A lightweight identification method for asphalt mixture wear state area

Abstract

The application discloses a lightweight identification method for an asphalt mixture wear state area, and relates to the technical field of road engineering.The application explains the distribution characteristics of the elevation information of the asphalt mixture section from the perspective of the gradient direction vector, adopts a cross replacement of horizontal and vertical lines to replace the point-by-point representation of the direction characteristics, and combines a clustering sampling operation mode, so that the computing power is greatly saved, the one-dimensional clustering calculation of DBscan is used to calculate the norm boundary, classification basis for the vector change type is provided, a Gaussian filter is used to draw a cross point density diagram, and the visualization of the area with large vector change is realized.

Description

Technical Field

[0001] This invention relates to the field of road engineering technology, specifically to a method for identifying lightweight areas of asphalt mixture wear state. Background Technology

[0002] Accurate identification of wear zones and clear analysis of wear stages in asphalt pavements are crucial for early warning and indication of subtle signs of asphalt pavement damage. Extensive engineering practice and literature review show that asphalt pavements exhibit varying degrees of wear loss before significant morphological and functional damage occurs. For example, the wear depth on the wheel track side increases sharply before rutting occurs, and the wear texture brought out by the wheel track may converge along the driving direction before anti-skid cracking. Therefore, failure to scientifically and accurately identify wear zones or reliably and reasonably define wear stages will not only exacerbate the damage to asphalt pavements and severely shorten their service life, but may even create significant hidden dangers affecting vehicle handling control and driving safety. Secondly, the analysis of asphalt pavement wear stages is crucial for predicting pavement life and formulating reasonable maintenance cycles. Without clearly identifying the specific stages of road wear, it is impossible to effectively predict the remaining service life of the pavement, which will affect the formulation of overall maintenance strategies and budget allocation. Without stage-specific wear information, road maintenance work may be overly concentrated or delayed until the most urgent maintenance periods, leading to wasted resources or a decline in road service performance.

[0003] Currently, cross-sectional texture depth is mainly used to characterize the wear state of thin layers at different time stages, while the rate of change of texture depth index is used as the main evaluation method for changes in different wear stages. However, due to the small amplitude of texture differences at different time stages in depth dimension measurement, which is difficult to capture, only subtle depth changes can be produced under a large number of repeated loads. Although some scholars have used sophisticated characterization methods such as three-dimensional laser and stereo vision to characterize or describe the apparent texture depth of thin-layer coating regions, the measurement of the degree of change in wear state based on depth measurement still suffers from problems such as small index amplitude and poor measurement accuracy. Summary of the Invention

[0004] This invention addresses the problem that measuring the degree of wear state change based on depth measurement still suffers from small index variation and poor measurement accuracy. By utilizing indoor three-dimensional laser scanning equipment and fully considering the change of cross-sectional elevation gradient vector of indoor asphalt mixture before and after wear, this invention proposes a lightweight method for identifying different wear state regions of indoor asphalt mixture based on elevation gradient distribution, with a row-column interactive structure and probability density distribution. This method initially realizes the scientific identification and segmentation of wear state regions of indoor asphalt mixture.

[0005] The present invention provides a method for lightweight identification of wear state zones in asphalt mixtures, comprising the following steps:

[0006] Collect asphalt mixture cross-sectional elevation data, reduce noise in the asphalt mixture cross-sectional elevation data, and extract three-dimensional laser cross-sectional data of the wheel-rolled area before and after wear based on the correspondence between the horizontal and vertical intervals of the collected laser cross-sectional data and the actual dimensions, while retaining the elevation data of the areas unaffected by wheel rolling before and after wear.

[0007] The elevation of the unworn area in the cross-sectional data before and after wear was used as the registration object for point cloud registration.

[0008] Based on the three-dimensional laser cross-sectional data of the wear area before and after wear, the difference between the row and column gradient direction vectors is calculated, and the L2 norm of the difference between the row and column gradient direction vectors is calculated respectively.

