A method for predicting road service life based on road construction indentation distribution

Abstract

The application discloses a kind of based on road construction indentation distribution prediction road surface service life method, belong to road data processing technical field, comprising: obtaining the three-dimensional point cloud data of road surface, the three-dimensional point cloud data is separated out indentation area and generates first data set;Image data is obtained from first data set, and second feature set containing indentation boundary data is generated;According to the height data of first data set, the point cloud height difference is calculated, and the third feature set containing depth feature data is obtained;The second feature set of current containing indentation boundary information and the third feature set containing depth feature are input to fourth prediction model, and material hardness prediction value is output;According to material hardness prediction value, in combination with road load data, the road surface life prediction value is calculated.The based on road construction indentation distribution prediction road surface service life method solves the problem that traditional method is difficult to accurately evaluate road indentation condition and predict life.

Description

Technical Field

[0001] This invention relates to the field of road data processing technology, and in particular to a method for predicting pavement service life based on the distribution of road construction indentations. Background Technology

[0002] In the field of road engineering, the durability and safety of infrastructure are paramount, and accurate prediction of pavement life directly impacts road maintenance costs and traffic efficiency. Indentation, a common pavement damage feature during road construction, significantly influences pavement life due to its distribution patterns, forming the basis for high-quality road management. However, existing methods for predicting pavement life primarily rely on simple indentation statistics or macroscopic damage assessments, neglecting the diverse characteristics of indentation morphology, depth, and road loads. This results in insufficient accuracy in prediction models, making them ill-suited to complex construction environments and varying traffic loads. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention provides a method for predicting pavement service life based on road construction indentation distribution, thereby solving the aforementioned problems.

[0004] The technical solution adopted by this invention to solve its technical problem is: a method for predicting pavement service life based on road construction indentation distribution, comprising the following steps:

[0005] S1: Acquire three-dimensional point cloud data of the road surface, separate the indentation area from the three-dimensional point cloud data and generate a first dataset;

[0006] S2: Obtain image data from the first dataset and use an edge detection mechanism to generate a second feature set containing indentation boundary data;

[0007] S3: Calculate the point cloud height difference based on the height data of the first dataset to obtain a third feature set containing depth feature data;

[0008] S4: Input the second feature set containing indentation boundary data, the third feature set containing depth feature data, and the measured material hardness value corresponding to the same historical time into the convolutional neural network for training, establish the mapping relationship between indentation features and material hardness, and generate the fourth prediction model, where the indentation features include indentation boundary data and depth feature data; input the current second feature set containing indentation boundary information and the third feature set containing depth features into the fourth prediction model, and output the predicted material hardness value;

[0009] S5: Calculate the pavement life prediction value based on the material hardness prediction value output by the fourth prediction model and the road load data.

[0010] Preferably, step S1 includes the following sub-steps:

[0011] S11: Use laser scanning technology to acquire three-dimensional point cloud data containing spatial information of the road surface to obtain the first point cloud;

[0012] S12: By setting the growth threshold in the region growth segmentation algorithm, multiple indentation regions are separated in the first point cloud. Each indentation region corresponds to a subset of data containing indentation features, thus obtaining the first dataset. The growth threshold includes the angle between the normal vectors of the point clouds and the distance between two adjacent point clouds.

[0013] Optionally, step S13 is included after step S12, and step S13 includes:

[0014] The number of points in each subset is compared with a preset threshold. If the number of points in a subset is lower than the preset threshold, the missing points in the subset are supplemented by a bilinear interpolation algorithm to generate an optimized subset.

[0015] It is worth noting that step S2 includes the following sub-steps:

[0016] S21: Obtain image data for each subset from the first dataset;

[0017] S22: Use the cv2.cvtColor function in the OpenCV library to convert the image data into a grayscale image to obtain the first grayscale image set;

[0018] S23: For each grayscale image in the first grayscale image set, the cv2.Canny function in the OpenCV library is used for edge detection. The high threshold and low threshold are preset as the first threshold and the second threshold, respectively, to extract the indentation boundary data containing the indentation boundary location;

[0019] S24: Integrate the indentation boundary data into its corresponding subset, and combine all subsets to obtain the second feature set.

