Water body extraction method based on the integration of multiple artificial intelligence algorithms

Abstract

The present invention discloses a water body extraction method based on the integration of multiple artificial intelligence algorithms, including: S1, obtaining a remote sensing image and preprocessing it; S2, performing multi-feature extraction on the image of the experimental area; S3, respectively using a water body index algorithm, a random forest RF algorithm, a support vector machine SVM algorithm, a K-means algorithm, and a convolutional neural network algorithm to extract the water body from the image of the experimental area; S4, taking the binary images of the extraction results of the five algorithms in S3 as the input of the data parsing model to obtain the true accuracy corresponding to the water body extraction result; S5, calculating the weights corresponding to the five algorithms according to the true accuracy; using the weights to calculate the water body probability of each pixel in the binary image to obtain a water body probability set; setting a classification threshold to optimize and integrate the water body probability set to obtain the final binary water body classification map. The method of the present invention enables the advantages of different water body extraction algorithms to complement each other and improves the accuracy of water body extraction.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image water body extraction, and in particular to a water body extraction method based on the integration of multiple artificial intelligence algorithms. Background Technique

[0002] As an important part of the ecological cycle, water resources play an important role in the hydrological and biogeochemical cycles at the regional and global scales. However, due to climate change, land use, land cover change, and other forms of environmental change, surface water bodies are also constantly changing in time and space. Therefore, it is of great significance to carry out research on high-precision extraction of large-scale water bodies and achieve timely and accurate monitoring of surface water bodies such as rivers. Most traditional extraction methods analyze through field measurements and on-site sampling. Although more accurate information on surface water bodies can be obtained through this method, it requires a large amount of manpower and material resources, and the experimental cycle is long, which cannot meet the actual needs of large-scale and multi-frequency extraction of surface water bodies. With the rapid development of satellite remote sensing technology and its characteristics of fast speed, all-weather, and high resolution, satellite remote sensing technology has become the main means of surface water body monitoring.

[0003] Satellite remote sensing technologies for surface water body extraction are mainly divided into three categories: water body index method, machine learning method, and deep learning method.

[0004] Among them:

[0005] The water body index method constructs a mathematical model by establishing a band ratio operation based on the different reflectivities of water bodies in different visible light bands, and sets corresponding thresholds for water body identification. The water body index method is widely used because of its simple and efficient calculation. Commonly used water body indices include: Normalized Difference Water Index (NDWI), Modified NDWI (MNDWI), etc. However, when dealing with images with higher resolutions, the more visible spatial details there are, the greater the spectral variability of the water body, which in turn leads to poor generalization ability of the water body index method.

[0006] The water body extraction method based on machine learning includes the supervised classification water body extraction method and the unsupervised classification water body extraction method. For the supervised classification water body extraction method, Support Vector Machine (SVM) and Random Forest (RF) are more commonly used. It mainly establishes a training data set by collecting water body samples and non-water body samples, and performs water body extraction on the identified area based on the training model. The clustering algorithm (Cluster) is a common unsupervised classification water body extraction method. Its principle is to divide the pixels in the image into different categories according to specific criteria, with the pixels in the same category having as high a similarity as possible and the pixels between different categories having as large a difference as possible.

[0007] The water body extraction method based on deep learning mainly uses a Convolutional Neural Network (CNN) to learn texture, spectral and other high-dimensional features from an artificially annotated data set (training samples), and then uses the convolutional neural network to perform water body extraction on the region of interest in the image.

[0008] Although the water body extraction methods based on machine learning and deep learning have improved the extraction accuracy compared with the water body index method, there are still the following two problems:

[0009] 1. The recognition results of machine learning algorithms such as Random Forest are extremely prone to confusing buildings, bridges, roads and rivers, resulting in misjudgment, especially in complex urban scenes.

[0010] 2. The water body extraction method based on deep learning has extremely high requirements for training samples and requires the entire image to be marked, otherwise it will seriously affect the water body recognition accuracy.

[0011] In addition, the water body index method, the machine learning method and the deep learning method all have poor recognition effects on narrow water bodies, and there are often situations of broken flow or missed judgment. Summary of the Invention

[0012] To solve the problems existing in the prior art, the present invention provides a high-precision, time-continuous and space-seamless water body extraction method based on the integration of multiple artificial intelligence algorithms.

