CN116303662A - Flood automatic detection and dynamic monitoring method and device - Google Patents

Abstract

The embodiment of the specification provides a method and a device for automatically detecting and dynamically monitoring flood, wherein the method comprises the following steps: preprocessing earth observation data based on GEE; carrying out water space-time distribution coarse extraction on the pretreated earth observation data; performing first-stage water body fine extraction based on water body space-time distribution coarse extraction results; and carrying out flood event detection and inundation analysis based on the primary water body fine extraction result.

Description

Flood automatic detection and dynamic monitoring method and device

technical field

This document relates to the field of computer technology, in particular to a flood automatic detection and dynamic monitoring method and device.

Background technique

Floods are one of the most destructive natural disasters in the world, killing more than 500,000 people worldwide and causing more than $3 billion in economic losses in the past 30 years. With the continuous acceleration of global warming and the continuous deepening of human activities, the probability and intensity of extreme flood events continue to rise. While the risk of floods has risen sharply, the uncertainty of time and space of floods has increased, and its impact is becoming more serious. Difficult to estimate. Accurately sorting out historical flood changes is essential for mastering flood dynamics, focusing on large-scale, long-term, and dynamic monitoring targets to develop reliable and effective methods for accurate collection and feedback of spatiotemporal information on flood disasters, and to carry out work such as identification of spatiotemporal dynamic features of flood disasters and their impacts It is of great significance to improve flood control and disaster reduction capabilities and reduce flood losses, especially for strengthening regional food security, building a solid economic and social cornerstone, and optimizing the ecological security barrier system. It is of strategic support.

Existing methods for research on flood impact, causes, and prevention and control are mostly limited by observational data, mainly using statistical data, or relying on models with large uncertainties for analysis and research, and only select a small part of the study area. A detailed analysis of the region fails to analyze from a larger spatial scale. However, the long-sequence, high-spatial coverage remote sensing data products used in this project have natural advantages in the extraction of submerged areas of surface water bodies, and have great potential in the study of global floods. Satellite remote sensing observation technology has been widely used in various disaster monitoring due to its advantages of wide observation range, periodic revisit, and continuous space coverage. For remote sensing monitoring of flood disasters, it is necessary to take into account the particularity of flood disasters, such as wide spatial and temporal distribution, process dynamic changes, and often accompanied by heavy rainfall, thick cloud weather, etc. Synthetic aperture radar (SAR) with all-weather and all-weather observation capabilities Aperture Radar) images are more suitable for flood dynamic monitoring than optical remote sensing images limited by cloudy and rainy weather. The Sentinel-1 satellite consists of A and B double stars in the same orbit, and is equipped with a C-band SAR sensor, forming a global 12d revisit cycle observation capability. Sentinel-1 data has a wide coverage, high spatial resolution, and is free and open to the world, bringing more opportunities and challenges for regular, accurate and systematic flood remote sensing monitoring.

Synthetic Aperture Radar (SAR) has become one of the indispensable data sources for flood dynamic monitoring under the limitation of cloudy and rainy weather due to its all-weather and all-weather advantages in earth observation. Due to the rise of the GEE (Google Earth Engine) cloud computing platform and the availability of short-revisit Sentinel-1 data, flood monitoring and disaster assessment are currently developing towards dynamic and wide-area development. Considering the complexity of land cover change in the flooded area and the uncertainty of occurrence time, based on the time-series Sentinel-1A satellite data, an automatic detection and dynamic monitoring method for large-scale, continuous and long-term flood conditions is proposed. First, the method uses image binarization to segment time-series SAR data to realize a rough map of water body spatio-temporal distribution, and calculates the frequency of water body candidate points in the time series pixel by pixel. Then, the Sentinel-2 optical image was used to correct the rough water body extraction results of the initial SAR to obtain a fine water body distribution map. Finally, according to the inundation characteristics of different frequency intervals, a differentiated timing anomaly detection strategy is used to identify the submerged range: for low-frequency water-covered areas, Euclidean distance is used to detect timing breakpoints to extract flood disaster areas with large disturbance intensity and short submersion time.

