H3 grid-based urban low-altitude airspace digital modeling method and system

Abstract

The application belongs to the technical field of low-altitude airspace management, and relates to a city low-altitude airspace digital modeling method and system based on an H3 grid, comprising: constructing a multi-height layer H3 hexagonal airspace grid; obtaining a risk index based on the multi-height layer H3 hexagonal airspace grid; obtaining a traffic index according to the multi-height layer H3 hexagonal airspace grid and a preset safety buffer distance; performing adaptive merging on the multi-height layer H3 hexagonal airspace grid based on the risk index and the traffic index to obtain a multi-level airspace grid; and performing merging quality evaluation on the multi-level airspace grid to generate a grid-level low-altitude airspace digital model. The application solves the problem of low-altitude airspace digital model distortion by using the scale-invariant characteristics of the H3 hexagonal grid, merging homogenous grids through a double-attribute sequence similarity index, and performing merging quality evaluation through a structural similarity index and an extreme value preservation index.

Description

Technical Field

[0001] This invention relates to the field of low-altitude airspace management technology, and specifically discloses a digital modeling method and system for urban low-altitude airspace based on H3 grids. Background Technology

[0002] With the rapid popularization of drones in urban logistics, emergency response, and infrastructure inspection, the strategic value of urban low-altitude airspace (within 120 meters above ground) as a three-dimensional urban transportation resource is becoming increasingly prominent. The International Civil Aviation Organization (ICAO) has incorporated unmanned aerial vehicle (UTM) traffic management into its global aviation planning system, and China also explicitly listed the low-altitude economy as a new growth engine in 2024. How to conduct a refined and quantitative assessment of urban low-altitude airspace resources has become a key technical issue in the construction of low-altitude economic infrastructure.

[0003] Currently, most existing low-altitude airspace digitization schemes use spatial grid discretization to represent the airspace. Among them, the scheme based on rectangular grids discretizes the airspace into square units with a fixed latitude and longitude step size. However, the area of ​​the rectangular grid varies significantly at different latitudes, and the distance between the neighborhood centers is not uniform, which causes systematic bias in spatial aggregation analysis and leads to distortion of the digitization model.

[0004] However, to accurately characterize the spatial heterogeneity and time-varying properties of urban low-altitude airspace, gridded modeling typically requires high-resolution grids and the overlay of attribute sequences across multiple time periods and buffer distances. This leads to a dramatic increase in the number of grid cells within the city-level area, resulting in a grid explosion problem and posing a severe challenge to storage overhead and subsequent path planning computations. To reduce the scale, existing solutions often employ uniform downsampling or simple hierarchical coarsening. However, these methods do not distinguish between the attribute differences of the grids and easily smooth adjacent heterogeneous grids together, causing the loss of key spatial features. In particular, local features crucial to flight safety, such as high-risk area boundaries and traffic bottlenecks, are idealized and smoothed out, thus distorting the digital model. How to significantly compress the grid size, save storage and computing resources, while maintaining the spatial fidelity of airspace attributes and avoiding the idealization of high-risk, low-traffic areas, has become a pressing technical problem to be solved in the digital modeling of low-altitude airspace. Summary of the Invention

[0005] The technical problem to be solved by this invention is the distortion of digital models of low-altitude airspace.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A digital modeling method for urban low-altitude airspace based on H3 grids includes: S1. Construct a multi-height-layer H3 hexagonal spatial grid; S2. Based on the multi-height layer H3 hexagonal spatial grid, risk indicators are obtained, which include risk values ​​and risk attribute sequences. S3. Based on the multi-height layer H3 hexagonal airspace grid and the preset safety buffer distance, obtain the access index, which includes the airspace accessibility index and the accessibility attribute sequence. S4. Based on risk indicators and accessibility indicators, adaptive merging of multi-height layer H3 hexagonal spatial grids is performed to obtain multi-level spatial grids. S5. Merge and assess the quality of multi-level airspace grids to generate a grid-level low-altitude airspace digital model.

[0007] Furthermore, a multi-height-layer H3 hexagonal spatial grid is constructed, including: Obtain the vector boundary data of the target city and generate the target city region; Based on the target city area, a two-dimensional hexagonal grid set of the target city is generated by filling the sphere with the H3 hexagonal grid system at a preset resolution level. Based on a set of two-dimensional hexagonal meshes and a preset set of height layers, a three-dimensional spatial mesh is obtained through Cartesian product. Obtain parameter data and digital elevation model of the target city, spatially associate the parameter data and digital elevation model with the three-dimensional spatial grid, construct multi-source geographic attribute tables for each grid unit, and generate a multi-height layer H3 hexagonal spatial grid.

