Pavement skid resistance detection device and detection method

Abstract

The invention relates to the technical field of road surface detection, in particular to a road surface skid resistance detection device and method, and the method comprises the steps: obtaining three-dimensional point cloud data of a to-be-detected road surface, and carrying out the down-sampling processing to obtain each point cloud voxel; curvature features, normal vector features and density features of the point cloud voxels are determined, and geometric features of the point cloud voxels are formed; dividing all the point cloud voxels into various types, and extracting each connected region of the point cloud voxels in the various types; determining a self-similarity value of each connected region through the distribution of the multi-scale fractal dimension of each connected region; obtaining the noisy point saliency of each connected region; performing statistical filtering on the point cloud voxels of the connected regions, and correcting a filtering threshold value in the statistical filtering by using noisy point saliency; and constructing a three-dimensional digital model of the road surface through the denoised point cloud voxels and the optical image data of the road surface, and detecting the skid resistance of the road surface. Therefore, the precision of pavement skid resistance detection is improved.

Description

Technical Field

[0001] This application relates to the field of road surface testing technology, specifically to a road surface skid resistance testing device and testing method. Background Technology

[0002] The anti-skid performance of road surfaces is a crucial indicator of road safety. It refers to the anti-skid force generated during tire-road slippage when a vehicle brakes. Anti-skid performance is primarily determined by the macroscopic and microscopic structure of the road surface, namely the texture between road material particles and the surface texture of the particles themselves. In today's society with rapidly increasing traffic volume, road surfaces with good anti-skid performance are a key factor in reducing the number of traffic accidents caused by slippery road surfaces. Traditional road surface anti-skid testing techniques are mostly contact-based methods, which are complex, time-consuming, and significantly disruptive to traffic. Furthermore, these methods often only assess the actual condition of a single road surface area, failing to meet the modern traffic demand for rapid testing of large-area road surface performance.

[0003] Currently, some emerging detection technologies, such as 3D laser scanning and digital image construction depth detection, can use optical equipment to reconstruct the 3D topography of the road surface texture at the micron level, analyze the road surface construction depth information, and achieve non-contact, large-area road surface anti-skid performance testing. However, 3D laser scanning is easily affected by external factors when collecting road surface texture features. The vibration effect of the vehicle carrying the equipment during movement can cause the acquired 3D laser point cloud data to shift, and the point cloud data contains a large amount of high-frequency abnormal noise data, resulting in poor quality of the 3D point cloud data of the road surface. This affects the accuracy of the assessment of the road surface construction depth information and reduces the accuracy of road surface anti-skid performance testing. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a road surface skid resistance testing device and method, the specific technical solution of which is as follows: In a first aspect, embodiments of this application provide a method for testing the anti-skid performance of road surfaces, the method comprising the following steps: The three-dimensional point cloud data of the road surface to be detected is acquired, and the point cloud voxels are obtained after downsampling processing. Determine the curvature features, normal vector features, and density features of each point cloud voxel to form the geometric features of each point cloud voxel; based on the similarity between the geometric features of the point cloud voxels, classify all point cloud voxels into categories and extract each connected region of the point cloud voxels in each category. By analyzing the distribution of the multi-scale fractal dimension of each connected region, the self-similarity value of each connected region is determined; the fitted surface of all point cloud voxels in each connected region is obtained, the dispersion of the fitting residuals of point cloud voxels in each connected region is analyzed, and the noise saliency of each connected region is determined by combining the self-similarity value. Statistical filtering is performed on the point cloud voxels of each connected region, and the filtering threshold in the statistical filtering is corrected using the noise saliency. A three-dimensional digital model of the road surface is constructed by combining the denoised point cloud voxels with the optical image data of the road surface, and the anti-skid performance of the road surface is tested.

[0005] In one embodiment, the curvature feature is obtained by using the covariance matrix method to obtain the curvature of the centroid point cloud in the point cloud voxel, and this curvature feature is used as the curvature feature.

[0006] In one embodiment, the normal vector feature is the eigenvector corresponding to the smallest eigenvalue in the covariance matrix.

[0007] In one embodiment, the density feature is the percentage of the number of three-dimensional point clouds in each point cloud voxel.

[0008] In one embodiment, determining the self-similarity value of each connected region includes: The degree of dispersion of the multi-scale fractal dimension of each connected region is calculated and denoted as the first degree of dispersion. The mean of the multi-scale fractal dimension of each connected region is calculated. The self-similarity value is negatively correlated with the mean and negatively correlated with the first degree of dispersion.