GEE is a cloud computing platform designed to store, process, analyze and visualize geospatial data. The emergence of this platform has changed the processing and analysis mode of remote sensing big data, and is widely used in the field of environmental remote sensing such as time series analysis or large-scale mapping. The advantage of GEE is that it provides: (1) massive remote sensing open data sets, (2) numerous remote sensing image processing algorithms, (3) powerful data cloud computing capabilities, (4) support for general programming languages (JavaScript, Python) . GEE can greatly improve the efficiency of time-series SAR data processing and analysis, and provides great opportunities for wide-area, dynamic, and long-term flood monitoring research.

In the existing technology, (1) how to improve the accuracy of water body monitoring, avoid the interference of rainy weather and clouds during flood periods, how to realize the automatic detection and dynamic extraction of large-scale and multi-period flood submerged areas, and how to evaluate the impact of floods on cities has become an urgent task. technical issues to be resolved.

Contents of the invention

The object of the present invention is to provide a flood automatic detection and dynamic monitoring method and device, aiming to solve the above-mentioned problems in the prior art.

The invention provides a flood automatic detection and dynamic monitoring method, comprising:

Earth observation data preprocessing based on GEE;

Coarsely extract the spatio-temporal distribution of water bodies from the preprocessed earth observation data;

Based on the rough extraction results of the spatio-temporal distribution of the water body, the first phase of fine water extraction is carried out;

Flood event detection and inundation analysis are carried out based on the results of the first phase of fine water extraction.

The invention provides a flood automatic detection and dynamic monitoring device, comprising:

The preprocessing module is used for preprocessing the earth observation data based on GEE;

The rough extraction module is used to perform rough extraction of the spatio-temporal distribution of water bodies on the preprocessed earth observation data;

The fine extraction module is used for the first stage of fine water extraction based on the rough extraction results of the spatiotemporal distribution of water;

The analysis module is used for flood event detection and inundation analysis based on the fine extraction results of the first phase of water.

The embodiment of the present invention highlights the advantages of the cloud computing platform (GEE) for long-term dynamic monitoring of large-scale floods, and can efficiently process and analyze time-series data to support continuous monitoring of floods in a large area during the flood season. Realized the dynamic monitoring of the global flood situation within the theoretical scope, accurately extracted the spatio-temporal distribution of the flood inundation range in the required areas since 2015, and revealed the differences in the development patterns of flood conditions in different regions. Creatively integrating the vulnerability analysis into the assessment of the city's ability to suffer from floods, the vulnerability assessment can be performed while looking at the flooded areas, and then guidance suggestions can be given for local flood control construction.

Description of drawings

In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, in the following description The accompanying drawings are only some embodiments described in this specification, and those skilled in the art can also obtain other drawings according to these drawings without any creative work.

Fig. 1 is the flow chart of the flood automatic detection and dynamic monitoring method of the embodiment of the present invention;

Fig. 2 is the schematic diagram of the basic parameter of SENTINEL-1A data of the embodiment of the present invention;

Fig. 3 is the schematic diagram of the SENTINEL2 product band information of the embodiment of the present invention;

Fig. 4 is the detailed processing flowchart of the flood automatic detection and dynamic monitoring method of the embodiment of the present invention;

Fig. 5 is a schematic diagram of an automatic flood detection and dynamic monitoring device according to an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will describe the technical solutions in one or more embodiments of this specification in conjunction with the drawings in one or more embodiments of this specification The technical solution is clearly and completely described, and obviously, the described embodiments are only a part of the embodiments in this specification, rather than all the embodiments. Based on one or more embodiments in this specification, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the scope of protection of this document.

method embodiment

According to an embodiment of the present invention, a flood automatic detection and dynamic monitoring method is provided, and Fig. 1 is a flow chart of the flood automatic detection and dynamic monitoring method according to an embodiment of the present invention. As shown in Fig. 1, the flood according to the embodiment of the present invention Automatic detection and dynamic monitoring methods specifically include:

Step 101, perform earth observation data preprocessing based on GEE; specifically include: determine data screening rules for monitoring requirements, group time series data based on the data screening rules, filter, mosaic and crop using median filter operators, construct A time-series SAR dataset covering the entire study area.

