Water conservancy special-purpose terrain extraction method based on laser radar and satellite remote sensing

Abstract

The invention discloses a special terrain extraction method for water conservancy based on a laser radar and satellite remote sensing. The method comprises the following steps: acquiring satellite radar data and satellite remote sensing data of a preset area; performing comparison and discrimination area division according to the radar data and the remote sensing data to obtain a live area and a fuzzy area, and performing water conservancy boundary and terrain extraction on the live area to obtain first data; carrying out collaborative complementary measurement on the fuzzy region to obtain covered water conservancy data, and carrying out water conservancy boundary and terrain extraction on the covered water conservancy data by adopting a closed contour line to obtain second data; according to the first data and the second data, constructing a time sequence water conservancy topology dedicated terrain extraction model, and extracting water conservancy terrain data DEM1 of a preset area; and constructing a two-dimensional hydrodynamic model on the basis of the topographic data and the water conservancy boundary of the preset area to perform flood inundation simulation, extracting a wading area according to an inundation range, and generating water-land integrated HDEM data of the preset area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of terrain extraction, and in particular to a water conservancy special terrain extraction method based on laser radar and satellite remote sensing. BACKGROUND

[0002] Terrain data is an important data support for water conservancy engineering design, flood simulation calculation of basin flood control, water resources management and allocation, soil erosion analysis of soil conservation, and is the key to the construction of digital scenes of basins, major rivers and main tributaries.

[0003] The precise design of water conservancy projects and the dynamic monitoring of basins highly depend on the spatial and temporal accuracy of terrain data, especially the extraction of key water conservancy elements such as river boundaries, inundation simulation, and riverbed evolution. Traditional terrain mapping methods have the defects of low efficiency, limited coverage, and inability to capture dynamic changes, which cannot meet the real-time, global, and high-precision needs of modern water conservancy informatization. The development of laser radar and satellite remote sensing technology provides a new way for terrain data collection: laser radar can accurately invert the three-dimensional shape of the ground through point cloud data, and satellite remote sensing can achieve large-scale synchronous observation through radar, multispectral images, etc. The combination of the two provides the possibility for rapid and refined extraction of water conservancy terrain.

[0004] However, the existing technology faces the following challenges in fusing multi-source data: the imaging mechanisms of laser radar data and satellite remote sensing data are significantly different, the former is sensitive to the geometric features of the ground, and the latter is sensitive to the spectral features, direct fusion can easily lead to feature misplacement; in water conservancy scenarios, terrain data has strong timeliness, and existing single data source models cannot balance spatial resolution and temporal continuity; water conservancy topology models need to reflect the evolution law of terrain over time, but traditional models are mostly based on static data, lack of dynamic fusion capability for "radar-remote sensing" time series data chain, and are difficult to simulate dynamic processes such as water flow direction and inundation range expansion.

[0005] In addition, high-precision DEM of buildings in water areas is produced by technologies such as unmanned aerial photography and laser radar. The current method of producing high-precision DEM of buildings in water areas has the following problems: conventional DEM does not consider the water passing and blocking functions of buildings in water areas, leading to distortion of terrain water connectivity, and manual classification is generally used for building point cloud classification, which is low in efficiency; conventional DEM generally does not include underwater terrain information, leading to unreasonable expression of water quantity and water flow in hydrological simulation, and inability to support accurate calculation of basin runoff and accurate simulation of basin flood.

[0006] Therefore, there is an urgent need for a water conservancy terrain extraction method based on multi-source data collaboration to fill the gap in the production method of watershed underlying surface terrain considering water connectivity, solve the problem that existing conventional terrain cannot truly express water connectivity and underwater terrain is missing, and improve the terrain extraction accuracy in complex terrain and dynamic scenarios, thereby improving the accuracy of runoff production, flood simulation, etc. on the watershed surface. SUMMARY

[0007] The purpose of the present application is to provide a water conservancy special terrain extraction method based on laser radar and satellite remote sensing.

[0008] To achieve the above purpose, the present application is implemented according to the following technical solutions: The present application comprises the following steps: Satellite radar data and satellite remote sensing data of a preset area are collected, and the radar data and the remote sensing data are preprocessed; According to the radar data and the remote sensing data, the region is divided by comparison and discrimination, the consistent region is taken as the live region, the inconsistent region is taken as the fuzzy region, the water conservancy boundary and the terrain of the live region are extracted to obtain first data; The fuzzy region is cooperatively measured to obtain coverage water conservancy data, and the coverage water conservancy data is extracted by water conservancy boundary and terrain extraction using closed contour lines to obtain second data; the second data includes second region terrain and second water conservancy boundary; According to the first data and the second data, a time-series water conservancy topology special terrain extraction model is constructed to extract water conservancy terrain data DEM1 of the preset area; On the basis of the terrain data and the water conservancy boundary of the preset area, a two-dimensional water dynamic model is constructed to simulate flood inundation, and a water area is extracted according to the inundation range; For water-involved building details, laser radar is used to obtain point cloud data of the water-involved building region, which is divided into water-blocking buildings and water-passing buildings according to functional characteristics, and combined with graphical processing and supplementary removal operation, and fused with underwater DEM, and embedded in the preset area water conservancy terrain data DEM1 by using weighted correction and Gaussian smoothing algorithm, to generate preset area water-land integrated HDEM data.

