Highway engineering original ground retest automatic calculation method and system based on laser point cloud and cloth simulation

Abstract

The invention provides a highway engineering original ground retest automatic calculation method and system based on laser point cloud and distribution simulation, and relates to the technical field of earthwork calculation, and the method comprises the steps: segmenting ground points through a deep learning point cloud segmentation model, and improving the robustness; point clouds are refined through the pose map, point cloud splicing errors are eliminated, and the absolute and relative precision of the large-range point clouds is ensured; road center line strong constraint cloth is adopted to simulate the cross section, and the problems that a free simulation result may deviate from the actual road trend, center line deviation and slope collapse are often generated due to lack of road geometric constraint in free deformation cloth simulation, and the cross section modeling error is large are solved. Meanwhile, the complex road environment can be effectively processed through the deep learning point cloud segmentation model and the constraint distribution simulation, the adaptability to the specific cross section form of the road is improved, terrain distortion is avoided, and the precision of road volume remeasurement is improved.

Description

An automated calculation method and system for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation. Technical Field

[0001] This invention relates to the field of earthwork calculation technology, and in particular to an automated calculation method and system for re-surveying the original ground surface of highway engineering based on laser point cloud and material placement simulation. Background Technology

[0002] The core basis for earthwork calculation is the cross-sectional area and the distance between adjacent cross sections, and these data must be obtained through road engineering resurvey. Currently, road engineering resurvey mainly relies on single-point measurement such as total station and GPS-RTK or traditional photogrammetry, which has problems such as low efficiency, limited coverage and susceptibility to environmental influences. Although the use of lidar technology can quickly obtain high-density three-dimensional point clouds, the point cloud data processing is complicated, specifically in the following aspects: (1) Ground point separation: Traditional point cloud data processing uses filtering methods (such as slope method, morphology), which are not effective in complex terrain and when there are a large number of non-ground points (vegetation, vehicles), and have poor robustness. (2) Point cloud stitching error: There is a cumulative error in mobile scanning. When the cumulative error of mobile lidar point cloud stitching (>0.5% / km) causes cross-section position drift, it affects the absolute and relative accuracy of large-scale point clouds. (3) Cross section extraction: Traditional methods (such as slicing by mileage, manual sketching) are inefficient and subjective, and it is difficult to accurately capture complex terrain (such as slopes).

[0003] CSF (Content Filtering) is an effective ground point filtering and terrain extraction method, but it has limitations in road engineering applications: (1) Free simulation results may deviate from the actual road direction. Free deformation CSF often results in centerline offset (>1m) and slope collapse due to lack of road geometric constraints, causing cross-section modeling errors to exceed 15%. (2) It is not adaptable to the unique cross-section morphology of roads (road camber, slope gradient). In complex scenarios (such as those with slope gravel and shoulder vegetation), the ground point extraction error rate of traditional filtering algorithms exceeds 30%, resulting in terrain distortion. (3) Existing earthwork calculation software usually requires a lot of manual intervention for cross-section drawing, matching and calculation, with low automation and low efficiency.

[0004] The aforementioned defects result in insufficient accuracy in remeasured volume calculations (generally >8%) and frequent manual intervention (correction required every 100 meters), failing to meet the audit requirements for high-grade roads (error ≤3%). Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems mentioned in the background art above, and provides an automated calculation method for re-measurement of the original ground surface of highway engineering based on laser point cloud and cloth simulation, so as to improve the accuracy of road volume re-measurement.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] An automated calculation method for resurveying the original ground surface in highway engineering based on laser point cloud and cloth simulation includes the following steps:

[0008] S1, Data Acquisition: Using drones and lidar equipment, collect three-dimensional point cloud data within the road construction width along the road centerline and obtain drone flight trajectory data including drone position and attitude information;

[0009] S2, Ground point segmentation: Automatically identify and classify ground point clouds in the 3D point cloud data using a pre-trained deep learning point cloud segmentation model;

[0010] S3, Point Cloud Optimization Processing: Based on the UAV flight trajectory data and point cloud features, loop closure detection is performed on the currently collected 3D point cloud data to obtain loop closure constraints; a pose graph integrating loop closure constraints and odometry information is constructed to optimize the UAV pose; the optimized pose is used to transform the ground point cloud to reposition the ground point cloud, thereby obtaining a globally consistent global ground point cloud.