Step 102, perform rough extraction of temporal and spatial distribution of water bodies on the preprocessed earth observation data; specifically include: refer to the maximum inter-class variance method Otsu to determine the SAR image water and land segmentation threshold in the time series SAR data set, and based on the water and land segmentation threshold, use the slope Remove the virtual detection water body in the shaded area with the HAND data, realize the rough extraction of time series information of the water body, and obtain the rough extraction result of the spatio-temporal distribution of the water body.

Step 103, based on the rough extraction results of the spatio-temporal distribution of the water body, the fine extraction of the first phase of the water body is carried out; specifically includes: combined with the Sentinel-2 optical data, the normalized water index NDWI is used to correct the rough extraction results of the spatio-temporal distribution of the water body, and the water body information of the initial phase is accurately extracted to obtain The results of the first phase of fine water extraction.

In step 104, flood event detection and inundation analysis are performed based on the results of the first phase of fine water body extraction. Specifically include:

Based on the spatio-temporal distribution data of the water body in the first phase of water body fine extraction results, the frequency of each pixel being identified as a water body is calculated one by one, and different flood event detection strategies are constructed based on different frequency intervals, and the corresponding flood event detection strategies are used to extract short-term submergence. area and continuous inundation area, and analyze the temporal change characteristics of the inundation area.

The technical solutions of the embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

Data source: Sentinel-1 is a new generation of dual-polarization C-band spaceborne SAR system developed by ESA’s Copernicus global earth observation mission. Sentinel-1A launched in 2014-04 and Sentinel-1B launched in 2016-04 composition. GEE stores Sentinel-1GRD data in interferometric wide mode (IW), extra wide mode (EW) and strip mode (SM), and updates and releases the latest production data daily. Before the release of each phase of data, the ESA SNAP software package was used for preprocessing including orbit file import, thermal noise removal, radiometric calibration, and orthorectification (Tiwari et al., 2020). The coverage of Sentinel-1B satellite in the middle and lower reaches of the Yangtze River is less than 10% and the strips are discontinuous, while the Sentinel-1A data covers the entire range and forms a 12d fixed cycle data accumulation, so the Sentinel-1A data is used, and the Sentinel-1A The main parameters of the data are shown in Figure 2. The Sentinel-1A GRD data covering the middle and lower reaches of the Yangtze River involve 8 relative orbit numbers.

As shown in Figure 3, Sentinel-2 is a wide-scan, medium-high-resolution, multi-spectral imaging satellite of the ESA Copernicus program, including two satellites 2A and 2B. The Sentinel-2 satellite carries the Multispectral Instrument (MSI) covering 13 spectral bands with ground resolutions of 10m, 20m and 60m. GEE supports the processing and analysis of Sentinel-2Level-2A surface reflectance products (Sen2Cor correction). Based on this data, the flood submersion range is analyzed, so the Sentinel-2 data observation time used should be close to the Sentinel-1A data observation time of the T1 period. In order to obtain full-coverage optical images, images taken in different periods need to be stitched together. At the same time, taking into account the possible impact of cloud cover, the shooting time of the required Sentinel-2 images is extended to a period when the water body is relatively stable.

The technical solution of the embodiment of the present invention uses a digital elevation model (DEM) and a hydrological model to remove misidentification caused by mountain shadows. The DEM uses 30m spatial resolution data from the US Space Shuttle Radar Topography Mission SRTM (ShuttleRadar Topography Mission). The hydrological model uses the watershed relative elevation model HAND (HeightAbove the Nearest Drainage).

The sudden change of land cover type caused by flooding makes the radar backscatter coefficient of surface scatterers fluctuate abnormally in time series. Therefore, based on the timing anomaly detection theory, an embodiment of the present invention proposes a flood automatic detection and dynamic monitoring method. As shown in Figure 4, the main contents of this method include: 1. GEE-based data preprocessing; 2. Rough extraction of temporal and spatial distribution of water bodies; 3. Fine extraction of first-phase water bodies; 4. Flood event detection and inundation analysis. In the data preprocessing stage, the data screening rules are determined according to the monitoring requirements, and then the time series data are grouped, filtered, mosaicked and clipped to construct a time series SAR data set with full coverage of the study area; in the stage of rough extraction of the temporal and spatial distribution of water bodies, refer to the maximum inter-class variance method ( Otsu) determined the SAR image water and land segmentation threshold, and used the slope and HAND data to remove the falsely detected water body in the shadow area to realize the rough extraction of time series information of the water body; in the stage of fine water body extraction, combined with Sentinel-2 optical data to improve the rough water body extraction results of the first phase, to Accurately extract the water body information in the initial phase; in the flood detection and inundation analysis stage, use the temporal and spatial distribution data of the water body to calculate the frequency of each pixel being recognized as a water body one by one, construct different flood event detection strategies based on different frequency intervals, and finally extract short-term Inundation area and continuous inundation area, and analyze the time series change characteristics of inundation area.