[0009] Further, the method for dividing the region by comparison and discrimination according to the radar data and the remote sensing data comprises: Edge features are extracted from radar data and satellite remote sensing data respectively, sensor registration is performed through edge feature fusion, time-series weight is introduced for multi-time phase matching of the spatially registered radar data and satellite remote sensing data, and the expression is: Wherein, the time-series weight of time point t is , the comprehensive similarity score is adjusted according to seasonal changes , the gradient component of the template image in the x direction at the vth point is , the gradient component of the search image in the x direction at the vth point is , the gradient component of the template image in the y direction at the vth point is , the gradient component of the search image in the y direction at the vth point is , the total number of features is , the number of time series templates is ; Grouping radar data and satellite remote sensing data, grouping radar data and satellite remote sensing data with a matching degree greater than 0.673, based on the registered multi-source data, calculating the spatio-temporal consistency of different groups: Wherein the spatio-temporal consistency is , the satellite remote sensing derived point cloud at time t is , the radar point cloud at time t is , the upper limit of the observation time is , the point cloud structure similarity is Project the point cloud onto a 2.5D grid, calculate the elevation mean and intensity variance within the grid, and obtain the SSIM comparison item according to the elevation mean and intensity variance; Sum the time series standard deviation of the spatio-temporal consistency to obtain the time series variance, when the pixel-level consistency is greater than or equal to 0.786 and the time series variance is less than 0.352, the height is stable and the intensity-vegetation index is strongly correlated, the location of the group is the live area; when the pixel-level consistency is less than 0.786 and the time series variance is greater than 0.352, the height is suddenly changed and the intensity-vegetation index is weakly correlated, the location of the group is the fuzzy area.

[0010] Further, the method for extracting water conservancy boundaries and terrain in the live area comprises: Extracting terrain features in the live area based on SAR technology to obtain the first regional terrain; locating the potential water area of the live area through the slope threshold, and using the maximum inter-class variance method to binarize the NDWI index in the potential area to extract the initial water boundary; for the cloud-covered area of the optical image, filling the boundary missing using the water backscattering coefficient of the SAR image, merging to generate the complete water contour of the live area as the first water conservancy boundary, and outputting the first water conservancy boundary and the first regional terrain as the first data.

[0011] Further, the method for obtaining the second regional terrain and the water conservancy data covering the fuzzy area through collaborative supplementary measurement comprises: An "air-space-ground" integrated observation network is constructed to realize data space-time complementation, obtain medium-density laser radar point cloud data, radar backscattering coefficient and optical vegetation index, identify the type of a fuzzy area through a random forest model, and obtain a high-priority area, a medium-priority area and a low-priority area; For a high-priority area, unmanned aerial vehicle data is introduced, the height of the unmanned aerial vehicle is equal to the maximum height of vegetation plus a safety margin, the overlap rate is proportional to the vegetation complexity, and a mesh flight path is generated based on a laser radar scanning angle; The unmanned aerial vehicle scans the same area from different directions, and the data is fused through a point cloud registration algorithm, and the expression is as follows: wherein a rotation matrix is A, a translation vector is w, a u-th point in a source point cloud is , the source point cloud is point cloud data to be registered to a target point cloud obtained from different scanning angles, a u-th point in the target point cloud is , a directional weight coefficient is , a normal vector is , and fused data is G; Multi-temporal radar images and laser data of the same area are spatio-temporally registered, a real terrain is estimated through Kalman filtering as second area terrain data of a fuzzy area; for a multi-echo laser radar point cloud area, first echoes are separated, echo intensity differences are calculated, and in a high-value area of the echo intensity differences, a potential river center line is identified through local minimum point searching; for a multi-source feature fusion area, linear shadow features are extracted from optical images, abnormal strips of backscattering coefficients are analyzed from radar images, and combined with water conservancy planning data, position information of water-related buildings is located through spatial superposition analysis; and the second area terrain data, the potential river center line and the position information of the water-related buildings are output as water conservancy data.