[0013] For this reason, the present invention adopts the following technical solutions:

[0014] The water body extraction method based on the integration of multiple artificial intelligence algorithms includes the following steps:

[0015] S1. Obtain remote sensing images and perform preprocessing:

[0016] Obtain a set of remote sensing images that contain the water body area to be extracted within a specified time range; perform cloud amount screening and cloud removal on the set of remote sensing images to reduce the influence of cloud cover; then crop the image set according to the experimental area to obtain an image set of the water body area to be extracted; perform median synthesis on the image set to obtain the final image of the water body area to be extracted;

[0017] S2. Perform multi-feature extraction on the final image:

[0018] Obtain the optical band information of each pixel in the final image, and calculate the index band information of each pixel in the final image according to the optical band information. The index band information includes NDWI, MNDWI, NDVI, EVI, NDBI, and BSI;

[0019] S3. Using the optical band information and index band information obtained in S2, use five water body extraction algorithms to perform water body extraction on the final image respectively, and obtain binary images of the water body extraction results, which are x a 、x b 、x c 、x d and x e ;

[0020] S4. Solve the true accuracy of the water body extraction result based on mathematical uncertainty analysis:

[0021] According to the binary images x a 、x b 、x c 、x d and x e obtained in S3, calculate the covariance between two different water body extraction results, and then calculate the accuracy vector y; use the accuracy vector y as the input of the data analysis model to obtain the true accuracy corresponding to the water body extraction result;

[0022] S5. Multi-algorithm integrated learning to obtain the optimal water body extraction algorithm, including the following steps:

[0023] S51. According to the true accuracy of the binary images obtained by the data analysis model, calculate the weights w a 、w b 、w c 、w d and w e corresponding to the five algorithms in S3 respectively. The calculation formula is as follows:

[0024]

[0025] where v i is the binary image x iThe true accuracy, where i = a to e;

[0026] S52. According to the weights corresponding to the five algorithms, traverse all pixels in the binary image, calculate the water body probability of each pixel in the binary image, and obtain the water body probability set M = {m1, m2... m n}, where the water body probability m j of the j-th pixel in the binary image is calculated as follows:

[0027] m j = w a a j + w b b j + w c c j + w d d j + w e e j (8)

[0028] In the formula, a j , b j , c j , d j and e j are the pixel values of the j-th pixel in the binary image x a , binary image x b , binary image x c , binary image x d and binary image x e respectively; w a , w b , w c , w d and w e are the weights corresponding to the five algorithms respectively; j = 1, 2,... n;

[0029] S53. Set the classification threshold to 0; perform masking processing on the pixels with water body probability greater than the classification threshold according to the water body probability set M, and optimize the integration to obtain the final water body classification binary image, completing the accurate extraction of the water body in the experimental area image.

[0030] The remote sensing image set described in S1 is the Sentinel-2 satellite image set.

[0031] The method for water body extraction using the water body index algorithm is: taking the experimental area image as the input, performing NDWI ratio operation on each pixel of the experimental area image using the water body index method and setting a threshold. If the NDWI of the pixel is greater than or equal to the threshold, it is a water body; if the NDWI of the pixel is less than the threshold, it is a non-water body. The water body index algorithm outputs a binary image x a with only water body and non-water body, where the threshold is set to 0.

[0032] The binary image x obtained by S3 a , x b , x c , x d and x e have the same total number of pixels, all of which are n.

[0033] The ensemble learning algorithm described in S3 is the random forest algorithm RF, the supervised machine learning algorithm is the support vector machine SVM algorithm, the clustering algorithm is the K-means algorithm, and the supervised deep learning algorithm is the convolutional neural network algorithm CNN.

[0034] The method for water body extraction using the random forest algorithm RF and the support vector machine SVM algorithm is as follows: Combining the actual situation of the experimental area, each pixel in the experimental area image is divided into 5 land cover types: water body, crop, building, vegetation, and bare land. Manually select sample pixels by visual interpretation on the experimental area image; Use the optical band information and index band information of each sample pixel as the feature variables in the training sets of the RF and SVM algorithms, and input them into the random forest RF algorithm and the support vector machine SVM algorithm for training respectively to obtain the training results. And mask the sample pixels with the land cover type of water body. Both the random forest algorithm RF and the support vector machine SVM algorithm obtain binary images with only water body and non-water body, which are x b and x c .

[0035] The sample pixels include 150 water pixels and 382 non-water pixels in the experimental area image.

