A method, system, device and medium for identifying key influencing factors of road performance

Abstract

The present invention belongs to the technical field related to road engineering. In order to solve the problems of inaccurate identification of key influencing factors of road performance status and lack of specific analysis, a method, system, equipment and medium for identifying key influencing factors of road state are proposed. Boundary boxes and segmentation masks are used to determine whether road diseases are regular diseases, so as to achieve accurate calculation of the disease size; pavement state influencing factors of pavement damage condition indicators are predicted using multiple regression prediction models, and weights are assigned to multiple regression prediction models based on MACBETH method to identify key influencing factors of road state; and a logistic regression model is introduced to further analyze the specific impact of key influencing factors of road state on road performance at different levels, which is helpful to extend the service life of roads and optimize resource allocation.

Description

Technical Field

[0001] The present invention belongs to the technical field related to road engineering, and in particular relates to a method, system, equipment and medium for identifying key influencing factors of road performance. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] During the use of highway pavement, under the influence of various uncertain factors such as environmental climate, traffic load, maintenance and repair level, the road performance status will show a decay trend, and the road surface will be damaged to varying degrees, resulting in reduced pavement performance and shortened service life. At present, the evaluation of road performance status mainly relies on traditional detection methods and standards, calculating the pavement performance index based on pavement diseases, and then evaluating the road performance status.

[0004] However, existing road condition assessment methods often lack detailed analysis of road disease size, resulting in a lack of key data support for road management personnel when making maintenance decisions. Although existing road maintenance decisions focus on the identification of key factors affecting road performance, there are inaccurate identifications and a lack of detailed analysis of the impact of different levels of key factors on road performance, which affects the scientific preventive maintenance of road conditions. Summary of the invention

[0005] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a method, system, equipment and medium for identifying key influencing factors of road performance, so as to achieve a detailed analysis of the size of road defects and the impact of different levels of key influencing factors of road performance on road performance, and subsequently provide scientific and reasonable maintenance decision-making suggestions for road management personnel.

[0006] To achieve the above object, a first aspect of the present invention provides a method for identifying key influencing factors of road performance, comprising:

[0007] For the acquired road remote sensing images, based on the trained road recognition model, the bounding box and segmentation mask for road damage recognition are obtained;

[0008] Based on the obtained bounding box and segmentation mask of road damage recognition, determine whether the road damage is a regular damage; if it is a regular road damage, calculate the damage area using the obtained bounding box; if it is an irregular road damage, calculate the road damage area using the obtained segmentation mask;

[0009] Calculate the road surface damage index based on the road disease area obtained;

[0010] Based on the pavement damage condition index and the factors affecting pavement performance, multiple regression prediction models are used for prediction, and weights are assigned to the multiple regression prediction models based on the MACBETH method to identify key factors affecting road performance;

[0011] The Logistic model is used to analyze the impact of different levels of the identified key influencing factors of road performance on the pavement damage condition indicators.

[0012] A second aspect of the present invention provides a road performance key influencing factor identification system, comprising:

[0013] The recognition module is used to obtain the bounding box and segmentation mask of road damage recognition based on the trained road recognition model for the acquired road remote sensing image;

[0014] A judgment module is used to judge whether the road damage is a regular damage based on the obtained bounding box and segmentation mask of the road damage identification. If it is a regular road damage, the damage area is calculated using the obtained bounding box; if it is an irregular road damage, the road damage area is calculated using the obtained segmentation mask;

[0015] A calculation module, used to calculate the road surface damage condition index according to the obtained road disease area;

[0016] An identification module is used to make predictions based on the pavement damage condition index and the factors affecting pavement performance using multiple regression prediction models, assign weights to the multiple regression prediction models based on the MACBETH method, and identify key factors affecting road performance;

[0017] The analysis module is used to analyze the influence of different levels of the identified key influencing factors of road performance on the road damage condition indicators by using the Logistic model.

[0018] The third aspect of the present invention provides a computer device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, a method for identifying key influencing factors of road performance is performed.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, a method for identifying key influencing factors of road performance is executed.