[0012] Further, a method for extracting the second data comprises: According to the water conservancy data, a river center line neighborhood elevation gradient and a water flow direction are obtained, and a robust fuzzy local information C-means clustering algorithm is used to preliminarily separate water bodies and non-water bodies; A target function of fuzzy clustering is given, and the expression is as follows: wherein the target function is , a prior probability that a pixel a belongs to a cluster c is , a topographic dissimilarity measure of the pixel a and the cluster c is , a fuzzy local variable of the pixel a and the cluster c is , a ridge regression coefficient is , a total number of pixels is , and a membership degree of the pixel a to the cluster c is , the spectral dissimilarity measure between pixel a and cluster c is , the equilibrium parameter is , balancing spectral and terrain characteristics, the elevation variation coefficient of pixel a is , the elevation variation coefficient of cluster c is ; Compute the mean, covariance, and dissimilarity measure: The mean of cluster c is , the covariance is , the transpose of the matrix is , the spectral measure of pixel a is , the data dimension is , the mean of pixel a in cluster c is ; Compute local variation density and fuzzy local variables: The local variation density of pixel a is , the neighboring pixels of pixel a are , the neighborhood pixel set is The average test value of the neighborhood pixel spectrum centered on pixel a is , the test variance of the neighborhood pixel spectrum centered on pixel a is , neighborhood pixels The local variation density is ; Compute prior probabilities and memberships: The index clustering is , the field effect intensity is , the attribute of pixel a is , the number of clusters is , the prior probability that pixel a belongs to cluster i is , the spectral dissimilarity measure between pixel a and cluster i is , the fuzzy local variables of pixel a and cluster i are ; If the maximum error of the objective function from time point t to time point t+1 is less than the error threshold, the membership matrix is ​​defuzzified to obtain a binary segmentation result. Morphological closing operations are used to fill small holes in the binary segmentation result and extract the contour of the largest connected domain. The water collection constraint is applied to the maximum connected domain contour, and the water cluster center is forced to satisfy , the isolated cluster with an area less than the area threshold is removed, and the segmentation data is obtained; wherein the neighborhood elevation gradient is , and the flow direction is ; The gradient and the gradient amplitude of the fuzzy water area are obtained according to the segmentation data, anisotropic diffusion filtering is applied to the segmentation data, and the river channel boundary sharpness is retained, and the expression is: The edge stopping function is , the diffusion strength is controlled, the edge sensitive coefficient in the anisotropic diffusion filtering is , the boundary gradient is adaptively adjusted, the gradient amplitude is s, and the gradient of the image intensity is ; The segmentation data boundary is taken as the second water boundary data, and the second data is output in combination with the second regional terrain data.

[0013] Further, a method for constructing a time-series water topological special terrain extraction model according to the first data and the second data, comprising: The terrain data based on the first data and the second data constructs a terrain base grid, extracts static parameters, performs time-series difference analysis on the water boundaries of the first data and the second data to identify the boundary change area, performs vector-raster conversion on the dynamic boundary to generate a time-series raster layer, and extracts a dynamic topological node set; the static parameters include terrain elevation, slope, and slope direction; The dynamic nodes are matched with the static nodes, a spatio-temporal joint node set is generated through spatial superposition analysis, a node attribute table is established, and the spatial coordinates, time attributes, and water properties of the nodes are recorded; Based on the spatio-temporal joint node set, a dynamic topological edge set is constructed through a point set triangulation algorithm, wherein the adjacency relationship represents the direct connectivity state between nodes, and the time-series correlation represents the state change of the same node at different time steps; the edge attributes include distance, flow direction, and water carrying capacity; The distance, water level difference, and time step are used as the time-series weight assignment for the dynamic topological edge, a time-series water topological special terrain extraction model is constructed, and the preset regional water terrain data DEM1 is output.

[0014] Further, a method for obtaining the water area, comprising: The first water boundary and the second water boundary are merged into a complete water boundary to form river network data; The preset area water conservancy terrain data DEM1 is meshed, the water conservancy boundary adjacent area and the covered water conservancy data area are meshed according to a grid size of 5m*5m, and other areas far away from the river are meshed with a grid size of 20m*20m; the meshed grids are assigned with water dynamic parameters such as roughness, area coefficient and time step, and the mesh model result is output; According to historical flow data of the preset area, a flood frequency curve is fitted, a design flood peak flow corresponding to a target frequency is determined as a inflow boundary condition; Based on the control volume method, a two-dimensional Saint-Venant shallow water equation is used to perform flood inundation preplay with the preset area river design maximum flood peak flow as a boundary condition, and the inundation range is output as a wading area.

[0015] The beneficial effects of the present application are: The present application is a water conservancy special terrain extraction method based on laser radar and satellite remote sensing, which has the following technical effects compared with the prior art: The water conservancy special terrain HDEM construction method proposed in the present application supplements and removes the terrain in detail on the basis of the conventional DEM considering the water passing and water blocking characteristics of water-related structures such as dikes, reservoir dams, sluices, dams, rubber dams, water retaining roads, bridges, culverts, aqueducts and inverted siphons, and simultaneously fuses underwater terrain data, which can express the true water flow connectivity and improve the accuracy of hydrological simulation analysis, meet the application requirements of digital twin watershed such as three-dimensional scene construction of watershed underlying surface, hydrological modeling and calculation, accurate calculation of runoff and convergence, precise simulation of watershed flood, assessment of flood inundation range and loss degree, analysis of water balance relationship in the watershed, and lay a solid foundation for realizing fine scheduling of the four predictions of the watershed flood control, i.e., one flow, one cubic reservoir capacity and one centimeter water level, and greatly improve the level of water conservancy digitalization, precision and practicality.

