Method for rapid extraction of flood range based on satellite-borne synthetic aperture radar band operation

Abstract

The present application relates to the technical field of remote sensing detection, and more particularly to a flood range rapid extraction method based on satellite-borne synthetic aperture radar band operation. The method comprises: obtaining pre-disaster first data, post-disaster second data and digital elevation model data; performing band operation on the first and second data respectively to obtain background maps; pixel-by-pixel difference between the two background maps to obtain a difference map, and fusion with a water body polarization confidence map obtained based on polarization decomposition to generate a water body determination map; identifying a perennial water body area according to the first background map to generate a mask; extracting a flood inundation preliminary map from the water body determination map and the perennial water body mask; and finally removing terrain shadow using a digital elevation model to obtain a flood range map. The present application determines newly added water bodies by fusing the difference map and the polarization confidence, removes terrain shadow interference in combination with the digital elevation model, realizes rapid and automatic extraction of the flood range under complex ground surface, and significantly improves the precision and timeliness.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, and in particular to a method for rapid extraction of flood range based on satellite synthetic aperture radar band calculation. Background Technology

[0002] Every summer, floods are frequent, and flood control, disaster relief, and post-disaster reconstruction urgently require the rapid acquisition of flood-affected areas. However, due to the generally thick cloud cover during floods, optical satellites and drones cannot image the ground. Spaceborne synthetic aperture radar (SAR) technology, with its advantages of all-weather, all-time, wide-area coverage, and ability to penetrate clouds, provides a new technical means for rapidly extracting flood-affected areas. However, existing methods for extracting flood-affected areas using spaceborne SAR have the following shortcomings: slow speed and low efficiency. Whether using manual interpretation, the magic wand method, or other methods, all require human observation and manual operation, resulting in slow speed and low efficiency; prone to omissions. When there are many flood-affected areas, especially some small ones, they are easily missed; low accuracy. The extracted flood-affected area boundaries have low precision.

[0003] Therefore, there is an urgent need for a method for rapidly extracting the extent of flood disasters using spaceborne SAR that can specifically address the above problems, thereby improving the efficiency, accuracy, and reliability of flood disaster range extraction. Summary of the Invention

[0004] To address this, the present invention provides a method for rapid extraction of flood range based on satellite-borne synthetic aperture radar band computation, which overcomes the problems of low efficiency and poor accuracy in the extraction of flood range caused by reliance on manual interpretation, strong subjectivity of thresholds, and difficulty in distinguishing between constant water bodies and terrain shadows in the existing technology.

[0005] To achieve the above objectives, this invention provides a method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band computation, comprising: Step S1: Obtain the first data of the disaster-stricken area before the flood, the second data of the disaster-stricken area after the flood, and the digital elevation model data of the disaster-stricken area. Step S2: Perform band operations on the first data to obtain the first background image data, and perform band operations on the second data to obtain the second background image data; Step S3: Subtract the second background image data from the first background image data pixel by pixel to obtain the difference image data, and obtain the water body polarization confidence map of the second data based on the polarization decomposition method. Then, fuse the difference image data with the water body polarization confidence map to generate water body judgment image data. Step S4: Identify the normal water body areas in the disaster-stricken area before the flood based on the first background map data, and generate normal water body mask data. Step S5: Based on the water body determination map data and the normal water body mask data, extract the flood-inundated area and generate the initial map data of the flood-inundated area; Step S6: Using the digital elevation model data, remove the terrain shadow areas from the initial map data of the flood-inundated area to obtain the flood range map data.

[0006] Further, in step S2, the first background image data and the second background image data are obtained, including: Perform band operations on the first data and set its background color to white to obtain the first background image data; Perform band operations on the second data and set its background color to white to obtain the second background image data.

[0007] Further, in step S3, water body determination map data is generated, including: Calculate the grayscale difference between the second background image data and the first background image data pixel by pixel, and combine all grayscale differences into difference image data according to the position of the corresponding pixel. An adaptive first threshold map is generated by fusing local statistics using a sliding window with the global Otsu method. Pixels with a grayscale difference greater than the first threshold are identified as candidate new water body pixels; Apply neighborhood spatial consistency constraints to candidate newly added water body pixels and remove candidate newly added water body pixels whose proportion of newly added water body in the neighborhood is less than a preset proportion. The retained candidate new water body pixels are combined into water body determination map data.

[0008] Further, in step S4, real-time water mask data is generated, including: Iterate through each pixel in the first background image data and compare the grayscale value of each pixel with a preset second threshold. If the gray value is less than the second threshold, the pixel is determined to be a normal water body pixel; All constant water body pixels are combined according to their spatial location to form constant water body mask data.

[0009] Further, in step S5, preliminary map data of the flood-inundated area is generated, including: The water body determination map data is compared pixel by pixel with the normal water body mask data; If a pixel is identified as a newly added water body pixel in the water body determination map data, but is not marked as a normal water body pixel in the normal water body mask data, then the pixel is identified as a flood inundation zone pixel. All pixels of the flood-inundated area are combined according to their spatial location to form the initial map data of the flood-inundated area.

[0010] Further, in step S6, flood extent map data is obtained, including: The digital elevation model data is used to identify terrain shadow areas and generate terrain shadow mask data. The initial map data of the flood-inundated area is compared pixel by pixel with the terrain shadow mask data; If a pixel is identified as a flood-inundated pixel in the initial map data of the flood-inundated area, but is not marked as a terrain shadow pixel in the terrain shadow mask data, then the pixel is identified as a flood-affected area pixel. All flood-affected area pixels are combined according to their spatial location to form flood-affected area map data.

