Pavement pit slot crack three-dimensional model building method based on point cloud data

Abstract

The invention discloses a pavement pit slot crack three-dimensional model establishment method based on point cloud data, and belongs to the technical field of engineering measurement and pavement detection, and the method comprises the following steps: S1, data acquisition and processing; s2, preprocessing the original point cloud data; s3, identifying the point cloud of the crack of the pit slot; and S4, establishing a three-dimensional model of the pit slot crack. According to the method, three-dimensional modeling is carried out by using the point cloud data, pits and cracks on the road surface can be accurately identified and positioned, and the problem that tiny damage is easily ignored in a traditional manual detection method is avoided; through the automatic point cloud data processing and clustering thought and the triangulation algorithm, manual intervention is reduced, the data processing efficiency is improved, and an accurate three-dimensional model can be quickly generated.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering measurement and road surface detection, and in particular to a method for establishing a three-dimensional model of road surface potholes and cracks based on point cloud data. Background Art

[0002] With the continuous increase in traffic volume, the problem of road damage has become increasingly prominent, especially on urban roads and highways. Problems such as potholes and cracks not only affect driving safety, but may also lead to traffic accidents and economic losses. Traditional road surface inspection methods, such as manual inspections and simple measurements, are inefficient and subjective, easily miss some minor damage, and are difficult to provide a comprehensive assessment of road conditions. These defects make it difficult for traditional methods to meet the needs of modern road management.

[0003] In recent years, 3D laser scanning technology, as an emerging measurement method, has gradually been applied to the field of road surface inspection. Its main advantages are high precision and high efficiency. 3D laser scanning can quickly capture a large amount of point cloud data to form a high-resolution 3D model. These point cloud data can not only accurately record the shape and characteristics of the road surface, but also effectively identify damage types such as potholes and cracks. In addition, laser scanning technology also has the characteristics of non-contact measurement, which can be carried out without interfering with traffic, greatly improving measurement efficiency and safety.

[0004] Compared with traditional detection technology, 3D laser scanning provides a more comprehensive and accurate road surface assessment method. Through point cloud data, engineers can analyze the distribution and severity of road damage in detail and then formulate corresponding maintenance plans. This data-based analysis method makes road surface management more scientific and reasonable.

[0005] However, although 3D laser scanning technology has many advantages, it still faces some challenges in practical applications. First, the processing and analysis of point cloud data requires professional software and technical support, which places high demands on the technical level of operators. Secondly, the amount of point cloud data is large and it takes a long time to process, especially when real-time feedback is required, which may affect the detection efficiency. In addition, how to effectively extract road damage information from complex point cloud data is also a hot topic in current technical research. To this end, the present invention proposes a method for establishing a three-dimensional model of road potholes and cracks based on point cloud data. Summary of the invention

[0006] The technical problem to be solved by the present invention is: how to solve the problem that traditional road surface detection methods are inefficient and prone to errors, and are difficult to meet the requirements of modern road management, and then efficiently and accurately use point cloud data to identify road potholes and cracks, and establish their three-dimensional models, so as to provide a scientific basis for road maintenance and management. A method for establishing a three-dimensional model of road potholes and cracks based on point cloud data is provided. This method improves the accuracy and efficiency of road damage identification through automated processing and intelligent analysis.

[0007] The present invention solves the above technical problems through the following technical solutions, and the present invention comprises the following steps:

[0008] Step S1: Data collection and processing

[0009] Collect the original laser point cloud data of the road according to the pre-planned collection route and collection sequence;

[0010] Step S2: Preprocessing of original point cloud data

[0011] Preprocess the original laser point cloud data of the road to obtain the local road surface point cloud with potholes and cracks;

[0012] Step S3: Pothole and Crack Point Cloud Identification

[0013] Perform local plane fitting on each point of the local road surface point cloud with potholes and cracks, calculate the angle between the local plane normal vector of each point and the reference normal vector according to the reference normal vector of the road surface and assign a eigenvalue, identify the flat road surface point cloud based on the eigenvalue of each point and use the clustering algorithm, then remove the flat road surface point cloud to obtain the pothole and crack point cloud;

[0014] Step S4: Establishment of three-dimensional model of pits and cracks

[0015] The Delaunay triangulation algorithm is used to generate the triangular mesh of the pits and cracks on the point cloud, and then the surface is reconstructed to finally obtain the three-dimensional solid model of the pits and cracks.

[0016] Furthermore, in step S1, the original laser point cloud data collection process of the road is as follows:

[0017] Step S11: Arrange a set number of target control points around the road;

[0018] Step S12: using an RTK measuring instrument and a leveling measuring instrument to obtain the plane coordinates and elevation of the target control point;

[0019] Step S13: using a laser scanning system to obtain original laser point cloud data of the road;

[0020] Step S14: Select preset target control points to check the plane and elevation accuracy of the original laser point cloud data of the road.

