Original ground line extraction method based on inverse distance weighted interpolation

Abstract

The invention discloses an original ground line extraction method based on inverse distance weighted interpolation, and relates to the technical field of road engineering topographic mapping. The method comprises the following steps: acquiring airborne LiDAR original point cloud data and CAD design line data of a target road area; statistical filtering denoising, voxel grid downsampling simplification and cloth simulation filtering ground point extraction are carried out in sequence; generating cross section line segments based on CAD design line data, and obtaining a cross section point set; an inverse distance weighted interpolation algorithm for dynamic weight index adjustment is adopted, the cross section point set serves as a known point, and a continuous original ground line is generated through interpolation; and evaluating the extraction precision through a cross validation method, and adjusting a parameter optimization result. According to the method, through a systematic preprocessing process of statistical filtering, voxel down-sampling and CSF ground point extraction, the point cloud data volume is reduced, point cloud preprocessing and IDW interpolation are combined, efficient operation of more than ten kilometers of measurement within one day is realized, the rapid surveying and mapping requirements of large-scale engineering are met, and the ground line extraction efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of road engineering topographic surveying technology, and in particular relates to a method for extracting original ground lines based on inverse distance weighted interpolation (IDW), which achieves efficient, high-precision and widely adaptable extraction of original ground lines to meet the needs of large-scale road engineering topographic surveying. Background Technology

[0002] In road engineering design, the extraction of the original ground line is fundamental for earthwork volume calculation and longitudinal profile design. Traditional methods largely rely on RTK (Real-Time Kinematic) surveying technology, which, while highly accurate, is inefficient, requiring an average of one day per kilometer, making it difficult to meet the needs of rapid measurement in large-scale projects. In recent years, airborne LiDAR technology has been increasingly applied to topographic mapping, acquiring surface elevation information from point cloud data, offering advantages such as high efficiency and wide coverage. However, accurately and efficiently extracting continuous original ground lines from discrete point clouds remains a technical challenge. Existing methods for extracting original ground lines mainly suffer from the following problems:

[0003] (1) Inefficient: Traditional RTK measurement methods are time-consuming and labor-intensive, and cannot meet the needs of rapid construction;

[0004] (2) Insufficient interpolation accuracy: Existing interpolation algorithms have large errors in areas with complex terrain (such as mountains and hills), making it difficult to accurately reflect the characteristics of abrupt changes in terrain;

[0005] (3) Incomplete point cloud processing flow: The lack of a systematic point cloud preprocessing flow leads to large noise interference and high data redundancy, which affects the final extraction accuracy.

[0006] (4) Poor adaptability: Existing methods are sensitive to changes in terrain complexity and are difficult to maintain stable accuracy under different landform conditions.

[0007] Therefore, there is an urgent need for an efficient, high-precision, and highly adaptable method for extracting original ground lines to meet the needs of modern road engineering for rapid and accurate terrain modeling. Summary of the Invention

[0008] The purpose of this invention is to provide a method for extracting original ground lines based on inverse distance weighted interpolation. Through a systematic preprocessing process of statistical filtering, voxel downsampling, and CSF ground point extraction, the amount of point cloud data is reduced. By combining point cloud preprocessing with IDW interpolation, efficient operation of measuring more than ten kilometers in a day can be achieved, meeting the needs of rapid surveying for large-scale engineering projects. This solves the problems of low efficiency, poor accuracy, and weak adaptability of existing ground line extraction methods.

[0009] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0010] This invention relates to a method for extracting original ground lines based on inverse distance weighted interpolation, comprising the following steps:

[0011] Step S1, Data Acquisition: Acquire raw point cloud data from airborne LiDAR and CAD design line data for the target road area;

[0012] Step S2, Point Cloud Data Preprocessing: Statistical filtering for noise reduction, voxel mesh downsampling for simplification, and cloth simulation filtering for ground point extraction are performed sequentially.

[0013] Step S3, Cross-section data extraction: Generate cross-section line segments based on CAD design line data and obtain cross-section point sets;

[0014] Step S4: Generate original ground line by inverse distance weighted interpolation: Using the inverse distance weighted interpolation algorithm with dynamic weighted index adjustment, a continuous original ground line is generated by interpolation with the cross-section point set as known points.