[0036] The method for water body extraction using the K-means algorithm is as follows: Divide all the pixels in the experimental area image into K given clusters, and mask the pixels with the land cover type of water body to obtain a binary image x d , where K = 5.

[0037] The method for water body extraction using the convolutional neural network algorithm CNN is as follows: Perform water body masking on the Dynamic World (DW) dataset, extract the pixels with the classification category of water body, and obtain a binary image x e .

[0038] S4 includes the following steps:

[0039] S41, According to the water body extraction results obtained by the five algorithms, calculate the covariance r ad , covariance r ae , covariance r de , covariance rbd Covariance r be Covariance r ce and covariance r cd ;

[0040] Taking covariance r ad as an example, its calculation formula is:

[0041]

[0042] In the formula, a j is the pixel value of the j-th pixel in the binary image x a , d j is the pixel value of the j-th pixel in the binary image x d , is the mean value of all pixel values of the binary image x a , is the mean value of all pixel values of the binary image x d , and n is the total number of pixels in the binary image;

[0043] S42. Combine and calculate the covariance obtained in S41 to obtain the accuracy vector y. The calculation formula is as follows:

[0044]

[0045] S43. Construct a data analysis model according to the accuracy vector y:

[0046] Av = y(3)

[0047] where v represents the true accuracy of the binary image obtained by five algorithms, y is the accuracy vector, A is a constant vector matrix connecting the true accuracy v and the accuracy vector y. The calculation formulas for the true accuracy v and the constant vector matrix A are as follows:

[0048] v = [v a v b v c v d v e T (4)

[0049]

[0050] In formula (4), v a is the true accuracy of the binary image x a in the data analysis model; v b is the true accuracy of the binary image x b in the data analysis model; v c is the true accuracy of the binary image x c in the data analysis model; v​d is the binary image x d is the true accuracy in the data parsing model; v e is the binary image x e is the true accuracy in the data parsing model;

[0051] S44, calculate the least squares solution of equation (3) to obtain the true accuracy v, and the calculation formula is as follows:

[0052] v = (A T A) -1 A T y (6).

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. Based on the mathematical uncertainty analysis theory, the present invention realizes the accuracy evaluation and ensemble learning of multiple intelligent algorithms without relying on "ground truth", overcomes the problem of water body misjudgment or missed judgment of a single water body extraction algorithm, makes the advantages of different water body extraction algorithms complement each other, and improves the water body recognition accuracy under complex backgrounds.

[0055] 2. The water body extraction method of the present invention aims at the problems of incomplete water body boundary extraction and discontinuous narrow water bodies of some water body extraction algorithms, determines the corresponding weight ratio through the fusion of different water body extraction algorithms, accurately extracts slender water bodies, and maintains high accuracy of the water body boundary.

[0056] 3. The water body extraction method of the present invention effectively avoids the influence of buildings, bridges, roads and mountain shadows, and this advantage is particularly obvious in urban areas.

[0057] 4. The water body extraction method of the present invention can realize high-precision, time-continuous and space-seamless water body extraction, become an important means for quickly and accurately monitoring the flood inundation range, has important significance for mastering the evolution of flood disasters, and provides data support and guarantee for disaster assessment, post-disaster reconstruction and water resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is the flowchart of the water body extraction method according to an embodiment of the present invention;

[0059] Figure 2 is the result diagram of the narrow water bodies extracted by the present invention and other algorithms in the same area respectively, wherein:

[0060] Figure (a) is the GF6 image;

[0061] Figure (b) is the true value image for inspection;

[0062] Figure (c) is the image of the experimental area;

[0063] Figure (d) shows the water body extraction result of the convolutional neural network algorithm;

[0064] Figure (e) shows the water body extraction result of the water index algorithm;

[0065] Figure (f) shows the water body extraction result of the SVM algorithm;

[0066] Figure (g) shows the water body extraction result of the K - means algorithm;

[0067] Figure (h) shows the water body extraction result of the RF algorithm;

[0068] Figure (i) shows the water body extraction result of the water body extraction method of the present invention. Detailed implementation manners

[0069] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0070] Embodiment

[0071] As Figure 1 shown, a water body extraction method based on the integration of multiple artificial intelligence algorithms specifically includes the following steps:

[0072] S1. Obtain remote sensing images and preprocess them:

[0073] Based on the Google Earth Engine (GEE) cloud platform, obtain the Sentinel - 2 satellite image collection within the specified time range including the experimental area; perform cloud amount screening and cloud removal processing on the Sentinel - 2 satellite image collection to reduce the influence of cloud cover; then perform image cropping on the Sentinel - 2 satellite image collection according to the experimental area to obtain the experimental area image collection; perform median synthesis on the experimental area image collection to obtain the final domain image of the experimental area.

