Floodwater extraction method and system based on time-series spaceborne SAR images

Abstract

The application provides a floodwater extraction method and system based on time-series spaceborne SAR images, radar image preprocessing is carried out, including coarse registration of the image by using spaceborne SAR platform parameters; registration of time-series spaceborne SAR images; construction of an abnormal value detection factor based on statistical characteristics, the abnormal value detection factor adopts a time-series SAR image Z-score value in a comprehensive polarization mode, orbit parameters and imaging mode difference; automatic image block screening based on GPM and Z-score, including obtaining a completely rain-free baseline CRB image set based on GPM, then obtaining a rain-affected baseline RRB based on Z-score and CRB combination, and realizing automatic screening of sequence image blocks by using RRB; large-scale flood automatic extraction based on a classification tree, including dividing urban areas and non-urban areas in the image by using world settlement footprint WSF data, and extracting urban waterlogging and natural area flood in a decision tree mode based on the difference between building area and natural area flood period scattering mechanism.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of polarimetric radar remote sensing image processing, and is a new technical scheme for realizing accurate detection of urban flood water based on time-series spaceborne SAR (Synthetic Aperture Radar) images. BACKGROUND

[0002] Flood is one of the natural disasters with the highest occurrence frequency, the widest range of influence and the most severe economic losses in the world.

[0003] Extreme rainfall is the main factor causing flood. However, due to cloud cover, it is difficult for traditional visible light and infrared band remote sensing satellites to obtain the actual situation of the ground, so it is difficult to draw a complete flood range in time, and the losses caused by flood disasters are often underestimated. Unlike optical remote sensing, synthetic aperture radar (SAR) is an active microwave remote sensing method, which uses the microwave band to ensure that it can obtain ground information under any weather conditions, and has all-weather and all-day imaging capability. SAR satellites play an indispensable role in emergency response and are an important part of the global earth observation satellite system. The Gaofen-3 satellite of China has obtained multiple scenes of emergency observation images during heavy rain, which shows that SAR has become one of the preferred tools for flood monitoring. In addition, the rapid development of commercial SAR satellites has brought different resolutions, different bands and polarization modes of data for flood monitoring, greatly improving the ability of SAR images to draw dynamic changes of the ground in emergency response.

[0004] The backscattering characteristics of different land cover types in SAR images are significantly different, and the intensity changes of natural areas and building areas are completely different when flood occurs. The traditional methods based on threshold or change detection have shortcomings in dealing with complex scene flood monitoring with multiple types of mixed distribution. The reason is that: threshold segmentation mainly uses single time phase SAR images during flood disasters, and uses global or local spatial statistical information to determine the threshold, but the differences in scattering behavior of different ground targets and the differences in scattering changes caused by flood in different areas will lead to the inability to determine a universal threshold for complex scenes; change detection usually compares the differences between time phase images before and after the occurrence of flood disasters, but the high time variability of natural areas, the strong heterogeneity of building areas and the inevitable multiplicative noise of SAR and other factors are coupled, which aggravates the generation of false changes. Based on the defects of the prior art, the application proposes to use time-series spaceborne SAR images to provide more detailed observation information in the time dimension, which can reduce the interference of time variability and other factors, and can determine the existence of flood based on the statistical information in the time dimension, which can avoid the problem of universality of spatial statistical information. SUMMARY

[0005] The present application aims at solving the problem that the current flood water body change detection scheme based on spaceborne SAR images cannot fully utilize the time dimension information between time series images, and each pixel of the time series images is usually detected without distinction, resulting in high false detection rate.

[0006] The technical scheme adopted by the present application provides a flood water body extraction method based on time series spaceborne SAR images, including the following processing:

[0007] Radar image preprocessing, including coarse registration of the image using spaceborne SAR platform parameters;

[0008] Registration of time series spaceborne SAR images;

[0009] Construction of an abnormal value detection factor based on statistical characteristics, which uses a comprehensive polarization method, orbit parameters and imaging mode difference Z-score value of time series SAR images;

[0010] Automatic image block screening based on GPM and Z-score, including obtaining a completely rain-free baseline CRB image set based on GPM, then obtaining a rain-affected baseline RRB based on Z-score and CRB, and using RRB to realize automatic screening of sequence image blocks;

[0011] Large-scale flood automatic extraction based on classification trees, including dividing urban and non-urban areas in the image using world settlement footprint WSF data, and using decision tree method to extract building area waterlogging and natural area flood based on the difference in scattering mechanism between building area and natural area during flood period.

