral image data of the flood - affected area, and time - series Sentinel - 1 data within the monitoring time period; wherein, the monitoring time period is the time period from before the flood disaster occurred to after the flood disaster occurred in the flood - affected area; Determining the terrain features of the flood - affected area according to the digital elevation data; and selecting multiple water body sample areas and multiple non - water body sample areas from the multispectral image data as training samples; For each Sentinel - 1 data in the time - series Sentinel - 1 data, using a multi - resolution segmentation algorithm to perform image segmentation on this Sentinel - 1 data, obtaining the corresponding image segmentation result; wherein, the image segmentation result contains the segmented images of each ground object in the Sentinel - 1 data; Extracting the polarization band features, texture features, water body index features, and time - SAR features of this Sentinel - 1 data; Taking the terrain features, the image segmentation result, the polarization band features, the texture features, the water body index features, and the time - SAR features of this Sentinel - 1 data as the input of the initial random forest model corresponding to this Sentinel - 1 data, and using the training samples to perform model training on this initial random forest model, so that the trained random forest model corresponding to this Sentinel - 1 data has the ability to distinguish the water body area and the non - water body area in this Sentinel - 1 data.

2. The method according to claim 1, wherein Before the step of using the multi - resolution segmentation algorithm to perform image segmentation on this Sentinel - 1 data and obtaining the corresponding image segmentation result, it further includes: Performing pre - processing on this Sentinel - 1 data to obtain pre - processed Sentinel - 1 data; wherein, the pre - processing includes: applying an orbit file, removing GRD boundary noise, removing thermal noise, radiometric calibration, filtering, terrain correction, and converting the backscatter intensity to decibels; The step of using the multi - resolution segmentation algorithm to perform image segmentation on this Sentinel - 1 data and obtaining the corresponding image segmentation result includes: Invoking an optimal segmentation scale estimation tool to determine the best segmentation scale parameter, the optimal value of the shape factor, and the compactness parameter of the pre - processed Sentinel - data; Inputting the determined best segmentation scale parameter, the optimal value of the shape factor, and the compactness parameter into the multi - resolution segmentation algorithm, and using the multi - resolution segmentation algorithm to perform image segmentation on the pre - processed Sentinel - 1 data to obtain the image segmentation result.

3. The method according to claim 1, wherein The step of extracting the polarization band features, texture features, water body index features, and time - SAR features of this Sentinel - 1 data includes: Based on the VV polarization and VH polarization of this Sentinel - 1 data, calculating the polarization ratio, polarization difference, polarization sum, and polarization product to obtain the polarization band features of this Sentinel - 1 data; SAR dual-polarization texture features are extracted from the VV polarization and VH polarization of the Sentinel-1 data respectively based on the gray-level co-occurrence matrix, and the texture features of the Sentinel-1 data are obtained; wherein, the texture features are used to reflect the spatial differences of ground objects. Based on the VV polarization and VH polarization of the Sentinel-1 data, a dual-polarization water index is calculated, and the dual-polarization water index is used as the water index feature of the Sentinel-1 data; the water index feature is used to highlight water bodies to distinguish them from other ground objects. According to the backscattering data of the Sentinel-1 data and the time-series Sentinel-1 data, the time SAR features of the Sentinel-1 data are calculated; wherein, the time SAR features include backscattering anomalies and Z-scores; the backscattering anomalies are used to characterize the difference between the backscattering during flood disasters and the average backscattering during non-flood periods; the Z-scores are used to characterize the difference between the backscattering during flood disasters and the average backscattering during the monitoring period.

