An asphalt paving segregation detection method, system and device based on sensitivity analysis

Abstract

The present application belongs to the technical field of asphalt paving segregation detection, and particularly relates to an asphalt paving segregation detection method, system and device based on sensitivity analysis, which comprises the following steps: S1, acquiring an image of an asphalt pavement; S2, extracting multiple features of the image, normalizing the features and splicing the normalized features to obtain one-dimensional features; S3, performing sensitivity grading and fusion processing on the one-dimensional features to obtain a fusion matrix; S4, dividing the fusion matrix of the image into a training set and a test set; S5, inputting the training set into a support vector machine to train a classifier; S6, inputting the test set into the trained classifier to obtain the paving segregation state of the asphalt pavement. The present application constructs fusion features as characteristic values for paving segregation detection through sensitivity analysis, thereby improving the detection accuracy and stability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of asphalt paving segregation detection, and particularly relates to an asphalt paving segregation detection method, system and device based on sensitivity analysis. BACKGROUND

[0002] For asphalt pavement, segregation is one of the important reasons for pavement damage. At present, the commonly used method for asphalt pavement detection mainly includes non-contact detection and contact detection.

[0003] The non-contact detection mainly includes visual observation method, construction depth method and core sampling method. The visual observation method lacks clear standards and has great subjectivity, and is suitable for asphalt aggregates with large particle size. The construction depth method mainly uses sand paving method, which can realize asphalt paving segregation detection, but has large test amount, consumes time and effort, and its precision is greatly affected by the operator, and thus is only suitable for asphalt paving segregation detection in a small range. The core sampling method extracts and screens the core sample to calculate the percentage of each component in the mixture to judge the paving segregation, and the detection precision of the method is the highest, but the method has irreversible damage to the pavement and consumes time and effort in the extraction process.

[0004] The contact detection mainly includes gray image method, laser measurement method and CT scanning method. The gray image method needs to vertically collect the pavement image to directly reflect the pavement concave-convex condition through the gray value, and thus is greatly affected by the light and the shooting angle. The laser measurement method directly obtains the pavement texture information through the laser range finder scanning, but has high requirements for the dryness and cleanliness of the pavement, and the instrument is expensive. The CT scanning method can obtain high-precision asphalt pavement aggregate distribution and texture three-dimensional information through the CT technology, but can only be detected indoors and needs to be core sampled first.

[0005] In summary, the current asphalt paving segregation detection mainly has two technical problems. First, the cost of asphalt paving segregation detection is high, the precision and stability of the detection method are poor, and the asphalt paving segregation state cannot be accurately obtained. Second, the judgment of asphalt paving segregation is based on expert knowledge, and needs the support of professional personnel, which is a huge consumption of manpower. Finally, the features of the existing image-based asphalt paving segregation detection cannot effectively support online real-time analysis, and it is also difficult to meet the demand of convenient on-site application. Therefore, it is of great significance to study how to improve the online and convenience of the asphalt paving segregation detection method for prolonging the road life and protecting the driving safety. SUMMARY

[0006] The technical problem to be solved by the application is to provide an asphalt paving segregation detection method, system and device based on sensitivity analysis, which can improve the detection precision to meet the requirements of asphalt segregation detection in the paving process.

[0007] The application is implemented as follows: the application provides an asphalt paving segregation detection method based on sensitivity analysis, comprising the following steps:

[0008] S1, acquiring an image of an asphalt pavement;

[0009] S2, extracting a plurality of features of the image, and splicing a one-dimensional feature after normalization processing;

[0010] S3, performing sensitivity classification and fusion processing on the one-dimensional feature to obtain a fusion matrix;

[0011] S4, dividing the fusion matrix of the image into a training set and a test set;

[0012] S5, inputting the training set into a support vector machine to train a classifier;

[0013] S6, inputting the test set into the trained classifier to obtain the paving segregation state of the asphalt pavement.

