Water body merging model construction method and system based on different water body fragment types

Abstract

The invention belongs to the field of information statistics of lake and river water resources, and particularly discloses a water body merging model construction method and system based on different water body fragment types. The method comprises the following steps: extracting water body fragment vector surface data in global surface water remote sensing data; extracting fragment water body data in direct spatial association with the river network data set, and combining according to different water body types in the river network data set to obtain water body fragment vector surface data after primary combination; establishing a buffer area for the fragment water body which has no direct spatial relationship with the river network data set, and merging again according to the spatial relationship to obtain water body fragment vector surface data after secondary merging; performing neighbor analysis on the water body fragment vector surface data after the secondary combination, and combining the fragment water body data with a neighbor relation to obtain water body fragment vector surface data which are combined again; and counting the quantity of the fragmented water bodies, carrying out error analysis, and carrying out related sensitivity experiments. The accuracy of water body combination can be effectively improved.

Description

Technical Field

[0001] This application belongs to the field of information statistics of lake and river water resources. More specifically, it relates to a method and system for constructing a water body merging model based on different water body fragment types. Background Art

[0002] In the field of information statistics of lake and river water resources, professional departments such as meteorology and hydrology focus on the spatio-temporal changes in the number of water bodies within the basin. These data provide important bases for disaster prevention and control and policy formulation. Remote sensing technology has become the core means for dynamic monitoring of water bodies. Global surface water datasets based on satellites such as Landsat and Sentinel have achieved high-precision water body extraction. Fragmented water bodies are defined as discrete water body units with an area less than 0.03 km 2 in remote sensing images. Their formation causes can be divided into physical mechanisms and technical limitations:

[0003] Physically, lakes with significant seasonal water level changes will expose the lake bottom topography during the dry season, dividing the originally continuous water area into several isolated sub-lakes. There are also some fragments in the form of pseudo-fragments along the river, which often exist in narrow river channels. Such fragments are caused by the fact that when satellite imaging, the shadows projected by riverside vegetation or buildings at a specific solar altitude angle result in local misjudgment of the river channel as an independent water body; and the reflectance of high-turbidity water bodies in the near-infrared band is similar to that of land, resulting in the segmentation algorithm being unable to identify the true boundary. Water body fragments caused by technical reasons are due to defects in image segmentation algorithms and the mixed pixel effect. The traditional threshold segmentation method based on the normalized difference water index is not sensitive enough to shallow water areas. When the water depth is less than 0.5 meters, the reflectance of the water body in the short-wave infrared band drops significantly, resulting in missed detections; the wetland vegetation has a similar spectrum to the water body in the visible light band, causing over-segmentation; in medium- and low-resolution images, narrow river channels and surrounding vegetation are mixed to form fragments. For example, the width of the tributaries of the Amazon is only 20 meters during the dry season, and it is shown as 1-2 mixed pixels in Sentinel-2 images, and traditional classification algorithms cannot accurately extract them.

[0004] Although certain progress has been made in current research on the processing of fragmented water body data, due to the combined influence of satellite resolution limitations and hydrological seasonal fluctuations, more than 30% of small water bodies show significant fragmentation characteristics, resulting in an area statistical error of up to 15-20%. Therefore, the problem of river breaks at the global scale cannot be solved yet. In addition, existing research lacks a processing mechanism for pseudo-fragments along the river and isolated water body fragments.

[0005] Therefore, how to solve the water body merging error caused by fragmented water bodies due to satellite resolution limitations and hydrological seasonal fluctuations is a difficult problem that urgently needs to be studied. Summary of the Invention

[0006] Aiming at the defects of the prior art, the purpose of this application is to provide a method and system for constructing a water body merging model based on different water body fragment types, which can improve the accuracy of water body merging and effectively enhance the continuity of river channels and the integrity of lakes.

[0007] To achieve the above object, in the first aspect, this application provides a method for constructing a water body merging model based on different water body fragment types, including the following steps:

[0008] S10. Obtain and extract the water body fragment vector surface data in the global surface water remote sensing data according to the global surface water remote sensing data, and the water body fragment vector surface data includes a number of fragmented water body data;

[0009] S20. Use the river network data set to extract the fragmented water body data that has a direct spatial association with the river network data set, and merge them according to different water body types in the river network data set to obtain the water body fragment vector surface data after the first merge;

[0010] S30. Establish a buffer for the fragmented water body data in the water body fragment vector surface data after the first merge that has no direct spatial relationship with the river network data set, and merge them again according to the spatial relationship to obtain the water body fragment vector surface data after the second merge;

[0011] S40. Perform a nearest neighbor analysis on the water body fragment vector surface data after the second merge, and merge the fragmented water body data with a neighboring relationship to obtain the water body fragment vector surface data after the re-merge;

[0012] S50. Count the number of fragmented water bodies in the water body fragment vector surface data in steps S20, S30, and S40, perform error analysis, and conduct relevant sensitivity experiments.

[0013] The method for constructing a water body merging model based on different water body fragment types provided by this application, through three fragmented water body merges based on spatial relationships, namely river network merge, buffer analysis merge, and nearest neighbor analysis merge, and through three-step progressive fragmentation merge, can effectively solve problems such as water body merging deviation caused by pseudo-fragment data along rivers, data around lakeshores, and data extraction errors, and improve the accuracy of water body merging.

