A flood detection method and system based on radar remote sensing without prior knowledge support

Abstract

The present invention relates to the field of flood detection, and more specifically, to a flood detection method and system based on radar remote sensing without prior knowledge support, wherein the method comprises: obtaining an image to be detected and a historical radar remote sensing image, preprocessing the image to be detected and the historical radar remote sensing image to obtain a processed image to be detected and a processed historical image; constructing a non-flood reference image and a standard deviation image according to the processed image to be detected and the processed historical image; obtaining a potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard deviation image; constructing a perennial water body and a water-like reference layer and a seasonal water body and a water-like reference layer; excluding the perennial water body and the water-like and the seasonal water body and the water-like in the potential flood layer, and determining the final flood range. The present invention obtains the potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard deviation image, so it is not necessary to know in advance which dates and places are flood images, and no prior knowledge support is required. In addition, the present invention excludes the influence of the perennial water body and the water-like and the seasonal water body and the water-like on the final flood detection, thereby improving the flood detection effect.

Description

Technical Field

[0001] The present invention relates to the field of flood detection, and more specifically, to a flood detection method and system based on radar remote sensing without prior knowledge support. Background Art

[0002] Flood disasters are one of the most frequent and serious natural disasters in the world, causing a huge impact on people's production and life, economic development, and ecosystems. At present, more than one-third of the world's land surface and 82% of the population face varying degrees of flood risk. Flood disasters usually have the characteristics of a large impact range and strong suddenness. Radar remote sensing technology has the characteristics of penetrating clouds and fog and is not affected by weather conditions. It can observe the surface all day and all weather, so it has important value in flood disaster monitoring research. Existing flood detection methods based on radar remote sensing usually include threshold method, change detection method, deep learning method or cluster analysis method. However, all of these existing methods require the support of prior knowledge, that is, before flood detection, it is necessary to determine in advance the location or approximate location of the flood, as well as flood period images and non-flood period images. These methods can only be used to detect known flood events, and cannot be used to mine historical flood events; secondly, the existing methods cannot distinguish between floods and perennial water bodies, seasonal water bodies, and quasi-water bodies, and will also judge normal water bodies and quasi-water bodies as floods, thereby affecting the final detection effect.

[0003] The prior art discloses a lightweight flood detection method and device based on a soft fusion strategy, the method comprising: obtaining a current time series satellite remote sensing image, processing the current time series satellite remote sensing image using a lightweight flood detection neural network, obtaining a probability matrix of the flood area, obtaining a preliminary detection result of the current time series satellite remote sensing image based on the probability matrix of the flood area, and determining a weighted constraint probability matrix of the current time series satellite remote sensing image; determining short-term memory and long-term memory according to the flood detection results of the historical time series satellite remote sensing image; using the maximum inter-class variance algorithm, according to the adaptive weighting factor, the short-term memory, the long-term memory and the weighted constraint probability matrix of the current time series satellite remote sensing image are fused and calculated to obtain the final flood detection result of the current time series satellite remote sensing image. And this method needs to know the flood area image in advance before using the lightweight flood detection neural network to process the current time series satellite remote sensing image to obtain the probability matrix of the flood area. This method requires preliminary work based on prior knowledge, and the method still cannot distinguish between floods and perennial water bodies, seasonal water bodies, and quasi-water bodies. It will also judge perennial water bodies, seasonal water bodies, and quasi-water bodies as floods, thus affecting the final detection effect. Summary of the invention

[0004] The purpose of the present invention is to disclose a flood detection method and system based on radar remote sensing and without prior knowledge support, which has better detection effect and does not require prior knowledge.

[0005] In order to achieve the above object, the present invention provides a flood detection method and system based on radar remote sensing without prior knowledge support, comprising:

[0006] S1: Acquire the image to be detected and the historical radar remote sensing image, pre-process the image to be detected and the historical radar remote sensing image, and obtain the processed image to be detected and the processed historical image;

[0007] S2: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images;

[0008] S3: Obtaining a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image;

[0009] S4: Construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers;

[0010] S5: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, exclude the perennial water body and quasi-water body and the seasonal water body and quasi-water body in the potential flood layer to obtain the final flood range.

