Method for Regionally Reconstructing Digital Elevation Model of Highway Slope Based on UAV

Abstract

The present invention discloses a method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle, which includes: First, rasterize the point cloud data of the reconstruction area, and determine the feature coding corresponding to the raster according to the terrain features within the area corresponding to the raster, so as to generate a global feature coding raster of the reconstruction area. Then, classify all the point clouds covered by all the rasters with the same feature coding in the feature coding raster into the same point cloud subset, and form point cloud subsets corresponding to different feature codings. After that, perform interpolation processing on the point cloud subsets corresponding to each feature coding by using the optimal interpolation parameters matching the feature coding, and generate the corresponding local digital elevation model. Finally, merge all the local digital elevation models to obtain the digital elevation model of the reconstruction area. It realizes the independent and optimized interpolation of different regions in the reconstruction area, solves the limitations of the traditional global single interpolation method for DEM reconstruction, and significantly improves the terrain reconstruction accuracy and the ability to restore details.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle remote sensing data processing, and particularly relates to a method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle. Background Art

[0002] Existing digital elevation model interpolation methods (such as TIN, IDW, Kriging, etc.) have obvious limitations in complex terrain areas. The global single interpolation method is difficult to adapt to the terrain feature differences in flat areas, gentle slope areas, and steep slope areas of highway slopes. For example, there are elevation mutations at the junction of steep slopes and flat areas, resulting in terrain continuity fractures and loss of terrain details. Traditional hybrid interpolation, on the other hand, has the problem of fixed parameters and lack of a dynamic adjustment mechanism, resulting in insufficient matching between terrain partitions and interpolation methods, affecting the reconstruction accuracy and detail restoration ability of the digital elevation model. Summary of the Invention

[0003] In view of the deficiencies of the existing technology, the present invention proposes a method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle, which can improve the reconstruction accuracy and detail restoration ability of the digital elevation model. The specific technical solutions are as follows:

[0004] A method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle is provided. In a first realizable manner, it includes:

[0005] Dividing the point cloud data of the reconstruction area into grids, and determining the feature code corresponding to each grid to form a global feature code grid;

[0006] Segmenting the point cloud data based on the feature code grid to generate point cloud subsets corresponding to each feature code;

[0007] Interpolating each point cloud subset according to the optimal interpolation parameters matched by the feature code, and generating corresponding local digital elevation models;

[0008] Merging all the local digital elevation models to generate a digital elevation model of the reconstruction area.

[0009] Combined with the first realizable manner, in a second realizable manner, determining the feature code corresponding to each grid includes:

[0010] Classifying the feature categories of each grid cell according to different classification criteria respectively to generate corresponding feature classification mask grids;

[0011] Overlaying all the feature classification mask grids, and combining all the feature categories corresponding to each grid cell to generate corresponding feature codes.

[0012] Combined with the first implementation method, in the third implementation method, segmenting the point cloud data based on the feature-encoded grid includes:

[0013] Merging the geographical ranges corresponding to all grid cells with the same feature encoding in the feature-encoded grid into corresponding encoded regions, and dividing the point clouds located within the encoded regions into the same point cloud subset.

[0014] Combined with the third implementation method, in the fourth implementation method, merging the geographical ranges corresponding to all grid cells with the same feature encoding into an encoded region includes:

[0015] Converting the geographical ranges of each grid cell with the same feature encoding into geographical polygons respectively;

[0016] Using a spatial merging algorithm to merge the scattered geographical polygons into an encoded region.

[0017] Combined with the third implementation method, in the fifth implementation method, segmenting the point cloud data based on the feature-encoded grid includes:

[0018] When the point cloud is at the junction of multiple encoded regions, dividing the point cloud subset to which the point cloud belongs according to the feature encoding of the grid cell to which the point cloud belongs;

[0019] When the number of point clouds in the encoded region is 0, recording the point cloud subset as an empty set, and using default parameters or the optimal interpolation parameters matching the point cloud subsets of adjacent encoded regions for interpolation during interpolation;

[0020] When the number of point clouds in the encoded region is less than the quantity threshold, merging the point cloud subset with the point cloud subsets of adjacent encoded regions.