[0015] Step S5, Accuracy Verification and Parameter Optimization: Evaluate the extraction accuracy using cross-validation and adjust the parameters to optimize the results;

[0016] Step S6: Data Export: Export the original ground line data according to the engineering standard format.

[0017] As a preferred technical solution, in step S1, the airborne LiDAR raw point cloud data and the corresponding CAD design line data of the target road area are acquired; wherein, the airborne LiDAR raw point cloud data includes ground elevation information, and the CAD design line data includes design parameter information such as road baseline and design cross section.

[0018] As a preferred technical solution, in step S2, statistical filtering denoising uses a statistical filtering algorithm to process the raw point cloud data of the airborne LiDAR. It calculates the average distance and standard deviation of points within a fixed neighborhood around each point cloud, sets a distance threshold as the sum of the average distance and three times the standard deviation, and identifies points with a distance greater than this threshold as noise points and removes them. The formula is as follows:

[0019] ;

[0020] In the formula, Distance threshold The average distance within the neighborhood of the point cloud. The distance is the standard deviation; if point cloud Average distance to all points in its neighborhood ,but These are noise points and should be removed.

[0021] As a preferred technical solution, in step S2, the voxel grid downsampling simplification involves constructing a three-dimensional voxel grid, dividing the preprocessed point cloud data into various voxel units, retaining a representative point within each voxel unit (this point is the geometric center of all points within the voxel unit), and eliminating redundant point clouds. This reduces the data volume while ensuring that the geometric features of the point cloud are not lost. The voxel grid side length s is adaptively set according to the point cloud density, as shown in the following formula:

[0022] ;

[0023] In the formula, The bounding box volume of the point cloud data. This represents the total number of point clouds after statistical filtering.

[0024] As a preferred technical solution, in step S2, the fabric simulation filtering ground point extraction treats the point cloud data as particles in three-dimensional space, simulating the process of a virtual piece of fabric naturally sinking to the ground under the action of gravity, by setting a time step. Number of iterations The parameters are used to make the virtual cloth conform to the shape of the ground, and finally the point cloud within a certain distance below the cloth is identified as the ground point cloud. The formula is as follows:

[0025] ;

[0026] In the formula, For the fabric after settling The elevation of the location This is the initial elevation of the fabric. For point clouds, the acceleration due to gravity is considered. satisfy , If P is the threshold for determining ground points, then P is a ground point.

[0027] As a preferred technical solution, the specific process for extracting cross-sectional data in step S3 is as follows:

[0028] Step S31: Import CAD design line data and set the sampling interval according to road condition design requirements. and cross-sectional width ;

[0029] Step S32: Generate samples uniformly along the baseline of the CAD design line. Calculate the tangent direction of the baseline at each sampling point, and then obtain the normal direction of the cross section. ;

[0030] Step S33: Using sampling points Extending outwards from the center along the normal direction Distance, generating cross-sectional line segments;

[0031] Step S34: Sample at equal intervals along the cross-sectional line segment, with the sampling interval being... Obtain the cross-sectional point set Each point contains three-dimensional coordinate information. .

[0032] As a preferred technical solution, the specific process of generating the original ground line by inverse distance weighted interpolation in step S4 is as follows:

[0033] Step S41: Using the cross-sectional points obtained in step S3 Using the points in the middle as known elevation points, interpolation calculations are performed on all positions to be interpolated on the cross-sectional line segment;

[0034] Step S42: An improved inverse distance weighted interpolation algorithm is adopted, and a dynamic weight index adjustment mechanism is introduced. The formula is as follows: (Adaptive adjustment based on terrain slope)

[0035] ;

[0036] In the formula, The average slope of the triangle formed by the three nearest known points around the point to be interpolated, with a range of values. When the terrain is flat ( )hour, To maintain the stability of the classic IDW algorithm; when the terrain is steep ( )hour, Enhance the weight of nearby known points to improve interpolation accuracy in areas of abrupt terrain changes;

[0037] Step S43: The distance decay function adopts the inverse square of the distance form, and a distance threshold is introduced at the same time. When the distance between the known point and the point to be interpolated exceeds In this case, the known point is not included in the interpolation calculation to suppress long-distance noise interference. The interpolation formula is as follows:

[0038] ;

[0039] In the formula, Points to be interpolated elevation, The number of known points participating in the interpolation ( ), For the point to be interpolated and the first The Euclidean distance between known points, and , For the first The elevation of a known point The weighted index is dynamically adjusted.