[0009] Clustering of the gradient direction vectors of the row and column vectors using a one-dimensional array with the L2 norm as the index;

[0010] Based on the calculated cluster boundaries, the WHERE function is used to mark the scattered points in different cluster intervals and return the planar coordinates of the original dataset row and column vectors corresponding to the points in different cluster intervals;

[0011] The row coordinates returned by the row gradient direction vector difference norm and the column coordinates returned by the column gradient direction vector difference norm are plotted on the two-dimensional image drawn from the three-dimensional laser cross-section data before or after wear, and the row line corresponding to each row coordinate and the column line corresponding to each column coordinate are plotted respectively.

[0012] Based on the row lines corresponding to each row coordinate and the column lines corresponding to each column coordinate, create the intersection coordinates of the strips that intersect the same clustering features that aggregate each norm index;

[0013] A cluster sampler is generated, and a large number of intersection points returned from the cluster intervals are collected. Based on the distribution density and distribution clustering characteristics, samples are taken, and the density function of the intersection points returned from the cluster intervals after processing by the cluster sampler is calculated.

[0014] Based on the cross-point density function returned by the clustering intervals after processing by the clustering sampler, a density map of the clustering intervals is drawn. The high-density region of each cluster feature is segmented by the watershed algorithm to obtain a visualization result, showing the wear status of regions with different wear states.

[0015] Furthermore, the process of collecting asphalt mixture cross-sectional elevation data, denoising the asphalt mixture cross-sectional elevation data, and extracting three-dimensional laser cross-sectional data of the wheel-rolled area before and after wear based on the correspondence between the horizontal and vertical intervals of the collected laser cross-sectional data and the actual dimensions, while retaining the elevation data of areas unaffected by wheel rolling before and after wear, specifically includes:

[0016] An indoor 3D laser scanning system was used to collect elevation data of asphalt mixture cross sections. Based on a Python compilation environment, a Gaussian low-pass filter function was used to reduce noise in the data. According to the correspondence between the horizontal and vertical intervals of the collected laser cross section data and the actual dimensions, the 3D laser cross section data before and after wear in the wheel-type accelerated loading equipment rolling area were extracted. At the same time, the elevation data of the area unaffected by the wheel rolling before and after wear, i.e. the edge area outside the wheel rolling, was retained to provide a basis for registration.

[0017] Among them, a unidirectional wheel-type accelerated loading device is used to wear down the thin layer, and the three-dimensional laser scanning system collects data at intervals of 0.5mm*0.5mm. The three-dimensional laser scanning system collects elevation data of the state before and after several wear cycles, and extracts the elevation data of the state before and after wear based on the rolling width of the rubber wheel in the wheel-type wear device.

[0018] Furthermore, the method of clustering one-dimensional arrays of gradient direction vectors of row and column vectors using the DBscan clustering method with the L2 norm as the index, based on the Python compilation environment, also includes:

[0019] A larger norm indicates a greater change in vector, and a greater change in the direction or modulus of the gradient direction vector. This indicates that the elevation gradient of a certain region is easily changed, and thus the wear resistance of the thin-layer coating in that region is worse.

[0020] Furthermore, the method of clustering one-dimensional arrays of gradient direction vectors of row and column vectors using the DBscan clustering method with the L2 norm as the index, based on the Python compilation environment, also includes:

[0021] The DBSCAN clustering method is used to cluster the gradient direction vector into a one-dimensional array according to two high-density expectations. The expectation is to obtain the high-density clustering intervals a and b corresponding to the large norm and small norm values. The remaining scattered points that do not belong to any density clustering interval are regarded as noise and are used as another interval called buffer c. The data is divided into three intervals in total.

[0022] Furthermore, the method of clustering one-dimensional arrays of gradient direction vectors of row and column vectors using the DBscan clustering method with the L2 norm as the index, based on the Python compilation environment, also includes:

[0023] The clustering density of the DBSCAN clustering method is set to 0.5, and the minimum capacity is set to 1 / 4 of the total computing units.

