A road waterlogging point prediction method integrating historical waterlogging points and spatial clustering

Abstract

The present application discloses a method for predicting road waterlogging points that integrates historical waterlogging points and spatial clustering, including the following steps: Step S1: Rapidly estimate rainfall runoff and drainage volume, perform rapid numerical simulation of surface waterlogging, and obtain an estimated value of surface waterlogging depth; Step S2: Based on the location of historical waterlogging points, integrate the estimated value of surface waterlogging depth and determine the high-risk area for waterlogging; Step S3: Use the improved DBSCAN clustering algorithm to predict the waterlogging point with the high-risk area for waterlogging as input; Step S4: Based on geocoding, output the predicted waterlogging point with latitude and longitude coordinates, geographic address, and maximum waterlogging depth information. The present application obtains location address information based on the surface waterlogging depth prediction matrix, thereby forming point-based waterlogging address information, which facilitates emergency response personnel to quickly arrive at the scene.

Description

Technical Field

[0001] The present application relates to the field of road waterlogging point prediction, and in particular to a road waterlogging point prediction method that integrates historical waterlogging points and spatial clustering. Background Art

[0002] With the acceleration of urbanization, the problem of urban waterlogging has become increasingly prominent. Urban waterlogging not only affects the normal lives of residents, but may also cause serious economic losses and social problems. Given the sudden occurrence of waterlogging and the complexity of its spatial distribution, how to effectively prevent and respond to urban waterlogging has become a key issue that needs to be urgently addressed in urban management. Against this background, waterlogging point prediction technology has become an important research hotspot. In the early days, waterlogging point prediction mainly relied on manual field surveys. This method had high accuracy but low efficiency. Later, with the development of waterlogging numerical simulation technology, the use of hydrological models based on waterlogging numerical simulation for waterlogging point prediction has become the mainstream. At present, waterlogging numerical simulation technology is combined with other technologies (GIS technology, artificial intelligence, big data) to further improve the prediction effect.

[0003] Although the numerical simulation technology of waterlogging has made some progress, there are still some problems in its practical application. At present, the output of the hydrological model based on numerical simulation of waterlogging is still based on the surface of waterlogging results. In practical application, it is difficult to accurately obtain point information and cannot be directly used to guide the front-line emergency response personnel to quickly arrive at the scene for disposal. Therefore, it needs to be improved. Summary of the invention

[0004] In view of the shortcomings of the existing technology, the present application provides a road waterlogging point prediction method that integrates historical waterlogging points and spatial clustering, aiming to solve the above problems.

[0005] A road waterlogging point prediction method integrating historical waterlogging points and spatial clustering includes the following steps:

[0006] Step S1: Rapidly estimate the rainfall runoff and drainage volume, and perform a rapid numerical simulation of surface waterlogging to obtain an estimated value of the depth of surface waterlogging;

[0007] Step S2: Based on the historical waterlogging location and the estimated value of surface waterlogging depth, determine the high-risk area for waterlogging;

[0008] Step S3: using the improved DBSCAN clustering algorithm to predict waterlogging points with high-risk areas of waterlogging as input;

[0009] Step S4: Based on the geocoding, output the predicted waterlogging point with latitude and longitude coordinates, geographic address, and maximum waterlogging depth information.

[0010] Optionally, the rapid numerical simulation of landmark waterlogging in step S1 specifically includes the following steps:

[0011] Step S1.1: Based on the watershed division of the generalized drainage pipe points, the regional sub-watersheds are formed by establishing Thiessen polygons;

[0012] Step S1.2: Calculate the rainfall runoff depth of each pixel through the rainfall-runoff curve number model (SCS-CN), convert it into runoff, accumulate the runoff of each catchment pixel, and obtain the runoff of each catchment;

[0013] Step S1.3: According to the runoff of each catchment area and the drainage volume of the drainage network, the total amount of surface water in each catchment area is obtained;

[0014] Step S1.4: Based on the total amount of surface water in each catchment area and the digital elevation model (DEM), the water depth of all pixels is estimated through passive inundation calculation, and the surface water depth prediction matrix is ​​output.