[0011] S4, obtain the road centerline;

[0012] S5, Constrained cloth simulation cross-section extraction: Using the global ground point cloud as a reference ground, initialize a virtual cloth. The particles of the cloth are strongly constrained by the road centerline. Set physical parameters for optimizing the road cross-section shape. Then execute the cloth simulation process. The cloth falls onto the global ground point cloud under the action of gravity and constraint and reaches equilibrium. Connect the particle positions of the final shape of the cloth to form the cross-section line of the original ground.

[0013] S6, Cross-section matching and fusion: Obtain the cross-section line data of the original design ground of the road project, match and overlay the cross-section line of the re-measured original ground with the cross-section line of the original design ground at the corresponding mileage to obtain the matched cross-section.

[0014] S7, Volume Calculation: Calculate the cut and fill area between two adjacent matching cross sections, calculate the earth and rock volume between two adjacent matching cross sections based on the cross section spacing, and sum up the earth and rock volumes between all two adjacent matching cross sections to obtain the total remeasured volume.

[0015] Furthermore, in step S3, loop closure detection is triggered when the overlap between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight is greater than the set threshold and the distance between the pose of the current UAV and the pose in the UAV flight trajectory data acquired in the previous flight is less than a preset range.

[0016] Furthermore, the loop closure detection method in step S3 is as follows: first, the ICP algorithm based on point cloud features is used to perform a coarse match between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight; then, NDT is used to perform a fine match between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight to obtain loop closure constraints.

[0017] Further, in step S3, the pose graph consists of nodes including position and orientation, and edges including odometry constraints and loop closure constraints. The pose optimization objective function of the pose graph is:

[0018] ;

[0019] In the formula, e ij It is the error between the observed value and the predicted value. It is the covariance matrix, (x i , x j ) represents the planar coordinates of the pose.

[0020] Furthermore, in step S4, the method for obtaining the road centerline is to directly import the road design centerline from the design data as the road centerline.

[0021] Furthermore, in step S5, the physical parameters for optimizing the road cross-section morphology are set as follows:

[0022] ① Set the cloth mesh resolution, i.e., the particle spacing, to 0.1-0.5m, and adjust it according to the point cloud density and road width;

[0023] ② The acceleration due to gravity is set to 9.8 m / s², with the direction vertically downwards;

[0024] ③ The fabric stiffness is set to 50-200, which can be adjusted according to the degree of terrain undulation. A larger value is used when the terrain undulation is large.

[0025] ④ The damping coefficient is set to 0.2-0.5. The damping coefficient is used to control the convergence speed.

[0026] ⑤ Set the number of iterations to 50-200, or until the energy converges, i.e., the change in kinetic energy is <1e-5.

[0027] Further, step S5 includes the following steps:

[0028] First, using the global ground point cloud as a reference ground, cross-sectional positions are set at preset fixed intervals along the centerline of the road.

[0029] Subsequently, at each set cross-sectional position, a cloth grid is initialized. The particles of the cloth are strongly constrained by the road centerline, so that the width of the cloth grid covers the road design width and extends outward by a preset safety distance. The length of the cloth grid along the road centerline direction is 1 particle spacing, and the length along the cross-sectional direction is the grid width.

[0030] Secondly, physical parameters for optimizing the cross-sectional shape of the road are set;

[0031] Finally, the cloth mesh is placed above the global ground point cloud, and cloth simulation is performed. During the simulation, the particles of the cloth mesh fall under gravity and collide with the global ground point cloud. After reaching equilibrium, the shape of the cloth mesh is the cross-sectional line of the re-measured original ground.

[0032] Further, in step S6, the method for matching and overlaying the cross-section line of the re-measured original ground with the cross-section line of the design original ground at the corresponding mileage is as follows: align the cross-section line of the design original ground with the cross-section line of the re-measured original ground at the corresponding mileage in the same coordinate system, and then superimpose the cross-section line of the design original ground at the same mileage with the cross-section line of the re-measured original ground to obtain a matching cross-section.

[0033] Further, in step S7, the earthwork volume V between two adjacent matching cross sections is calculated using the average cross section method, as shown in the following formula:

[0034] ;

[0035] In the formula, A1 and A2 are the fill area and cut area of ​​two adjacent matching cross sections, respectively, and L is the cross section distance between two adjacent matching cross sections.