[0021] Furthermore, in step S2, the process of preprocessing the original laser point cloud data of the road is as follows:

[0022] Step S21: using the CSF simulated cloth filtering algorithm to separate the road surface points and non-road surface points in the original laser point cloud data of the road to obtain a road surface point cloud;

[0023] Step S22: using a statistical filtering algorithm to remove high-frequency noise points in the road surface point cloud;

[0024] Step S23: using a Gaussian filter to smooth the road surface point cloud processed in step S22;

[0025] Step S24: downsampling the smoothed road surface point cloud using a voxel filter to obtain a simplified point cloud;

[0026] Step S25: Manually search and cut out the local road surface point cloud with potholes and cracks.

[0027] Furthermore, in step S21, the process of using the CSF simulated cloth filter algorithm to separate the road surface points and the non-road surface points is as follows:

[0028] Step S211: invert the original laser point cloud data of the road, project the inverted point cloud data and the cloth nodes simulated by the user onto the same horizontal plane, and find the point corresponding to the cloth node in the point cloud data, and record its height as h;

[0029] Step S212: During each iteration, the current height of the cloth node is compared with the size of h, until the absolute value of the difference between h and the current height is less than a preset value or the maximum number of iterations is reached, the iteration is terminated;

[0030] Step S213: Calculate the distance between the point cloud data point at the end of the inverted iteration and the cloth node. When the distance is greater than the set threshold, the point is recorded as a road point and retained. Otherwise, it is recorded as a non-road point and filtered out, thereby obtaining a road surface point cloud.

[0031] Furthermore, in step S22, the process of removing high-frequency noise points in the road surface point cloud using a statistical filtering algorithm is as follows:

[0032] Step S221: Perform statistical analysis on the neighborhood of each road point. The distances between all data points in the road point cloud data are assumed to form a Gaussian distribution, whose shape is determined by the mean and standard deviation Decision: Assume that the coordinates of the nth data point in the road point cloud data are , from this data point to any data point The distance is:

[0033] ;

[0034] Step S222: Calculate the average distance between each data point and any data point. The formula is:

[0035] ;

[0036] Step S223: Set the standard deviation to:

[0037] ;

[0038] Step S224: Set the standard deviation multiple to , when the average distance of a data point to k data points is within the standard range If it is within the range, the point is retained; if it is not, it is defined as an outlier and deleted.

[0039] Furthermore, in step S23, the process of smoothing the road point cloud using a Gaussian filter is as follows:

[0040] Step S231: Select the neighborhood range of each point to determine the neighborhood points for smoothing, and the neighborhood range is defined by setting a radius value r;

[0041] Step S232: For each point, use the Gaussian function to calculate the weight of each point in the neighborhood. The weight is determined according to the distance from the point to the center point. The closer the distance, the greater the weight.

[0042] Step S233: For each point, replace it with the weighted average of all points in its neighborhood to obtain a new smooth point. The position of the new smooth point is obtained by the following formula:

[0043] ;

[0044] in, is the smoothed point, is a point in the neighborhood, is the corresponding weight;

[0045] Step S234: repeating steps S231 to S233 until smoothing of all points is completed;

[0046] Step S235: Output the road point cloud data smoothed by the Gaussian filter.

[0047] Furthermore, in step S24, the process of downsampling the smoothed road surface point cloud using a voxel filter is as follows:

[0048] Step S241: Initialize filter parameters, including the size of voxels;

[0049] Step S242: Divide the smoothed road surface point cloud into corresponding voxels, traverse each data point in the road surface point cloud, determine the voxel to which it belongs according to the coordinates of the data point, map the coordinates of the data point to the voxel grid, and each voxel contains data points within a set range;

[0050] Step S243: Process the data points in each voxel. For each non-empty voxel, use the point closest to the voxel center to represent the data point in the voxel.

[0051] Step S244: generating a new point cloud data set according to the processing result;

[0052] Step S245: Output the point cloud data set processed by the voxel filter as input of step S25.

[0053] Furthermore, in step S3, the process of identifying the pit and crack point cloud is as follows:

[0054] Step S31: using the eigenvalue method to calculate the local plane normal vector for each point in the local road surface point cloud;

[0055] Step S32: Calculate the angle between the local plane normal vector of each point and the reference normal vector and assign an eigenvalue, use the eigenvalue to perform the first screening, identify the preliminary flat road surface point cloud, and then use the clustering algorithm to perform the second screening, and identify the final flat road surface point cloud from the preliminary flat road surface point cloud;

[0056] Step S33: The flat road surface point cloud finally identified in step S32 is eliminated, and the remaining part is the pothole and crack point cloud.