[0020] Specifically, step S3 includes the following sub-steps:

[0021] S31: Obtain the point cloud height data of each subset in the first dataset, calculate the height difference between adjacent point clouds for each subset, and obtain the subset height difference set;

[0022] S32: Extract the maximum, minimum and median values ​​from the height difference set of the subset datasets and integrate them into the corresponding subset datasets to obtain subset datasets containing depth feature data;

[0023] S33: Combine the subset containing deep feature data into a third feature set.

[0024] Specifically, in step S32, the K-means algorithm is used to classify the subset containing deep feature data, determine the category of deep feature data of the subset, and obtain a set of classification features.

[0025] It is worth noting that step S4 includes the following sub-steps:

[0026] S41: Combine the indentation boundary data in the second feature set and the depth feature data in the third feature set into a feature input matrix;

[0027] S42: The feature input matrix is ​​propagated forward through a convolutional neural network. The ReLU activation function and Adam optimizer are used to calculate the feature mappings of the convolutional layer and the pooling layer to obtain an initial feature mapping set.

[0028] S43: Using the mean squared error loss function, calculate the error between the initial feature mapping set and the actual value of material hardness, adjust the network parameters through backpropagation to obtain the optimized network parameters, and obtain the mapping relationship based on the optimized network parameters and convergence conditions;

[0029] S44: Based on the mapping relationship, construct a fourth prediction model and calculate the predicted value of material hardness by inputting new indentation boundary data and depth feature data.

[0030] Optionally, in step S43, if the optimized network parameters meet the preset convergence condition, the current feature map in the initial feature map set is associated with the material property label to generate a mapping relationship.

[0031] If the optimized network parameters do not meet the preset convergence conditions, the forward and backward propagation are repeated to adjust the network parameters until the convergence conditions are met and the mapping relationship is obtained.

[0032] Preferably, step S5 includes the following sub-steps:

[0033] S51: Obtain the predicted material hardness value output by the fourth prediction model and the traffic flow intensity in the load data to obtain the pavement life prediction dataset; perform standardization processing on the pavement life prediction dataset to obtain a standardized dataset;

[0034] S52: Use the random forest algorithm to extract features from the standardized dataset to obtain a set of feature vectors;

[0035] S53: Based on the set of feature vectors, calculate the importance of each feature, and determine the weight of the weighted fusion mechanism based on the importance of each feature; process the set of feature vectors through the weighted fusion mechanism to obtain a set of road surface life prediction values;

[0036] S54: Select the median from the set of predicted pavement life values ​​to determine the final predicted pavement life value.

[0037] The beneficial effects of this invention are as follows: In the method for predicting pavement service life based on road construction indentation distribution, three-dimensional point cloud data of the pavement is acquired, and the indentation area is separated using a stereo segmentation algorithm. Based on the indentation area, the corresponding indentation boundary data and depth feature data are obtained. The feature set is input into a convolutional neural network to establish a mapping relationship between indentation features and material hardness values. Finally, the predicted material hardness value output by the fourth prediction model is combined with load data for analysis, achieving accurate prediction of pavement service life. This invention effectively solves the problem that traditional methods are difficult to accurately assess pavement indentation conditions and predict service life through multi-dimensional feature extraction and deep learning models, providing a reliable basis for pavement maintenance decisions. Attached Figure Description

[0038] Figure 1 This is a flowchart of a method for predicting pavement service life based on road construction indentation distribution in one embodiment of the present invention;

[0039] Figure 2 This is a flowchart of step S1 in one embodiment of the present invention;

[0040] Figure 3 This is a flowchart illustrating the steps of step S2 in one embodiment of the present invention;

[0041] Figure 4 This is a flowchart illustrating the steps of step S3 in one embodiment of the present invention;

[0042] Figure 5 This is a flowchart illustrating the steps of step S4 in one embodiment of the present invention;

[0043] Figure 6 This is a flowchart of step S5 in one embodiment of the present invention. Detailed Implementation

[0044] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0045] like Figure 1-6 As shown, a method for predicting pavement service life based on road construction indentation distribution includes the following steps:

[0046] S1: Acquire three-dimensional point cloud data of the road surface, separate the indentation area from the three-dimensional point cloud data and generate a first dataset;

[0047] S2: Obtain image data from the first dataset and use an edge detection mechanism to generate a second feature set containing indentation boundary data;

[0048] S3: Calculate the point cloud height difference based on the height data of the first dataset to obtain a third feature set containing depth feature data;

[0049] S4: Input the second feature set containing indentation boundary data, the third feature set containing depth feature data, and the measured material hardness value corresponding to the same historical time into the convolutional neural network for training, establish the mapping relationship between indentation features and material hardness, and generate the fourth prediction model, where the indentation features include indentation boundary data and depth feature data; input the current second feature set containing indentation boundary information and the third feature set containing depth features into the fourth prediction model, and output the predicted material hardness value;

[0050] S5: Calculate the pavement life prediction value based on the material hardness prediction value output by the fourth prediction model and the road load data.