[0008] Furthermore, based on the multi-height layer H3 hexagonal spatial grid, risk indicators were obtained, including: Based on land cover data and building vector data in the multi-height H3 hexagonal spatial grid, the ground within each grid cell is functionally partitioned to obtain multiple ground functional zones; the area proportion of each ground functional zone within the grid cell is calculated to obtain the area proportion of each ground functional zone. Based on the building vector data, obtain the building morphology indicators within each grid unit; Dynamic population density data is acquired, and the dynamic population density data is mounted to each grid cell of the multi-height layer H3 hexagonal spatial grid through the spatial interpolation method to generate a time-varying population density sequence. Based on time-varying population density sequences and various surface functional zones, as well as the acquired dynamic population data, the population distribution patterns are extracted to obtain the indoor and outdoor population. Based on the obtained international unmanned aerial vehicle (UAV) flight safety assessment standards, UAV type parameters, and flight altitude h, the critical impact area on the ground after the UAV crashes is obtained. The ground shading coefficient is obtained based on ground functional zones and building form indicators; Population exposure risk is obtained based on the area ratio of ground functional zones, indoor population, outdoor population, ground shielding coefficient, and area of ​​the ground critical impact zone; Based on population exposure risk, risk attribute sequences are calculated for each preset time period; and risk values ​​are obtained based on the risk attribute sequences.

[0009] Furthermore, based on the multi-height layer H3 hexagonal spatial grid and the preset safety buffer distance, passage indicators are obtained, including: Based on the multi-height H3 hexagonal spatial grid, multiple directional connections are obtained for each grid cell; According to the preset safety buffer distance, spatial expansion obstacles are screened in the multi-height layer H3 hexagonal spatial grid to obtain the effective obstacle expansion area; the number of intersections between the lines in each direction and the effective obstacle expansion area is counted to obtain the collision count; Based on the number of collisions, the airspace traversability index is obtained; Calculate the airspace mobility index for all altitude layers and the safe buffer distance, and generate a sequence of mobility attributes.

[0010] Furthermore, based on risk and accessibility indicators, the multi-height H3 hexagonal spatial grids are adaptively merged to obtain a multi-level spatial grid, including: S41. Construct candidate merging clusters based on the multi-height layer H3 hexagonal spatial grid; S42. Based on the candidate merge clusters and the corresponding risk attribute sequence and commonality attribute sequence, construct an attribute matrix and perform row-level standardization on the attribute matrix; S43. Based on the row-level standardized attribute matrix, obtain the single-attribute heterogeneity measure; S44. Based on the single-attribute heterogeneity measure, nonlinear fusion is performed using the Minkowski norm to obtain the dual-attribute sequence similarity index. S45. Determine whether the similarity index of the dual-attribute sequence is less than the preset merging threshold, whether the maximum value of the risk value is less than the preset high-risk threshold, and whether the minimum value of the spatial traversability index is greater than the preset low traversability threshold. If so, adopt a bottom-up aggregation mechanism to adaptively merge adjacent grid cells within the candidate merging cluster; otherwise, merging is prohibited. S46. Repeat steps S41 to S46 until no new merges are generated or the preset number of iterations is reached, to obtain the merged multi-level spatial grid.

[0011] Furthermore, the formula for calculating the dual-attribute sequence similarity index is as follows: , in, It is a two-attribute sequence similarity index. The weights of the risk attributes, Here, p represents the weight of the commutability attribute, and p is the Minkowski parameter. As a measure of heterogeneity of risk attributes, It is a measure of heterogeneity of commonality attributes.

[0012] Furthermore, a quality assessment is performed on the merged multi-level airspace grids to generate a grid-level low-altitude airspace digital model, including: Structural similarity is evaluated for multi-level spatial grids to obtain a structural similarity index; An extremum preservation index is obtained by evaluating the extremum preservation of multi-level spatial grids. Determine whether the structural similarity index is greater than the structural similarity threshold and whether the extreme value preservation index is greater than the extreme value preservation threshold. If so, generate a grid-level low-altitude airspace digital model; otherwise, return to S4.

[0013] Furthermore, the formula for calculating the structural similarity index is: , in, Let X be the structural similarity index, and Y be the attribute spatial distribution matrix before merging and Y be the attribute spatial distribution matrix after merging. The mean before the merger. This is the combined mean. The variance before merging. The variance after aggregation. For covariance, and All of these are set stability constants.

[0014] Furthermore, the formula for calculating the extreme value preservation index is: , in, To preserve the extremum, This is a set of extreme value grids identified based on a preset percentile threshold before merging. This is a set of extreme value grids identified after merging based on a preset percentile threshold. For the symmetric difference of sets, Indicates the number of elements in the set. This is the attenuation parameter.

[0015] This invention also relates to an H3 grid-based digital modeling system for urban low-altitude airspace, used in the aforementioned H3 grid-based digital modeling method for urban low-altitude airspace, comprising: The H3 mesh building module is used to construct multi-height-layer H3 hexagonal spatial meshes. The risk value generation module is used to obtain risk indicators based on the multi-height layer H3 hexagonal spatial grid. The risk indicators include risk values ​​and risk attribute sequences. The airspace accessibility index generation module is used to obtain accessibility indicators based on the multi-height layer H3 hexagonal airspace grid and the preset safety buffer distance. The accessibility indicators include the airspace accessibility index and the accessibility attribute sequence. The grid merging module is used to adaptively merge multi-height H3 hexagonal spatial grids based on risk indicators and accessibility indicators to obtain multi-level spatial grids. The digital model generation module is used to perform quality assessment on multi-level airspace grids and generate grid-level low-altitude airspace digital models.