[0014] As a further preference, step S20 is specifically:

[0015] Perform a first buffer analysis on the water body fragment vector surface data and the river network data set to obtain the buffer of the water body fragment vector surface data and the buffer of the river network data set;

[0016] Through spatial connection, perform a spatial association analysis on the buffer zone of the water body fragment vector surface data and the buffer zone of the river network dataset to determine whether there is an intersection between each fragment water body data and the buffer zone of the river network dataset. If so, mark the intersecting part in the water body fragment vector surface data as the same type of river data. If not, execute step S30;

[0017] According to different water body types in the river network dataset, merge the same type of river fragment data to obtain the water body fragment vector surface data after the first merge.

[0018] As a further optimization, the buffer zone of the water body fragment vector surface data is obtained by constructing a 100m - 500m buffer zone with the water body fragment vector surface data as the benchmark through ArcGIS; the buffer zone of the river network dataset is obtained by constructing a 100m - 500m buffer zone with the river network vector as the benchmark through ArcGIS.

[0019] As a further optimization, in step S20, the GRNWRZ V2.0 river network dataset is used for the river network dataset.

[0020] As a further optimization, step S30 is specifically as follows:

[0021] Perform a second buffer zone analysis on the fragment water bodies in the water body fragment vector surface data after the first merge that have no direct spatial relationship with the river network dataset, and identify the buffer zones with intersections;

[0022] Mark the fragment water bodies with buffer zone intersections as homologous water bodies and perform the same type of merge to obtain the water body fragment vector surface data after the second merge.

[0023] As a further optimization, in step S30, the step of identifying the buffer zones with intersections is specifically as follows:

[0024] Calculate the area within a range of 100m - 500m around the fragment water bodies in the water body fragment vector surface data after the first merge that have no direct spatial relationship with the river network dataset through ArcGIS software, so as to form corresponding buffer zone polygons;

[0025] Perform a spatial relationship judgment on the buffer zone polygons to identify the buffer zone combinations with intersections.

[0026] As a further optimization, in step S50, use Arcgis to count the number of fragment water bodies, introduce the reference true value for error analysis, and perform a sensitivity experiment.

[0027] As a further optimization, the reference true value uses the annual maximum water area data of WaterExtent, and the sensitivity experiment calculates the influence of the buffer zone radius at intervals of 50m within a range of 100m - 500m on the merging effect.

[0028] In a second aspect, the present application provides a water body merging method based on different water body fragment types, including the following steps:

[0029] Preprocess the global surface water remote sensing image data to obtain the vector surface data of water body fragments in the target area;

[0030] Input the vector surface data of water body fragments in the target area into the water body merging model constructed by the water body merging model construction method described above to obtain the water body merging result of the target area.

[0031] In a third aspect, the present application provides a water body merging model construction system based on different water body fragment types, including:

[0032] An extraction module for obtaining and extracting the vector surface data of water body fragments in the global surface water remote sensing data according to the global surface water remote sensing data, where the vector surface data of water body fragments includes several fragment water body data;

[0033] A primary merging module for using the river network dataset to extract the fragment water body data that has a direct spatial association with the river network dataset, and merging according to different water body types in the river network dataset to obtain the vector surface data of the water body fragments after primary merging;

[0034] A secondary merging module for establishing a buffer zone for the fragment water body data in the vector surface data of the water body fragments after primary merging that has no direct spatial relationship with the river network dataset, and merging again according to the spatial relationship to obtain the vector surface data of the water body fragments after secondary merging;

[0035] A re-merging module for performing a nearest neighbor analysis on the vector surface data of the water body fragments after secondary merging, and merging the fragment water body data with a neighboring relationship to obtain the vector surface data of the water body fragments after re-merging;

[0036] An error analysis module for counting the number of fragment water bodies in the vector surface data of the water body fragments in the primary merging module, secondary merging module, and re-merging module, performing error analysis, and conducting relevant sensitivity experiments.

[0037] It can be understood that the beneficial effects of the above second aspect and third aspect can be referred to the relevant descriptions in the above first aspect, and will not be elaborated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flowchart of a water body merging model construction method based on different water body fragment types provided by an embodiment of the present application;

[0039] Figure 2 is a principle block diagram of a water body merging model construction method based on different water body fragment types provided by an embodiment of the present application;

[0040] Figure 3 It is a schematic diagram of a data merging algorithm based on a river network database provided by an embodiment of the present application;

[0041] Figure 4 It is a schematic diagram of a secondary merging algorithm based on buffer analysis and marking provided by an embodiment of the present application;

[0042] Figure 5 It is a schematic diagram of a proximity analysis algorithm provided by an embodiment of the present application;

[0043] Figure 6 It is a river network buffer analysis diagram provided by a specific embodiment of the present application; wherein, (a) and (c) are schematic diagrams of local river networks, and (b) and (d) are schematic diagrams of local river networks after corresponding buffer analysis;

[0044] Figure 7 It is a marked fusion diagram of fragmented water body data provided by a specific embodiment of the present application; wherein, (a), (c), (e), (g), (i), and (k) are schematic diagrams of local water body fragment vector surface data before marking, and (b), (d), (f), (h), (j), and (l) are schematic diagrams after marking in the corresponding areas (light blue in the figure represents the actual water body, dark blue represents the original water body data, and yellow represents the marked data);

[0045] Figure 8 It is a comparison of fragmented buffer analysis provided by a specific embodiment of the present application; wherein, (a) and (c) are distribution diagrams of local water body fragment vector surface data before buffer analysis, and (b) and (d) are buffer analysis result diagrams of the corresponding local areas;

