International boundary water body extraction method based on composite index

Abstract

The invention discloses an international boundary water extraction method based on a composite index, which comprises the following steps: determining the range, position and name of cross-border water by using water system and boundary line data, and constructing a database; a satellite image of a research area is obtained on a GEE platform, images covering cross-border water are screened in combination with the database, and peripheral non-border water bodies are removed; the method comprises the following steps of: preprocessing an image, extracting a water body by using a multi-index and condition combination, screening each pixel meeting a combination condition as a water body pixel, performing noise reduction by using a composite mask to obtain image data, and identifying and extracting pixels meeting a water body characteristic condition by using a pixel classification function to form a water pixel set; through visual interpretation and experience judgment, cross-border water buffer areas of different grades are drawn, water pixels outside the range of the buffer areas are removed, and a result is obtained through comparison and integration. The method has the advantages that the calculation process of the index method is simple, and the accuracy and reliability of water body extraction can be greatly improved.

Description

Technical Field

[0001] The invention belongs to the technical field of water body extraction, and more specifically, the invention relates to an international boundary water body extraction method based on a composite index. Background Art

[0002] In the field of global geography and resource management, accurate processing of international boundary water bodies is very important. Their complex environment, diverse terrain, and complicated land use types increase the difficulty of water body identification. However, traditional water body extraction technology has defects. The single threshold method is susceptible to noise interference. The combined method of mNDWI, NDVI and EVI has poor extraction effect in complex water bodies and vegetation interlaced areas, and the extraction accuracy is insufficient. It is also difficult to determine the name of cross-border water bodies and eliminate surrounding non-boundary water bodies. Therefore, there is an urgent need for an innovative and effective method to solve key problems such as accurately identifying the name and location of cross-border boundary water bodies, eliminating surrounding non-boundary water bodies, and improving the accuracy of water body extraction, so as to fill the gaps in existing technologies and promote related research and management development. Summary of the invention

[0003] An object of the present invention is to address at least the above-mentioned disadvantages and to provide at least the advantages which will be described hereinafter.

[0004] The present invention provides an international boundary water body extraction method based on a composite index. It starts with constructing a database using water system and boundary line data, providing an accurate basis for subsequent image screening. It is closely related to steps such as water body extraction and preprocessing by combining multiple remote sensing indices and specific conditions. It has the advantages of a simple index calculation process and greatly improving the accuracy and reliability of water body extraction.

[0005] The present invention provides a method for extracting international boundary water bodies based on a composite index, comprising the following steps:

[0006] Step 1: Use water system and boundary data to determine the scope, location and name of cross-border waters and build a database;

[0007] Step 2: Obtain multiple Landsat satellite images of the study area on the GEE platform, filter images covering cross-border waters based on the database, and exclude surrounding non-boundary water bodies;

[0008] Step 3: Preprocess the image and extract water bodies using multiple indexes and condition combinations. Each pixel that meets the combination conditions is selected as a water pixel, and a composite mask is used for noise reduction to obtain image data and preliminarily determine the water area.

[0009] Step 4: Further utilize the pixel classification function of the GEE platform for the image data of the preliminarily determined water body area to identify and extract the pixels that meet the water body feature conditions, forming a water pixel set; through visual interpretation and empirical judgment, draw cross-border water buffers of different levels in the GEE platform, remove the water pixels outside the buffer range, and compare and integrate to generate the final result.

[0010] Preferably, in Step 1, specifically: use the water system of the basic geographic database and the administrative boundary line of the basic geographic database to perform buffer overlay analysis in the geographic information system software; set the buffer distance according to the administrative boundary line, identify the intersection part of the area with the buffer distance and the water system data, determine the cross-border water range, and record its geographical location information and name to construct a database.

[0011] Preferably, the preprocessing of the image in Step 3 includes cloud removal, filtering operation, mean synthesis, and boundary noise removal.