[0057] Furthermore, in step S31, the calculation process of the local plane normal vector is as follows:

[0058] Step S311: Each point in the local road point cloud The neighborhood point set of ; The neighborhood point set includes the point and k neighboring points;

[0059] Step S312: Using the neighborhood point set The eigenvalue method is used to calculate all points in the local plane to obtain the local plane normal vector set ;

[0060] Step S313: The calculated local plane normal vector With point cloud Associated.

[0061] Furthermore, in step S32, the specific processing process is as follows:

[0062] Step S321: Angle calculation

[0063] Calculate the angle between each point and the reference normal vector. The angle calculation formula is as follows:

[0064]

[0065] in, is the reference normal vector, ;

[0066] Step S322: Eigenvalue conversion

[0067] Calculate the eigenvalue of each point. The eigenvalue calculation formula is as follows:

[0068] ;

[0069] Step S323: Primary screening

[0070] Compare the eigenvalues ​​of each point, select the points with eigenvalues ​​≤ T, and mark them as preliminary flat road surface point clouds;

[0071] Step S324: Secondary screening

[0072] For each point marked as a preliminary smooth road point cloud, consider its nearest 50 neighboring points. If the number of points marked as preliminary smooth road point clouds among this point and its 50 neighboring points reaches or exceeds 90%, then the point is marked as the final smooth road point cloud. Otherwise, the corresponding point is deleted from the preliminary smooth road point cloud to obtain the final smooth road point cloud.

[0073] Furthermore, in step S4, the process of establishing the three-dimensional model of pits and cracks is as follows:

[0074] Step S41: applying a Delaunay triangulation algorithm to the pit, groove and crack point cloud to generate a triangular mesh of the pit, groove and crack, wherein the pit, groove and crack point cloud includes specific location and height information of the pit, groove and crack;

[0075] Step S42: using an interpolation or fitting method to optimize the generated triangular mesh to obtain a surface model and achieve surface reconstruction;

[0076] Step S43: performing post-processing on the surface model, the post-processing including smoothing, simplification or refinement;

[0077] Step S44: Obtain a three-dimensional solid model of the pits and cracks and output it.

[0078] Compared with the prior art, the present invention has the following advantages:

[0079] 1. Enhanced damage identification accuracy: The present invention analyzes point cloud data through a clustering algorithm, which can more accurately identify flat road surface point clouds, thereby more accurately locating potholes and cracks on the road surface, thereby improving the accuracy of damage identification.

[0080] 2. Automated and intelligent processing: The present invention not only realizes the automated processing of point cloud data, but also reduces manual intervention through intelligent clustering methods, further improving the efficiency and accuracy of data processing.

[0081] 3. High efficiency and real-time performance: After adopting the clustering algorithm, the present invention can process a large amount of point cloud data in a short time, realize rapid detection and evaluation of road damage, provide support for timely maintenance, and enhance real-time performance.

[0082] 4. Non-contact measurement: The present invention adopts non-contact measurement methods such as laser scanning, which can complete road surface detection without interfering with traffic, avoiding the impact on traffic flow and ensuring the safety of operation.

[0083] 5. Visual analysis: The three-dimensional model of road potholes and cracks generated by the present invention provides an intuitive visual basis for subsequent road repair and maintenance, enabling managers to better analyze the distribution and severity of road damage.

[0084] 6. Adapt to complex terrain: The present invention is based on three-dimensional point cloud data processing technology, which can adapt to road damage detection under various complex terrains and provide a wider range of application scenarios.

[0085] 7. Accurately assess road damage characteristics: Through precise three-dimensional models, characteristic parameters such as pothole depth, area and volume can be calculated in detail, providing a more accurate basis for road maintenance.

[0086] 8. Improve the robustness of data processing: Through the clustering algorithm, the present invention can better handle the noise and outliers in the point cloud data, thereby improving the robustness of data processing.

[0087] 9. Reduce misjudgment and missed judgment: The present invention greatly reduces the possibility of misjudgment and missed judgment and improves the accuracy of damage detection by considering the similarity of the eigenvalues ​​of each point and its nearest 50 neighboring points.