Specific related algorithms:

1. Median filter operator.

This operator is to reduce the error caused by coherent speckle noise in SAR images; median filtering is a nonlinear signal processing technique based on sorting statistics theory that can effectively suppress noise. The basic principle of median filtering is to convert digital images or digital The value of a point in the sequence is replaced by the median value of each point value in a neighborhood of the point, so that the surrounding pixel values are close to the real value, thereby eliminating isolated noise points. The technical solution of the embodiment of the present invention is to use a 3×3 matrix template to sort the pixels in the template according to the size of the pixel values, and generate a monotonously rising (or falling) two-dimensional data sequence.

yi=Med{fi-v,...,fi,...fi+v}i∈N, v=(m-1)/2#(1)

In the formula (1), set a one-dimensional sequence f1, f2...fn, take the window length m (m is an odd number), and perform median filtering on it, that is, take out m numbers from the input sequence successively, in The m numbers are sorted by size, and the number whose serial number is the center point is taken as the filtering output. In the embodiment of the present invention, m is equal to 3.

2. Water and land segmentation threshold algorithm.

Cross-polarized images mainly reflect volume scattering information and are less sensitive to specular reflection. Due to the smooth surface and strong homogeneity of the water body, the noise level in the cross-polarization image is low, and the intra-class variance is small. Compared with co-polar data, the overlapping area between water body and non-water body is smaller, and the separability is higher, which is more suitable for water body information extraction. Therefore, the embodiment of the present invention adopts the Otsu method (otsu) algorithm for water and land segmentation.

The Otsu method, also known as the maximum inter-class variance method, is an efficient algorithm for binarizing images proposed by the Japanese scholar Otsu Zhanyuki in 1979. The most basic formula (2) is established:

ω ₀ +ω ₁ ＝1#(2)

Then the total average gray value of the image is:

u ₀ ω ₀ +u ₁ ω ₁ ＝u#(3)

The between-class variance is:

g＝W ₀ (u ₀ -u) ² +w ₁ (u ₁ -u) ² #(4)

The optimal threshold for the image is T, and T divides the image into target and background. Among them, the proportion of target points to the total image is W0, and the average gray value is U0, and the proportion of background points to the image is W1, and the average gray value is U1.

3. Initial fine extraction of water body.

The spatio-temporal distribution results of the water body extracted by the above method include the possible flood potential area. When the finely extracted area of the water body is subtracted, the place where the flood occurs can be obtained; taking into account the consequences of land cover changes before and after the flood According to the significant change characteristics of the scattering coefficient, the time series anomaly detection method can be used to identify the flood submersion range and submersion time. Therefore, combined with the Sentinel-2 optical data, the normalized difference water index NDWI (Normalized Difference Water Index) is used to correct the initial SAR water body extraction results to retrieve reliable and accurate initial water body distribution information.

Due to the high reflectance of water in the green light band and the low reflectance in the near-infrared band, the ratio of the difference between the two reflectance values to the sum can effectively reflect the surface water body information. The ratio is NDWI, and the calculation is shown in formula (5):

In the formula, the green light band (Green) and the near-infrared band (NIR) correspond to the 3rd and 8th bands of the Sentinel-2MSI image, respectively.

4. Automatic detection of flood events.

Identifying flood events by detecting water body changes is limited by the low accuracy of SAR water body extraction. At the same time, cloudy and rainy weather also limits the water body extraction results of optical image synchronous correction time series. Using the time-series frequency-based flood event detection method to adapt to large-scale flood dynamic monitoring First, use formula (6) to calculate the frequency of pixels identified as water bodies in the time series:

In the formula, fw represents the frequency of recognition as water body, Nw represents the number of recognition as water body, and Na represents the total number of phases of water body extraction.

