Method for automatically extracting flood inundation range of power grid equipment based on water body index

Abstract

The invention discloses a water body index-based power grid equipment flood inundation range automatic extraction method, and belongs to the technical field of power system safety monitoring and image processing. According to the invention, a water body index value is obtained through specific mathematical operation by using reflectivity data of different wavebands based on spectral characteristics of a water body on a remote sensing image. The water body index can remarkably highlight the difference between the water body and other ground features, and the water body part in the remote sensing image can be accurately extracted by setting a proper threshold value. And carrying out spatial overlay analysis on known geographic position information of the power grid equipment and the extracted water body range, and matching coordinate data of the power grid equipment with vector data of the water body range by virtue of a geographic information system technology so as to determine which power grid equipment is in a flood submerging range. According to the method, whether the large-scale power grid equipment is submerged or not can be conveniently and rapidly detected, the disaster damage degree can be dynamically sensed in real time, and real-time dynamic monitoring and accurate evaluation of the flood submerged information of the power grid equipment are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system safety monitoring and image processing, and particularly relates to an automatic extraction method for the flood inundation range of power grid equipment based on a water body index. Background Art

[0002] The safe and stable operation of power grid equipment has long been threatened by floods. Especially, floods beyond the standard (exceeding the defense standard of the basin flood control project system) caused by large-scale continuous heavy rainfall may even cause devastating blows to power grid equipment. When floods pass by, a large number of power grid equipment are washed away and submerged, resulting in reduced production and shutdown of factories, interruption of transportation and communication, and the normal social order will be disrupted, and all aspects of society will be severely impacted. Therefore, it is crucial for the power department to timely and accurately master the flood inundation range of power grid equipment to take effective emergency measures and reduce disaster losses.

[0003] Although China has compiled flood risk maps for some protected areas of important rivers to guide the daily flood control management and actual flood defense response, there are serious problems of insufficient coverage. From the technical method perspective, the existing flood risk maps can only simulate the flood risks of specified-level floods and fixed breach floods, with a low matching degree with the actual dynamic flood process, and a small possibility that the actual flood matches the design conditions, greatly reducing the guiding value for flood control decision-making. Traditional flood disaster assessments are mostly carried out by local manual data statistics and reporting after the disaster occurs, lacking real-time monitoring and dynamic assessment of the flood disaster occurrence process. In addition, the current attention to the flood impact carriers mostly focuses on the direct damage of houses, agriculture, industry and commerce, water conservancy projects, transportation, water supply, etc. within the inundation range, while less work has been carried out on the flood risk analysis and assessment of important lifeline projects such as power supply facilities. Generally speaking, the traditional manual monitoring method has low efficiency and is difficult to implement in harsh environments such as floods, and cannot meet the requirements of real-time and accuracy. Some existing methods for extracting the flood inundation range based on remote sensing images have problems such as low accuracy and poor adaptability to complex terrain and water body characteristics, and it is difficult to accurately identify the flood inundation situation in the area where power grid equipment is located. The power grid control center has insufficient real-time control ability for the damage of power grid equipment caused by extreme weather and the operation status of equipment, and the information obtained by monitoring means is not intuitive and specific, resulting in low response efficiency of power grid equipment in the face of flood disasters, which becomes a problem restricting the safe operation of the power grid.

[0004] To consolidate the basic support position of power facilities in flood control and disaster relief, it is urgent to conduct systematic research on the technology for quickly extracting the flood inundation information of a large range of power grid equipment, draw the flood inundation degree zoning map of power grid equipment in real time, improve the emergency response ability to flood disasters and the active disaster prevention ability of the power grid, and provide technical support for the safe and reliable operation of the power grid.

[0005] Therefore, there is an urgent need for a new technical solution in the existing technology to solve this problem. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide an automatic extraction method for the flood inundation range of power grid equipment based on water body index, which is used to solve the technical problems of low accuracy, poor adaptability to complex terrain and water body characteristics, and inability to monitor and dynamically evaluate in real time in the existing flood inundation range extraction methods.

[0007] The technical solution adopted by the present invention is to provide an automatic extraction method for the flood inundation range of power grid equipment based on water body index, including the following steps:

[0008] Step 1: Data collection and processing: Collect satellite remote sensing image data according to requirements, and process the collected images.

[0009] Step 2: Inundation information extraction: Extract the flood inundation information of power grid equipment from the processed data based on the water body index extraction algorithm, and verify and analyze the extraction results.

[0010] Step 3: Thematic element production: Produce thematic map data and form a data analysis report on the flood and waterlogging disasters of power grid equipment.

[0011] The data collection and processing in the above Step 1 include remote sensing image data collection, remote sensing image data quality inspection, and optical data generation.

[0012] The remote sensing image data collection includes data collection and data arrangement:

[0013] The collection content includes remote sensing image data with full coverage of the project area range and meeting the project requirements for resolution, such as optical and unmanned aerial vehicle remote sensing images, supplemented by collection and arrangement of other available remote sensing image data and relevant basic data to support remote sensing image data processing, ensuring that the remote sensing image data meeting the project resolution covers the entire area range;

[0014] Perform image quality inspection on the existing standard scene remote sensing image data, perform secondary correction on the data that does not meet the quality requirements, generate a mapping file after ensuring that all data quality is qualified, perform spatial overlay analysis on the mapping file and the regional range boundary file, judge the area covered by each scene of data, and based on this, sort the image data according to the organizational structure directory of area range - time phase - sensor.

