Pavement anti-skid performance evaluation method and device based on multi-factor coupling

Abstract

The present invention discloses a method and device for evaluating the anti-skid performance of a pavement based on multi-factor coupling, which relates to the field of road traffic safety technology and solves the problem in the prior art that there is no anti-skid performance evaluation method that can be both accurate and efficient. The method comprises: obtaining an asphalt pavement sample set; extracting two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature data set of the sample set; constructing an evaluation model, and training the evaluation model in combination with the multimodal feature data set to obtain a trained evaluation model; wherein the evaluation model comprises: a basis learner, a probability function of a priori distribution, and a scoring function; and utilizing the trained evaluation model to evaluate the anti-skid performance of the asphalt pavement. The method achieves the goal of improving the efficiency and accuracy of the anti-skid performance evaluation of the pavement, and provides a good data basis for road driving safety and pavement maintenance decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of road traffic safety, and in particular to a method and device for evaluating the anti-skid performance of a road surface based on the coupling effect of multiple factors. Background Art

[0002] Pavement skid resistance is a crucial factor in ensuring road safety and a key focus of road maintenance. Therefore, accurately and efficiently assessing pavement skid resistance is crucial. Current methods for assessing pavement skid resistance primarily rely on three-dimensional texture data, but fail to consider the impact of non-depth texture information on the measured area on pavement skid resistance.

[0003] Therefore, there is an urgent need for an accurate and efficient anti-slip performance evaluation method. Summary of the Invention

[0004] The present invention provides a pavement anti-skid performance evaluation method based on multi-factor coupling to improve the efficiency and accuracy of pavement anti-skid performance evaluation, providing a good data basis for road driving safety and pavement maintenance decision-making.

[0005] In a first aspect, the present invention provides a method for evaluating the anti-skid performance of a pavement based on the coupling of multiple factors, the method comprising:

[0006] Obtain asphalt pavement sample set;

[0007] Extracting two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature dataset of the sample set;

[0008] Constructing an evaluation model and training the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model; wherein the evaluation model includes: a base learner, a probability function of a prior distribution, and a scoring function;

[0009] The trained evaluation model is used to evaluate the anti-skid performance of the asphalt pavement.

[0010] In combination with the first aspect, in a possible implementation, the two-dimensional features include: local binary pattern features, variance, energy, and entropy;

[0011] The three-dimensional features include: material volume, skewness, root mean square height and spatial volume.

[0012] In conjunction with the first aspect, in one possible implementation, training the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model includes:

[0013] Defining a base learner, a probability function of a prior distribution, and a scoring function in constructing the evaluation model;

[0014] Using a local surrogate model analysis method to filter the multimodal feature dataset to obtain filtered features;

[0015] The evaluation model is trained using the screened features to obtain a trained evaluation model.

[0016] In conjunction with the first aspect, in one possible implementation, the using a local surrogate model analysis method to filter the multimodal feature dataset to obtain filtered features includes:

[0017] Setting a black box model and a local proxy model, and determining a first input data set of the black box model and a second input data set of the local proxy model; wherein the first input data set belongs to a neighborhood of the second input data set;

[0018] Calculating the weight of each data in the first input data set in the neighborhood to obtain the weight corresponding to each data in the first data set;

[0019] perturbing each data in the first data set to obtain a perturbed first data set;

[0020] Inputting each data in the first data set and each data in the perturbed first data set into the black box model respectively to obtain a first output data set and a first perturbed output data set;

[0021] Calculating the difference between each data in the first output data set and each data in the corresponding first perturbation output data set using the weight to obtain the difference between the corresponding data;

[0022] Training the local proxy model using the second input data set, and adjusting parameters of the local proxy model according to output data of the local proxy model to obtain a trained local proxy model;

[0023] The trained local proxy model and the weight corresponding to each node are used to calculate and evaluate the regression coefficient of each feature, and the multimodal feature dataset is screened according to the regression coefficient to obtain the screened features.