[0015] Optionally, the calculation formula in step S1.2 specifically includes:

[0016] According to SCS-CN, the rainfall-runoff relationship is obtained:

[0017] ,

[0018] Through empirical formula Calculate the S value, and then get the runoff of each pixel in the region, where the CN value is an empirical value based on the land use type, P is the rainfall depth, Q is the runoff depth, and I a is the initial loss;

[0019] The rainfall runoff of the catchment area is the total rainfall runoff of all pixels in the area, that is:

[0020] ,

[0021] Where Q pix is the runoff depth of a single pixel in catchment area i, Area pix is the actual area of ​​the pixel.

[0022] Optionally, the calculation formula in step S1.3 specifically includes:

[0023] The surface water volume of the catchment area is equal to the rainfall runoff in the catchment area minus the drainage volume of the drainage network, that is,

[0024] ,

[0025] Where V remain_i is the total surface water volume in catchment area i, V catch_i is the total rainfall runoff of catchment area i, V drainage_i is the total discharge of the drainage system in catchment area i;

[0026] Total displacement V drainage_i The average flow rate Q i And the drainage time t is estimated, that is

[0027] ,

[0028] Average flow rate Q i The Manning formula is simplified through the average pipe diameter, Manning roughness coefficient, and average slope, namely:

[0029] ,

[0030] Where n is the Manning roughness coefficient, D avg is the average diameter of all pipelines in catchment area i, S avg is the average slope of all pipelines in catchment area i, is the flow rate estimation constant.

[0031] Optionally, the calculation formula in step S1.4 specifically includes:

[0032] First, sort all pixel DEM values ​​from small to large to get a sequence ,

[0033] Maximum water accumulation height H in the catchment area, surface water volume V remain 、Pixel area pix The following relationship exists:

[0034] ,

[0035] So, according to the sequence , start from the minimum elevation and obtain m and D that satisfy the above formula m , directly use it as the highest elevation point of the submerged area, and the maximum submerged depth of the area is D m ,

[0036] Finally, the surface water depth of all pixels is calculated, that is, , output surface water depth prediction matrix M pre .

[0037] Optionally, step S2 specifically includes the following steps:

[0038] Step S2.1: Establish a historical waterlogging point database containing location information longitude and latitude based on the waterlogging points obtained through prior knowledge and historical predictions of this algorithm;

[0039] Step S2.2: The longitude and latitude coordinates of historical flooding points ( , i=1,2...n) into the surface water depth prediction matrix Mpre The target pixel coordinate position in , i=1,2...n), set the matrix area with the target pixel coordinate position as the center as the historical waterlogging high risk area matrix M his , where the matrix pixel values ​​of historical high-risk waterlogging areas are set to 1, and the matrix pixel values ​​of non-historical high-risk waterlogging areas are set to 0;

[0040] Step S2.3: Calculate the surface water depth prediction matrix M output by step S1.4 pre , the historical high-risk area matrix M output by the dot multiplication step S2.2 his , and obtain the matrix M of predicted high-risk areas for waterlogging hrisk :

[0041] ,

[0042] That is M pre There is water accumulation, and M his Pixels with a value of 1 (historically at risk) retain the water depth value, while the remaining M his Pixel points with value 0 are assigned value 0.

[0043] Optionally, step S3 specifically includes the following steps:

[0044] Step S3.1: Set the maximum expected value (MaxNumCl), minimum expected value (MinNumCl), initial search radius (Eps(0)), radius adjustment gain (ε) of the target water accumulation point clusters, and set the minimum element contained in the same cluster to 1;

[0045] Step S3.2: Use the predicted high-risk area M of waterlogging in step S2.3 hrisk , the initial search radius Eps(0) and the minimum number of included elements 1 are used as input to complete a DBSCAN algorithm clustering;

[0046] Step S3.3: Remove the category clusters with 1 element as noise, and determine the current actual number of category clusters NumCl, that is:

[0047] ,

[0048] Among them, NumCl is the current actual number of category clusters, NumCl step NumCl is the number of initial category clusters after clustering is completed. 1pts is the number of category clusters with 1 element;