[0051] In the method for predicting pavement service life based on road construction indentation distribution, three-dimensional point cloud data of the pavement is acquired, and an indentation region is separated using a stereo segmentation algorithm. Based on the indentation region, its corresponding indentation boundary data and depth feature data are obtained. The feature set is input into a convolutional neural network to establish a mapping relationship between indentation features and material hardness values. Finally, the predicted material hardness value output by the fourth prediction model is combined with load data for analysis to achieve accurate prediction of pavement service life. This invention effectively solves the problem that traditional methods struggle to accurately assess pavement indentation conditions and predict service life through multi-dimensional feature extraction and deep learning models, providing a reliable basis for pavement maintenance decisions.

[0052] It is worth noting that, such as Figure 2 As shown, step S1 includes the following sub-steps:

[0053] S11: Use laser scanning technology to acquire three-dimensional point cloud data containing spatial information of the road surface to obtain the first point cloud; use Gaussian filtering to denoise and downsample the first point cloud to generate a smooth point cloud dataset to obtain the preprocessed first point cloud;

[0054] S12: By setting the growth threshold in the region growth segmentation algorithm, multiple indentation regions are separated in the first point cloud. Each indentation region corresponds to a subset of data containing indentation features, thus obtaining the first dataset. The growth threshold includes the angle between the normal vectors of the point clouds and the distance between two adjacent point clouds.

[0055] For example, laser scanning technology can be used to acquire three-dimensional point cloud data of road surfaces, which can be achieved through a vehicle-mounted LiDAR system. LiDAR emits millions of laser pulses per second, recording the positions of reflection points on the road surface to generate a high-precision spatial coordinate dataset. For instance, in a highway road surface scan, LiDAR collects data at a point spacing of 0.01 meters, forming the first point cloud containing tens of millions of points, accurately reflecting the geometric features of the road surface. This high-density data provides a reliable foundation for subsequent analysis.

[0056] In one possible implementation, Gaussian filtering is used for denoising and downsampling the first point cloud. In principle, Gaussian filtering smooths the point cloud through weighted averaging, reducing noise interference, while downsampling reduces the amount of data. For example, applying Gaussian filtering to the first point cloud with a standard deviation of 0.02 meters filters out random noise points, and then reducing the point spacing from 0.01 meters to 0.05 meters yields the preprocessed first point cloud. The data volume is reduced by approximately 80%, while still retaining the main road surface features, thus improving computational efficiency.

[0057] Specifically, a region growing segmentation algorithm processes the first point cloud to separate indentation regions. This algorithm segments the point cloud into different regions based on spatial proximity and normal vector differences. For example, for road indentations, a growing threshold is set, such as grouping points with a normal vector angle less than 15 degrees and a distance less than 0.03 meters into the same region. In a certain scan, multiple indentation regions were successfully separated, resulting in the first dataset. After segmentation, the point cloud of the indentation regions is independent of other road surface features, facilitating subsequent analysis.

[0058] Preferably, step S12 is followed by step S13, which includes:

[0059] The number of points in each subset is compared with a preset threshold. If the number of points in a subset is lower than the preset threshold, the missing points in the subset are supplemented by a bilinear interpolation algorithm to generate an optimized subset.

[0060] It should be noted that if the number of indentation points in a subset of the first dataset is below a threshold (e.g., less than 1000 points), bilinear interpolation is used to fill the missing points. Bilinear interpolation generates new points to fill sparse regions based on a weighted average of neighboring points. For example, if a certain indentation region has only 800 points, interpolation can be used to supplement it to 1500 points, forming an optimized subset. This method enhances the continuity of the point cloud and improves the accuracy of feature extraction.