[0011] Furthermore, the removal of terrain shadow areas using the digital elevation model data includes: Based on the digital elevation model data and the orbital parameters of the spaceborne SAR sensor, a SAR grayscale image is generated, and overlapping, shadows, and slope shortening regions are identified in the SAR grayscale image. The SAR grayscale image and the grayscale image of the second data are analyzed pixel by pixel. Only when a pixel is marked as a shadow area in the SAR grayscale image and the increment of the grayscale value of the pixel in the grayscale image of the second data relative to the grayscale value of its corresponding position in the SAR grayscale image is less than a preset grayscale recovery threshold, the pixel is marked as a terrain shadow pixel. The terrain shadow mask data is generated based on the terrain shadow region.

[0012] Furthermore, it also includes: By introducing prior data on building distribution, we can identify areas of bright pores inside water bodies caused by secondary scattering from buildings. Morphological closing operations were used to preprocess the initial map data of the flood-inundated area to close the gaps in the holes; Using the pre-processed flooded area boundary as the seed point, region growth is performed on the regions in the difference map data where the gray-level difference is less than the first threshold of the corresponding pixel position in the adaptive first threshold map to repair missing pixels in the holes.

[0013] Furthermore, when the first data and the second data come from different spaceborne SAR sensors or the difference in imaging geometry parameters exceeds a preset range, the method further includes the following step before step S2: Stable and unchanging ground feature sample sets are selected from the first data and the second data, respectively; Based on the stable and invariant ground feature sample set, construct a grayscale mapping function from the second data to the first data; The grayscale mapping function is used to perform relative radiometric normalization on the second data.

[0014] Furthermore, it also includes: Isolated pixel blocks with an area smaller than a preset area threshold are removed from the flood disaster area map data, and smoothing filtering is performed. When removing isolated pixel blocks, for each connected region with an area smaller than a preset area threshold, its roundness or compactness feature is calculated, and the digital elevation model data is used to determine whether the connected region is located in a local depression. If the roundness is greater than the preset roundness threshold and it is located in a local depression, it is determined to be a real small water body and is retained.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: by uniformly setting the background color of the pre-disaster and post-disaster SAR images to white and then performing pixel-by-pixel differential calculation, the gray values ​​of non-effective observation areas are consistent, avoiding misjudgment caused by background interference; at the same time, the white background and the low gray values ​​of the water body form a strong contrast, which significantly improves the sensitivity and accuracy of newly added water body identification, thereby achieving high-precision extraction of flood inundation range.

[0016] Furthermore, this invention determines newly added water bodies by fusing the band-operated difference map of SAR images before and after floods with the polarization decomposition water body confidence map, and combines this with DEM to remove terrain shadow interference. This enables rapid and automated extraction of flood inundation ranges in complex surface environments, significantly improving extraction accuracy and timeliness. This method effectively suppresses isolated noise points through adaptive threshold segmentation and neighborhood spatial consistency constraints, avoiding the missed and false detection problems of traditional fixed threshold methods under heterogeneous surface conditions. Simultaneously, by using DEM to simulate SAR imaging and identify terrain shadow areas, it eliminates spurious change information caused by overlay and shadows, further improving the reliability of flood range extraction.

[0017] Furthermore, this invention uses a statistical method or adaptive Otsu method based on known constant water body regions to determine the second threshold, thereby achieving accurate judgment of constant water body pixels in pre-disaster images. This avoids the subjective bias and poor regional adaptability caused by manually setting fixed thresholds, thus generating accurate and reliable constant water body mask data, laying a solid foundation for subsequent accurate removal of constant water bodies and extraction of real flood-inundated areas.

[0018] Furthermore, through the aforementioned pixel-by-pixel comparison and logical judgment, this invention combines all pixels identified as flood-prone areas according to their spatial location to form initial flood-prone area data. This initial image data is also a binarized image used to indicate the preliminary flooding range caused by this flood, eliminating interference from pre-disaster water bodies. This ensures that the flooding area extraction results focus on the actual newly flooded areas caused by the flood, providing accurate input data for subsequent removal of terrain shadow interference based on digital elevation model data.

[0019] Furthermore, this invention determines newly added water bodies by fusing the band-operated difference map of SAR images before and after floods with the polarization decomposition water body confidence map, and combines this with DEM to remove terrain shadow interference, achieving rapid and automated extraction of flood inundation range in complex surface environments, significantly improving extraction accuracy and timeliness. This method effectively suppresses isolated noise points through adaptive threshold segmentation and neighborhood spatial consistency constraints, avoiding the missed detections and false detections of traditional fixed threshold methods under heterogeneous surface conditions. Through joint analysis of simulated SAR grayscale images and actual images, it accurately identifies and removes spurious changes caused by terrain shadows based on grayscale recovery thresholds, further improving the reliability of flood range extraction. For densely built-up urban areas, it introduces prior data on building distribution to repair internal holes in water bodies caused by secondary scattering, making the extraction results more complete and accurate.

[0020] Furthermore, this invention determines newly added water bodies by fusing band-operated difference maps of SAR images before and after floods with polarization decomposition water body confidence maps, and combines this with DEM to remove terrain shadow interference, achieving rapid and automated extraction of flood extent in complex surface environments, significantly improving accuracy and timeliness. This method employs adaptive threshold segmentation and neighborhood spatial consistency constraints to suppress isolated noise, and accurately identifies and removes spurious changes caused by terrain shadows through joint analysis of simulated SAR grayscale images and actual images. For densely built-up urban areas, prior data on building distribution is introduced to locate internal pores in water bodies caused by secondary scattering, which are then repaired using morphological closure operations and region growing constrained by difference maps. When there are significant differences in sensor or imaging geometry between two image periods, a grayscale mapping function is constructed using stable ground object samples for relative radiometric normalization to eliminate systematic bias. In the post-processing stage, area and roundness screening combined with DEM depression discrimination remove noise while protecting real small water bodies. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method for rapid extraction of flood range based on spaceborne synthetic aperture radar band calculation in this embodiment; Figure 2 This is a flowchart of the process for generating water body determination map data in the method for rapid extraction of flood range based on spaceborne synthetic aperture radar band calculation in this embodiment; Figure 3 This is a flowchart illustrating the process of generating real-time water mask data in the method for rapid extraction of flood range based on spaceborne synthetic aperture radar band computation in this embodiment. Detailed Implementation

[0022] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0023] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] Please see Figure 1 As shown, it is a flowchart of the method for rapid extraction of flood range based on spaceborne synthetic aperture radar band calculation in this embodiment.