[0021] Combined with the first implementation method, in the sixth implementation method, interpolating the point cloud subset according to the optimal interpolation parameters matched by the feature encoding includes:

[0022] Defining a parameter mapping table between the feature encoding and the optimal interpolation parameters;

[0023] Matching the feature encoding corresponding to the point cloud subset with the predefined parameter mapping table to determine the optimal interpolation parameters matched by the feature encoding.

[0024] Combined with the sixth implementation method, in the seventh implementation method, defining a parameter mapping table between the feature encoding and the optimal interpolation parameters includes:

[0025] Respectively obtaining the point cloud data corresponding to different feature encodings to construct corresponding sample data sets;

[0026] Interpolate each sample dataset using different interpolation parameters respectively, and evaluate the interpolation performance of each sample dataset with different interpolation parameters;

[0027] Compare the interpolation performances of different interpolation parameters corresponding to each sample dataset respectively, determine the optimal interpolation parameters for each sample dataset, and generate a parameter mapping table between the feature encoding and the optimal interpolation parameters.

[0028] Combined with the seventh implementation manner, in the eighth implementation manner, compare the interpolation performances of different interpolation parameters corresponding to the sample dataset, including:

[0029] Calculate various interpolation performance indexes of different interpolation parameters respectively according to the interpolation results obtained after interpolating the sample dataset with different interpolation parameters;

[0030] Calculate the performance quantization index corresponding to each interpolation parameter respectively according to the corresponding various interpolation performance indexes;

[0031] Compare the performance quantization indexes corresponding to each interpolation parameter, and determine the optimal interpolation parameters of the sample dataset.

[0032] Combined with the first implementation manner, in the ninth implementation manner, generate a digital elevation model of the reconstruction area, including:

[0033] Perform intersection detection on the digital elevation model to determine the intersection area in the digital elevation model;

[0034] Convert the intersection area into a binary raster mask, and determine the overlapping raster cells in the binary raster mask;

[0035] Obtain the Z values of the overlapping raster cells in each local digital elevation model, and calculate the fused Z value of the overlapping raster cells in combination with the weights corresponding to each local digital elevation model;

[0036] Update the Z value of the corresponding area in the digital elevation model according to the fused Z value.

[0037] Combined with the ninth implementation manner, in the tenth implementation manner, determine the weight corresponding to the local digital elevation model according to the interpolation accuracy of the corresponding point cloud subset.

[0038] Beneficial effects: By using the method for partitioning and reconstructing the digital elevation model of highway slopes based on drones according to the present invention, the point cloud data of the reconstruction area can be partitioned according to the terrain features, the point cloud subsets corresponding to different terrain feature areas can be determined, and then the optimal interpolation parameters matching the terrain features can be used to perform interpolation processing on each point cloud subset, realizing independent optimized interpolation for each partition, thereby solving the limitations of the traditional global single interpolation method for DEM reconstruction and significantly improving the terrain reconstruction accuracy and the ability to restore details. Brief Description of the Drawings

[0039] In order to more clearly illustrate the specific embodiments of the present invention, the drawings required for the specific embodiments will be briefly introduced below. In all the drawings, the elements or parts are not necessarily drawn to actual scale.

[0040] Figure 1 It is a flowchart of a method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle provided by an embodiment of the present invention. Specific Embodiments

[0041] The embodiments of the technical solution of the present invention will be described in detail below with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, so they are only examples and cannot be used to limit the protection scope of the present invention.

[0042] As Figure 1 shown in the flowchart of the method for reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle, the reconstruction method includes:

[0043] Step 1: Perform grid division on the point cloud data of the reconstruction area, and determine the feature code corresponding to each grid to form a global feature code grid.

[0044] Step 2: Segment the point cloud data based on the feature code grid to generate point cloud subsets corresponding to each feature code.

[0045] Step 3: Interpolate each point cloud subset according to the optimal interpolation parameters matched by the feature code, and generate corresponding local digital elevation models.

[0046] Step 4: Merge all the local digital elevation models to generate a digital elevation model of the reconstruction area.