[0046] Figure 9 It is a comparison of secondary merging results provided by a specific embodiment of the present application; wherein, (a), (c), (e), (g), (i), and (k) are schematic diagrams of local water body fragment vector surface data after primary merging, and (b), (d), (f), (h), (j), and (l) are schematic diagrams of local water body fragment vector surface data after secondary merging in the corresponding areas;

[0047] Figure 10 It is the result of proximity analysis merging provided by a specific embodiment of the present application; wherein, (a), (c), (e), (g), (i), and (k) are schematic diagrams of local areas before proximity analysis, and (b), (d), (f), (h), (j), and (l) are schematic diagrams of local areas after proximity analysis in the corresponding areas;

[0048] Figure 11 It is the effect of fragmented merging provided by a specific embodiment of the present application; wherein, (a) and (b) are schematic diagrams before and after merging of pseudo-water body fragments along the river, (c) is a schematic diagram of isolated small fragments after merging, and (d) is a schematic diagram of a repaired fractured river channel;

[0049] Figure 12 It is the combined effect of the buffer radius sensitivity experiment provided by the specific embodiment of the present application. Specific embodiments

[0050] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0051] It should be understood that in the description of the present application, the meaning of the term "a number of" is at least one, for example, one, two, etc., unless otherwise specifically defined; the meaning of the term "a plurality of" is two or more, unless otherwise specifically defined; the terms "first" and "second" etc. are used to distinguish different objects, rather than to describe a specific order of the objects; the term "and / or" includes any and all combinations of one or more of the related listed items.

[0052] In addition, the reference to "an embodiment" throughout this specification; the language such as "an embodiment", "an example" or the like means that the specific features, structures or characteristics described in connection with the embodiment are included in at least one embodiment of the present application. Therefore, the appearance of the phrase "in an embodiment;" throughout this specification, "in an embodiment" and similar language may or may not all refer to the same embodiment.

[0053] Aiming at the problem that small water bodies present significant fragmentation characteristics, resulting in significant errors in water body quantity statistics, the present application conducts research on how to restore the morphological integrity of lakes and the continuity of rivers, providing data support for lake-river classification; a method for constructing a water body merging model based on different water body fragment types is proposed, and by using river network data, the spatial relationships among fragmented water bodies, lakes and rivers are deeply explored. Aiming at the problems of pseudo-fragments along the river and river breaks in remote sensing images, according to different fragment types, combined with the existing river network data set and the spatial feature relationships among fragmented water bodies, the fragments are identified and the water bodies are merged.

[0054] As Figure 1 and 2 shown, the method for constructing a water body merging model based on different water body fragment types provided by the present application includes steps S10 to S50, which are described in detail as follows:

[0055] Step S10, obtain and extract the vector surface data of water body fragments in the global surface water remote sensing data according to the global surface water remote sensing data, and the vector surface data of water body fragments includes a number of fragmented water body data.

[0056] In step S10, remote sensing images of global surface water remote sensing data are processed through operations such as vector rasterization, projection transformation, union splitting, and fragment combination to obtain the required global water body fragment vector surface data. A river network dataset is introduced to assist in discriminating the watershed boundary and watershed hierarchy. Finally, based on the vectorization of the global surface water remote sensing image data of the Joint Research Center, the water body fragment vector surface data is extracted, and a buffer zone of 100m to 500m is established to obtain the buffer zone of the water body fragment vector surface data.

[0057] Step S20: Using the river network dataset, extract the water body fragment vector surface data that has a direct spatial association with the river network dataset, and merge them according to different water body types in the river network dataset to obtain the water body fragment vector surface data after the initial merge, which includes water body fragments classified as rivers and other unclassified independent water body fragments.

[0058] As Figure 3 shown, step S20 provided in this embodiment can specifically be:

[0059] Step S21: Conduct a first buffer analysis on the water body fragment vector surface data and the river network dataset to obtain the buffer zone of the water body fragment vector surface data and the buffer zone of the river network dataset.

[0060] Specifically, the method for constructing the buffer zone of the water body fragment vector surface data can be: using the Buffer tool in ArcGIS to construct a buffer zone of 100m to 500m based on the water body fragment vector surface data to generate the corresponding buffer polygon. The method for constructing the buffer zone of the river network dataset can be: using the Buffer tool in ArcGIS to construct a buffer zone of 100m to 500m based on the river network vector to generate the corresponding buffer polygon, so as to more comprehensively capture the water body data related to the river network.

[0061] Step S22: Through spatial join, conduct a spatial association analysis on the buffer zone of the water body fragment vector surface data and the buffer zone of the river network dataset to determine whether each fragment water body data intersects with the buffer zone of the river network dataset. If so, mark the fragment water body data as the same type of river data; if not, execute step S30.

[0062] Step S23: For the fragment water body data in the same buffer zone, recognize that they belong to the same type of river fragment dataset, and perform an operation of merging the same type of these river fragment data marked as the same type; during the merging process, fully consider factors such as the flow direction of the river and the topological structure of the water system, and according to different water body types in the river network dataset, that is, water system grading and independent water systems, integrate the river fragments in the same buffer zone through an algorithm. Finally, form a continuous and complete river dataset to ensure that the merged river data has higher accuracy and reliability in reflecting the actual water system structure.

