Method for improving precision of remote sensing monitoring of river surface characteristic parameters based on hydrological knowledge

Abstract

The application discloses a remote sensing monitoring river water surface feature parameter precision improving method based on hydrological knowledge, which comprises the following steps: extracting a river channel and a ridge line by using DEM, determining a valley+river channel range, and accurately cutting a remote sensing image; using a water body index to set a threshold interval to clearly define a preliminary water surface, a mixed area and a non-water body range; then, combining a remote sensing image original resolution, constructing a high-resolution NDWI grid map, quantitatively estimating a water body proportion in a mixed pixel by analyzing NDWI values of surrounding pixels, and improving water body identification precision in the mixed pixel; subsequently, using hydrological knowledge to divide the river channel into a single flow path or a multi-flow path structure, and accurately judging a water body distribution position in the mixed pixel in combination with a confluence accumulation number; finally, merging the preliminary water surface and the water surface refined based on the hydrological knowledge to form a high-precision water surface vector file, and extracting water surface feature parameters; and the application improves identification precision of water body boundaries and feature parameters in small rivers and complex landforms.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image information extraction, and in particular to a method for improving the accuracy of remote sensing monitoring river surface feature parameters based on hydrological knowledge. Background Technology

[0002] Remote sensing technology has become a key means of water body information extraction, especially with the support of high-resolution satellite remote sensing, and is widely used in the identification and measurement of surface features of various water bodies such as rivers and lakes. Currently, various remote sensing water body extraction methods have been developed and applied, mainly including thresholding methods (such as the widely used water index method, which uses specific band ratios to emphasize water body features and simplify the extraction process), classifier methods (such as decision tree methods and object-oriented methods), and other complex extraction techniques based on principles such as multi-band spectral relationships and spectral knowledge. These methods use pixels as the basic unit for water body extraction, resulting in a relatively coarse judgment of mixed pixels, which leads to a loss of accuracy in water body boundary identification. Mixed pixel decomposition technology can specifically address the problem of mixed pixels near water body boundaries, improving the accuracy of water body extraction through quantitative analysis.

[0003] Despite continuous improvements in remote sensing resolution, even high-resolution imagery can be limited by spatial resolution when dealing with complex scenes such as narrow rivers, making it difficult to accurately identify the boundaries of small water bodies. This is especially true during the dry season, when the spectral complexity caused by the shrinking water area further complicates accurate detection. Furthermore, existing research often focuses on foreign remote sensing products and areas with relatively simple land cover types, or large water bodies. The extraction methods used, such as water body and vegetation indices, often neglect factors like actual river flow, resulting in extracted river width information that fails to fully reflect hydrological characteristics such as characteristic flow and flatland flow. In the extraction of small water bodies in complex terrains such as mountainous areas, the problem of mixed pixels can lead to reduced accuracy in water body boundary identification due to the severe issue of mixed pixel extraction. Therefore, future research urgently needs to develop more refined mixed pixel decomposition techniques and water body extraction methods adapted to complex geographical conditions to improve the accuracy and completeness of water body information extraction under various sizes and terrain conditions. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a method for improving the accuracy of remote sensing monitoring of river surface characteristic parameters based on hydrological knowledge, so as to improve the accuracy and completeness of water body information extraction under various sizes and topographic conditions.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for improving the accuracy of remote sensing monitoring of river water surface characteristic parameters based on hydrological knowledge, which includes the following steps:

[0006] S1: Based on the DEM, extract the river channel and the ridgeline on both sides of the river channel to determine the valley range;

[0007] S2: Use vector files of ridgelines on both sides of the river to crop the remote sensing image to obtain remote sensing images within the river channel and valley area;

[0008] S3: Calculate the acquired remote sensing images of the river channel and valley using the water body index; determine the water surface 1, the water surface plus other features, and the range of other features;

[0009] S4: Determine the proportion of water in the mixed pixels within the water body + other land features area based on the NDWI values ​​of the surrounding pixels;

[0010] S5: Generate the river centerline based on the initial water surface 1, detect the cross-section in the vertical direction of the river centerline, and determine the water distribution location in single flow path, multi-flow path and mixed pixels based on the NDWI value changes of pixels on the cross-section.