[0009] In one embodiment, the expression for the self-similarity value is: In the formula, Let be the self-similarity value of the h-th connected region, and exp() be an exponential function with base 10 ... Let be the mean of the multiscale fractal dimension of the h-th connected region. denoted as the degree of dispersion of the multiscale fractal dimension of the h-th connected region.

[0010] In one embodiment, determining the noise saliency of each connected region includes: Divide the range of minimum and maximum values ​​of the fitting residuals of point cloud voxels in each connected region into local intervals. Count the number of point cloud voxels whose fitting residuals fall within each local interval. Calculate the information entropy of the number of point cloud voxels in all local intervals corresponding to each connected region. Combine this with the self-similarity value to obtain the noise saliency of each connected region.

[0011] In one embodiment, the expression for correcting the filter threshold in the statistical filtering using the noise saliency is: In the formula, The threshold value is the statistically corrected filter threshold, and e is the natural constant. Let h be the noise saliency of the h-th connected region. This is the initial threshold for statistical filtering.

[0012] In one embodiment, the testing of the anti-skid performance of the road surface includes: The average construction depth of the road surface is calculated using a three-dimensional digital model. If the average construction depth meets the road surface construction standards, the anti-skid performance of the road surface is qualified; otherwise, the anti-skid performance is unqualified.

[0013] Secondly, embodiments of this application also provide a road surface skid resistance testing device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] This application has at least the following beneficial effects: This application acquires 3D point cloud data of the road surface and, by combining downsampling and voxel segmentation, reduces the computational load while accurately capturing minute geometric features of the road surface. Through comprehensive analysis of the curvature, normal vector, and density features of each point cloud voxel, the morphological characteristics of the road surface can be fully evaluated. This multi-dimensional geometric feature analysis helps reveal the relationship between the road surface microstructure and anti-skid performance, improving the comprehensiveness and depth of anti-skid performance detection and avoiding misjudgments that may be caused by a single feature. Based on the similarity between the geometric features of point cloud voxels, automatic classification of point cloud voxels is performed, effectively distinguishing different types of road surface areas and enabling anti-skid performance assessments for different types of road surfaces. Analysis enhances the adaptability and efficiency of anti-skid performance testing. By analyzing the multi-scale fractal dimension distribution of each connected region, the self-similarity value can be quantitatively determined, which helps improve the accuracy of noise influence analysis within the connected region, determine the noise significance of each connected region, and adaptively adjust the threshold of statistical filtering, avoiding over-filtering or under-filtering that can easily occur with a fixed threshold. This improves the accuracy of filtering point cloud data. By combining the denoised road surface point cloud voxels with optical image data, a three-dimensional digital model is constructed, which not only provides road surface morphology information but also enables a comprehensive analysis of road surface anti-skid performance, thus improving the reliability and accuracy of road surface anti-skid performance testing. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the steps of a method for testing the anti-skid performance of a road surface, as provided in one embodiment of this application; Figure 2 The flowchart shows the adaptive filtering threshold determination process for statistical filtering algorithms. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a road surface skid resistance testing device and testing method proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the road surface skid resistance testing device and testing method provided in this application.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for testing the anti-skid performance of a road surface according to an embodiment of this application. The method includes the following steps: S1: Acquire the three-dimensional point cloud data of the road surface to be detected, and obtain each point cloud voxel after downsampling processing.

[0021] A 3D laser scanner is used to scan the road surface to be inspected, and 3D point cloud data of the road surface to be inspected is obtained.

[0022] Because 3D laser scanners are susceptible to external interference, such as vibration, the accuracy and validity of point cloud data at the microscopic level can be compromised by noise. Traditional statistical filtering algorithms are sensitive to changes in point cloud density when filtering and denoising point cloud data. Since texture variations in the road surface can easily be misjudged as noise, the filtering process can lead to over-smoothing of the point cloud data. In the process of eliminating noise, some normal road surface textures are also excessively smoothed, making it difficult for the point cloud data to effectively reflect the true texture, structural depth, and other characteristics of the road surface. This, in turn, affects the accuracy of the road surface skid resistance performance test.