[0016] The beneficial effects of this invention are: This invention utilizes the scale-invariant properties of H3 hexagonal meshes, where each cell area is approximately equal at the same resolution, six neighboring cells are equidistant, and multi-level aggregation with proportional aperture is supported. It constructs a multi-height layer H3 hexagonal airspace mesh and intelligently identifies and adaptively merges adjacent meshes with similar attributes using a dual-attribute sequence similarity index. This significantly compresses the mesh size, saves storage space and subsequent path planning computational overhead, while maintaining the spatial fidelity of airspace attributes. By setting merging constraints for high-risk, low-traffic areas and a conservative aggregation method for risk values, it ensures that key local features such as high-risk area boundaries and traffic bottlenecks are not smoothed or idealized during merging, avoiding the underestimation of risk in the digital model. The Structural Similarity Index (SSIM) and the Extremum Preservation Index (EPI) quantify the merging quality from two dimensions: global spatial distribution patterns and local key extreme features. This ensures that the digital model maintains high fidelity while significantly compressing, providing a streamlined, reliable, and directly usable grid-level digital foundation for low-altitude airspace organization, airspace capacity assessment, and path planning. This solves the problem of distortion in low-altitude airspace digital models. Attached Figure Description

[0017] Figure 1 This is a flowchart of a digital modeling method for urban low-altitude airspace based on H3 grids in an embodiment of the present invention; Figure 2 This is a block diagram of an urban low-altitude airspace digital modeling system based on H3 grids in an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of grid merging in an embodiment of the present invention.

[0019] The attached diagram lists the components represented by each number as follows: 1. Candidate merged clusters; 2. Upper-level grid cells. Detailed Implementation

[0020] The principles and features of the present invention are described below. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0021] A digital modeling method for urban low-altitude airspace based on H3 grids, such as Figure 1 As shown, it includes the following steps: S1. Construct a multi-height-layer H3 hexagonal spatial grid; S11. Obtain the vector boundary data of the target city and generate the target city region; Using the CGCS-2000 coordinate system as a spatial reference datum, the vector boundary data of the target city is obtained, and the target city region is generated based on the vector boundary data.

[0022] S12. Based on the target city area, use the H3 hexagonal grid system to fill the sphere with a preset resolution level to generate a two-dimensional hexagonal grid set for the target city. The H3 hexagonal grid system is used by Uber H3. The H3 hexagonal grid system is used to fill the target city area with a sphere at a resolution level r, generating a two-dimensional hexagonal grid set covering the city area.

[0023] The resolution level r can be set to 10, corresponding to a two-dimensional hexagonal grid with a side length of approximately 76m and an area of ​​approximately 0.03km². In the H3 hexagonal grid system, the areas of all hexagons at the same resolution level are approximately equal, and each two-dimensional hexagonal grid has exactly 6 equidistant neighbors, facilitating neighborhood topology analysis.

[0024] S13. Based on the two-dimensional hexagonal mesh set and the preset height layer set, obtain the three-dimensional spatial mesh through Cartesian product; Define the set of height layers L={h1,h2…,hn} for the airspace to be evaluated. The airspace to be evaluated refers to the altitude AGL above the ground, corresponding to the 0-120 meter range of the Class W low-altitude airspace in China. Take the Cartesian product of the two-dimensional hexagonal grid set and the set of height layers to obtain the three-dimensional airspace grid G(index, h). The three-dimensional airspace grid includes multiple grid cells, and each grid cell is uniquely identified by the hexagonal grid index and the height layer i.

[0025] The Cartesian product is a basic set operation used to describe all possible combinations of ordered pairs (or ordered n-tuples) between two or more sets.

[0026] S14. Obtain parameter data and digital elevation model of the target city, spatially associate the parameter data and digital elevation model with the three-dimensional spatial grid, construct the multi-source geographic attribute table of each grid unit, and generate a multi-height layer H3 hexagonal spatial grid.

[0027] The target city's parameter data includes land cover data, building vector data, road network data, population raster data, and no-fly zone boundary data.

[0028] This invention uses a multi-height-layer H3 hexagonal airspace grid to discretize urban low-altitude airspace, making the area of ​​each grid cell approximately consistent, the neighborhood relationship regular, and the directional connection uniform, thereby improving the uniformity and comparability of airspace resource assessment results between different regions and different height layers.

[0029] S2. Risk indicators are obtained based on the multi-height layer H3 hexagonal spatial grid. Risk indicators include risk values ​​and risk attribute sequences.

[0030] S21. Based on land cover data and building vector data in the multi-height layer H3 hexagonal spatial grid, the ground within each grid cell is functionally partitioned to obtain multiple ground functional zones; the area ratio of each ground functional zone within the grid cell is calculated to obtain the area ratio of each ground functional zone. The surface functional zones include open water areas (referring to water body surfaces), vegetated green areas (referring to vegetated surfaces), building-covered areas (referring to the superimposed area of ​​building outlines), vehicular road areas (referring to the area of ​​roads used by motor vehicles), and pedestrian road areas (referring to the area of ​​pedestrian walkways). The area proportions Aj of each surface functional zone are normalized so that the sum of the area proportions of each surface functional zone is 1.