[0049] Step S3.4: Adjust the search radius according to the relationship between the actual number of category clusters and the maximum and minimum expected values; if the actual number of category clusters NumCl is less than the expected minimum MinNumCl, the search radius is reduced according to the adjustment gain ε; if the actual number of category clusters NumCl is greater than the expected maximum MaxNumCl, the search radius is increased according to the adjustment gain ε, and steps S3.2-S3.4 are repeated, that is, the search radius is first calculated:

[0050]

[0051] Then M hrisk ,Eps(step),the same cluster contains the minimum element 1 as input, and iterates the DBSCAN algorithm clustering;

[0052] Step S3.5: According to the number of cluster categories, when the number of categories is between the expected maximum and minimum values, the algorithm is exited to obtain the final cluster category.

[0053] Optionally, the step S4 specifically includes the following steps:

[0054] Step S4.1: According to the cluster category output in step S3.5, obtain the pixel location coordinate set of the cluster center point ;

[0055] Step S4.2: Combine the coordinate position with the predicted high-risk area matrix M output from step S2.3 hrisk , get the maximum water depth information ( ,i=1,2... NumCl final );

[0056] Step S4.3: Based on the obtained cluster center pixel position coordinate set, convert it into a longitude and latitude coordinate position set, obtain the geographic address information, and convert the pixel row and column number coordinate set ( , i=1,2...NumCl final ), first converted into longitude and latitude coordinates ; Using the longitude and latitude coordinates as parameters, obtain the geographical address information of the location, thereby forming the geographical address information of the waterlogging point (Address i , i=1,2... NumCl final );

[0057] Step S4.4: The latitude and longitude coordinates, geographic addresses, and maximum water depth information of all waterlogging points are integrated into a one-dimensional array set, and the complete attribute information of the predicted waterlogging points is output, namely:

[0058] .

[0059] In summary, the present application includes at least one of the following beneficial technical effects:

[0060] 1. This application obtains location address information based on the surface water depth prediction matrix, thereby forming point-based water address information, which facilitates emergency response personnel to quickly arrive at the scene.

[0061] 2. The total amount of surface water is quickly assessed by simplifying the Manning formula through rainfall runoff and pipe network drainage in the catchment area, and then the depth of water accumulation is quickly predicted based on DEM and passive inundation of the total amount of water in the catchment area.

[0062] 3. Through geographic inverse coding, the geographic coordinates, address information and water depth of the point are integrated to output the waterlogging point information.

[0063] 4. This application uses the historical waterlogging database to determine the high-risk areas for waterlogging, and combines it with the prediction results to form a highly accurate prediction of high-risk areas for waterlogging. Secondly, an improved DBSCAN clustering algorithm is used to solve the problem that the simulation results of the waterlogged surface are difficult to cluster. The clustering effect is optimized by dynamically adjusting the search radius to improve the accuracy of waterlogging point extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a flowchart of an embodiment of the present application.

[0065] Figure 2 is the surface water depth prediction matrix M in the embodiment of the present application pre Schematic diagram.

[0066] Figure 3 is the historical high-risk area matrix M in the embodiment of the present application his Schematic diagram.

[0067] Figure 4 is the matrix M for predicting high-risk areas for waterlogging in the embodiment of the present application hrisk Schematic diagram. DETAILED DESCRIPTION

[0068] In order to facilitate the understanding of the present application, the present application is described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that when an element is described as "fixed to" another element, it can be directly on the other element, or there can be one or more centered elements therebetween. When an element is described as "connected to" another element, it can be directly connected to the other element, or there can be one or more centered elements therebetween. The terms "vertical", "horizontal", "left", "right" and similar expressions used in this specification are for illustrative purposes only.

[0069] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0070] The present application embodiment discloses a method for predicting road waterlogging points by integrating historical waterlogging points and spatial clustering. Figure 1 , including the following steps: Step S1: Through rapid estimation of rainfall runoff and drainage, rapid numerical simulation of surface waterlogging is performed to obtain an estimated value of surface waterlogging depth; Step S2: Based on the historical waterlogging point locations, the estimated value of surface waterlogging depth is integrated to determine the high-risk areas for waterlogging; Step S3: Using the improved DBSCAN clustering algorithm, the high-risk areas for waterlogging are used as input to predict the waterlogging points; Step S4: Based on geographic coding, the predicted waterlogging points with latitude and longitude coordinates, geographic address, and maximum waterlogging depth information are output.