[0015] The quality inspection of the remote sensing image data is carried out by using image data processing software, mainly in an automatic human-computer interaction method, and the results are recorded. Check whether the spatial resolution meets the requirements of the satellite remote sensing image of the project, whether the spectral bands meet the requirement that the satellite remote sensing image contains at least 4 bands of blue, green, red, and near-infrared, and can accurately reflect the main ground object content such as vegetation, soil, water body, and building, and the information with clear boundaries. Review whether the remote sensing image fully covers the project implementation scope, the shooting area, without missing corners or gaps, check whether there is cloud or snow coverage in the image coverage area, whether the total cloud amount meets the requirements, check that the satellite images of different strips should have an overlapping area of 10%, check that the side view angle of the satellite image result is not greater than 30 degrees, check whether the image has rich levels, clear texture details, normal tone, no obvious noise, spots, bad lines, seams, and deformations, and there is no abnormal highlight in the urban buildings.

[0016] For the generation of the optical data, the orthorectification is carried out on the panchromatic and multispectral images of the received single-piece satellite image data respectively, and then the multispectral image is geographically registered by using the corrected panchromatic image. The registered panchromatic and multispectral images are subjected to image fusion, and the fused image is subjected to color homogenization output by using professional software. Finally, image mosaicking is carried out, and the mosaic line should make the main ground objects complete as much as possible, and the image texture should be naturally connected to achieve a balanced and consistent rendering color tone and no obvious seam marks, so as to generate an orthophoto image (DOM) product.

[0017] In the extraction of inundation information in Step 2, a water body index inversion model is used. The water body index inversion model is based on the spectral differences between water bodies and other ground objects, including the Normalized Difference Water Index (NDWI) and the Modified Normalized Difference Water Index (MNDWI). The preprocessed remote sensing image is calculated to highlight the characteristics of water bodies in the image, which is used for the extraction and identification of water body information in the remote sensing image. The input of the water body index inversion model is the optical surface reflectance data, and the output is the water body index image, which is in raster format. According to the statistical characteristics of the water body index image, an adaptive threshold segmentation method is used to segment the water body index image into two parts: water body and non-water body.

[0018] The water body index inversion model constructs a variety of water body indices based on multi-source remote sensing images. The specific process includes water body index model construction, quality inspection and accuracy verification, and extraction of water-covered surfaces.

[0019] (1) Water body index model construction

[0020] Based on the analysis of the spectral characteristics of water bodies in multiple bands, multiple bands closely related to water body identification are selected, the mapping relationship between water bodies and remote sensing spectral values is analyzed, and a mathematical model of the water body index is constructed:

[0021] ① Normalized Difference Water Index (NDWI)

[0022]

[0023] Where ρ GREEN represents the brightness value or reflectance in the green light band, and ρ NIR represents the brightness value or reflectance in the near-infrared band;

[0024] ② Modified Normalized Difference Water Index (MNDWI)

[0025]

[0026] Where ρ SWIR represents the brightness value in the mid-infrared band;

[0027] ③ Multi-Band Water Index (MBWI)

[0028] Based on a full analysis of the differences between water and other low-reflectance surfaces, the MBWI water index is constructed. It can suppress non-water features in mountain shadows and dark built-up areas, enhance surface water information, and reduce the seasonal and daily impacts caused by changes in solar conditions. The calculation formula is as follows:

[0029] MBWI = 2×ρ Green - ρ Red - ρ NIR - ρ SWIR1 - ρ SWIR2

[0030] Where ρ Red represents the brightness value in the red light band, and ρ SWIR1 , ρ SWIR2 are the reflectances of two short-wave infrared bands;

[0031] ④ Multi-Band Combined Water Index (MBCWI)

[0032] Taking advantage of the characteristics that the reflectance of water gradually decreases with the increase of wavelength in the visible to mid-infrared bands, and the reflectance of water is significantly lower than that of other ground objects in the near-infrared and mid-infrared bands, the MBCWI index is constructed, which can suppress the impacts of mountain shadows, building shadows, clouds, bare soil, and some low-reflectance objects. The calculation formula is as follows:

[0033] MBCWI = 2×(ρ NIR - ρ SWIR1 ) - (ρ Blue + ρ Green + ρ Red)

[0034] In the formula, ρ Blue represents the brightness value or reflectance in the blue light band;

[0035] ⑤ Automated Water Extraction Index (AWEI)

[0036] The arithmetic combination of spectral bands in AWEI is determined based on a rigorous examination of the reflectance characteristics of various land cover types. By band differencing, adding, and applying different coefficients, the separability between water and non-water pixels is maximized. The algorithm model coefficients are empirical results determined based on the reflectance patterns observed in the pure pixel datasets of various land cover types. The ultimate goal is to enhance the separability between water and dark surfaces, such as shadows and building structures, and improve the identification effect of surface water in areas including deep shadows and dark surfaces caused by mountainous terrain. The calculation formula is as follows:

[0037] AWEI nsh = 4×(ρ Creen -ρ SWIR1 ) - (0.25×ρ NIR + 2.75×ρ SWIR2 )

[0038] AWEI sh = ρ Blue + 2.5×ρ Green - 1.5×(ρ NIR + ρ SWIR1 ) - 0.25×ρ SWIR2

[0039] "nsh" is used to specify that the index is applicable to cases where shadows are not a major problem. The subscript "sh" indicates that this formula is designed to effectively eliminate shadow pixels and improve the water extraction accuracy in areas with shadows and / or other dark surfaces;

[0040] (2) Quality Inspection and Accuracy Verification

[0041] According to the above synthesis principle, the information on the data quality related to the dekadal, monthly, quarterly, and annual synthesis periods is stored in one band. The synthesized data consists of all the original data corresponding to the selected pixels. According to the value range corresponding to different water body indices, outliers are screened and uniformly replaced;

[0042] The accuracy inspection of the water body index is mainly verified by comparing with the remotely sensed data with high spatial resolution in the same period. Representative regional samples with different water areas are selected to verify the reliability of the water body index product;