[0012] Moreover, the radar image preprocessing includes reading head information, orbit interpolation, coarse registration, thermal noise removal, terrain radiation correction and delicate Lee filtering.

[0013] Moreover, the registration of time series spaceborne SAR images is realized by selecting one SAR image as the source image and the other images as the target images, extracting the accurate registration points of the target images and the source images in the set window based on the correlation coefficient, and using the quadratic polynomial to perform geometric rectification of the target images to realize the relatively accurate registration of the images.

[0014] Moreover, in the automatic image block screening based on GPM and Z-score, the frequency distribution curve of the Z-score value of the CRB slice set is calculated based on the CRB slice set as a reference; the Z-score value of the comprehensive difference score is recalculated according to the mean and standard deviation of the CRB slice set, and the frequency distribution curve of the Z-score value is counted; the difference degree between the two curves is judged based on the curve similarity of the Manhattan distance, and when the similarity meets the threshold condition, the rule set satisfied by the image at this time is recorded, and a new rain-free baseline image set is obtained by using the obtained new rule set.

[0015] Moreover, the large-scale flood automatic extraction based on the classification tree is implemented as follows,

[0016] The building mask data set is used to distinguish the building area and the natural area.

[0017] In the natural area, whether a pixel is a permanent water body is judged according to the water body coverage time of the pixel, the permanent water body pixel no longer participates in subsequent classification, and for the non-permanent water body, if the Z-score values of the co-polarization and cross-polarization of a given pixel are both lower than the corresponding threshold value, the pixel is assigned a dual-polarization flood label; if only one is lower than the corresponding threshold value, the pixel is assigned a single-polarization flood label.

[0018] In the building area, if the Z-score values of the co-polarization and cross-polarization of a given pixel are both greater than the corresponding threshold value, the pixel is assigned a dual-polarization flood label; if only one is greater than the threshold value, the pixel is assigned a single-polarization flood label.

[0019] On the other hand, the application also provides a flood water body extraction system based on time-series spaceborne SAR images, which is used to implement the flood water body extraction method based on time-series spaceborne SAR images as described above.

[0020] Moreover, the system comprises the following modules,

[0021] The first module is used for radar image preprocessing, including using the spaceborne SAR platform parameters to realize coarse registration of the image;

[0022] The second module is used for registration of time-series spaceborne SAR images;

[0023] The third module is used for constructing an abnormal value detection factor based on statistical characteristics, and the abnormal value detection factor adopts a Z-score value of a time-series SAR image in a comprehensive polarization mode, an orbit parameter and an imaging mode difference;

[0024] The fourth module is used for automatic image block screening based on GPM and Z-score, including obtaining a completely rain-free baseline CRB image set based on GPM, then obtaining a rain-free baseline RRB not affected by rainfall based on Z-score and CRB, and realizing automatic screening of the sequence image block by using the RRB.

[0025] The fifth module is used for wide range flood automatic extraction based on classification tree, including image urban and non-urban division by using world settlement footprint (WSF) data, and flood extraction in building area and natural area by using decision tree method based on the difference of scattering mechanism in flood period.

[0026] Alternatively, the method comprises a processor and a memory, the memory is used for storing program instructions, and the processor is used for calling the stored instructions in the memory to execute the flood water extraction method based on time-series spaceborne SAR images.

[0027] Alternatively, the method comprises a readable storage medium, and the readable storage medium stores a computer program, and the computer program is executed to realize the flood water extraction method based on time-series spaceborne SAR images.

[0028] The innovation of the present application lies in that:

[0029] 1) Based on time-series SAR images, a statistical outlier detection scheme considering the difference of polarization mode, orbit parameter and imaging mode is proposed;

[0030] 2) Based on time-series SAR images, an automatic baseline image screening algorithm combined with GPM is proposed;

[0031] 3) Based on the difference of scattering mechanism in flood period in building area and natural area, a scheme for simultaneously extracting building area waterlogging and natural area flood is invented.

[0032] The present application is based on the all-weather and all-day imaging capability of synthetic aperture radar, uses time-series SAR images to provide detailed time dimension observation information, enhances the feature description capability of high time-varying natural targets, proposes a scheme based on time-series statistical information, reduces the interference of high spatial heterogeneity and other factors in complex scenes, and obtains higher precision of flood monitoring results. Since the synthetic aperture radar can perform ground surface observation under all-weather and all-day conditions, the flood monitoring results can provide sufficient guarantee for disaster emergency response, and meet the precision and timeliness requirements of disaster emergency response. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flowchart of an embodiment of the present application.