[0031] S22. Based on the building vector data, obtain the building morphology indicators within each grid unit; Building form indicators include building coverage (the ratio of building footprint to grid cell area), weighted average building height (weighted by building footprint area), and maximum building height; these building form indicators are used as input data for subsequent shading coefficient calculations.

[0032] S23. Obtain dynamic population density data, and use the spatial difference method to mount the dynamic population density data to each grid cell of the multi-height layer H3 hexagonal spatial grid to generate a time-varying population density sequence. Using data from operator communication base stations or population heat maps, dynamic population density data ρ(index,t) is obtained for 30 time periods covering weekdays and weekends from 08:00 to 22:00, with a time resolution of 1 hour.

[0033] S24. Based on the time-varying population density sequence and various surface functional zones, as well as the acquired dynamic population data, extract the population distribution pattern to obtain the indoor population and outdoor population. Based on the time-varying population density sequence and the acquired dynamic population data, the population distribution patterns of various ground functions are extracted, and the total population in each grid unit is divided into indoor population (referring to those mainly distributed inside the building coverage area and protected by the building structure) and outdoor population (referring to those distributed in open areas such as roads and green spaces and directly exposed to the risk of falling).

[0034] If dynamic population data cannot extract more detailed distribution patterns, then a uniformly distributed ρ(index,t) is used as the population density in different functional areas.

[0035] S25. Based on the obtained international unmanned aerial vehicle (UAV) flight safety assessment standards, UAV type parameters, and flight altitude, obtain the area of ​​the critical impact zone on the ground after the UAV crashes due to a malfunction. Obtain international drone flight safety assessment standards, drone type parameters, and flight altitude h; the international drone flight safety assessment standards can be the JARUS SORA model; drone type parameters include fuselage size, flight speed, and mass.

[0036] S26. Based on the ground functional zones and building form indicators, obtain the ground shading coefficient; Ground shading coefficient ,in These correspond to five types of ground functional zones; the ground shielding coefficient ranges from [0,1]; the ground shielding coefficient is highest in the building-covered area because the building structure has a significant protective effect against falling impacts; the ground shielding coefficient is lowest in the open water area and the vegetated green area; the specific value of the ground shielding coefficient is determined by the protection efficiency of each type of ground functional zone against typical impact energy.

[0037] S27. Obtain population exposure risk based on the area ratio of ground functional zones, indoor population, outdoor population, ground shielding coefficient, and area of ​​ground critical impact zone; The formula for calculating population exposure risk is: , in, For population exposure risk; The accident rate of drones flying within the grid cell; The area of ​​the critical influence zone on the ground; The area ratio of the ground functional zone j; The effective population density corresponding to ground functional zone j within the grid cell at time t (assigned values ​​according to indoor / outdoor attributes); For shielding penetration rate, it represents the probability that a person is actually threatened by falling in this type of ground functional area; This is the ground shading coefficient.

[0038] S28. Based on population exposure risk, calculate time-varying risk sequences for each preset time period; obtain risk values ​​based on the time-varying risk sequences.

[0039] Based on population exposure risk, time-varying risk sequences are calculated for 30 time periods on weekdays and weekends respectively; based on the time-varying risk sequences, the average value for all time periods is obtained, and the average value for all time periods is used as a static representative risk value and output.

[0040] Each grid cell calculates a time-varying risk sequence for each preset time period, which is a risk attribute sequence.

[0041] This invention incorporates ground functional zoning, building morphology indicators, ground shading coefficient, indoor population, outdoor population, and dynamic population density data into population exposure risk calculation, enabling risk assessment to reflect actual exposure differences under different surface types, time periods, and spatial locations, thereby improving the accuracy of low-altitude operation safety risk assessment.

[0042] S3. Obtain passage indicators based on the multi-height layer H3 hexagonal spatial grid and the preset safety buffer distance; Traffic indicators include airspace trafficability index and trafficability attribute sequence.

[0043] S31. Based on the multi-height layer H3 hexagonal spatial grid, obtain multiple directional connection lines for each grid cell; For each grid cell, based on its six-neighbor topology, the center point of the current grid cell is connected to the center points of each of the adjacent grid cells to obtain 6 evenly distributed directional lines, and the included angle between any two of the 6 directional lines is 60°.

[0044] S32. According to the preset safety buffer distance, the spatial expansion obstacle screening is carried out on the multi-height layer H3 hexagonal spatial grid to obtain the effective obstacle expansion area; the number of intersections between the connecting lines in each direction and the effective obstacle expansion area is counted to obtain the collision count; Set a safe buffer distance d, where The system is divided into 11 levels; the building outline, no-fly zone boundary, and obstacles are spatially expanded according to a safe buffer distance d, and obstacles above the height level ld are screened to obtain the effective obstacle expansion area; the number of intersections between each directional line and the effective obstacle expansion area is counted to obtain the collision count for that direction. (k = 1,...,6).