For flood events in the rapid inundation-retreat mode, the submerged area was submerged for a short time, and the frequency of water bodies was identified as low (fw<0.5) throughout the time series. However, due to the seasonal growth of crops, the backscatter coefficient of cultivated land shows seasonal fluctuations, and the backscatter coefficient is low in May and June, so it is easy to be misjudged as a water body change event. Flood inundation and flood retreat cause changes in the land cover type, and the variation range of the backscatter coefficient is usually higher than that of crop seasonal growth. Therefore, this characteristic can be used to extract the transient inundation area of the rapid inundation-recession flood mode. Use the Euclidean distance to measure the change intensity of the backscatter coefficient of the two adjacent images in turn, and the calculation formula is as follows (7):

为像元p在前后两期间的距离；分别表示在VV和VH极化方式下的后向散射系数,D用来判定变化的方向，利用i和j两期影像中vh的差值和差值绝对值的比值计算，正值表示vh增加，负值表示vh减小。In the formula, i and j represent the phases of the two images before and after;

is the distance of the pixel p in the two periods before and after; it represents the backscattering coefficient in the VV and VH polarization modes respectively, and D is used to determine the direction of change, using the difference and difference of vh in the i and j images The ratio calculation of the absolute value, the positive value indicates that vh increases, and the negative value indicates that vh decreases.

The technical solution of the embodiment of the present invention is developed on the GEE platform based on the JavaScript language. The whole process basically does not involve the back end, because as a remote sensing big data platform, the data can be used as needed. In addition, for various remote sensing to be processed Data, the GEE platform has many encapsulated functions, which are very fast and convenient to call.

The beneficial effects of the embodiments of the present invention are as follows:

It highlights the advantages of the cloud computing platform (GEE) for long-term dynamic monitoring of large-scale floods, which can efficiently process and analyze time-series data to support continuous monitoring of large-scale floods during the flood season. Realized the dynamic monitoring of the global flood situation within the theoretical scope, accurately extracted the spatio-temporal distribution of the flood inundation range in the required areas since 2015, and revealed the differences in the development patterns of flood conditions in different regions. Creatively integrating the vulnerability analysis into the assessment of the city's ability to suffer from floods, the vulnerability assessment can be performed while looking at the flooded areas, and then guidance suggestions can be given for local flood control construction.

System embodiment

According to an embodiment of the present invention, a flood automatic detection and dynamic monitoring device is provided. FIG. 5 is a schematic diagram of a flood automatic detection and dynamic monitoring device according to an embodiment of the present invention. As shown in FIG. The detection and dynamic monitoring devices specifically include:

The preprocessing module 50 is used to preprocess the earth observation data based on GEE; specifically, it is used to: determine data screening rules for monitoring requirements, group time series data based on the data screening rules, use median filter operators to filter, Mosaicking and clipping to construct a time-series SAR dataset with full coverage of the study area.

The rough extraction module 52 is used to roughly extract the spatio-temporal distribution of water bodies from the preprocessed earth observation data; specifically, it is used to: refer to the maximum inter-class variance method Otsu to determine the SAR image water and land segmentation threshold in the time series SAR data set, and based on the water and land Segment the threshold, use the slope and HAND data to remove false detection of water bodies in shadow areas, realize rough extraction of time series information of water bodies, and obtain rough extraction results of time and space distribution of water bodies.

The fine extraction module 54 is used to perform the first-stage fine water body extraction based on the rough extraction results of the spatiotemporal distribution of the water body; it is specifically used for: combining the Sentinel-2 optical data with the normalized water body index NDWI to correct the rough extraction results of the spatiotemporal distribution of the water body, and accurately extract the initial phase The water information of the first phase of the water body is obtained.

The analysis module 56 is used to perform flood event detection and inundation analysis based on the results of the first phase of fine water body extraction. Specifically for:

Based on the spatio-temporal distribution data of the water body in the first phase of water body fine extraction results, the frequency of each pixel being identified as a water body is calculated one by one, and different flood event detection strategies are constructed based on different frequency intervals, and the corresponding flood event detection strategies are used to extract short-term submergence. area and continuous inundation area, and analyze the temporal change characteristics of the inundation area.