[0014] Further, the features include statistical features, edge features, and texture features, and the feature values are normalized to the interval [-1, 1] according to the value range.

[0015] Further, the S2 comprises the following sub-steps:

[0016] S201, for the upper and lower limits of the features with infinite value intervals, selecting a hyperbolic tangent function to map them to the interval [-1, 1];

[0017] S202, for the upper and lower limits of the features without infinite value intervals, performing normalization according to the following formula;

[0018] ,

[0019] X’ denotes the normalized value, X denotes the original value, X min denotes the minimum value in the data set, X max denotes the maximum value in the data set.

[0020] Further, the S3 comprises the following sub-steps:

[0021] S301, calculating the sensitivity effect of the one-dimensional feature through Morris sensitivity analysis;

[0022] S302, calculating the sensitivity of the one-dimensional feature according to the sensitivity effect;

[0023] S303. Select one-dimensional features whose sensitivity meets preset conditions to construct a model;

[0024] S304. Calculate the main effect coefficient of each of the one-dimensional features based on the Sobol sensitivity analysis and the model;

[0025] S305. Calculate the fusion weight corresponding to each of the one-dimensional features based on the main effect coefficients;

[0026] S306. The one-dimensional features are fused according to the fusion weights to obtain a fusion matrix.

[0027] Furthermore, the formula for calculating the sensitivity effect is as follows:

[0028] ;

[0029] In the above formula, Δ represents the disturbance increment of the i-th variable, and the function f(X) represents a variable with... k A model with independent inputs, i.e. They are defined in k 3D input space middle;

[0030] The formula for calculating the sensitivity is:

[0031] ;

[0032] In the above formula, Indicates a sensitive effect;

[0033] The expression for the model is:

[0034] ;

[0035] In the above formula, Indicates the model output, This indicates the parameters selected for Morris sensitivity analysis;

[0036] The formula for calculating the primary effect coefficient is as follows:

[0037] ;

[0038] In the above formula, Representing one-dimensional features The main effect, ; ; ;

[0039] The expression for the fusion weight is:

[0040] ;

[0041] The expression of the fusion matrix is:

[0042] ;

[0043] Wherein, i represents the number of features for feature fusion, is the fusion weight of the feature .

[0044] Further, S5 includes the following steps:

[0045] S501, the image is marked as two kinds of segregation asphalt pavement and normal asphalt pavement, and 0 and 1 represent segregation asphalt pavement and normal asphalt pavement respectively;

[0046] S502, select the training set input support vector machine for training, the kernel function selects the radial basis kernel function, the penalty coefficient is 30, and the Gamma kernel function is 0.45.

[0047] Based on the same inventive concept, an asphalt paving segregation detection system based on sensitivity analysis comprises:

[0048] An acquisition module is configured to acquire an image of an asphalt pavement.

[0049] A normalization module is configured to extract multiple features of the image, normalize the features, and splice the normalized features to obtain one-dimensional features.

[0050] A fusion module is configured to perform sensitivity grading and fusion processing on the one-dimensional features to obtain a fusion matrix.

[0051] A division module is configured to divide the fusion matrix of the image into a training set and a test set.

[0052] A training module is configured to input the training set into a support vector machine to train a classifier.

[0053] A test module is configured to input the test set into the trained classifier to obtain the paving segregation state of the asphalt pavement.

[0054] Based on the same inventive concept, the present application also provides an asphalt paving segregation detection device based on sensitivity analysis, comprising a processor and a memory, the memory is used for storing computer programs, and the processor is used for executing computer programs to realize the steps of the above method.

[0055] The present application has the following beneficial effects:

[0056] (1) The present application constructs a fusion feature as a feature value for paving segregation detection through sensitivity analysis, which improves the detection accuracy and stability.

[0057] (2) The present invention uses support vector machine (SVM) to detect paving segregation, which can be achieved without professional personnel and expertise.