[0020] One or more of the above technical solutions have the following beneficial effects:

[0021] In the present invention, a bounding box and a segmentation mask are used to determine whether a road defect is a regular defect. If it is a regular road defect, the bounding box is used to calculate the defect area; if it is an irregular road defect, the segmentation mask is used to calculate the road defect area, so as to achieve accurate calculation of the defect size; multiple regression prediction models are used to predict the pavement performance influencing factors of the pavement damage condition index, and weights are assigned to multiple regression prediction models based on the MACBETH method to identify key road performance influencing factors; and a Logistic model is introduced to further analyze the specific impact of key road performance influencing factors on road performance at different levels, so as to provide road management personnel with more scientific and reasonable maintenance decisions, which is helpful to extend the service life of the road and optimize resource allocation.

[0022] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0024] Figure 1 This is a flow chart of a method for identifying key influencing factors of road performance in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0025] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0026] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0027] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0028] Terminology explanation:

[0029] YOLACT model: A single-stage, real-time instance segmentation method that achieves efficient instance segmentation through parallel object detection and instance segmentation branches.

[0030] MACBETH method: A judgment-based, multi-criteria decision-making method for making relative assessments of the attractiveness of different options or criteria.

[0031] Decision tree: A method of approximating the value of a discrete function and a typical classification method.

[0032] Random Forest: Combines multiple decision trees together, each time the data set is randomly selected with replacement, and some features are randomly selected as input.

[0033] XGBoost: An ensemble learning algorithm based on gradient boosting decision trees, introducing regularization terms and second-order derivative information.

[0034] LightGBM: A distributed gradient boosting framework based on decision tree algorithm.

[0035] Score standardization method: convert the original data with different means and standard deviations into a standard normal distribution with the same mean of 0 and standard deviation of 1.

[0036] Ordered Logistic Model: A statistical analysis method for dealing with ordered categorical variables.

[0037] Embodiment 1

[0038] This embodiment discloses a method for identifying key influencing factors of road performance, including:

[0039] For the acquired road remote sensing images, based on the trained road recognition model, the bounding box and segmentation mask for road damage recognition are obtained;

[0040] Based on the obtained bounding box and segmentation mask of road damage recognition, determine whether the road damage is a regular road damage. If it is a regular road damage, calculate the damage area using the obtained bounding box; if it is an irregular road damage, calculate the road damage area using the obtained segmentation mask;

[0041] Calculate the road surface damage index based on the road disease area obtained;

[0042] Based on the pavement damage condition index and the factors affecting pavement performance, multiple regression prediction models are used for prediction, and weights are assigned to the multiple regression prediction models based on the MACBETH method to identify key factors affecting road performance;

[0043] The Logistic model is used to analyze the impact of different levels of the identified key influencing factors of road performance on the pavement damage condition indicators.

[0044] This embodiment uses a bounding box and a segmentation mask to determine whether a road defect is a regular defect. If it is a regular road defect, the bounding box is used to calculate the defect area; if it is an irregular road defect, the segmentation mask is used to calculate the road defect area, thereby achieving accurate calculation of the defect size; multiple regression prediction models are used to predict the pavement performance influencing factors of the pavement damage condition index, and weights are assigned to multiple regression prediction models based on the MACBETH method to identify key road performance influencing factors; and a Logistic model is introduced to further analyze the specific impact of key road performance influencing factors on road performance at different levels, thereby providing road management personnel with more scientific and reasonable maintenance decisions, which helps to extend the service life of roads and optimize resource allocation.

[0045] Combine the following Figure 1 A method for identifying key influencing factors of road performance proposed in this embodiment is described in detail, specifically including:

[0046] Step 1: Identify pavement defects based on low-altitude remote sensing images of road surfaces and calculate pavement performance indicators.

[0047] This embodiment uses drones to collect low-altitude remote sensing images of road surfaces, and uses the YOLACT deep learning model to identify detailed information on road surface defects, including the length and width of the defects, and calculates road surface performance indicators.

[0048] In step 1 of this embodiment, the following steps are specifically included:

[0049] Step 11: Use a drone equipped with a high-definition camera to track the center line of the road to collect low-altitude remote sensing images of the road surface, and perform pre-processing such as denoising and image enhancement on the collected images.

[0050] Step 12: Based on the public road damage dataset on the Internet, train the YOLACT deep learning model to identify road pavement damage and collect detailed information on road pavement damage, including the location, length, width, and other information of the damage.