[0040] Step S44: Smoothly stitch together the interpolation results of all cross sections in order of road mileage to form a continuous original ground line of the target road.

[0041] As a preferred technical solution, in step S5, the point set of each cross-section is... The data is randomly divided into a training set and a validation set. The interpolated elevation corresponding to the validation set is calculated using the interpolation algorithm in step S4 using the training set. This interpolated elevation is then compared with the actual elevation of the validation set to calculate the root mean square error (RMSE) and mean absolute error (MAE), as shown in the following formulas:

[0042] ;

[0043] ;

[0044] In the formula, To verify the number of collection points, For the verification set The actual elevation of each point The elevation is obtained through interpolation.

[0045] As a preferred technical solution, the root mean square error or Adjust the voxel mesh side length CSF algorithm parameters, interpolation weight index adjustment coefficient or distance threshold ,repeat Continue until the error meets the requirements.

[0046] As a preferred technical solution, in step S6, the optimized original ground line data is exported in an engineering standard format, supporting formats such as DXF, DWG, and CSV, and can be directly imported into road design software for subsequent engineering applications such as earthwork calculation and longitudinal profile design.

[0047] The present invention has the following beneficial effects:

[0048] (1) This invention reduces the amount of point cloud data through a systematic preprocessing process of statistical filtering, voxel downsampling and CSF ground point extraction. Combined with point cloud preprocessing and IDW interpolation, it enables efficient operation of measuring more than ten kilometers in a day, meets the needs of rapid mapping for large-scale projects, and improves the efficiency of ground line extraction.

[0049] (2) This invention can effectively capture the terrain abrupt changes such as steep slopes and side slopes through CSF ground point extraction and IDW interpolation. It is stable in various terrains such as plains, hills and mountains. Voxel downsampling and statistical filtering effectively remove noise and preserve geometric structure. The IDW interpolation parameters are adjustable to adapt to different landform conditions, thus improving the accuracy and adaptability of ground line extraction.

[0050] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0051] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart of a method for extracting original ground lines based on inverse distance weighted interpolation according to the present invention;

[0053] Figure 2 This is a diagram showing the specific site data collection situation in Example 2;

[0054] Figure 3 This is a cross-sectional view of Example 2;

[0055] Figure 4 This is a schematic diagram of the algorithm extraction results in Example 2. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0058] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-4 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0059] Example 1

[0060] Please see Figure 1 As shown, this invention is a method for extracting original ground lines based on inverse distance weighted interpolation, comprising the following steps:

[0061] Step S1, Data Acquisition: Acquire raw point cloud data from airborne LiDAR and CAD design line data for the target road area;

[0062] Step S2, Point Cloud Data Preprocessing: Statistical filtering for noise reduction, voxel mesh downsampling for simplification, and cloth simulation filtering for ground point extraction are performed sequentially.

[0063] Step S3, Cross-section data extraction: Generate cross-section line segments based on CAD design line data and obtain cross-section point sets;

[0064] Step S4: Generate original ground line by inverse distance weighted interpolation: Using the inverse distance weighted interpolation algorithm with dynamic weighted index adjustment, a continuous original ground line is generated by interpolation with the cross-section point set as known points.

[0065] Step S5, Accuracy Verification and Parameter Optimization: Evaluate the extraction accuracy using cross-validation and adjust the parameters to optimize the results;

[0066] Step S6: Data Export: Export the original ground line data according to the engineering standard format.

[0067] In step S1, the airborne LiDAR raw point cloud data and the corresponding CAD design line data of the target road area are obtained; wherein, the airborne LiDAR raw point cloud data includes ground elevation information, and the CAD design line data includes design parameter information such as road baseline and design cross section.