[0063] Through step S20 of this application, river fragments can be deeply integrated with the authoritative river network dataset, making the integration of river data more scientific and reasonable, and providing more accurate data support for work such as water system analysis and water resource flow simulation in hydrological research.

[0064] Preferably, the river network dataset can adopt the GRNWRZ V2.0 high-precision river network dataset.

[0065] Step S30, as Figure 4 shown, establish a buffer for the fragmented water bodies in the vector surface data of the water bodies after the initial merger that have no direct spatial relationship with the river network dataset, and merge them again according to the spatial relationship to obtain the vector surface data of the water bodies after the secondary merger, which includes the fragmented water bodies classified as rivers and other sets of fragmented water bodies merged into one category according to the spatial relationship.

[0066] The core principle of step S30 is that when there is an intersection in the buffers of two or more fragmented water bodies, it implies logically in the geographical space that they are very likely to belong to the same continuous water body system in the natural state. During the buffer analysis process, the ArcGIS software calculates the area within a range of 100m to 500m around each fragmented water body through an accurate algorithm, thereby forming corresponding buffer polygons. Then, the spatial relationship of these buffer polygons is judged to identify the buffer combinations with intersections. Once it is determined that there is an intersection in the buffers, the corresponding fragmented water body data is marked as the same water body data, and after merging the same type of water bodies, the vector surface data of the water bodies after the secondary merger is obtained.

[0067] This process involves operations such as polygon merging and topological relationship reconstruction to ensure the rationality of the merged water body data in terms of geometric shape and spatial relationship. Through this step, the degree of fragmentation of the data is initially reduced, and the originally scattered fragmented water bodies are gradually integrated into relatively continuous water body units, laying a solid foundation for further subsequent merger work.

[0068] Step S40, perform a proximity analysis on the set of fragmented water bodies in the vector surface data of the water bodies after the secondary merger, and merge them according to the proximity relationship to obtain the vector surface data of the water bodies after the re-merger.

[0069] Although through the operations of the above steps S20 and S30, most of the fragmented water body data can be incorporated into the river network dataset, there are still areas not covered, and this area contains sets of fragmented water bodies that have been independently classified. Therefore, it is necessary to further improve the integrity of the water bodies through a proximity analysis method that expands a certain range, as Figure 5 shown.

[0070] In step S40, nearest neighbor analysis is an important method in geographic information science for identifying associations between spatial objects based on distance relationships. This application uses nearest neighbor analysis technology. By setting a dynamic value of 100m to 500m for the distance threshold, multiple classes of water body fragments can be divided, and a comprehensive and detailed search can be conducted on the vector surface data of water body fragments. This method relies on the basic theory in geographic information science for identifying spatial associations based on distance relationships. Through the Euclidean distance calculation module built into ArcGIS, nearest neighbor searches are performed on each water body fragment.

[0071] During the search process, the ArcGIS software calculates the distances between each surface data and other surface data, and filters out the surface data with the closest distance as its nearest neighbor. Once the nearest neighbor relationship is determined, these surface data with nearest neighbor relationships are marked. This marking method based on nearest neighbor relationships is based on such a geographical space cognition: water body fragments that are close in distance within a certain range may have some kind of association in the natural formation process and are very likely to belong to the same water body system. After marking all the nearest neighbor surface data, the same-class merging operation is performed again. Through the merging operation, the fragmented data with nearest neighbor relationships are merged together to obtain the vector surface data of the water body fragments after re-merging, further optimizing the continuity and integrity of the water body data and enhancing the logic of the data in spatial distribution. After this step, the fragmentation situation of the water body data is further improved, and the overall quality of the river network data is significantly enhanced.

[0072] Step S50: Count the number of fragmented water bodies in the vector surface data of water body fragments obtained in each of steps S20, S30, and S40, conduct error analysis, and conduct relevant sensitivity experiments.

[0073] Specifically, Arcgis can be used to count the number of fragmented water bodies, a reference true value is introduced for error analysis, and sensitivity experiments are conducted.

[0074] Among them, the reference true value uses the annual maximum water area data of WaterExtent; the sensitivity experiment calculates the impact of the buffer radius at intervals of 50m within the range of 100m to 500m on the merging effect.

[0075] The method for constructing a water body merging model based on different water body fragment types provided in this embodiment can effectively solve problems such as water body merging deviation caused by pseudo-fragmented data along the river, data around the lake shore, and data extraction errors through three fragment water body mergings based on spatial relationships, namely river network merging, buffer analysis merging, and nearest neighbor analysis merging. Through three-step progressive fragment merging, the accuracy of water body merging is improved.

[0076] In addition, this application also provides a water body merging method based on different water body fragment types, including step 1 and step 2, which are described in detail as follows:

[0077] Step 1: Preprocess the global surface water remote sensing image data to obtain the vector surface data of water body fragments in the target area.

[0078] Specifically, the preprocessing can be carried out by operations such as vector rasterization, projection transformation, union splitting, and fragment combination on the global surface water remote sensing data to obtain the required global vector surface data of water body fragments.

[0079] Step 2: Input the vector surface data of water body fragments in the target area into the water body merging model constructed by the above water body merging model construction method to obtain the water body merging result in the target area.