[0024] Furthermore, the step of using the WHERE function to label the scatter points in different clustering intervals based on the calculated clustering boundaries, and returning the planar coordinates of the original dataset row and column vectors corresponding to the points in different clustering intervals, also includes:

[0025] Since the number of scatter points is equal to the x-axis and the norm is equal to the y-axis in a coordinate system, each scatter point represents the norm value of the difference between the gradient directions of a row vector or column vector. Therefore, the returned coordinates are either row coordinates or column coordinates.

[0026] Based on the Python compilation environment, retrieve the corresponding positions of these scattered point coordinates in the column or row vectors of the dataset and return the corresponding row and column coordinates of these vectors. The row coordinates refer to the coordinates of class (x, 0), and the column coordinates refer to the coordinates of class (0, y).

[0027] Furthermore, the step of creating the intersection coordinates of the strips that intersect the same clustering features aggregating each norm index also includes:

[0028] The function returns the coordinates of the intersection point of the row line corresponding to the row gradient direction vector difference and the column line corresponding to the column gradient direction vector difference for cluster interval a. The same method is used to return the intersection point for cluster intervals b and c, as well as the noise interval.

[0029] Furthermore, the step of creating the intersection coordinates of the strips that intersect the same clustering features aggregating each norm index also includes:

[0030] By using a weaving algorithm to find the intersection points of horizontal and vertical lines, the first step of weight reduction is achieved, reducing the computational load from m*n to m+n operations.

[0031] Furthermore, the generation of the clustering sampler, based on the distribution density and clustering degree characteristics of the massive number of intersection points returned from the clustering intervals, takes samples and calculates the density function of the intersection points returned from the clustering intervals after processing by the clustering sampler, and also includes:

[0032] By employing the K-means clustering algorithm, the density characteristics of the intersection distribution corresponding to cluster intervals a, b, and c are preserved, while the computational load is reduced for the second time as a sampler.

[0033] Compared with the prior art, the present invention provides a method for lightweight identification of wear state zones in asphalt mixtures, which has the following advantages:

[0034] This invention explains the distribution characteristics of asphalt mixture cross-sectional elevation information from the perspective of gradient direction vectors. It adopts a computational mode combining horizontal and vertical line intersections with cluster sampling, replacing point-by-point representation of directional features, significantly saving computational resources. One-dimensional clustering using DBscan calculates norm boundaries, providing a classification basis for vector change types. Simultaneously, a Gaussian filter is used to draw an intersection density map, enabling visualization of areas with significant vector changes. The computational structure used in this invention, which calculates the intersection points of horizontal and vertical lines based on the difference between row and column gradient direction vectors, initially reveals and extracts a large number of samples with different clustering characteristics of vector difference norms. Subsequently, a density sampler is used to optimize and filter the samples corresponding to intersection points within each vector difference norm feature interval. The "scaling" concept clarifies the characteristics of gradient direction vector changes before and after wear, avoiding the ambiguity and randomness of using vertical indices such as cross-sectional depth with small amplitudes to calculate different wear state regions. It also avoids the problems of small amplitude and poor measurement accuracy in existing technologies that use depth measurements to measure the degree of wear state change. Attached Figure Description

[0035] Figure 1 A flowchart of the lightweight identification method for wear state zones of asphalt mixtures provided by the present invention;

[0036] Figure 2 This is a schematic diagram of the scatter distribution of the L2 norm of the column gradient direction vector difference in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of the scatter distribution of the row gradient direction vector difference norm in an embodiment of the present invention;

[0038] Figure 4 This is a schematic diagram showing the intersection of the horizontal and vertical lines corresponding to the row and column gradient direction vectors in Embodiment 1 of the present invention.