[0088] 10. Improve resource utilization efficiency: By accurately identifying the level road surface point cloud, the present invention can more effectively eliminate non-damaged point clouds, saving time and computing resources for subsequent analysis and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1It is a schematic flow chart of a method for establishing a three-dimensional model of road pothole cracks based on point cloud data in an embodiment of the present invention;

[0090] Figure 2 1 is a schematic diagram of the process of collecting original point cloud data of roads in an embodiment of the present invention;

[0091] Figure 3 It is a schematic diagram of the preprocessing process of the original point cloud data of the road in an embodiment of the present invention;

[0092] FIG4( a ) is an example diagram of point cloud data before Gaussian smoothing in data preprocessing in an embodiment of the present invention, wherein (1) is a plan view and (2) is an oblique view;

[0093] FIG4( b ) is an example diagram of point cloud data after Gaussian smoothing in data preprocessing in an embodiment of the present invention, wherein (1) is a plan view and (2) is an oblique view;

[0094] FIG5 (a) is an oblique view example of road point cloud data after data preprocessing in an embodiment of the present invention;

[0095] FIG5( b ) is a close-up enlarged example of a pothole in the road surface point cloud data after data preprocessing in an embodiment of the present invention;

[0096] FIG6 (a) is an oblique view of an example of a pit point cloud data recognition result according to an embodiment of the present invention;

[0097] FIG6( b ) is an example diagram of a magnified view of a pit portion of a pit point cloud data recognition result according to an embodiment of the present invention;

[0098] Figure 7 is an example diagram of the pit point cloud data identified in an embodiment of the present invention;

[0099] FIG8 (a) is an oblique view of a three-dimensional model result of a pit in an embodiment of the present invention;

[0100] FIG8( b ) is a plan view of an example of a three-dimensional model result of a pit in an embodiment of the present invention. DETAILED DESCRIPTION

[0101] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0102] like Figure 1 As shown, this embodiment provides a technical solution: a method for establishing a three-dimensional model of road pothole cracks based on point cloud data, comprising the following steps:

[0103] Step 1: Data collection and processing

[0104] Collect the original laser point cloud data of the road according to the pre-planned collection route and collection sequence;

[0105] Step 2: Preprocessing of original point cloud data

[0106] Preprocessing the original laser point cloud data of the road to obtain a local road surface point cloud with potholes and cracks;

[0107] Step 3: Pothole and Crack Point Cloud Identification

[0108] Perform local plane fitting on each point of the local road surface point cloud with potholes and cracks, calculate the angle between the local plane normal vector of each point and the reference normal vector according to the reference normal vector of the road surface and assign a eigenvalue, identify the flat road surface point cloud based on the eigenvalue of each point and use the clustering algorithm, then remove the flat road surface point cloud to obtain the pothole and crack point cloud;

[0109] Step 4: Establishment of three-dimensional model of pits and cracks

[0110] The Delaunay triangulation algorithm is used to generate the triangular mesh of the pits and cracks on the point cloud, and then the surface is reconstructed to finally obtain the three-dimensional solid model of the pits and cracks.

[0111] In step 1, the original laser point cloud data collection process of the road is as follows: Figure 2 As shown:

[0112] (1) Deploy a set number of target control points around the road;

[0113] (2) Use RTK measuring instruments and leveling instruments to obtain the plane coordinates and elevations of the target control points;

[0114] (3) Use a laser scanning system to obtain original laser point cloud data of the road;

[0115] (4) Select preset target control points to check the plane and elevation accuracy of the original laser point cloud data of the road.

[0116] In this embodiment, the target control points refer to specific locations that actually exist in the field or are artificially set, and they play a key role in the positioning and calibration process of the scan data.

[0117] The layout of target control points should follow the following guidelines:

[0118] Layout density: Target control points should be evenly arranged in the scanning area to ensure the measurement and calibration accuracy of the entire area. The layout density should be determined according to the scanning objectives and accuracy requirements;

[0119] Layout location: Target control points should be placed at key locations, such as significant feature points, intersections or edges of the ground to provide a stable reference frame to facilitate data calibration and positioning;

[0120] Number of target control points: A sufficient number of target control points should be arranged to meet the measurement requirements. It is usually recommended to set at least 3 to ensure that sufficient positioning and calibration information is provided.

[0121] Layout accuracy: The spatial position accuracy of the target control point must reach the millimeter or sub-millimeter level to ensure the accuracy of the scanning data;

[0122] In step 2, the preprocessing process of the original laser point cloud data of the road is as follows: Figure 3 As shown, the specific process is as follows:

[0123] The collected original road point cloud data includes not only road surface point clouds, but also useless point clouds such as surrounding buildings and trees. The premise for establishing a three-dimensional model is to determine and filter out the pothole and crack point clouds. In this embodiment, the following only describes the specific process of this embodiment with potholes as the target.

[0124] First, the CSF simulated cloth filtering algorithm is used to separate road points and non-road points. The principle of the CSF simulated cloth filtering algorithm is to invert the point cloud first, then project the inverted point cloud data and the cloth nodes simulated by the user onto the same horizontal plane, and find the point corresponding to the cloth node in the point cloud data, and record its height as h; in each iteration, compare the current height of the cloth node and the size of h, until the absolute value of the difference between h and the current height is less than a certain value, or the maximum number of iterations is reached, the iteration is terminated. Finally, the distance between the inverted point cloud data point at the end of the iteration and the cloth node is calculated. When the distance is less than a specific threshold, the point is recorded as a road point, otherwise it is recorded as a non-road point and filtered.