[0025] This embodiment provides a method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation, including: Step S1: Obtain the first data of the disaster-stricken area before the flood, the second data of the disaster-stricken area after the flood, and the digital elevation model data of the disaster-stricken area. Step S2: Perform band operations on the first data to obtain the first background image data, and perform band operations on the second data to obtain the second background image data; Step S3: Subtract the second background image data from the first background image data pixel by pixel to obtain the difference image data, and obtain the water body polarization confidence map of the second data based on the polarization decomposition method. Then, fuse the difference image data with the water body polarization confidence map to generate water body judgment image data. Step S4: Identify the normal water body areas in the disaster-stricken area before the flood based on the first background map data, and generate normal water body mask data. Step S5: Based on the water body determination map data and the normal water body mask data, extract the flood-inundated area and generate the initial map data of the flood-inundated area; Step S6: Using the digital elevation model data, remove the terrain shadow areas from the initial map data of the flood-inundated area to obtain the flood range map data.

[0026] In this embodiment, the first radar image data of the disaster-stricken area taken by a spaceborne synthetic aperture radar sensor before the flood is acquired, serving as the first data; a second radar image data of the same area after the flood is acquired, serving as the second data; and digital elevation model (DEM) data covering the area is acquired simultaneously. Both the first and second data undergo preprocessing such as radiometric calibration, multi-view processing, speckle filtering, and geocoding to ensure accurate spatial registration and a consistent geographic coordinate system with the DEM data. The DEM data is used for subsequent terrain shadow identification and removal.

[0027] Specifically, in step S2, the first background image data and the second background image data are obtained, including: Perform band operations on the first data and set its background color to white to obtain the first background image data; Perform band operations on the second data and set its background color to white to obtain the second background image data.

[0028] In this embodiment, the acquired pre-disaster first data and post-disaster second data are processed using band operations. Specifically, the first data is a pre-disaster single-polarization or dual-polarization SAR image, stored in grayscale value form after radiometric calibration, with the grayscale value typically ranging from 0 to 255. When performing band operations on the first data, pixels with a grayscale value of 0 or marked as null are identified as background pixels, and their grayscale values ​​are uniformly assigned to 255; for pixels with grayscale values ​​between 1 and 255, they are identified as ground feature pixels containing real surface backscattering information, and their original grayscale values ​​are kept unchanged, forming the first background map data. Similarly, the same band operation is performed on the second data, uniformly assigning a grayscale value of 255 to background pixels marked as 0 or null, while keeping the original grayscale values ​​of ground feature pixels unchanged, to obtain the second background map data.

[0029] This invention achieves consistent grayscale values ​​in non-effective observation areas by uniformly setting the background color of pre-disaster and post-disaster SAR images to white before performing pixel-by-pixel differential calculations, thus avoiding misjudgments caused by background interference. At the same time, the white background contrasts sharply with the low grayscale values ​​of water bodies, significantly improving the sensitivity and accuracy of identifying newly added water bodies, thereby realizing high-precision extraction of flood inundation range.

[0030] Please see Figure 2 As shown, it is a flowchart of the method for rapidly extracting flood range based on spaceborne synthetic aperture radar band calculation in this embodiment, which generates water body determination map data.

[0031] Specifically, in step S3, water body determination map data is generated, including: Step S21: Calculate the grayscale difference between the second background image data and the first background image data pixel by pixel, and combine all grayscale differences into difference image data according to the position of the corresponding pixel; Step S22: The sliding window local statistics and the global Otsu method are fused to generate an adaptive first threshold map; Step S23: Pixels with grayscale differences greater than the first threshold are identified as candidate new water body pixels; Step S24: Apply neighborhood spatial consistency constraints to candidate newly added water body pixels and remove candidate newly added water body pixels whose proportion of newly added water body in the neighborhood is less than a preset proportion. Step S25: Combine the retained candidate new water body pixels into water body determination map data.

[0032] In this embodiment, firstly, the grayscale difference between the second background image data and the first background image data is calculated pixel by pixel, and all grayscale differences are combined into a difference map data according to the spatial position of the corresponding pixels. For example, if the grayscale value of a pixel in the second background image is significantly higher than the corresponding value in the first background image, a larger grayscale difference is generated at that position; if the grayscale values ​​of the two images are close, the grayscale difference at that position is smaller. Arranging the grayscale differences of all pixels according to their original spatial positions forms the difference map data.

[0033] Secondly, an adaptive first threshold map is generated by fusing local statistics using a sliding window with the global Otsu method. Specifically, for each pixel in the difference map data, a neighborhood window of appropriate size is taken centered on that pixel. The overall distribution characteristics of the gray-level differences of all pixels within this window are statistically analyzed, and the local judgment benchmark of the central pixel of the window is calculated. Simultaneously, the global judgment benchmark is calculated using the Otsu method across the entire difference map data. Finally, for each pixel position, the local and global judgment benchmarks are combined to form the final adaptive threshold for that pixel position. The adaptive thresholds of all pixels are combined according to their spatial positions to form the adaptive first threshold map. Each pixel in the difference map data is traversed, and the gray-level difference of each pixel is compared with the adaptive threshold at the corresponding position in the adaptive first threshold map. If the gray-level difference of a pixel is greater than the adaptive threshold at that position, the pixel is determined as a candidate for new water body; otherwise, the pixel is not determined as a new water body.