[0011] S6: Based on the proportion and location of water bodies in the mixed pixels determined by S4 and S5, and the NDWI value of the original resolution a, create a high-resolution NDWI raster map of the corresponding location and extract the water surface 2.

[0012] S7: Use the new water surface file to obtain water surface feature parameters.

[0013] Preferably, the specific process of S1 is as follows:

[0014] First, the flow direction of each grid is obtained on the depression-free DEM using the maximum slope method. Then, the number of grids accumulated in the flow direction for each grid is calculated using the flow direction grid data, i.e., the flow accumulation. The obtained flow accumulation represents how many grids have flow directions passing through a grid at a grid location. The flow accumulation of a grid represents the water flow of that grid. A threshold is set to distinguish river channels from other smaller ditches or drainage paths, and areas with larger flow accumulation are selected as river networks.

[0015] Based on the DEM, depression filling is performed through neighborhood analysis to eliminate the influence of depressions and generate a depression-free DEM. Hydrological analysis is then performed using the depression-free DEM to calculate water flow direction data, reflecting the slope of the terrain and the water flow path. The runoff accumulation is calculated using the water flow direction data. This index reflects the total amount of water received by each pixel. When it is zero, it indicates that the point is located at the watershed. Finally, the intersection of the points with zero runoff accumulation and the positive terrain is obtained to obtain the ridgeline.

[0016] Preferably, the specific process of S2 is as follows:

[0017] The image data was subjected to radiometric calibration, geometric correction, atmospheric correction, and image fusion to obtain preprocessed image data. For each identified flow path, the remote sensing image was cropped using the vector file of the ridgeline on both sides of the river channel to remove background noise, reduce the study area, and obtain remote sensing images of only the river channel and valley.

[0018] Preferably, the specific process of S3 is as follows:

[0019] The water index map is obtained from the preprocessed image data using the normalized water index method, as shown in the following formula:

[0020]

[0021] Wherein, NDWI is the normalized water index, G is the green band data in the preprocessed image data, and NIR is the near-infrared band data in the preprocessed image data;

[0022] In the preprocessed remote sensing image, n samples of each land feature are selected to determine the maximum and minimum values ​​of each land feature in the green band and near-infrared band.

[0023] Determine the threshold range for NDWI for water bodies, bare soil, vegetation, and rocky surface cover;

[0024] This determines the threshold ranges for initial water surface 1, water surface + other land features, and other land feature categories. The overlapping part of the threshold ranges of water bodies and other land features is defined as the threshold range of water surface + other land features, and the non-overlapping part is defined as pure land features.

[0025] Initial water level 1:

[0026] Water surface + other land features:

[0027] Other land features:

[0028] Where V represents the set of pixels in the NDWI raster image, I ij Let I be the water index value of the pixel in the i-th row and j-th column of the raster image. Pmin I Pmax Let I be the maximum and minimum thresholds for segmenting pure water bodies, respectively. Mmax The maximum threshold for non-water body pixel segmentation is determined separately; W1 is the initial water surface range, M is the range where water bodies and other land features are mixed, and N is the range of other land features.

[0029] Preferably, the specific process of S4 is as follows:

[0030] First, the land cover to which the neighboring pixels of the mixed pixel (i,j) belong is determined based on the threshold range of water bodies, bare soil, vegetation, and rock surface cover in the green band, near-infrared band, and NDWI. This determines the types of land cover other than water bodies contained in the mixed pixel.

[0031] The NDWI value of the mixed pixel and the NDWI value of the neighboring pixel show a linear superposition relationship. The relationship expression is constructed as follows:

[0032]

[0033] Among them, I ij For the NDWI value of the mixed pixel, I W I N These represent the NDWI values ​​of water bodies and non-water bodies in the neighborhood pixels, respectively, and P water (M), P nowater (M) represents the proportion of water and non-water bodies in the mixed pixel, and the sum of the two is 1.

[0034] Preferably, the specific process of S4 is as follows: Define a weight function w(I ij Construct an n×n filter kernel based on the NDWI values ​​of neighboring pixels (M). ij Weights are calculated based on the water body baseline and the NDWI value of the mixed pixel itself;

[0035]

[0036] Where α is a positive adjustment parameter used to control the steepness of the slope of the weighting function; max diff It is the maximum difference between the NDWI values ​​of non-water pixels in the dataset and the water body reference value, used to normalize the difference; the tanh function ensures that the weight value varies between -1 and 1. When the difference from the water body reference value is large, the weight value tends to -1; when the difference is small, the weight value tends to 1.