[0045] S33. Based on the number of collisions, obtain the airspace traversability index; The formula for calculating the airspace passability index is: , in, AAI is the airspace accessibility index, with a value range of [0,1]. The larger the AAI value, the better the accessibility conditions of the grid cell at the corresponding height and safe buffer distance. The proportion of collision-free directions to the total number of directions reflects the degrees of freedom for passage in each direction. This represents the normalized average collision intensity. This is the maximum value for the collision count, which can be 10.

[0046] S34. Calculate the airspace traversability index for all altitude layers and the safe buffer distance, and generate a multi-buffer traversability sequence.

[0047] The multi-buffered accessibility sequence calculated by each grid cell for each preset safety buffer distance is the accessibility attribute sequence.

[0048] S4. Based on risk indicators and accessibility indicators, adaptive merging of multi-height layer H3 hexagonal spatial grids is performed to obtain multi-level spatial grids. To address the problem of exploding high-resolution mesh counts, this invention intelligently identifies and merges adjacent meshes with similar attributes by measuring the consistency of their attributes in two dimensions: risk and accessibility. This allows for the reduction of mesh size while maintaining spatial fidelity. Figure 3 As shown, it specifically includes: S41. Based on the multi-height layer H3 hexagonal spatial grid, construct candidate merging cluster 1; For any central grid cell within the same height layer and its six equidistant adjacent grid cells, a candidate merging cluster 1 containing up to seven grid cells is formed. Utilizing the scale-invariant property of the H3 hexagonal grid, this candidate merging cluster 1 maintains a consistent topology across different resolution levels, facilitating bottom-up, hierarchical merging.

[0049] S42. Based on candidate merge cluster 1 and the corresponding risk attribute sequence and commonality attribute sequence, construct an attribute matrix and perform row-level standardization on the attribute matrix. For the risk attribute sequence and the traversability attribute sequence corresponding to candidate merge cluster 1, respectively, an attribute matrix is ​​constructed based on the attribute sequence of each grid cell within candidate merge cluster 1 according to the sequence position. The rows of the matrix correspond to the sequence position (time period or safety buffer distance), and the columns correspond to the grid cells within candidate merge cluster 1. To eliminate the dimensional differences between different sequence positions and enhance robustness, the attribute matrix is ​​row-level standardized. , in, This represents the attribute value of the i-th grid cell within candidate merge cluster 1 at sequence position k. This represents the mean of the attribute values ​​of all grid cells at sequence position k. Let k be the standard deviation of all grid cell attribute values ​​at sequence position k. For numerically stable terms, These are the standardized attribute values.

[0050] The standardization is performed row by row according to the sequence position, which eliminates the influence of the absolute numerical scale at each position while preserving the relative differences between grid cells.

[0051] S43. Based on the row-level standardized attribute matrix, obtain the single-attribute heterogeneity measure; S431. Based on the row-level standardized attribute matrix, the deviation of attributes within candidate merged cluster 1 is obtained.

[0052] This bias refers to the difference between the local statistics and global statistics of the attributes of each grid cell within candidate merge cluster 1, quantifying the heterogeneity of attributes within the cluster. For each grid cell i, the mean and standard deviation of its standardized attribute sequence are calculated as local statistics; the dispersion of the local means of each grid cell is then statistically analyzed. The degree of dispersion of the local standard deviation of each grid cell : Describe the differences in the center position of each grid attribute. Spatial differences reflecting the variability of attributes of each grid.

[0053] S432. The bias is mapped to the [0,1] interval by quantile normalization, and a single-attribute heterogeneity measure is obtained by equal-weight fusion.

[0054] The formula for calculating the single-attribute heterogeneity measure is: , in, These represent risk attributes and commonality attributes, respectively, where R represents risk attributes and T represents commonality attributes. Indicates quantile normalization; It is a measure of heterogeneity for a single attribute, with a value range of [0,1]. The smaller the value, the more homogeneous the attribute is within the cluster.

[0055] S44. Based on the single-attribute heterogeneity measure, nonlinear fusion is performed using the Minkowski norm to obtain the dual-attribute sequence similarity index. Based on the single-attribute heterogeneity measure, a nonlinear fusion of risk and commonality attributes is performed using the Minkowski norm to obtain a two-attribute sequence similarity index. The formula for calculating the two-attribute sequence similarity index is as follows: , in, The DSSI is a dual-attribute sequence similarity index. The smaller the DSSI value, the more consistent the grid cells within candidate merged cluster 1 are in terms of both risk and accessibility. The weights of the risk attributes, In this embodiment, equal weights are used for the versatility attribute. = =0.5 to consider both dimensions equally; p is the Minkowski parameter, with a value between 1 and 2 (e.g., 1.5), providing a compromise metric between Manhattan distance and Euclidean distance, maintaining computational efficiency while taking into account a balanced characterization of the heterogeneity of the two dimensions; As a measure of heterogeneity of risk attributes, It is a measure of heterogeneity of commonality attributes.