[0074] Among them, in this embodiment, the Jiuxuanzha area in Jinghai District, Tianjin is used as the experimental area, the pixel size of the experimental area image is 10m×10m, and the Sentinel - 2 satellite image collection in October 2022 is selected for preprocessing.

[0075] S2. Perform multi - feature extraction on the final image of the experimental area:

[0076] Obtain the optical band information of each pixel in the final image of the experimental area, and calculate the index band information of each pixel in the final image of the experimental area according to the optical band information and the calculation method provided in Table 1. The index band information includes NDWI, MNDWI, NDVI, EVI, NDBI, and BSI. Among them, the calculation formula of the index band is shown in Table 1.

[0077] In Table 1, Red represents the red band (B4), Green represents the green band (B3), Blue represents the blue band (B2), NIR represents the near-infrared band (B8), and SWIR1 represents the short-wave infrared band (B11).

[0078] Feature variables in Table 1

[0079]

[0080] S3. The water body extraction of the experimental area image is carried out by using the water body index algorithm, the random forest RF algorithm, the support vector machine SVM algorithm, the K-means algorithm and the convolutional neural network algorithm respectively, as follows:

[0081] Water body index algorithm: Taking the experimental area image as the input, using the water body index method to perform the NDWI ratio operation on each pixel of the experimental area image and setting a threshold. If the NDWI of a pixel is greater than or equal to the threshold, it is a water body; if the NDWI of a pixel is less than the threshold, it is a non-water body. The water body index algorithm outputs a binary image with only water body and non-water body x a . In this embodiment, the threshold is set to 0.

[0082] Random forest algorithm (RF) and support vector machine SVM algorithm: Combining the actual situation of the experimental area, each pixel in the experimental area image is divided into 5 land cover types: water body, crop, building, vegetation, and bare land. Manually select sample pixels by visual interpretation on the experimental area image; Use the optical band information and index band information of each sample pixel as the feature variables in the training sets of the RF and SVM algorithms, and input them into the random forest RF algorithm and the support vector machine SVM algorithm for training respectively to obtain the training results, and perform masking processing on the sample pixels with the land cover type of water body. Both the random forest algorithm (RF) and the support vector machine SVM algorithm obtain binary images with only water body and non-water body, which are x b and x c . In this embodiment, the sample pixels include 150 water pixels and 382 non-water pixels in the experimental area image.

[0083] K-means algorithm: Divide all pixels in the experimental area image into K given clusters. In the present invention, combining the land use / cover characteristics of the study area, the value of the classification cluster K is set to 5, and the pixels with the land cover type of water body are masked to obtain a binary image with only water body and non-water body x d .

[0084] Convolutional Neural Network (CNN): The Dynamic World (DW) dataset (a near real-time LULC product developed by Google based on Sentinel-2 data with a resolution of 10 meters, which uses a fully convolutional neural network (FCNN) to classify Sentine-2 images containing the experimental area into nine categories: water bodies, trees, grass, flooded vegetation, crops, bushes / shrubs, built-up areas, bare land, snow / ice) is processed with water body masks, pixels classified as water bodies are extracted, and a binary image x containing only water bodies and non-water bodies is obtained. a .

[0085] Among them, the binary images obtained by the above five algorithms are x a 、x b 、x c 、x d and x e , their total number of pixels is equal, both are n.

[0086] S4, solve the true accuracy of water extraction results based on mathematical uncertainty analysis:

[0087] The binary images x obtained by the five algorithms in S3 a 、x b 、x c 、x d and x e As the input of the classification triple correction fusion method (data analysis model) based on mathematical uncertainty analysis, the true accuracy corresponding to the water body extraction result is obtained. The specific steps are as follows:

[0088] S41, based on the water body extraction results obtained by the five algorithms, calculate the covariance r between two different water body extraction results (binary images) ad , covariance r ae , covariance r de , covariance r bd , covariance r be , covariance r ce and covariance r cd .

