Crack evaluation method based on three-dimensional ground penetrating radar and deflection value

Abstract

The invention discloses a crack evaluation method based on a three-dimensional ground penetrating radar and a deflection value. The crack evaluation method comprises the following steps: S1, obtaining a processed pavement crack image and a corresponding deflection value data set; s2, establishing a pre-trained crack classification model to obtain a classified pavement crack image; s3, converting the classified pavement crack image into a binary image, obtaining the average crack energy of the pavement crack image, and carrying out dynamic weight fusion on the average crack energy and deflection value data to obtain comprehensive crack energy; and S4, determining a plurality of static parameters influencing the energy distribution of the pavement crack image, and respectively analyzing the influence degree of the plurality of static parameters on the comprehensive crack energy. According to the method, the relationship between each static parameter and the crack development is established to determine the influence degree of different static parameters on the crack, the static parameters with relatively large influence are determined, and targeted guidance is provided for subsequent maintenance according to the static parameters with relatively large influence, so that the maintenance efficiency is improved, and the cost is saved.

Description

Technical Field

[0001] This invention relates to the field of crack analysis technology, and in particular to a crack evaluation method based on three-dimensional ground penetrating radar and deflection values. Background Technology

[0002] During road use, road surfaces often develop problems such as cracks, potholes, and ruts due to factors such as vehicle loads, environmental influences, and material aging. Linear cracks, in particular, not only affect the lifespan of the road surface and driving comfort but can also lead to serious structural damage.

[0003] Current technologies primarily rely on image processing to identify cracks and conduct road maintenance based on crack length and width. However, these two parameters only reflect geometric characteristics and do not fully reveal the risk of crack propagation. Furthermore, crack evaluation methods often depend on single visual data and lack quantitative analysis of pavement structural strength. The correlation between deflection value, a key indicator of pavement bearing capacity, and crack development has not been effectively explored. Simultaneously, while static parameters such as surface layer gradation, thickness, and materials are considered in road design and maintenance, the specific relationship between these data and crack development is often overlooked. Analysis of crack development capacity and the impact of environmental factors on crack energy are not addressed, thus failing to accurately determine the causes of crack formation to provide effective reference for later road operation and maintenance, reducing the efficiency of the maintenance process. Summary of the Invention

[0004] This invention provides a crack evaluation method based on three-dimensional ground-penetrating radar and deflection values. This method overcomes the technical problems of existing crack evaluation methods that rely heavily on single visual data, lack quantitative analysis of pavement structural strength, ignore the specific relationship between static parameters and crack development, do not involve crack development capacity analysis and the influence of environmental factors on crack energy, and cannot accurately obtain the causes of crack formation to provide effective reference for the later operation and maintenance of roads, thus reducing the efficiency of the maintenance process.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A crack evaluation method based on three-dimensional ground-penetrating radar and deflection values, the specific steps of which include:

[0007] S1: Obtain road surface crack images using three-dimensional ground-penetrating radar and simultaneously collect deflection value data at the crack locations using a deflectometer; preprocess the road surface crack images and deflection value data to obtain processed road surface crack images and corresponding deflection value datasets.

[0008] S2: Establish a pre-trained crack classification model, and use the pre-trained crack classification model to classify the processed road surface crack image to obtain the classified road surface crack image.

[0009] S3: Convert the classified pavement crack image into a binary image, and obtain the average crack energy of the pavement crack image based on the binary image. Dynamically weight and fuse the average crack energy with the deflection value data to obtain the comprehensive crack energy.

[0010] S4: Determine several static parameters that affect the energy distribution of pavement crack images, analyze the degree of influence of several static parameters on the comprehensive crack energy, and judge the risk level of the crack area based on the collected deflection data and the preset deflection tolerance value, so as to achieve the evaluation of different types of cracks.

[0011] The static parameters include surface layer gradation, pavement layer thickness, humidity, traffic volume, base course materials, diurnal temperature range, subgrade soil materials, service life, and precipitation.

[0012] Furthermore, in S3, the specific steps for converting the classified road surface crack image into a binary image include:

[0013] The classified road surface crack map is converted into a grayscale image;

[0014] The operation of distinguishing linear crack regions from background regions in the grayscale image includes:

[0015] Count the pixel values ​​of all grayscale images, determine the distribution range of pixel values, and denote the minimum value as d. min The maximum value is denoted as d. max The pixel distribution range is represented as [d] min d max ];

[0016] Let the grayscale threshold be d. t With grayscale threshold d t Using the initial critical point, the pixel distribution range is divided into interval C [d min d t ] and interval D (d t d max ];

[0017] Wherein, the grayscale threshold d t The range of values ​​for is (d min d max -1), and the pixel increment step is 1 pixel;