[0047] Specifically, first, the point cloud data of the reconstruction area can be grid-divided, and the feature code corresponding to the grid can be determined according to the terrain features within the area range corresponding to each grid to generate a global feature code grid of the reconstruction area. Then, all the point clouds covered by all the grids belonging to the same feature code in the feature code grid can be grouped into the same point cloud subset, so as to form point cloud subsets corresponding to different feature codes. After that, the optimal interpolation parameters matched with the feature code can be used to perform interpolation processing on the point cloud subsets corresponding to each feature code and generate corresponding local digital elevation models. Finally, all the local digital elevation models are merged to obtain the digital elevation model of the reconstruction area. In this way, the partition-independent optimal interpolation of the reconstruction area can be realized, thus solving the limitations of the traditional global single interpolation method for DEM reconstruction and significantly improving the terrain reconstruction accuracy and detail restoration ability.

[0048] In this embodiment, optionally, determining the feature encoding corresponding to each grid includes:

[0049] Classifying the feature categories of each grid cell according to different classification criteria respectively to generate corresponding feature classification mask grids;

[0050] Overlaying all the feature classification mask grids and combining all the feature categories corresponding to each grid cell to generate corresponding feature encodings.

[0051] Specifically, in step 1, first, the feature categories of each grid can be classified according to different feature classification criteria to generate corresponding feature classification mask grids. For example, the point cloud data in the reconstruction area can be divided into terrain slope grids and point cloud density grids according to the same spatial resolution first. The spatial resolution is consistent with the digital elevation model. Then, a terrain partition discrimination function can be used to classify each grid in the terrain slope grid and mark the terrain category to which each grid belongs, so as to form a terrain mask grid. The terrain partition discrimination function can be set as:

[0052] ;

[0053] Among them, is the slope angle corresponding to the grid. The flat area emphasizes smoothness and efficiency, the gentle slope area balances continuity and details, and the steep slope area gives priority to retaining mutation features.

[0054] Similarly, the number of point clouds in each grid in the point cloud density grid can be counted, and each grid can be classified according to the set density partition standard and the density category of each grid can be marked, so as to form a density mask grid. The specific classification standard is:

[0055] ;

[0056] Among them, D is the number of point clouds in the grid.

[0057] Then, the density mask grid and the terrain mask grid can be overlaid, and the density category and the terrain category corresponding to the overlapping grids can be combined to form the feature encoding of the grid. The formed feature encoding situation is as follows:

[0058]

[0059] In this embodiment, optionally, segmenting the point cloud data based on the feature encoding grid includes:

[0060] Merging the geographical ranges corresponding to all the grid cells with the same feature encoding in the feature encoding grid into corresponding encoded regions, and dividing the point clouds located in the encoded regions into the same point cloud subset.

[0061] Specifically, first, according to the geographical coordinates of each grid respectively, the geographical ranges corresponding to each grid cell with the same feature code in the feature-coded grid can be converted into geographical polygons. Then, an existing spatial merging algorithm is used to merge the geographical polygons of adjacent grids, so as to merge the scattered geographical polygons into coded regions. Finally, the point clouds falling within the coded region are retrieved, and all the retrieved point clouds are divided into subsets of point clouds corresponding to the same feature code, thereby realizing the topographic feature partitioning of the point clouds in the reconstruction region and providing a basis for subsequent independent optimization and interpolation of each partition.

[0062] In this embodiment, optionally, segmenting the point cloud data based on the feature-coded grid includes:

[0063] When the point cloud is located at the junction of coded regions with multiple different feature codes, the point cloud subset to which the point cloud belongs is divided according to the feature code of the grid cell to which the point cloud belongs;

[0064] When the number of point clouds in the coded region is 0, the point cloud subset is recorded as an empty set, and when interpolating, the default parameters or the optimal interpolation parameters matching the point cloud subset of the adjacent coded region are used for interpolation;

[0065] When the number of grids in the coded region is less than the quantity threshold, the point cloud subset is merged with the point cloud subset of the adjacent coded region.

[0066] Specifically, when dividing the point cloud, it is possible that the point cloud is located at the junction of coded regions corresponding to multiple feature codes. For this situation, the grid to which the point cloud belongs can be determined according to the geographical coordinates of the point cloud, and the point cloud subset to which the point cloud belongs can be determined according to the feature code corresponding to the grid.