The embodiment of the present invention is an apparatus embodiment corresponding to the above method embodiment, and the specific operations of each module can be understood by referring to the description of the method embodiment, and will not be repeated here.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. scope.

Claims (10)

1. A flood automatic detection and dynamic monitoring method, characterized in that, comprising: Earth observation data preprocessing based on GEE; Coarsely extract the spatio-temporal distribution of water bodies from the preprocessed earth observation data; Based on the rough extraction results of the spatio-temporal distribution of the water body, the first phase of fine water extraction is carried out; Flood event detection and inundation analysis are carried out based on the results of the first phase of fine water extraction. 2. The method according to claim 1, wherein the preprocessing of earth observation data based on GEE specifically comprises: The data screening rules are determined for monitoring needs, and the time series data are grouped based on the data screening rules, and the median filter operator is used for filtering, mosaic and clipping to construct a time series SAR data set with full coverage of the study area. 3. The method according to claim 1, wherein the rough extraction of water body spatiotemporal distribution to the preprocessed earth observation data specifically comprises: Refer to the maximum inter-class variance method Otsu to determine the water and land segmentation threshold of SAR images in the time series SAR data set, and based on the water and land segmentation threshold, use the slope and HAND data to remove the false detection of water bodies in the shadow area, realize the rough extraction of time series information of water bodies, and obtain the temporal and spatial distribution of water bodies Rough extraction results. 4. The method according to claim 1, characterized in that, based on the rough extraction results of the spatiotemporal distribution of the water body, the fine extraction of the first phase of the water body specifically includes: Combined with the Sentinel-2 optical data, the normalized water index NDWI is used to correct the rough extraction results of the temporal and spatial distribution of the water body, and the water body information of the initial phase is accurately extracted to obtain the fine extraction results of the first phase of the water body. 5. The method according to claim 1, characterized in that, performing flood event detection and submersion analysis based on the results of the first-phase fine water body extraction specifically includes: Based on the spatio-temporal distribution data of the water body in the first phase of water body fine extraction results, the frequency of each pixel being identified as a water body is calculated one by one, and different flood event detection strategies are constructed based on different frequency intervals, and the corresponding flood event detection strategies are used to extract short-term submergence. area and continuous inundation area, and analyze the temporal change characteristics of the inundation area. 6. A flood automatic detection and dynamic monitoring device, characterized in that it comprises: The preprocessing module is used for preprocessing the earth observation data based on GEE; The rough extraction module is used to perform rough extraction of the spatio-temporal distribution of water bodies on the preprocessed earth observation data; The fine extraction module is used for the first stage of fine water extraction based on the rough extraction results of the spatiotemporal distribution of water; The analysis module is used for flood event detection and inundation analysis based on the fine extraction results of the first phase of water. 7. The device according to claim 6, wherein the preprocessing module is specifically used for: The data screening rules are determined for monitoring needs, and the time series data are grouped based on the data screening rules, and the median filter operator is used for filtering, mosaic and clipping to construct a time series SAR data set with full coverage of the study area. 8. The device according to claim 6, wherein the rough extraction module is specifically used for: Refer to the maximum inter-class variance method Otsu to determine the water and land segmentation threshold of SAR images in the time series SAR data set, and based on the water and land segmentation threshold, use the slope and HAND data to remove the false detection of water bodies in the shadow area, realize the rough extraction of time series information of water bodies, and obtain the temporal and spatial distribution of water bodies Rough extraction results. 9. The device according to claim 6, wherein the fine extraction module is specifically used for: Combined with the Sentinel-2 optical data, the normalized water index NDWI is used to correct the rough extraction results of the temporal and spatial distribution of the water body, and the water body information of the initial phase is accurately extracted to obtain the fine extraction results of the first phase of the water body. 10. The device according to claim 8, wherein the analysis module is specifically used for: Based on the spatio-temporal distribution data of the water body in the first phase of water body fine extraction results, the frequency of each pixel being identified as a water body is calculated one by one, and different flood event detection strategies are constructed based on different frequency intervals, and the corresponding flood event detection strategies are used to extract short-term submergence. area and continuous inundation area, and analyze the temporal change characteristics of the inundation area.

Images