[0024] In conjunction with the first aspect, in one possible implementation, constructing the base learner includes:

[0025] Determining a training data set, a loss function, a maximum depth of the base learner, a minimum number of sample splits, and a minimum number of leaf node samples;

[0026] Initializing the root node of the base learner and using the training data set as a sample set of the root node;

[0027] Determine whether the number of samples in the sample set is less than the minimum number of sample splits, or determine whether a judgment condition is met. If so, mark the root node as a leaf node, and calculate the mean of the target variable in the sample set corresponding to the leaf node as the predicted value of the node and return it;

[0028] Selecting an optimal split point in the sample set, dividing the sample set into a first subset and a second subset according to the optimal split point, and constructing a left child node and a right child node;

[0029] Using the first subset as the sample set of the left child node, using the second subset as the sample set of the right child node, and calculating the predicted value of each node according to the judgment condition through recursive calling;

[0030] Return the base learner based on the predicted value of each node.

[0031] In conjunction with the first aspect, in a possible implementation, the probability function of the prior distribution is expressed as:

[0032]

[0033] Where μ represents the location parameter, b represents the scale parameter, e represents the natural constant, and x represents a single observation.

[0034] In conjunction with the first aspect, in a possible implementation, the scoring function is expressed as:

[0035]

[0036] Where μ represents the location parameter, b represents the scale parameter, and x represents a single observation.

[0037] In a second aspect, the present invention provides a device for evaluating the anti-skid performance of a road surface based on the coupling of multiple factors, the device comprising:

[0038] A sample acquisition unit, used for acquiring an asphalt pavement sample set;

[0039] A data set acquisition unit, configured to extract two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature data set of the sample set;

[0040] A training module is used to construct an evaluation model and train the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model;

[0041] The evaluation module is used to evaluate the anti-skid performance of the asphalt pavement using the trained evaluation model.

[0042] In a third aspect, the present invention provides a road surface anti-skid performance evaluation server based on multi-factor coupling, the server including a memory and a processor;

[0043] The memory is used to store computer-executable instructions;

[0044] The processor is used to execute the computer executable instructions to implement a road surface anti-skid performance evaluation method based on multi-factor coupling.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having executable instructions, which can implement a road surface anti-skid performance evaluation method based on multi-factor coupling when a computer executes the executable instructions.

[0046] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0047] (1) The present invention provides a regularization mechanism for model parameters by using a prior distribution. Different prior distributions impose different constraints on the distribution of model parameters, thereby affecting the complexity and generalization ability of the model. Selecting an appropriate prior distribution helps prevent overfitting and improves the performance of the model on unknown data.

[0048] (2) The present invention adopts two-dimensional features and three-dimensional features that are different from traditional methods. The two-dimensional features (two-dimensional texture information) complement the three-dimensional features (three-dimensional structural information). This comprehensive method can more accurately reflect the actual situation of the road surface, avoid the deviations and limitations that may be brought about by a single method, and has higher accuracy and universality than traditional methods.

[0049] (3) The present invention uses a local proxy model analysis method to select important features that affect anti-skid performance, and assists in building a more accurate anti-skid performance evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments of the present invention or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 A schematic flow chart of a method for evaluating the anti-skid performance of a road surface based on multi-factor coupling provided by an embodiment of the present invention;

[0052] Figure 2 This is an example diagram of multimodal data collection for asphalt pavement texture provided by an embodiment of the present invention;

[0053] Figure 3 A diagram showing the construction of an anti-skid performance evaluation model based on a natural gradient boosting tree according to an embodiment of the present invention;

[0054] Figure 4 A schematic diagram of a priori probability distribution function of the coefficient of kinetic friction provided by an embodiment of the present invention;

[0055] Figure 5 A schematic diagram of a local proxy model analysis method provided by an embodiment of the present invention;

[0056] Figure 6 This is a test result diagram of the anti-skid performance evaluation model based on the natural gradient boosting tree provided by an embodiment of the present invention;

[0057] Figure 7 The anti-slip performance evaluation results of the measured data on the representative specimen 1-AC16 provided in the embodiment of the present invention;

[0058] Figure 8 The anti-slip performance evaluation results of the measured data on the representative specimen 2-OGFC13 provided in the embodiment of the present invention;

[0059] Figure 9 The anti-slip performance evaluation results of the measured data on the representative specimen 3-SMA13 provided in the embodiment of the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0061] In the first aspect, the present invention provides a method for evaluating the anti-skid performance of a road surface based on the coupling of multiple factors. Figure 1 The method includes the following steps S101 to S104.