[0034] Then, a neighborhood spatial consistency constraint is applied to the candidate newly added water body pixels. Specifically, for each candidate newly added water body pixel, a neighborhood window of appropriate size is taken centered on it, and the proportion of pixels within this window that are determined to be candidate newly added water bodies is counted. If this proportion is lower than a preset proportion, it means that the candidate pixel lacks support from other newly added water body pixels around it, and it is determined to be an isolated noise point and is removed; conversely, if the proportion reaches or exceeds the preset proportion, it means that the pixel is located in a large area of ​​newly added water bodies, and it is retained as a newly added water body pixel.

[0035] The preset ratio is determined based on the clustering distribution characteristics of newly added water body pixels in typical flood disaster area samples to ensure that isolated noise points can be effectively distinguished from areas of newly added water bodies. If the preset ratio is set too small, scattered isolated noise points may be misjudged as newly added water bodies, increasing false alarms in the water body determination map. If the preset ratio is set too large, scattered pixels in the actual water body edge areas are difficult to identify, reducing the integrity of the water body boundary. Therefore, this embodiment sets the preset ratio in a transitional range between the distribution of isolated noise points and the clustering distribution of large water bodies, ensuring that subsequent flood inundation area extraction is based on the effective suppression of false alarms in the determination of newly added water bodies, while also considering the integrity and accuracy of the water body boundary. Finally, all newly added water body pixels retained after the above screening are combined according to their spatial location to form water body determination map data. This water body determination map data is a binarized image used to indicate the spatial distribution range of newly added water bodies after the disaster.

[0036] This invention determines newly added water bodies by fusing the band-operated difference map of SAR images before and after floods with the polarization decomposition water body confidence map, and combines this with DEM to remove terrain shadow interference. This enables rapid and automated extraction of flood inundation areas in complex surface environments, significantly improving extraction accuracy and timeliness. The method effectively suppresses isolated noise points through adaptive threshold segmentation and neighborhood spatial consistency constraints, avoiding the missed and false detection problems of traditional fixed threshold methods on heterogeneous surfaces. Simultaneously, by using DEM to simulate SAR imaging and identify terrain shadow regions, it eliminates spurious changes caused by overlay and shadows, further enhancing the reliability of flood extent extraction.

[0037] Please see Figure 3 As shown, it is a flowchart of the method for rapidly extracting flood range based on spaceborne synthetic aperture radar band calculation in this embodiment, which generates constant water body mask data.

[0038] Specifically, in step S4, the generation of constant-time water mask data includes: Step S31: Traverse each pixel in the first background image data and compare the gray value of each pixel with a preset second threshold. Step S32: If the gray value is less than the second threshold, then the pixel is determined to be a normal water body pixel; Step S33: Combine all the pixels of the constant water body according to their spatial location to form constant water body mask data.

[0039] In this embodiment, each pixel in the first background image data is traversed, and the grayscale value of each pixel is compared with a preset second threshold one by one. Since in the pre-disaster first background image data, permanent water bodies typically exhibit low backscatter echo characteristics in SAR images, their grayscale values ​​are significantly lower than those of non-water features such as vegetation, bare soil, and buildings. Therefore, when the grayscale value of a pixel is less than the second threshold, the pixel is determined to be a permanent water body pixel; conversely, if the grayscale value is greater than or equal to the second threshold, it is determined to be a non-water body pixel. After traversing all pixels, all pixels determined to be permanent water bodies are combined according to their spatial location into binary permanent water body mask data. Permanent water body regions are assigned a specific marker value, and non-water body regions are assigned another marker value, used in subsequent steps to distinguish between newly added water bodies after the disaster and existing permanent water bodies.

[0040] The second threshold is determined as follows: It is set based on the statistical characteristics of the grayscale distribution of typical water body samples and typical non-water body samples in the pre-disaster background map data. Specifically, areas known as normal water bodies in the first background map data are selected as the water body sample set, while areas known as vegetation, bare soil, buildings, and other non-water body features are selected as the non-water body sample set. The mean grayscale value and dispersion of the two types of samples are statistically analyzed. The second threshold is set in the transition range between the grayscale distribution of water body samples and the grayscale distribution of non-water body samples, enabling water body pixels to be effectively distinguished from non-water body pixels. If the second threshold is set too low, some water bodies with high grayscale values ​​will be missed, such as water bodies with enhanced echoes due to increased surface roughness caused by wind speed; this can result in incomplete normal water body masks. If the second threshold is set too high, some non-water bodies with low grayscale values ​​will be misidentified as water bodies, increasing the error in subsequent flood-inundated area extraction.

[0041] This invention uses a statistical method or adaptive Otsu method based on known constant water body regions to determine the second threshold, thereby achieving accurate judgment of constant water body pixels in pre-disaster images. This avoids the subjective bias and poor regional adaptability caused by manually setting a fixed threshold, thus generating accurate and reliable constant water body mask data, laying a solid foundation for subsequent accurate removal of constant water bodies and extraction of real flood-inundated areas.

[0042] Specifically, in step S5, initial map data of the flood-inundated area is generated, including: The water body determination map data is compared pixel by pixel with the normal water body mask data; If a pixel is identified as a newly added water body pixel in the water body determination map data, but is not marked as a normal water body pixel in the normal water body mask data, then the pixel is identified as a flood inundation zone pixel. All pixels of the flood-inundated area are combined according to their spatial location to form the initial map data of the flood-inundated area.