[0061] Optional, such as Figure 3 As shown, step S2 includes the following sub-steps:

[0062] S21: Obtain image data for each subset from the first dataset;

[0063] S22: Use the cv2.cvtColor function in the OpenCV library to convert the image data into a grayscale image to obtain the first grayscale image set;

[0064] S23: For each grayscale image in the first grayscale image set, the cv2.Canny function in the OpenCV library is used for edge detection. The high threshold and low threshold are preset as the first threshold and the second threshold, respectively, to extract the indentation boundary data containing the indentation boundary location;

[0065] S24: Integrate the indentation boundary data into its corresponding subset, and combine all subsets to obtain the second feature set.

[0066] Specifically, obtaining image data for each subset from the first dataset typically involves projecting the 3D point cloud onto a 2D plane to generate a corresponding image representation. For example, point cloud projection can be achieved through orthogonal projection, mapping 3D coordinates to a 2D pixel grid. For instance, in the analysis of indentations on a highway surface, a subset containing 1500 indentation areas is projected onto an image with a resolution of 512×512 pixels, where the pixel values ​​reflect the height information of the points. This projection preserves the geometric features of the indentations, facilitating subsequent image processing.

[0067] In one possible implementation, the `cv2.cvtColor` function from the OpenCV library is used to convert the image data into a grayscale image, generating a first set of grayscale images. In principle, `cv2.cvtColor` converts an RGB image into a single-channel grayscale image, reducing color information interference and highlighting the brightness differences in the indentation area. For example, a projected image of an indentation area may contain road texture and shadows; by converting it to a grayscale image, texture details are simplified, and the outline of the indentation area is more easily identified. Preferably, the image data can be preprocessed before conversion, such as adjusting the contrast, to enhance the feature representation of the grayscale image.

[0068] It should be noted that, for the first set of grayscale images, the cv2.Canny function is used for edge detection to extract the indentation boundary data. The Canny algorithm is based on gradient calculation and combines high and low thresholds to filter edge points. For example, the first threshold is set to 200 and the second threshold is set to 100 to detect significant edges in the grayscale images. For instance, if a grayscale image contains the concave contour of an indentation area, the Canny algorithm generates a binary image after detection, retaining only the edge pixels. This method effectively separates the indentation boundary from the background, improving the accuracy of boundary localization. Preferably, for different indentation shapes, the Canny detection threshold can be dynamically adjusted. If the indentation contour is blurred, the second threshold can be reduced to 80 to increase the sensitivity of edge detection.

[0069] Understandably, determining the boundary location information corresponding to the indentation boundary data and generating the second feature set requires mapping the edge pixels back to three-dimensional space. In one embodiment, by recording the coordinate correspondence during projection, the two-dimensional coordinates of the edge pixels are back-projected onto the three-dimensional point cloud to obtain the spatial location of the boundary points. For example, if an indentation edge contains 300 pixels, back-projection generates 300 three-dimensional coordinates, forming the second feature set. These coordinates accurately describe the location of the indentation boundary, facilitating subsequent geometric analysis.

[0070] Specifically, such as Figure 4 As shown, step S3 includes the following sub-steps:

[0071] S31: Obtain the point cloud height data of each subset in the first dataset, calculate the height difference between adjacent point clouds for each subset, and obtain the subset height difference set;

[0072] S32: Extract the maximum, minimum and median values ​​from the height difference set of the subset datasets and integrate them into the corresponding subset datasets to obtain subset datasets containing depth feature data;

[0073] S33: Combine the subset containing deep feature data into a third feature set.

[0074] Specifically, in road surface point cloud data processing, each subset of the first dataset contains the height information of a 3D point cloud, typically represented by Z-axis coordinates. The process of acquiring point cloud height data involves extracting Z-values ​​from the point cloud file to form a height dataset. For example, a highway road surface subset contains 2000 points, each with X, Y, and Z coordinates, where the Z-value reflects changes in road surface height, such as differences in the height of indentations or raised areas. In one possible implementation, the Z-values ​​are directly extracted by traversing the point cloud data in each subset of the first dataset, forming a set of height differences for each subset, facilitating subsequent analysis.