[0071] Reference Figure 1 and Figure 2 The rapid numerical simulation of landmark waterlogging in step S1 specifically includes the following steps:

[0072] Step S1.1: Based on the watershed division of the generalized drainage pipe points, the regional sub-watersheds are formed by establishing Thiessen polygons; Step S1.2: Through the rainfall runoff curve number model (SCS-CN), the rainfall runoff depth (unit: mm) of each pixel is calculated and converted into runoff volume, and the runoff volume of each watershed pixel is accumulated to obtain the runoff volume of each watershed.

[0073] According to SCS-CN, the rainfall-runoff relationship is obtained:

[0074] ,

[0075] Through empirical formula Calculate the S value, and then get the runoff of each pixel in the region, where the CN value is an empirical value based on the land use type, P is the rainfall depth, Q is the runoff depth, and I a is the initial loss; the rainfall runoff in the catchment area is the total value of rainfall runoff in all pixels in the area, that is:

[0076] ,

[0077] Where Q pix is the runoff depth of a single pixel in catchment area i, Area pix is the actual area of ​​the pixel.

[0078] Step S1.3: According to the runoff of each catchment area and the drainage volume of the drainage network, the total amount of surface water in each catchment area is obtained; the surface water volume of the catchment area is equal to the rainfall runoff of the catchment area minus the drainage volume of the drainage network, that is,

[0079] ,

[0080] Where V remain_i is the total surface water volume in catchment area i, V catch_i is the total rainfall runoff of catchment area i, V drainage_i is the total discharge of the drainage system in catchment area i;

[0081] Total displacement V drainage_i The average flow rate Q i And the drainage time t is estimated, that is

[0082] ,

[0083] Average flow rate Q i The Manning formula is simplified through the average pipe diameter, Manning roughness coefficient, and average slope, namely:

[0084] ,

[0085] Where n is the Manning roughness coefficient (estimated based on the material of all pipelines in catchment area i), D avg is the average diameter of all pipelines in catchment area i, S avg is the average slope of all pipelines in catchment area i, is the flow estimation constant (calculated according to Manning's formula).

[0086] Step S1.4: Based on the total surface water volume of each catchment area and the digital elevation model (DEM), the water depth of all pixels is estimated through passive inundation calculation, and the surface water depth prediction matrix is ​​output. First, according to the DEM values ​​of all pixels, the sequence is obtained by sorting from small to large.

[0087] Maximum water accumulation height H in the catchment area, surface water volume V remain 、Pixel area pix The following relationship exists:

[0088] ,

[0089] The above formula can be transformed into:

[0090] ,

[0091] According to the above formula and The relationship can be deduced:

[0092] ,

[0093] So, according to the sequence , start from the minimum elevation and obtain m and D that satisfy the above formula m , directly use it as the highest elevation point of the submerged area, and the maximum submerged depth of the area is D m ,

[0094] Finally, the surface water depth of all pixels is calculated, that is, , output surface water depth prediction matrix M pre .

[0095] Reference Figure 1 and Figure 3 , step S2 specifically includes the following steps: step S2.1: using the prior knowledge and the waterlogging points predicted by the algorithm, a historical waterlogging point database is established, including the location information longitude and latitude; step S2.2: the historical waterlogging point location information ( , i=1,2...n) into the surface water depth prediction matrix M pre The target pixel coordinate position in , i=1,2...n), set the matrix area with the target pixel coordinate position as the center as the historical waterlogging high risk area matrix M his , and then set a 50m×50m matrix area as the historical waterlogging high risk area with the above pixel as the center (if the matrix grid is 10m resolution, set a 5×5 matrix area as the high risk area), where the matrix pixel value of the historical waterlogging high risk area is set to 1, and the matrix pixel value of the non-historical waterlogging high risk area is set to 0;

[0096] Reference Figure 3 The red pixel is the historical waterlogging point. The pixel value of the red pixel and its surrounding 5×5 matrix area (yellow pixel) is set to 1, and the remaining pixels are set to 0.