[0058] (3) The present invention can realize online segregation detection during the asphalt paving process, which fundamentally reduces the probability of asphalt segregation events. Attached Figure Description

[0059] Figure 1 This is a flowchart of the method in this invention;

[0060] Figure 2 This is a flowchart of S3 in this invention;

[0061] Figure 3 The images show partial images of segregated asphalt pavement and normal asphalt pavement (where Figure a represents the image of paved segregated asphalt pavement, and Figure b represents the image of normal asphalt pavement).

[0062] Figure 4 This is a schematic diagram of the detection results of the present invention. Detailed Implementation

[0063] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0064] Example 1: As Figures 1 to 4 As shown, a method for detecting asphalt segregation in paving based on sensitivity analysis includes the following steps:

[0065] S1. Obtain an image of the asphalt pavement.

[0066] S2. Extract various features from the image, normalize them, and then concatenate them to obtain one-dimensional features. In this step, features that may be related to the separation are extracted from the image, including statistical features, edge features, and texture features. The feature values ​​are normalized to the interval [-1, 1] according to their value range. Table 1 shows the names of the extracted features and their corresponding value ranges.

[0067] Table 1 Image Feature Table

[0068]

[0069] In this step, some features have infinite value ranges, while others do not. Therefore, normalization can be performed separately according to the characteristics of the value ranges. Specifically, S2 includes the following sub-steps:

[0070] S201. For the upper and lower limits of the feature, there is an infinite range of values. We choose to use the hyperbolic tangent function to map it to the interval [-1, 1].

[0071] S202, for the upper and lower limits of the feature does not exist infinite value interval, according to the following formula is normalized: ;

[0072] X’ denotes the normalized value, X denotes the original value, X min denotes the minimum value in the data set, X max denotes the maximum value in the data set.

[0073] S3, the sensitivity of one-dimensional features is graded and fused to obtain a fusion matrix. This step is based on the Morris sensitivity analysis of the sensitivity of the features, and the features are fused. As shown in Figure 3 , specifically, S3 includes the following sub-steps:

[0074] S301, the sensitive effect of one-dimensional features is calculated by Morris sensitivity analysis. In this step, the function f(X) represents a model with k independent inputs, that is , which is defined in k dimensional input space . Let be the value assigned to X in . If the given value only changes , while the values of all other inputs remain unchanged, the corresponding model output is , where . Therefore, the formula for calculating the sensitive effect of feature X i is:

[0075] ;

[0076] In the above formula, Δ represents the perturbation increment of the i th variable.

[0077] S302, the sensitivity of one-dimensional features is calculated according to the sensitive effect. This step calculates the sensitive effect r times for each feature to study the sensitivity of X i , and the calculation formula of the sensitivity is:

[0078] ;

[0079] In the above formula, denotes the sensitive effect. The following formula is used to measure the dispersion degree of data:

[0080] ;

[0081] wherein, σ i denotes the standard deviation, μ i denotes the sensitivity of the i-th data, i N denotes the sample number.

[0082] Select multiple asphalt images for calculation, and the specific calculation results are shown in Table 2:

[0083] Table 2 Sensitivity of different features

[0084]

[0085] S303, select a one-dimensional feature with sensitivity meeting a preset condition to construct a model. In this step, a feature with high sensitivity is selected to construct a model, that is, a feature with a high sensitivity value. According to Table 2, the mean value, Sobol operator and LBP_GLCM homogeneity feature can be selected as the subsequent fusion feature to construct a model.

[0086] S304, calculate the main effect coefficient of each one-dimensional feature according to the Sobol sensitivity analysis and the model. The expression of the model is:

[0087] .

[0088] In the above formula, denotes the model output, denotes the parameter selected by the Morris sensitivity analysis.

[0089] S305, calculate the fusion weight corresponding to each one-dimensional feature according to the main effect coefficient. The calculation formula of the main effect coefficient is:

[0090] ;

[0091] In the above formula, denotes the main effect of the one-dimensional feature ; ; ; .