[0023] Based on the above analysis, firstly, to avoid excessive computational resource consumption from processing massive point cloud data and affecting the modeling speed of the 3D digital model of the road surface, the acquired point cloud data is downsampled using voxel meshing, with the size of each voxel mesh set to... The size of the voxel mesh can be set by the implementer according to the actual situation. This embodiment does not impose any special restrictions on this. .

[0024] After voxelizing the point cloud data, the road surface point cloud data can be represented as follows: ,in, Indicates the first Each point cloud voxel contains a number of points, and the number of points contained in each voxel varies depending on the characteristics of different areas of the road surface.

[0025] S2, determine the curvature features, normal vector features, and density features of each point cloud voxel to form the geometric features of each point cloud voxel; based on the similarity between the geometric features of the point cloud voxels, classify all point cloud voxels into categories and extract each connected region of the point cloud voxels in each category.

[0026] In this embodiment, the geometric features of the centroid point cloud within each point cloud voxel are taken as the geometric features of that voxel. Finding the centroid of a point cloud voxel is a well-known technique, and its specific process will not be elaborated further. In this embodiment, the geometric features include curvature features, normal vector features, and density features. The curvature feature is obtained by calculating the covariance matrix of the centroid point cloud within the point cloud voxel, and then performing eigenvalue decomposition on the covariance matrix to obtain each eigenvalue. The curvature of the point cloud voxel is then the ratio of the smallest eigenvalue to the sum of all eigenvalues. Calculating the curvature of the point cloud using the covariance matrix method is a well-known technique, and its specific process will not be elaborated further. The normal vector feature of a point cloud voxel is obtained by taking the eigenvector corresponding to the smallest eigenvalue in the eigenvalue decomposition of its covariance matrix as the normal vector of that point cloud voxel. Indicates the first The normal vector features of a point cloud voxel; the first Density characteristics of point cloud voxels for ,in, For the first The number of points in a point cloud voxel Let the side length of a point cloud voxel be the size of the voxel. Indicates the first Point cloud density in each point cloud voxel. Furthermore, the curvature features, normal vector features, and density features of each point cloud voxel are standardized to obtain the standardized curvature features. Normal vector characteristics Density characteristics , then the first Geometric features of a point cloud voxel Represented as ,in, For the first The curvature feature of the point cloud voxel after standardization. The spatial coordinate information of the point cloud voxel uses the coordinates of the centroid point cloud. A point cloud voxel Its spatial location information is the position coordinates of the point cloud in the grid coordinate system defined during gridding. .

[0027] Traditional statistical filtering algorithms are highly sensitive to point cloud density. While normal, homogeneous surface textures can effectively identify some noise points in road surface voxels, surface defects such as cracks and potholes cause continuous reflections of laser light in the shallow layers of these defects, leading to point cloud accumulation. This results in some defective areas exhibiting higher density compared to normal surface textures due to their geometric characteristics. Under this influence, even without noise points, statistical filtering, based on a fixed neighborhood size, can easily misclassify areas as noise due to the differences in features between homogeneous surface textures and defective areas within the neighborhood. This can lead to over-smoothing and loss of detail, ultimately degrading the overall quality of the point cloud data after denoising.

[0028] Based on the above analysis, considering the geometric feature similarity of point clouds, connected component merging is performed on each point cloud voxel. In subsequent filtering and denoising, adaptive filtering is performed on homogeneous regions with similar geometric features, thereby avoiding the problem of over-filtering and smoothing caused by traditional statistical filtering in this scenario.

[0029] Specifically, the geometric feature similarity between point cloud voxels can be expressed as: In the formula, This represents the nth point cloud voxel. This represents the m-th point cloud voxel. Let be the similarity between the nth point cloud voxel and the mth point cloud voxel. Represents the geometric features of the nth point cloud voxel. Let m represent the geometric features of the m-th point cloud voxel, and let cosine() be the cosine similarity calculation function.

[0030] Based on the geometric feature similarity between point cloud voxels, this embodiment uses the K-means clustering algorithm to classify all point cloud voxels. In this embodiment, K=5, but implementers can set it according to their actual situation; this embodiment does not impose any restrictions on this. The K-means clustering algorithm is a well-known existing technology, and implementers can choose other feasible clustering algorithms; this embodiment does not impose any restrictions on this. Each point cloud voxel cluster can be obtained using the K-means clustering algorithm. Then, for each point cloud voxel cluster, connectivity is established according to the spatial location information between the point cloud voxels to obtain each connected region.