[0012] Preferably, in Step 3, multiple indices and conditional combinations are used to extract the water body. Specifically: use multiple remote sensing indices such as the Normalized Difference Water Index (NDWI), the Modified Normalized Difference Water Index (MNDWI), the Normalized Difference Vegetation Index (NDVI), and the Enhanced Vegetation Index (EVI), and combine conditional combinations to extract the boundary water body; among them, the combined conditions are NDWI>0, MNDWI>0.1, EVI<0.1, MNDWI>NDVI, and MNDWI>EVI.

[0013] Preferably, in Step 4, the buffer levels of different levels are 0.5 km, 1 km, 2 km, 3 km, 5 km, 8 km, and 10 km; use the buffer to screen the extracted water pixels. Specifically: first judge whether the water pixel is within the buffer range, if not, remove it; then for the water pixels within the buffer, use the small area batch screening method to remove small water patches.

[0014] Preferably, it further includes Step 5: Verify the accuracy of the image data of the final result.

[0015] The present invention has at least the following beneficial effects:

[0016] The present invention constructs a database using water system and boundary line data, which facilitates verification and screening using database information during the extraction process, laying a foundation for the accuracy of the extraction results. Then, by combining the use of multiple remote sensing indices (NDWI, MNDWI, NDVI, and EVI) and specific condition combinations (NDWI>0, MNDWI>0.1, EVI<0.1, MNDWI>NDVI, and MNDWI>EVI) for water body extraction, in the image preprocessing stage, comprehensive operations such as cloud removal, filtering, mean synthesis, and boundary noise removal are covered, improving the image quality. Combining with the composite mask noise reduction technology effectively avoids the problems of the traditional single-threshold method being interfered by noise and the traditional combined method having poor extraction effects in complex water body and vegetation intersection areas, greatly improving the accuracy of water body extraction. It also combines the pixel classification function and sets different levels of buffer zones to screen and integrate water pixels, ensuring the reliability of the final result.

[0017] Other advantages, objectives, and features of the present invention will be partially reflected by the following description and partially understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the international boundary water body extraction method based on the composite index according to the present invention.

[0019] Figure 2 It is a detailed comparison diagram of the actual water body distribution and the water body extracted by the method used in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] The following further describes the present invention in detail with reference to the embodiments, so that those skilled in the art can implement it according to the description in the specification.

[0021] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial sources unless otherwise specified; in the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "set" should be understood in a broad sense, for example, they can be fixedly connected, set, or detachably connected, set, or connected and set in one piece. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood in specific circumstances. The orientation or position relationship indicated by the terms "lateral", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. is based on the orientation or position relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention.

[0022] Figure 1 The flowchart of the method for extracting international boundary water bodies based on the composite index is presented, including the following steps:

[0023] Step S1, using water system and boundary data to determine the scope, location and name of cross-border boundary waters, and construct a database, specifically: using the water system of the 1:250,000 national basic geographic database and the administrative boundary of the 1:1 million national basic geographic database, a buffer overlay analysis is performed in the geographic information system (GIS) software. According to the administrative boundary, a suitable buffer distance is set (such as 500 meters to 5 kilometers, adjusted according to the characteristics of the research area and the accuracy requirements), the intersection of the area of ​​the buffer distance and the water system data is identified, the possible cross-border boundary water scope is preliminarily determined, and its geographical location information and name (latitude and longitude range, administrative region, etc.) are recorded to construct a database. In the process of constructing the database, it also includes the verification and improvement of treaties and local chronicles: organizing professionals to conduct detailed reading and analysis of these materials, and for the preliminarily determined potential cross-border boundary water scope, according to the boundary description, water system name record and other information in the data, the accurate name and precise location of the water body are checked and confirmed one by one.

[0024] Step S2: obtain multiple Landsat5 or Landsat8 series satellite images of the study area on the GEE platform. The time span of the images should cover different seasons and years to reflect the status of water bodies in different periods. Combined with the database, the images covering cross-border boundary waters are screened and the surrounding non-boundary water bodies are eliminated.

[0025] Step S3: perform the following pre-processing operations on the acquired images in sequence:

[0026] Deblurring processing: Apply the advanced deblurring algorithms built into the GEE platform, such as the deblurring method based on multi-spectral features and texture analysis. This algorithm identifies and removes cloud layers and their shadow parts in the image by analyzing the reflectance and transmittance characteristics of cloud layers in different bands, as well as the shape, direction, and spectral characteristics of shadows, ensuring the integrity and accuracy of the image data.