[0089] With covariance r ad For example, the calculation formula is:

[0090]

[0091] In the formula, a j is a binary image x a The pixel value of the jth pixel in , d j is a binary image x d The pixel value of the jth pixel in , is a binary image xa The mean value of all pixel values of is the binary image x d The mean value of all pixel values of, and n is the total number of pixels of the binary image.

[0092] S42. Perform a combined calculation on the covariance obtained in S41 to obtain the precision vector y. The calculation formula is as follows:

[0093]

[0094] S43. Construct a data analysis model based on the precision vector y:

[0095] Av = y (3)

[0096] Among them, v represents the true precision of the binary image obtained by five algorithms, y is the precision vector, A is a constant vector matrix connecting the true precision v and the precision vector y. The calculation formulas for the true precision v and the constant vector matrix A are as follows:

[0097] v = [v a v b v c v d v e T (4)

[0098]

[0099] In formula (4), v a is the true precision of the binary image x a in the data analysis model; v b is the true precision of the binary image x b in the data analysis model; v c is the true precision of the binary image x c in the data analysis model; v d is the true precision of the binary image x d in the data analysis model; v e is the true precision of the binary image x e in the data analysis model.

[0100] S44. Calculate the least squares solution of equation (3) to obtain the true precision v. The true precision v reflects the accuracy of the water body extraction results of the five algorithms. The calculation formula is as follows:

[0101] v = (A T A) -1 A T y (6)

[0102] S5. Perform multi-algorithm integrated learning to obtain the optimal water body extraction algorithm:

[0103] Calculate the weights corresponding to the five algorithms according to the true accuracy v of the binary image obtained by the data parsing model; traverse all pixels in the binary image, calculate the water body probability of each pixel in the binary image using the weights, and obtain the water body probability set M; set a classification threshold, optimize and integrate the water body probability set M to obtain the final binary water body classification map, and complete the accurate extraction of the water body in the experimental area image. The specific steps are as follows:

[0104] S51. According to the true accuracy v of the binary image obtained by the data parsing model, calculate the weights w a 、w b 、w c 、w s and w e corresponding to the five algorithms respectively. The calculation formula is as follows:

[0105]

[0106] where v i is the true accuracy of the binary image x i obtained by different algorithms, and i = a to e.

[0107] S52. According to the weights corresponding to the five algorithms, traverse all pixels in the binary image, calculate the water body probability of each pixel in the binary image, and obtain the water body probability set M = {m1, m2... n n}. Among them, the calculation formula for the water body probability m j of the jth pixel in the binary image is as follows:

[0108] m j = w a a j + w b b j + w c c j + w d d j + w e e j (8)

[0109] In the formula, a j 、b j 、c j 、d j and e j are the pixel values of the jth pixel in the binary image x a 、the binary image x b 、the binary image x c 、the binary image x d and the binary image x e respectively; w a 、w b, w c , w d and w e are the weights corresponding to five algorithms respectively; j = 1, 2, …… n.

[0110] S53. Set the classification threshold to 0, that is, the threshold for water body discrimination is 0; according to the water body probability set M, perform masking processing on the pixels with water body probability greater than the classification threshold, and optimize the integration to obtain the final binary water body classification map, completing the accurate extraction of the water body in the experimental area image.

[0111] To verify the accuracy of this method, the accuracy of the water body extraction result (binary water body classification map) of this method is evaluated from different perspectives as follows:

[0112] (1) Evaluate the accuracy of the water body extraction result from the perspective of visual interpretation:

[0113] The water body extraction results of the five algorithms in the embodiment are as Figure 2 shown in (d) - (h). It can be seen from the figure that the NDWI algorithm has insufficient extraction ability for water bodies in small - area regions and thin - long river water bodies, and there is a serious phenomenon of missed extraction. The SVM algorithm and RF algorithm mis - extract a large number of buildings, roads, bridges, and farmlands as water bodies. The K - means algorithm mis - extracts roads as water bodies, and the boundary of the water body extraction result of the CNN algorithm is blurred.

[0114] The water body extraction result obtained by the method of the present invention is as Figure 2 shown in (i). It can be seen from the figure that the water body extraction method of the present invention has an ideal extraction result at the river bank edge. Compared with the water body extraction results of the five algorithms, the water body extraction result of the water body extraction method of the present invention is smoother and more accurate. The water body extraction method of the present invention can not only well distinguish the water body and non - water body regions, but also more accurately extract small targets, small - area rivers, and thin - long rivers, and distinguish the water body from the shadow.