[0080] It should be noted that the water body merging method provided in this embodiment can be applicable to the statistics of the global scope, and can also be applicable to the dynamic monitoring of the water body quantity within a specific area, such as the dynamic monitoring of the number of ice lakes in the glacier area. The dynamic monitoring of the number of ice lakes in the glacier area is mainly considered for the following reasons:

[0082] By dynamically monitoring the number of ice lakes in the glacier area, the precursors of ice lake outburst flood disasters can be identified. Accurate lake and river classification is not only the basic support for constructing an intelligent water ecosystem, but also can provide a scientific basis for optimizing the drought prevention and control emergency dispatching plan by evaluating the lake spatial distribution and water resource carrying capacity.

[0083] Based on the same inventive concept, the present application also provides a water body merging construction system based on different water body fragment types.

[0084] Among them, an extraction module is used to obtain and extract the vector surface data of water body fragments in the global surface water remote sensing data according to the global surface water remote sensing data. The vector surface data of water body fragments includes several fragment water body data.

[0085] The primary merging module is used to extract the fragment water body data that has a direct spatial association with the river network dataset by using the river network dataset, and merge them according to different water body types in the river network dataset to obtain the vector surface data of water body fragments after primary merging.

[0086] The secondary merging module is used to establish a buffer zone for the fragment water body data in the vector surface data of water body fragments after primary merging that has no direct spatial relationship with the river network dataset, and merge them again according to the spatial relationship to obtain the vector surface data of water body fragments after secondary merging.

[0087] The re-merging module is used to perform a proximity analysis on the vector surface data of water body fragments after secondary merging, and merge the fragment water body data with a proximity relationship to obtain the vector surface data of water body fragments after re-merging;

[0088] An error analysis module is used to count the number of fragmented water bodies in the water body debris vector surface data in the initial merging module, the secondary merging module, and the re-merging module, perform error analysis, and conduct relevant sensitivity experiments.

[0089] Specifically, for the functions of each module provided in this embodiment, reference can be made to the detailed description in the foregoing embodiment of the water body merging model construction method, and details are not described herein again.

[0090] To more clearly illustrate the water body merging model construction method provided in this application based on different water body debris types, the following is described in conjunction with specific embodiments:

[0091] The fragmented water body merging model construction method based on different water body debris types provided in this specific embodiment includes the following steps:

[0092] Step 1, identification and extraction of fragmented water bodies. Set an area threshold less than 1 km 2 Specify the size of fragmented water bodies, and at the same time introduce a river network dataset to assist in discriminating the watershed boundary and watershed hierarchy. Finally, based on the vectorization of the global surface water remote sensing image data of the Joint Research Centre, the water body debris vector surface data is extracted, and a buffer zone of 100 m to 500 m is established to obtain the buffer zone of the water body debris vector surface data, thereby helping to realize the subsequent merging process of fragmented water bodies.

[0093] Secondly, problems such as unclear watershed boundary division and unclear sub-watershed hierarchy relationship directly affect the accurate extraction of river spatial distribution characteristics. For this reason, this embodiment introduces the GRNWRZ V2.0 river network dataset as an auxiliary discrimination basis. This dataset constructs a seven-level classification system based on the differences in river basin characteristics: L1 is the set of rivers flowing into the same ocean or inland basin, L2 is divided according to the drainage area of external rivers or the catchment area flowing into inland basins, and L3-L7 are hierarchically subdivided based on the decreasing drainage area. Through comparative analysis, the river network data at the L5 level can not only ensure that the main rivers in the study area are covered, but also effectively separate connected lakes and independent rivers.

[0094] Step 2, Initial merging of the GRNWRZ V2.0 river network dataset. Conduct buffer analysis on the GRNWRZ V2.0 river line dataset globally, with the buffer distance ranging from 100 to 500 m. Then perform a spatial join operation to conduct a spatial association analysis between the water body fragment vector polygon data and the buffer of the river network dataset, and determine whether each water body fragment polygon data intersects with the river network buffer. If there is an intersection, mark the water body data as river data. Then, for the water body data within the same buffer, it is considered that they belong to the same type of river fragment dataset, and perform an operation to merge these river fragments marked as the same type; during the merging process, fully consider factors such as the flow direction of the river and the topological structure of the water system, and integrate the river fragments within the same buffer through an algorithm. Finally, obtain the vector polygon data of the water body fragments after the initial merging, ensuring that the merged river data has higher accuracy and reliability in reflecting the actual water system structure.

[0095] The core of merging using the river network dataset is to grasp the relationships of intersection, inclusion, and a certain surrounding area between the river network buffer and the water body fragment vector polygon data, mark all water bodies that meet the above relationships with the GRNWRZ V2.0 dataset, and thus determine homologous fragment water bodies. That is, perform a one-by-one spatial join operation between each water body dataset and the river network dataset in its corresponding area. Then perform a fusion operation on it to obtain the vector polygon data of the water body fragments after the initial merging.

[0096] Step 3, Secondary merging of buffer analysis and marking. In Step 2, through the spatial join and fusion operations of the river network dataset, only part of the vector polygon data of the water body fragments can be merged, that is, the part on the river network path, but the part not on the path has not been merged, such as unrecorded rivers and lake fragments. Therefore, further merging operations are required. First, conduct buffer analysis on each polygon feature in the vector polygon data of the water body fragments after the initial merging generated in the previous step, with the buffer distance set to 100 - 500 m. Calculate the intersection relationships of each buffer, assign classification values according to the intersection relationships, and buffers with the same value can be considered homologous water bodies; after calculating all the intersecting buffer water body fragments, perform a fusion operation according to the CLASS value to obtain the vector polygon data of the water body fragments after the secondary merging.