[0097] Reference Figure 1 and Figure 4 , Step S2.3: The surface water depth prediction matrix M output by step S1.4 pre , the historical high-risk area matrix M output by the dot multiplication step S2.2 his , and obtain the matrix M of predicted high-risk areas for waterlogging hrisk :

[0098]

[0099] That is M pre There is water accumulation, and M his Pixels with a value of 1 (historically at risk) retain the water depth value, while the remaining M hisPixel points with value 0 are assigned value 0.

[0100] Reference Figure 1 , step S3 specifically includes the following steps: step S3.1: set the maximum expected value (MaxNumCl), minimum expected value (MinNumCl), initial search radius (Eps(0)), radius adjustment gain (ε) of the target waterlogging point cluster number, and set the minimum element contained in the same cluster to 1; step S3.2: use the predicted high-risk area M of waterlogging in step S2.3 hrisk , initial search radius Eps(0), and the minimum number of included elements 1 are used as input to complete a DBSCAN algorithm clustering; Step S3.3: remove the category clusters with the number of elements of 1 as noise, and determine the current actual number of category clusters NumCl, that is:

[0101] ,

[0102] Among them, NumCl is the current actual number of category clusters, NumCl step NumCl is the number of initial category clusters after clustering is completed. 1pts is the number of category clusters with 1 element; step S3.4: adjust the search radius according to the relationship between the actual number of category clusters and the maximum and minimum expected values; if the actual number of category clusters NumCl is less than the expected minimum MinNumCl, the search radius is reduced according to the adjustment gain ε; if the actual number of category clusters NumCl is greater than the expected maximum MaxNumCl, the search radius is increased according to the adjustment gain ε, and steps S3.2-S3.4 are repeated, that is, the search radius is first calculated:

[0103]

[0104] Then M hrisk , Eps(step), the same cluster contains the minimum element 1 as input, and iteratively performs DBSCAN algorithm clustering; step S3.5: according to the number of cluster categories, when the number of categories is between the expected maximum and minimum values, jump out of the algorithm to obtain the final cluster category.

[0105] Reference Figure 1 Step S4 specifically includes the following steps: Step S4.1: According to the cluster category output in step S3.5, obtain the cluster center pixel position coordinate set ; Step S4.2: Combine the coordinate position with the predicted high-risk area matrix M output from step S2.3 hrisk , get the maximum water depth information ( ,i=1,2... NumCl final); Step S4.3: Convert the obtained cluster center pixel position coordinate set into a longitude and latitude coordinate position set, and use the Tiandi Map reverse geocoding service to obtain the geographic address information. , i=1,2...NumCl final ), first converted into longitude and latitude coordinates ( ,i=1,2... NumCl final ); Using the longitude and latitude coordinates as parameters, use the Tiandi Map Reverse Geocoding API service (http: / / api.tianditu.gov.cn / geocoder?) to obtain the geographic address information of the location according to the "formatted_address" field of the return value, thereby forming the geographic address information of the waterlogging point (Address i , i=1,2... NumCl final ); Step S4.4: The latitude and longitude coordinates, geographic addresses, and maximum water depth information of all waterlogging points are integrated into a one-dimensional array set, and the complete attribute information of the predicted waterlogging points is output, namely:

[0106] .

[0107] Reference Figures 2 to 4 The implementation principle of the embodiment of the present application is as follows: first, according to step S1, the surface water rapid numerical simulation result is calculated to obtain the surface water depth prediction matrix M pre (Reference Figure 2 ), and then through steps S2.1 and S2.2, the historical waterlogging high risk area matrix M is obtained. his (Reference Figure 3 ); M pre With M his Perform point multiplication calculation to obtain the predicted high-risk area M for waterlogging hrisk (Reference Figure 4 ), obtain the cluster point result according to step S3, obtain the pixel set {[3, 3], [10, 12], [38, 31]} through step S4.1, obtain the maximum water depth information (Deeps=[2,3,5]) through step S4.2, and then obtain the longitude and latitude coordinate set {[119.967515, 30.231346], [119.983432,30.213423], [120.213432, 30.187653]} through S4.3, call the Tiandi Map reverse geocoding interface service, and take the formatted_address attribute returned by the interface as the geographic address information; obtain the final waterlogging point prediction result containing longitude and latitude coordinates, geographic address, and maximum water depth information

[0108] Waterlogging_Point=

[0109] .