[0056] S45. Determine whether the similarity index of the dual-attribute sequence is less than the preset merging threshold, whether the maximum value of the risk value is less than the preset high-risk threshold, and whether the minimum value of the spatial traversability index is greater than the preset low traversability threshold. If so, adopt a bottom-up aggregation mechanism to adaptively merge adjacent grid cells in candidate merging cluster 1; otherwise, merging is prohibited. When the dual-attribute sequence similarity index (DSSI) meets the preset merging condition, i.e., the DSSI is less than the preset merging threshold τ, candidate cluster 1 is initially determined to be mergingable. To avoid idealizing key features such as high-risk area boundaries and traffic bottlenecks, and to avoid underestimating the risk, a high-risk, low-traffic constraint is further applied to the initially merging candidate cluster 1: only when the maximum risk value of each grid cell within candidate cluster 1 is reached... Less than the preset high-risk threshold And the minimum value of the spatial accessibility index of each grid cell. Greater than the preset low passage threshold If the candidate cluster 1 is determined not to belong to a high-risk, low-traffic area, merging is allowed; otherwise, even if the DSSI meets the merging conditions, merging is prohibited, and the original mesh resolution of the area is preserved.

[0057] The merge indicator function is expressed as: Merge(C) = 1 if and only if DSSI < and and Otherwise, Merge(C) = 0; where C is candidate merge cluster 1.

[0058] This constraint ensures that high-risk, low-access critical grids are always preserved at the highest resolution, making the digital model's portrayal of risk more conservative.

[0059] For candidate merge clusters 1 that are allowed to be merged, a bottom-up aggregation mechanism is used to merge them into a higher-level grid cell 2. A conservative inheritance strategy is applied to aggregate their dual-attribute sequences, meaning that the merged grid cell inherits the worse attribute value from the sub-grids—a higher value for the risk dimension and a lower value for the traversability dimension. This approach prioritizes overestimating risk and underestimating traversability over an overly optimistic assessment. The risk attribute sequence is determined by the maximum value of each grid cell within candidate merge cluster 1, based on its sequence position. , The traversability attribute sequence is determined by taking the minimum value of each grid cell within candidate merge cluster 1 according to sequence position: , in, Let i be the risk value of grid cell i at sequence position k. Let i be the risk value and the accessibility value of grid cell i at sequence position k. This refers to the risk value inherited by the merged grid cell at sequence position k. These are the accessibility values ​​inherited by the merged grid cells at sequence position k. Correspondingly, the representative risk value of the merged grid cells is the maximum risk value of all grid cells within candidate merged cluster 1. This conservative inheritance ensures that the risk of any upper-level grid after merging is not underestimated and the accessibility is not overestimated, thus avoiding the idealization of high-risk and access bottleneck areas.

[0060] S46. Repeat steps S41 to S46 until no new merges are generated or the preset number of iterations is reached, to obtain the merged multi-level spatial grid.

[0061] In city-level examples such as Shenzhen, by adjusting the merging threshold τ, this strategy can achieve a grid compression rate of approximately 60%-90% while maintaining the main characteristics of the spatial distribution of risk and accessibility, significantly reducing storage overhead and the computational complexity of subsequent path planning.

[0062] S5. Merge and assess the quality of multi-level airspace grids to generate a grid-level low-altitude airspace digital model.

[0063] To ensure that the merged grid can still accurately represent the original spatial characteristics, this invention quantifies the merging quality from two dimensions: global spatial distribution pattern and local key extreme value characteristics.

[0064] Merger quality assessment includes structural similarity assessment and extreme value preservation assessment.

[0065] S51. Perform structural similarity evaluation on multi-level spatial grids to obtain structural similarity index; The Structural Similarity Index (SSIM) quantifies the degree of preservation of attribute spatial distribution patterns before and after merging by comprehensively evaluating three dimensions: brightness, contrast, and structure. The formula for calculating the SSIM is as follows: , Where X is the attribute space distribution matrix before merging, and Y is the attribute space distribution matrix after merging; The mean before the merger. This is the combined mean, which reflects the consistency of the overall attribute level. The variance before merging. The variance is the combined variance, which measures the degree to which attribute variability is preserved. To measure covariance, the similarity of spatially correlated patterns is quantified; and All values ​​are stability constants set based on the dynamic range of the attributes to prevent numerical instability when the denominator approaches zero. The SSIM value ranges from [-1, 1]. When SSIM exceeds 0.85, it indicates that the merged mesh can effectively maintain the main characteristics of the original spatial distribution.

[0066] S52. Perform an extremum preservation evaluation on a multi-level spatial grid to obtain the extremum preservation index; Extreme areas in the urban low-altitude environment often correspond to key management elements, such as high-risk hotspots in densely populated areas or traffic bottlenecks in densely built-up areas. The Extreme Value Preservation Index (EPI) specifically assesses the degree to which the merging process retains these key characteristics: , in, and These are the extreme value grid sets identified before and after the merger based on preset percentile thresholds (such as the top 5% being high-risk or the bottom 5% being low-accessibility). The symmetric difference of a set measures the magnitude of change in the distribution of extreme values. This represents the number of elements in the set, and the denominator is normalized to ensure comparability of data of different sizes; EPI is the attenuation parameter, which controls the sensitivity of the exponential function to extreme value loss. The value range of EPI is (0,1], and the closer it is to 1, the more completely the extreme value characteristics are preserved.

[0067] Through the joint evaluation of SSIM and EPI, this invention establishes a comprehensive merged quality assurance mechanism: SSIM ensures the stability of the global spatial pattern, while EPI protects local extremal features that are critical to security and efficiency.