[0075] For example, when calculating the height difference between adjacent point clouds, the definition of adjacent points needs to be determined. This is typically based on the spatial adjacency relationships of the point clouds, such as using the k-nearest neighbor algorithm to determine the nearest neighbor of each point. For instance, a point in a subset of the dataset has a Z-value of 10.5 cm, and its nearest neighbor has a Z-value of 10.8 cm, resulting in a height difference of 0.3 cm. The set of height differences is generated by traversing all points in the subset. It should be noted that the height difference reflects local road surface undulations, such as the steepness of the indentation edges. In one embodiment, the subset contains 1000 height difference values, ranging from -0.5 cm to 0.7 cm, indicating slight unevenness in the road surface.

[0076] Understandably, after extracting the maximum, minimum, and median values ​​from the height difference set of a subset, these values ​​are integrated into their corresponding subsets to obtain subsets containing depth feature data. For example, a subset might have a maximum height difference of 0.7 cm, a minimum of -0.5 cm, and a median of 0.1 cm. These statistical characteristics summarize the magnitude and distribution of local height variations on the road surface.

[0077] It is worth noting that in step S32, the K-means algorithm is used to classify the subset containing deep feature data, determine the category of deep feature data of the subset, and obtain the classification feature set.

[0078] In one possible implementation, the K-means algorithm is used to classify subsets containing depth feature data to determine the categories of depth feature data. For example, the depth feature data categories can be divided into three types: flat, slight indentation, and significant indentation. The feature vector of a subset containing depth feature data (maximum value 0.7, minimum value -0.5, median 0.1) is classified as the slight indentation category. The classification feature set records the category label of each subset containing depth feature data, thus more intuitively reflecting the depth feature data of each subset. It should be noted that K-means iteratively optimizes the cluster centers to ensure that the classification results reflect the inherent differences in road surface features.

[0079] Preferred, such as Figure 5 As shown, step S4 includes the following sub-steps:

[0080] S41: Merge the indentation boundary data in the second feature set and the depth feature data in the third feature set into a feature input matrix; adjust the distribution of the feature input matrix using the Z-score normalization method to obtain the normalized feature input matrix;

[0081] S42: The feature input matrix is ​​propagated forward through a convolutional neural network. The ReLU activation function and Adam optimizer are used to calculate the feature mappings of the convolutional layer and the pooling layer to obtain an initial feature mapping set.

[0082] S43: Using the mean squared error loss function, calculate the error between the initial feature mapping set and the actual value of material hardness, adjust the network parameters through backpropagation to obtain the optimized network parameters, and obtain the mapping relationship based on the optimized network parameters and convergence conditions;

[0083] S44: Based on the mapping relationship, construct a fourth prediction model and calculate the predicted value of material hardness by inputting new indentation boundary data and depth feature data.

[0084] In one embodiment, a highway point cloud subset contains 1000 points. Boundary data records the X, Y, and Z coordinates of the indentation edge points, such as edge point coordinates (50, 30, 20) and (52, 32, 22), and depth features such as a median of 0.2 cm. During merging, the boundary coordinates and depth features are aligned point by point to form a feature input matrix, with each row containing the boundary and depth information of the points.

[0085] Specifically, Z-score standardization adjusts the distribution of the feature input matrix to ensure that features of different dimensions are comparable. For example, boundary coordinate values ​​range from 50 to 100, while depth feature values ​​are smaller, such as 0.1 to 0.5 cm. By calculating the mean and standard deviation of each feature, the standardized values ​​are concentrated in [-1, 1], which is convenient for neural network processing and thus reflects the relative changes in boundary and depth.

[0086] In one embodiment, a convolutional neural network processes the normalized feature matrix, employs the ReLU activation function to enhance the nonlinear expression, and uses the Adam optimizer to accelerate gradient descent. Convolutional layers extract local features, such as the continuity of indentation edges; pooling layers compress data while retaining key information. For example, the feature input matrix is ​​processed by a 3x3 convolutional kernel to generate a feature map reflecting the spatial relationships of edge points.

[0087] Subsequently, based on this mapping, new data is input into the fourth prediction model to predict material hardness. For example, the feature mapping of a certain indentation area corresponds to a predicted material hardness of 50 MPa, indicating that the road surface material is relatively strong.

[0088] Specifically, in step S43, if the optimized network parameters meet the preset convergence condition, the current feature map in the initial feature map set is associated with the material property label to generate a mapping relationship;

[0089] If the optimized network parameters do not meet the preset convergence conditions, the forward and backward propagation are repeated to adjust the network parameters until the convergence conditions are met and the mapping relationship is obtained.