[0115] (2) Evaluate the accuracy of the water body extraction result from the perspective of evaluation indicators:

[0116] Calculate the evaluation indicators: Obtain the GF6 (Gaofen - 6) image that includes the experimental area and has an imaging time of October 7, 2022, as Figure 2 shown in (a). Among them, the pixel size of the GF6 (Gaofen - 6) image is 2m × 2m. Manually create a water body mask, mark the water body pixel value as 1 and the non - water body pixel value as 0. This water body mask will be used as the true - value image for inspection, as Figure 2As shown in Fig. (b). Water body sample points and non-water body sample points are selected in the true value image, and a confusion matrix is constructed based on the selected water body sample points and non-water body sample points; evaluation indicators are calculated using the confusion matrix, and the evaluation indicators include Kappa coefficient (KC), F1 score, and Balanced Accuracy.

[0117] From the calculated evaluation indicators, it can be seen that the results of the RF algorithm and the water body extraction method of the present invention have better accuracy. The evaluation indicators of the water body extraction results of the water body index algorithm are the lowest, while the evaluation indicators of the water body extraction results of the K-means algorithm and the SVM algorithm are the same. The F1 score of the water body extraction results of the water body extraction method of the present invention is the highest, all exceeding 0.95; the F1 score of the water body extraction results of the method of the present invention is 4.49% and 15.14% higher than that of the RF method and the CNN method respectively; the Kappa coefficient of the water body extraction results of the method of the present invention is increased by 11.90% and 30.56% compared with the RF algorithm and the convolutional neural network algorithm respectively; the Balanced accuracy of the water body extraction results of the method of the present invention is increased by 5.43% and 12.79% respectively.

[0118] In summary, the water body extraction method of the present invention has the highest accuracy for water body extraction in remote sensing images, and the water body extraction results are the best.

[0119] The water body extraction method of the present invention is applicable to water body extraction data sets that are continuous in time and space in any region and any year.

Claims

1. A water body extraction method based on the integration of multiple artificial intelligence algorithms, characterized in that It includes the following steps: S1. Obtain remote sensing images and perform preprocessing: Obtain a set of remote sensing images containing the water area to be extracted within a specified time range; perform cloud amount screening and cloud removal processing on the set of remote sensing images to reduce the influence of cloud cover; then crop the image set according to the experimental area to obtain an image set of the water area to be extracted; perform median synthesis on the image set to obtain the final image of the water area to be extracted; S2. Perform multi-feature extraction on the final image: Obtain the optical band information of each pixel in the final image, and calculate the index band information of each pixel in the final image according to the optical band information. The index band information includes the Normalized Difference Water Index (NDWI), the Modified Normalized Difference Water Index (MNDWI), the Normalized Difference Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI), the Normalized Difference Built-up Index (NDBI), and the Bare Soil Index (BSI); S3. Using the optical band information and index band information obtained in S2, respectively apply the water body index method, support vector machine (SVM) algorithm, random forest (RF) algorithm, K-means algorithm, and convolutional neural network (CNN) algorithm to the final image for water body extraction, and obtain binary images of the water body extraction results, namely x a , x b , x c , x d , and x e ; S4. Solve the true accuracy of the water extraction result based on mathematical uncertainty analysis: Binary image x obtained according to S3 a , x b , x c , x d and x e Calculate the covariance between two different water body extraction results, and then calculate the accuracy vector y; use the accuracy vector y as the input of the data analysis model to obtain the true accuracy corresponding to the water body extraction result. S5. Perform multi-algorithm integrated learning to obtain the optimal water extraction algorithm, including the following steps: S51. Calculate the weights w corresponding to the five algorithms in S3 respectively according to the true accuracy of the binary image obtained by the data parsing model. a , w b , w c , w d and w e , and the calculation formula is as follows: Among them, v i is the true accuracy of the binary image x i obtained by different algorithms, where i = a to e; S52. Traverse all pixels in the binary image according to the weights corresponding to the five algorithms, calculate the water body probability of each pixel in the binary image, and obtain the water body probability set M = {m1, m2... m n}, where the water body probability m j of the j-th pixel in the binary image is calculated as follows: m j = w a a j + w b b j + w c c j + w d d j + w e e j (8) Wherein, a j 、b j 、c j 、d j and e j are respectively the pixel values of the j-th pixel in the binary image x a , the binary image x b , the binary image x c , the binary image x d and the binary image x e ; w a , w b , w c , w d and w e are respectively the weights corresponding to five algorithms; j = 1, 2, …… n; S53. Set the classification threshold to 0; perform masking processing on the pixels with water probability greater than the classification threshold according to the water probability set M, and optimize the integration to obtain the final binary water classification map, and complete the accurate extraction of water bodies in the experimental area image.