[0039] Figure 5 This is a schematic diagram illustrating the visualization of different wear regions based on the watershed algorithm in Embodiment 1 of the present invention;

[0040] Figure 6 This is a schematic diagram illustrating the visualization of different wear regions based on the watershed algorithm in Embodiment 2 of the present invention. Detailed Implementation

[0041] The following is in conjunction with the appendix Figure 1 -Appendix Figure 6 The following describes specific embodiments of the present invention in further detail. These embodiments are merely for illustrating the technical solutions of the present invention more clearly and should not be construed as limiting the scope of protection of the present invention.

[0042] Example 1: This example addresses the problem that measuring the degree of wear state change based on depth measurement still suffers from small index variation and poor measurement accuracy. It provides a method for lightweight identification of wear state areas in asphalt mixtures, including the following steps:

[0043] Step 1: Prepare indoor asphalt mixture rutting slabs in accordance with the standard "Test Procedures for Asphalt and Asphalt Mixtures in Highway Engineering".

[0044] Step 2: An indoor 3D laser scanning system is used to collect elevation data of the rut slab cross-section. Based on the Python compilation environment, a Gaussian low-pass filter function is used to reduce noise in the data. Then, based on the correspondence between the horizontal and vertical intervals of the collected laser cross-section data and the actual dimensions, the 3D laser cross-section data of the wheel-driven accelerated loading equipment's rolling area before and after wear is extracted. At the same time, the elevation data of the areas unaffected by the wheel rolling before and after wear (i.e., the edge areas outside the wheel rolling area) are retained to provide a basis for registration.

[0045] Step 3: Using the elevation of the unworn area in the cross-sectional data before and after wear as the registration object for point cloud registration, registration is performed using the Python-based ICP compilation environment.

[0046] Step four: To reduce computational load, the gradient direction vectors of the row and column vectors of the dataset are calculated separately. Then, based on the three-dimensional laser cross-sectional data of the wear area before and after wear, the difference between the row and column gradient direction vectors is calculated separately, and the L2 norm of the difference between the row and column gradient direction vectors is calculated separately.

[0047] Step 5: Based on the Python compilation environment, the DBscan clustering method is used to cluster the gradient direction vectors of the row and column vectors into a one-dimensional array with the L2 norm as the index. The clustering results are as follows: Figure 2 and Figure 3 Indication.

[0048] Step 6: Based on the calculated cluster boundaries, use the WHERE function to mark the scattered points in different cluster intervals, and return the planar coordinates of the original dataset row and column vectors corresponding to the points in different cluster intervals.

[0049] Step 7: Using the Python compilation environment, plot the row coordinates returned by the row gradient direction vector difference norm and the column coordinates returned by the column gradient direction vector difference norm on the 2D image generated from the 3D laser cross-section data before or after wear. Then, plot the row lines corresponding to each row coordinate, and then plot the column lines corresponding to each column coordinate. Using the compilation environment, similar to "weaving," plot the horizontal and vertical lines on the 2D image and set a certain transparency to visualize regions with different clustering features / different gradient direction change features. Figure 4This is a visualization of the intersection of horizontal and vertical lines corresponding to the row and column gradient direction vectors. The stripes represent clusters a, b, and the noise interval, respectively.

[0050] Step 8: First, use the set function to create the coordinates of the intersection points of the strips that aggregate the same clustering features of each norm index.

[0051] Step 9: Using the K-means clustering algorithm in the Python compilation environment, a cluster sampler is generated. The massive number of intersection points returned from the clustering intervals a, b, and c are used. Based on the distribution density and distribution clustering characteristics of the K-means algorithm, a total of 100 samples are taken. Then, the intersection point density function of the 100 clustering intervals a, b, and c after processing by the cluster sampler is calculated according to the Gaussian filter.

[0052] Step 10: Based on the clustering intervals a, b, and c calculated by the Gaussian filter, return the density function corresponding to the intersection points after sampler processing, plot the density maps of the three clustering intervals, and use the watershed algorithm to segment the high-density regions of each cluster feature, thereby visualizing regions with different wear states, such as... Figure 5 As shown.