[0036] This invention further provides an automated calculation system for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, comprising:

[0037] The point cloud segmentation module is configured to: receive three-dimensional point cloud data and UAV flight trajectory data within the collected road construction width range, and automatically identify and classify the ground point cloud in the three-dimensional point cloud data using a pre-trained deep learning point cloud segmentation model.

[0038] The point cloud optimization module is configured to: perform loop closure detection on the currently collected 3D point cloud data based on the UAV flight trajectory data and point cloud features to obtain loop closure constraints; construct a pose graph that integrates loop closure constraints and odometry information to optimize the UAV pose; and use the optimized pose to transform the ground point cloud to relocate the ground point cloud, thereby obtaining a globally consistent global ground point cloud.

[0039] The centerline acquisition module is configured to acquire the road centerline;

[0040] The constrained cloth simulation module is configured to: use the global ground point cloud as a reference ground, initialize a virtual cloth, the particles of the cloth are strongly constrained by the road centerline, receive physical parameters for optimizing the road cross-section shape input by the user, and then execute the cloth simulation process. The cloth falls onto the global ground point cloud under the action of gravity and constraint and reaches equilibrium. The particle positions of the final shape of the cloth are connected to form the cross-section line of the original ground.

[0041] The cross-section matching module is configured to: acquire the cross-section line data of the original design ground of the road project, match and overlay the cross-section line of the re-measured original ground with the cross-section line of the original design ground at the corresponding mileage to obtain the matching cross-section.

[0042] The volume calculation module is configured to: calculate the cut and fill area between two adjacent matching cross sections, calculate the earth and rock volume between two adjacent matching cross sections based on the cross section spacing between them, and sum up the earth and rock volumes between all two adjacent matching cross sections to obtain the total remeasured volume.

[0043] By adopting the above technical solution, the present invention has the following beneficial effects:

[0044] This invention presents an automated calculation method for re-surveying the original ground surface of highway engineering based on laser point cloud and cloth simulation. It improves robustness by segmenting ground points using a deep learning point cloud segmentation model; optimizes point cloud processing through pose graphs to eliminate point cloud splicing errors and ensure the absolute and relative accuracy of large-scale point clouds; and employs a strongly constrained cloth simulation of the road centerline to avoid deviations from the actual road direction in free simulations. Free deformation cloth simulations often suffer from centerline offsets (>1m) and slope collapses due to a lack of road geometric constraints, leading to large errors in cross-section modeling. Simultaneously, the deep learning point cloud segmentation model and constrained cloth simulation effectively handle complex road environments (slopes, shoulders, and a certain degree of vegetation / debris interference), improving adaptability to unique road cross-sectional morphologies (road camber, slope gradient) and avoiding terrain distortion. Therefore, this invention significantly improves the accuracy of each step by using a three-in-one technology of robust ground segmentation through deep learning, pose graph point cloud optimization, and centerline strong constraint cloth simulation, ultimately improving the accuracy of road volume remeasurement and meeting the requirements of high-grade road audit (error ≤ 3%).

[0045] The present invention provides an automated calculation method for re-surveying the original ground surface of highway engineering based on laser point cloud and cloth simulation. This method realizes fully automated re-surveying of road volume, greatly reduces manual intervention, and shortens the re-surveying cycle. Attached Figure Description

[0046] Figure 1 is a flowchart of an automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, provided by a preferred embodiment of the present invention.

[0047] Figure 2 shows the three-dimensional point cloud data of the road site collected by the automated calculation method for resurveying the original ground of highway engineering based on laser point cloud and cloth simulation according to a preferred embodiment of the present invention.

[0048] Figure 3 is a schematic diagram of the cross-sectional line of the resurveyed original ground in the automated calculation method for resurveying the original ground of highway engineering based on laser point cloud and cloth simulation according to a preferred embodiment of the present invention. In the figure, red represents the cross-sectional line of the resurveyed original ground, and blue represents the road centerline.

[0049] Figure 4 is a schematic diagram of superimposing the cross-sectional lines of the designed original ground and the cross-sectional lines of the remeasured original ground in the automated calculation method for remeasurement of original ground in highway engineering based on laser point cloud and cloth simulation according to a preferred embodiment of the present invention. In the figure, white represents the cross-sectional lines of the designed original ground, green represents the cross-sectional lines of the designed roadbed, and red represents the cross-sectional lines of the remeasured original ground.