[0011] Further, in step S1, the radar remote sensing image is a Sentinel-1 radar remote sensing image.

[0012] Furthermore, the image to be detected and the historical radar remote sensing image are preprocessed to obtain the processed image to be detected and the processed historical image including:

[0013] Thermal noise removal: Using filtering signal processing technology to remove or reduce thermal noise, improving the clarity and interpretability of Sentinel-1 data;

[0014] Orbit correction: The long-term series of Sentinel-1 data is corrected by precise satellite orbit data to obtain the corrected long-term series of Sentinel-1 data;

[0015] Radiometric calibration: The brightness value DN of Sentinel-1 data is calibrated relative to the standard cross-sectional area and converted into a backscattering coefficient;

[0016] De-banding: Remove the thin black bands with no signal between adjacent pulse bands in Sentinel-1 data;

[0017] Multi-Look: Improve the data radiometric resolution in Sentinel-1 data;

[0018] Terrain correction: mainly includes geocoding and terrain radiation correction, which converts Sentinel-1 data into actual coordinate information and eliminates the impact of terrain on Sentinel-1 data.

[0019] Furthermore, constructing a non-flood reference image and a standard deviation image according to the processed image to be detected and the processed historical image includes: calculating the median of the time series radar image according to the processed historical images of 9 months, namely m-1 month, m month and m+1 month in each year of the past three years in the processed historical images, and taking the median of the time series radar image as the non-flood reference image, wherein m is the month where the image to be detected is located;

[0020] Calculate the standard deviation image based on the processed historical images of these 9 months:

[0021]

[0022] Where s is the standard deviation of the pixel, n is the number of images, is the mean value of pixel I in the entire image set, I i is the image value of pixel I in the i-th image.

[0023] Further, in step S3, obtaining a potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard variance image includes: calculating the difference between the processed image to be detected and the non-flood reference image to obtain a difference image, and then calculating the ratio of the difference image to the standard variance image; identifying pixels whose backscatter intensity is lower than a set water body threshold on the processed image to be detected, and judging whether the ratio of the difference image to the standard variance image is lower than the change threshold on these pixels, if it is higher, no processing is performed, and if it is lower, the pixel is regarded as a potential flood pixel:

[0024]

[0025] PF is a potential flood judgment mark, I i is the image value of pixel I in the i-th image, I median The median value of pixel I in the entire image set, s is the standard deviation of pixel I in the entire image set.

[0026] Furthermore, the water body threshold is an empirical threshold.

[0027] Further, in step S4, constructing the perennial water body and quasi-water body reference layer includes: defining the pixels in the processed historical image corresponding to the image to be detected whose backscattering of the current year is lower than the set water body threshold with a probability of more than 80% as perennial water body and quasi-water body pixels:

[0028] PW=(P I>0.8)

[0029] PW is a permanent water body or a perennial water body, P I is the probability of pixel I being a water body in that year.

[0030] Further, in step S4, constructing a seasonal water body and quasi-water body reference layer includes: defining pixels with a water body probability of 1 in the processed historical images of months m-1, m, and m+1 in each of the past three years as seasonal water body and quasi-water body pixels, wherein m is the month of the image to be detected;

[0031] SW=(P1 I =1)&(P2 I =1)&(P3 I =1)

[0032] SW is seasonal water body, P1 I , P2 I , P3 I are the water body probabilities of pixel I in the 1st, 2nd and 3rd year respectively.

[0033] Further, in step S5, according to the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, the seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range, including: masking the non-flood area to determine the flood range:

[0034] Based on the potential flood layer obtained in step S3, remove the perennial water bodies and quasi-water bodies, and then remove the seasonal water bodies and quasi-water bodies to determine the flood range;

[0035] F=PF-PW-SW

[0036] Among them, F is the flood range, PF is the potential flood pixel, PW is the perennial water body and quasi-water body, and SW is the seasonal water body and quasi-water body.