[0110] This application uses the Tiandi Map geocoding interface service to obtain location address information based on the surface waterlogging depth prediction matrix, thereby forming point-based waterlogging address information; in view of the fact that the existing numerical simulation methods for waterlogging are complex and have large data requirements, a passive inundation algorithm is used to quickly estimate the waterlogging depth of all pixels, and a surface waterlogging depth prediction matrix is ​​constructed to achieve rapid simulation of surface waterlogging conditions; in addition, this application uses the historical waterlogging point database to determine high-risk areas for waterlogging, and combines it with the prediction results to form a highly accurate prediction of high-risk areas for waterlogging. Secondly, an improved DBSCAN clustering algorithm is used to solve the problem that it is difficult to cluster the simulation results of the waterlogging surface, and the clustering effect is optimized by dynamically adjusting the search radius to improve the accuracy of waterlogging point extraction.

[0111] The above are all preferred embodiments of the present application, and the protection scope of the present application is not limited thereto. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be included in the protection scope of the present application.

Claims

1. A road waterlogging point prediction method integrating historical waterlogging points and spatial clustering, characterized by: The following steps are involved: Step S1: Rapidly estimate the rainfall runoff and drainage volume, and perform a rapid numerical simulation of surface waterlogging to obtain an estimated value of the depth of surface waterlogging; Step S2: Based on the historical waterlogging location and the estimated value of surface waterlogging depth, determine the high-risk area for waterlogging; Step S3: using the improved DBSCAN clustering algorithm to predict waterlogging points with high-risk areas of waterlogging as input; Step S4: Based on the geocoding, output the predicted waterlogging point with latitude and longitude coordinates, geographic address, and maximum waterlogging depth information; The rapid numerical simulation of landmark waterlogging in step S1 specifically includes the following steps: Step S1.1: Based on the watershed division of the generalized drainage pipe points, the regional sub-watersheds are formed by establishing Thiessen polygons; Step S1.2: Calculate the rainfall runoff depth of each pixel through the rainfall runoff curve number model SCS-CN, convert it into runoff, accumulate the runoff of each catchment pixel, and obtain the runoff of each catchment; Step S1.3: According to the runoff of each catchment area and the drainage volume of the drainage network, the total amount of surface water in each catchment area is obtained; Step S1.4: Based on the total amount of surface water in each catchment area and the digital elevation model DEM, the water depth of all pixels is estimated through passive inundation calculation, and the surface water depth prediction matrix is ​​output; The step S2 specifically includes the following steps: Step S2.1: Establish a historical waterlogging point database based on the waterlogging points obtained through prior knowledge and historical predictions, including location information longitude and latitude; Step S2.2: convert the longitude and latitude coordinates of the historical waterlogging points into the target pixel coordinates in the surface waterlogging depth prediction matrix, and set a matrix area with the target pixel coordinates as the center as the historical waterlogging high-risk area matrix, where the pixel values ​​of the historical waterlogging high-risk area matrix are set to 1, and the pixel values ​​of the non-historical waterlogging high-risk area matrix are set to 0; Step S2.3: By multiplying the surface waterlogging depth prediction matrix output by step S1.4 with the historical waterlogging high-risk area matrix output by step S2.2, the predicted waterlogging high-risk area matrix is ​​obtained: The step S3 specifically comprises the following steps: Step S3.1: Set the maximum expected value MaxNumCl, the minimum expected value MinNumCl, the initial search radius Eps(0), the radius adjustment gain ε, and the minimum element contained in the same cluster is set to 1; Step S3.2: Taking the predicted high-risk area of ​​waterlogging in step S2.3, the initial search radius Eps(0), and the minimum number of included elements 1 as input, complete a DBSCAN algorithm clustering; Step S3.3: Remove the category clusters with 1 element as noise, and determine the current actual number of category clusters NumCl, that is: ， Among them, NumCl is the current actual number of category clusters, NumCl step NumCl is the number of initial category clusters after clustering is completed. 1pts is the number of category clusters with 1 element; Step S3.4: Adjust the search radius according to the relationship between the actual number of category clusters and the maximum and minimum expected values; if the actual number of category clusters NumCl is less than the expected minimum MinNumCl, the search radius is reduced according to the adjustment gain ε; if the actual number of category clusters NumCl is greater than the expected maximum MaxNumCl, the search radius is increased according to the adjustment gain ε, and steps S3.2-S3.4 are repeated, that is, the search radius is first calculated: ； Then, the predicted high-risk area for waterlogging, Eps (step), and the minimum element 1 in the same cluster are used as input, and the DBSCAN algorithm clustering is performed iteratively; Step S3.5: According to the number of cluster categories, when the number of categories is between the expected maximum and minimum values, the algorithm is exited to obtain the final cluster category.