[0051] In step 12 of this embodiment, the following steps are specifically included:

[0052] Step 121: Obtain a road damage dataset publicly available on the Internet, which includes well-labeled road damage images and covers a variety of common damage types, such as cracks, repairs, and cracks; and clean and enhance the obtained road damage dataset publicly available on the Internet.

[0053] Step 122: Divide the preprocessed online public road disease dataset into a training set, a validation set, and a test set. Use the training set to train the road disease recognition model based on the YOLACT deep learning model, use the validation set to tune the model parameters, and use the test set to evaluate the model performance to ensure the high accuracy of the model in disease recognition.

[0054] Step 123: Use the trained YOLACT pavement defect recognition model to identify defects in the low-altitude remote sensing image of the road surface preprocessed in step 11, and output the boundary box, classification, and size information of the defect as basic data for calculating the pavement performance index.

[0055] In step 123 of this embodiment, the following steps are specifically included:

[0056] Step 123-1: Input the preprocessed road surface remote sensing image into the trained YOLACT model to detect defects in the image.

[0057] The YOLACT model first identifies the diseased areas in the image and generates a bounding box for each disease, corresponding to a classification label indicating the disease type.

[0058] Bounding box coordinates: ,in, is the coordinate of the upper left corner, is the lower right corner coordinate.

[0059] Step 123-2: The bounding box cannot reflect the actual shape of the disease, especially for irregularly shaped diseases. At the same time, the YOLACT model generates a segmentation mask, in which the pixel value of the pixel belonging to the disease area is 1, and the pixel value of the pixel belonging to the background is 0. The segmentation mask accurately describes the specific shape and location of the disease in the image, and outputs .in and are the pixel coordinates of the diseased area in the mask, respectively.

[0060] Step 13: Use the YOLACT model results to calculate the detailed information on the length and width of each type of road pavement defects and establish a defect dataset for the road section.

[0061] In step 13 of this embodiment, the following steps are specifically included:

[0062] Step 131: Determine whether the defect in the image is a regular-shaped defect.

[0063] For regular-shaped defects, the coordinates of the bounding box and the segmentation mask are relatively consistent, and the bounding box coordinates of the defect are directly used to calculate the defect size.

[0064] Length calculation formula:

[0065] (1)

[0066] Width calculation formula:

[0067] (2)

[0068] The length and width are the pixel distances of the disease in the image, which can be converted into the actual size of the disease by multiplying it by the spatial resolution of the image (such as the actual distance represented by each pixel).

[0069] The calculation formula for the coverage area of ​​regular shape diseases is:

[0070] (3)

[0071] Step 132: For irregular shaped diseases, the bounding box cannot accurately describe the actual shape and size of the disease, and a segmentation mask is required for more accurate calculation. The minimum enclosing rectangle method is used to calculate the size of the disease.

[0072] Length calculation formula:

[0073] (4)

[0074] Width calculation formula:

[0075] (5)

[0076] The pixel area of ​​the disease can be calculated by counting the total number of pixels with a value of 1 in the segmentation mask, and the actual area of ​​the disease can be obtained by multiplying the pixel area by the actual area corresponding to a single pixel (i.e., the square of the resolution).

[0077] Step 133: The calculated defect size is reasonably verified by comparing it with the results of traditional detection methods. After the effectiveness of the verification method is verified, the pavement defect size is output.

[0078] Step 14: Based on the road section disease dataset, calculate the pavement damage condition PCI in accordance with the Highway Technical Condition Assessment Standard JTG 5210-2018.

[0079] Further preferably, the specific implementation steps of step 14 are as follows:

[0080] Step 141: The calculation formula for the pavement condition index (PCI) is as follows:

[0081] (6)

[0082] (7)

[0083] in, is the road surface damage rate, unit: ; The value of: 15.00 for asphalt pavement and 10.66 for cement concrete pavement; The value of is: 0.412 for asphalt pavement and 0.461 for cement concrete pavement; It is The cumulative area of ​​road damage of this type, unit: ; is the road surface inspection or investigation area, unit: ; It is The weight or conversion factor of the road surface damage is as follows: in detail, the weight of cracking of asphalt pavement is 1.0, the weight of block crack is 0.8, the weight of longitudinal crack is 2.0, the weight of transverse crack is 2.0, the weight of subsidence is 1.0, the weight of wave bulging is 1.0, the weight of pothole is 1.0, the weight of looseness is 1.0, the weight of oil spill is 0.2, the weight of block repair is 0.1, and the weight of strip repair is 0.2. The weight of broken plate of cement concrete pavement is 1.0, the weight of crack is 10, the weight of plate corner fracture is 1.0, the weight of misalignment is 10, the weight of arch is 1.0, the weight of edge peeling is 10, the weight of joint material damage is 6, the weight of pothole is 1.0, the weight of mud is 10, the weight of exposed is 0.3, the weight of block repair is 0.1, and the weight of strip repair is 0.2; is the type of pavement damage, including the degree of damage (light, moderate, or severe); is the total number of damage types, 21 for asphalt pavement.