[0018] The number of pixels in intervals C and D are counted as n1 and n2 respectively, and the weights of the number of pixels in intervals C and D in the entire image are calculated as e1 and e2 respectively. 1＝ n1 / (n1+n2), e 2＝ n2 / (n1+n2), the average pixel value in intervals C and D is dc1 and d d2 The inter-class variance of image pixels in intervals C and D is calculated as E = e1·e2· / (d c1 -d d2 ) 2 ;

[0019] Iterate through the grayscale value thresholds to obtain the grayscale value threshold d that maximizes the E value. tm The grayscale threshold d that maximizes the E value. tm As the critical pixel value between the linear crack region and the background region;

[0020] With the final grayscale value threshold d tm As the final dividing point, gray values ​​less than d are... tm The pixel value of the pixel is changed to 0, and the gray value is greater than or equal to d. tm The pixel value of each pixel is changed to 1, thus obtaining a binarized image.

[0021] Furthermore, the specific steps for obtaining the average energy of the road surface crack image based on the binarized image include:

[0022] The binarized image is transformed into a coordinate system: a Cartesian coordinate system is established with the lower left corner of the binarized image as the origin, the horizontal rightward direction is the positive x-axis, and the vertical upward direction is the positive y-axis, forming a binarized image matrix, represented as:

[0023]

[0024] In the formula, j is the imaginary unit, e = 0, 1, 2, ..., J-1; f = 0, 1, 2, ..., K-1; J is the number of pixels in the image matrix along the x-axis, and K is the number of pixels in the image matrix along the y-axis; J S K is the angular frequency in the x-axis direction. S Angular frequency in the y-axis direction;

[0025] The binarized image matrix is ​​split into several combinations of sine and cosine functions with known frequencies and amplitudes;

[0026] The energy distribution of the road surface crack image is obtained based on the combination of the split sine and cosine functions. The energy distribution of the road surface crack image in one period signal is calculated using the following formula:

[0027]

[0028] In the formula, L is the image signal period in the x-axis direction, and M is the image signal period in the y-axis direction; P i H represents the i-th frequency; i For P iThe corresponding energy values, and the energy distribution of the road surface crack image, are the energy values ​​corresponding to each frequency.

[0029] Calculate the average crack energy H based on the energy distribution of road crack images. a The calculation formula is:

[0030]

[0031] In the formula, N represents the number of road surface crack images.

[0032] Furthermore, the specific steps for dynamically weighting and fusing the average crack energy with the deflection value data to obtain the comprehensive crack energy include:

[0033] Kriging interpolation is performed on the deflection data to generate a continuous deflection field surface, as shown in the following formula:

[0034]

[0035] Where, d i Let λ be the deflection value at the measuring point. i These are the weighting coefficients, and they satisfy the unbiasedness property. and the minimum variance condition;

[0036] The formula for calculating the overall fracture energy is as follows:

[0037]

[0038] Where D(x,y) is the deflection value at the current position, D crit Let α be the standard value of pavement deflection, and α be the coupling coefficient determined by regression of historical data.

[0039] Furthermore, in S4, the specific steps for determining several static parameters that affect the energy distribution of the pavement crack image and analyzing the influence of these static parameters on the overall crack energy include:

[0040] The static parameters are standardized using the following formula:

[0041]

[0042] Among them, X i (k) represents the k-th sample value of the i-th parameter; X i '(k) represents the standardized value;

[0043] The correlation between surface layer gradation, pavement layer thickness, humidity, traffic volume, base course material, diurnal temperature range, subgrade soil material, service life, precipitation, and comprehensive crack energy was calculated to obtain the influence of different static parameters on comprehensive crack energy, including:

[0044] Calculate each static parameter and the overall fracture energy H c The correlation coefficient at sample k is calculated using the following formula:

[0045]

[0046] Among them, H' c (k) and N' i (k) represents the standardized value; ρ is the resolution coefficient;

[0047] The correlation between each static coefficient and the overall fracture energy is calculated based on the correlation coefficient. The calculation formula is as follows:

[0048]

[0049] Furthermore, in S1, the specific steps for preprocessing the road surface crack image include:

[0050] Determine the location of the survey lines on the road surface to be inspected and mark the station numbers;

[0051] The antenna detection parameters of the 3D ground-penetrating radar are set according to the condition of the road surface to be detected, so as to obtain images of road surface cracks.

[0052] The three-dimensional ground-penetrating radar data processing software rS1icer was used to sequentially perform time zero-point, interference suppression, inverse Fourier transform, background removal, bandpass filtering, gain compensation and offset imaging processing on the road surface crack image to highlight the location of the abnormal body in the road surface crack image.