[0043] (3) Water Extraction Based on the OSTU Method

[0044] The OSTU method uses the idea of clustering to divide the gray levels of an image into two parts according to the gray level, so that the gray value difference between the two parts is the largest and the gray difference within each part is the smallest. By calculating the variance, a suitable gray level is found for division. Assuming that the threshold range of the NDWI or MNDWI image is from x to y, where -1 ≤ x < y ≤ 1, the threshold t obtained based on the OSTU algorithm can divide the NDWI or MNDWI image into two categories: water body (from t to y) and non-water body (from x to t). The calculation of the optimal threshold t of the OSTU algorithm is as follows:

[0045]

[0046] where δ is the between-class variance of the water body and the non-water body, P w and P nw are the probabilities that a single pixel belongs to the water body and the non-water body respectively, M w and M nw are the average values of the water body and the non-water body, and M is the average value of the NDWI or MNDWI image;

[0047] After obtaining the optimal threshold through calculation, the threshold method is directly used for classification to obtain the water body extraction result;

[0048] (4) Verification of the accuracy of the water body extraction result

[0049] Set the reference sample as the result of manual water body extraction and the sample data made manually; the sample to be evaluated is the water body result automatically extracted by the water body extraction algorithm;

[0050] The verification evaluation indicators mainly include accuracy, precision, recall, F1 score, G score, AP and MAP, and ROC curve.

[0051] The production of thematic elements in step 3 is range extraction:

[0052] Overlay and analyze the segmented water body range with the location information of power grid equipment, extract the range of power grid equipment flooded by floods, and display it in a visual way to generate a thematic map. Combine GIS and historical remote sensing images to establish a basic geographic information database of power facilities and a flood inundation element database, and conduct spatial association to quickly evaluate the inundation range, water depth, duration and loss of power grid equipment under different scenarios of different breaches, overflows, positions and scales in different years, and draw a "single map" of flood risk zoning.

[0053] Through the above design scheme, the present invention can bring the following beneficial effects:

[0054] (1) High accuracy: By constructing a targeted water body index, it is possible to more accurately identify water bodies, reduce misjudgments and missed judgments, and improve the accuracy of flood inundation area extraction.

[0055] (2) Strong real-time performance: The entire algorithm process has a high degree of automation and can quickly process a large amount of remote sensing image data to meet the power department's need for real-time monitoring of flood disasters.

[0056] (3) Good adaptability: It has good adaptability to different terrains, different types of water bodies, and complex remote sensing image data, and can accurately extract the flood inundation area of power grid equipment in various scenarios.

[0057] (4) Using rolling updated remote sensing images and flood inundation area extraction algorithms to real-time extract the flood disaster situation of large-scale power grid equipment. The method is simple and fast, avoiding the inconsistency and lag of data reported by various departments after the disaster, and effectively improving the flood control and disaster resistance support ability.

[0058] (5) The algorithm of the present invention can accurately evaluate the flood risk of the location of power grid equipment, timely detect emerging disasters, thereby providing more accurate and timely early warning information for decision-making departments and relevant personnel, enhancing the emergency management ability of power enterprises, and improving their rapid response ability and decision-making level in dealing with natural disasters and other emergencies. Description of the Drawings

[0059] Figure 1 It is the overall technical flow chart of an automatic extraction method for the flood inundation area of power grid equipment based on a water body index according to the present invention;

[0060] Figure 2 It is the water body index processing flow chart of an automatic extraction method for the flood inundation area of power grid equipment based on a water body index according to the present invention;

[0061] Figure 3 It is a schematic diagram of the inundation area of the Lalin River Basin of an automatic extraction method for the flood inundation area of power grid equipment based on a water body index according to the present invention;

[0062] Figure 4 It is the monitoring distribution map of the water body inundation area in Jilin City, Jilin Province in 1994 of an automatic extraction method for the flood inundation area of power grid equipment based on a water body index according to the present invention;

[0063] Figure 5 It is the monitoring distribution map of the water body and substation equipment in Jilin City, Jilin Province in 1994 of an automatic extraction method for the flood inundation area of power grid equipment based on a water body index according to the present invention;

[0064] Figure 6This is a monitoring distribution map of the water body inundation range and transmission pole equipment in Jilin City, Jilin Province in 1994 for an automatic extraction method of the flood inundation range of power grid equipment based on the water body index of the present invention. Detailed implementation manners

[0065] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation manners.

[0066] Extracting power grid equipment flood disaster data based on satellite remote sensing generally includes three parts: data preparation and processing, inundation information extraction, and thematic element production. First, collect remote sensing image data according to requirements and process the collected images; second, extract the flood inundation information of power grid equipment based on the water body index extraction algorithm for the processed data, and verify and analyze the extraction results; finally, produce thematic map data to form a data analysis report on power grid equipment flood disasters. The overall technical flow chart is shown in Figure 1 .

[0067] Data collection and processing include:

[0068] (1) Remote sensing image collection

[0069] The collection and arrangement of remote sensing images are the first step in the successful implementation and development of the project. It includes two stages: the collection of remote sensing image data and the inspection of data quality.

[0070] ① Data collection

[0071] Power grid equipment is widely distributed, and remote sensing image data during the flood period from multiple sources need to be used, such as Sentinel series satellite images, high-resolution series satellite images in our country, and unmanned aerial vehicle (UAV) remote sensing images. The collection content includes remote sensing image data that fully covers the project area and meets the project's required resolution, such as optical and UAV remote sensing images; supplementary collection and arrangement of other available remote sensing image data and relevant basic data to support remote sensing image data processing to ensure that the remote sensing image data meeting the project resolution covers the entire area; after collecting relevant materials, conduct data inspection, and after confirmation, it is convenient for subsequent data processing.

[0072] ② Data arrangement

[0073] Conduct image quality inspection on the existing standard scene remote sensing image data, perform secondary correction on the data that does not meet the quality requirements, and generate a mapping file after ensuring that all data quality is qualified. Then, perform spatial overlay analysis on the mapping file and the regional boundary file to determine the area covered by each scene of data. Based on this, organize the image data according to the organizational structure directory of region - time phase - sensor.