[0053] In this embodiment, in step two, the three-dimensional laser detection system can acquire data at intervals of up to 0.5mm*0.5mm; a unidirectional wheel-type accelerated loading device is used to wear down the thin layer; the three-dimensional laser detection device is used to acquire elevation data of the state before and after several wear cycles, and the elevation data of the state before and after wear is extracted based on the rolling width of the rubber wheel in the wheel-type wear device.

[0054] In this embodiment, in step five, regardless of the difference between the row and column gradient direction vectors, the larger the norm of the interval, the greater the vector change, the greater the change in direction and modulus, and the greater the change in the direction and modulus of the gradient direction vector, which also indicates that the wear resistance of a certain area of ​​the asphalt mixture is worse.

[0055] In this embodiment, in step five, the DBSCAN clustering method is used to cluster the gradient direction vector into a one-dimensional array according to two high-density expectations. The expectation is to obtain the high-density clustering intervals a and b corresponding to the large norm value and the small norm value. The remaining scattered points that do not belong to any density clustering interval are regarded as noise and are used as another interval called buffer interval c. As a result, the data is divided into three intervals.

[0056] In this embodiment, in step five, DBSCAN needs to set parameters regarding cluster density and minimum cluster size. According to the literature review and user manual, the cluster density is set to 0.5 and the minimum size is set to 1 / 4 of the total computing units.

[0057] In this embodiment, in step six, since the number of scattered points is x-axis and the norm is y-axis, each scattered point represents the norm value of the gradient direction vector difference of a row vector or column vector. Therefore, the returned coordinates are row coordinates or column coordinates. Based on the Python compilation environment, the positions of these scattered point coordinates in the corresponding column vectors or row vectors of the dataset are retrieved and the corresponding row and column coordinates of these vectors are returned. The row coordinates refer to the coordinates of the (x, 0) class, and the column coordinates refer to the coordinates of the (0, y) class.

[0058] In this embodiment, in step eight, the coordinates of the intersection of the row coordinates returned by the row gradient direction vector difference of cluster interval a and the column coordinates returned by the column gradient direction vector difference are returned; the intersection points of cluster intervals b and c and the noise interval are returned using the same method.

[0059] In this embodiment, in step eight, the first lightweighting is achieved by obtaining the intersection point through the weaving algorithm of horizontal and vertical lines, which reduces the amount of computation. Specifically, the amount of computation of m*n is reduced to m+n operations.

[0060] In this embodiment, in step nine, the K-means clustering algorithm is used to retain the density characteristics of the intersection distribution corresponding to the cluster intervals a, b, and c, while simultaneously reducing the computational load for the second time as a sampler.

[0061] Example 2: This example further explains and demonstrates the lightweight identification method for asphalt mixture wear state zones provided in Example 1. The lightweight identification method for asphalt mixture wear state zones provided in this example includes the following steps:

[0062] Step 1: Prepare one AC-13 asphalt mixture rutting slab in the laboratory, referring to the standard "Test Procedures for Asphalt and Asphalt Mixtures in Highway Engineering".

[0063] Step two involves acquiring elevation data of the rut slab cross-section using an indoor 3D laser scanning system. The scanned rut slab is 300mm x 300mm in size. Based on a Python compilation environment, a Gaussian low-pass filter is used to reduce noise in the data. 3D laser cross-sectional data of the 200mm x 200mm area rolled by the isostatic wheel-driven accelerator are extracted before and after wear. Elevation data of the unaffected area (i.e., the 50mm x 50mm x 2mm edge area outside the wheel-driven accelerator) measured in the wear direction are also retained to provide a basis for registration. The laser cross-sectional data corresponding to the wear area is a 400-row x 400-column elevation dataset, while the cross-sectional data used for registration is 100-row x 100-column x 2.

[0064] The three-dimensional laser detection system can acquire data at intervals of up to 0.5mm*0.5mm; a unidirectional wheel-type accelerated loading device is used to wear down the thin layer; the three-dimensional laser detection device is used to acquire elevation data of the state before and after several wear cycles, and the elevation data of the state before and after wear is extracted based on the rolling width of the rubber wheel in the wheel-type wear device.