[0037] Determine the neighborhood of the intermediate mixed cell M, for each neighboring cell I of the mixed cell M. ij Calculate its NDWI value w(I) after adjustment by the weighting function. ij ,M)·I ij The weighted NDWI values ​​of all neighboring pixels are summed and divided by the sum of all weight values ​​to obtain an estimate P of the proportion of water in the mixed pixel M. water (M), whose theoretical formula is:

[0038]

[0039] Preferably, the specific process of S5 is as follows:

[0040] The acquired raster initial water surface 1 is converted into a water surface vector file, and the river centerline is generated by morphological skeleton extraction.

[0041] For each point along the river centerline, calculate the flow direction angle α at that point; based on this angle, determine the direction perpendicular to the centerline, i.e., α+90°, and determine the river cross-section;

[0042] Along the river cross-section, NDWI values ​​are extracted pixel by pixel. This index can effectively distinguish between water and land. Pixels are classified according to preset thresholds to determine the land cover types such as water, vegetation, and soil, and to identify the specific distribution of water in single-flow-path areas, multi-flow-path areas, and mixed pixels.

[0043] Preferably, the specific process of S6 is as follows:

[0044] Create a new a / n 2 For high spatial resolution raster files, the NDWI values ​​of the NDWI raster generated by S3 are written into the newly created raster file to ensure consistency with the spatial reference coordinate system of the original image.

[0045] Using the water body proportion data obtained from S4 blended pixel decomposition and the water body location determined in S5 blended pixels, the pixel values ​​in the high-resolution NDWI raster are adjusted to ensure that the processed NDWI values ​​are correctly mapped to the new high-resolution raster.

[0046] Water bodies were extracted again from the high-resolution NDWI raster image to obtain water surface 2.

[0047] Preferably, the specific process of S7 is as follows:

[0048] Based on the water surface vector file and the perpendicular river channel vector line file, calculate the river width and water surface characteristic parameters such as the length-to-width ratio in the river morphology;

[0049] The gradient of the river channel:

[0050]

[0051] In the formula, i is the gradient; h1 and h2 are the cross-sectional elevations; and L is the horizontal distance between the two points.

[0052] The ratio of the water surface area to the length of a river segment is used as the average width of that segment.

[0053]

[0054] In the formula, W is the generalized river section width; A is the water surface area of ​​the river section under study; and L is the length of the river section under study.

[0055] Beneficial effects of this invention:

[0056] 1. This invention can improve the accuracy and completeness of water body information extraction under various sizes and terrain conditions.

[0057] 2. This invention addresses the challenges of extracting long-term low-resolution images and small water bodies by integrating digital river network extraction technology, valley extraction technology, and remote sensing interpretation technology. This ensures the integrity of the extraction of boundaries for both small and regular water bodies and effectively improves the extraction accuracy of water body feature parameters such as water surface width. Attached Figure Description

[0058] Figure 1 A schematic diagram of land feature classification;

[0059] Figure 2 Diagram showing single-path / multi-path flow;

[0060] Figure 3 For high spatial resolution NDWI raster images;

[0061] Figure 4 This refers to the preprocessed image data;

[0062] Figure 5 Water body index chart;

[0063] Figure 6 Image 1 of the initial water surface;

[0064] Figure 7 Map showing the area containing water surface and other land features;

[0065] Figure 8 This is a map showing the proportion of water bodies in the mixed pixels within the water body + other land features area determined according to Method 2.

[0066] Figure 9 This is a map showing the location of water bodies in the mixed pixels in the high-resolution raster image.

[0067] Figure 10 This is a diagram showing the result of merging the initial water surface with the water bodies in the mixed pixels. Detailed Implementation

[0068] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0069] Example 1: S1: Based on DEM (Digital Elevation Model), extract the river channel and the ridgelines on both sides of the river channel to determine the valley range;

[0070] First, the flow direction of each grid is obtained on a depression-free DEM using the maximum slope method. Then, the number of grids accumulated in the flow direction for each grid is calculated using the flow direction grid data, i.e., the flow accumulation. The obtained flow accumulation represents how many grids have water flowing through that grid at a given grid location. The flow accumulation of a grid represents the water flow of that grid. An appropriate threshold is set to distinguish river channels from other smaller ditches or drainage paths, and areas with larger flow accumulation are selected as river networks.