[0074] (2) Remote sensing data quality inspection

[0075] After the collection of image data is completed, quality inspection should be carried out on the data to ensure that the data entered into the database meets the requirements. An image data processing software is used to conduct the inspection mainly by automatic human-computer interaction, and the results are recorded.

[0076] ① Inspection content

[0077] According to the information in the header file of the image, count and produce the image time-phase distribution map. Check whether the spatial resolution meets the requirements of the satellite remote sensing images for the project. Whether the spectral bands meet the requirements that the satellite remote sensing images should at least include 4 bands of blue, green, red, and near-infrared, and can accurately reflect the main ground features such as vegetation, soil, water bodies, and buildings and the information with clear boundaries. Review whether the remote sensing images fully cover the project implementation scope, with no missing corners or gaps in the shooting area. Check whether there are phenomena such as cloud and snow coverage in the image coverage area, and whether the total cloud amount meets the requirements. Check that the satellite images of different strips should have an overlapping area of 10%. Check that the side view angle of the satellite image results is not greater than 30 degrees. Check whether the image has rich levels, clear texture details, normal tone, no obvious noise, spots, bad lines, seams, and deformations, and there is no abnormal highlight in the urban buildings.

[0078] ② Remote sensing data quality inspection and processing

[0079] For the image data collected within the remote sensing monitoring range of the project area, mainly including the optical and unmanned aerial vehicle remote sensing data source images within the project area. According to the requirements of the remote sensing monitoring of the project area, conduct quality inspection on the original data, and remove the areas covered by unqualified data; process and analyze the data that meet the requirements.

[0080] ③ Integrity inspection

[0081] Decompress the original satellite image compression package and check whether the image data, RPC files, XML meta-files, etc. are missing and whether the files are readable.

[0082] Check the orthorectification satellite orbit parameters or RPC parameters and other files, and the data format is a common standard format such as IMG.

[0083] ④ Quality inspection

[0084] Check the quality of the original image and the cloud and snow coverage situation, and feedback the unqualified images to the project organizing and implementing unit. The main inspection contents include:

[0085] Check the cloud, snow, and fog amount in units of scenes. Whether the cloud and snow coverage area in each scene of data is less than 20%, and the cloud and snow cannot fall on the key monitored ground features.

[0086] Check whether the spatial resolution of the original image is better than 2m, whether it has panchromatic and multispectral data, and whether the spectrum has band information such as blue, green, red, and near-infrared.

[0087] Check whether there are problems such as dropped lines, bad lines, missing bands, striping, speckle noise, and flares in the original image, and use the highest quality remote sensing image data as much as possible.

[0088] Check whether the original image is clear, whether the ground features are distinct, and whether the image tone is uniform.

[0089] Check whether the image header files are complete, including checking other information such as the image shooting time, sensor type, solar altitude angle, solar radiation angle, and central point longitude and latitude.

[0090] While ensuring that each interpretation data can fully cover the distribution area of the monitoring area, use a single data source as much as possible.

[0091] Other checks, such as checking whether the side view angle is less than 25° in plain areas and less than 20° in mountainous areas, and checking whether there is an overlapping area of not less than 10% between adjacent scenes, and not less than 5% in special cases.

[0092] ⑤ Reception time phase check

[0093] Check whether the reception time phase of the original image meets the project requirements, and whether each phase of the image meets the time range required by the user.

[0094] ⑥ Original image processing

[0095] For remote sensing data sources with poor quality, such as too large side view angles or too large cloud and snow coverage areas, timely feedback to the project organization and implementation unit to determine the processing and solution methods.

[0096] The original data inspection is carried out on a scene-by-scene basis. Use remote sensing image processing software to open the image data, and use the method of manual visual inspection to conduct quality inspection on each scene of data and make written records.

[0097] ⑦ Special data processing

[0098] For multi-source and multi-scale remote sensing data, the processing methods are different. Select the processing method suitable for the original data for remote sensing monitoring processing. For data with good synchronization performance of each band, select appropriate bands for band combination, and select images with better combination effects for color and brightness processing. For data with mismatched bands, during the image processing process, band registration can be carried out first.

[0099] (3) Optical data generation

[0100] The orthorectification is performed on the panchromatic and multispectral images of the received single - piece satellite image data respectively. Then, the corrected panchromatic image is used to perform georegistration on the multispectral image. The registered panchromatic and multispectral images are subjected to image fusion. The fused image is color - homogenized and output using professional software. Finally, image mosaicking is carried out. The mosaic line should keep the main features intact as much as possible, and the image texture should be naturally connected, achieving a balanced and consistent rendering color tone and no obvious seam marks, and generating an orthoimage (DOM) product.

[0101] Optical remote sensing image processing includes the processing of medium - resolution, meter - level and other optical remote sensing data. The processing process mainly includes radiometric calibration, orthorectification, panchromatic and multispectral image registration, panchromatic and multispectral image fusion, true - color output of the fused image, color - homogenization and mosaicking of the image, and final result cutting and arrangement, etc. Among them, for medium - resolution image data such as Landsat - 5, Landsat - 7, and Landsat - 8 image data, only radiometric calibration and band combination operations are required; for GF - 1 16 - meter data, since its data only contains multispectral data, operations such as panchromatic and multispectral image registration and panchromatic and multispectral image fusion are not required. The remaining meter - level optical remote sensing images, because they all include high - resolution panchromatic images and four - band multispectral image data, need to be processed such as panchromatic and multispectral image registration and panchromatic and multispectral image fusion.