[0067] It is also possible that the number of point clouds in the coded region is 0, such as water areas. For this situation, the point cloud subset corresponding to this region can be recorded as an empty set. When interpolating the empty set, the optimal interpolation parameters corresponding to the feature codes of the adjacent regions of this region can be referred to for interpolation, or the default parameters can be used.

[0068] It is also possible that the number of grids in the coded region is small, only 1 or 2 grids. For this situation, these two grids can be merged into other adjacent coded regions, and the point clouds in the grids can be incorporated into the point cloud subsets corresponding to the adjacent coded regions, or merged into the point cloud subsets with similar feature codes, so as to avoid over-parameterization of interpolation.

[0069] In this embodiment, optionally, interpolating the point cloud subset according to the optimal interpolation parameters matched by the feature code includes:

[0070] Define a parameter mapping table between the feature code and the optimal interpolation parameter;

[0071] Match the feature encoding corresponding to the point cloud subset with a predefined parameter mapping table to determine the optimal interpolation parameters matched by the feature encoding.

[0072] Specifically, first, it is necessary to determine the optimal interpolation parameters matched by different feature encodings, determine the mapping relationship between the feature encoding and the optimal interpolation parameters, and form a corresponding parameter mapping table. Then, the feature encodings corresponding to each point cloud subset can be matched with the parameter mapping table to determine the optimal interpolation parameters corresponding to each point cloud subset. The optimal interpolation parameters include the interpolation method most suitable for the terrain feature corresponding to the feature encoding and the parameter values corresponding to the various parameters required by the interpolation method.

[0073] In this embodiment, optionally, define a parameter mapping table between the feature encoding and the optimal interpolation parameters, including:

[0074] Obtain the point cloud data corresponding to different feature encodings respectively to construct corresponding sample data sets;

[0075] Interpolate each sample data set with different interpolation parameters respectively, and evaluate the interpolation performance of each sample data set with different interpolation parameters;

[0076] Compare the interpolation performances of different interpolation parameters corresponding to each sample data set respectively, determine the optimal interpolation parameters of each sample data set, and generate a parameter mapping table between the feature encoding and the optimal interpolation parameters.

[0077] Specifically, when determining the mapping relationship between the feature encoding and the optimal interpolation parameters, first, the point cloud data corresponding to different terrain features can be obtained respectively, and the sample data sets corresponding to the feature encoding can be constructed respectively. Then, the sample data is interpolated with different interpolation parameters respectively, and the interpolation performance of each interpolation parameter for the sample data set is evaluated. Finally, the interpolation performances of different interpolation parameters for the sample data set are compared, and the interpolation parameter with the best interpolation performance is selected as the optimal interpolation parameter matched with the feature encoding. Repeat this process to obtain the optimal interpolation parameters corresponding to different encoded features and form a parameter mapping table between the feature encoding and the optimal interpolation parameters.

[0078] In this embodiment, optionally, comparing the interpolation performances of different interpolation parameters corresponding to the sample data set includes:

[0079] Calculate the various interpolation performance indexes of different interpolation parameters respectively according to the interpolation results obtained by interpolating the sample data set with different interpolation parameters;

[0080] Calculate the performance quantization index corresponding to each interpolation parameter respectively according to the corresponding various interpolation performance indexes;

[0081] Compare the performance quantization indexes corresponding to each interpolation parameter to determine the optimal interpolation parameter for the sample data set.

[0082] Specifically, when comparing the interpolation performances of different interpolation parameters, the interpolation performance indexes of the interpolation parameters can be calculated first according to the interpolation results of the sample data under the action of the interpolation parameters. The interpolation performance indexes can include MAE, RMSE, and R². Among them, R² can be used to evaluate the fitting degree between the interpolation surface and the true terrain, RMSE can emphasize the penalty for large errors, and MAE reflects the average absolute deviation between the interpolation result and the true value. To ensure the accuracy of the evaluation results, it can be repeated multiple times and the average value of each interpolation performance index can be taken.