[0016] The present application can improve the accuracy of the water conservancy special terrain HDEM by the steps of preprocessing, region division by comparison and discrimination, first region terrain data and first water conservancy boundary extraction, multi-angle joint inversion, collaborative supplementary survey, second region terrain data and second water conservancy boundary extraction, model construction, obtaining the water-related area and obtaining the HDEM result, optimize the extraction of the water conservancy special terrain HDEM, greatly save resources and improve work efficiency, realize intelligent extraction of the water conservancy special terrain HDEM, real-time multi-angle joint inversion and second region terrain water conservancy boundary extraction of the water conservancy special terrain HDEM extraction, and have important significance for the extraction of the water conservancy special terrain HDEM, can meet the extraction requirements of the water conservancy special terrain HDEM of different standards, have certain universality, and have important significance for the popularization and application of the water conservancy special terrain HDEM. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a flow chart of steps of a water conservancy special terrain extraction method based on laser radar and satellite remote sensing of the present application; Figure 2 is a bridge area HDEM result schematic diagram in an embodiment of the present application; Figure 3 is a reservoir dam area HDEM result schematic diagram in an embodiment of the present application; Figure 4 is a water gate area HDEM result schematic diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The present application will be further described below through specific embodiments, and the schematic embodiments of the present application and the description are used to explain the present application, but do not limit the present application.

[0019] The water conservancy special terrain extraction method based on laser radar and satellite remote sensing of the present application comprises the following steps: As shown in Figures 1-2 in the present embodiment, the following steps are included: Satellite radar data and satellite remote sensing data of a preset area are collected, and the radar data and the remote sensing data are preprocessed; In actual evaluation, XX water conservancy tributaries are taken as research objects, satellite radar data: Sentinel-1 satellite-borne SAR, L-band, incident angle 35°, resolution 10 m, flood period and dry period images are acquired, airborne laser radar points, cloud density 12 points / m², 0.5 m resolution three-dimensional structure containing dikes and river channels are generated; FMCW ground-based radar is used, with a resolution of 0.1 m, to locally scan typical areas; Satellite remote sensing data, extract NDWI index, threshold 0.4, identify water body; According to the radar data and the remote sensing data, a comparison and discrimination region division is performed, consistent regions are taken as live regions, and inconsistent regions are taken as fuzzy regions, water conservancy boundaries and terrain extraction are performed on the live regions to obtain first data; In actual evaluation, 32% of the regions are fuzzy regions, and 68% of the regions are live regions; The fuzzy regions are cooperatively resurveyed to obtain covered water conservancy data, and the closed contour lines are used to extract water conservancy boundaries and terrain from the covered water conservancy data to obtain second data; the second data includes second region terrain and second water conservancy boundaries; According to the first data and the second data, a time-series water conservancy topology special terrain extraction model is constructed, and water conservancy terrain data DEM1 of a preset area is extracted; Constructing a two-dimensional hydrodynamic model based on the topographic data and water conservancy boundaries of the preset area to simulate flooding, and extracting the wading area according to the flooding range; For the details of water-related buildings, lidar is used to obtain point cloud data of water-related building areas, which are divided into water-retaining buildings and water-passing buildings according to their functional characteristics. Graphic processing and combined operations of addition and removal are performed, and the data are fused with the underwater DEM. The weighted correction and Gaussian smoothing algorithm are used to embed the data into the water conservancy terrain data DEM1 of the preset area to generate the integrated water and land HDEM data of the preset area.

[0020] In this embodiment, the method for comparing and determining region division based on the radar data and the remote sensing data includes: Edge features are extracted from the optical images of radar data and satellite remote sensing data respectively, and sensor registration is performed by edge feature fusion. Temporal weights are introduced to perform multi-temporal matching on the radar data and satellite remote sensing data after spatial registration. The expression is: The temporal weight of time point t is , adjusted for seasonal variations, the overall similarity score is , the gradient component of the template image in the x direction at point v is , the gradient component of the search image in the x direction at point v is , the gradient component of the template image in the y direction at point v is , the gradient component of the search image in the y direction at point v is The total number of features is , the number of timing templates is ; Radar data and satellite remote sensing data are grouped, and radar data and satellite remote sensing data with a matching degree greater than 0.673 are grouped together. Based on the registered multi-source data, the spatiotemporal consistency of different groups is calculated: The spatiotemporal consistency is , the satellite remote sensing derived point cloud at time t is , the radar point cloud at time t is The upper limit of observation time is , the point cloud structure similarity is , project the point cloud to a 2.5D grid, calculate the elevation mean and intensity variance within the grid, and obtain the SSIM comparison item based on the elevation mean and intensity variance; The time series variance is obtained by summing the time series standard deviation of the spatio-temporal consistency, when the pixel-level consistency is greater than or equal to 0.786 and the time series variance is less than 0.352, the height is stable, the intensity-vegetation index is strongly correlated, and the location of the grouping is the live area; when the pixel-level consistency is less than 0.786 and the time series variance is greater than 0.352, the height is mutated, the intensity-vegetation index is weakly correlated, and the location of the grouping is the fuzzy area.