[0071] Based on the DEM (Digital Equation Model), depression filling is performed using neighborhood analysis to eliminate the influence of depressions and generate a depression-free DEM. Hydrological analysis is then conducted using the depression-free DEM to calculate water flow direction data, reflecting the topographic slope and water flow path. The runoff accumulation is calculated using the water flow direction data; this indicator reflects the total water volume received by each pixel, and a value of zero indicates that the point is located at the watershed watershed. Finally, the intersection of points with zero runoff accumulation and the positive topographic features yields the ridgeline.

[0072] S2: Use vector files of ridgelines on both sides of the river to crop the remote sensing image to obtain remote sensing images within the river channel and valley area;

[0073] The image data was processed by radiometric calibration, geometric correction, atmospheric correction and image fusion to obtain preprocessed image data; for each identified flow path, the remote sensing image was cropped using the vector file of the ridgeline on both sides of the river to remove background noise, reduce the scope of the study area and obtain remote sensing images of only the river channel and valley.

[0074] S3: Calculate the river channel and valley area remote sensing imagery obtained in S2 using water body indices; determine the boundaries of water surface 1, water surface plus other features, and other features, see [link to relevant documentation]. Figure 1 ;

[0075] The water index map is obtained from the preprocessed image data using the normalized water index method, as shown in the following formula:

[0076]

[0077] Wherein, NDWI is the normalized water index, G is the green band data in the preprocessed image data, and NIR is the near-infrared band data in the preprocessed image data.

[0078] In the preprocessed remote sensing image, n samples of each land feature are selected to determine the maximum and minimum values ​​of each land feature in the green band and near-infrared band:

[0079]

[0080] Equations (1) and (2) can be used to determine the threshold range of NDWI for water bodies, bare soil, vegetation, and rock surface cover.

[0081] For example, the threshold range of water bodies:

[0082] Maximum value:

[0083] Minimum value:

[0084]

[0085] This determines the threshold ranges for categories such as initial water surface 1, water surface + other land features, and other land features. The overlapping part of the threshold ranges of water bodies and other land features is defined as the threshold range of water surface + other land features, and the non-overlapping part is defined as pure land features.

[0086] Initial water level 1:

[0087] Water surface + other land features:

[0088] Other land features:

[0089] Where V represents the set of pixels in the NDWI raster image, I ij Let I be the water index value of the pixel in the i-th row and j-th column of the raster image. Pmin I Pmax Let I be the maximum and minimum thresholds for segmenting pure water bodies, respectively. Mmax The maximum threshold for non-water body pixel segmentation is determined separately; W1 is the initial water surface range, M is the range where water bodies and other land features are mixed, and N is the range of other land features.

[0090] S4: Based on the NDWI (Normalized Water Index) values ​​of surrounding pixels, determine the proportion of water in the mixed pixels within the water body + other land features area in S3.

[0091] Method 1: First, determine the land cover to which the neighboring pixels of the mixed pixel (i,j) belong based on the threshold range of each land cover in the green band, near-infrared band and NDWI according to equations (2) and (3), thereby determining the land cover types excluding water bodies contained in the mixed pixel.

[0092] The NDWI value of the mixed pixel and the NDWI value of the neighboring pixel show a linear superposition relationship. The relationship expression is constructed as follows:

[0093]

[0094] Among them, I ij For the NDWI value of the mixed pixel, IW I N , respectively, are the NDWI values ​​of water and non-water bodies in the domain pixels, and Pwater(M) and Pnowater(M) are the proportions of water and non-water bodies in the mixed pixels, respectively, and their sum is 1.

[0095] Method 2: Define a weight function w(I) ij Construct a filter kernel of size n×n (e.g., 3x3) based on the NDWI values ​​of neighboring pixels. ij Weights are calculated based on the water body baseline and the NDWI value of the mixed pixel itself.