[0083] To quantify the interpolation performance with the same dimension, interval normalization can be performed on MAE, RMSE, and R², and the performance quantization index can be calculated using the following calculation formula:

[0084] ;

[0085] Among them, , , are the weight coefficients corresponding to each interpolation performance index respectively, , , are the maximum values of each interpolation performance index corresponding to all data sample sets under the current feature encoding respectively. is the set of interpolation parameters corresponding to the current feature encoding.

[0086] Finally, compare the performance quantization indexes of different interpolation parameters for the sample data set, and select the interpolation parameter with the largest performance quantization index as the optimal interpolation parameter matching the feature encoding. Repeat this process to obtain the optimal interpolation parameters corresponding to different encoded features. Based on the currently collected sample data, the parameter mapping table between the formed feature encoding and the optimal interpolation parameter is as shown in the following table:

[0087]

[0088] It should be understood that the parameters recorded in the above table are only the evaluation results based on the currently collected sample data, and do not represent that the optimal interpolation parameters and methods corresponding to each feature encoding are as shown in the above table. With the increase of sample data, the optimal interpolation parameters and methods corresponding to each feature encoding in the above table may also change.

[0089] In this embodiment, optionally, generating the digital elevation model of the reconstruction area includes:

[0090] Perform intersection detection on the digital elevation model to determine the intersection area in the digital elevation model;

[0091] Convert the intersection area into a binary raster mask to determine the overlapping raster cells in the binary raster mask;

[0092] Obtain the Z values of the overlapping raster cells in each local digital elevation model, and combine the weights corresponding to each local digital elevation model to calculate the fused Z value of the overlapping raster cells;

[0093] Update the Z value of the corresponding area in the digital elevation model according to the fused Z value.

[0094] Specifically, after determining the optimal interpolation parameters for each point cloud subset matching, a distributed computing framework can be used to perform interpolation optimization processing on each point cloud subset in parallel to improve data processing efficiency. Then, generate the corresponding local digital elevation models according to the optimized point cloud subsets. Finally, preliminarily merge all the local digital elevation models according to their spatial positions to form an initial global digital elevation model.

[0095] Since there may be intersecting parts between local digital elevation models, the initial global digital elevation model generated by merging has an intersection area. Therefore, a topological retrieval tool can be used to perform intersection detection on the initial global digital elevation model to determine the intersection area in the initial global digital elevation model. Then, convert the intersection area into a binary raster mask, where the raster value of 1 represents the overlapping area of the local digital elevation models, and 0 represents the non-overlapping area. For the overlapping raster cells marked as 1 in the binary raster mask, all the local digital elevation models covering this raster can be determined through a spatial index or a hash table, and according to the Z values of this raster in each local digital elevation model, combined with the weights corresponding to the local digital elevation models, calculate the fused Z value corresponding to this overlapping raster cell. The specific calculation formula is as follows:

[0096] ;

[0097] Wherein, is the Z value corresponding to the overlapping raster cell in the local digital elevation model, is the weight corresponding to the local digital elevation model, is the number of local digital elevation models covering the overlapping raster cell. Finally, the fused Z value corresponding to the overlapping raster cell can be written into the initial global digital elevation model to obtain the final digital elevation model.

[0098] In this embodiment, optionally, determine the weight corresponding to the local digital elevation model according to the interpolation accuracy of the corresponding point cloud subset.

[0099] Specifically, the weight corresponding to the local digital elevation model can be set according to the interpolation accuracy of the point cloud subset for constructing the local digital elevation model. The higher the interpolation accuracy, the greater the weight corresponding to the local digital elevation model. The specific calculation formula for the weight is as follows:

[0100] ;

[0101] wherein, is the loop variable in the summation process, representing the traversal from the first local digital elevation model (k = 1) to the th local digital elevation model (k = n).

[0102] Since there may be multiple Z values for the same grid cell, to improve the accuracy of the reconstructed digital elevation model, it is necessary to select the Z value with the highest interpolation accuracy, or perform weighted summation according to the interpolation accuracy weights. The higher the interpolation accuracy, the greater the weight, and the lower the interpolation accuracy, the smaller the weight. Only in this way can the accuracy of the fused Z value be ensured.