[0102] The extraction of inundation information uses a water - body index inversion model. The water - body index inversion model is based on the spectral differences between water bodies and other features, including the Normalized Difference Water Index (NDWI) and the Modified Normalized Difference Water Index (MNDWI). The pre - processed remote sensing image is calculated to highlight the features of water bodies in the image for the extraction and identification of water - body information in remote sensing images. The input of the water - body index inversion model is optical surface reflectance data, and the output is a water - body index image in raster format. According to the statistical characteristics of the water - body index image, an adaptive threshold segmentation method is used to divide the water - body index image into two parts: water bodies and non - water bodies.

[0103] To meet the requirements of different application scenarios, the model constructs multiple water - body indices based on multi - source remote sensing images. The specific process includes three parts: construction of the water - body index model, quality inspection and accuracy verification, and extraction of water surfaces, as Figure 2 shown.

[0104] (1) Construction of the water - body index model

[0105] Based on the analysis of the multi - band spectral characteristics of water bodies, multiple bands closely related to water - body identification are selected, the mapping relationship between water bodies and remote - sensing spectral values is analyzed, and a mathematical model of the water - body index is constructed, which can more accurately characterize the water content information of features. The commonly used water - body indices are as follows:

[0106] ①Normalized Difference Water Index (NDWI)

[0107] Based on the strong reflection and absorption characteristics of water body information in the green light band and the near-infrared band, and the fact that the reflectance of vegetation is generally the strongest in the near-infrared band, the ratio of the green light band to the near-infrared band is used to suppress vegetation information to the greatest extent, so as to highlight water body information. However, the spectral characteristics of soil / buildings in the green light and near-infrared bands are almost the same as those of water bodies, that is, the reflectance in the green light is higher than that in the near-infrared band and some have a large contrast, so the model is prone to overestimate the identification of water areas with more building backgrounds. Its calculation formula is as follows:

[0108]

[0109] In the formula, ρ GREEN represents the brightness value or reflectance of the green light band, and ρ NIR represents the brightness value or reflectance of the near-infrared band.

[0110] ②Modified Normalized Difference Water Index (MNDWI)

[0111] MNDWI utilizes the characteristic that the reflectance of buildings suddenly increases from the near-infrared to the short-wave infrared, while the water body continuously decreases. It replaces the near-infrared band in NDWI with the short-wave infrared band, improving the difference between buildings and water bodies and greatly reducing background noise. Therefore, MNDWI can reveal the fine features of water bodies, such as the distribution of suspended sediments, and the normalized ratio operation can easily distinguish shadows and water bodies, solving the problem of difficult elimination of shadows in water body extraction. Its calculation formula is as follows:

[0112]

[0113] In the formula, ρ SWIR represents the brightness value of the mid-infrared band.

[0114] ③Multi-Band Water Index (MBWI)

[0115] As the wavelength increases from the visible light band to the infrared band, the water surface reflectance also shows a decreasing trend, while the non-water surface reflectance shows no obvious trend. In addition, the maximum reflectance of surface water appears in the visible light band, while the maximum reflectance of non-water appears in the infrared band. Based on a full analysis of the differences between water and other low-reflectance surfaces, the MBWI water index is constructed. It can effectively suppress non-water features in mountain shadows and dark built-up areas, enhance surface water information, and reduce seasonal and daily impacts caused by changes in solar conditions. The calculation formula is as follows:

[0116] MBWI = 2×ρ Green -ρ Red -ρ NIR -ρ SWIR1 -ρ SWIR2

[0117] In the formula, ρ Red represents the brightness value in the red light band, and ρ SWIR1 , ρ SWIR2 are the reflectances of two short-wave infrared bands.

[0118] ④ Multi-Band Combined Water Index (MBCWI)

[0119] Taking advantage of the fact that the reflectance of water gradually decreases as the wavelength increases from the visible light to the mid-infrared band, and in the near-infrared and mid-infrared bands, the reflectance of water is significantly lower than that of other ground objects, the MBCWI index is constructed, which can better suppress the influence of mountain shadows, building shadows, clouds, bare soil, and some low-reflectance objects. The calculation formula is as follows:

[0120] MBCWI = 2×(ρ NIR -ρ SWIR1 ) - (ρ Blue +ρ Green +ρ Red )

[0121] In the formula, ρ Blue represents the brightness value or reflectance in the blue light band.

[0122] ⑤ Automated Water Extraction Index (AWEI)

[0123] The arithmetic combination of AWEI's medium spectral bands is determined based on a strict examination of the reflectance characteristics of various land cover types, maximizing the separability of water and non-water pixels through band differencing, addition, and application of different coefficients. The algorithm model coefficients are empirical results determined based on the reflectance patterns observed in the pure pixel datasets of various land cover types. The ultimate goal is to enhance the separability of water and dark surfaces, such as shadows and building structures, so that it can improve the identification effect of surface water in areas including deep shadows and dark surfaces caused by mountainous terrain. The calculation formula is as follows:

[0124] AWEI nsh = 4 × (ρ Green - ρ SWIR1 ) - (0.25 × ρ NIR + 2.75 × ρ SWIR2 )

[0125] AWEI sh = ρ Blue + 2.5 × ρ Creen - 1.5 × (ρ NIR + ρ SWIR1 ) - 0.25 × ρ SWIR2

[0126] "nsh" is used to specify that the index applies to cases where shadows are not a major problem. The subscript "sh" indicates that the formula is designed to effectively eliminate shadow pixels and improve the water extraction accuracy in shaded and / or other dark surface areas.

[0127] (2) Quality inspection and accuracy verification

[0128] According to the above synthesis principle, information on the data quality related to dekadal, monthly, quarterly, and annual synthesis periods is stored in one band. The synthesized data consists of all the original data corresponding to the selected pixels. According to the value range corresponding to different water body indices, outlier values are screened and uniformly replaced.