[0037] In addition, the present invention provides a flood detection system based on radar remote sensing without prior knowledge support, comprising:

[0038] Acquisition module: acquires the image to be detected and the historical radar remote sensing image, pre-processes the image to be detected and the historical radar remote sensing image, and obtains the processed image to be detected and the processed historical image;

[0039] Non-flood module: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images;

[0040] Potential flood module: obtain the potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image;

[0041] Other water body modules: construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers;

[0042] Flood determination module: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range.

[0043] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0044] The present invention constructs a non-flood reference image and a standard deviation image based on the processed image to be detected and the processed historical image; then obtains a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image. Therefore, it is not necessary to know in advance which dates or places are flood images, and no prior knowledge is required, making the flood detection of the present invention simpler.

[0045] The present invention constructs perennial water body and quasi-water body reference layers and seasonal water body and quasi-water body reference layers; thereby eliminating the influence of perennial water body and quasi-water body and seasonal water body and quasi-water body on the final flood detection, thereby improving the flood detection effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of a flood detection method based on radar remote sensing without prior knowledge support as described in Example 1;

[0047] Figure 2 This is a framework diagram of a flood detection system based on radar remote sensing without prior knowledge support as described in Example 2; DETAILED DESCRIPTION

[0048] The drawings are for illustrative purposes only and should not be construed as limiting the present patent;

[0049] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0050] Embodiment 1:

[0051] This embodiment provides Figure 1 The flowchart of a flood detection method based on radar remote sensing without prior knowledge support is shown, comprising the following steps:

[0052] S1: Acquire the image to be detected and the historical radar remote sensing image, pre-process the image to be detected and the historical radar remote sensing image, and obtain the processed image to be detected and the processed historical image;

[0053] S2: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images;

[0054] S3: Obtaining a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image;

[0055] S4: Construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers;

[0056] S5: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, exclude the perennial water body and quasi-water body and the seasonal water body and quasi-water body in the potential flood layer to obtain the final flood range.

[0057] This embodiment constructs a non-flood reference image and a standard deviation image based on the processed image to be detected and the processed historical image; then obtains a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image. Therefore, it is not necessary to know in advance which dates or places are flood images, and no prior knowledge is required, making flood detection in the present invention simpler.

[0058] This embodiment constructs a perennial water body and a quasi-water body reference layer and a seasonal water body and a quasi-water body reference layer; thereby eliminating the influence of the perennial water body and the quasi-water body and the seasonal water body and the quasi-water body on the final flood detection, thereby improving the flood detection effect.

[0059] Embodiment 2:

[0060] This embodiment further discloses on the basis of the first embodiment:

[0061] In step S1, the radar remote sensing image is a Sentinel-1 radar remote sensing image.

[0062] Preprocess the image to be detected and the historical radar remote sensing image to obtain the processed image to be detected and the processed historical image including:

[0063] Thermal noise removal: Using filtering signal processing technology to remove or reduce thermal noise, improving the clarity and interpretability of Sentinel-1 data;

[0064] Orbit correction: The long-term series of Sentinel-1 data is corrected by precise satellite orbit data to obtain the corrected long-term series of Sentinel-1 data;

[0065] Radiometric calibration: The brightness value DN of Sentinel-1 data is calibrated relative to the standard cross-sectional area and converted into a backscattering coefficient;

[0066] De-banding: Remove the thin black bands with no signal between adjacent pulse bands in Sentinel-1 data;

[0067] Multi-Look: Improve the data radiometric resolution in Sentinel-1 data;

[0068] Terrain correction: mainly includes geocoding and terrain radiation correction, which converts Sentinel-1 data into actual coordinate information and eliminates the impact of terrain on Sentinel-1 data.