4. The method according to claim 1, wherein The terrain features include elevation, slope, aspect, and mountain shadow.

5. The method according to claim 1, wherein Using the terrain features, the image segmentation result of the Sentinel-1 data, the polarization band features, the texture features, the water index features, and the time SAR features as the input of the initial random forest model corresponding to the Sentinel-1 data, and using the training samples to train the initial random forest model includes: Inputting the terrain features, the polarization band features, the texture features, the water index features, and the time SAR features of the Sentinel-1 data into the initial random forest model corresponding to the Sentinel-1 data, and determining the importance ranking of each input feature through the initial random forest model. According to the importance ranking of each feature, add one feature to the initial random forest model in order from high to low to determine the classification accuracy of the initial random forest model after each feature is added. According to the change of the classification accuracy of the initial random forest model, determine the features added to the initial random forest model when the classification accuracy is the highest, and determine the features added when the classification accuracy is the highest as the target features corresponding to the Sentinel-1 data. Using the image segmentation result and each of the target features corresponding to the Sentinel-1 data as the input of the initial random forest model corresponding to the Sentinel-1 data, and using the training samples to train the initial random forest model.

6. The method according to claim 5, wherein The method further includes: For each Sentinel-1 data, after obtaining the trained random forest model corresponding to this Sentinel-1 data, fuse the feature values of each of the target features corresponding to this Sentinel-1 data into this Sentinel-1 data to obtain the fused Sentinel-1 data; Input the fused Sentinel-1 data and the image segmentation result corresponding to this Sentinel-1 data into the random forest model, and based on the random forest model, identify the water area and non-water area in the fused Sentinel-1 data.

7. The method according to claim 6, wherein The method further includes: Based on the identified water area and non-water area in the fused Sentinel-1 data, generate a binary classification image for distinguishing the water area and non-water area; After obtaining the binary classification image corresponding to each Sentinel-1 data, for each pixel in the binary classification image, calculate the frequency at which the pixel is identified as water; wherein, the frequency is used to reflect the duration of water at this pixel during the monitoring time period; Select multiple pixels from all the pixels as sample pixels, and determine whether each of the sample pixels is a permanent water pixel or a flood pixel under real circumstances; For each preset frequency threshold, compare the frequency corresponding to each of the sample pixels with the preset frequency threshold to infer whether each of the sample pixels is a permanent water pixel or a flood pixel based on the preset frequency threshold; According to the inferred permanent water pixel or flood pixel for each of the sample pixels and the determined permanent water pixel or flood pixel for each of the sample pixels under real circumstances, calculate the accuracy of the preset frequency threshold for distinguishing permanent water pixels and flood pixels; According to the accuracy corresponding to each preset frequency threshold, screen out the preset frequency threshold with the highest accuracy as the target preset frequency threshold; For each pixel of the binary classification image, compare the frequency of the pixel with the target preset frequency threshold. If the frequency of the pixel is greater than the target preset frequency threshold, it means that the pixel is a permanent water pixel; if the frequency of the pixel is less than or equal to the target preset frequency threshold, it means that there is a flood pixel at this pixel.

8. A flood monitoring device, characterized in that, Includes: An acquisition module, configured to acquire digital elevation data, multispectral image data, and time series Sentinel-1 data during a monitoring time period for the flood-occurring area; wherein, the monitoring time period is the time period from before the flood disaster occurs to after the flood disaster occurs in the flood-occurring area; A determination module, configured to determine the topographic features of the flood-occurring area according to the digital elevation data; and select multiple water sample areas and multiple non-water sample areas from the multispectral image data as training samples; A segmentation module, configured to perform image segmentation on each piece of Sentinel-1 data in the time series Sentinel-1 data by using a multi-resolution segmentation algorithm, so as to obtain a corresponding image segmentation result; wherein, the image segmentation result includes the segmentation images of each ground object in the Sentinel-1 data; An extraction module, configured to extract the polarization band features, texture features, water body index features, and temporal SAR features of this piece of Sentinel-1 data; A training module, configured to use the terrain features, the image segmentation result of this piece of Sentinel-1 data, the polarization band features, the texture features, the water body index features, and the temporal SAR features as the input of the initial random forest model corresponding to this piece of Sentinel-1 data, and use the training samples to perform model training on the initial random forest model, so that the trained random forest model corresponding to this piece of Sentinel-1 data has the ability to distinguish the water body area and the non-water body area in this piece of Sentinel-1 data.

9. An electronic device, characterized in that, Comprising: A processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, the steps of the method according to any one of claims 1 to 7 are executed.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium. When the computer program is run by a processor, the steps of the method according to any one of claims 1 to 7 are executed.

Images