[0092] The expression of the fusion weight is:

[0093] .

[0094] The main effect coefficient and the corresponding weight of the three selected features are shown in Table 3:

[0095] Table 3 Main effect coefficient and fusion weight of different features

[0096] ​

[0097] S306、According to the fusion weight, the one-dimensional features are fused to obtain a fusion matrix. The expression of the fusion matrix is:

[0098] .

[0099] wherein i represents the number of features for feature fusion, is the fusion weight of the feature .

[0100] S4, the fusion matrix of the image is divided into a training set and a test set.

[0101] S5, the training set is input into a support vector machine to train a classifier. S5 includes the following steps:

[0102] S501, the image is labeled as two types of segregated asphalt pavement and normal asphalt pavement, and 0 and 1 are used to represent the segregated asphalt pavement and the normal asphalt pavement, respectively;

[0103] S502, the training set is selected to be input into a support vector machine for training, a radial basis (RBF) kernel function is selected as the kernel function, the penalty coefficient is 30, and the Gamma kernel function is 0.45.

[0104] S6, the fusion feature matrix of the asphalt image to be detected is extracted and input into the trained classifier to obtain the paving segregation state of the asphalt pavement. As shown in Figure 3 , the present application selects a total of 100 images with segregation and 100 images without segregation for classification detection, and the detection result is as shown in Figure 4 . Among the 200 asphalt images, 197 images are correctly classified, and the accuracy of the detection result can reach 98.5%. Among them, label 1 represents that the asphalt pavement in the image has segregation, and label 0 represents that the image is a normal asphalt pavement.

[0105] Embodiment 2: Based on the same inventive concept, an asphalt paving segregation detection system based on sensitivity analysis includes an acquisition module, a normalization module, a fusion module, a training module, and a test module.

[0106] The acquisition module is used to acquire the image of the asphalt pavement.

[0107] The normalization module is used to extract various features of the image, and the one-dimensional features are obtained by splicing after normalization.

[0108] The fusion module is used for sensitivity grading and fusion processing of the one-dimensional features to obtain a fusion matrix.

[0109] The division module is used to divide the fusion matrix of the image into a training set and a test set.

[0110] The training module is configured to input the training set into the support vector machine to train the classifier.

[0111] The testing module is configured to input the testing set into the trained classifier to obtain the paving segregation state of the asphalt pavement.

[0112] In the same inventive concept, the application also provides an asphalt paving segregation detection device based on sensitivity analysis, comprising a processor and a memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program to realize the steps of the method in the embodiment one.

[0113] The above is only the preferred embodiment of the application, and it should be pointed out that for ordinary skilled in the art, without departing from the technical principles of the application, a number of improvements and modifications can be made, and these improvements and modifications should be considered as the protection scope of the application.

Claims

1. A method for asphalt paving segregation detection based on sensitivity analysis, characterized in that, The method comprises the following steps: S1, acquiring an image of an asphalt pavement; S2, extracting a plurality of features of the image, and splicing one-dimensional features after normalization processing; S3, performing sensitivity classification and fusion processing on the one-dimensional features to obtain a fusion matrix; the S3 comprises the following sub-steps: S301, calculating a sensitive effect of the one-dimensional features by Morris sensitivity analysis; the calculation formula of the sensitive effect is: ； In the above equation, Δ represents the perturbation increment of the ith variable, and the function f(X) represents a model with k independent inputs, i.e. , which are defined in k -dimensional input space . S302, calculating a sensitivity of the one-dimensional features according to the sensitive effect; the calculation formula of the sensitivity is: ； In the above formulae, represents a sensitive effect; S303, selecting one-dimensional features with a sensitivity meeting a preset condition to construct a model; the expression of the model is: ； In the above formula, denotes the model output, denotes the parameters selected for the Morris sensitivity analysis; S304, calculating a main effect coefficient of each one-dimensional feature according to Sobol sensitivity analysis and the model; the calculation formula of the main effect coefficient is: ； In the above formulae, represents the main effect of the one-dimensional feature , ; ; ; S305, calculating a fusion weight corresponding to each one-dimensional feature according to the main effect coefficient; the expression of the fusion weight is: ； S306, fusing the one-dimensional features according to the fusion weight to obtain a fusion matrix; the expression of the fusion matrix is: ； wherein i represents the number of features used for feature fusion, is a fusion weight of the feature . S4, dividing the fusion matrix of the image into a training set and a test set; S5, inputting the training set into a support vector machine to train a classifier; S6, inputting the test set into the trained classifier to obtain a paving segregation state of the asphalt pavement.