[0031] S3. By analyzing the distribution of the multi-scale fractal dimension of each connected region, the self-similarity value of each connected region is determined; the fitted surface of all point cloud voxels in each connected region is obtained, the dispersion of the fitting residuals of point cloud voxels in each connected region is analyzed, and the noise saliency of each connected region is determined by combining the self-similarity value.

[0032] Traditional statistical filtering typically relies on uniform point cloud features within a region to set a threshold. However, for defective regions with anomalous geometric features, a uniform threshold may lead to difficulty in accurately removing anomalous noise data within the region, leaving isolated noise that affects the quality of the point cloud. To better characterize the degree to which each connected region is affected by noise factors, this paper considers the fitting residual features and fractal features of homogeneous textured pavement, defective pavement, and noise data to construct a noise saliency index for adaptive filtering of each region.

[0033] First, a self-similarity index is constructed by considering the fractal features in each connected region. Fractal features can represent the texture structure of road surface point cloud data at a certain scale. In this embodiment, the fractal dimension is used to represent its fractal features, and the range of the fractal dimension is (2, 3). For road surfaces with homogeneous textures, there is a certain regularity and stability under different scale windows, so its fractal dimension is low. However, for road surface point cloud data with defects or noise interference, the distribution is more chaotic and lacks continuous texture features at the micro level. For example, the distribution and geometric features of point clouds at the edge of cracks will lead to an increase in the fractal dimension under its local window, while the distribution of noise points in the local space is random and does not have regular statistical characteristics, so its fractal dimension is also high.

[0034] Based on the above analysis, for the obtained first... Connected regions The fractal dimension of the connected region is obtained using the difference box dimension algorithm. The difference box dimension algorithm is a well-known existing technology, and its specific process will not be elaborated here. A noise-free homogeneous textured road surface exhibits regular texture in windows of different scales. Therefore, this regularity is maintained across multiple scale windows. This embodiment evaluates the first-order difference box dimension by calculating the difference between the fractal dimensions of the multiple scale windows. Connected regions Self-similarity at multiple scales. Specifically, the size of the multi-scale window in this embodiment is set to a value of... In the multi-scale window, 2 indicates that the first scale is calculated using a window of size 0.2 cm. Connected regions The fractal dimension allows implementers to set the size of multi-scale windows according to actual conditions. Therefore, the first... Connected regions Five multi-scale fractal dimensions can be obtained. , construct the first Connected regions self-similarity value Specifically: Calculate the first Connected regions The degree of dispersion of the multiscale fractal dimension is denoted as the first degree of dispersion. The second degree of dispersion is calculated. Connected regions The mean of the multi-scale fractal dimension, wherein the self-similarity value is negatively correlated with the mean and negatively correlated with the first dispersion.

[0035] In this embodiment, the first Connected regions self-similarity value The expression is: In the formula, Let be the self-similarity value of the h-th connected region, and exp() be an exponential function with base 10 ... Let be the mean of the multiscale fractal dimension of the h-th connected region. Let be the degree of dispersion of the multiscale fractal dimension of the h-th connected region, denoted as the first degree of dispersion.

[0036] It should be noted that the degree of dispersion can be calculated using methods such as standard deviation and coefficient of variation. In this embodiment, standard deviation is used as the calculation method for the first degree of dispersion.

[0037] When self-similarity value A value close to 1 indicates a higher self-similarity of the h-th connected region, suggesting that the h-th connected region is a homogeneous surface region with less noise; conversely, a lower self-similarity value indicates a lower self-similarity. The closer it is to 0, the lower the self-similarity of the h-th connected region, which may be a defective road surface or a region containing a lot of noise information.

[0038] To further identify noise interference in defective regions, the fitting residual characteristics of each connected region are considered. Although defective regions are difficult to distinguish from noisy data in terms of fractal dimension self-similarity, they possess a certain degree of structure compared to regions heavily affected by noise.

[0039] For the h-th connected region Using the spatial location information of all point cloud voxels in the region, the least squares surface fitting method is used to fit the approximate surface equation of the region. Least squares surface fitting is a well-known technique, and its specific process will not be elaborated here. Then, the h-th connected region is calculated. The Middle A point cloud voxel The smaller the residual, the better the point cloud voxel fits the fitted surface; conversely, the larger the residual, the more the point cloud voxel deviates from the fitted surface. The calculation of the fitting residual is a well-known technique, and the specific process will not be elaborated here.