[0084] Step 2: Create a regression prediction model for pavement performance indicators and identify key influencing factors of road performance.

[0085] This embodiment uses the pavement damage condition PCI obtained in step 1 as the dependent variable, and combines the pavement structure design, climate conditions, traffic volume, maintenance history, etc. of the road section as independent variables to construct a complete pavement disease data set for identifying key influencing factors of road performance.

[0086] Based on the constructed complete pavement disease data set, the Logistic algorithm is used to create a regression prediction model for pavement performance indicators and identify key influencing factors of road performance.

[0087] Step 2 of this embodiment specifically includes the following steps:

[0088] Step 21: Summarize the disease information of the identified road sections and other features that affect the pavement performance, construct a complete pavement disease data set for identifying key factors affecting road performance, clean and process the data, and divide it into training set, validation set and test set. Among them, other features that affect pavement performance include but are not limited to pavement structure design, climate conditions, traffic volume, maintenance history, etc.

[0089] Step 22: Create a regression prediction model for pavement damage condition PCI.

[0090] Step 22 of this embodiment specifically includes:

[0091] Step 221: Use the training set to create a regression prediction model for the pavement damage condition PCI. The independent variables of the model are the characteristics that affect the pavement properties, such as pavement structure design, climate conditions, traffic volume, maintenance history, etc. The dependent variable is the calculated pavement performance index, that is, the pavement damage condition index PCI.

[0092] A regression prediction model for PCI of road damage condition is created based on decision tree, random forest, XGBoost, and LightGBM. The error between the true value and the predicted value is compared, and the mean absolute error (MAE) is output as an indicator for evaluating model performance. MAE is the absolute error between the predicted value and the actual value, taking the average of all errors. MAE measures the prediction accuracy of the model. The smaller the MAE, the better the prediction effect of the model.

[0093] In the process of model construction, the data set is divided into training set, validation set and test set. The training set is used for model learning. The model adjusts internal parameters through iterative learning on the training set to minimize the prediction error. The validation set is used to evaluate the model performance. The test set is used to evaluate the performance of the trained model on new data to measure the generalization ability of the model. During the training process, each round of iteration will output a MAE value. As the model learns, the MAE usually decreases gradually. After the training is completed, a complete prediction will be made based on the test set, and the final MAE value will be calculated to evaluate the global performance of different models on the test set as an indicator for the final report.

[0094] MAE formula:

[0095] (8)

[0096] in, is the sample size, is the true value, Represents the predicted value.

[0097] Step 222: Using the MAE evaluation indicator, weights are calculated for the decision tree, random forest, XGBoost and LightGBM models based on the MACBETH method.

[0098] In step 222 of this embodiment, specifically, the following steps are included:

[0099] Step 222-1: Select the MAE indicator as the basis for constructing the judgment matrix, and use the Z-score standardization method to normalize the evaluation performance indicator.

[0100] Z-score normalization formula:

[0101] (9)

[0102] in, is the normalized MAE value; is the MAE value of the model; is the mean, that is, the mean of the MAE values ​​of the four models; is the standard deviation, that is, the standard deviation of the MAE values ​​of these four models.

[0103] Step 222-2: Calculate weights for decision tree, random forest, XGBoost and LightGBM models based on the MACBETH method.

[0104] The normalized performance evaluation standard MAE obtained in step 222-1 is defined as a scoring standard and will be used to compare the advantages and disadvantages of different models. Then, a judgment matrix is ​​constructed, and each element of the matrix represents the comparison result of the two models on the evaluation standard of mean absolute error MAE. The performance difference between model A and model B is judged according to the scoring standard, and the performance difference is evaluated using the classification scale of MACBETH (such as extreme, strong, moderate, weak, no). By solving the linear programming in the comparison matrix, the weight of each model, that is, the relative importance of the model, is obtained.