[0062] S101: Obtain an asphalt pavement sample set.

[0063] For example, refer to Figure 2 , Figure 2This is an example of multimodal data collection for asphalt pavement texture. First, a 200mm*100mm test area was selected on the surface of a 500mm*500mm specimen. This test area was ensured to be as consistent as possible with the sliding area of ​​the rubber slider on the chassis of the dynamic friction coefficient tester. Surface image data, surface texture data, test temperature data, and dynamic friction coefficient data were then measured in these test areas to create an asphalt pavement sample set. The sample set contains 937 samples collected from 18 asphalt specimens of different gradations.

[0064] S102 , extracting two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature dataset of the sample set.

[0065] For example, two-dimensional features such as local binary pattern features, mean, variance, energy, entropy, etc. are extracted through two-dimensional image samples; three-dimensional features such as height parameters, functional parameters, volume parameters, etc. are extracted through three-dimensional point cloud samples; and then a multimodal feature data set is constructed by combining two-dimensional features, three-dimensional features, dynamic friction coefficient, etc.

[0066] S103, constructing an evaluation model, and training the evaluation model in combination with the multimodal feature data set to obtain a trained evaluation model; wherein the evaluation model includes: a base learner, a probability function of a prior distribution, and a scoring function. Figure 3 Constructing a graph for the anti-skid performance evaluation model based on natural gradient boosting trees.

[0067] Specifically, in step S103, the evaluation model is trained in combination with the multimodal feature dataset to obtain a trained evaluation model, which includes the following steps S1031 to S1033.

[0068] S1031, define the base learner, the probability function of the prior distribution, and the scoring function in constructing the evaluation model.

[0069] Specifically, the construction of the base learner includes:

[0070] (1) Determine the training dataset D, loss function, maximum depth of the base learner, minimum number of sample splits, and minimum number of leaf node samples.

[0071] (2) Initialize the root node root of the base learner and use the training dataset D as the sample set of the root node.

[0072] (3) Determine whether the number of samples in the sample set D is less than the minimum number of sample splits min_samples_split, or determine whether the judgment condition is met. If so, mark the root node as a leaf node, and calculate the mean of the target variable in the sample set corresponding to the leaf node as the predicted value of the node and return it. Here, the judgment conditions mentioned are: one is to determine whether the number of samples in the training data set D of the current node is less than the minimum number of sample splits min_samples_split; the other is: the depth of the current node reaches the maximum depth max_depth. If either of the two judgment conditions is met, it is recorded as meeting the judgment condition.

[0073] (4) Select the optimal split point in the sample set, divide the sample set into the first subset D1 and the second subset D2 according to the optimal split point, and construct the left child node left and the right child node right;

[0074] (5) The first subset D1 is used as the sample set of the left child node left, and the second subset D2 is used as the sample set of the right child node right. The predicted value of each node is calculated according to the judgment condition based on the recursive call;

[0075] (6) Return the base learner based on the predicted value of each node. Specifically, the predicted value of the current node is the weighted average of the predicted values ​​of the left and right child nodes, where the weights are the proportion of the number of samples in the left and right child node sample sets D1 and D2 to the total number of samples.

[0076] In addition, the main goal of the natural gradient boosting tree is to improve the prediction accuracy by predicting the distribution of the target variable, rather than just predicting a point estimate. Therefore, it is important to estimate the prior distribution of the target variable. Different prior distribution estimates will have the following effects on the model:

[0077] (1) Regularization effect: The prior distribution actually provides a regularization mechanism for model parameters. Different prior distributions impose different constraints on the distribution of model parameters, thereby affecting the complexity and generalization ability of the model. Choosing an appropriate prior distribution helps prevent overfitting and improve the performance of the model on unknown data.