[0125] Secondly, the statistical filtering algorithm is used to remove high-frequency noise points in the road point cloud data. Statistical filtering performs statistical analysis on the neighborhood of each data point to remove obvious outliers. Outliers are often introduced by measurement noise, which is characterized by sparse distribution in space. The average distance from each data point to its nearest K (hypothetical) data points is calculated. The distances of all data points in the point cloud should form a Gaussian distribution. By giving the mean and variance, the standard deviation can be eliminated. Data points outside of .

[0126] In this embodiment, the calculation process of statistical filtering is as follows:

[0127] The neighborhood of each data point is statistically analyzed, and the distances of all data points in the point cloud data are assumed to form a Gaussian distribution, whose shape is determined by the mean and standard deviation Determine; let the coordinates of the nth data point in the point cloud data be , from this data point to any data point The distance is:

[0128] ;

[0129] The formula for calculating the average distance between each data point and any data point is:

[0130] ;

[0131] Let the standard deviation be:

[0132] ;

[0133] Let the standard deviation be , only K and Two thresholds, when the average distance of a point to K points is within the standard range If it is within the range, the point is retained; if it is not, it is defined as an outlier and deleted.

[0134] Then, a Gaussian filter is used to smooth the road point cloud data. Gaussian filtering is a commonly used point cloud data smoothing technology. It uses a Gaussian kernel to reduce noise in the data to achieve a smoothing effect. In the processing of road point cloud data, Gaussian filtering can effectively smooth the road point cloud and reduce data fluctuations caused by uneven road surfaces, which is crucial for subsequent tasks such as pothole identification. Figures 4 (a) and 4 (b) are comparisons before and after the Gaussian smoothing effect.

[0135] In this embodiment, the calculation process of Gaussian smoothing is as follows:

[0136] Set a radius value r to define the neighborhood range for selecting each point.

[0137] For each point, a Gaussian function is used to calculate the weight of each point in the neighborhood. The weight is determined by the distance from the point to the center point. The closer the distance, the greater the weight.

[0138] The new smoothed point position is the weighted average of all points in its neighborhood, using the following formula:

[0139] ;

[0140] in, is the smoothed point, is a point in the neighborhood, is the corresponding weight.

[0141] Repeat the above steps until all points are smoothed and output, ensuring that the generated point cloud data meets the required smoothing effect.

[0142] Next, due to the high density of the point cloud, the point cloud data is downsampled using a voxel filter to obtain a simplified point cloud of the road. This method divides the point cloud into a regular voxel grid, where each voxel represents a volume unit in the point cloud data. Within each voxel, a representative point (the point closest to the voxel center) is selected and retained, while other points are ignored, thereby simplifying the data. Through the voxel filter, the points in the dense area of ​​the point cloud data are effectively reduced, while retaining the overall shape and surface features of the point cloud at the voxel scale.

[0143] In this embodiment, the specific process of downsampling point cloud data using a voxel filter is as follows:

[0144] 1): Initialize the filter parameters, including the size of the voxel; a voxel is a small cube in three-dimensional space, and the size of the voxel determines the granularity of the filter; smaller voxels can retain more details, but the amount of calculation is larger; larger voxels can reduce the number of points, but may lose details;

[0145] 2): Divide the smoothed point cloud data into corresponding voxels, traverse each data point in the point cloud data, determine the voxel to which it belongs according to the coordinates of the data point, and map the coordinates of the data point to the voxel grid. Each voxel contains data points within a certain range.

[0146] 3): Process the data points in each voxel. For each non-empty voxel, use the point closest to the voxel center to represent the data point in the voxel;

[0147] 4): Generate a new point cloud dataset based on the processing results. This new point cloud dataset will contain fewer data points while retaining the main features of the original point cloud as much as possible;

[0148] 5): Output the point cloud dataset processed by the voxel filter as input for subsequent processing steps.

[0149] Finally, a small part of the road surface point cloud containing potholes is manually found and selected, which can significantly improve the computational efficiency of subsequent processing steps and reduce storage requirements. The preprocessing results are shown in Figure 5 (a) and Figure 5 (b).

[0150] In step 3, the process of pothole point cloud recognition is as follows:

[0151] (1) Calculate the local plane normal vector

[0152] First, a local plane is fitted to each point in the point cloud. This usually involves selecting a local neighborhood around each point and using the least squares method or other statistical methods to estimate a best-fit plane within this neighborhood. By fitting the resulting plane, we can calculate the local plane normal vector for each point.