[0103] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. A method for reconstructing a digital elevation model partition of a highway slope based on an unmanned aerial vehicle, characterized in that Including: Performing grid division on the point cloud data of the reconstruction area, classifying the feature categories of each grid cell according to different classification criteria respectively, and generating corresponding feature classification mask grids; Overlaying all the feature classification mask grids, combining all the feature categories corresponding to each grid cell, generating corresponding feature codes, and forming a global feature code grid; Merging the geographical ranges corresponding to all the grid cells with the same feature code in the feature code grid into corresponding coded areas, and dividing the point clouds located within the coded areas into the same point cloud subsets, generating point cloud subsets corresponding to each feature code; Interpolating each point cloud subset according to the optimal interpolation parameters matched by the feature code, and generating corresponding local digital elevation models; Merging all the local digital elevation models to generate a digital elevation model of the reconstruction area.

2. The method for regionally reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle according to claim 1, wherein Merging the geographical ranges corresponding to all the grid cells with the same feature code into a coded area, including: Converting the geographical ranges of each grid cell with the same feature code into geographical polygons respectively; Using a spatial merging algorithm to merge the scattered geographical polygons into a coded area.

3. The method for regionally reconstructing a digital elevation model of a highway slope based on an unmanned aerial vehicle according to claim 1, wherein Segmenting the point cloud data based on the feature code grid, including: When the point cloud is located at the junction of multiple coded areas, dividing the point cloud subset to which the point cloud belongs according to the feature code of the grid cell to which the point cloud belongs; When the number of point clouds in the coded area is 0, recording the point cloud subset as an empty set, and using default parameters or the optimal interpolation parameters matched by the point cloud subsets of adjacent coded areas for interpolation during interpolation; When the number of point clouds in the coded area is less than the quantity threshold, merging the point cloud subset with the point cloud subsets of adjacent coded areas.

4. The method for reconstructing a digital elevation model partition of a highway slope based on an unmanned aerial vehicle according to claim 1, wherein Interpolating the point cloud subset according to the optimal interpolation parameters matched by the feature code, including: Defining a parameter mapping table between the feature code and the optimal interpolation parameters; Matching the feature code corresponding to the point cloud subset with the predefined parameter mapping table to determine the optimal interpolation parameters matched by the feature code.

5. The method for reconstructing the digital elevation model partition of the highway slope based on the drone according to claim 4, wherein, Defining a parameter mapping table between the feature code and the optimal interpolation parameters, including: Respectively obtaining the point cloud data corresponding to different feature codes to construct corresponding sample data sets; Interpolating each sample data set with different interpolation parameters respectively, and evaluating the interpolation performance of each sample data set with different interpolation parameters; Comparing the interpolation performances of different interpolation parameters corresponding to each sample data set respectively, determining the optimal interpolation parameters of each sample data set, and generating a parameter mapping table between the feature code and the optimal interpolation parameters.

6. The method for reconstructing the digital elevation model partition of the highway slope based on the unmanned aerial vehicle according to claim 5, wherein Comparing the interpolation performances of different interpolation parameters corresponding to the sample data set, including: Respectively calculating various interpolation performance indexes of different interpolation parameters according to the interpolation results obtained by interpolating the sample data set with different interpolation parameters; Calculating the performance quantization indexes corresponding to each interpolation parameter respectively according to the corresponding various interpolation performance indexes; Comparing the performance quantization indexes corresponding to each interpolation parameter to determine the optimal interpolation parameters of the sample data set.

7. The method for reconstructing a digital elevation model partition of a highway slope based on an unmanned aerial vehicle according to claim 1, wherein Generating a digital elevation model of the reconstruction area, including: Performing intersection detection on the digital elevation model to determine the intersection area in the digital elevation model; Convert the intersection area into a binary raster mask and determine the overlapping raster cells in the binary raster mask; Obtain the Z values of the overlapping raster cells in each local digital elevation model, and calculate the fused Z value of the overlapping raster cells by combining the weights corresponding to each local digital elevation model; Update the Z value of the corresponding area in the digital elevation model according to the fused Z value.

8. The method for reconstructing a digital elevation model partition of a highway slope based on an unmanned aerial vehicle according to claim 7, characterized in that Determine the weight corresponding to the local digital elevation model according to the interpolation accuracy of the corresponding point cloud subset.

Images