[0078] (2) Uncertainty estimation: The choice of prior distribution directly affects the accuracy and reliability of uncertainty estimation. Certain prior distributions may be more suitable for processing specific types of data or problems, thereby providing more accurate uncertainty estimates.

[0079] (3) Shape of the predictive distribution: The prior distribution determines the basic shape and characteristics of the predictive distribution. Different prior distributions will lead to different predictive distribution shapes, which may affect the model's ability to handle extreme values ​​or outliers. Choosing an appropriate prior distribution helps to better fit the actual distribution of the data and improve the model's prediction accuracy.

[0080] For example, the base learner is defined as a decision tree model with a maximum depth of 4 and a loss function of Feldman mean square error. The decision tree model inputs are: training data set D, loss function FMSE, maximum depth max_depth, minimum number of sample splits min_samples_split, and minimum number of leaf node samples min_samples_leaf. Where n is the number of samples, is the true value, y i is the predicted value.

[0081] Decision tree model output: decision tree model Tree.

[0082] S1032: Use a local surrogate model analysis method to filter the multimodal feature dataset to obtain filtered features.

[0083] S1033: Train the evaluation model using the filtered features to obtain a trained evaluation model.

[0084] In a specific embodiment provided by the present invention, the data distribution of the dynamic friction coefficient is inferred by using existing sample data. For details, please refer to Figure 4 , Figure 4 Figure 2 is a schematic diagram of the prior probability distribution function of the kinetic friction coefficient. The prior distribution inferred from the measured kinetic friction coefficient sample is closest to the Laplace distribution with a position parameter of 0.32 and a scale parameter of 0.05. Therefore, in this embodiment, the probability distribution function of the prior distribution is defined as the Laplace distribution, and the probability density function of the Laplace distribution is defined as:

[0085]

[0086] Where μ is the location parameter, b is the scale parameter, and x is a single observation. Random variables with a scale parameter that follows a Laplace distribution have a much higher probability of occurring extremely large values ​​than those with a normal distribution.

[0087] In this embodiment, the scoring function uses a logarithmic score. The logarithmic score is generally used to evaluate the degree of match between the predicted probability distribution and the observed value. The higher the logarithmic score, the closer the predicted distribution is to the true distribution, and the better the predictive performance of the model. In the context of the Laplace distribution, the logarithmic score will be the log-likelihood function of the Laplace distribution. For a single observation x, the log-likelihood function of the Laplace distribution can be expressed as:

[0088]

[0089] Where μ represents the location parameter, b represents the scale parameter, and x represents a single observation.

[0090] Specifically, in step S1032, the multimodal feature dataset is screened using the local proxy model analysis method to obtain the filtered features. Figure 5 , including the following steps.

[0091] (1) Set up a black box model and a local proxy model, and determine the first input data set of the black box model and the second input data set D of the local proxy model. For each data instance x to be explained i ∈D; where the first input dataset belongs to the neighborhood Z of the second input dataset, z i ∈Z, this neighborhood contains the i Similar data points.

[0092] (2) Calculate the weight of each data in the neighborhood of the first input data set to obtain the weight corresponding to each data in the first data set. The specific calculation formula is expressed as:

[0093]

[0094] Where D(x,z) represents the distance metric between x and z (such as Euclidean distance), and σ represents a parameter that controls the width of the weight distribution.

[0095] Weight π x (z) measures the similarity between z and x, and the more similar the points are, the greater the weight.

[0096] (3) Perform perturbation on each data in the first data set to obtain the perturbed first data set.

[0097] (4) Input each data point in the first data set and each data point in the perturbed first data set into the black box model to obtain a first output data set and a first perturbed output data set. For example, for each data point z in the neighborhood Z i , generate its perturbed version z i ′ The perturbation is achieved by changing z iThis is achieved by changing certain feature values ​​of to observe how these changes affect the prediction results of the black box model.