[0021] In the embodiment, the method for extracting the water conservancy boundary and the terrain of the live area comprises: The terrain feature of the live area is extracted based on the SAR technology to obtain the first area terrain; the potential water area of the live area is located through the slope threshold value, the NDWI index is binarized in the potential area by using the maximum inter-class variance method to extract the initial water boundary, the cloud coverage area of the optical image is filled with the water backscattering coefficient of the SAR image to generate the complete water contour of the live area, which is taken as the first water conservancy boundary; the first water conservancy boundary and the first area terrain are output as the first data.

[0022] In the embodiment, the method for obtaining the second area terrain and the water conservancy data covering the fuzzy area through cooperative supplementary surveying comprises: An "air-space-ground" integrated observation network is constructed to realize the time-space complementation of data, obtain the medium-density laser radar point cloud data, radar backscattering coefficient and optical vegetation index, identify the fuzzy area type through the random forest model, and obtain the high-priority area, the medium-priority area and the low-priority area; For the high-priority area, the unmanned aerial vehicle data is introduced, the height of the unmanned aerial vehicle is equal to the maximum height of the vegetation plus a safety margin, the overlap rate is proportional to the vegetation complexity, and the mesh flight route is generated based on the laser radar scanning angle; The unmanned aerial vehicle scans the same area from different directions, and the data is fused through the point cloud registration algorithm, and the expression is: wherein the rotation matrix is A, the translation vector is w, the u-th point in the source point cloud is , the source point cloud is the point cloud data to be registered to the target point cloud obtained from different scanning angles, the u-th point in the target point cloud is , the direction weight coefficient is , the normal vector is , and the fused data is G. The multi-temporal radar image and laser data of the same region are spatio-temporally registered, and the real terrain is estimated by Kalman filtering as the second regional terrain data of the fuzzy region; for the multi-echo laser radar point cloud region, the first echo is separated, the echo intensity difference is calculated, and in the high value area of the echo intensity difference, the potential river center line is identified by searching the local minimum point; for the multi-source feature fusion region, linear shadow features are extracted from the optical image, abnormal strips of backscattering coefficients are analyzed from the radar image, combined with water conservancy planning data, the location information of the water-related building is positioned by spatial overlay analysis; and the second regional terrain data, the potential river center line and the water-related building location information are output as the overlay water conservancy data.

[0023] In the embodiment, the method for extracting the second data comprises: According to the overlay water conservancy data, the elevation gradient and the flow direction of the river center line neighborhood are obtained, and a robust fuzzy local information C-means clustering algorithm is used to preliminarily separate the water body and the non-water body; The objective function of fuzzy clustering is given, and the expression is: The objective function is The prior probability of pixel a belonging to cluster c is The terrain dissimilarity measure of pixel a and cluster c is The fuzzy local variable of pixel a and cluster c is The ridge regression coefficient is The total number of pixels is The membership degree of pixel a belonging to cluster c is The spectral dissimilarity measure of pixel a and cluster c is The balance parameter is The balance of spectral and terrain features, the elevation variation coefficient of pixel a is The elevation variation coefficient of cluster c is ; The mean, covariance and dissimilarity measure are calculated: The mean of cluster c is The covariance is The transpose of the matrix is The spectral measure of pixel a is The data dimension is The mean of pixel a of cluster c is ; The local variation density and the fuzzy local variable are calculated: wherein the local variation density of pixel a is , the neighborhood pixels of pixel a are , the neighborhood pixel set is , the neighborhood pixel spectral test mean centered at pixel a is , the neighborhood pixel spectral test variance centered at pixel a is , the neighborhood pixel , the local variation density of pixel a is ; The prior probability and the membership are calculated: wherein the index cluster is , the field action intensity is , the pixel a attribute is , the cluster number is , the prior probability of pixel a belonging to cluster i is , the spectral dissimilarity measure of pixel a and cluster i is , the fuzzy local variable of pixel a and cluster i is ; If the maximum error of the objective function from time point t to time point t+1 is less than the error threshold, the binary segmentation result is obtained by defuzzifying the membership matrix, small holes in the binary segmentation result are filled by using morphological closing operation, and the maximum connected domain contour is extracted; The water collection constraint is applied to the maximum connected domain contour, and the water cluster center is forced to satisfy , the isolated cluster with an area less than the area threshold is removed, and the segmentation data is obtained; wherein the neighborhood elevation gradient is , and the flow direction is ; According to the gradient and the gradient amplitude of the fuzzy water area obtained from the segmentation data, an anisotropic diffusion filter is applied to the segmentation data to preserve the sharpness of the river channel boundary, and the expression is: wherein the edge stopping function is , the control diffusion intensity, the edge sensitive coefficient in the anisotropic diffusion filter is , which is adaptively adjusted according to the boundary gradient, the gradient amplitude is s, and the gradient of the image intensity is ; The boundary of the segmentation data is taken as the second water boundary data, and the second regional terrain data is combined to output the second data.