[0027] Filtering operation: Adopt filtering techniques, such as median filtering or Gaussian filtering. Median filtering can effectively remove salt-and-pepper noise by replacing the value of each pixel with the median of the values of its neighboring pixels; Gaussian filtering performs weighted averaging of pixel values according to the Gaussian function, which can remove noise while preserving the edge and detail information of the image. The size of the filtering window is reasonably set according to the image resolution and noise conditions (usually between 3x3 and 9x9 pixels).

[0028] Mean synthesis: Perform mean synthesis on multiple images of the same area according to the time series. Determine the synthesis time interval according to research needs (such as monthly, quarterly, or annually), calculate the average of the corresponding pixel values of the images within this time period, and generate a new image with higher signal-to-noise ratio and stability, reducing the impact of possible outliers or noise in a single image on subsequent analysis.

[0029] Boundary noise removal: For possible data anomalies or noise in the boundary area of the image, use an edge detection algorithm (such as the Canny edge detection algorithm) to identify the image boundary, and then judge and remove possible noise pixels according to the spectral characteristics and spatial distribution rules of the boundary pixels. For example, boundary pixels with too large spectral differences from surrounding pixels and not conforming to the spectral characteristics of water bodies or common ground objects are regarded as noise and removed.

[0030] Step S4: Extract water bodies from the preprocessed image using multiple indices and conditional combinations. Each pixel that meets the combined conditions is screened as a water body pixel, and composite masking is used to reduce noise to obtain image data and preliminarily determine the water body area. Specifically:

[0031] Based on the band and data characteristics of the Landsat series satellites, on the GEE platform, multiple remote sensing indices such as the Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Normalized Difference Vegetation Index (NDVI), and Enhanced Vegetation Index (EVI) are applied to the preprocessed images, and combined with specific multi - condition combinations (such as NDWI>0, MNDWI>0.1, EVI<0.1, MNDWI>NDVI, and MNDWI>EVI) to extract boundary waters. During the extraction process, the names and location information of transboundary waters are used as references to preliminarily verify and screen the extraction results. For example, if the location of the extracted water body is significantly inconsistent with the location recorded in the database, or the name of the extracted water body cannot correspond to the known name of the transboundary water, then the extraction result is further analyzed and corrected.

[0032] The calculation formulas of the above - mentioned multiple remote sensing indices are as follows:

[0033]

[0034] In the above formulas, the actual calculation bands of different satellites are different. For Landsat 8 satellite, the Green band, Red band, SWIR band, and NIR band are SR_B3, SR_B4, SR_B6, and SR_B5 respectively. For Landsat 5 satellite, the Green band, Red band, SWIR band, and NIR band are SR_B2, SR_B3, SR_B5, and SR_B4 respectively.

[0035] Set NDWI>0 to exclude the shadow interference caused by most mountains, vegetation, and buildings; at the same time, require MNDWI>0.1, EVI<0.1, MNDWI>NDVI, and MNDWI>EVI. Through these conditions, comprehensively judge whether each pixel is a water pixel, and screen and extract the pixels that meet the above conditions as water pixels. After the extraction is completed, noise is removed through the composite mask noise reduction technology to obtain the image data of the preliminarily determined water area.

[0036] Among them, the composite mask noise reduction technology is to create a composite mask file on the GEE platform based on the topographic data of the study area, land use data, or the previous understanding of the water body and noise distribution in the study area. For example, if there are known noise source areas (such as industrial mining areas, large construction sites, etc.) or areas prone to misjudgment (such as mountainous shadow - complex areas, urban built - up area edges, etc.) in the study area, these areas are marked as areas that need noise reduction processing in the mask file. The created mask file is applied to the water - extraction result image, and through mask operation, only the pixels in the marked areas that need to be processed are denoised to improve the accuracy of water - pixel extraction.

[0037] Step S5: Further utilize the pixel classification function of the GEE platform for the image data of the preliminarily determined water body area to identify and extract the pixels that meet the water body characteristic conditions, forming a water pixel set.