[0105] In this embodiment, the four selected models, namely decision tree, random forest, XGBoost and LightGBM models, are represented by A, B, C and D respectively, and the judgment matrix is ​​constructed in sequence, similar to constructing an upper triangular matrix, by calculating the six values ​​of A and B, A and C, A and D, B and C, B and D, and obtaining the classification scale according to these six values. Among them, if A and B, A and C, A and D are calculated, then A is used as the denominator of formula (11); B and C, B and D, B is used as the denominator of formula (11); C and D, C is used as the denominator of formula (11).

[0106] Specifically, the rules for obtaining the classification scale of MACBETH based on the mean absolute error MAE are as follows:

[0107] Assume that the MAEs of model A and model B are and , the differences between them are evaluated and mapped to the MACBETH classification scale by the following steps:

[0108] Calculate the MAE difference between the two models:

[0109] (10)

[0110] Determine the relative difference percentage:

[0111] (11)

[0112] Based on the relative difference percentage , combined with Table 1, we get the classification scale of MACBETH.

[0113] Table 1:

[0114]

[0115] Solving the linear programming problem will get the weight of each model. The solution formula is as follows:

[0116] (12)

[0117] in, and Respectively represent the model and Model The weight of ; Representation Model and Model The differences in the mean absolute error (MAE) evaluation criterion are quantified into numerical values ​​using classification scales (such as extreme, strong, moderate, weak, and no). .

[0118] The specific numerical quantification principles of the classification scale are shown in Table 2:

[0119] Table 2:

[0120]

[0121] Step 23: Using the weights of the decision tree, random forest, XGBoost and LightGBM models obtained in step 22, the feature importance of all models is integrated to calculate the feature importance values ​​and rankings of all road performance influencing factors, and the cumulative contribution rate method is used to identify the key factors affecting road performance.

[0122] The specific implementation steps of step 23 are:

[0123] Step 231: In step 22, MACBETH and the judgment matrix are used to obtain the weights of the decision tree, random forest, XGBoost and LightGBM models respectively.

[0124] Step 232: In the output of the above models, they output feature importance, which evaluates the contribution of each feature to the model's prediction. Decision trees and random forests focus on information gain and reduction of impurities, while XGBoost and LightGBM prefer to use gain, coverage, and frequency to evaluate feature importance. Through these feature importance indicators, you can understand which features contribute the most to the model's predictive ability.

[0125] Step 233: Using the model weights and the feature importance of each model, calculate the feature importance values ​​and rankings of all road behavior influencing factors.

[0126] (13)

[0127] in, is the first j The importance value of a feature, It is i The weight of the model, m is the number of models; It is The model The importance value of a feature.

[0128] Each model used generates a set of feature importances to measure the impact of each input feature on the model output. Since different models calculate feature importance in different ways and the output values ​​have different scales, the feature importances obtained by each model are standardized to obtain the feature importance values ​​of each model.

[0129] The standardized formula is as follows:

[0130] (14)

[0131] in, is the standardized feature importance value; is the feature importance value of the original output of the model; is the minimum value of feature importance in the model; is the maximum value of feature importance in the model.

[0132] Step 234: Using the cumulative contribution rate method, identify key factors affecting road performance.

[0133] Specifically, the cumulative contribution rate refers to the sum of the contribution rates of each factor during the factor extraction process. The cumulative contribution rate ratio method can help determine the number of factors that need to be retained to achieve the best effect on the explanation of the original variables.

[0134] Formula (13) is the sum of the importance of the jth feature obtained based on all models. According to formula (13), the sum of the feature importance of different features obtained based on all models can be calculated, and the feature importance values ​​of different features can be sorted. According to the sorting results, for the feature variables ranked in the top k, the cumulative importance scores are accumulated, and then the cumulative contribution ratio is calculated according to the cumulative importance scores according to formula (15). For example, the cumulative contribution ratio threshold is set to 80%. When the cumulative contribution ratio reaches the threshold (80%), the selection is stopped and these features are used as key influencing factors. For the feature variables ranked in the top k, the value of k is obtained according to the set threshold. If the top 8 features reach the set threshold of 80%, then k is the calculated value of 8.

[0135] The formula is as follows:

[0136] (15)

[0137] Among them, the sum of all feature importances is calculated as:

[0138] (16)

[0139] in, is the number of features.