[0068] S53. Determine whether the structural similarity index is greater than the structural similarity threshold and whether the extreme value preservation index is greater than the extreme value preservation threshold. If so, generate a grid-level low-altitude airspace digital model; otherwise, return to S4.

[0069] If the merge quality does not meet the preset requirements, the corresponding merge operation can be rolled back or the merge threshold τ can be adjusted.

[0070] When the merging quality meets the preset requirements, the grid index, height layer, dual-attribute sequence (time-varying risk sequence and multi-buffered traversability sequence), merging level and aggregation attribute of the merged multi-level airspace grid are associated to generate and output a grid-level low-altitude airspace digital model.

[0071] The grid-level low-altitude airspace digital model supports querying and statistics by region, altitude layer, risk level, and traffic conditions, and can be further transformed into a graph structure with topological connections, providing a simplified, reliable, and directly callable grid-level digital foundation for low-altitude route organization, airspace capacity assessment, take-off and landing point selection, and route planning.

[0072] This invention utilizes the scale-invariant property of the H3 hexagonal grid to intelligently identify and merge adjacent grids with similar attributes through a dual-attribute sequence similarity index. This significantly reduces grid size and saves storage and computational costs while maintaining the spatial fidelity of airspace attributes. Furthermore, it avoids idealizing key features through high-risk, low-traffic constraints and conservative aggregation of risk values, thereby providing a streamlined and efficient digital foundation for upper-level applications in low-altitude airspace.

[0073] This invention also relates to an H3 grid-based digital modeling system for urban low-altitude airspace, used in the aforementioned H3 grid-based digital modeling method for urban low-altitude airspace, such as... Figure 2 As shown, it includes: The H3 mesh building module is used to construct multi-height-layer H3 hexagonal spatial meshes. The risk value generation module is used to obtain risk indicators based on the multi-height layer H3 hexagonal spatial grid. The risk indicators include risk values ​​and risk attribute sequences. The airspace accessibility index generation module is used to obtain accessibility indicators based on the multi-height layer H3 hexagonal airspace grid and the preset safety buffer distance. The accessibility indicators include the airspace accessibility index and the accessibility attribute sequence. The grid merging module is used to adaptively merge multi-height H3 hexagonal spatial grids based on risk indicators and accessibility indicators to obtain multi-level spatial grids. The digital model generation module is used to perform quality assessment on multi-level airspace grids and generate grid-level low-altitude airspace digital models.

[0074] In other implementations, the hexagonal mesh can be replaced with a GeoSOT spherical mesh, S2 spherical subdivision, or BeiDou grid code; the H3 mesh resolution level can be adjusted within the range of 5-15 levels; the number of height layers can be set within the range of 2-10 layers; the Minkowski parameter p in the dual-attribute sequence similarity index can take values ​​within the range of 1-2, and the weights... and Unweighted allocation can be made based on the relative importance of risk and accessibility; in addition to SSIM and EPI, peak signal-to-noise ratio, root mean square error, and other indicators can also be used for quality assessment.

[0075] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A digital modeling method for urban low-altitude airspace based on H3 grids, characterized in that, include: S1. Construct a multi-height-layer H3 hexagonal spatial grid; S2. Based on the multi-height layer H3 hexagonal spatial grid, risk indicators are obtained, which include risk values ​​and risk attribute sequences. S3. Based on the multi-height layer H3 hexagonal airspace grid and the preset safety buffer distance, obtain the access index, which includes the airspace accessibility index and the accessibility attribute sequence. S4. Based on risk indicators and accessibility indicators, adaptive merging of multi-height layer H3 hexagonal spatial grids is performed to obtain multi-level spatial grids. S5. Merge and assess the quality of multi-level airspace grids to generate a grid-level low-altitude airspace digital model.

2. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 1, characterized in that, The construction of the multi-height-layer H3 hexagonal spatial mesh includes: Obtain the vector boundary data of the target city and generate the target city region; Based on the target city area, a two-dimensional hexagonal grid set of the target city is generated by filling the sphere with the H3 hexagonal grid system at a preset resolution level. Based on a set of two-dimensional hexagonal meshes and a preset set of height layers, a three-dimensional spatial mesh is obtained through Cartesian product. Obtain parameter data and digital elevation model of the target city, spatially associate the parameter data and digital elevation model with the three-dimensional spatial grid, construct multi-source geographic attribute tables for each grid unit, and generate a multi-height layer H3 hexagonal spatial grid.

3. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 1, characterized in that, The risk indicators obtained based on the multi-height-layer H3 hexagonal spatial grid include: Based on land cover data and building vector data in the multi-height H3 hexagonal spatial grid, the ground within each grid cell is functionally partitioned to obtain multiple ground functional zones; the area proportion of each ground functional zone within the grid cell is calculated to obtain the area proportion of each ground functional zone. Based on the building vector data, obtain the building morphology indicators within each grid unit; Dynamic population density data is acquired, and the dynamic population density data is mounted to each grid cell of the multi-height layer H3 hexagonal spatial grid through the spatial interpolation method to generate a time-varying population density sequence. Based on time-varying population density sequences and various surface functional zones, as well as the acquired dynamic population data, the population distribution patterns are extracted to obtain the indoor and outdoor population. Based on the obtained international unmanned aerial vehicle (UAV) flight safety assessment standards, UAV type parameters, and flight altitude h, the critical impact area on the ground after the UAV crashes is obtained. The ground shading coefficient is obtained based on ground functional zones and building form indicators; Population exposure risk is obtained based on the area ratio of ground functional zones, indoor population, outdoor population, ground shielding coefficient, and area of ​​the ground critical impact zone; Based on population exposure risk, risk attribute sequences are calculated for each preset time period; and risk values ​​are obtained based on the risk attribute sequences.

4. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 1, characterized in that, The passage indicators are obtained based on the multi-height layer H3 hexagonal spatial grid and the preset safety buffer distance, including: Based on the multi-height H3 hexagonal spatial grid, multiple directional connections are obtained for each grid cell; According to the preset safety buffer distance, spatial expansion obstacles are screened in the multi-height layer H3 hexagonal spatial grid to obtain the effective obstacle expansion area; the number of intersections between the lines in each direction and the effective obstacle expansion area is counted to obtain the collision count; Based on the number of collisions, the airspace traversability index is obtained; Calculate the airspace mobility index for all altitude layers and the safe buffer distance, and generate a sequence of mobility attributes.

5. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 1, characterized in that, The method involves adaptively merging multi-height-layer H3 hexagonal spatial grids based on risk and accessibility indicators to obtain multi-level spatial grids, including: S41. Construct candidate merging clusters based on the multi-height layer H3 hexagonal spatial grid; S42. Based on the candidate merge clusters and the corresponding risk attribute sequence and commonality attribute sequence, construct an attribute matrix and perform row-level standardization on the attribute matrix; S43. Based on the row-level standardized attribute matrix, obtain the single-attribute heterogeneity measure; S44. Based on the single-attribute heterogeneity measure, nonlinear fusion is performed using the Minkowski norm to obtain the dual-attribute sequence similarity index. S45. Determine whether the similarity index of the dual-attribute sequence is less than the preset merging threshold, whether the maximum value of the risk value is less than the preset high-risk threshold, and whether the minimum value of the spatial traversability index is greater than the preset low traversability threshold. If so, adopt a bottom-up aggregation mechanism to adaptively merge adjacent grid cells within the candidate merging cluster; otherwise, merging is prohibited. S46. Repeat steps S41 to S46 until no new merges are generated or the preset number of iterations is reached, to obtain the merged multi-level spatial grid.

6. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 5, characterized in that, The formula for calculating the dual-attribute sequence similarity index is: ， in, It is a two-attribute sequence similarity index. The weights of the risk attributes, Here, p represents the weight of the commutability attribute, and p is the Minkowski parameter. As a measure of heterogeneity of risk attributes, It is a measure of heterogeneity of commonality attributes.

7. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 1, characterized in that, The process of merging and evaluating the quality of multi-level airspace grids to generate a grid-level low-altitude airspace digital model includes: Structural similarity is evaluated for multi-level spatial grids to obtain a structural similarity index; An extremum preservation index is obtained by evaluating the extremum preservation of multi-level spatial grids. Determine whether the structural similarity index is greater than the structural similarity threshold and whether the extreme value preservation index is greater than the extreme value preservation threshold. If so, generate a grid-level low-altitude airspace digital model; otherwise, return to S4.

8. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 7, characterized in that, The formula for calculating the structural similarity index is: ， in, Let X be the structural similarity index, and Y be the attribute spatial distribution matrix before merging and Y be the attribute spatial distribution matrix after merging. The mean before the merger. This is the combined mean. The variance before merging. The variance after aggregation. For covariance, and All of these are set stability constants.

9. The method for digital modeling of urban low-altitude airspace based on H3 grids according to claim 7, characterized in that, The formula for calculating the extreme value preservation index is: ， in, To preserve the extremum, This is a set of extreme value grids identified based on a preset percentile threshold before merging. This is a set of extreme value grids identified after merging based on a preset percentile threshold. For the symmetric difference of sets, Indicates the number of elements in the set. This is the attenuation parameter.

10. A digital modeling system for urban low-altitude airspace based on H3 grids, characterized in that, A digital modeling method for urban low-altitude airspace based on H3 grids for any one of claims 1-9 includes: The H3 mesh building module is used to construct multi-height-layer H3 hexagonal spatial meshes. The risk value generation module is used to obtain risk indicators based on a multi-height layer H3 hexagonal spatial grid. The risk indicators include risk values ​​and risk attribute sequences. The airspace accessibility index generation module is used to obtain accessibility indicators based on the multi-height layer H3 hexagonal airspace grid and the preset safety buffer distance. The accessibility indicators include airspace accessibility index and accessibility attribute sequence. The grid merging module is used to adaptively merge multi-height H3 hexagonal spatial grids based on risk indicators and accessibility indicators to obtain multi-level spatial grids. The digital model generation module is used to perform quality assessment on multi-level airspace grids and generate grid-level low-altitude airspace digital models.

Images