[0096]

[0097] Where α is a positive adjustment parameter used to control the steepness of the slope of the weighting function; max diff This is the maximum difference between the NDWI values ​​of non-water pixels in the dataset and the water body reference value, used to normalize the difference. The tanh function ensures that the weight value varies between -1 and 1; when the difference from the water body reference value is large, the weight value tends to -1; when the difference is small, the weight value tends to 1.

[0098] Determine the neighborhood of the intermediate mixed cell M, for each neighboring cell I of the mixed cell M. ij Calculate its NDWI value w(I) after adjustment by the weighting function. ij ,M)·I ij The weighted NDWI values ​​of all neighboring pixels are summed and divided by the sum of all weight values ​​to obtain the estimated proportion of water in the mixed pixel M, Pwater(M). The theoretical formula is as follows:

[0099]

[0100] S5: Generate the river centerline based on the initial water surface 1, detect the cross-section in the vertical direction of the river centerline, and determine the water distribution location in single flow path, multi-flow path, and mixed pixels based on the NDWI value changes of pixels on the cross-section.

[0101] The acquired raster initial water surface 1 is converted into a water surface vector file, and the river centerline can be generated by morphological skeleton extraction.

[0102] For each point along the river's centerline, calculate the flow direction angle α. Based on this angle, determine the direction perpendicular to the centerline, i.e., α + 90°, and thus determine the river cross-section.

[0103] Along the river cross-section, NDWI values ​​are extracted pixel by pixel. This index can effectively distinguish between water bodies and land. Pixels are classified according to a preset threshold formula (3) to determine the land cover types such as water bodies, vegetation, and soil, and to identify single-flow-path areas and multi-flow-path areas. Figure 2 (and the specific distribution of water bodies in the mixed pixels.)

[0104] S6: Based on the proportion and location of water bodies in the mixed pixels determined by S4 and S5, and the NDWI value of the original resolution a, create a high-resolution NDWI raster map of the corresponding location and extract the water surface 2.

[0105] Create a new a / n 2 For high spatial resolution raster files, the NDWI values ​​of the NDWI raster generated by S3 are written into the newly created raster file to ensure consistency with the spatial reference coordinate system of the original image.

[0106] Using the water body proportion data obtained from S4 blended pixel decomposition and the water body location determined in S5 within the blended pixels, the pixel values ​​in the high-resolution NDWI raster image are adjusted. This ensures that the processed NDWI values ​​are correctly mapped to the new high-resolution raster. Figure 3 ).

[0107] Water bodies were extracted again from the high-resolution NDWI raster image to obtain water surface 2.

[0108] S7: Use the new water surface file to obtain water surface feature parameters.

[0109] Based on the water surface vector file, the river width and other water surface characteristic parameters such as the length-to-width ratio in the river morphology are calculated using the river channel vector line file perpendicular to the river channel vector line file.

[0110] The gradient of the river channel:

[0111]

[0112] In the formula, i is the gradient; h1 and h2 are the cross-sectional elevations; and L is the horizontal distance between the two points.

[0113] The ratio of the water surface area to the length of a river segment is used as the average width of that segment.

[0114]

[0115] In the formula, W is the generalized river section width; A is the water surface area of ​​the river section under study; and L is the length of the river section under study.

[0116] Example 2:

[0117] This example uses 30m resolution Landsat 8 multispectral remote sensing image data to extract water surface information; the obtained Landsat 8 remote sensing image is cropped using the ridgelines on both sides of the river, and the cropped image data is preprocessed, mainly including radiometric calibration and atmospheric correction, to obtain the preprocessed image data, such as... Figure 4 ;

[0118] A water index map is obtained by calculating the preprocessed image data using the normalized water index method, such as... Figure 5 And determine the threshold ranges for categories such as initial water surface 1, water surface + other land features, and other land features; such as Figure 6 The initial water level is 1. Figure 7 The area consists of water surface and other land features.