[0097] Step 4, Deep merging assisted by nearest neighbor analysis. In Step 3, most fragmented data can be merged through buffer analysis, but there are still some uncovered areas. Therefore, it is necessary to further improve the integrity of water bodies by means of expanding a certain range of nearest neighbor analysis. First, use the vector surface data of the water body fragments after the secondary merging generated in the previous step to perform nearest neighbor analysis on each surface element, calculate the distance relationship between these small fragmented water bodies and other surrounding surface elements, and find the adjacent water body fragments. Next, according to the results of the nearest neighbor analysis, for each fragmented water body, classify it according to its nearest neighbor relationship with other water body fragments. The classification rule is: if the distance between two water body fragments is relatively close, it is considered that they belong to the same type of water body fragments. By setting a dynamic value according to the distance threshold, multiple classes of water body fragments can be divided. For each classification group, assign a unified classification value, and the water body fragments within the same group can be further merged. Fuse the water body fragments of the same classification through the Dissolve operation to obtain a fused small water body dataset, and perform a spatial join operation with the vector surface data of the water body fragments after the secondary merging. Finally, perform a fusion operation to obtain the vector surface data of the water body fragments merged again.

[0098] Step 5, Evaluate the difference between the experimental results and the true values. Discuss the experimental processes and result analysis of different methods. To better explain the influence of different methods on the merging efficiency of fragmented water body data each time, six representative regions will be selected for analysis, and the merging rate, merging accuracy, and area change will be used as evaluation indicators. The test regions are typical regions (high-density fragmented areas) of the six continents in 2020, the input data is the GSW vector water body data, and the WaterExtent annual maximum water area data is used as the benchmark true value. The parameter settings of the merging algorithm are shown in Table 1 below.

[0099] Table 1 Parameter setting table for each step

[0100]

[0101] By comparing the quantitative relationship of the water body surface elements before and after fragment merging, the merging ability of this method can be reflected to a certain extent. The merging rate is an indicator reflecting the merging effect. The calculation formula is:

[0102]

[0103] In the formula, Merge R represents the merging rate; Water true represents the number of correctly merged water bodies; Water merged represents the number of water bodies to be merged.

[0104] To evaluate the effectiveness of the merging algorithm, the false merge rate is introduced as an evaluation index. The false merge rate is the ratio of the number of pairs of fragments that are wrongly merged to the number of pairs of fragments that are merged (which should actually be independent water bodies). The formula can be expressed as:

[0105]

[0106] In the formula, FMR represents the false merge rate; Water false represents the number of water bodies that are wrongly merged; Water merged represents the number of water bodies that are correctly merged.

[0107] (1) Preliminary merging experiment results and analysis based on the GRNWRZV2.0 river network dataset

[0108] First, it is necessary to perform buffer analysis on the river network line data to simulate the width of the river channel so that it can cover the fragmented water body data within the river range. The river network buffer analysis is as Figure 6 shown.

[0109] Based on the merging method of the river network data, it mainly merges the pseudo-fragments along the river. Select typical regions containing river pseudo-fragments on six continents of the world from the 2020 GSW data as experimental objects, perform river network merging operations in sequence, compare the merging results with the WaterExtent data, and calculate the merging rate and false merge rate of each region. The quantities, merging rates, and false merge rates before and after merging are shown in Table 2 below.

[0110] Table 2 Merging quantities of river network fragments in typical regions

[0111]

[0112] Testing the fragment merging ability of the river network dataset in typical regions, it is found that the highest merging rate is 92.3%, the lowest is 83.24%, and the average merging rate is 87.96%; the highest false merge rate is 20.48%, the lowest is 9.05%, and the average false merge rate is 13.5%.

[0113] From Figure 7It can be seen from the comparison that the initially unmerged water bodies contain many unconnected and separate water body surface data, and the merging operation significantly reduces the number of fragmented water bodies. Through spatial joining and buffer analysis, the original water body fragments are effectively joined into larger water body areas, especially on the river network paths where the water body connection is more obvious. In addition, the increase in the yellow area indicates that the continuity of the water body data has been improved through the merging method, and the distribution of the fragmented water bodies tends to be smoother. The merged water body areas show a more consistent shape. However, although the merging method effectively processes most of the fragments, some water body fragments in the edge areas are still not fully merged. These fragments are mainly located at the boundaries of the buffer zones, probably because the buffer zone settings are not sufficient to cover all the fragmented water bodies.

[0114] Table 3 Comparison of the number of river network fragments before and after merging

[0115]

[0116] River network merging operations were performed on the water body data in the study area in 2020, and a comparison table of the number of river network fragments before and after merging was obtained. As can be seen from Table 3, the number of water body surface elements before merging the river network data was 31,514,783, and the number after merging was 22,905,547. A total of 34,831 surface element data were merged, and the merging ratio was 27.32%.

[0117] (2) Results and analysis of the secondary merging experiment based on buffer analysis and marking

[0118] In the experiment in (1), there were some fragmented water body surface elements that could not be covered by the river network. Therefore, buffer analysis was performed on the remaining fragmented water body vector surface elements as Figure 8 shown, and then the spatial relationships between the buffer zones of each water body fragmented vector surface element were calculated. Finally, a fusion and merging operation was carried out, and the secondary merging results are as Figure 9 shown.