[0050] Figure 5 is a block diagram of an automated calculation system for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, provided by a preferred embodiment of the present invention. Detailed Implementation

[0051] 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, not all, of the embodiments of the present invention. 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.

[0052] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0053] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0054] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0055] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0056] Please refer to Figures 1 to 4 simultaneously. A preferred embodiment of the present invention provides an automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, comprising the following steps:

[0057] S1, Data Acquisition: Using drones and lidar equipment, collect three-dimensional point cloud data within the road construction width along the road centerline and obtain drone flight trajectory data including drone position and attitude information.

[0058] In this embodiment, a drone combined with a lidar system is used to collect 3D point cloud data within the road construction width along the road centerline and obtain drone flight trajectory data. Specifically, data collection can be performed using a drone equipped with a lidar system and a positioning and attitude determination system. During collection, the lidar system scans and collects 3D point cloud data of the road site. The positioning and attitude determination system obtains drone flight trajectory data including the drone's position and attitude information. Specifically, in this embodiment, the core hardware configuration parameters of the drone, lidar system, and positioning and attitude determination system are as follows:

[0059] Unmanned aerial vehicle (UAV) platform: VTOL fixed-wing UAV or multi-rotor UAV with a flight time of ≥1.5 hours and wind resistance of ≥8m / s.

[0060] LiDAR system: Integrated mapping-grade LiDAR (such as Riegl VUX-1LR or equivalent), with the following performance characteristics:

[0061] Ranging capability: Effective measurement range ≥ 500 meters (for targets with 60% reflectivity), with a maximum range of up to 1350 meters.

[0062] Point frequency: Maximum effective measurement rate ≥750,000 points / second.

[0063] Scanning field of view: The horizontal field of view can reach 330°.

[0064] Distance measurement accuracy: ±1.5 cm.

[0065] Positioning and Attitude System (POS):

[0066] GNSS (Global Navigation Satellite System): Supports positioning using multiple systems including GPS, GLONASS, GALILEO, and BeiDou.

[0067] IMU (Inertial Measurement Unit): Attitude accuracy (post-processing) can reach 0.006°.

[0068] Absolute elevation accuracy: up to 5 centimeters at an altitude of 200 meters.

[0069] Operation parameters:

[0070] Flight altitude: Set according to accuracy requirements and terrain undulations, adopting a terrain-following flight method, with a typical operating flight altitude of 200-350 meters.

[0071] Point cloud density: Controlled by adjusting flight altitude and flight speed, it is required to be ≥50 points / square meter.

[0072] Flight speed: Cruising speed 70-80 km / h.

[0073] Flight route planning: Design parallel flight routes based on the shape and accuracy requirements of the survey area, ensuring a lateral overlap rate of ≥20%.

[0074] The lidar system and positioning and attitude determination system can connect with the UAV platform via 5G / 4G and other communication methods to transmit the collected information.

[0075] S2, Ground Point Segmentation: Automatically identify and classify ground point clouds in 3D point cloud data using a pre-trained deep learning point cloud segmentation model.

[0076] In this embodiment, before ground point segmentation, the original 3D point cloud data acquired in step S1 undergoes denoising and downsampling preprocessing. During denoising, outliers are removed through statistical analysis. During downsampling, a voxel grid filtering method is used to divide the point cloud space into voxel grids, retaining only one representative point (such as the voxel center or centroid within the voxel) in each voxel. The voxel size is set to 0.05m-0.1m. This denoising and downsampling preprocessing improves the quality and usability of the point cloud data.

[0077] In this implementation, RandLA-Net or KPConv is used as the deep learning point cloud segmentation model. RandLA-Net or KPConv possesses efficient processing capabilities for large-scale point clouds and strong local feature learning abilities. During training, a custom-annotated road scene point cloud dataset is used as the training dataset. This dataset includes various road types (highways, municipal roads, rural roads) and complex scenes (slopes, shoulders, vegetation, buildings, vehicles, etc.). The training dataset contains ≥100,000 point cloud segments, each containing ≥100,000 points, with annotation categories including at least ground, vegetation, buildings, vehicles, and curbs. During model training, the Adam optimizer is used with an initial learning rate of 0.01, employing a cosine annealing strategy, and a training cycle of 100 epochs. The batch size is adjusted based on GPU memory. Data augmentation includes random rotation, translation, scaling, and point dropping. After training, the deep learning point cloud segmentation model's accuracy is evaluated on the test set. The results show that the IoU (Intersection over Union) for ground point classification is ≥90%, and the overall accuracy is ≥95%.