[0069] Constructing a non-flood reference image and a standard deviation image based on the processed image to be detected and the processed historical image includes: calculating the median of the time series radar image based on the processed historical images of 9 months, namely, month m-1, month m, and month m+1 in each year of the past three years, and using the median of the time series radar image as the non-flood reference image, wherein m is the month of the image to be detected;

[0070] Calculate the standard deviation image based on the processed historical images of these 9 months:

[0071]

[0072] Where s is the standard deviation of the pixel, n is the number of images, is the mean value of pixel I in the entire image set, I i is the image value of pixel I in the i-th image.

[0073] In a specific embodiment, the image to be detected is March 2024, and the processed historical images are respectively selected from February, March and April of 2023, 2022 and 2021, and the processed historical images of February, March and April of each of the three years 2023, 2022 and 2021, a total of 9 months, are calculated to obtain the median of the time series radar image.

[0074] In step S3, obtaining a potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard variance image includes: calculating the difference between the processed image to be detected and the non-flood reference image to obtain a difference image, and then calculating the ratio of the difference image to the standard variance image; identifying pixels whose backscatter intensity is lower than a set water body threshold on the processed image to be detected, and judging whether the ratio of the difference image to the standard variance image is lower than the change threshold on these pixels, if it is higher, no processing is performed, and if it is lower, the pixel is regarded as a potential flood pixel:

[0075]

[0076] PF is a potential flood judgment mark, I i is the image value of pixel I in the i-th image, Imedian The median value of pixel I in the entire image set, s is the standard deviation of pixel I in the entire image set.

[0077] The water body threshold is an empirical threshold.

[0078] In step S4, constructing the perennial water body and quasi-water body reference layer includes: defining the pixels in the processed historical image corresponding to the image to be detected whose backscattering of the current year is lower than the set water body threshold with a probability of more than 80% as perennial water body and quasi-water body pixels:

[0079] PW=(P I >0.8)

[0080] PW is a permanent water body or a perennial water body, P I is the probability of pixel I being a water body in that year.

[0081] In step S4, constructing a seasonal water body and quasi-water body reference layer includes: defining pixels with a water body probability of 1 in the processed historical images of months m-1, m, and m+1 in each of the past three years as seasonal water body and quasi-water body pixels, wherein m is the month of the image to be detected;

[0082] SW=(P1 I =1)&(P2 I =1)&(P3 I =1)

[0083] SW is seasonal water body, P1 I , P2 I , P3 I are the water body probabilities of pixel I in the 1st, 2nd and 3rd year respectively.

[0084] In a specific embodiment, the image to be detected is March 2024, then the processed historical images select February, March and April of 2023, 2022 and 2021 respectively, and the pixels with a water body probability of 1 in the processed historical images of February, March and April of each of the three years of 2023, 2022 and 2021, a total of 9 months, are defined as seasonal water bodies and quasi-water body pixels.

[0085] In step S5, according to the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, the seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range, including: Masking the non-flood area to determine the flood range:

[0086] Based on the potential flood layer obtained in step S3, remove the perennial water bodies and quasi-water bodies, and then remove the seasonal water bodies and quasi-water bodies to determine the flood range;

[0087] F=PF-PW-SW

[0088] Among them, F is the flood range, PF is the potential flood pixel, PW is the perennial water body and quasi-water body, and SW is the seasonal water body and quasi-water body.

[0089] The Sentinel-1 radar remote sensing image used in this embodiment is microwave frequency band data, which can penetrate water vapor and fog in the atmosphere. Therefore, it can overcome the influence of bad weather and fog cover, has all-day and all-weather monitoring capabilities, and is sensitive to the shape, structure, and dielectric constant of vegetation scatterers while having a certain penetration ability, which can provide a reliable data source for flood detection.

[0090] Embodiment three:

[0091] In this embodiment, Figure 2 A flood detection system based on radar remote sensing without prior knowledge support is provided, comprising:

[0092] Acquisition module: acquires the image to be detected and the historical radar remote sensing image, pre-processes the image to be detected and the historical radar remote sensing image, and obtains the processed image to be detected and the processed historical image;

[0093] Non-flood module: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images;

[0094] Potential flood module: obtain the potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image;

[0095] Other water body modules: construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers;

[0096] Flood determination module: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range.