[0129] The accuracy inspection of the water body index is mainly verified by comparing with high-spatial-resolution remote sensing data of the same period. Representative regional quadrats of different water areas are selected to verify the reliability of the water body index product.

[0130] (3) Water body extraction based on the Otsu method (OSTU)

[0131] The OSTU method uses the idea of clustering to divide the gray levels of an image into two parts according to the gray level, so that the gray value difference between the two parts is the largest, and the gray difference within each part is the smallest. A suitable gray level for division is found through variance calculation. Suppose the threshold range of the NDWI or MNDWI image is from x to y, where -1 ≤ x < y ≤ 1. Based on the threshold t obtained by the OSTU algorithm, the NDWI or MNDWI image can be divided into two categories: water body (from t to y) and non-water body (from x to t). The calculation of the optimal threshold t of the OSTU algorithm is as follows:

[0132]

[0133] where δ for the water body is the between-class variance of the water body and the non-water body, P w and P nw are the probabilities that a single pixel belongs to the water body and the non-water body respectively, M w and M nw are the averages of the water body and the non-water body, and M is the average of the NDWI or MNDWI image.

[0134] After obtaining the optimal threshold through calculation, the threshold method is directly used for classification to obtain the water body extraction result.

[0135] (4) Accuracy verification of the water body extraction result

[0136] The main purpose of carrying out the accuracy verification of the water body extraction result is to evaluate the water body recognition result obtained by the intelligent algorithm, and appropriately optimize it through manual post-processing methods, so as to achieve the water body recognition result that meets the accuracy requirements of users. Based on the verification sample points within the selected range, by calculating evaluation indicators and analyzing the differences between the reference sample and the sample to be evaluated, the accuracy evaluation of the water body extraction result is realized. When using remote sensing data to extract water bodies, the manual extraction method has the highest accuracy. Assuming that the water bodies extracted in the target area are accurate, the reference sample is set as the result of manual water body extraction and the sample data made manually; the sample to be evaluated is the water body result automatically extracted by the water body extraction algorithm. Select the accuracy verification area, and perform overlay analysis on the reference sample and the sample to be evaluated within the verification area for verification and evaluation.

[0137] The verification and evaluation indicators mainly include Accuracy, TPR, FPR, Recall, Precision, F-score, MAP, ROC curve, and AUC, etc. The calculation of the indicators is mainly based on the confusion matrix generated by the overlay analysis of the reference sample and the sample to be evaluated, and the indicator parameters are shown in Table 1.

[0138] Table 1 Indicator parameters

[0139] Index parameter Explanation TP (True Positive) Indicates that the sample is actually a positive class and the prediction result is a positive class FP (False Positive) Indicates that the sample is actually a negative class and the prediction result is a positive class TN (True Negative) Indicates that the sample is actually a negative class and the prediction result is a negative class FN (False Negative) Indicates that the sample is actually a positive class and the prediction result is a negative class

[0140] ① Accuracy: It measures the proportion of correct classifications. Its calculation formula is simple and direct, and the expression method is as follows:

[0141]

[0142] Among them, TP: True Positive, the number of positive classes predicted as positive classes; TN: True Negative, the number of negative classes predicted as negative classes; FP: False Positive, the number of misreports of negative classes predicted as positive classes; FN: False Negative, the number of missed reports of positive classes predicted as negative classes.

[0143] ② Precision: Also known as the precision rate, it refers to the proportion of correctly predicted ones among the detection frames predicted as positive samples, and is expressed as follows:

[0144]

[0145] ③ Recall: Also known as the recall rate, it is for the original samples. Its meaning is the probability of being predicted as positive samples among the actually positive samples, and is expressed as follows:

[0146]

[0147] ④ F1 score: Sometimes it is necessary to balance between precision and recall. One option is to draw a precision-recall curve (Precision-Recall Curve), and the area under the curve is called the AP score (Average precision score); another option is to calculate the F β score.

[0148]

[0149] When β = 1, it is called the F1 score. F1 is the harmonic mean of precision and recall. The F1 score takes into account both the precision rate and the recall rate, making both reach the highest at the same time and achieving a balance. When β > 1, the weight of recall is higher than that of precision. On the contrary, when β < 1, the weight of precision is higher than that of recall.

[0150] ⑤ G score is another system performance evaluation standard that unifies precision and recall. The F score is the harmonic mean of accuracy and recall, while the G score is defined as the geometric mean of accuracy and recall. It is expressed as follows:

[0151]

[0152] ⑥AP and MAP: AP measures the quality of the learned model for each category, and MAP (Mean Average Precision) measures the quality of the learned model for all categories. AP measures the quality of the system for a single category, and MAP measures the quality of the learned model for all categories. After obtaining the AP for each category, MAP is the average of all APs.

[0153] ⑦ROC curve: Also known as the Receiver Operating Characteristic curve, the two main indicators in the ROC curve are the True Positive Rate (TPR) and the False Positive Rate (FPR). The abscissa is the False Positive Rate (FPR), and the ordinate is the True Positive Rate (TPR), which are expressed as follows:

[0154]

[0155] Power grid equipment is vulnerable to damage during flood disasters, which may lead to power outages and other situations that affect people's production and life. To address the issues of unclear flood risk base and unknown degree of power grid equipment, the superimposed analysis of the segmented water body range and the location information of power grid equipment is carried out to extract the range of power grid equipment flooded by floods and display it in a visual way to generate a thematic map. Specifically, by combining GIS and historical remote sensing images, a basic geographic information database of power facilities and a flood inundation factor database are established and spatially correlated to quickly evaluate the inundation range, water depth, duration, and losses of power grid equipment under different scenarios of breaches, overflows, positions, and scales in different years, draw a "single map" of flood risk zoning, and accurately grasp the flood control safety situation of power facilities. This can guide power grid enterprises to take corresponding protective measures, such as strengthening equipment and transferring equipment, in areas with different risk levels before disasters occur, thereby improving the safety and stability of power grid equipment, reducing the likelihood and harm of disasters, and effectively ensuring the safety of people's lives and property.