[0140] Step 3: Create an ordered logistic model to analyze the impact of specific levels of key factors affecting road performance.

[0141] This embodiment performs a level analysis on the identified key influencing factors of road behavior, creates an OLM model to further quantify the specific impact of each level on the pavement performance index, and thus analyzes the specific impact of different levels of the key influencing factors of road behavior on the pavement performance.

[0142] Step 3 of this embodiment specifically includes:

[0143] Step 31: Summarize the road damage information of the road section identified in step 3 and other features that affect the pavement performance, construct a complete pavement damage data set for identifying key factors affecting road performance, clean and process the data, and divide it into training set, validation set and test set. Among them, other features that affect pavement performance include but are not limited to pavement structure design, climate conditions, traffic volume, maintenance history, etc.

[0144] The key influencing factors of road performance and the PCI of road damage condition in the complete pavement data set used for identification of key influencing factors of road performance in step 21 are selected as a new data set, that is, as a data set for analysis of key influencing factors of road performance, including ordered target dependent variables and independent variables. Among them, the target dependent variable is an ordered categorical variable, and the independent variable is the key influencing factor identified in step 2, including numerical variables and categorical variables. The target dependent variable is the PCI of road damage condition, and its ordered classification is five levels: "poor", "secondary", "medium", "good" and "excellent"; the independent variable is the key influencing factor of road performance identified in step 2.

[0145] Specifically, the independent variables are the key influencing factors identified, such as the pavement structure design, climate conditions, traffic volume, maintenance history, etc. of the road section. Numerical variables include the pavement thickness and base material strength in the pavement structure design of the road section, the annual average precipitation, annual average temperature, number of freeze-thaw cycles, etc. in the climate conditions, the annual average daily traffic volume in the traffic volume, and the maintenance frequency in the maintenance history; categorical variables include the pavement material type in the pavement structure design of the road section and the maintenance type in the maintenance history.

[0146] Step 32: Create an ordered logistic model to further quantify the specific impact of different levels of key influencing factors of road behavior on pavement performance indicators.

[0147] Step 32 of this embodiment specifically includes:

[0148] Step 321: Process missing values ​​and outliers in the data set for the key influencing factors analysis of road performance, encode all categorical variables, and encode the ordered dependent variables; standardize the numerical independent variables; and analyze the correlation between the variables to ensure that there is no high collinearity between them.

[0149] Step 322: Create an ordered logistic model to analyze in detail the specific impact of different levels of key influencing factors of road performance on road damage condition indicators.

[0150] Ordered Logistic Regression is a regression model used to process ordered dependent variables. The mathematical expression of the model is as follows:

[0151]

[0152] in, is an ordered dependent variable and can be An ordered category, the pavement condition index (PCI), has five levels: poor, poor, fair, good, and excellent. is the independent variable, representing the key influencing factor of road performance; Given the independent variable The dependent variable The value is The probability of class and below; is with category associated thresholds to distinguish between categories, i.e., ordered categories for classifying pavement damage conditions; is the regression coefficient vector, , represents the influence of independent variables on the pavement damage condition, is the number of key influencing factors of the independent variable road performance.

[0153] Step 33: Based on the results of the ordered logistic model, analyze the specific impact of different levels of key influencing factors of road behavior on pavement performance.

[0154] Step 33 of this embodiment specifically includes:

[0155] Step 331: In the ordered logistic model, each independent variable The coefficient of Indicates the impact of changes in the level of the independent variable on the dependent variable. Specifically, in the specific analysis of the key influencing factors of road performance, the coefficient Indicates that the independent variable The change in the logit of the dependent variable when the value of increases by one unit. Here, one unit, for a numeric independent variable, It indicates the change in the logarithmic probability (logit) of the dependent variable when the value of the numerical variable increases by 1; for categorical independent variables, Represents the change in the log odds (logit) of the dependent variable when switching from one category to another, such as from a low-level category to a high-level category. Interpreting the coefficients in the model output , represents the impact of the level change of key influencing factors of road performance on the pavement damage condition index, among which, It means that the increase in the level of the key influencing factors of the road behavior tends to push the pavement performance to a higher level; An increase in the level of the key influencing factor representing the road behavior tends to push the pavement performance to a lower level.