[0068] Specifically, the point cloud density of the raw point cloud data from airborne LiDAR is no less than 8 points / square meter, including the three-dimensional coordinates of the ground surface. Information covering areas extending beyond both sides of the road design boundary. CAD design line data includes design parameters such as road baseline, design mileage, design cross-section width, and slope ratio, and is in DXF or DWG standard engineering format.

[0069] In step S2, statistical filtering denoising uses a statistical filtering algorithm to process the raw point cloud data from the airborne LiDAR. It calculates the average distance and standard deviation of points within a fixed neighborhood around each point cloud. A distance threshold is set as the sum of the average distance and three times the standard deviation. Points with a distance greater than this threshold are identified as noise points and removed. The formula is as follows:

[0070] ;

[0071] In the formula, Distance threshold The average distance within the neighborhood of the point cloud. The distance is the standard deviation; if point cloud Average distance to all points in its neighborhood ,but These are noise points and should be removed.

[0072] Specifically, iterate through the original point cloud data, for each point cloud... Search its radius is ( Range of values Calculate the average distance between all points in the neighborhood of a given point. and standard deviation If you dot the clouds Average distance to all points in its neighborhood Then determine Noise points (such as isolated tree points or bird points) are removed.

[0073] Voxel mesh downsampling simplification involves constructing a 3D voxel mesh to divide the preprocessed point cloud data into individual voxel cells. Each voxel cell retains a representative point (the geometric center of all points within the cell), eliminating redundant point clouds. This reduces the data volume while preserving the geometric features of the point cloud. The voxel mesh side length *s* is adaptively set according to the point cloud density, as shown in the following formula:

[0074] ;

[0075] In the formula, The bounding box volume of the point cloud data. This represents the total number of point clouds after statistical filtering.

[0076] Specifically, the statistically filtered point cloud is divided into voxel units, and a representative point (the geometric center of all points in the voxel unit) is retained in each voxel unit. Redundant point clouds are removed, reducing the amount of data while preserving the geometric features of the terrain.

[0077] In step S2, the cloth simulation filtering ground point extraction treats the point cloud data as particles in three-dimensional space, simulating the process of a virtual cloth naturally sinking to the ground under the action of gravity, by setting the time step. Number of iterations The parameters are used to make the virtual cloth conform to the shape of the ground, and finally the point cloud within a certain distance below the cloth is identified as the ground point cloud. The formula is as follows:

[0078] ;

[0079] In the formula, For the fabric after settling The elevation of the location This is the initial elevation of the fabric. For point clouds, the acceleration due to gravity is considered. satisfy , If the threshold is used for determining ground points, then... For ground points.

[0080] Specifically, the simplified point cloud is treated as particles in three-dimensional space, simulating the process of a virtual piece of cloth naturally sinking to the ground under the influence of gravity. Ground points are separated using the following parameters and formulas:

[0081] Fabric stiffness coefficient : Range of values The more complex the terrain, The smaller the value;

[0082] Time step : Range of values ;

[0083] Number of iterations : Range of values Second-rate;

[0084] Ground point determination: Set ground point determination threshold (range of values) If dotted clouds satisfy Then determine If a point is identified as a ground point, it is considered a non-ground point (such as a building or tree) and is discarded to obtain clean ground point cloud data.

[0085] In step S3, the specific process for extracting cross-sectional data is as follows:

[0086] Step S31: Import CAD design line data and set the sampling interval according to road condition design requirements. and cross-sectional width ;

[0087] Step S32: Generate samples uniformly along the baseline of the CAD design line. Calculate the tangent direction of the baseline at each sampling point, and then obtain the normal direction of the cross section. ;

[0088] Step S33: Using sampling points Extending outwards from the center along the normal direction Distance, generating cross-sectional line segments;

[0089] Step S34: Sample at equal intervals along the cross-sectional line segment, with the sampling interval being... (range of values) ), obtain the cross-sectional point set Each point contains three-dimensional coordinate information. .