[0043] In this embodiment, the water body determination map data generated in step S3 and the constant water body mask data generated in step S4 are spatially aligned and compared pixel by pixel. The water body determination map data is a binary image, where pixel value 1 represents newly added water body pixels after the disaster, and pixel value 0 represents non-newly added water body pixels; the constant water body mask data is also a binary image, where pixel value 1 represents constant water body pixels before the disaster, and pixel value 0 represents non-water body pixels.

[0044] The marker values ​​of the two data images at the same spatial location are read pixel by pixel. If a pixel is marked as a newly added water body pixel in the water body determination map data, but is not marked as a normal water body pixel in the normal water body mask data, then the pixel is determined to be a pixel of the newly added flooded area in this flood. Conversely, if a pixel is marked as a newly added water body pixel in the water body determination map data, but is also marked as a normal water body pixel in the normal water body mask data, it means that the water body existed before the disaster and was not caused by this flood, so it is not included in the flood flooded area. If a pixel is not marked as a newly added water body in the water body determination map data, then no matter what the marker value is in the normal water body mask data, it is not determined to be a pixel of the flood flooded area.

[0045] This invention, through the aforementioned pixel-by-pixel comparison and logical judgment, combines all pixels identified as flood-prone areas according to their spatial location to form initial flood-prone area data. This initial image data is also a binarized image used to indicate the preliminary flood range caused by this flood, eliminating interference from pre-disaster water bodies. This ensures that the flood-prone area extraction results focus on the actual newly flooded areas caused by the flood, providing accurate input data for subsequent removal of terrain shadow interference based on digital elevation model data.

[0046] Specifically, in step S6, flood extent map data is obtained, including: The digital elevation model data is used to identify terrain shadow areas and generate terrain shadow mask data. The initial map data of the flood-inundated area is compared pixel by pixel with the terrain shadow mask data; If a pixel is identified as a flood-inundated pixel in the initial map data of the flood-inundated area, but is not marked as a terrain shadow pixel in the terrain shadow mask data, then the pixel is identified as a flood-affected area pixel. All flood-affected area pixels are combined according to their spatial location to form flood-affected area map data.

[0047] Specifically, removing terrain shadow areas using the digital elevation model data includes: Based on the digital elevation model data and the orbital parameters of the spaceborne SAR sensor, a SAR grayscale image is generated, and overlapping, shadows, and slope shortening regions are identified in the SAR grayscale image. The SAR grayscale image and the grayscale image of the second data are analyzed pixel by pixel. Only when a pixel is marked as a shadow area in the SAR grayscale image and the increment of the grayscale value of the pixel in the grayscale image of the second data relative to the grayscale value of its corresponding position in the SAR grayscale image is less than a preset grayscale recovery threshold, the pixel is marked as a terrain shadow pixel. The terrain shadow mask data is generated based on the terrain shadow region.

[0048] In this embodiment, firstly, terrain shadow regions are identified using the digital elevation model data, generating terrain shadow mask data. Specifically, based on the digital elevation model data and the orbital parameters of the spaceborne SAR sensor, including but not limited to the satellite flight direction, side view, and incident angle, a radar geometric simulation algorithm is used to generate a simulated SAR grayscale image. In this simulated image, overlapping, shadows, and slope shortening regions caused by terrain undulations are identified. Among these, shadow regions appear as continuous dark areas without echo signals in the simulated image and are a key source of interference to focus on in subsequent processing.

[0049] The simulated SAR grayscale image and the actual acquired grayscale image of the second data are analyzed pixel-by-pixel. For each pixel marked as a shadow area in the simulated image, the grayscale value of its corresponding position in the second data grayscale image is checked, and the increment of this grayscale value relative to the grayscale value of that pixel in the simulated image is calculated. If the increment is less than a preset grayscale rise threshold, it indicates that the pixel did not show a significant backscattering intensity rise in actual observation and was not filled by high-echo features such as water bodies, so it is marked as a real terrain shadow pixel; conversely, if the pixel shows a significant grayscale rise in the second data grayscale image and reaches or exceeds the preset grayscale rise threshold, it indicates that there are other features generating echoes in the actual ground surface and it does not belong to the terrain shadow that needs to be removed. After traversing all pixels identified as shadow areas in the simulated image, all pixels marked as terrain shadows are combined according to their spatial location to form terrain shadow mask data. This mask data is a binary image, where terrain shadow pixels are assigned a first label value and non-shadow pixels are assigned a second label value.

[0050] The preset grayscale recovery threshold is determined based on the statistical characteristics of the grayscale distribution of typical shaded and non-shaded areas in the second data. Specifically, typical areas known to be affected by terrain shadows in the second data are selected as the shadow sample set, while non-shaded areas adjacent to the shadowed areas, with flat terrain and similar land cover types are selected as the non-shaded sample set. The mean and dispersion of grayscale values ​​for the two types of samples are statistically analyzed. The grayscale recovery threshold is set in the transition range between the grayscale distribution of shadowed and non-shaded samples, so that the real terrain shadows can be effectively distinguished from areas that recover due to water echoes. When the threshold is set too low, some shadow areas with slight grayscale recovery due to weak water echoes will be misjudged as terrain shadows and removed, resulting in missed detection of flood-inundated areas. When the threshold is set too high, some real terrain shadow pixels will be retained in the flooded area because their grayscale has not recovered sufficiently, resulting in misjudgment of false change information. By setting the grayscale recovery threshold in the above-mentioned transition range, the terrain shadow mask can effectively remove false flooding information while preserving as much of the real flood-inundated area as possible.