2. The method for predicting road waterlogging points by integrating historical waterlogging points and spatial clustering according to claim 1, characterized in that: The calculation formula in step S1.2 specifically includes: According to SCS-CN, the rainfall-runoff relationship is obtained: ， Through empirical formula Calculate the S value, and then get the runoff of each pixel in the region, where the CN value is an empirical value based on the land use type, P is the rainfall depth, Q is the runoff depth, and I a is the initial loss; The rainfall runoff of the catchment area is the total rainfall runoff of all pixels in the area, that is: ， Where Q pix is the runoff depth of a single pixel in catchment area i, Area pix is the actual area of ​​the pixel.

3. The method for predicting road waterlogging points by integrating historical waterlogging points and spatial clustering according to claim 2, characterized in that: The calculation formula in step S1.3 specifically includes: The surface water volume of the catchment area is equal to the rainfall runoff in the catchment area minus the drainage volume of the drainage network, that is, , Where V remain_i is the total surface water volume in catchment area i, V catch_i is the total rainfall runoff of catchment area i, V drainage_i is the total discharge of the drainage system in catchment area i; Total displacement V drainage_i The average flow rate Q i And the drainage time t is estimated, that is , Average flow rate Q i The Manning formula is simplified through the average pipe diameter, Manning roughness coefficient, and average slope, namely: ， Where n is the Manning roughness coefficient, D avg is the average diameter of all pipelines in catchment area i, S avg is the average slope of all pipelines in catchment area i, is the flow rate estimation constant.

4. The method for predicting road waterlogging points by integrating historical waterlogging points and spatial clustering according to claim 3 is characterized by: The calculation formula in step S1.4 specifically includes: First, sort all pixel DEM values ​​from small to large to get a sequence , Maximum water accumulation height H in the catchment area, surface water volume V remain 、Pixel area pix The following relationship exists: ， So, according to the sequence , start from the minimum elevation and obtain m and D that satisfy the above formula m , directly use it as the highest elevation point of the submerged area, and the maximum submerged depth of the area is D m , Finally, the surface water depth of all pixels is calculated, that is, , output the surface water depth prediction matrix.

5. The method for predicting road waterlogging points by integrating historical waterlogging points and spatial clustering according to claim 1, characterized in that: The step S4 specifically comprises the following steps: Step S4.1: According to the cluster category outputted in step S3.5, obtain the cluster center pixel position coordinate set; Step S4.2: Combining the coordinate position with the predicted high-risk area matrix of waterlogging outputted in step S2.3, the maximum waterlogging depth information is obtained; Step S4.3: According to the obtained cluster center pixel position coordinate set, it is converted into a longitude and latitude coordinate position set to obtain geographic address information. The pixel row and column number coordinate set is first converted into longitude and latitude coordinates; the longitude and latitude coordinates are used as parameters to obtain the geographic address information of the position, thereby forming the geographic address information of the waterlogging point; Step S4.4: The longitude and latitude coordinates, geographic addresses, and maximum water depth information of all waterlogging points are integrated into a one-dimensional array set, and the complete attribute information of the predicted waterlogging points is output.

Images