[0034] Figure 2 is an illustration of screening each SAR image without strict rainfall screening rules, and leaving the image blocks meeting the rules, wherein the window size is GPM pixel size.

[0035] Figure 3is a schematic diagram of baseline generation based on a Z-score value rule set according to an embodiment of the application.

[0036] Figure 4 is a schematic diagram of a flood extraction process based on a classification tree according to an embodiment of the application. DETAILED DESCRIPTION

[0037] The technical solutions of the application will be described in detail below with reference to the accompanying drawings and embodiments.

[0038] The application utilizes an outlier detection scheme, fully utilizes the time dimension statistical information of the historical accumulated time series SAR image, and combines the GPM multi-satellite integrated retrieval (IMERG) precipitation product corrected by ground station data and the proposed Z-score value to remove unreliable areas in the image. By analyzing the change of SAR information scattering mechanism before and after the flood, the world settlement footprint (WSF) is used to distinguish urban and non-urban areas, and then the outlier detection and classification tree method is used to realize high-precision extraction of water body change information.

[0039] The technical solutions of the application can be realized by computer technology to automatically run, as shown in the figure, Figure 1 The flood water extraction method based on time series spaceborne SAR images provided by the embodiment of the application includes the following steps:

[0040] Step 1, radar image preprocessing: the application proposes to use spaceborne SAR platform parameters to realize image coarse registration, which ensures that the computer can completely automatically complete the processing process,

[0041] Further, the preferred radar image preprocessing specifically includes the radar image preprocessing, including reading header information, orbit interpolation, coarse registration, thermal noise removal, terrain radiation correction and delicate Lee filtering.

[0042] The radar image preprocessing process of step 1 in the embodiment includes the following sub-steps:

[0043] Step 1.1 reads the SAR satellite image header information to obtain the distance digital-to-analog (A / D) sampling rate, pulse repetition time, imaging time, wavelength, imaging center point slant range, satellite orbit position, satellite speed information, and Doppler frequency;

[0044] Step 1.2 uses a quadratic polynomial to interpolate the satellite orbit and satellite speed information to obtain the satellite position and speed information corresponding to each row of the image;

[0045] Step 1.3 uses the range-Doppler model (equation 2) to solve the spaceborne SAR image center point coordinates;

[0046]

[0047] In the formula f D Let λ be the Doppler frequency, λ be the radar wavelength, and R be the slant range corresponding to the image point being solved. t ,y t ,z t ( ) represents the coordinates of the image point to be determined. and These are the position vectors of the satellite and the target, respectively. and These are the velocity vectors of the satellite and the target, respectively. and These are the satellite's major and minor axes, respectively.

[0048] Step 1.4 converts the center point coordinate difference of each image into image coordinate difference to achieve coarse image registration;

[0049] Step 1.5 Calculate the corresponding thermal noise value for each pixel based on the thermal noise estimate provided in the SAR image metadata, and subtract the thermal noise value to complete the thermal noise removal.

[0050] Step 1.6 uses the sensor calibration parameter metadata built into the image to calculate the pixel values ​​into backscatter intensity with physical meaning.

[0051] Step 1.7 Use DEM to correct the effect of surface undulation on backscattering intensity, obtain the normalized backscattering coefficient, and realize topographic correction radiation.

[0052] Step 1.8 Based on refined Lee filtering, set the filtering window (the window size is generally between 3 and 7) to perform filtering and noise reduction processing on the spaceborne SAR image.

[0053] Step 2, Registration of Time-Series Spaceborne SAR Images:

[0054] In this embodiment, step 2, the registration process of the temporal SAR image, includes the following sub-steps:

[0055] Step 2.1 Select one SAR image as the source image and other images as target images. Based on the correlation coefficient (Equation 3), extract the accurate registration points between the other images (target images) and the source image within a set window (e.g., 30*30).

[0056]

[0057] Where i and j are the image row and column numbers; m and n are the width and height of the corresponding window.

[0058] Step 2.2. Based on the extracted registration points, geometric correction of the target image is performed using a quadratic polynomial to achieve relative accurate registration of the images.

[0059] Step 3. Constructing an abnormal value detection factor based on statistical characteristics: the present application proposes a time-series SAR image Z-score value considering the differences in polarization modes, orbital parameters and imaging modes as a detection factor.