[0024] In the embodiment, the method for constructing a time-series water conservancy topology special terrain extraction model according to the first data and the second data comprises the following steps: A terrain base grid is constructed based on the first data and the second data, static parameters are extracted, time-series difference analysis is performed on the water conservancy boundaries of the first data and the second data to identify boundary change areas, vector-raster conversion is performed on the dynamic boundaries to generate a time-series raster layer, and a dynamic topology node set is extracted; the static parameters include terrain elevation, slope, and slope direction; The dynamic nodes are matched with the static nodes, a space-time joint node set is generated through space superposition analysis, a node attribute table is established to record the spatial coordinates, time attributes, and water conservancy attributes of the nodes; Based on the space-time joint node set, a dynamic topology edge set is constructed through a point set triangulation algorithm, wherein: the adjacency relationship represents the direct connectivity state between nodes, and the time-series correlation represents the state change of the same node at different time steps; the edge attributes include distance, flow direction, and water carrying capacity; The distance, water level difference, and time step are used as time-series weight assignment for the dynamic topology edges to construct a time-series water conservancy topology special terrain extraction model, and the water conservancy terrain data DEM1 of a preset area is output.

[0025] In the embodiment, the method for obtaining the water-involved area comprises the following steps: The first water conservancy boundary and the second water conservancy boundary are merged into a complete water conservancy boundary to form river network data; The water conservancy terrain data DEM1 of the preset area is meshed, the water conservancy boundary adjacent area and the water conservancy data covered area are meshed according to a 5m*5m grid size, and other areas far away from the river are meshed according to a 20m*20m grid size; the meshed grids are assigned with water dynamic parameters such as roughness, area coefficient, and time step, and a mesh model result is output; According to historical flow data of the preset area, a flood frequency curve is fitted, a design flood peak flow corresponding to a target frequency is determined as an inflow boundary condition; Based on the control volume method, a two-dimensional Saint-Venant shallow water equation is used to perform flood inundation simulation using the design maximum flood peak flow of the preset area as a boundary condition, and the inundation range is output as the water-involved area.

[0026] In the embodiment, the method for obtaining the HDEM result comprises the following steps: Within the scope of the water-related area, the high-density lidar point cloud data of water-related buildings is obtained in combination with the water conservancy planning data and the covered water conservancy data. The regional growth building classification filtering algorithm based on the plane fitting technology is used to extract the dikes, dams, culverts, bridges, sluices, flood control walls, flood avoidance platforms, aqueducts, and inverted siphon water-related buildings: within the scope of the water-related area, the regional growing algorithm is used to segment the building footpoints of the point cloud data. The regional growing algorithm of the plane fitting technology is used to realize the segmentation of the ground feature footpoints and extract the building footpoint blocks: ① Create a regional set of adjacent building footpoints , randomly select a non-ground foot point as the initial seed point to start the creation of the region, and recursively connect the non-ground foot points in the eight neighborhoods of the seed point to form a region, and repeat the creation operation until each non-ground foot point is divided into a region ② Each area Divide and i two parts, of which Indicates the foot points inside the block, A series of edge points representing the block. If at least one laser foot point in the eight neighborhoods of a point is a ground point, the point is defined as an edge point; otherwise, the point is defined as an internal point. If the area is smaller than the preset minimum area threshold or the number of laser foot points inside is zero, then Remove ④Region Each small area Do the following loop: Given an interior point , get The best-fit plane of eight adjacent points, plane equation ,parameter By solving the deviation Minimum sum of squares Obtain; ⑤Define As the small block obtained through the above process The collection of , ,in It's a small piece The If the represented small block is smaller than the preset minimum area threshold, Is not completely contained by other blocks, then from Remove; ⑥ Merge The connected blocks in the block and get a new set ; ⑦ Get the point cloud block of the top surface of the water-related building, if Smaller than the preset minimum building area threshold, remove ; final The blocks remaining in the middle are all used as the top surface point cloud blocks of the water-related building; ⑧ Combined with artificial correction processing, ensure that the terrain form of the water-related building area is complete and fine expression; The TIN is constructed by using the top surface point cloud of the water-related building and the supplementary collected terrain feature lines, the regular grid is interpolated to produce the water-related building DEM2, according to the functional characteristics of the water-related building, the embankment, dam, culvert, bridge, sluice, flood wall, water-avoiding platform, aqueduct and inverted siphon are stored as DEM3 file as the terrain to be supplemented of DEM1; The culvert, bridge and other water-passing buildings are stored as DEM4 file as the terrain to be removed of DEM1; The DEM1 and the water-related building area DEM3 and DEM4 are processed by imaging, the water-stopping building DEM3 is added on the basis of DEM1, and the water-passing building DEM4 is removed, and the land part of the water conservancy special land is obtained by fusion, and the expression is: The obtained land water conservancy land result DEM5 is fused with the underwater DEM, and the expression is: The adaptive weighted least square method is used for smoothing processing, the unreasonable terrain and incomplete building are checked, edited and modified, and the HDEM is generated, Figure 2 is a schematic diagram of the HDEM result of the bridge area, Figure 3 is a schematic diagram of the HDEM result of the reservoir dam area, Figure 4 is a schematic diagram of the HDEM result of the sluice area.