[0153] As a bit more specific, the local plane normal vector is calculated as follows:

[0154] 1): Each point in the point cloud The neighborhood point set of ; The neighborhood point set includes the point and k neighboring points;

[0155] 2): Using neighborhood point sets The eigenvalue method is used to calculate all points in the local plane to obtain the local plane normal vector set ;

[0156] 3): The calculated local plane normal vector With point cloud The associated data are used for subsequent pothole point cloud recognition.

[0157] (2) Pothole point cloud recognition

[0158] First, set a corresponding relationship between eigenvalue and angle, assuming that the angle corresponding to 100 eigenvalues ​​is 1″;

[0159] Then the angle between the local plane normal vector of each point and the reference vector perpendicular to the ground is calculated and converted into an eigenvalue;

[0160] Then, for each point, consider its nearest 50 neighboring points. If the number of points with similar feature values ​​among the point and its 50 neighboring points reaches or exceeds 90%, these points are classified into the same category, that is, the normal vectors of these points are considered to be basically perpendicular to the ground and are identified as flat road surface point clouds.

[0161] As more specific, the recognition process of the pothole point cloud is as follows:

[0162] 1): Calculate the angle between each point and the reference normal vector. The angle calculation formula is as follows:

[0163]

[0164] in, is the reference normal vector, .

[0165] 2): Calculate the eigenvalue of each point. The eigenvalue calculation formula is as follows:

[0166]

[0167] A threshold T is set to filter out the flat road surface point cloud. The threshold T is determined according to the characteristics of the point cloud data and specific requirements. In this embodiment, T=200.

[0168] 3): Compare the eigenvalues ​​of each point, filter out points with eigenvalues ​​≤ 200, and mark them as preliminary flat road surface point clouds; through the first screening, points that meet the eigenvalue conditions can be accurately screened out from the local road surface point cloud.

[0169] 4): Further screening, for each point marked as preliminary flat road point cloud, consider its nearest 50 neighboring points. If the number of points marked as preliminary flat road point cloud in this point and its 50 neighboring points reaches or exceeds 90%, these points are divided into the same category, that is, the local plane normal vectors of these points are considered to be basically perpendicular to the ground, and they are identified as the final flat road point cloud. The purpose of clustering is to accurately remove some points that are not flat road point clouds but exist in the results of the first screening (preliminary flat road point clouds). These points are located at the bottom of potholes (or cracks) and are relatively "flat" partial point clouds. Through the second screening, these points can be accurately removed to obtain the final flat road point cloud.

[0170] 5): Inversely select the point cloud, identify all the pit point clouds, and build an index list of the pit point clouds to prepare for the subsequent modeling steps. The recognition results are shown in Figure 6 (a) and Figure 6 (b).

[0171] Finally, the flat road surface point cloud is removed, and the remaining point clouds are regarded as pothole point clouds. The results are as follows: Figure 7 As shown, these points represent the pavement damage areas and provide a basis for subsequent modeling.

[0172] In step 4, the process of establishing the three-dimensional solid model of the pits and cracks is as follows:

[0173] 1): First, the Delaunay triangulation algorithm is applied to the pit point cloud to generate a triangular mesh of the pit. This algorithm connects the points in the point cloud to form a series of triangles, ensuring that the generated mesh has good geometric properties;

[0174] In this step, the process of generating the triangular mesh of the pit is as follows:

[0175] (1) First, it is necessary to determine the input data of the Delaunay triangulation algorithm, that is, the pit point cloud data after preprocessing and identification. These point cloud data contain the specific location and height information of the pit.

[0176] (2) The input pit point cloud data is processed using the Delaunay triangulation algorithm. Delaunay triangulation is a method for triangulating scattered data points on a plane. It can ensure that the triangles formed after triangulation are as close to equilateral triangles as possible, thereby reducing errors. During the triangulation process, the algorithm creates a triangular mesh in which the circumscribed circle of each triangle does not contain other data points, thus ensuring the quality of the triangular mesh. Finally, a high-quality triangular mesh is generated, laying the foundation for subsequent surface reconstruction and the establishment of a three-dimensional solid model.

[0177] 2): Use interpolation or fitting methods to optimize the generated triangular mesh to obtain a surface model and achieve surface reconstruction. Triangles are surface elements, and giving each triangle the same color completes the basic surface reconstruction. This model conforms to the actual point cloud characteristics;

[0178] In this step, the process of optimizing the generated triangular mesh using interpolation or fitting methods is as follows:

[0179] (1) First, it is necessary to select a suitable interpolation or fitting method based on the characteristics of the pits and the data characteristics of the triangular mesh. Common methods include linear interpolation, polynomial fitting, spline interpolation, etc. The selected method should be able to accurately reflect the geometric characteristics of the cracks and improve the smoothness and accuracy of the model.