[0040] For homogeneous textured pavements, which are generally flat with only slight surface complexity due to the homogeneous texture at varying depths, the fitting residual distribution in the corresponding area is stable and exhibits low dispersion. For defective areas, taking cracks as an example, while the overall structure is complex, the residuals from point cloud voxels at different depth levels within the crack to the fitting surface are uniformly distributed according to their depth levels. Although the dispersion of the fitting residuals is relatively large, the distribution is a structured distribution with depth levels, thus maintaining stability. For areas containing noisy data, due to the randomness of noise distribution, the corresponding fitting residual distribution in these areas is more discrete.

[0041] Based on the above analysis, taking the h-th connected region as an example... For example, find the region The fitting residuals of all point cloud voxels are divided into intervals of equal length by the range of maximum and minimum values ​​of the fitting residuals. For each local interval, count the number of point cloud voxels whose fitting residuals fall within each local interval, and calculate the h-th connected region. Information entropy of the number of point cloud voxels in all corresponding local intervals Information entropy The larger the entropy, the more discrete and complex the fitting residual distribution of the point cloud voxels in the connected region, containing a large amount of noise interference; conversely, the smaller the entropy, the more complex the information entropy. A smaller value indicates that the fitting residual distribution of point cloud voxels in the connected region is stable, with fewer noisy data points, and the structural integrity of the fitting residual distribution is not destroyed by the noisy data. In this embodiment, B=10 is set, but implementers can set it according to their actual situation; this embodiment does not impose any restrictions on this.

[0042] Based on the above analysis, the noise saliency of each connected region is determined by combining the information entropy and the self-similarity value of each connected region. Specifically, the noise saliency of each connected region is negatively correlated with the self-similarity value and positively correlated with the information entropy.

[0043] In this embodiment, the h-th connected region noise saliency The expression is: In the formula, Let h be the self-similarity value of the h-th connected region. Let be the information entropy of the h-th connected region.

[0044] Noise saliency is modulated by both the self-similarity value and information entropy of the connected region, reflecting the noise characteristics of the connected region. When the self-similarity value... Close to 1, and information entropy The smaller the size, the more homogeneous the textured surface area, and the less noise interference it has, resulting in lower noise saliency. The overall value is relatively small, so the filtering strength needs to be reduced during subsequent filtering to avoid over-smoothing of the point cloud data; when the self-similarity value Approaching 0, while information entropy When the value is large, then the noise data in that connected region has a significant impact. The noise level is relatively high, so the filtering strength needs to be increased in subsequent filtering to achieve a better noise reduction effect.

[0045] S4. Perform statistical filtering on the point cloud voxels of each connected region, and use the noise saliency to correct the filtering threshold in the statistical filtering; construct a three-dimensional digital model of the road surface using the denoised point cloud voxels and the optical image data of the road surface, and detect the anti-skid performance of the road surface.

[0046] Finally, adaptive statistical filtering is performed on each connected region based on the noise saliency of each region. In this embodiment, the number of point cloud voxels in the neighborhood of each point cloud voxel is set to 50 during the statistical filtering process. Implementers can set this number according to their actual situation; this embodiment does not impose any restrictions on this. Then, the adaptive filtering threshold for each connected region is obtained based on the noise saliency of each region. The adaptive filtering threshold for each connected region is: In the formula, The threshold value is the statistically corrected filter threshold, and e is the natural constant. is the noise significance of the h-th connected region, is the initial threshold of statistical filtering. Based on the corrected filtering thresholds of each connected region, the statistical filtering algorithm is used to denoise the point cloud voxels of each connected region, and the results after denoising of each connected region are obtained. Both the statistical filtering algorithm and the calculation of the initial threshold in the statistical filtering algorithm are well-known existing technologies, and the specific process will not be elaborated. The flowchart for determining the adaptive filtering threshold of the statistical filtering algorithm is as shown in Figure 2 shown.

[0047] Furthermore, when acquiring the three-dimensional point cloud data of the road surface, the optical image data of the road surface is acquired synchronously, and the optical image data is preprocessed, including geometric correction and bilateral filtering denoising, to improve the quality and data consistency of the road surface optical image. Implementers can use other existing filtering algorithms to preprocess the optical image data by themselves, and this embodiment does not limit this.