[0051] Then, the initial map data of the flood-inundated area and the generated terrain shadow mask data are spatially aligned and compared pixel by pixel. The initial map data of the flood-inundated area is a binary image, indicating the pixels of the initially determined flooded area; the terrain shadow mask data is also a binary image, indicating the terrain shadow pixels that need to be removed. The comparison logic is as follows: if a pixel is determined to be a flood-inundated pixel in the initial map data of the flood-inundated area, but is not marked as a terrain shadow pixel in the terrain shadow mask data, then the pixel is determined to be a final flood range pixel and is retained; conversely, if the pixel is determined to be a flood-inundated pixel in the initial map data of the flood-inundated area, but is marked as a terrain shadow pixel in the terrain shadow mask data, it indicates that the flood determination of this area may be caused by a false change due to terrain shadows, and the pixel is removed from the flood range and not included in the final result. Finally, all the flood range pixels retained after the above filtering are combined according to their spatial location to form flood range map data.

[0052] This invention determines newly added water bodies by fusing the band-operated difference map of SAR images before and after floods with the polarization decomposition water body confidence map, and combines this with DEM to remove terrain shadow interference. This enables rapid and automated extraction of flood inundation ranges in complex surface environments, significantly improving extraction accuracy and timeliness. The method effectively suppresses isolated noise points through adaptive threshold segmentation and neighborhood spatial consistency constraints, avoiding the missed detections and false detections of traditional fixed threshold methods on heterogeneous surfaces. Through joint analysis of simulated SAR grayscale images and actual images, it accurately identifies and removes spurious changes caused by terrain shadows based on grayscale recovery thresholds, further improving the reliability of flood range extraction. For densely built-up urban areas, it introduces prior data on building distribution to repair internal holes in water bodies caused by secondary scattering, making the extraction results more complete and accurate.

[0053] Specifically, the method further includes: By introducing prior data on building distribution, we can identify areas of bright pores inside water bodies caused by secondary scattering from buildings. Morphological closing operations were used to preprocess the initial map data of the flood-inundated area to close the gaps in the holes; Using the pre-processed flooded area boundary as the seed point, region growth is performed on the regions in the difference map data where the gray-level difference is less than the first threshold of the corresponding pixel position in the adaptive first threshold map to repair missing pixels in the holes.

[0054] In this embodiment, for areas with man-made structures, such as densely built-up urban areas, the initial map data of flood-prone areas often shows bright holes within the water body due to the secondary scattering effect of buildings. This means the water area is fragmented into discontinuous patches by strong building echoes, affecting the integrity of the flooded area. Therefore, prior data on building distribution is introduced. This prior data can be obtained from existing urban building vector layers or building footprint information extracted from high-resolution optical images. It is used to locate areas with bright holes within the water body caused by building secondary scattering. Spatial overlay analysis is performed on the initial map data of the flood-prone area and the prior data on building distribution to identify the locations of low-grayscale water cavities located within densely built-up areas and surrounded by bright building echoes.

[0055] Then, morphological closing operations were used to preprocess the initial map data of the flood-inundated area. Morphological closing operations are a combination of expansion and erosion operations, which can close small pores and gaps within the water body and connect adjacent broken water patches, making the boundary of the inundated area more continuous and complete. After morphological closing preprocessing, most of the small pores inside the inundated area were filled, and the boundary of the inundated area became more regular.

[0056] Finally, using the preprocessed flooded area boundary as the seed point, region growing is performed on regions in the difference map data whose gray-level difference is less than the adaptive threshold of the corresponding pixel position in the adaptive first threshold map, to repair the missing pixels within the remaining holes. The basic logic of region growing is as follows: starting from the seed pixel at the flooded area boundary, it expands pixel by pixel to adjacent unmarked pixels. If the gray-level difference of an adjacent pixel in the difference map data is less than the adaptive threshold corresponding to that pixel position, it is included in the flooded area; the above expansion process is repeated until no new pixels meet the growth conditions. Since the real water pixels inside the holes are represented by low gray-level differences in the difference map data, while non-water pixels such as buildings are represented by high gray-level differences.

[0057] Specifically, when the first data and the second data come from different spaceborne SAR sensors or the difference in imaging geometric parameters exceeds a preset range, the process includes the following steps before step S2: Stable and unchanging ground feature sample sets are selected from the first data and the second data, respectively; Based on the stable and invariant ground feature sample set, construct a grayscale mapping function from the second data to the first data; The grayscale mapping function is used to perform relative radiometric normalization on the second data.

[0058] In this embodiment, when the first data and the second data come from different spaceborne SAR sensors, or although they come from the same sensor but the geometric parameter differences between the two images exceed a preset range, in addition to the actual surface changes caused by the flood, a systematic grayscale deviation caused by sensor characteristics or imaging geometric differences is superimposed between the two images. This deviation will be directly transmitted to the difference map data, causing the grayscale difference in non-changed areas to be non-zero, thereby interfering with the accurate determination of newly added water body pixels. Therefore, before performing band calculation in step S2, the second data needs to be subjected to relative radiometric normalization processing, and the specific implementation process is as follows.