[0027] The above only describes the preferred embodiments of the present application, and does not limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for extracting water conservancy special terrain based on laser radar and satellite remote sensing, characterized in that, The following steps are involved: Collecting satellite radar data and satellite remote sensing data of a preset area, and preprocessing the radar data and the remote sensing data; Comparing the radar data and the remote sensing data to determine the region division, taking the consistent region as the actual region and the inconsistent region as the fuzzy region, and extracting the water conservancy boundary and terrain of the actual region to obtain the first data; Performing collaborative supplementary survey on the fuzzy area to obtain covered water conservancy data, and extracting water conservancy boundaries and terrain from the covered water conservancy data using closed contour lines to obtain second data; the second data includes the terrain of the second area and the second water conservancy boundary; Construct a time series water conservancy topology-specific terrain extraction model based on the first data and the second data, and extract the water conservancy topography data DEM1 of the preset area; Constructing a two-dimensional hydrodynamic model based on the topographic data and water conservancy boundaries of the preset area to simulate flooding, and extracting the wading area according to the flooding range; For the details of water-related buildings, lidar is used to obtain point cloud data of water-related building areas, which are divided into water-retaining buildings and water-passing buildings according to their functional characteristics. Graphic processing and combined operations of addition and removal are performed, and the data are fused with the underwater DEM. The weighted correction and Gaussian smoothing algorithm are used to embed the data into the water conservancy terrain data DEM1 of the preset area to generate the integrated water and land HDEM data of the preset area.

2. The method according to claim 1, wherein, The method for comparing and determining area division based on the radar data and the remote sensing data includes: Edge features are extracted from the optical images of radar data and satellite remote sensing data respectively, and sensor registration is performed by edge feature fusion. Temporal weights are introduced to perform multi-temporal matching on the radar data and satellite remote sensing data after spatial registration. The expression is: The temporal weight of time point t is , adjusted for seasonal variations, the overall similarity score is , the gradient component of the template image in the x direction at point v is , the gradient component of the search image in the x direction at point v is , the gradient component of the template image in the y direction at point v is , the gradient component of the search image in the y direction at point v is The total number of features is , the number of timing templates is ; Radar data and satellite remote sensing data are grouped, and radar data and satellite remote sensing data with a matching degree greater than 0.673 are grouped together. Based on the registered multi-source data, the spatiotemporal consistency of different groups is calculated: wherein the spatio-temporal consistency is , the satellite remote sensing derived point cloud at time t is , the radar point cloud at time t is , the upper limit of the observation time is , the point cloud structure similarity is projecting the point cloud to a 2.5D grid, calculating the elevation mean and intensity variance within the grid, and obtaining the SSIM comparison item according to the elevation mean and the intensity variance; The temporal variance is obtained by summing the temporal standard deviation of spatiotemporal consistency. When the pixel-level consistency is greater than or equal to 0.786 and the temporal variance is less than 0.352, the elevation is stable, the intensity-vegetation index is strongly correlated, and the location of the group is the actual area; when the pixel-level consistency is less than 0.786 and the temporal variance is greater than 0.352, the elevation changes suddenly, the intensity-vegetation index is weakly correlated, and the location of the group is a fuzzy area.

3. The method according to claim 1, wherein, The method for extracting water conservancy boundaries and terrain from the actual area comprises: Based on SAR technology, terrain elements of the actual area are extracted to obtain the terrain of the first area; the potential water body area of ​​the actual area is located by using the slope threshold, the NDWI index is binarized using the maximum inter-class variance method in the potential area, and the initial water body boundary is extracted. For the cloud-covered area of ​​the optical image, the water body backscatter coefficient of the SAR image is used to fill the boundary missing, and the complete water body outline of the actual area is generated as the first water conservancy boundary. The first water conservancy boundary and the terrain of the first area are output as the first data.

4. The method according to claim 1, wherein, The method for collaboratively supplementing the survey of the fuzzy area to obtain terrain and covering water conservancy data of the second area includes: An "air-space-ground" integrated observation network is constructed to achieve data space-time complementarity, obtain medium-density laser radar point cloud data, radar backscattering coefficient and optical vegetation index, identify the type of fuzzy area through a random forest model, and obtain high-priority area, medium-priority area and low-priority area; For high-priority areas, unmanned aerial vehicle data is introduced, the height of the unmanned aerial vehicle is equal to the maximum height of the vegetation plus a safety margin, the overlap rate is proportional to the vegetation complexity, and a mesh flight path is generated based on the scanning angle of the laser radar; The unmanned aerial vehicle scans the same area from different directions, and the data is fused through a point cloud registration algorithm, and the expression is: wherein the rotation matrix is A, the translation vector is w, the u-th point in the source point cloud is , the source point cloud is a point cloud data to be registered to the target point cloud acquired from different scanning angles, the u-th point in the target point cloud is , the directional weight coefficient is , the normal vector is , and the fusion data is G; The multi-temporal radar image and laser data of the same area are spatio-temporally registered, the real terrain is estimated through Kalman filtering as the second area terrain data of the fuzzy area; for the multi-echo laser radar point cloud area, the first echo is separated, the echo intensity difference is calculated, and in the high value area of the echo intensity difference, the potential river center line is identified through local minimum point search; for the multi-source feature fusion area, linear shadow features are extracted from optical images, abnormal strips of backscattering coefficients are analyzed from radar images, combined with water conservancy planning data, the position information of water-related buildings is located through spatial overlay analysis; combined with the second area terrain data, the potential river center line and the position information of the water-related buildings, the output is covered water conservancy data.