[0065] Step 3: Using the elevation of the unworn area in the cross-sectional data before and after wear as the registration object for point cloud registration, registration is performed using the Python-based ICP compilation environment.

[0066] Step 4: Calculate a total of 800 gradient direction vectors for the row and column vectors of the dataset, and then calculate a total of 800 differences between the row and column gradient direction vectors of the dataset before and after wear, and calculate the L2 norm of the differences between the row and column gradient direction vectors respectively.

[0067] Step 5: Based on the Python compilation environment, the DBSCAN clustering method is used to cluster the gradient direction vector differences of row and column vectors into a one-dimensional array with the L2 norm as the index. In the clustering of column and row gradient direction vector differences, regardless of whether it's row or column gradient direction vector differences, the larger the norm of the interval, the greater the vector change, indicating a greater change in direction and modulus. This means a greater change in the direction and modulus of the gradient direction vector, and consequently, a worse wear resistance in a certain area of ​​the asphalt mixture. The DBSCAN clustering method clusters the one-dimensional array of gradient direction vectors according to two high-density expectations, aiming to obtain high-density clustering intervals a and b corresponding to the large and small norm values. Scattered points not belonging to any density clustering interval are considered noise and treated as another interval called buffer c. The result is that the data is divided into three intervals.

[0068] Step six: Based on the calculated cluster boundaries, the WHERE function is used to label the scatter points in different cluster intervals, and the planar coordinates of the original dataset row and column vectors corresponding to the points in different cluster intervals are returned. The number of coordinate points returned for the three intervals with decreasing moduli are 214, 39, 147 and 208, 41, 151 respectively; at the same time, we can obtain the number of intersection points of the three intervals as follows: 44512 for 214*208, 1599 for 39*41, and 22197 for 151*147.

[0069] Step 7: Using the Python compilation environment, plot the row coordinates returned by the row gradient direction vector difference norm and the column coordinates returned by the column gradient direction vector difference norm on the 2D image generated from the 3D laser cross-section data before or after wear. Draw row lines corresponding to each row coordinate, and then draw column lines corresponding to each column coordinate. Using the compilation environment, similar to "weaving," draw the horizontal and vertical lines on the 2D image and set a certain transparency to initially visualize regions with different clustering features / different gradient direction change features. The stripes corresponding to clusters a, b, and buffer c are shown.

[0070] Step eight: First, the `set` function is used to create the intersection coordinates of the strips that aggregate the same clustering features for each norm index. The coordinates of the intersection points of the row lines corresponding to the row gradient direction vector difference and the column lines corresponding to the column gradient direction vector difference for cluster interval a are returned. The same method is used to return the intersection points for cluster intervals b and c, as well as the noise interval. Therefore, the number of intersection coordinates for the three intervals mentioned in step six are: 44512 for 214*208, 1599 for 39*41, and 22197 for 151*147.

[0071] Step 9: Using the Kmeans clustering algorithm in the Python compilation environment, a cluster sampler is generated, focusing on the distribution density and clustering characteristics of scattered points. A total of 100 samples are collected from the massive number of intersection points returned from clustering intervals a, b, and c. Then, the intersection point density function of the 100 clustering intervals a, b, and c after processing by the cluster sampler is calculated based on the Gaussian filter.

[0072] Step 10: Based on the clustering intervals a, b, and c calculated using the Gaussian filter, return the density function corresponding to the intersection points after sampler processing, plot the density maps of the three clustering intervals, and use the watershed algorithm to segment the high-density regions of each cluster feature, thereby visualizing regions with different wear states, such as... Figure 6 As shown, cluster1 represents the region of type a wear state, that is, the region with a large gradient direction vector difference or easy wear; cluster2 represents the region of type b wear state, that is, the region with a small gradient direction vector difference or not easy wear; and cluster3 represents the region of type c wear state, that is, the region with random gradient direction vector difference or between the wear resistance characteristics.