[0059] First, stable and unchanging land cover sample sets are selected from the first and second data sets, respectively. Stable and unchanging land cover refers to natural or artificial land cover types that have not changed before and after the disaster, including but not limited to stable bare soil areas, rooftops of man-made buildings unaffected by floods, road surfaces, stable rock or concrete surfaces, etc. The selection of sample sets is accomplished through manual interpretation or automatic extraction: manual interpretation is based on the texture features and prior knowledge of optical imagery or SAR imagery, selecting multiple sample areas at the same geographical location on pre- and post-disaster images that have not undergone significant visual changes; automatic extraction is based on the initial screening results of the difference map, automatically classifying pixels with the smallest gray-level difference changes into the stable sample candidate set, and further filtering is performed in conjunction with texture features and spatial continuity constraints to ensure the purity and representativeness of the sample sets. The number and spatial distribution of the sample sets should cover different land cover types and different gray-level ranges throughout the entire image as much as possible to ensure the global applicability of the subsequent gray-level mapping function.

[0060] Then, a grayscale mapping function from the second data to the first data is constructed based on the stable and invariant ground feature sample set. The grayscale values ​​of each sample pixel in the stable and invariant ground feature sample set in the first data are used as target values, and the grayscale values ​​in the second data are used as source values, forming several sets of one-to-one corresponding grayscale value pairs. Based on these grayscale value pairs, a linear regression method is used to establish the grayscale mapping function from the second data to the first data, ensuring that the mapped grayscale values ​​of the second data have as consistent a grayscale representation as possible with the first data on the same ground feature. Finally, the grayscale mapping function is used to perform relative radiometric normalization on the second data. The original grayscale value of each pixel in the second data is substituted into the grayscale mapping function to calculate the normalized grayscale value, which replaces the original grayscale value of that pixel, forming the normalized second data.

[0061] Specifically, it also includes: Isolated pixel blocks with an area smaller than a preset area threshold are removed from the flood disaster area map data, and smoothing filtering is performed. When removing isolated pixel blocks, for each connected region with an area smaller than a preset area threshold, its roundness or compactness feature is calculated, and the digital elevation model data is used to determine whether the connected region is located in a local depression. If the roundness is greater than the preset roundness threshold and it is located in a local depression, it is determined to be a real small water body and is retained.

[0062] In this embodiment, firstly, isolated pixel blocks with areas smaller than a preset area threshold are removed from the flood extent map data. The flood extent map data is a binary image, where pixels marked as flood extents are often presented as connected components. All independent connected components are identified using an eight-neighbor connectivity labeling algorithm, and the number of pixels contained in each connected component is counted, which is used as the area of ​​that connected component. Isolated pixel blocks with areas smaller than the preset area threshold are usually noise or sporadic non-real flooded areas, and are removed from the flood extent map data to avoid interference from minor noise in the final result. However, in real-world scenarios, there may be small bodies of water that are indeed caused by floods.

[0063] Secondly, when removing isolated pixel blocks, not all connected components with an area smaller than a preset area threshold are discarded. Instead, each connected component with an area smaller than the preset area threshold is further evaluated to identify whether it is a real small body of water and to retain it. Specifically, for each connected component with an area smaller than the preset area threshold, its roundness feature is calculated, which is the ratio of the area of ​​the connected component to the area of ​​the circle corresponding to its perimeter. The closer the roundness value is to 1, the closer the shape of the connected component is to a circle. Real small bodies of water in SAR images usually appear as regular geometric shapes that are approximately circular or elliptical, while noise spots or non-water body interference usually appear as irregular fragments. Therefore, the roundness feature can effectively distinguish between the two.

[0064] Simultaneously, the digital elevation model (DEM) data is used to determine whether the connected region is located in a local depression. A local depression refers to a low-lying area in the DEM where the elevation value is lower than that of surrounding pixels. These depressions typically form in closed, concave terrains without outlets and represent favorable topographic conditions for natural water accumulation. The spatial location of the connected region is overlaid with the DEM data, and the elevation value of the location and its neighboring elevation distribution are extracted to determine if the location is within a local depression. If the roundness of a connected region is greater than a preset roundness threshold and the region is located within a local depression, it indicates that the region conforms to the characteristics of a real small water body in both shape and topographic location, and is thus retained as a real small water body, not included in the rejection range. Conversely, if the roundness of the connected region is less than or equal to the preset roundness threshold, or if the roundness meets the condition but the region is not located in a local depression, it indicates that the region does not possess the complete characteristics of a real small water body and is rejected as noise.

[0065] The preset area threshold and preset roundness threshold are determined based on the distribution characteristics of real small water bodies in typical flood disaster area samples to ensure effective differentiation between noise and real small water bodies. If the preset area threshold is set too low, a large number of noise spots will be retained in the results, reducing the data quality of the flood disaster area map; if the preset area threshold is set too high, real small water bodies are easily mistakenly removed, resulting in information loss. Similarly, if the preset roundness threshold is set too low, irregular noise is easily misjudged as real water bodies and retained, increasing false alarms in the results; if the preset roundness threshold is set too high, some connected regions that are slightly distorted in shape due to SAR imaging geometry but are indeed real water bodies will be missed, resulting in information loss. By setting the preset area threshold and preset roundness threshold within the transition range between noise distribution and real water body distribution, the post-processing process can effectively suppress noise interference while preserving as much real small water body information as possible.

[0066] Finally, the flood extent map data, after the isolated pixel block removal and preservation of real small water bodies, is subjected to smoothing filtering to eliminate any small jagged edges and isolated spikes in the resulting image, making the boundaries of the flooded area smoother and more continuous. After post-processing optimization, the final flood extent map data is obtained, which is used to indicate the true spatial distribution range of this flood.