[0180] For linear interpolation, you can calculate the straight line segment between two vertices to determine the location of the middle point; for polynomial fitting, you can build a polynomial model based on a set of points in the grid to predict and adjust the location of other points. For spline interpolation, you can use a smooth curve (such as a cubic spline) to fit the points in the grid to maintain overall smoothness.

[0181] (2) Apply the selected interpolation or fitting algorithm to the vertices and edges of the triangular mesh; this involves mathematically processing each point or edge in the mesh to ensure the continuity and smoothness of the mesh.

[0182] (3) Finally, an optimized triangular mesh is generated.

[0183] 3): Then the surface model is post-processed. In this implementation, the mesh doctor function in Geomagic Wrap software is used to smooth, simplify, and repair holes on the model to further improve the visual effect and accuracy of the model. The post-processing results are shown in Figure 8 (a) and Figure 8 (b);

[0184] 4): Finally, the 3D solid model of the pothole is output as the basis for subsequent analysis and application. This model can be used to evaluate the depth, area and other characteristics of the pothole and provide data support for road maintenance.

[0185] It should be noted that if "cracks" are used as the target instead of "pits", the implementation process of the entire method will be different. The specific differences are as follows:

[0186] 1. In step 1, data collection for crack targets requires higher resolution and finer scanning parameters than for pit targets, because cracks are usually finer than pits and require more accurate data to capture their shape and position. When laying out target control points, a denser arrangement may be required to ensure that the details of the cracks can be accurately captured and calibrated.

[0187] 2. In step 4, when using the Delaunay triangulation algorithm to generate the triangular mesh of the crack, the algorithm parameters need to be adjusted to adapt to the slender shape of the crack.

[0188] In summary, the method for establishing a three-dimensional model of road potholes and cracks based on point cloud data in the above-mentioned embodiment, by accurately collecting road point cloud data, combining an efficient preprocessing algorithm for data smoothing and denoising, and then generating a high-precision three-dimensional model through plane fitting and Delaunay triangulation, can accurately identify and quantify potholes and cracks on the road surface. Compared with traditional methods, the present invention has higher detection accuracy, automation level and processing efficiency, and effectively solves the problems of difficulty in comprehensive evaluation and slow detection speed in road surface detection. In addition, the non-contact laser scanning method also ensures the safety and convenience of the measurement process. The final generated three-dimensional model can intuitively display the depth, area and volume of road damage, providing a scientific basis for road maintenance decisions, thereby improving the pertinence and cost-effectiveness of maintenance work.

[0189] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for establishing a three-dimensional model of road potholes and cracks based on point cloud data, characterized in that: The following steps are involved: Step S1: Data collection and processing Collect the original laser point cloud data of the road according to the pre-planned collection route and collection sequence; Step S2: Preprocessing of original point cloud data Preprocess the original laser point cloud data of the road to obtain the local road surface point cloud with potholes and cracks; Step S3: Pothole and Crack Point Cloud Identification Perform local plane fitting on each point of the local road surface point cloud with potholes and cracks, calculate the angle between the local plane normal vector of each point and the reference normal vector according to the reference normal vector of the road surface and assign a eigenvalue, identify the flat road surface point cloud based on the eigenvalue of each point and use the clustering algorithm, then remove the flat road surface point cloud to obtain the pothole and crack point cloud; Step S4: Establishment of three-dimensional model of pits and cracks The Delaunay triangulation algorithm is used to generate the triangular mesh of the pits and cracks on the point cloud, and then the surface is reconstructed to finally obtain the three-dimensional solid model of the pits and cracks.

2. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 1, characterized in that: In step S1, the original laser point cloud data collection process of the road is as follows: Step S11: Arrange a set number of target control points around the road; Step S12: using an RTK measuring instrument and a leveling measuring instrument to obtain the plane coordinates and elevation of the target control point; Step S13: using a laser scanning system to obtain original laser point cloud data of the road; Step S14: Select preset target control points to check the plane and elevation accuracy of the original laser point cloud data of the road.

3. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 2, characterized in that: In step S2, the process of preprocessing the original laser point cloud data of the road is as follows: Step S21: using the CSF simulated cloth filtering algorithm to separate the road surface points and non-road surface points in the original laser point cloud data of the road to obtain a road surface point cloud; Step S22: using a statistical filtering algorithm to remove high-frequency noise points in the road surface point cloud; Step S23: using a Gaussian filter to smooth the road surface point cloud processed in step S22; Step S24: downsampling the smoothed road surface point cloud using a voxel filter to obtain a simplified point cloud; Step S25: Manually search and cut out the local road surface point cloud with potholes and cracks.

4. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 3, characterized in that: In step S21, the process of using the CSF simulated cloth filter algorithm to separate road surface points and non-road surface points is as follows: Step S211: invert the original laser point cloud data of the road, project the inverted point cloud data and the cloth nodes simulated by the user onto the same horizontal plane, and find the point corresponding to the cloth node in the point cloud data, and record its height as h; Step S212: During each iteration, compare the current height of the cloth node with the size of h, until the absolute value of the difference between h and the current height is less than a preset value or the maximum number of iterations is reached, the iteration is terminated; Step S213: Calculate the distance between the inverted point cloud data point at the end of the iteration and the cloth node. When the distance is greater than the set threshold, the point is recorded as a road point and retained. Otherwise, it is recorded as a non-road point and filtered out, thereby obtaining a road surface point cloud.

5. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 3, characterized in that: In step S22, the process of removing high-frequency noise points in the road surface point cloud using a statistical filtering algorithm is as follows: Step S221: Perform statistical analysis on the neighborhood of each road point. The distances between all data points in the road point cloud data are assumed to form a Gaussian distribution, whose shape is determined by the mean and standard deviation Decision: Assume that the coordinates of the nth data point in the road point cloud data are , from this data point to any data point The distance is: ； Step S222: Calculate the average distance between each data point and any data point. The formula is: ； Step S223: Set the standard deviation to: ； Step S224: Set the standard deviation multiple to , when the average distance of a data point to k data points is within the standard range If it is within the range, the point is retained; if it is not, it is defined as an outlier and deleted.

6. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 3, characterized in that: In step S23, the process of smoothing the road point cloud using a Gaussian filter is as follows: Step S231: Select the neighborhood range of each point to determine the neighborhood points for smoothing, and the neighborhood range is defined by setting a radius value r; Step S232: For each point, use the Gaussian function to calculate the weight of each point in the neighborhood. The weight is determined according to the distance from the point to the center point. The closer the distance, the greater the weight. Step S233: For each point, replace it with the weighted average of all points in its neighborhood to obtain a new smooth point. The position of the new smooth point is obtained by the following formula: ； in, is the smoothed point, is a point in the neighborhood, is the corresponding weight; Step S234: repeating steps S231 to S233 until smoothing of all points is completed; Step S235: Output the road point cloud data smoothed by the Gaussian filter.

7. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 1, characterized in that: In step S3, the process of identifying the pit, crack and point cloud is as follows: Step S31: using the eigenvalue method to calculate the local plane normal vector for each point in the local road surface point cloud; Step S32: Calculate the angle between the local plane normal vector of each point and the reference normal vector and assign an eigenvalue, use the eigenvalue to perform the first screening, identify the preliminary flat road surface point cloud, and then use the clustering algorithm to perform the second screening, and identify the final flat road surface point cloud from the preliminary flat road surface point cloud; Step S33: The flat road surface point cloud finally identified in step S32 is eliminated, and the remaining part is the pothole and crack point cloud.

8. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 7, characterized in that: In step S31, the calculation process of the local plane normal vector is as follows: Step S311: Each point in the local road point cloud The neighborhood point set of ; The neighborhood point set includes the point and k neighboring points; Step S312: Using the neighborhood point set The eigenvalue method is used to calculate all points in the local plane to obtain the local plane normal vector set ; Step S313: The local plane normal vector of each point is calculated With point cloud Associated.

9. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 8, characterized in that: In step S32, the specific processing process is as follows: Step S321: Angle calculation Calculate the angle between each point and the reference normal vector. The angle calculation formula is as follows: in, is the reference normal vector, ; Step S322: Eigenvalue conversion Calculate the eigenvalue of each point. The eigenvalue calculation formula is as follows: ； Step S323: Primary screening Compare the eigenvalues ​​of each point, select the points with eigenvalues ​​≤ T, and mark them as preliminary flat road surface point clouds; Step S324: Secondary screening For each point marked as a preliminary smooth road point cloud, consider its nearest 50 neighboring points. If the number of points marked as preliminary smooth road point clouds among this point and its 50 neighboring points reaches or exceeds 90%, then the point is marked as the final smooth road point cloud. Otherwise, the corresponding point is deleted from the preliminary smooth road point cloud to obtain the final smooth road point cloud.

10. The method for establishing a three-dimensional model of road potholes and cracks based on point cloud data according to claim 9, characterized in that: In step S4, the process of establishing the three-dimensional model of pits and cracks is as follows: Step S41: applying the Delaunay triangulation algorithm to the pit, groove and crack point cloud to generate a triangular mesh of the pit, groove and crack, wherein the pit, groove and crack point cloud includes specific location and height information of the pit, groove and crack; Step S42: using an interpolation or fitting method to optimize the generated triangular mesh to obtain a surface model and achieve surface reconstruction; Step S43: post-processing the surface model, the post-processing including smoothing, simplification or refinement; Step S44: Obtain a three-dimensional solid model of the pits and cracks and output it.

Images