[0078] Subsequently, the ground point cloud in the 3D point cloud data is automatically identified and classified using a pre-trained deep learning point cloud segmentation model: In this embodiment, the pre-processed 3D point cloud data is input into the trained model, the semantic label of each point is output, and the ground points are extracted, thus completing the ground point segmentation.

[0079] S3, Point Cloud Optimization Processing: Based on the UAV flight trajectory data and point cloud features, loop closure detection is performed on the currently collected 3D point cloud data to obtain loop closure constraints; a pose graph integrating loop closure constraints and odometry information is constructed to optimize the UAV pose; the optimized pose is used to transform the ground point cloud to relocate the ground point cloud, thereby obtaining a globally consistent global ground point cloud.

[0080] In this embodiment, loop closure detection is triggered when the overlap (calculated by point cloud density or feature matching) between the currently acquired 3D point cloud data and the previously acquired 3D point cloud data is greater than a set threshold (e.g., 30%), and the distance between the current UAV pose and the pose in the UAV flight trajectory data acquired in the previous flight is less than a preset range (e.g., 50m). The loop closure detection method is as follows: first, the ICP (Iterative Closest Point) algorithm based on point cloud features (e.g., FPFH) is used to perform a coarse match between the currently acquired 3D point cloud data and the previously acquired 3D point cloud data; then, the NDT (Normal Distribution Transform) is used to perform a fine match between the currently acquired 3D point cloud data and the previously acquired 3D point cloud data to obtain loop closure constraints.

[0081] In step S3, the pose graph consists of nodes including position and orientation, and edges including odometry constraints and loop closure constraints. The pose optimization objective function of the pose graph is:

[0082] ;

[0083] In the formula, e ij It is the error between the observed value and the predicted value. It is the covariance matrix, (x i , x j ) represents the planar coordinates of the pose.

[0084] In this embodiment, graph optimization libraries such as g2o or Ceres Solver can be used to solve the pose optimization objective function of the pose graph.

[0085] By optimizing the pose graph, a globally consistent ground point cloud can be obtained, which can significantly reduce the cumulative error of point cloud stitching and obtain a high-precision, consistent global point cloud.

[0086] S4, obtain the road centerline.

[0087] Specifically, there are three methods for obtaining the road centerline: First, smoothing the trajectory data using a moving average filter to obtain the road centerline; second, fitting the road boundary using the RANSAC algorithm based on the global ground point cloud, and then calculating the road centerline; third, directly importing the road design centerline from the design data as the road centerline. In this embodiment, the method for obtaining the road centerline is to directly import the road design centerline from the design data as the road centerline.

[0088] S5, Constrained Cloth Simulation Cross-Section Extraction: Using the global ground point cloud as a reference ground, a virtual cloth is initialized. The particles of the cloth are strongly constrained by the road centerline. Physical parameters optimized for the road cross-section shape are set. Then, the cloth simulation process is executed. Under the action of gravity and constraints, the cloth falls onto the global ground point cloud and reaches equilibrium. The particle positions of the final cloth shape are connected to form the cross-section line of the original ground. In this embodiment, step S5 specifically includes the following steps:

[0089] First, using the global ground point cloud as a reference ground, set cross-sectional positions along the road centerline at preset fixed intervals (e.g., 20m).

[0090] Subsequently, at each set cross-sectional position, a fabric grid is initialized. The particles of the fabric are strongly constrained by the road centerline, so that the width of the fabric grid covers the road design width and extends beyond a preset safety distance (such as 5m). The length of the fabric grid along the road centerline direction is 1 particle spacing, and the length along the cross-sectional direction is the grid width.

[0091] Secondly, physical parameters for optimizing the road cross-section shape are set. In this embodiment, the physical parameters for optimizing the road cross-section shape are set as follows:

[0092] ① Set the cloth mesh resolution, i.e., the particle spacing, to 0.1-0.5m, and adjust it according to the point cloud density and road width;

[0093] ② The acceleration due to gravity is set to 9.8 m / s², with the direction vertically downwards;

[0094] ③ The fabric stiffness is set to 50-200, which can be adjusted according to the degree of terrain undulation. A larger value is used when the terrain undulation is large.