[0119] The secondary merging method based on buffer analysis mainly merges the water body fragments caused by terrain and shadows. Typical areas containing fragments on six continents of the world were selected from the 2020 GSW data as experimental objects. Buffer analysis and merging operations were carried out in sequence, and the merged results were compared with the WaterExtent data to calculate the merging rate and mismerging rate of each area. The quantity, merging rate, and mismerging rate before and after merging are shown in Table 4 below.

[0120] Table 4 Number of river network fragments merged in typical areas

[0121]

[0122]

[0123] The fragmentation merging ability of the buffer analysis in the typical area was tested, and it was found that the highest merging rate was 93.4%, the lowest was 83.47%, and the average merging rate was 88.76%; the highest mis-merging rate was 19.8%, the lowest was 7.06%, and the average mis-merging rate was 12.59%.

[0124] As Figure 9 shown, the pink area represents the water surface data after the first merging, and the cyan area represents the independent water body fragment surface elements that have not been merged yet. These data are spatially scattered, forming multiple discontinuous fragments. By performing buffer analysis on the spatially discrete water body fragment surface data, it plays a certain positive role in the merging of water body fragments, merging the originally scattered small water body fragments into larger areas, merging independent and adjacent water body fragments into a whole, and the water body fragments near the river are also incorporated into the river. This shows that through this method, the water bodies become no longer independent, and the correlation between water body fragments and lakes or rivers becomes stronger, improving the integrity of the water body area.

[0125] Table 5 Comparison of the number of water bodies after secondary merging

[0126]

[0127] After performing the secondary merging operation on the data after all river network fragment merging operations, a comparison table of the number before and after the secondary merging was obtained. As can be seen from Table 5, the number of water surface elements in this area before the secondary merging of buffer analysis and marking was 22,905,547, and the number after merging was 4,064,639. A total of 18,840,908 surface element data were merged, and the merging ratio was 82.23%.

[0128] (3) Results and analysis of the re-merging experiment based on the nearest neighbor analysis

[0129] The merging method based on the nearest neighbor analysis mainly merges isolated water body fragments and broken rivers. Select typical areas containing fragments on six continents of the world from the 2020 GSW data as the experimental objects, perform buffer analysis merging operations in sequence, compare the merging results with the WaterExtent data, and calculate the merging rate and mis-merging rate of each area. The quantity, merging rate, and mis-merging rate before and after merging are shown in Table 6 below:

[0130] Table 6 Table of the number of river network fragment mergings in typical areas

[0131]

[0132] The fragmentation merging ability of the nearest neighbor analysis in the typical area was tested, and it was found that the highest merging rate was 87.31%, the lowest was 72.6%, and the average merging rate was 81.33%; the highest mis-merging rate was 37.7%, the lowest was 14.53%, and the average mis-merging rate was 23.4%.

[0133] Most of the spatially discrete water body fragments and pseudo-water body fragments along the river course can be solved by the first two experimental methods, but the water body fragment problem of discontinuous rivers cannot be solved. Through the nearest neighbor analysis, the broken main river course can be repaired and the nearest neighbor table of the water body fragment data that are close to each other after merging can be obtained. The connection results after analysis based on the nearest neighbor table are as Figure 10 shown.

[0134] Table 7 Comparison of the number of water bodies merged again by nearest neighbor analysis

[0135]

[0136] 3 Perform the nearest neighbor analysis and merging operation on all the data after buffer analysis and secondary merging processing, and obtain the quantity comparison table before and after the nearest neighbor analysis. As can be seen from Table 7, the number of water body surface elements in this area before the buffer analysis and the secondary merging of markings is 4,064,639, and the number after merging is 3,764,204. A total of 300,435 surface element data are merged, and the merging ratio is 7.39%.

[0137] (4) Error analysis

[0138] In order to evaluate the difference between the experimental results and the true values, and analyze the sources and impacts of errors. In this subsection, the method of this study will be used to compare the merging effects with the traditional buffer analysis and multi-temporal analysis, and analyze the reasons for the errors.

[0139] The comparison of the merging effects of different methods is shown in Table 8.

[0140] Table 8 Comparison of merging effects

[0141]

[0142] As shown in Table 8 above, the mis-merging rate of the method used in this study is 6.7% lower than that of the traditional method and 3.1% lower than that of the multi-temporal method. This is mainly due to the prior constraint of the river network. By GRNWRZV2.0, the over-connection of non-river channel areas is avoided. And by using the distance threshold for dynamic control to repair the broken river channels, the mis-merging rate of the fragments is reduced. During the process of water body fragment processing, the wetland patches in the riparian zone are wrongly integrated into the river system due to the method of this study, resulting in a significant increase in errors.

[0143] The progressive optimization method solves various water body fragment problems in different aspects. Among them, 27.32% of the pseudo-water body fragments along the river are merged in the river network fusion stage ( Figure 11 a->b), and the buffer analysis and marking method merges 82.23% of the isolated small water body fragments ( Figure 11 c), and 7.39% of the broken river channels are repaired by using the nearest neighbor analysis method (Figure 11 d).

[0144] Regarding the influence of different buffer radii on the merging effect, this study systematically evaluated the influence of different buffer radii on the merging effect of water body debris through sensitivity experiments. The results show that ( Figure 12 ), when the analysis radius of the result buffer is set to 300 m, the best balance is achieved in terms of the merging rate and recall rate. It should be noted that when the radius exceeds 300 m, the system's false merging rate shows an increasing trend, which is mainly due to the abnormal cross-basin connection phenomenon caused by the excessive radius. When the buffer range expands excessively, the algorithm will wrongly aggregate water body debris belonging to different basins, thus destroying its natural association characteristics.