[0060] In the embodiment, step 3 performs abnormal value detection based on the statistical characteristics of the registered time-series SAR images, and the specific process is as follows:

[0061] Step 3.1. Using the registered sequence SAR images, one SAR image in the flood period is removed, and then

[0062] Formula 4 and formula 5 are used to calculate the mean value mean() and the variance std().

[0063]

[0064]

[0065] where X i represents the normalized backscatter value of the sequence SAR image at each time i represents the mean value, and n is the number of images, wherein p represents the VV or VH polarization mode, o represents the orbital number, m represents the imaging mode, and pre represents the backscatter value of the SAR image acquired before the flood, and then represents the normalized backscatter coefficient of the corresponding (p, o, m) acquired before the flood.

[0066] Step 3.2. According to the mean value map, the standard deviation map and the SAR image in the flood period, formula 6 is used to calculate the comprehensive difference score value Z-score considering the differences in polarization modes, orbital parameters and imaging modes:

[0067]

[0068] wherein p represents the VV or VH polarization mode, o represents the orbital number, m represents the imaging mode, flood represents the image backscatter acquired during the flood, pre represents the backscatter value of the SAR image acquired before the flood, and sigma is the normalized backscatter coefficient, and then represents the normalized backscatter coefficient of the corresponding (p, o, m) acquired during the flood, mean() is the mean value calculation, and std() is the standard deviation calculation.

[0069] Step 4, GPM and Z-score based automatic image block screening: first, get the completely rain-free baseline (CRB) image set based on GPM, then get the rule set based on Z-score and CRB joint statistical analysis, based on the new rule set, form the rain-free baseline (RRB), and use RRB to realize the sequence image block screening.

[0070] The specific process of step 4 in the embodiment based on GPM and Z-score automatic image screening is as follows:

[0071] Step 4.1 extracts the time and range of each time series SAR image, extracts the corresponding GPM image according to the time and range to form a GPM image set (for example, including N GPM images), blocks the sequence SAR image according to the GPM pixel size, forms a SAR image mask based on the strict rainfall screening rule set (for example, rules 1, 2, 3, etc., Table 1), and uses the inundation map to inundate the SAR image to form a new sequence SAR image set, called the completely rain-free baseline (CRB) image set, the flow is shown in FIG. 1. Figure 2 .

[0072] The amount of rainfall will change the backscattering intensity of the building area to varying degrees, which can be quantified by the difference in its statistical characteristics. The present application proposes a strict rainfall screening rule set (rules 1, 2, 3, etc.), in which each rainfall level and the number of days are combined as a rule. Here is an example to describe the screening process: if the rainfall in a certain area on August 1st is 80mm, the rainfall level is heavy rain, and the number of days to be excluded is 24 days, then the Sentienl-1 image pixels of the area imaging date from August 1st to August 24th need to be inundated, and if 20mm of rainfall occurs on August 23rd, the screening date is delayed to August 28th. Combined with the rule set, the CRB image set is obtained by using the GPM based automatic image screening algorithm in this paper.

[0073] Table 1 Strict rainfall screening rule set

[0074]

[0075]

[0076] Step 4.2, taking the CRB slice set as the reference, calculate the frequency distribution curve of the Z-score value of the CRB slice set;

[0077] Step 4.3, calculate the mean and standard deviation of the CRB slice set, add a SAR image in the sequence to recalculate the mean and variance of each point according to formula 5, and correspondingly recalculate the comprehensive difference score Z-score value, and calculate the frequency distribution curve of the Z-score value;

[0078] Step 4.4. Determine the difference between two curves based on Manhattan distance-based curve similarity. The Manhattan distance-based curve similarity is measured by calculating the straight-line distance between the same x-axis points on the curve, i.e., measuring the area between the two curves. The formula is as shown in Equation 7

[0079] D = ∑|y1-y2| (7)

[0080] where y1 and y2 represent the corresponding y-coordinate values when the x-coordinates are the same.

[0081] Determine the similarity between the Z-score frequency distribution curves of the image pixels based on Manhattan distance-based curve similarity. When the similarity meets the threshold condition, record the rule set satisfied by the image at that time. The parameters are a series of rainfall grades and corresponding interval days.

[0082] Step 4.5. Repeat step 4.4 to obtain the rule set. The flowchart is shown in Figure 4. Figure 3 .

[0083] Step 4.6. Use the new rule set obtained to repeat step 4.1 to obtain a new set of rain-free baseline (RRB) images.