[0038] Through visual interpretation and empirical judgment, draw cross-border water buffers of different levels on the GEE platform, remove the water pixels outside the buffer range, and compare and integrate to generate the final result. Specifically:

[0039] According to the actual characteristics of the boundary water body, such as width, braided degree, and fluctuation range, draw cross-border water buffers of different levels on the GEE platform through visual interpretation and empirical judgment. The buffer levels can be set to 0.5 km, 1 km, 2 km, 3 km, 5 km, 8 km, 10 km, etc., and use the constructed buffers to further screen the extracted water pixels. First, by writing code or using the spatial analysis function of the software, judge whether the water pixels are within the buffer range. If not, remove them. Then, for the water pixels within the buffer, adopt a small-area batch screening method. According to the set minimum area threshold (such as determined to be between 10 square meters and 1000 square meters according to the image resolution and the scale of the water body in the study area), remove the water patches with too small areas. These small patches may be noise or non-main water body parts to ensure the accuracy and practicality of the extracted water body.

[0040] Step S6: Conduct accuracy verification on the image data of the final result. Specifically:

[0041] Step S6.1: Open the image data after water body extraction processing on the GEE platform and perform stratified random sampling operations using the "ee.Image.stratifiedSample" function. When stratifying, fully consider the distribution characteristics of water bodies and non-water bodies, geographical environment factors (such as topography and land use types), and the types of water bodies (such as rivers, lakes, reservoirs, etc.) in the study area, and divide the image into multiple representative layers. For example, it can be divided into layers such as mountainous areas and plains according to topography, into layers such as farmland, forest, and cities according to land use types, and into layers such as large rivers, small rivers, and lakes according to water body types. In each layer, use the random sampling function of the GEE platform to respectively extract multiple water body and non-water body sample pixels. Ensure the randomness and independence of the sampling process to avoid the correlation and deviation between samples. During the sampling process, record the position information (such as row and column numbers or longitude and latitude coordinates) of each sample pixel and its characteristic values in the original image (such as the reflectance of each band, the calculated index values, etc.) for subsequent comparison and analysis.

[0042] Step S6.2: In the QGIS 3.34 software platform, load the Esri Wayback historical images and Sentinel-2 images respectively, and register them with the sampled pixel positions. Ensure that the geographic coordinate systems of different images are consistent so as to accurately compare the feature information of the same location. For each sample pixel, judge its true category (water body or non-water body) in the Esri Wayback historical images and Sentinel-2 images through a combination of visual interpretation and spectral analysis. During visual interpretation, use features such as the color, texture, and shape of the images for identification; spectral analysis is judged based on the spectral reflectance differences of different ground objects in different bands. For example, water bodies usually have a lower reflectance in the near-infrared band, while vegetation has a higher reflectance in the near-infrared band. At the same time, refer to the known ground object spectral library and the ground object feature knowledge of the study area to improve the accuracy of judgment. For sample pixels that are difficult to determine, relevant materials can be further consulted. During the determination process, record in detail the feature performance and judgment basis of each sample pixel in different images, and establish a sample determination database.

[0043] Step S6.3: According to the results of sample determination, count the numbers of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN). TP is the number of sample pixels that are determined to be water bodies in the water body extraction results and are also confirmed to be water bodies in the multi-source image determination; TN is the number of sample pixels that are determined to be non-water bodies in the water body extraction results and are also confirmed to be non-water bodies in the multi-source image determination; FP is the number of sample pixels that are determined to be water bodies in the water body extraction results but are confirmed to be non-water bodies in the multi-source image determination; FN is the number of sample pixels that are determined to be non-water bodies in the water body extraction results but are confirmed to be water bodies in the multi-source image determination. Use the formula to calculate the overall accuracy (OA) and F1 score:

[0044]

[0045] First, calculate the precision and recall:

[0046]

[0047] Then calculate:

[0048]

[0049] Through these accuracy indicators, comprehensively evaluate the accuracy and reliability of the boundary water body extraction method based on the composite index. According to the accuracy evaluation results, analyze the possible problems and error sources in the extraction process in order to further optimize and improve the method.