[0097] This embodiment constructs a non-flood reference image and a standard deviation image based on the processed image to be detected and the processed historical image; then obtains a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image. Therefore, it is not necessary to know in advance which dates or places are flood images, and no prior knowledge is required, making flood detection in the present invention simpler.

[0098] This embodiment constructs a perennial water body and a quasi-water body reference layer and a seasonal water body and a quasi-water body reference layer; thereby eliminating the influence of the perennial water body and the quasi-water body and the seasonal water body and the quasi-water body on the final flood detection, thereby improving the flood detection effect.

[0099] The embodiment of the present invention provides a flood detection method and system based on radar remote sensing without prior knowledge support, wherein the method includes: obtaining an image to be detected and a historical radar remote sensing image, preprocessing the image to be detected and the historical radar remote sensing image to obtain a processed image to be detected and a processed historical image; constructing a non-flood reference image and a standard deviation image according to the processed image to be detected and the processed historical image; obtaining a potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard deviation image; constructing a perennial water body and a water-like reference layer and a seasonal water body and a water-like reference layer; excluding the perennial water body and the water-like and the seasonal water body and the water-like in the potential flood layer, and determining the final flood range. The embodiment of the present invention obtains the potential flood layer according to the processed image to be detected, the non-flood reference image, and the standard deviation image, so it is not necessary to know in advance which dates and places are flood images, and no prior knowledge support is required. In addition, the embodiment of the present invention eliminates the influence of the perennial water body and the water-like and the seasonal water body and the water-like on the final flood detection, thereby improving the flood detection effect.

[0100] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A flood detection method based on radar remote sensing without prior knowledge support, characterized in that: The following steps are involved: S1: Acquire the image to be detected and the historical radar remote sensing image, pre-process the image to be detected and the historical radar remote sensing image, and obtain the processed image to be detected and the processed historical image; S2: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images; The median of the time series radar image is calculated based on the processed historical images of the 9 months m-1, m, and m+1 in each year of the past three years, and the median of the time series radar image is used as the non-flood reference image, where m is the month of the image to be detected; Calculate the standard deviation image based on the processed historical images of these 9 months: Where s is the standard deviation of the pixel, n is the number of images, is the mean value of pixel I in the entire image set, I i is the image value of pixel I in the i-th image; S3: Obtaining a potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image; Calculate the difference between the processed image to be detected and the non-flood reference image to obtain the difference image, and then calculate the ratio of the difference image to the standard variance image; identify the pixels whose backscatter intensity is lower than the set water body threshold on the processed image to be detected, and judge whether the ratio of the difference image to the standard variance image is lower than the change threshold on these pixels. If it is higher, no processing is performed; if it is lower, the pixel is regarded as a potential flood pixel: PF is a potential flood judgment mark, I i is the image value of pixel I in the i-th image, I median The median value of pixel I in the entire image set, s is the standard deviation of pixel I in the entire image set; S4: Construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers; S5: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, exclude the perennial water body and quasi-water body and the seasonal water body and quasi-water body in the potential flood layer to obtain the final flood range.

2. A flood detection method based on radar remote sensing without prior knowledge support according to claim 1, characterized in that: In step S1, the radar remote sensing image is a Sentinel-1 radar remote sensing image.

3. A flood detection method based on radar remote sensing without prior knowledge support according to claim 2, characterized in that: Preprocess the image to be detected and the historical radar remote sensing image to obtain the processed image to be detected and the processed historical image including: Thermal noise removal: Using filtering signal processing technology to remove or reduce thermal noise, improving the clarity and interpretability of Sentinel-1 data; Orbit correction: The long-term series of Sentinel-1 data is corrected by precise satellite orbit data to obtain the corrected long-term series of Sentinel-1 data; Radiometric calibration: The brightness value DN of Sentinel-1 data is calibrated relative to the standard cross-sectional area and converted into a backscattering coefficient; De-banding: Remove the thin black bands with no signal between adjacent pulse bands in Sentinel-1 data; Multi-Look: Improve the data radiometric resolution in Sentinel-1 data; Terrain correction: mainly includes geocoding and terrain radiation correction, which converts Sentinel-1 data into actual coordinate information and eliminates the impact of terrain on Sentinel-1 data.