[0090] The mean squared error loss function measures the difference between the feature map and the actual material hardness value; for example, if the predicted material hardness value deviates from the actual value by 0.05, then backpropagation is used to adjust the network parameters until the loss converges, thus constructing the fourth prediction model.

[0091] It is worth noting that, such as Figure 6 As shown, step S5 includes the following sub-steps:

[0092] S51: Obtain the predicted material hardness value output by the fourth prediction model and the traffic flow intensity in the load data to obtain the pavement life prediction dataset; perform standardization processing on the pavement life prediction dataset to obtain a standardized dataset;

[0093] S52: Use the random forest algorithm to extract features from the standardized dataset to obtain a set of feature vectors;

[0094] S53: Based on the set of feature vectors, calculate the importance of each feature, and determine the weight of the weighted fusion mechanism based on the importance of each feature; process the set of feature vectors through the weighted fusion mechanism to obtain a set of road surface life prediction values;

[0095] S54: Select the median from the set of predicted pavement life values ​​to determine the final predicted pavement life value.

[0096] For example, in the pavement life prediction dataset, the predicted material hardness reflects the pavement's compressive strength, while the traffic flow intensity in the load data is statistically analyzed by a road monitoring system, such as an average of 5,000 vehicles per day. In one possible implementation, standardization aims to unify the feature scales of different dimensions. The numerical ranges of hardness and flow characteristics vary considerably, for example, hardness values ​​range from 40 to 60, and flow values ​​range from 1,000 to 10,000. Standardization uses the Z-score method to transform each feature into a distribution with a mean of 0 and a standard deviation of 1. For example, a road segment with a hardness value of 50 MPa becomes 0.5 after standardization, and a flow rate of 8,000 vehicles becomes 0.8. This processing facilitates subsequent algorithmic analysis.

[0097] Specifically, the random forest algorithm is used for feature extraction, evaluating the relationships between features by constructing multiple decision trees. In one embodiment, for 1000 road segment samples, each tree analyzes the impact of material hardness and traffic flow characteristics on lifespan, generating a set of feature vectors. For example, the feature vector for a certain road segment might be [0.5, 0.8], corresponding to material hardness and traffic flow, respectively. This method can capture the nonlinear relationships between features.

[0098] It should be noted that the feature importance calculation is based on the output of the random forest, reflecting the contribution of each feature to the pavement life prediction. For example, the importance of material hardness is 0.6, and the importance of traffic flow is 0.4. The weights are determined accordingly, such as a weight of 0.6 for material hardness and a weight of 0.4 for traffic flow. This weighting method emphasizes the influence of key features. Preferably, the weighted fusion mechanism integrates feature vectors to generate a pavement life prediction set. Specifically, an initial lifespan is set as X. For each road segment, its corresponding pavement life prediction value S = X + 0.6A - 0.4B is calculated, where A is the predicted material hardness value and B is the traffic flow intensity in the load data. For example, after weighted calculation, the feature vector of a certain road segment yields a pavement life prediction value of 15 years. The pavement life prediction set contains pavement life prediction values ​​for 1000 road segments, such as 12 to 18 years. The median of these 1000 pavement life prediction values ​​is selected as the final pavement life prediction value, which can effectively avoid the interference of extreme values. For example, a median of 16 years is used as the final life prediction for a certain road segment.

[0099] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A method for predicting pavement service life based on road construction indentation distribution, characterized in that, Includes the following steps: S1: Acquire three-dimensional point cloud data of the road surface, separate the indentation area from the three-dimensional point cloud data and generate a first dataset; S2: Obtain image data from the first dataset and use an edge detection mechanism to generate a second feature set containing indentation boundary data; S3: Calculate the point cloud height difference based on the height data of the first dataset to obtain a third feature set containing depth feature data; S4: Input the second feature set containing indentation boundary data, the third feature set containing depth feature data, and the measured material hardness value corresponding to the same historical time into the convolutional neural network for training, establish the mapping relationship between indentation features and material hardness, and generate the fourth prediction model, where the indentation features include indentation boundary data and depth feature data; input the current second feature set containing indentation boundary information and the third feature set containing depth features into the fourth prediction model, and output the predicted material hardness value; S5: Calculate the pavement life prediction value based on the material hardness prediction value output by the fourth prediction model and the road load data.