[0084] Step 5. Large-scale flood automatic extraction based on classification tree: The present application uses the world settlement footprint (WSF) data to divide the urban and non-urban areas in the image. Based on the differences in scattering mechanisms between building areas and natural areas during flood periods, a decision tree method is proposed to extract waterlogging in building areas and floods in natural areas.

[0085] The specific process of step 5 in the embodiment based on classification tree large-scale flood automatic extraction is as follows (taking dual-polarization SAR image as an example):

[0086] Step 5.1. Use the building inundation data set (take urban WSF = 255 in the embodiment) to distinguish between building areas and natural areas.

[0087] Step 5.2. In the natural area, determine whether a pixel is a permanent water body based on its water body coverage time. Determine whether a pixel is a seasonal water body based on its water body coverage time. The criterion is occurrence greater than 25, which means that the pixel is a water body in 25% of the satellite images within a year. Permanent water body pixels are no longer involved in subsequent classification. For non-permanent water bodies, if the Z-score values of the given pixel for the same polarization and cross-polarization, Z VV , VH , are both lower than the threshold value th N (recommended value is 3 times the standard deviation, denoted as 3*std()), the pixel is assigned a dual-polarization flood label; if only one is lower than the threshold value th N (3*std()), it is assigned a single-polarization flood label;

[0088] Step 5.3 In the building area, if the Z-score values Z VV 、Z VH of the given pixel of the same polarization and cross-polarization are both greater than the threshold th s (3*std()) (also preferably recommended to be 3*std()), the pixel is assigned a dual-polarization flood label; if only one is greater than the threshold th s (3*std()), it is assigned a single-polarization flood label. The flowchart is shown in FIG. 4. In specific implementation, preferably, the last surface slope information obtained based on the digital elevation model (DEM) is used to further screen and remove the flood pixels with a slope greater than the corresponding threshold, which can reduce the risk of false detection caused by the superposition effect in mountainous areas. Figure 4

[0089] In specific implementation, the method proposed in the technical solution of the present application can be automatically run by a person skilled in the art using computer software technology, and the system device of the method, such as a computer readable storage medium storing the corresponding computer program of the technical solution of the present application and a computer device including running the corresponding computer program, should also be within the protection scope of the present application.

[0090] In some possible embodiments, a flood water extraction system based on time-series spaceborne SAR images is provided, including the following modules,

[0091] The first module is used for radar image preprocessing, including using the spaceborne SAR platform parameters to realize coarse registration of the images;

[0092] The second module is used for registration of time-series spaceborne SAR images;

[0093] The third module is used for constructing an anomaly value detection factor based on statistical characteristics, and the anomaly value detection factor uses a time-series SAR image Z-score value that comprehensively considers polarization, orbit parameters and imaging mode differences;

[0094] The fourth module is used for automatic image block screening based on GPM and Z-score, including first obtaining a completely rain-free baseline CRB image set based on GPM, then obtaining a baseline RRB that is not affected by rainfall based on Z-score and CRB, and using RRB to realize automatic screening of the sequence image blocks;

[0095] The fifth module is used for large-scale flood automatic extraction based on classification trees, including using world settlement footprint WSF data to realize division of urban and non-urban areas in the images, and based on the differences in flood period scattering mechanisms between building areas and natural areas, using decision tree method to extract urban waterlogging in building areas and natural area floods.

[0096] ​In some possible embodiments, a system for extracting flood water body based on time-series spaceborne SAR images is provided, which comprises a processor and a memory. The memory is configured to store program instructions, and the processor is configured to invoke the program instructions stored in the memory to execute a method for extracting flood water body based on time-series spaceborne SAR images.

[0097] In some possible embodiments, a system for extracting flood water body based on time-series spaceborne SAR images is provided, which comprises a readable storage medium. The readable storage medium has a computer program stored thereon. The computer program, when executed, implements a method for extracting flood water body based on time-series spaceborne SAR images.

[0098] The specific embodiments described herein are merely illustrative of the spirit of the present application. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them. For example, the template pre-processing window size can be set to different values; the calculation of Z value can only consider single polarization; the curve similarity discrimination algorithm can select other methods instead of Manhattan distance, and so on. However, this will not exceed the framework of the algorithm proposed in the present application, will not deviate from the spirit of the present application, or will not exceed the scope defined by the appended claims.