[0095] ④ The damping coefficient is set to 0.2-0.5. The damping coefficient is used to control the convergence speed.

[0096] ⑤ Set the number of iterations to 50-200, or until the energy converges, i.e., the change in kinetic energy is <1e-5.

[0097] Finally, place the cloth mesh above the global ground point cloud and perform cloth simulation. During the simulation, the particles of the cloth mesh fall under gravity and collide with the global ground point cloud. After reaching equilibrium, the shape of the cloth mesh is the cross-sectional line of the remeasured original ground.

[0098] In this embodiment, after the constraint cloth simulation, the cross-sectional lines of the obtained re-measured original ground are smoothed and filtered (such as Savitzky-Golay filtering) to remove burrs, and outliers (such as sudden depressions caused by missing point clouds) are checked. If outliers are found, interpolation is used to repair them.

[0099] S6, Cross-section matching and fusion: Obtain the cross-section line data of the original design ground of the road project, match and overlay the cross-section line of the re-measured original ground with the cross-section line of the original design ground at the corresponding mileage to obtain the matched cross-section.

[0100] In this embodiment, when setting the extraction density of the cross-section lines of the original design ground and the cross-section lines of the re-measured original ground, a cross-section is typically set every 20m, and can be densified to 5m in areas with curves or large terrain changes. The method for matching and overlaying the cross-section lines of the re-measured original ground with the cross-section lines of the design ground at the corresponding mileage is as follows: align the cross-section lines of the design ground and the cross-section lines of the re-measured original ground at the corresponding mileage in the same coordinate system, and then superimpose the cross-section lines of the design ground and the re-measured original ground at the same mileage to obtain a matched cross-section.

[0101] During the processing, if there are local missing cross-sectional lines of the re-measured original ground, interpolation between adjacent cross-sections will be used; if there is a significant inconsistency between the cross-sectional lines of the original ground design and the cross-sectional lines of the re-measured original ground in a local location (such as due to construction changes), the area will be marked and manually verified.

[0102] S7, Volume Calculation: Calculate the cut and fill area between two adjacent matching cross sections. Calculate the earthwork volume between two adjacent matching cross sections based on the cross section distance (i.e., the mileage difference between two adjacent matching cross sections). Summate the earthwork volume between all two adjacent matching cross sections to obtain the total remeasured volume.

[0103] In step S7, the earthwork volume V between two adjacent matching cross sections is calculated using the average cross section method, as shown in the following formula:

[0104] ;

[0105] In the formula, A1 and A2 are the fill area and cut area of ​​two adjacent matching cross sections, respectively. Their calculation method is existing technology and will not be elaborated here for the sake of brevity; L is the cross section distance between two adjacent matching cross sections.

[0106] Please refer to Figure 5. This invention also provides an automated calculation system for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, comprising:

[0107] The point cloud segmentation module is configured to: receive three-dimensional point cloud data and UAV flight trajectory data within the collected road construction width range, and automatically identify and classify the ground point cloud in the three-dimensional point cloud data using a pre-trained deep learning point cloud segmentation model.

[0108] The point cloud optimization module is configured to: perform loop closure detection on the currently collected 3D point cloud data based on the UAV flight trajectory data and point cloud features to obtain loop closure constraints; construct a pose graph that integrates loop closure constraints and odometry information to optimize the pose; and use the optimized pose to transform the ground point cloud to relocate the ground point cloud, thereby obtaining a globally consistent global ground point cloud.

[0109] The centerline acquisition module is configured to acquire the road centerline;

[0110] The constrained cloth simulation module is configured to: use the global ground point cloud as a reference ground, initialize a virtual cloth, the particles of the cloth are strongly constrained by the road centerline, receive physical parameters for optimizing the road cross-section shape input by the user, and then execute the cloth simulation process. The cloth falls onto the global ground point cloud under the action of gravity and constraint and reaches equilibrium. The particle positions of the final shape of the cloth are connected to form the cross-section line of the original ground.

[0111] The cross-section matching module is configured to: acquire the cross-section line data of the original design ground of the road project, match and overlay the cross-section line of the re-measured original ground with the cross-section line of the original design ground at the corresponding mileage to obtain the matching cross-section.