2. The method for predicting pavement service life based on road construction indentation distribution according to claim 1, characterized in that, Step S1 includes the following sub-steps: S11: Use laser scanning technology to acquire three-dimensional point cloud data containing spatial information of the road surface to obtain the first point cloud; S12: By setting the growth threshold in the region growth segmentation algorithm, multiple indentation regions are separated in the first point cloud. Each indentation region corresponds to a subset of data containing indentation features, thus obtaining the first dataset. The growth threshold includes the angle between the normal vectors of the point clouds and the distance between two adjacent point clouds.

3. The method for predicting pavement service life based on road construction indentation distribution according to claim 2, characterized in that: Following step S12, a step S13 is further included, which includes: The number of points in each subset is compared with a preset threshold. If the number of points in a subset is lower than the preset threshold, the missing points in the subset are supplemented by a bilinear interpolation algorithm to generate an optimized subset.

4. The method for predicting pavement service life based on road construction indentation distribution according to claim 3, characterized in that, Step S2 includes the following sub-steps: S21: Obtain image data for each subset from the first dataset; S22: Use the cv2.cvtColor function in the OpenCV library to convert the image data into a grayscale image to obtain the first grayscale image set; S23: For each grayscale image in the first grayscale image set, the cv2.Canny function in the OpenCV library is used for edge detection. The high threshold and low threshold are preset as the first threshold and the second threshold, respectively, to extract the indentation boundary data containing the indentation boundary location; S24: Integrate the indentation boundary data into its corresponding subset, and combine all subsets to obtain the second feature set.

5. The method for predicting pavement service life based on road construction indentation distribution according to claim 4, characterized in that, Step S3 includes the following sub-steps: S31: Obtain the point cloud height data of each subset in the first dataset, calculate the height difference between adjacent point clouds for each subset, and obtain the subset height difference set; S32: Extract the maximum, minimum and median values ​​from the height difference set of the subset datasets and integrate them into the corresponding subset datasets to obtain subset datasets containing depth feature data; S33: Combine the subset containing deep feature data into a third feature set.

6. The method for predicting pavement service life based on road construction indentation distribution according to claim 5, characterized in that, In step S32, the K-means algorithm is used to classify the subset containing deep feature data, determine the category of deep feature data in the subset, and obtain the classification feature set.

7. The method for predicting pavement service life based on road construction indentation distribution according to claim 6, characterized in that, Step S4 includes the following sub-steps: S41: Combine the indentation boundary data in the second feature set and the depth feature data in the third feature set into a feature input matrix; S42: The feature input matrix is ​​propagated forward through a convolutional neural network. The ReLU activation function and Adam optimizer are used to calculate the feature mappings of the convolutional layer and the pooling layer to obtain an initial feature mapping set. S43: Using the mean squared error loss function, calculate the error between the initial feature mapping set and the actual value of material hardness, adjust the network parameters through backpropagation to obtain the optimized network parameters, and obtain the mapping relationship based on the optimized network parameters and convergence conditions; S44: Based on the mapping relationship, construct a fourth prediction model and calculate the predicted value of material hardness by inputting new indentation boundary data and depth feature data.

8. The method for predicting pavement service life based on road construction indentation distribution according to claim 7, characterized in that: In step S43, if the optimized network parameters meet the preset convergence condition, the current feature map in the initial feature map set is associated with the material property label to generate a mapping relationship. If the optimized network parameters do not meet the preset convergence conditions, the forward and backward propagation are repeated to adjust the network parameters until the convergence conditions are met and the mapping relationship is obtained.

9. A method for predicting pavement service life based on road construction indentation distribution according to claim 8, characterized in that, Step S5 includes the following sub-steps: S51: Obtain the predicted material hardness value output by the fourth prediction model and the traffic flow intensity in the load data to obtain the pavement life prediction dataset; perform standardization processing on the pavement life prediction dataset to obtain a standardized dataset; S52: Use the random forest algorithm to extract features from the standardized dataset to obtain a set of feature vectors; S53: Based on the set of feature vectors, calculate the importance of each feature, and determine the weight of the weighted fusion mechanism based on the importance of each feature; process the set of feature vectors through the weighted fusion mechanism to obtain a set of road surface life prediction values; S54: Select the median from the set of predicted pavement life values ​​to determine the final predicted pavement life value.

Images