[0090] In step S4, the specific process for generating the original ground line using inverse distance weighted interpolation is as follows:

[0091] Step S41: Using the cross-sectional points obtained in step S3 Using the points in the middle as known elevation points, interpolation calculations are performed on all positions to be interpolated on the cross-sectional line segment;

[0092] Step S42: An improved inverse distance weighted interpolation algorithm is adopted, and a dynamic weight index adjustment mechanism is introduced. The formula is as follows: (Adaptive adjustment based on terrain slope)

[0093] ;

[0094] In the formula, The average slope of the triangle formed by the three nearest known points around the point to be interpolated, with a range of values. When the terrain is flat ( )hour, To maintain the stability of the classic IDW algorithm; when the terrain is steep ( )hour, Enhance the weight of nearby known points to improve interpolation accuracy in areas of abrupt terrain changes;

[0095] Step S43: The distance decay function adopts the inverse square of the distance form, and a distance threshold is introduced at the same time. When the distance between the known point and the point to be interpolated exceeds In this case, the known point is not included in the interpolation calculation to suppress long-distance noise interference. The interpolation formula is as follows:

[0096] ;

[0097] In the formula, Points to be interpolated elevation, The number of known points participating in the interpolation ( ), For the point to be interpolated and the first The Euclidean distance between known points, and , For the first The elevation of a known point The weighted index is dynamically adjusted.

[0098] Step S44: Smoothly stitch together the interpolation results of all cross sections in order of road mileage to form a continuous original ground line of the target road.

[0099] In step S5, the point set of each cross-section is... The data is randomly divided into a training set and a validation set. The interpolated elevation corresponding to the validation set is calculated using the interpolation algorithm in step S4 using the training set. This interpolated elevation is then compared with the actual elevation of the validation set to calculate the root mean square error (RMSE) and mean absolute error (MAE), as shown in the following formulas:

[0100] ;

[0101] ;

[0102] In the formula, To verify the number of collection points, For the verification set The actual elevation of each point The elevation is obtained through interpolation.

[0103] Root mean square error or Adjust the voxel mesh side length CSF algorithm parameters ( , , , ), interpolation weight index adjustment coefficient or distance threshold ,repeat until the error meets the requirements ( or ).

[0104] In step S6, the optimized original ground line data is exported in the engineering standard format, supporting formats such as DXF, DWG, and CSV. It can be directly imported into road design software for subsequent engineering applications such as earthwork calculation and longitudinal profile design.

[0105] Example 2

[0106] See Figure 2 , 3 As shown in Figure 4, the technical solution of the present invention will be described in detail below with reference to specific embodiments and accompanying drawings. This embodiment takes a section of a highway from K0+000 to K0+1000 as an example. This section includes three types of terrain: plains, hills, and steep slopes, with a total length of 1 kilometer.

[0107] Step S1, Data Acquisition

[0108] The raw point cloud data from the airborne LiDAR was acquired using the RIEGLVQ-1560i airborne LiDAR device. The point cloud density was 12 points / square meter, and the coverage area was 80m on both sides of the road design red line. The data format was LAS1.4. The CAD design line data was in DWG format and included parameters such as the three-dimensional coordinates of the road baseline, design mileage, cross-section width of 60m, and slope ratio of 1:1.5.

[0109] Step S2: Point cloud data preprocessing

[0110] Statistical filtering for noise reduction: Setting the neighborhood search radius The average distance of a certain point cloud neighborhood is calculated. Standard deviation According to the formula The distance threshold is calculated. Remove distance greater than The noise points were removed, and the pass rate of the point cloud after denoising reached [percentage missing]. ;

[0111] Voxel mesh downsampling simplification: average point spacing of the original point cloud Set the target point cloud density Points per square meter, according to the formula The voxel side length was calculated. A voxel grid was constructed according to this side length. After simplification, the point cloud data volume was reduced from 1.2 million points to 380,000 points, a reduction of 68.3%, while retaining key terrain features.

[0112] CSF Ground Point Extraction: Setting Fabric Stiffness Coefficient Time step Number of iterations The ground point determination threshold is determined by the formula. Simulated fabric settlement (initial fabric elevation 6m higher than the highest point cloud) extracted 320,000 ground point clouds, successfully eliminating interference from non-ground points such as trees and bridges.