[0156] Step 332: Quantify the specific impact of different levels of a key influencing factor of road performance on the road damage condition index by using the marginal effect method, and describe the impact of a small change in the independent variable on the probability of each level of the dependent variable.

[0157] The calculation formula is as follows:

[0158]

[0159] in, Dependent variable Equal to The probability of a level, is the independent variable The coefficient of .

[0160] Step 333: Based on the interpretation of the model output coefficients in steps 331 and 332, make targeted maintenance decisions, consider the specific impact of key influencing factors of different road properties on the pavement damage condition, optimize the allocation of maintenance resources, extend the service life of the road and improve the overall pavement quality.

[0161] Taking the actual analysis as an example, the coefficient of heavy vehicles, one of the independent variable vehicle types, is a fraction, and the marginal effect shows that it increases the probability of low-grade pavement performance ("poor") by 20%. This shows that heavy vehicles significantly increase the risk of pavement performance deterioration. Therefore, managers can consider restricting the passage of heavy vehicles or increasing the maintenance frequency of these sections to improve the overall pavement performance.

[0162] This embodiment takes into account the lack of refined analysis of road damage size in existing road condition assessment methods, obtains a road damage dataset publicly available on the Internet, uses the YOLACT deep learning model to perform refined identification of pavement damage, and obtains pavement damage type and detailed size information through bounding boxes and segmentation masks to support subsequent pavement damage condition index prediction tasks.

[0163] This embodiment creates a regression prediction model of road damage condition PCI based on decision tree, random forest, XGBoost, and LightGBM, and uses the MACBETH method to obtain the model weight according to the mean absolute error (MAE) indicator for evaluating model performance, thereby identifying the key influencing factors of road performance.

[0164] This embodiment introduces an ordered logistic model. Taking into account the lack of in-depth analysis of key influencing factors of road properties in existing maintenance decisions, the specific impact of key influencing factors of road properties on road performance at different levels is further analyzed, thereby providing road management personnel with more scientific and reasonable maintenance decisions, which helps to extend the service life of roads and optimize resource allocation.

[0165] Embodiment 2

[0166] The purpose of this embodiment is to provide a road performance key influencing factor identification system, including:

[0167] The recognition module is used to obtain the bounding box and segmentation mask of road damage recognition based on the trained road recognition model for the acquired road remote sensing image;

[0168] A judgment module is used to judge whether the road damage is a regular damage based on the obtained bounding box and segmentation mask of the road damage identification. If it is a regular road damage, the damage area is calculated using the obtained bounding box; if it is an irregular road damage, the road damage area is calculated using the obtained segmentation mask;

[0169] A calculation module, used to calculate the road surface damage condition index according to the obtained road disease area;

[0170] An identification module is used to make predictions based on the pavement damage condition index and the factors affecting pavement performance using multiple regression prediction models, assign weights to the multiple regression prediction models based on the MACBETH method, and identify key factors affecting road performance;

[0171] The analysis module is used to analyze the influence of different levels of the identified key influencing factors of road performance on the road damage condition indicators by using the Logistic model.

[0172] In further embodiments, there is also provided:

[0173] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, no further description is given here.

[0174] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0175] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0176] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the method described in embodiment 1 is completed.

[0177] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0178] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in the first embodiment is implemented.

[0179] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0180] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.

[0181] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, etc. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0182] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0183] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for identifying key influencing factors of road performance, characterized in that: include: For the acquired road remote sensing images, based on the trained road recognition model, the bounding box and segmentation mask for road damage recognition are obtained; Based on the obtained bounding box and segmentation mask of road damage recognition, determine whether the road damage is a regular road damage. If it is a regular road damage, calculate the damage area using the obtained bounding box; if it is an irregular road damage, calculate the road damage area using the obtained segmentation mask; Calculate the road surface damage index based on the road disease area obtained; Based on the pavement damage indicators and combined with the factors affecting pavement performance, multiple regression prediction models are used for prediction. Based on the MACBETH method, weights are assigned to multiple regression prediction models to identify the key factors affecting road performance, which are: Normalize the evaluation performance indicators of multiple regression prediction models; The normalized evaluation performance index is used as the scoring standard to calculate the performance difference between different regression prediction models; The MACBETH method is used to divide the classification scales for the performance differences between different regression prediction models. The weight solution problem of each regression prediction model is combined with the divided classification scale and converted into a linear programming problem for solution to obtain the corresponding weight of each regression prediction model. According to the weights corresponding to each regression prediction model, the characteristic importance values ​​of all road performance influencing factors are calculated, and the key factors affecting road performance are identified using the cumulative contribution rate method; The Logistic model is used to analyze the impact of different levels of the identified key influencing factors of road performance on the pavement damage condition indicators.