4. The flood detection method based on radar remote sensing without prior knowledge support according to claim 1 is characterized in that: The water body threshold is an empirical threshold.

5. The flood detection method based on radar remote sensing without prior knowledge support according to claim 1 is characterized in that: In step S4, constructing the perennial water body and quasi-water body reference layer includes: defining the pixels in the processed historical image corresponding to the image to be detected whose backscattering of the current year is lower than the set water body threshold with a probability of more than 80% as perennial water body and quasi-water body pixels: PW=(P I >0.8) PW is a permanent water body or a perennial water body, P I is the probability of pixel I being a water body in that year.

6. The flood detection method based on radar remote sensing without prior knowledge support according to claim 1 is characterized in that: In step S4, constructing a seasonal water body and quasi-water body reference layer includes: defining pixels with a water body probability of 1 in the processed historical images of months m-1, m, and m+1 in each of the past three years as seasonal water body and quasi-water body pixels, wherein m is the month of the image to be detected; SW=(P1 I =1)&(P2 I =1)&(P3 I =1) SW is seasonal water body, P1 I , P2 I , P3 I are the water body probabilities of pixel I in the 1st, 2nd and 3rd year respectively.

7. The flood detection method based on radar remote sensing without prior knowledge support according to claim 1 is characterized in that: In step S5, according to the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, the seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range, including: Masking the non-flood area to determine the flood range: Based on the potential flood layer obtained in step S3, remove the perennial water bodies and quasi-water bodies, and then remove the seasonal water bodies and quasi-water bodies to determine the flood range; F=PF-PW-SW Among them, F is the flood range, PF is the potential flood pixel, PW is the perennial water body and quasi-water body, and SW is the seasonal water body and quasi-water body.

8. A flood detection system based on radar remote sensing without prior knowledge support, characterized in that: include: Acquisition module: acquires the image to be detected and the historical radar remote sensing image, pre-processes the image to be detected and the historical radar remote sensing image, and obtains the processed image to be detected and the processed historical image; Non-flood module: Construct non-flood reference images and standard deviation images based on the processed images to be detected and the processed historical images; The median of the time series radar image is calculated based on the processed historical images of the 9 months m-1, m, and m+1 in each year of the past three years, and the median of the time series radar image is used as the non-flood reference image, where m is the month of the image to be detected; Calculate the standard deviation image based on the processed historical images of these 9 months: Where s is the standard deviation of the pixel, n is the number of images, is the mean value of pixel I in the entire image set, I i is the image value of pixel I in the i-th image; Potential flood module: obtain the potential flood layer based on the processed image to be detected, the non-flood reference image, and the standard deviation image; Calculate the difference between the processed image to be detected and the non-flood reference image to obtain the difference image, and then calculate the ratio of the difference image to the standard variance image; identify the pixels whose backscatter intensity is lower than the set water body threshold on the processed image to be detected, and judge whether the ratio of the difference image to the standard variance image is lower than the change threshold on these pixels. If it is higher, no processing is performed; if it is lower, the pixel is regarded as a potential flood pixel: PF is a potential flood judgment mark, I i is the image value of pixel I in the i-th image, I median The median value of pixel I in the entire image set, s is the standard deviation of pixel I in the entire image set; Other water body modules: construct perennial water body and water-like reference layers and seasonal water body and water-like reference layers; Flood determination module: Based on the perennial water body and quasi-water body reference layer and the seasonal water body and quasi-water body reference layer, the perennial water body and quasi-water body, seasonal water body and quasi-water body are excluded from the potential flood layer to obtain the final flood range.

Images