[0119] Initial water surface 1: W1={(i,j)∈V|0 ij ≤0.201936};

[0120] Water surface + other land features: M = {(i,j)∈V|-0.15} ij ≤0};

[0121] Other land features: N={(i,j)∈V|I ij ≤-0.15,}; On the obtained water body index raster map of the water surface + other land features area, method two is used based on the NDWI value of neighboring pixels. ij The weights are calculated based on the water body reference value and the NDWI value of the mixed pixel itself. The NDWI value w(I) adjusted by the weighting function is then calculated. ij ,M)·I ij The weighted NDWI values ​​of all neighboring pixels are summed and divided by the sum of all weight values ​​to obtain an estimate of the proportion of water in the mixed pixel M, Pwater(M). Figure 8 This represents the proportion of water in the mixed pixels within the water body + other land features area determined according to Method 2.

[0122] Based on the initial water surface 1, a river centerline is generated. Cross-sections perpendicular to the river centerline are detected. The distribution of water in single-flow-path, multi-flow-path, and mixed-flow-path pixels is determined based on changes in NDWI values ​​of pixels on the cross-section. Furthermore, based on the 30m resolution NDWI raster image, a corresponding 3m high-resolution NDWI raster image is created to extract water surface 2, as shown below. Figure 9 This represents the location of the water body in the mixed pixel within the high-resolution raster image.

[0123] Figure 10 The result of merging the initial water surface with the water bodies in the mixed pixels is used to calculate the water body characteristic parameters on the merged water surface.

[0124] ​​The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for improving the accuracy of remote sensing monitoring of river surface characteristic parameters based on hydrological knowledge, characterized in that: It includes the following steps: S1: Based on the DEM, extract the river channel and the ridgeline on both sides of the river channel to determine the valley range; S2: Use vector files of ridgelines on both sides of the river to crop the remote sensing image to obtain remote sensing images within the river channel and valley area; S3: Calculate the acquired remote sensing images of the river channel and valley using the water body index; determine the water surface 1, the water surface plus other features, and the range of other features; S4: Determine the proportion of water in the mixed pixels within the water body + other land features area based on the NDWI values ​​of the surrounding pixels; S5: Generate the river centerline based on the initial water surface 1, detect the cross-section in the vertical direction of the river centerline, and determine the water distribution location in single flow path, multi-flow path and mixed pixels based on the NDWI value changes of pixels on the cross-section. S6: Based on the proportion and location of water bodies in the mixed pixels determined by S4 and S5, and the NDWI value of the original resolution a, create a high-resolution NDWI raster map of the corresponding location and extract the water surface 2. S7: Use the new water surface file to obtain water surface feature parameters; The specific process of S1 is as follows: First, the flow direction of each grid is obtained on the depression-free DEM using the maximum slope method. Then, the number of grids accumulated in the flow direction for each grid is calculated using the flow direction grid data, i.e., the flow accumulation. The obtained flow accumulation represents how many grids have flow directions passing through a grid at a grid location. The flow accumulation of a grid represents the water flow of that grid. A threshold is set to distinguish river channels from other smaller ditches or drainage paths, and areas with larger flow accumulation are selected as river networks. Based on the DEM, depression filling is performed through neighborhood analysis to eliminate the influence of depressions and generate a depression-free DEM. Hydrological analysis was performed using a depression-free DEM to calculate water flow direction data, reflecting the slope of the terrain and the water flow path. The runoff accumulation was calculated using the water flow direction data. This index reflects the total amount of water received by each pixel. When it is zero, it indicates that the point is located at the watershed. Finally, the intersection of the points with zero runoff accumulation and the positive terrain is obtained to obtain the ridgeline. The specific process of S3 is as follows: The water index map is obtained from the preprocessed image data using the normalized water index method, as shown in the following formula: Wherein, NDWI is the normalized water index, G is the green band data in the preprocessed image data, and NIR is the near-infrared band data in the preprocessed image data; In the preprocessed remote sensing image, n samples of each land feature are selected to determine the maximum and minimum values ​​of each land feature in the green band and near-infrared band. Determine the threshold range for NDWI for water bodies, bare soil, vegetation, and rocky surface cover; This determines the threshold ranges for initial water surface 1, water surface + other land features, and other land feature categories. The overlapping portion of the threshold ranges for water bodies and other land features is defined as the threshold range for water surface + other land features, and the non-overlapping portion is defined as pure land features. Initial water level 1: ; Water surface + other land features: ; Other land features: ; Where V represents the set of pixels in the NDWI raster image, I ij Let I be the water index value of the pixel in the i-th row and j-th column of the raster image. Pmin I Pmax Let I be the maximum and minimum thresholds for segmenting pure water bodies, respectively. Mmax The maximum threshold for non-water body pixel segmentation is determined separately; W1 is the initial water surface range, M is the range where water bodies and other land features are mixed, and N is the range of other land features.

2. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S2 is as follows: The image data was subjected to radiometric calibration, geometric correction, atmospheric correction, and image fusion to obtain preprocessed image data. For each identified flow path, the remote sensing image was cropped using the vector file of the ridgeline on both sides of the river channel to remove background noise, reduce the study area, and obtain remote sensing images of only the river channel and valley.

3. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S4 is as follows: First, the land cover to which the neighboring pixels of the mixed pixel (i,j) belong is determined based on the threshold range of water bodies, bare soil, vegetation, and rock surface cover in the green band, near-infrared band, and NDWI. This determines the types of land cover other than water bodies contained in the mixed pixel. The NDWI value of the mixed pixel and the NDWI value of the neighboring pixel show a linear superposition relationship. The relationship expression is constructed as follows: ； Among them, I ij For the NDWI value of the mixed pixel, I W I N These represent the NDWI values ​​of water bodies and non-water bodies in the neighborhood pixels, respectively, and P water (M), P nowater (M) represents the proportion of water and non-water bodies in the mixed pixel, respectively, and the sum of the two is 1.

4. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S4 is as follows: Define a weight function w(I ij Construct an n×n filter kernel based on the NDWI values ​​of neighboring pixels (M). ij Weights are calculated based on the water body baseline and the NDWI value of the mixed pixel itself; in, α It is a positive adjustment parameter used to control the steepness of the slope of the weighting function; It is the maximum difference between the NDWI values ​​of non-water pixels in the dataset and the water body reference value, used to normalize the difference; the tanh function ensures that the weight value varies between -1 and 1. When the difference from the water body reference value is large, the weight value tends to -1; when the difference is small, the weight value tends to 1. Determine the neighborhood of the intermediate mixed cell M, for each neighboring cell I of the mixed cell M. ij Calculate its NDWI value w(I) after adjustment by the weighting function. ij M) I ij The weighted NDWI values ​​of all neighboring pixels are summed and divided by the sum of all weight values ​​to obtain an estimate P of the proportion of water in the mixed pixel M. water (M), its theoretical formula is: 。 5. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S5 is as follows: The acquired raster initial water surface 1 is converted into a water surface vector file, and the river centerline is generated by morphological skeleton extraction. For each point along the river centerline, calculate the flow direction angle α at that point; based on this angle, determine the direction perpendicular to the centerline, i.e., α+90°, and determine the river cross-section; Along the river cross-section, NDWI values ​​are extracted pixel by pixel. This index can effectively distinguish between water and land. Pixels are classified according to preset thresholds to determine the land cover types such as water, vegetation, and soil, and to identify the specific distribution of water in single-flow-path areas, multi-flow-path areas, and mixed pixels.

6. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S6 is as follows: Create a new a / n 2 For high spatial resolution raster files, the NDWI values ​​of the NDWI raster generated by S3 are written into the newly created raster file to ensure consistency with the spatial reference coordinate system of the original image. Using the water body proportion data obtained from S4 blended pixel decomposition and the water body location determined in S5 blended pixels, the pixel values ​​in the high-resolution NDWI raster are adjusted to ensure that the processed NDWI values ​​are correctly mapped to the new high-resolution raster. Water bodies were extracted again from the high-resolution NDWI raster image to obtain water surface 2.

7. The method for improving the accuracy of remote sensing monitoring river surface characteristic parameters based on hydrological knowledge according to claim 1, characterized in that: The specific process of S7 is as follows: Based on the water surface vector file and the perpendicular river channel vector line file, calculate the river width and water surface characteristic parameters such as the length-to-width ratio in the river morphology; The gradient of the river channel: ； In the formula, For the sake of comparison; This refers to the cross-sectional elevation. The horizontal distance between the two points; The ratio of the water surface area to the length of a river segment is used as the average width of that segment. ； In the formula, W is the generalized river section width; A is the water surface area of ​​the river section under study; and L is the length of the river section under study.

Images