2. The water body extraction method according to claim 1, wherein: The set of remote sensing images in S1 is a set of Sentinel-2 satellite images.

3. The water body extraction method according to claim 1, characterized in that, The method for water body extraction using the water body index algorithm is as follows: taking the image of the experimental area as the input, performing the NDWI ratio operation on each pixel of the experimental area image using the water body index method and setting a threshold. If the NDWI of a pixel is greater than or equal to the threshold, it is a water body; if the NDWI of a pixel is less than the threshold, it is a non-water body. The water body index algorithm outputs a binary image x with only water bodies and non-water bodies a , where the threshold is set to 0.

4. The water body extraction method according to claim 1, wherein: The binary image x obtained by S3 a , x b , x c , x d and x e have the same total number of pixels, which is n for all of them.

5. The water body extraction method according to claim 1, characterized in that The method for water body extraction using the Random Forest algorithm RF and the Support Vector Machine algorithm SVM is as follows: Combining the actual situation of the experimental area, each pixel in the experimental area image is divided into 5 land cover types: water body, crop, building, vegetation, and bare land. Manually select sample pixels by visual interpretation on the experimental area image; Use the optical band information and index band information of each sample pixel as the feature variables in the training sets of the RF and SVM algorithms, and input them into the Random Forest RF algorithm and the Support Vector Machine SVM algorithm for training respectively to obtain the training results. And perform masking processing on the sample pixels with the land cover type of water body. Both the Random Forest algorithm RF and the Support Vector Machine SVM algorithm obtain binary images with only water body and non-water body, which are x b and x c .

6. The water body extraction method according to claim 5, wherein: The sample pixels include 150 water pixels and 382 non-water pixels in the experimental area image.

7. The water body extraction method according to claim 1, wherein The method for water body extraction using the K-means algorithm is as follows: all pixels in the experimental area image are divided into K given clusters, and pixels with the land cover type of water body are masked to obtain a binary image x with only water bodies and non-water bodies, d where K = 5.

8. The water body extraction method according to claim 1, characterized in that, The method for water body extraction using the Convolutional Neural Network algorithm CNN is as follows: perform water body masking on the Dynamic World (DW) dataset, extract the pixels with the classification category of water body, and obtain a binary image x with only water body and non-water body e .

9. The water body extraction method according to claim 1, characterized in that, S4 includes the following steps: S41. Calculate the covariance r between two different water body extraction results based on the water body extraction results obtained by the five algorithms ad , covariance r ae , covariance r de , covariance r bd , covariance r be , covariance r ce and covariance r cd ; With covariance r ad as an example, its calculation formula is as follows: where a j is the pixel value of the j-th pixel in the binary image x a and d j is also the pixel value of the j-th pixel in the binary image x d . is the mean value of all pixel values of the binary image x a , is also the mean value of all pixel values of the binary image x d , and n is the total number of pixels in the binary image; S42. Perform combined calculation on the covariance obtained in S41 to obtain the accuracy vector y, and the calculation formula is as follows: S43. Construct a data analysis model according to the accuracy vector y: Av = y (3) where v represents the true accuracy of the binary images obtained by five algorithms, y is the accuracy vector, A is a constant vector matrix connecting the true accuracy v and the accuracy vector y, and the calculation formulas for the true accuracy v and the constant vector matrix A are as follows: v = [v a v b v c v d v e T (4)​ In formula (4), v a is the true accuracy of the binary image x a in the data parsing model; v b is the true accuracy of the binary image x b in the data parsing model; v c is the true accuracy of the binary image x c in the data parsing model; v d is the true accuracy of the binary image x d in the data parsing model; v e is the true accuracy of the binary image x e in the data parsing model; S44. Calculate the least squares solution of equation (3) to obtain the true accuracy v, and the calculation formula is as follows: v = (A T A) -1 A T y (6).

Images