[0112] The volume calculation module is configured to: calculate the cut and fill area between two adjacent matching cross sections, calculate the earth and rock volume between two adjacent matching cross sections based on the cross section spacing, and sum up the earth and rock volume between all two adjacent matching cross sections to obtain the total remeasured volume.

[0113] This invention presents an automated calculation method for re-surveying the original ground surface of highway engineering based on laser point cloud and cloth simulation. It improves robustness by segmenting ground points using a deep learning point cloud segmentation model; optimizes point cloud processing through pose graphs to eliminate point cloud splicing errors and ensure the absolute and relative accuracy of large-scale point clouds; and employs a strongly constrained cloth simulation of the road centerline to avoid deviations from the actual road direction in free simulations. Free deformation cloth simulations often suffer from centerline offsets (>1m) and slope collapses due to a lack of road geometric constraints, leading to large errors in cross-section modeling. Simultaneously, the deep learning point cloud segmentation model and constrained cloth simulation effectively handle complex road environments (slopes, shoulders, and a certain degree of vegetation / debris interference), improving adaptability to unique road cross-sectional morphologies (road camber, slope gradient) and avoiding terrain distortion. Therefore, this invention significantly improves the accuracy of each step by using a three-in-one technology of robust ground segmentation through deep learning, pose graph point cloud optimization, and centerline strong constraint cloth simulation, ultimately improving the accuracy of road volume remeasurement and meeting the requirements of high-grade road audit (error ≤ 3%).

[0114] The present invention provides an automated calculation method for re-surveying the original ground surface of highway engineering based on laser point cloud and cloth simulation. This method realizes fully automated re-surveying of road volume, greatly reduces manual intervention, and shortens the re-surveying cycle.

[0115] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0117] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0118] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0119] The above description is a detailed description of the preferred embodiments of the present invention. However, the embodiments are not intended to limit the scope of the patent application of the present invention. All equivalent changes or modifications made under the technical spirit of the present invention should fall within the patent scope covered by the present invention.

Claims

1. An automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, characterized in that, Includes the following steps: S1, Data Acquisition: Utilizing a drone and lidar equipment, collect 3D point cloud data within the road construction width along the road centerline and acquire drone flight trajectory data including drone position and attitude information; S2, Ground Point Segmentation: Automatically identify and classify ground point clouds in the 3D point cloud data using a pre-trained deep learning point cloud segmentation model; S3, Point Cloud Optimization Processing: Based on the drone flight trajectory data and point cloud features, perform loop closure detection on the currently acquired 3D point cloud data to obtain loop closure constraints; construct a pose graph fusing loop closure constraints and odometry information to optimize the drone's pose; use the optimized pose to transform the ground point cloud to reposition the ground point cloud, thereby obtaining a globally consistent global ground point cloud; S4, Obtain the Road Centerline; S5, Constraint Fabric Simulation Cross-Section Extraction: Use the global ground point cloud as a reference... S6. Initialize a virtual cloth on the ground. The particles of the cloth are strongly constrained by the centerline of the road. Set physical parameters for optimizing the cross-sectional shape of the road. Then, execute the cloth simulation process. The cloth falls onto the global ground point cloud under the action of gravity and constraint and reaches equilibrium. Connect the particle positions of the final shape of the cloth to form the cross-sectional line of the remeasured original ground. S7. Cross-sectional matching and fusion: Obtain the cross-sectional line data of the original ground design of the road project. Match and overlay the cross-sectional line of the remeasured original ground with the cross-sectional line of the original ground design at the corresponding mileage to obtain the matched cross-section. S8. Volume calculation: Calculate the cut and fill area between two adjacent matched cross-sections. Calculate the earthwork volume between two adjacent matched cross-sections based on the cross-sectional distance between them. Accumulate the earthwork volume between all two adjacent matched cross-sections to obtain the total remeasured volume.

2. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S3, loop closure detection is triggered when the overlap between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight is greater than the set threshold and the distance between the pose of the current UAV and the pose in the flight trajectory data of the UAV acquired in the previous flight is less than the preset range.

3. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, The loop closure detection method in step S3 is as follows: First, the ICP algorithm based on point cloud features is used to perform a coarse match between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight. Then, NDT is used to perform a fine match between the currently acquired 3D point cloud data and the 3D point cloud data acquired in the previous flight to obtain loop closure constraints.

4. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S3, the pose graph consists of nodes including position and orientation, and edges including odometry constraints and loop closure constraints. The pose optimization objective function of the pose graph is: In the formula, e ij It is the error between the observed value and the predicted value. It is the covariance matrix, (x i ,x j ) represents the planar coordinates of the pose.

5. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S4, the method for obtaining the road centerline is to directly import the road design centerline from the design data as the road centerline.

6. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S5, the physical parameters for optimizing the road cross-section morphology are set as follows: ① The mesh resolution, i.e., the particle spacing, is set to 0.1-0.5m, adjusted according to the point cloud density and road width; ② The gravitational acceleration is set to 9.8m / s², with the direction vertically downward; ③ The mesh stiffness is set to 50-200, adjusted according to the degree of terrain undulation, with a larger value taken when the terrain undulation is large; ④ The damping coefficient is set to 0.2-0.5, which is used to control the convergence speed; ⑤ The number of iterations is set to 50-200 times, or until energy convergence, i.e., the kinetic energy change is <1e-5.

7. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, Step S5 includes the following steps: First, using the global ground point cloud as a reference ground, cross-section positions are set at preset fixed intervals along the road centerline; then, at each set cross-section position, a cloth grid is initialized. The particles of the cloth are strongly constrained by the road centerline, so that the width of the cloth grid covers the road design width and extends beyond a preset safety distance. The length of the cloth grid along the road centerline direction is 1 particle spacing, and the length along the cross-section direction is the grid width; second, physical parameters for optimizing the road cross-section shape are set; finally, the cloth grid is placed above the global ground point cloud, and cloth simulation is performed. During the simulation, the particles of the cloth grid fall under gravity and collide with the global ground point cloud. After reaching equilibrium, the shape of the cloth grid is the cross-section line of the re-measured original ground.

8. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S6, the method for matching and overlaying the cross-section line of the re-measured original ground with the cross-section line of the design original ground at the corresponding mileage is as follows: align the cross-section line of the design original ground with the cross-section line of the re-measured original ground at the corresponding mileage in the same coordinate system, and then superimpose the cross-section line of the design original ground at the same mileage with the cross-section line of the re-measured original ground to obtain a matching cross-section.

9. The automated calculation method for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation as described in claim 1, characterized in that, In step S7, the earthwork volume V between two adjacent matching cross sections is calculated using the average cross section method, as shown in the following formula: In the formula, A1 and A2 are the fill area and cut area of ​​two adjacent matching cross sections, respectively, and L is the cross section distance between two adjacent matching cross sections.

10. An automated calculation system for resurveying the original ground surface of highway engineering based on laser point cloud and cloth simulation, characterized in that, include: The point cloud segmentation module is configured to: receive 3D point cloud data and UAV flight trajectory data within the collected road construction width range; automatically identify and classify ground point clouds in the 3D point cloud data using a pre-trained deep learning point cloud segmentation model; and the point cloud optimization module is configured to: perform loop closure detection on the currently collected 3D point cloud data based on the UAV flight trajectory data and point cloud features to obtain loop closure constraints; construct a pose graph that integrates loop closure constraints and odometry information to optimize the UAV pose; and use the optimized pose to transform the ground point cloud to improve the ground point cloud structure. Cloud relocation is performed to obtain a globally consistent global ground point cloud; the centerline acquisition module is configured to acquire the road centerline; the constraint cloth simulation module is configured to use the global ground point cloud as a reference ground, initialize a virtual cloth, the particles of the cloth are strongly constrained by the road centerline, receive physical parameters for optimizing the road cross-section shape input by the user, and then execute the cloth simulation process. The cloth falls onto the global ground point cloud under the action of gravity and constraint and reaches equilibrium. The particle positions of the final shape of the cloth are connected to form the cross-section line of the original ground. The cross-section matching module is configured to: acquire the cross-section line data of the original design ground of the road project, match and overlay the cross-section line of the re-measured original ground with the cross-section line of the original design ground at the corresponding mileage to obtain a matching cross-section; the volume calculation module is configured to: calculate the cut and fill area between two adjacent matching cross-sections, calculate the earthwork volume between two adjacent matching cross-sections based on the cross-sectional distance between the two adjacent matching cross-sections, and sum up the earthwork volume between all two adjacent matching cross-sections to obtain the total re-measured volume.

Images