[0113] Step S3: Cross-sectional data extraction

[0114] Import CAD design line data and set the sampling interval. (Due to the inclusion of complex terrain), cross-sectional width ;

[0115] 101 sampling points (from K0+000 to K0+1000, one every 10m) are generated along the road baseline, and the normal direction of each sampling point is calculated;

[0116] With the sampling point as the center, extend 30m to each side along the normal direction to generate 101 cross-sectional line segments;

[0117] On each cross-sectional line segment, according to the sampling interval Sampling was performed, with 61 sampling points obtained from each cross section. The elevation of the corresponding ground point cloud was extracted through spatial matching, forming a set of 101 cross section points. (Each point set contains 61 three-dimensional coordinate points).

[0118] In step S4, inverse distance weighted interpolation is used to generate the original ground line.

[0119] Calculate the average slope of the triangle formed by the three nearest known points around each interpolation point. :

[0120] Plain area (K0+000-K0+300): Calculated according to the formula ;

[0121] Hilly area (K0+300-K0+700): Calculations yielded ;

[0122] Steep slope area (K0+700-K0+1000): Calculations yielded ≈2.349;

[0123] Set distance threshold Interpolation formula is used Interpolation calculations are performed on each cross-section segment to obtain the continuous elevation curve of each cross-section; the interpolation results of 101 cross-sections are spliced ​​together in mileage order to form the continuous original ground line of the K0+000-K0+1000 section.

[0124] Step S5: Accuracy Verification and Parameter Optimization

[0125] Cross-validation: 80% of the points in each cross section were used as the training set and 20% as the validation set (a total of 1232 validation points). The calculated RMSE=0.11m and MAE=0.078m, which meet the preset error threshold requirements.

[0126] Local optimization: In the steep slope area (K0+850-K0+900), the initial RMSE was 0.13m, and the distance threshold was adjusted. Afterwards, the RMSE dropped to 0.09m, further improving the accuracy.

[0127] Step S6: Data Export

[0128] The extracted original ground line data was exported as DXF format and imported into Civil3D software. It was successfully used for earthwork volume calculation (total earthwork volume error ≤ 2.5%) and longitudinal profile design of this highway section, meeting the engineering design requirements.

[0129] It is worth noting that the various units included in the above system embodiments are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0130] Furthermore, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the corresponding program can be stored in a computer-readable storage medium.

[0131] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for extracting original ground lines based on inverse distance weighted interpolation, characterized in that, Includes the following steps: Step S1, Data Acquisition: Acquire raw point cloud data from airborne LiDAR and CAD design line data for the target road area; Step S2, Point Cloud Data Preprocessing: Statistical filtering for noise reduction, voxel mesh downsampling for simplification, and cloth simulation filtering for ground point extraction are performed sequentially. Step S3, Cross-section data extraction: Generate cross-section line segments based on CAD design line data and obtain cross-section point sets; Step S4: Generate original ground line by inverse distance weighted interpolation: Using the inverse distance weighted interpolation algorithm with dynamic weighted index adjustment, a continuous original ground line is generated by interpolation with the cross-section point set as known points. Step S5, Accuracy Verification and Parameter Optimization: Evaluate the extraction accuracy using cross-validation and adjust the parameters to optimize the results; Step S6: Data Export: Export the original ground line data according to the engineering standard format.

2. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S1, the airborne LiDAR raw point cloud data and the corresponding CAD design line data of the target road area are acquired; wherein, the airborne LiDAR raw point cloud data includes ground elevation information, and the CAD design line data includes road baseline and design parameter information of the design cross section.

3. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S2, statistical filtering denoising uses a statistical filtering algorithm to process the raw point cloud data of the airborne LiDAR. It calculates the average distance and standard deviation of points within a fixed neighborhood around each point cloud, sets a distance threshold as the sum of the average distance and three times the standard deviation, and identifies points with a distance greater than this threshold as noise points and removes them. The formula is as follows: ； In the formula, Distance threshold The average distance within the neighborhood of the point cloud. The distance is the standard deviation; if point cloud Average distance to all points in its neighborhood ,but These are noise points and should be removed.

4. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S2, voxel mesh downsampling simplification involves constructing a three-dimensional voxel mesh, dividing the preprocessed point cloud data into individual voxel units, and retaining a representative point within each voxel unit (this point is the geometric center of all points within the voxel unit). Redundant point clouds are eliminated, reducing the data volume while ensuring that the geometric features of the point cloud are not lost. The voxel mesh side length s is adaptively set according to the point cloud density, as shown in the following formula: ； In the formula, The bounding box volume of the point cloud data. This represents the total number of point clouds after statistical filtering.

5. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S2, the fabric simulation filtering ground point extraction treats point cloud data as particles in three-dimensional space, simulating the process of a virtual piece of fabric naturally sinking to the ground under gravity, by setting a time step. Number of iterations The parameters are used to make the virtual cloth conform to the shape of the ground, and finally the point cloud within a certain distance below the cloth is identified as the ground point cloud. The formula is as follows: ； In the formula, For the fabric after settling The elevation of the location This is the initial elevation of the fabric. For point clouds, the acceleration due to gravity is considered. satisfy , If P is the threshold for determining ground points, then P is a ground point.

6. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S3, the specific process for extracting cross-sectional data is as follows: Step S31: Import CAD design line data and set the sampling interval according to road condition design requirements. and cross-sectional width ; Step S32: Generate samples uniformly along the baseline of the CAD design line. Calculate the tangent direction of the baseline at each sampling point, and then obtain the normal direction of the cross section. ; Step S33: Using sampling points Extending outwards from the center along the normal direction Distance, generating cross-sectional line segments; Step S34: Sample at equal intervals along the cross-sectional line segment, with the sampling interval being... Obtain the cross-sectional point set Each point contains three-dimensional coordinate information. .

7. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S4, the specific process for generating the original ground line using inverse distance-weighted interpolation is as follows: Step S41: Using the cross-sectional points obtained in step S3 Using the points in the middle as known elevation points, interpolation calculations are performed on all positions to be interpolated on the cross-sectional line segment; Step S42: An improved inverse distance weighted interpolation algorithm is adopted, and a dynamic weight index adjustment mechanism is introduced. The formula is as follows: (Adaptive adjustment based on terrain slope) ； In the formula, The average slope of the triangle formed by the three nearest known points around the point to be interpolated, with a range of values. When the terrain is flat ( )hour, ; where the terrain is steep ( )hour, Increase the weight of known points in close proximity; Step S43: The distance decay function adopts the inverse square of the distance form, and a distance threshold is introduced at the same time. When the distance between the known point and the point to be interpolated exceeds In this case, the known point is not included in the interpolation calculation to suppress long-distance noise interference. The interpolation formula is as follows: ； In the formula, Points to be interpolated elevation, The number of known points participating in the interpolation ( ), For the point to be interpolated and the first The Euclidean distance between known points, and , For the first The elevation of a known point The weighted index is dynamically adjusted. Step S44: Smoothly stitch together the interpolation results of all cross sections in order of road mileage to form a continuous original ground line of the target road.

8. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S5, each cross-sectional point set The data is randomly divided into a training set and a validation set. The interpolated elevation corresponding to the validation set is calculated using the interpolation algorithm in step S4 using the training set. This interpolated elevation is then compared with the actual elevation of the validation set to calculate the root mean square error and the mean absolute error, as shown in the following formulas: ； ； In the formula, To verify the number of collection points, For the verification set The actual elevation of each point The elevation is obtained through interpolation. The root mean square error, This represents the mean absolute error.

9. The original ground line extraction method based on inverse distance weighted interpolation according to claim 8, characterized in that, The root mean square error or Adjust the voxel mesh side length CSF algorithm parameters, interpolation weight index adjustment coefficient or distance threshold ,repeat Continue until the error meets the requirements.

10. The original ground line extraction method based on inverse distance weighted interpolation according to claim 1, characterized in that, In step S6, the optimized original ground line data is exported in the engineering standard format, supporting DXF, DWG, and CSV formats, and can be directly imported into road design software for subsequent engineering applications such as earthwork calculation and longitudinal profile design.

Images