5. The method according to claim 1, wherein, The method for extracting the second data comprises: According to the covered water conservancy data, the elevation gradient and water flow direction of the river center line neighborhood are obtained, and a robust fuzzy local information C-means clustering algorithm is used to preliminarily separate water bodies and non-water bodies; A target function of fuzzy clustering is given, and the expression is: where the objective function is , the prior probability of pixel a belonging to cluster c is , the topographic dissimilarity measure of pixel a to cluster c is , the fuzzy local variable of pixel a to cluster c is , the ridge regression coefficient is , the total number of pixels is , the membership of pixel a to cluster c is , the spectral dissimilarity measure of pixel a to cluster c is , the balancing parameter is , the balancing spectral and topographic features, the elevation coefficient of variation of pixel a is , the elevation coefficient of variation of cluster c is ; The mean, covariance and non-similarity measure are calculated: where the mean of cluster c is , the covariance is , the transpose of a matrix is , the spectral measure of pixel a is , the data dimensionality is , the mean of cluster c pixel a is ; The local variation density and fuzzy local variable are calculated: where the local variation density of pixel a is , the neighborhood pixels of pixel a are , the neighborhood pixel set is , the neighborhood pixel spectral test mean centered on pixel a is , the neighborhood pixel spectral test variance centered on pixel a is , the local variation density of neighborhood pixel is ; The prior probability and membership are calculated: wherein the index cluster is , the field strength is , the pixel a attribute is , the cluster number is , the prior probability of pixel a belonging to cluster i is , the spectral dissimilarity measure of pixel a and cluster i is , and the fuzzy local variable of pixel a and cluster i is ; If the maximum error of the target function from time point t to time point t+1 is less than the error threshold, the membership matrix is de-fuzzified to obtain a binary segmentation result, a morphological closing operation is used to fill small holes in the binary segmentation result, and the maximum connected domain contour is extracted; The water collection constraint is applied to the contour of the maximum connected domain, and the water cluster center is forced to satisfy , the isolated cluster with an area less than the area threshold is removed, and the segmentation data is obtained; wherein the neighborhood elevation gradient is , and the water flow direction is ; The gradient and gradient amplitude of the fuzzy water area are obtained according to the segmented data, an anisotropic diffusion filter is applied to the segmented data to preserve the sharpness of the river boundary, and the expression is: Wherein the edge stop function is , control diffusion intensity, the edge sensitive coefficient in anisotropic diffusion filtering is , according to the boundary gradient adaptive adjustment, the gradient amplitude is s, and the gradient of image intensity is ; The boundary of the segmented data is taken as the second water conservancy boundary data, and the second data is output in combination with the second area terrain data.

6. The method according to claim 1, wherein, The method for constructing a time-series water conservancy topology special terrain extraction model according to the first data and the second data comprises: A terrain base grid is constructed based on the terrain data of the first data and the second data, static parameters are extracted, time-series difference analysis is performed on the water conservancy boundaries of the first data and the second data to identify boundary change areas, dynamic boundaries are subjected to vector-to-raster conversion to generate a time-series raster layer, and a dynamic topology node set is extracted; the static parameters include terrain elevation, slope and slope direction; The dynamic nodes are matched with the static nodes, a space-time joint node set is generated through spatial overlay analysis, a node attribute table is established to record the spatial coordinates, time attributes and water conservancy attributes of the nodes; Based on the spatio-temporal joint node set, a dynamic topology edge set is constructed by a triangulation algorithm of the point set, wherein: the adjacency relation represents the direct connectivity state between nodes, the time sequence association represents the state change of the same node at different time steps; the edge attribute includes distance, flow direction, and water capacity; The distance, water level difference, and time step are used as the time sequence weight of the dynamic topology edge to construct a special terrain extraction model for the time sequence water conservancy topology, and the water conservancy terrain data DEM1 of a preset area is output.

7. The method according to claim 1, wherein, The method for obtaining the water area includes: The first water conservancy boundary and the second water conservancy boundary are merged into a complete water conservancy boundary to form river network data; The preset area water conservancy terrain data DEM1 is meshed, the water conservancy boundary adjacent area and the water conservancy data covered area are meshed according to a grid size of 5m*5m, and other areas far away from the river are meshed according to a grid size of 20m*20m; the meshed grids are assigned with water dynamic parameters such as roughness, area coefficient, and time step, and a mesh model result is output; According to historical flow data of the preset area, a flood frequency curve is fitted, a design flood peak flow corresponding to a target frequency is determined, and the design flood peak flow is used as an inflow boundary condition; Based on the control volume method, a two-dimensional Saint-Venant shallow water equation is used, the design maximum flood peak flow of the river in the preset area is used as a boundary condition for flood inundation simulation, and the inundation range is output as the water area.

Images