[0156] Based on WebGIS, a flood disaster loss assessment system for power grid facilities is established. By combining real-time remote sensing image data during the flood season, the flood disaster situation is analyzed and evaluated in real time and quickly. The distribution laws of elements such as the coverage range, working condition duration, and flood disaster losses of the power grid basin flood disaster in space and time are demonstrated. By drawing the flood risk zoning map of power grid equipment, the flood risk levels faced by power grid equipment in different regions can be clearly shown, providing targeted flood control and disaster reduction strategies and measures for relevant institutions, and thus providing a scientific and powerful basis for the flood control and disaster reduction work of power grid equipment.

[0157] Taking Jilin Province as an example, the specific implementation process of the present invention is elaborated in detail for the key links of the invention:

[0158] (1) Data collection and processing

[0159] Landsat satellite imagery data for the flood season (June - September) from 1994 to 2024 in the whole of Jilin Province were obtained, including Landsat 4 - 5 and Landsat 8 - 9. Meanwhile, geographical location information data of power grid equipment in Jilin Province were collected from Jilin Electric Power Research Institute Co., Ltd.

[0160] (2) Flood inundation information extraction

[0161] An automatic extraction algorithm for the flood inundation range of power grid equipment based on the water body index was used. According to the actual situation of each city in Jilin Province and the characteristics of the image data, parameters for calculating the water body index, threshold segmentation, etc. were reasonably set. For example Figure 3 As shown, taking the Lalin River as an example, remote sensing image data under the flood scenario in 2023 were selected for algorithm verification. By comparing with the actual flood inundation situation, the accuracy and reliability of the algorithm were verified.

[0162] (3) The production of thematic elements

[0163] As shown in Table 2, taking 1994 as an example, from the results of flood monitoring changes in Jilin Province, the top five affected cities are Baicheng City, Songyuan City, Jilin City, Changchun City, and Siping City.

[0164] Table 2 Monitoring change results of flood disasters in Jilin Province from June to September in 1994

[0165]

[0166] In 1994, the water inundation area in Jilin City, Jilin Province was 712.80 square kilometers, as Figure 4 shown.

[0167] In 1994, the number of substations in Jilin City, Jilin Province within the water inundation range was 831, as Figure 5 shown.

[0168] In 1994, the number of transmission poles in Jilin City within the water inundation range was 3, as Figure 6 shown.

[0169] The implementation mode of the present invention is not limited by the above - mentioned embodiments. Any changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. An automatic extraction method for the flood inundation range of power grid equipment based on a water body index, characterized in that: It includes the following steps: Step 1: Data acquisition and processing: Collect satellite remote sensing image data according to requirements and process the collected images; Step 2: Flooding information extraction: Extract the flood inundation information of power grid equipment from the processed data based on the water body index extraction algorithm, and verify and analyze the accuracy of the extraction results; Step 3: Thematic element production: Produce thematic map data and form a data analysis report on the flood disasters of power grid equipment.

2. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 1, characterized in that: The data acquisition and processing in Step 1 include remote sensing image data collection, remote sensing image data quality inspection, and optical data generation.

3. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 2, characterized in that: The remote sensing image data collection includes data collection and data arrangement: The collection content includes remote sensing image data that fully covers the project area range and meets the project requirement resolution, such as optical and unmanned aerial vehicle remote sensing images, supplemented by the collection and arrangement of other available remote sensing image data and relevant basic data to support remote sensing image data processing, ensuring that the remote sensing image data that meets the project resolution covers the entire area range; Conduct image quality inspection on the existing standard scene remote sensing image data, perform secondary correction on the data that does not meet the quality requirements, ensure that all data quality is qualified and then generate a landing map file, perform spatial overlay analysis on the landing map file and the regional range boundary file, judge the area covered by each scene of data, and based on this, organize the image data according to the organizational structure directory of region range - time phase - sensor.

4. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 2, wherein: The remote sensing image data quality inspection uses image data processing software, and is inspected mainly by an automatic human - machine interaction method, and the results are recorded. Check whether the spatial resolution of the satellite remote sensing image meets the project requirements, whether the spectral bands of the satellite remote sensing image meet the requirement that it should at least include 4 bands of blue, green, red, and near - infrared, and can accurately reflect the main ground objects such as vegetation, soil, water body, and buildings and the information with clear boundaries. Review whether the remote sensing image fully covers the project implementation scope, the shooting area, without missing corners or gaps, check whether there is cloud or snow coverage in the image coverage area, whether the total cloud amount meets the requirements, check that different strips of satellite images should have a 10% overlapping area, check that the side view angle of the satellite image result is not greater than 30 degrees, check whether the image is rich in layers, clear in texture details, normal in tone, without obvious noise, spots, bad lines, seams, and deformations, and there is no abnormal highlight in urban buildings.

5. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 2, wherein: The optical data generation performs orthorectification on the panchromatic and multispectral images of the received single - chip satellite image data respectively, then uses the corrected panchromatic image to perform georegistration on the multispectral image, fuses the registered panchromatic and multispectral images, performs color homogenization output on the fused image using professional software, and finally performs image mosaicing. The mosaic line should make the main ground objects complete as much as possible, and the image texture should be naturally connected, achieving a balanced and consistent rendering tone and no obvious seam marks, and generating an orthophoto image (DOM) product.

6. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 1, characterized in that: In step 2, the inundation information extraction uses a water body index inversion model. The water body index inversion model is based on the spectral differences between water bodies and other ground objects, including the Normalized Difference Water Index (NDWI) and the Modified Normalized Difference Water Index (MNDWI). It calculates the preprocessed remote sensing images to highlight the characteristics of water bodies in the images, and is used for the extraction and identification of water body information in remote sensing images. The input of the water body index inversion model is optical surface reflectance data, and the output is a water body index image in raster format. According to the statistical characteristics of the water body index image, an adaptive threshold segmentation method is used to divide the water body index image into two parts: water bodies and non-water bodies.

7. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 6, characterized in that: The water body index inversion model constructs multiple water body indices based on multi-source remote sensing images. The specific process includes water body index model construction, quality inspection and accuracy verification, and extraction of water surfaces. (1) Water body index model construction Based on the analysis of the spectral characteristics of water bodies in multiple bands, multiple bands closely related to water body identification are selected, and the mapping relationship between water bodies and remote sensing spectral values is analyzed to construct a mathematical model of the water body index: ① Normalized Difference Water Index (NDWI) where ρ GREEN represents the brightness value or reflectivity in the green light band, and ρ NIR represents the brightness value or reflectivity in the near-infrared band; ② Modified Normalized Difference Water Index (MNDWI) where ρ SWIR represents the brightness value in the mid-infrared band; ③ Multi-Band Water Index (MBWI) The MBWI water body index is constructed based on a full analysis of the differences between water and other low-reflectance surfaces. It can suppress non-water features in mountain shadows and dark built-up areas, enhance surface water information, and reduce seasonal and daily impacts caused by changes in solar conditions. The calculation formula is as follows: MBWI = 2×ρ Green -ρ Red -ρ NIR -ρ SWWIR1 -ρ SWIR2 where ρ Red represents the brightness value in the red light band, and ρ SWIR1 , ρ SWIR2 are the reflectivities of two short-wave infrareds; ④ Multi-Band Combined Water Index (MBCWI) Using the characteristics that the reflectance of water bodies gradually decreases with the increase of wavelength in the visible to mid-infrared bands, and the reflectance of water bodies is significantly lower than that of other ground objects in the near-infrared and mid-infrared bands, the MBCWI index is constructed, which can suppress the influences of mountain shadows, building shadows, clouds, bare soil, and some low-reflectance objects. The calculation formula is as follows: MBCWI = 2×(ρ NIR - ρ sWIR1 ) - (ρ Blue + ρ Green + ρ Red ) where ρ Blue represents the brightness value or reflectance in the blue light band; ⑤ Automated Water Extraction Index (AWEI) The arithmetic combination of spectral bands in AWEI is determined based on a strict examination of the reflectance characteristics of various land cover types. By band differencing, adding, and applying different coefficients, the separability of water and non-water pixels is maximized. The algorithm model coefficients are empirical results determined based on the reflectance patterns observed in pure pixel datasets of various land cover types. The ultimate goal is to enhance the separability of water and dark surfaces, such as shadows and building structures, and improve the identification effect of surface water in areas including deep shadows and dark surfaces caused by mountainous terrain. The calculation formula is as follows: AWEI nsh = 4×(ρ Green - ρ SWIR1 ) - (0.25×ρ NIR + 2.75×ρ sWIR2 ) AWEI sh = ρ Blue + 2.5×ρ Green - 1.5×(ρ NIR + ρ SWIR1 ) - 0.25×ρ SWIR2 "nsh" is used to specify that the index is applicable to cases where shadow is not a major issue. The subscript "sh" indicates that the formula is designed to effectively eliminate shadow pixels and improve the accuracy of water extraction in shaded and / or other dark surface areas. (2) Quality inspection and accuracy verification According to the above synthesis principle, information on the data quality of dekadal, monthly, quarterly, and annual synthesis periods is stored in one band. The synthesized data is composed of all the original data corresponding to the selected pixels. According to the value range corresponding to different water indices, outlier values are screened and uniformly replaced. The accuracy inspection of water indices is mainly verified by comparing with remotely sensed data with high spatial resolution in the same period. Representative regional samples with different water body sizes are selected to verify the reliability of water index products. (3) Water extraction based on the Otsu method (OSTU) The OSTU method uses the idea of clustering to divide the gray levels of an image into two parts according to the gray level, so that the gray value difference between the two parts is the largest, and the gray difference within each part is the smallest. By calculating the variance, a suitable gray level is found for division. Suppose the threshold range of the NDWI or MNDWI image is from x to y, where -1 ≤ x < y ≤ 1. The threshold t obtained based on the OSTU algorithm can divide the NDWI or MNDWI image into two categories: water bodies (from t to y) and non-water bodies (from x to t). The calculation of the optimal threshold t of the OSTU algorithm is as follows: where the water body δ is the between-class variance of the water body and the non-water body, P w and P nw are the probabilities that a single pixel belongs to the water body and the non-water body respectively, M w and M nw are the average values of the water body and the non-water body, and M is the average value of the NDWI or MNDWI image; After obtaining the optimal threshold, the threshold method is directly used for classification to obtain the water extraction result. (4) Accuracy verification of water extraction results The reference samples are set as the results of manual water extraction and the sample data made manually; the samples to be evaluated are the water body results automatically extracted using the water extraction algorithm. The verification evaluation indicators mainly include accuracy, precision, recall, F1 score, G score, AP and MAP, and ROC curve.

8. The automatic extraction method for the flood inundation range of power grid equipment based on the water body index according to claim 1, characterized in that: The thematic feature production in step 3 mentioned above is the range extraction: The superimposed analysis is carried out on the segmented water body range and the location information of power grid equipment to extract the range of power grid equipment flooded by floods and display it in a visual way to generate a thematic map. Combining GIS and historical remotely sensed images, a basic geographic information database of power facilities and a flood inundation factor database are established, and spatial association is carried out to quickly evaluate the flooded range, water depth, duration, and losses of power grid equipment under different scenarios of different breach and overtopping positions and scales in different years, and draw a "one map" of flood risk zoning.

Images