[0098] (5) Using the weights, calculate the difference between each data in the first output data set and each data in the corresponding first perturbation output data set to obtain the difference between the corresponding data. For example, calculate the black box model Local Proxy Model The difference between them, ξ(x), is defined as follows:

[0099]

[0100] This difference is obtained by summing the black box model predictions for each data point z (and its perturbed version z′) and the predicted value of the local surrogate model The difference between them, and multiplied by the corresponding weight π x (z). This difference reflects the error of the local proxy model in approximating the black-box model.

[0101] (6) The local proxy model is trained using the second input data set, and the parameters of the local proxy model are adjusted according to the output data of the local proxy model to obtain a trained local proxy model. For example, in order to minimize the difference ξ(x), a local proxy model needs to be trained. This model is usually simple, such as linear regression or decision tree, for ease of interpretation. Here, the local surrogate model is selected as ridge regression, which is defined as follows:

[0102]

[0103] in, Represents the regression coefficient vector. X represents the matrix of input data. T represents the transposed matrix of X. y represents the vector of target variables. λ represents the ridge coefficient, which is a regularization parameter. E represents the identity matrix.

[0104] During training, the local proxy model is adjusted parameters to make its prediction results as close as possible to the black box model Prediction results within the neighborhood Z.

[0105] (7) Using the trained local proxy model and the weight corresponding to each node, the regression coefficient of each feature is calculated and evaluated, and the multimodal feature data set is screened according to the regression coefficient to obtain the screened features.

[0106] For example, by analyzing the local agent model The coefficient or weight of can evaluate the importance of each feature to the prediction result. Specifically, the local surrogate model Each feature is assigned a regression coefficient, which reflects the i The degree of influence of ∈X on the target variable y.

[0107] Specifically, if the absolute value of a feature's regression coefficient is large, then the feature has a greater impact on the prediction results and is considered more important in the model. Conversely, if the absolute value of the regression coefficient is small, then the feature has a smaller impact on the prediction results and is considered less important. Based on these importance assessments, we can select the key features that have the greatest impact on the prediction results.

[0108] Through the local proxy model analysis method, the selected features include the test speed 'Speed', the material volume of the macro texture 'Vmma', the skewness of the macro texture 'Sskma', the root mean square height of the macro texture 'Sqma', the spatial volume of the micro texture 'Vvmi', the material height difference of the macro texture 'Sdcma', the core material volume of the macro texture 'Vmcma', the test temperature 'Temp', the kurtosis of the micro texture 'Skumi', the skewness of the micro texture 'Sskmi', etc., and the anti-slip performance evaluation model based on the natural gradient boosting tree is trained.

[0109] S104: Use the trained evaluation model to evaluate the anti-skid performance of the asphalt pavement.

[0110] For example, please refer to Figure 6 , Figure 6 This is a test result diagram of the anti-skid performance evaluation model based on the natural gradient boosting tree. The test set data is input into the trained anti-skid performance evaluation model based on the natural gradient boosting tree. The red broken line represents the actual measured value of the dynamic friction coefficient, and the blue broken line represents the predicted value of the anti-skid performance evaluation model based on the natural gradient boosting tree on the test set samples.

[0111] In addition, in this embodiment, the anti-skid performance evaluation model based on the natural gradient boosting tree is compared with LinearRegressor and XGBoostRegressor. Specifically, the obtained accuracy and error evaluation indicators are shown in Table 1. Table 1 is a comparison table of the anti-skid performance evaluation effects of multiple models. The comparative experimental results show that the determination coefficient of the anti-skid performance evaluation model based on the natural gradient boosting tree on the test set can reach 0.9835, which is significantly higher than the simple linear regressor, and is also slightly better than the existing mainstream machine learning method XGBoost model.