[0050] Verification example:

[0051] Using the extraction method of the present invention, a certain boundary water body is experimentally extracted, such as Figure 2 as shown in the comparison chart of the extracted water body. Taking Figure 2 the extraction results of the b river section, using the random sampling function of the GEE platform, 1000 water body and non-water body sample pixels are respectively extracted. It is found that there are 465 sample pixels of true positives (TP), 465 sample pixels of true negatives (TN), 35 sample pixels of false positives (FP), and 35 sample pixels of false negatives (FN).

[0052] The overall accuracy of OA is calculated according to the formula for calculation.

[0053] Substituting the above data into the formula, we get:

[0054]

[0055] That is, the overall accuracy is 92.9%. This indicates that among all samples, the proportion of correctly classified samples is 92.9%, reflecting that the water body extraction method has relatively high accuracy overall.

[0056] For the calculation of the F1 score, first calculate the precision and the recall

[0057] Substituting the above data into the formula, calculate the precision:

[0058]

[0059] The recall:

[0060]

[0061] Then, according to the formula calculate the F1 score:

[0062]

[0063] The F1 score comprehensively considers the precision and the recall, and its value is approximately 0.927, close to 1, further indicating that the water body extraction method has excellent performance in the classification of water bodies and non-water bodies, verifying the effectiveness and reliability of the method.

[0064] Although the embodiments of the present invention have been disclosed as above, it is not limited to only the applications listed in the specification and the embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily achieved.

Claims

1. The international boundary water body extraction method based on composite index is characterized by: The steps include: Step 1: Use water system and boundary data to determine the scope, location and name of cross-border waters and build a database; Step 2: Obtain multiple Landsat satellite images of the study area on the GEE platform, filter images covering cross-border waters based on the database, and exclude surrounding non-boundary water bodies; Step 3: Preprocess the image and extract water bodies using multiple indexes and condition combinations. Each pixel that meets the combination conditions is selected as a water pixel, and a composite mask is used for noise reduction to obtain image data and preliminarily determine the water area. Step 4: Use the pixel classification function of the GEE platform to identify and extract pixels that meet the water body characteristic conditions from the image data of the preliminarily determined water area, forming a water pixel set; through visual interpretation and empirical judgment, draw different levels of cross-border water buffer zones in the GEE platform, remove water pixels outside the buffer zone, and compare and integrate them to generate the final results.

2. The method for extracting international boundary water bodies based on a composite index according to claim 1, characterized in that: In the step one, specifically: using the basic geographic database water system and the administrative boundary line of the basic geographic database, a buffer overlay analysis is performed in the geographic information system software; a buffer distance is set according to the administrative boundary line, the intersection of the buffer distance area and the water system data is identified, the scope of cross-border boundary waters is determined, and its geographical location information and name are recorded to build a database.

3. The method for extracting international boundary water bodies based on a composite index as claimed in claim 2, characterized in that: The image preprocessing in step three includes cloud removal, filtering, mean synthesis, and boundary noise removal.

4. The method for extracting international boundary water bodies based on a composite index as claimed in claim 3, characterized in that: In the step three, multiple indexes and condition combinations are used to extract water bodies, specifically: multiple remote sensing indices including the normalized difference water index NDWI, the modified normalized difference water index MNDWI, the normalized vegetation index NDVI and the enhanced vegetation index EVI are used in combination with condition combinations to extract boundary water bodies; wherein the combination conditions are NDWI>0, MNDWI>0.1, EVI<0.1, MNDWI>NDVI and MNDWI>EVI.

5. The method for extracting international boundary water bodies based on a composite index as claimed in claim 4, characterized in that: In the step 4, the buffer zones of different levels are 0.5km, 1km, 2km, 3km, 5km, 8km and 10km; the extracted water pixels are screened using the buffer zones, specifically: first determine whether the water pixels are within the buffer zone, if not, remove them; then for the water pixels within the buffer zone, use a small-area batch screening method to remove small-area water patches.

6. The method for extracting international boundary waters based on a composite index according to any one of claims 1 to 5, characterized in that: It also includes step five, verifying the accuracy of the image data of the final result.

Images