[0073] In summary, compared with existing technologies, the lightweight identification method for wear state zones of asphalt mixtures provided by this invention has the following beneficial effects:

[0074] This invention explains the distribution characteristics of asphalt mixture cross-sectional elevation information from the perspective of gradient direction vectors. It adopts a computational mode combining horizontal and vertical line intersections with cluster sampling, replacing point-by-point representation of directional features, which significantly saves computational resources. One-dimensional clustering using DBscan calculates norm boundaries, providing a classification basis for vector change types. Simultaneously, a Gaussian filter is used to draw an intersection density map, enabling visualization of areas with large vector changes. The computational structure used in this invention, which calculates the intersection points of horizontal and vertical lines based on the difference between row and column gradient direction vectors, initially reveals and extracts a large number of samples with different clustering characteristics of vector difference norms. Subsequently, a density sampler is used to optimize and filter the samples corresponding to intersection points within each vector difference norm feature interval. The "scaling" concept clarifies the change characteristics of gradient direction vectors before and after wear, avoiding the ambiguity and randomness of using vertical indicators such as cross-sectional depth with small amplitude variations to calculate different wear state regions. This provides a new approach for characterizing the wear state of asphalt pavements and classifying different wear state regions.

[0075] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0076] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for identifying lightweight wear zones in asphalt mixtures, characterized in that, Includes the following steps: Collect asphalt mixture cross-sectional elevation data, reduce noise in the asphalt mixture cross-sectional elevation data, and extract three-dimensional laser cross-sectional data of the wheel-rolled area before and after wear based on the correspondence between the horizontal and vertical intervals of the collected laser cross-sectional data and the actual dimensions, while retaining the elevation data of the areas unaffected by wheel rolling before and after wear. The elevation of the unworn area in the cross-sectional data before and after wear was used as the registration object for point cloud registration. Based on the three-dimensional laser cross-sectional data of the wear area before and after wear, the row and column gradient direction vectors are obtained respectively, and the L2 norm of the difference between the row and column gradient direction vectors is calculated respectively. Based on the Python compilation environment, the DBscan clustering method is used to cluster the gradient direction vectors of row and column vectors into a one-dimensional array with the L2 norm as the index. Based on the calculated cluster boundaries, the WHERE function is used to mark the scattered points in different cluster intervals and return the planar coordinates of the original dataset row and column vectors corresponding to the points in different cluster intervals; The row coordinates returned by the row gradient direction vector difference norm and the column coordinates returned by the column gradient direction vector difference norm are plotted on the two-dimensional image drawn from the three-dimensional laser cross-section data before or after wear, and the row line corresponding to each row coordinate and the column line corresponding to each column coordinate are plotted respectively. Based on the row lines corresponding to each row coordinate and the column lines corresponding to each column coordinate, create the intersection coordinates of the strips that intersect the same clustering features that aggregate each norm index; Generate a cluster sampler, take samples from the massive number of intersections returned by the cluster intervals, based on the distribution density and distribution clustering characteristics, and calculate the density function of the intersections returned by the cluster intervals after processing by the cluster sampler; Based on the cross-point density function returned by the clustering intervals after processing by the clustering sampler, a density map of the clustering intervals is drawn. The high-density region of each cluster feature is segmented by the watershed algorithm to obtain a visualization result, showing the wear status of regions with different wear states.

2. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 1, characterized in that, The process involves collecting elevation data of asphalt mixture cross-sections, denoising the asphalt mixture cross-section elevation data, and extracting three-dimensional laser cross-section data of the wheel-rolled area before and after wear based on the correspondence between the horizontal and vertical intervals of the collected laser cross-section data and the actual dimensions. Elevation data of areas unaffected by wheel rolling before and after wear are retained. Specifically, this includes: An indoor 3D laser scanning system was used to collect elevation data of asphalt mixture cross sections. Based on a Python compilation environment, a Gaussian low-pass filter function was used to reduce noise in the data. According to the correspondence between the horizontal and vertical intervals of the collected laser cross section data and the actual dimensions, the 3D laser cross section data before and after wear in the wheel-type accelerated loading equipment rolling area were extracted. At the same time, the elevation data of the area unaffected by the wheel rolling before and after wear, i.e. the edge area outside the wheel rolling, was retained to provide a basis for registration. Among them, a unidirectional wheel-type accelerated loading device is used to wear down the thin layer, and the three-dimensional laser scanning system collects data at intervals of 0.5mm*0.5mm. The three-dimensional laser scanning system collects elevation data of the state before and after several wear cycles, and extracts the elevation data of the state before and after wear based on the rolling width of the rubber wheel in the wheel-type wear device.

3. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 1, characterized in that, The method, based on a Python compilation environment and employing the DBscan clustering method to cluster a one-dimensional array of row and column vector gradient direction vectors using the L2 norm as an index, also includes: A larger norm indicates a greater change in vector, and a greater change in the direction or modulus of the gradient direction vector. This indicates that the elevation gradient of a certain region is easily changed, and thus the wear resistance of the thin-layer coating in that region is worse.

4. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 3, characterized in that, The method, based on a Python compilation environment and employing the DBscan clustering method to cluster a one-dimensional array of row and column vector gradient direction vectors using the L2 norm as an index, also includes: The DBSCAN clustering method is used to cluster the gradient direction vector into a one-dimensional array according to two high-density expectations. The expectation is to obtain the high-density clustering intervals a and b corresponding to the large norm and small norm values. The remaining scattered points that do not belong to any density clustering interval are regarded as noise and are used as another interval called buffer c. The data is divided into three intervals in total.

5. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 4, characterized in that, The method, based on a Python compilation environment and employing the DBscan clustering method to cluster a one-dimensional array of row and column vector gradient direction vectors using the L2 norm as an index, also includes: The clustering density of the DBSCAN clustering method is set to 0.5, and the minimum capacity is set to 1 / 4 of the total computing units.

6. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 1, characterized in that, The step of using the WHERE function to label scatter points in different clustering intervals based on the calculated cluster boundaries, and returning the planar coordinates of the original dataset row and column vectors corresponding to the points in different clustering intervals, also includes: Since the number of scatter points is equal to the x-axis and the norm is equal to the y-axis in a coordinate system, each scatter point represents the norm value of the difference between the gradient directions of a row vector or column vector. Therefore, the returned coordinates are either row coordinates or column coordinates. Based on the Python compilation environment, retrieve the corresponding positions of these scattered point coordinates in the column or row vectors of the dataset and return the corresponding row and column coordinates of these vectors. The row coordinates refer to the coordinates of the (x, 0) class, and the column coordinates refer to the coordinates of the (0, y) class.

7. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 4, characterized in that, The step of creating the intersection coordinates of the strips that intersect the same clustering features aggregating each norm index also includes: The function returns the coordinates of the intersection point of the row line corresponding to the row gradient direction vector difference and the column line corresponding to the column gradient direction vector difference for cluster interval a. The same method is used to return the intersection point for cluster intervals b and c, as well as the noise interval.

8. The method for identifying lightweight wear zones in asphalt mixtures as described in claim 7, characterized in that, The step of creating the intersection coordinates of the strips that intersect the same clustering features aggregating each norm index also includes: By using a weaving algorithm to find the intersection points of horizontal and vertical lines, the first step of weight reduction is achieved, reducing the computational load from m*n to m+n operations.

9. The method for identifying lightweight wear state zones of asphalt mixtures as described in claim 4, characterized in that, The generation of the clustering sampler, which takes samples from the massive number of intersection points returned by the clustering intervals based on the distribution density and distribution clustering degree characteristics, and calculates the density function of the intersection points returned by the clustering intervals after processing by the clustering sampler, also includes: By employing the K-means clustering algorithm, the density characteristics of the intersection distribution corresponding to cluster intervals a, b, and c are preserved, while the computational load is reduced for the second time as a sampler.

Images