[0112] Table 1 Comparison of experimental results

[0113]

[0114] Finally, in this embodiment, through experimental verification, the measured data of the test pieces AC16, OGFC13, and SMA13 were verified, and the R2 values ​​were 0.9538, 0.882, and 0.8965 respectively. Figure 7 、 Figure 8 、 Figure 9 , respectively, show the anti-skid performance evaluation results based on measured data from representative specimens 1 (AC16), 2 (OGFC13), and 3 (SMA13). Dotted lines represent measured values, while asterisk-marked lines represent evaluated values. It's important to note that the data obtained from specimens AC16, OGFC13, and SMA13 were not used in the model training process. This data is used solely to evaluate the effectiveness of the natural gradient boosting tree-based anti-skid performance evaluation model on friction coefficient at different test speeds.

[0115] In a second aspect, the present invention provides a device for evaluating the anti-skid performance of a road surface based on the coupling of multiple factors. The device includes: a sample acquisition unit, a data set acquisition unit, a training module, and an evaluation module.

[0116] The sample acquisition unit is used to acquire an asphalt pavement sample set.

[0117] The data set acquisition unit is used to extract the two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature data set of the sample set.

[0118] The training module is used to build an evaluation model and train the evaluation model in combination with a multimodal feature dataset to obtain a trained evaluation model.

[0119] The evaluation module is used to evaluate the anti-skid performance of asphalt pavement using the trained evaluation model.

[0120] The devices or modules described in the above embodiments can be implemented as computer chips or physical devices, or as products with certain functions. For ease of description, the above devices are described separately by function, with each module described separately. When implementing the present invention, the functions of each module can be implemented in the same or multiple software and / or hardware components. Of course, a module that implements a certain function can also be implemented by combining multiple submodules or subunits.

[0121] The methods, devices, or modules described in the present invention can be implemented in the form of computer-readable program code. The controller can be implemented in any appropriate manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to implement the same function of the controller in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be considered as structures within the hardware component. Or even, the means for implementing various functions may be considered to be both a software module for implementing the method and a structure within a hardware component.

[0122] Some modules within the apparatus described herein may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, and the like that perform specific tasks or implement specific abstract data types. The present invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communications network. In a distributed computing environment, program modules may reside in local and remote computer storage media, including storage devices.

[0123] The present invention provides a road surface anti-skid performance evaluation server based on the coupling effect of multiple factors. The server includes a memory and a processor; the memory is used to store computer-executable instructions; the processor is used to execute the computer-executable instructions to implement a road surface anti-skid performance evaluation method based on the coupling effect of multiple factors.

[0124] The present invention provides a computer-readable storage medium having executable instructions. When a computer executes the executable instructions, a road surface anti-skid performance evaluation method based on multi-factor coupling can be implemented.

[0125] The above-mentioned storage medium includes, but is not limited to, random access memory (RAM), read-only memory (ROM), cache, hard disk drive (HDD), or memory card. The memory can be used to store computer program instructions.

[0126] Although the present invention provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative work. The order of steps listed in this embodiment is only one way of executing the order of many steps and does not represent the only execution order. When an actual device or client product is executed, it can be executed sequentially or in parallel according to the method shown in this embodiment or the accompanying drawings (for example, in a parallel processor or multi-threaded processing environment).

[0127] Through the description of the above implementation methods, it can be seen that those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, or can be embodied through the implementation process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention or certain parts of the embodiments.

[0128] The various embodiments in this specification are described in a progressive manner. References to the same or similar parts between the various embodiments are sufficient. Each embodiment focuses on the differences from other embodiments. All or part of the present invention can be used in a variety of general or specialized computer system environments or configurations. For example, personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments that include any of the above systems or devices.

[0129] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the present invention.

Claims

1. A method for evaluating the anti-skid performance of a road surface based on the coupling of multiple factors, characterized in that: include: Obtain asphalt pavement sample set; Extracting two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature dataset of the sample set; The two-dimensional features include: local binary pattern features, variance, energy and entropy; the three-dimensional features include: material volume, skewness, root mean square height and spatial volume; Constructing an evaluation model, and training the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model; wherein the evaluation model includes: a base learner, a probability function of a prior distribution, and a scoring function; training the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model includes: defining the base learner, the probability function of the prior distribution, and the scoring function in the constructed evaluation model; using a local surrogate model analysis method to filter the multimodal feature dataset to obtain filtered features; and using the filtered features to train the evaluation model to obtain a trained evaluation model; The trained evaluation model is used to evaluate the anti-skid performance of the asphalt pavement.