[0145] Those skilled in the art can easily understand that the above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for constructing a water body merging model based on different water body fragment types, characterized in that It includes the following steps: S10. Obtain and extract the vector surface data of water body fragments from the global surface water remote sensing data according to the global surface water remote sensing data. The vector surface data of water body fragments includes a number of fragmented water body data; S20. Use the river network dataset to extract the fragmented water body data that has a direct spatial association with the river network dataset, and merge them according to different water body types in the river network dataset to obtain the vector surface data of water body fragments after the first merge; S30. Establish a buffer for the fragmented water body data in the vector surface data of water body fragments after the first merge that has no direct spatial relationship with the river network dataset, and merge them again according to the spatial relationship to obtain the vector surface data of water body fragments after the second merge; S40. Perform a nearest neighbor analysis on the vector surface data of water body fragments after the second merge, and merge the fragmented water body data with a neighboring relationship to obtain the vector surface data of water body fragments after the re-merge; S50. Count the number of fragmented water bodies in the vector surface data of water body fragments in steps S20, S30, and S40, perform error analysis, and conduct relevant sensitivity experiments.

2. The method for constructing a water body merging model based on different water body debris types according to claim 1, wherein, Step S20 is specifically as follows: Perform a first buffer analysis on the vector surface data of water body fragments and the river network dataset to obtain the buffer of the vector surface data of water body fragments and the buffer of the river network dataset; Through spatial join, perform a spatial association analysis on the buffer of the vector surface data of water body fragments and the buffer of the river network dataset to determine whether each fragmented water body data intersects with the buffer of the river network dataset. If so, mark the intersecting part in the vector surface data of water body fragments as the same type of river data. If not, execute step S30; Merge the fragmented data of the same type of river according to different water body types in the river network dataset to obtain the vector surface data of water body fragments after the first merge.

3. The method for constructing a water body merging model based on different water body debris types according to claim 2, wherein The buffer of the vector surface data of water body fragments is obtained by constructing a 100m - 500m buffer based on the vector surface data of water body fragments through ArcGIS; the buffer of the river network dataset is obtained by constructing a 100m - 500m buffer based on the river network vector through ArcGIS.

4. The method for constructing a water body merging model based on different water body debris types according to claim 1, wherein In step S20, the GRNWRZ V2.0 river network dataset is used as the river network dataset.

5. The method for constructing a water body merging model based on different water body debris types according to claim 1, wherein Step S30 is specifically as follows: Perform a second buffer analysis on the fragmented water body data in the vector surface data of water body fragments after the first merge that has no direct spatial relationship with the river network dataset, and identify the buffers with intersections; Mark the fragmented water bodies with buffer intersections as homologous water bodies and perform a merge of the same type to obtain the vector surface data of water body fragments after the second merge.

6. The method for constructing a water body merging model based on different water body debris types according to claim 5, wherein, In step S30, the step of identifying the buffers with intersections is specifically as follows: Calculate the area within 100m - 500m around the fragmented water body data in the vector surface data of water body fragments after the first merge that has no direct spatial relationship with the river network dataset through ArcGIS software, so as to form the corresponding buffer polygon; Perform a spatial relationship judgment on the buffer polygon to identify the buffer combinations with intersections.

7. The method for constructing a water body merging model based on different water body debris types according to claim 1, wherein In step S50, use Arcgis to count the number of fragmented water bodies, introduce the reference true value for error analysis, and conduct sensitivity experiments.

8. The method for constructing a water body merging model based on different water body debris types according to claim 7, characterized in that, The reference true value uses the annual maximum water area data of WaterExtent, and the sensitivity experiment calculates the influence of the buffer radius at intervals of 50 m in the range of 100 m to 500 m on the merging effect.

9. A water body merging method based on different water body fragment types, characterized in that It includes the following steps: Preprocess the global surface water remote sensing image data to obtain the vector surface data of water body fragments in the target area; Input the vector surface data of water body fragments in the target area into the water body merging model constructed by the water body merging model construction method described in claim 1 to obtain the water body merging result of the target area.

10. A water body merging model construction system based on different water body fragment types, characterized in that, It includes: An extraction module, which is used to obtain and extract the vector surface data of water body fragments in the global surface water remote sensing data according to the global surface water remote sensing data, and the vector surface data of water body fragments includes several fragment water body data; A primary merging module, which is used to use the river network dataset to extract the fragment water body data that has a direct spatial association with the river network dataset, and merge them according to different water body types in the river network dataset to obtain the vector surface data of water body fragments after primary merging; A secondary merging module, which is used to establish a buffer for the fragment water body data in the vector surface data of water body fragments after primary merging that has no direct spatial relationship with the river network dataset, and merge them again according to the spatial relationship to obtain the vector surface data of water body fragments after secondary merging; A re-merging module, which is used to perform a proximity analysis on the vector surface data of water body fragments after secondary merging, and merge the fragment water body data with a proximity relationship to obtain the vector surface data of water body fragments after re-merging; An error analysis module, which is used to count the number of fragment water bodies in the vector surface data of water body fragments in the primary merging module, secondary merging module and re-merging module, perform error analysis, and conduct relevant sensitivity experiments.

Images