[0067] This invention achieves rapid and automated extraction of flood extent in complex surface environments by fusing band-operated difference maps of SAR images before and after floods with polarization decomposition water body confidence maps, and combining this with DEM to remove topographic shadow interference. This significantly improves accuracy and timeliness. The method employs adaptive threshold segmentation and neighborhood spatial consistency constraints to suppress isolated noise. It accurately identifies and removes spurious changes caused by topographic shadows through joint analysis of simulated SAR grayscale images and actual images. For densely built-up urban areas, prior data on building distribution is used to locate internal pores in water bodies caused by secondary scattering, which are then repaired using morphological closure operations and region growing constrained by difference maps. When there are significant differences in sensor or imaging geometry between two image periods, a grayscale mapping function is constructed using stable ground object samples for relative radiometric normalization to eliminate systematic bias. In the post-processing stage, area and roundness filtering, combined with DEM depression discrimination, remove noise while protecting real small water bodies.

Claims

1. A method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band computation, characterized in that, include: Step S1: Obtain the first data of the disaster-stricken area before the flood, the second data of the disaster-stricken area after the flood, and the digital elevation model data of the disaster-stricken area. Step S2: Perform band operations on the first data to obtain the first background image data, and perform band operations on the second data to obtain the second background image data; Step S3: Subtract the second background image data from the first background image data pixel by pixel to obtain the difference image data, and obtain the water body polarization confidence map of the second data based on the polarization decomposition method. Then, fuse the difference image data with the water body polarization confidence map to generate water body judgment image data. Step S4: Identify the normal water body areas in the disaster-stricken area before the flood based on the first background map data, and generate normal water body mask data. Step S5: Based on the water body determination map data and the normal water body mask data, extract the flood-inundated area and generate the initial map data of the flood-inundated area; Step S6: Using the digital elevation model data, remove the terrain shadow areas from the initial map data of the flood-inundated area to obtain the flood range map data.

2. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, In step S2, the first background image data and the second background image data are obtained, including: Perform band operations on the first data and set its background color to white to obtain the first background image data; Perform band operations on the second data and set its background color to white to obtain the second background image data.

3. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, In step S3, water body determination map data is generated, including: Calculate the grayscale difference between the second background image data and the first background image data pixel by pixel, and combine all grayscale differences into difference image data according to the position of the corresponding pixel. An adaptive first threshold map is generated by fusing local statistics using a sliding window with the global Otsu method. Pixels with a grayscale difference greater than the first threshold are identified as candidate new water body pixels; Apply neighborhood spatial consistency constraints to candidate newly added water body pixels and remove candidate newly added water body pixels whose proportion of newly added water body in the neighborhood is less than a preset proportion. The retained candidate new water body pixels are combined into water body determination map data.

4. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, In step S4, constant water mask data is generated, including: Iterate through each pixel in the first background image data and compare the grayscale value of each pixel with a preset second threshold. If the gray value is less than the second threshold, the pixel is determined to be a normal water body pixel; All constant water body pixels are combined according to their spatial location to form constant water body mask data.

5. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, In step S5, preliminary map data of the flood-inundated area is generated, including: The water body determination map data is compared pixel by pixel with the normal water body mask data; If a pixel is identified as a newly added water body pixel in the water body determination map data, but is not marked as a normal water body pixel in the normal water body mask data, then the pixel is identified as a flood inundation zone pixel. All pixels of the flood-inundated area are combined according to their spatial location to form the initial map data of the flood-inundated area.

6. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, In step S6, flood extent map data is obtained, including: The digital elevation model data is used to identify terrain shadow areas and generate terrain shadow mask data. The initial map data of the flood-inundated area is compared pixel by pixel with the terrain shadow mask data; If a pixel is identified as a flood-inundated pixel in the initial map data of the flood-inundated area, but is not marked as a terrain shadow pixel in the terrain shadow mask data, then the pixel is identified as a flood-affected area pixel. All flood-affected area pixels are combined according to their spatial location to form flood-affected area map data.

7. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 6, characterized in that, Removing terrain shadow areas using the digital elevation model data includes: Based on the digital elevation model data and the orbital parameters of the spaceborne SAR sensor, a SAR grayscale image is generated, and overlapping, shadows, and slope shortening regions are identified in the SAR grayscale image. The SAR grayscale image and the grayscale image of the second data are analyzed pixel by pixel. Only when a pixel is marked as a shadow area in the SAR grayscale image and the increment of the grayscale value of the pixel in the grayscale image of the second data relative to the grayscale value of its corresponding position in the SAR grayscale image is less than a preset grayscale recovery threshold, the pixel is marked as a terrain shadow pixel. The terrain shadow mask data is generated based on the terrain shadow region.

8. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, Also includes: By introducing prior data on building distribution, we can identify areas of bright pores inside water bodies caused by secondary scattering from buildings. Morphological closing operations were used to preprocess the initial map data of the flood-inundated area to close the gaps in the holes; Using the pre-processed flooded area boundary as the seed point, region growth is performed on the region where the gray-level difference in the difference map data is less than the first threshold of the corresponding pixel position in the adaptive first threshold map to repair the missing pixels in the hole.

9. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, When the first data and the second data come from different spaceborne SAR sensors or the difference in imaging geometry parameters exceeds a preset range, the method further includes the following step before step S2: Stable and unchanging ground feature sample sets are selected from the first data and the second data, respectively; Based on the stable and invariant ground feature sample set, construct a grayscale mapping function from the second data to the first data; The grayscale mapping function is used to perform relative radiometric normalization on the second data.

10. The method for rapid extraction of flood disaster range based on spaceborne synthetic aperture radar band calculation according to claim 1, characterized in that, Also includes: Isolated pixel blocks with an area smaller than a preset area threshold are removed from the flood disaster area map data, and smoothing filtering is performed. When removing isolated pixel blocks, for each connected region with an area smaller than a preset area threshold, its roundness or compactness feature is calculated, and the digital elevation model data is used to determine whether the connected region is located in a local depression. If the roundness is greater than the preset roundness threshold and it is located in a local depression, it is determined to be a real small water body and is retained.

Images