2. The method for evaluating the anti-skid performance of a road surface based on multi-factor coupling according to claim 1, characterized in that: The multimodal feature dataset is screened using a local surrogate model analysis method to obtain screened features, including: Setting a black box model and a local proxy model, and determining a first input data set of the black box model and a second input data set of the local proxy model; wherein the first input data set belongs to a neighborhood of the second input data set; Calculating the weight of each data in the first input data set in the neighborhood to obtain the weight corresponding to each data in the first data set; perturbing each data in the first data set to obtain a perturbed first data set; Inputting each data in the first data set and each data in the perturbed first data set into the black box model respectively to obtain a first output data set and a first perturbed output data set; Calculating the difference between each data in the first output data set and each data in the corresponding first perturbation output data set using the weight to obtain the difference between the corresponding data; Training the local proxy model using the second input data set, and adjusting parameters of the local proxy model according to output data of the local proxy model to obtain a trained local proxy model; The trained local proxy model and the weight corresponding to each node are used to calculate and evaluate the regression coefficient of each feature, and the multimodal feature dataset is screened according to the regression coefficient to obtain the screened features.

3. The method for evaluating the anti-skid performance of a road surface based on multi-factor coupling according to claim 1, characterized in that: The construction of the base learner includes: Determining a training data set, a loss function, a maximum depth of the base learner, a minimum number of sample splits, and a minimum number of leaf node samples; Initializing the root node of the base learner and using the training data set as a sample set of the root node; Determine whether the number of samples in the sample set is less than the minimum number of sample splits, or determine whether a judgment condition is met. If so, mark the root node as a leaf node, and calculate the mean of the target variable in the sample set corresponding to the leaf node as the predicted value of the node and return it; Selecting an optimal split point in the sample set, dividing the sample set into a first subset and a second subset according to the optimal split point, and constructing a left child node and a right child node; Using the first subset as the sample set of the left child node, using the second subset as the sample set of the right child node, and calculating the predicted value of each node according to the judgment condition through recursive calling; Return the base learner based on the predicted value of each node.

4. The method for evaluating the anti-skid performance of a road surface based on multi-factor coupling according to claim 1, characterized in that: The probability function of the prior distribution is expressed as: in, Represents positional parameters; represents the scale parameter; represents a natural constant; Represents a single observation.

5. The method for evaluating the anti-skid performance of a road surface based on multi-factor coupling according to claim 1, characterized in that: The scoring function is expressed as: in, Represents positional parameters; represents the scale parameter; Represents a single observation.

6. A road surface anti-skid performance evaluation device based on multi-factor coupling, characterized in that: include: A sample acquisition unit, used for acquiring an asphalt pavement sample set; A data set acquisition unit, configured to extract two-dimensional features and three-dimensional features of each sample in the sample set to obtain a multimodal feature data set of the sample set; The two-dimensional features include: local binary pattern features, variance, energy and entropy; the three-dimensional features include: material volume, skewness, root mean square height and spatial volume; A training module is used to construct an evaluation model and train the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model; the training of the evaluation model in combination with the multimodal feature dataset to obtain a trained evaluation model includes: defining a base learner, a probability function of a prior distribution, and a scoring function in the constructed evaluation model; using a local surrogate model analysis method to filter the multimodal feature dataset to obtain filtered features; and using the filtered features to train the evaluation model to obtain a trained evaluation model; The evaluation module is used to evaluate the anti-skid performance of the asphalt pavement using the trained evaluation model.

7. A road surface anti-skid performance evaluation server based on multi-factor coupling, characterized in that: including memory and processor; The memory is used to store computer-executable instructions; The processor is used to execute the computer-executable instructions to implement the road surface anti-skid performance evaluation method based on multi-factor coupling as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium has executable instructions, and when a computer executes the executable instructions, it can implement the road surface anti-skid performance evaluation method based on multi-factor coupling as described in any one of claims 1 to 5.

Images