2. A method for asphalt paving segregation detection based on sensitivity analysis according to claim 1, characterized in that, The features include statistical features, edge features and texture features, and feature values are normalized to the interval [-1, 1] according to a value range.

3. A method for asphalt paving segregation detection based on sensitivity analysis according to claim 2, characterized in that, The S2 comprises the following sub-steps: S201, for the feature with an infinite value range of upper and lower limits, selecting a hyperbolic tangent function to map it to the interval [-1, 1]; S202, for the feature with no infinite value range of upper and lower limits, performing normalization according to the following formula; ， X’ represents the normalized value, X represents the original value, X min represents the minimum value in the data set, X max represents the maximum value in the data set.

4. The asphalt paving segregation detection method based on sensitivity analysis according to claim 1, characterized in that, S5 comprises the following steps: S501, marking the image as two types of segregated asphalt pavement and normal asphalt pavement, and representing the segregated asphalt pavement and the normal asphalt pavement by 0 and 1 respectively; S502, selecting the training set to input into the support vector machine for training, selecting a radial basis kernel function as a kernel function, and setting a penalty coefficient to 30 and a Gamma kernel function to 0.

45.

5. A sensitivity analysis based asphalt paving segregation detection system characterized by, It comprises: An acquisition module, configured to acquire an image of an asphalt pavement; A normalization module, configured to extract a plurality of features of the image, and splice one-dimensional features after normalization processing; A fusion module, configured to perform sensitivity classification and fusion processing on the one-dimensional features to obtain a fusion matrix; the module performs the following sub-steps: calculating a sensitive effect of the one-dimensional features by Morris sensitivity analysis; the calculation formula of the sensitive effect is: ； In the above equation, Δ represents the perturbation increment of the i-th variable, and the function f(X) represents a model with k independent inputs, i.e. , which are defined in k a d-dimensional input space . calculating a sensitivity of the one-dimensional features according to the sensitive effect; the calculation formula of the sensitivity is: ； In the above formulae, represents a sensitive effect; selecting one-dimensional features with a sensitivity meeting a preset condition to construct a model; the expression of the model is: ； In the above formula, denotes the model output, denotes the parameters selected for the Morris sensitivity analysis; calculating a main effect coefficient of each one-dimensional feature according to Sobol sensitivity analysis and the model; the calculation formula of the main effect coefficient is: ； In the above formulae, represents the main effect of the one-dimensional feature ; ; ; ; calculating a fusion weight corresponding to each one-dimensional feature according to the main effect coefficient; the expression of the fusion weight is: ； Fusing the one-dimensional features according to the fusion weights to obtain a fusion matrix; an expression of the fusion matrix is: ； wherein i represents the number of features used for feature fusion, is a fusion weight of the feature . A division module, configured to divide the fusion matrix of the image into a training set and a test set; A training module, configured to input the training set into a support vector machine to train a classifier; A test module, configured to input the test set into the trained classifier to obtain the paving segregation state of the asphalt pavement.

6. A sensitivity analysis based asphalt paving segregation detection device, characterized by, A computer program product, comprising a processor and a memory, the memory being configured to store a computer program, and the processor being configured to execute the computer program to implement the steps of the method of claim 1.

Images