[0048] All the point cloud voxels after denoising the road surface to be detected are combined with the preprocessed optical image data to construct a three-dimensional digital model of the road surface through point cloud modeling software. In this embodiment, the point cloud modeling software uses Bentley Descartes, and implementers can choose other existing feasible point cloud modeling software by themselves.

[0049] Finally, analyze the three-dimensional digital model of the road surface through the point cloud modeling software, quantify the average texture depth of the road surface, and evaluate whether the texture depth of the road surface to be detected meets the standard requirements according to the highway construction standards. For example, for expressways and first-class highways, the texture depth needs to be between 0.8 mm and 1.2 mm, and for second-class, third-class, and fourth-class highways, the texture depth needs to be between 0.6 mm and 1.1 mm. If the average texture depth meets the construction standards of the road surface, the anti-skid performance of the road surface is qualified; otherwise, the anti-skid performance is unqualified. Calculating the average texture depth of the road surface based on the three-dimensional digital model is a well-known existing technology, and the specific process will not be elaborated.

[0050] Based on the same inventive concept as the above method, an embodiment of the present application also provides a road surface anti-skid performance detection device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above methods for a road surface anti-skid performance detection method.

[0051] It should be noted that: the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. ]>

[0052] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0053] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for testing the anti-skid performance of road surfaces, characterized in that, The method includes the following steps: The three-dimensional point cloud data of the road surface to be detected is acquired, and the point cloud voxels are obtained after downsampling processing. Determine the curvature features, normal vector features, and density features of each point cloud voxel to form the geometric features of each point cloud voxel; based on the similarity between the geometric features of the point cloud voxels, classify all point cloud voxels into categories and extract each connected region of the point cloud voxels in each category. By analyzing the distribution of the multi-scale fractal dimension of each connected region, the self-similarity value of each connected region is determined; the fitted surface of all point cloud voxels in each connected region is obtained, the dispersion of the fitting residuals of point cloud voxels in each connected region is analyzed, and the noise saliency of each connected region is determined by combining the self-similarity value. Statistical filtering is performed on the point cloud voxels of each connected region, and the filtering threshold in the statistical filtering is corrected using the noise saliency. A three-dimensional digital model of the road surface is constructed by combining the denoised point cloud voxels with the optical image data of the road surface, and the anti-skid performance of the road surface is tested.

2. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, The curvature feature is obtained by using the covariance matrix method to obtain the curvature of the centroid point cloud in the point cloud voxel, which is then used as the curvature feature.

3. The method for testing the anti-skid performance of road surfaces as described in claim 2, characterized in that, The normal vector feature is the eigenvector corresponding to the smallest eigenvalue in the covariance matrix.

4. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, The density feature is the percentage of the number of three-dimensional point clouds in each point cloud voxel.

5. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, Determining the self-similarity value of each connected region includes: The degree of dispersion of the multi-scale fractal dimension of each connected region is calculated and denoted as the first degree of dispersion. The mean of the multi-scale fractal dimension of each connected region is calculated. The self-similarity value is negatively correlated with the mean and negatively correlated with the first degree of dispersion.

6. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, The expression for the self-similarity value is: In the formula, Let be the self-similarity value of the h-th connected region, and exp() be an exponential function with base 10 ... Let be the mean of the multiscale fractal dimension of the h-th connected region. denoted as the degree of dispersion of the multiscale fractal dimension of the h-th connected region.

7. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, Determining the noise saliency of each connected region includes: Divide the range of minimum and maximum values ​​of the fitting residuals of point cloud voxels in each connected region into local intervals. Count the number of point cloud voxels whose fitting residuals fall within each local interval. Calculate the information entropy of the number of point cloud voxels in all local intervals corresponding to each connected region. Combine this with the self-similarity value to obtain the noise saliency of each connected region.

8. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, The expression for correcting the filter threshold in statistical filtering using the noise saliency is: In the formula, The threshold value is the statistically corrected filter threshold, and e is the natural constant. Let h be the noise saliency of the h-th connected region. This is the initial threshold for statistical filtering.

9. The method for testing the anti-skid performance of road surfaces as described in claim 1, characterized in that, The testing of the anti-skid performance of the road surface includes: The average construction depth of the road surface is calculated using a three-dimensional digital model. If the average construction depth meets the road surface construction standards, the anti-skid performance of the road surface is qualified; otherwise, the anti-skid performance is unqualified.

10. A road surface skid resistance testing device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-9.

Images