2. A method for identifying key influencing factors of road performance as claimed in claim 1, characterized in that: The YOLACT model is used as the road recognition model. Based on whether the coordinates of the identified bounding box and the segmentation mask match, it is judged whether the road damage is a regular damage; if the coordinates of the obtained bounding box and the segmentation mask match, it is a regular road damage.

3. A method for identifying key influencing factors of road performance as claimed in claim 1, characterized in that: According to the weights corresponding to each regression prediction model, the characteristic importance values ​​of all road performance influencing factors are calculated, and the key factors affecting road performance are identified using the cumulative contribution rate method, specifically: According to the importance values ​​of different road behavior influencing factors obtained by each regression prediction model and the weight corresponding to each regression prediction model, the total importance value of each road behavior influencing factor and the importance values ​​of all road behavior influencing factors are calculated; The total importance values ​​of each calculated road performance influencing factor are sorted, and the importance values ​​of the road performance influencing factors are accumulated according to the sorting results. The cumulative contribution ratio is calculated by combining the importance values ​​of all road performance influencing factors until the calculated cumulative contribution ratio reaches the set threshold, and the key road performance influencing factors are determined.

4. A method for identifying key influencing factors of road performance as claimed in claim 1, characterized in that: The Logistic model is used to analyze the impact of different levels of the identified key factors of road performance on the pavement damage indicators, specifically: The key factors affecting road performance are used as independent variables, and the ordered classification is used as the dependent variable; wherein the ordered classification includes five levels: "poor", "inferior", "medium", "good", and "excellent"; Based on the determined independent variables and dependent variables, a logistic analysis model is constructed; In the constructed Logistic analysis model, the specific impact of different levels of key influencing factors of road behavior on pavement performance is analyzed.

5. A method for identifying key influencing factors of road performance as claimed in claim 4, characterized in that: The marginal effect method is used to quantify the specific impact of different levels of a key influencing factor of road performance on the pavement damage condition index, and to describe the impact of a small change in the independent variable on the probability of each level of the dependent variable.

6. A method for identifying key influencing factors of road performance as claimed in claim 1, characterized in that: The factors affecting pavement properties include pavement structure design, climate conditions, traffic volume and maintenance history; the regression prediction models are decision tree, random forest, XGBoost and LightGBM models.

7. A road performance key influencing factor identification system, characterized in that: include: The recognition module is used to obtain the bounding box and segmentation mask of road damage recognition based on the trained road recognition model for the acquired road remote sensing image; A judgment module is used to judge whether the road damage is a regular damage based on the obtained bounding box and segmentation mask of the road damage identification. If it is a regular road damage, the damage area is calculated using the obtained bounding box; if it is an irregular road damage, the road damage area is calculated using the obtained segmentation mask; A calculation module, used to calculate the road surface damage condition index according to the obtained road disease area; The identification module is used to make predictions based on the pavement damage condition index and the factors affecting pavement performance using multiple regression prediction models, assign weights to the multiple regression prediction models based on the MACBETH method, and identify the key factors affecting road performance, which are: Normalize the evaluation performance indicators of multiple regression prediction models; The normalized evaluation performance index is used as the scoring standard to calculate the performance difference between different regression prediction models; The MACBETH method is used to divide the classification scales for the performance differences between different regression prediction models. The weight solution problem of each regression prediction model is combined with the divided classification scale and converted into a linear programming problem for solution to obtain the corresponding weight of each regression prediction model. According to the weights corresponding to each regression prediction model, the characteristic importance values ​​of all road performance influencing factors are calculated, and the key factors affecting road performance are identified using the cumulative contribution rate method; The analysis module is used to analyze the influence of different levels of the identified key influencing factors of road performance on the road damage condition indicators by using the Logistic model.

8. A computer device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, a method for identifying key influencing factors of road performance as described in any one of claims 1 to 6 is performed.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for identifying key influencing factors of road performance as claimed in any one of claims 1 to 6 is executed.

Images