Claims

1. A method for extracting flood water body based on time-series space-borne SAR images, characterized in that, The method comprises the following processes: Radar image preprocessing, including coarse registration of the image by using the parameters of the space-borne SAR platform; Registration of the time-series space-borne SAR image; the registration of the time-series space-borne SAR image is achieved by selecting one SAR image as a source image and other images as target images, extracting the accurate registration points of the target images and the source image in a set window based on the correlation coefficient, and performing geometric correction of the target images by using a quadratic polynomial based on the extracted registration points to achieve the relatively accurate registration of the images; An abnormal value detection factor based on statistical characteristics is constructed, the abnormal value detection factor adopts a time-series SAR image Z-score value of a comprehensive polarization mode, orbit parameters and imaging mode difference; Automatic image block screening based on GPM and Z-score, including obtaining a completely rain-free baseline CRB image set based on GPM, obtaining a rain-affected baseline RRB based on Z-score and CRB, and realizing automatic screening of the sequence image block by using the RRB; The implementation is to calculate the frequency distribution curve of the Z-score value of the CRB slice set based on the CRB slice set; the Z-score value of the comprehensive difference score is recalculated according to the mean and standard deviation of the CRB slice set, the frequency distribution curve of the Z-score value is counted; the difference degree between the two curves is judged based on the Manhattan distance curve similarity, when the similarity meets the threshold condition, the rule set satisfied by the image at this time is recorded, and a new rain-free baseline image set is obtained by using the obtained new rule set; Large-scale flood automatic extraction based on classification tree, including dividing the urban and non-urban areas in the image by using the world settlement footprint WSF data, and performing decision tree extraction of the building area waterlogging and the natural area flood based on the difference between the building area and the natural area in the scattering mechanism during the flood period. 2.The floodwater extraction method based on time-series spaceborne SAR images according to claim 1, characterized in that: The radar image preprocessing is implemented by reading the head information, orbit interpolation, coarse registration, thermal noise removal, terrain radiation correction and fine Lee filtering.

3. The method according to claim 1 or 2, characterized in that: The large-scale flood automatic extraction based on classification tree is implemented as follows, The building mask data set is used to distinguish the building area and the natural area; In the natural area, whether a pixel is a permanent water body is determined according to the water body coverage time of the pixel, the permanent water body pixel no longer participates in the subsequent classification, and for the non-permanent water body, if the Z-score values of the co-polarization and cross-polarization of a given pixel are both lower than the corresponding threshold value, the pixel is assigned a dual-polarization flood label; if only one is lower than the corresponding threshold value, the pixel is assigned a single-polarization flood label; In the building area, if the Z-score values of the co-polarization and cross-polarization of a given pixel are both greater than the corresponding threshold value, the pixel is assigned a dual-polarization flood label; If only one is greater than the threshold value, the pixel is assigned a single-polarization flood label.

4. A system for extracting flood water body based on time-series space-borne SAR images, characterized in that: A flood water body extraction method based on time-series space-borne SAR images is used to implement any one of claims 1-3.

5. The system according to claim 4, wherein the system further comprises: The method comprises the following modules, A first module is used for radar image preprocessing, including coarse registration of the image by using the parameters of the space-borne SAR platform; A second module is used for registration of the time-series space-borne SAR image; The third module is configured to construct an abnormal value detection factor based on statistical characteristics, and the abnormal value detection factor adopts a time-series SAR image Z-score value of a comprehensive polarization mode, an orbit parameter, and an imaging mode difference; The fourth module is configured to perform automatic image block screening based on GPM and Z-score, including obtaining a completely rain-free baseline CRB image set based on GPM, and then obtaining a rain-affected baseline RRB based on Z-score and CRB, and realizing automatic screening of a sequence image block by using the RRB; The fifth module is configured to perform large-scale flood automatic extraction based on a classification tree, including dividing urban areas and non-urban areas in an image by using world settlement footprint WSF data, and performing decision tree extraction of building area waterlogging and natural area flood based on the difference between building area and natural area flood scattering mechanisms in a flood period.

6. The system according to claim 4, wherein the system further comprises: The device comprises a processor and a memory, the memory is configured to store program instructions, and the processor is configured to call the stored instructions in the memory to execute the flood water body extraction method based on time-series satellite SAR images.

7. The system according to claim 4, wherein the system further comprises: The device comprises a readable storage medium, and the readable storage medium stores a computer program, and the computer program is configured to execute the flood water body extraction method based on time-series satellite SAR images.

Images