[0053] The location of the crack and the corresponding station number are determined from the anomaly.

[0054] Furthermore, in S2, the specific steps for establishing a pre-trained crack classification model include:

[0055] Based on the processed pavement crack image, horizontal slices of the surface at the crack, horizontal slices of the depth of the surface layer and the base layer, horizontal slices of the depth of the base layer and the base layer, and horizontal slices of the depth of the base layer and the subbase layer are obtained.

[0056] Based on the slicing of the acquired pavement crack images and the crack determination method, the pavement crack images are divided into the following types: top-bottom development without cracks in the middle, top-bottom development with cracks in the middle and partially penetrating, completely penetrating, and top-to-bottom development.

[0057] An original crack classification model is established, and the original crack classification model is trained based on the classified pavement crack images to obtain a pre-trained crack classification model.

[0058] Beneficial Effects: This invention classifies pavement crack images and dynamically weights and fuses average crack energy with deflection data, combining structural strength degradation with crack propagation, and integrating structural performance with visual features to obtain the comprehensive crack energy of the classified pavement crack images. By identifying several static parameters affecting the energy distribution of pavement crack images and analyzing the degree of influence of each of these static parameters on the comprehensive crack energy, crack evaluation can be achieved. This invention establishes the relationship between each static parameter and crack development to determine the degree of influence of different static parameters on cracks, identifies the static parameters with the greatest influence, and provides targeted guidance for subsequent maintenance based on these influential static parameters, thereby improving maintenance efficiency and saving costs. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart of a crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​in this invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] This embodiment provides a crack evaluation method based on three-dimensional ground-penetrating radar and deflection values, such as... Figure 1 As shown, the specific steps include:

[0063] S1: Obtain road surface crack images using three-dimensional ground-penetrating radar and simultaneously collect deflection value data at the crack locations using a deflectometer; preprocess the road surface crack images and deflection value data to obtain processed road surface crack images and corresponding deflection value datasets.

[0064] Specifically, in this embodiment, the deflectometer is a falling weight deflectometer (FWD), and the dynamic deflection value of the road surface area is collected at 0.2m intervals using the falling weight deflectometer.

[0065] In a specific embodiment, the preprocessing steps for the road surface crack image in S1 include:

[0066] Determine the location of the survey lines on the road surface to be inspected and mark the station numbers;

[0067] Based on the condition of the road surface to be inspected, the antenna detection parameters of the three-dimensional ground penetrating radar are set, non-destructive testing is performed, and the actual station number is recorded to obtain the road surface crack image;

[0068] Using the 3D ground-penetrating radar data processing software rS1icer, the road surface crack image is sequentially processed with time zeroing, interference suppression, inverse Fourier transform, background removal, bandpass filtering, gain compensation, and offset imaging to remove noise, clutter, and the influence of multiple waves in the road surface crack image, thereby suppressing strong signals and enhancing weak signals to highlight the location of abnormal bodies in the road surface crack image.

[0069] Determine the location of the crack and the corresponding station number from the anomaly.

[0070] Specifically, the deflection data is preprocessed, including: measuring each measuring point three times and taking the average value; and aligning the deflection value with the processed pavement crack image in spatial coordinates by matching the station number.

[0071] S2: Establish a pre-trained crack classification model, and use the pre-trained crack classification model to classify the processed road surface crack image to obtain the classified road surface crack image.

[0072] In a specific embodiment, the steps for establishing a pre-trained crack classification model include:

[0073] Based on the processed pavement crack image, horizontal slices of the surface at the crack, horizontal slices of the depth of the surface layer and the base layer, horizontal slices of the depth of the base layer and the base layer, and horizontal slices of the depth of the base layer and the subbase layer are obtained.

[0074] Specifically, in this embodiment, the selected slices include: a horizontal slice of the surface where the crack is located, a horizontal slice at the junction of the surface layer and the base layer (approximately 18 cm deep), a horizontal slice inside the base layer (approximately 36 cm deep), and a horizontal slice at the junction of the base layer and the subbase layer (approximately 54 cm deep) (the structural layer interfaces of conventional highways are located 1 cm below the surface, at the junction of the surface layer and the base layer, in the middle of the base layer, and at the junction of the base layer and the subbase layer, respectively).

[0075] Based on the slicing of the acquired pavement crack images and the crack determination method, the pavement crack images are divided into the following types: top-bottom development without cracks in the middle, top-bottom development with cracks in the middle and partially penetrating, completely penetrating, and top-to-bottom development.

[0076] An original crack classification model is established, and the original crack classification model is trained based on the classified pavement crack images to obtain a pre-trained crack classification model.

[0077] Specifically, in this embodiment, the original crack classification model is built based on the YOLOv8 model. YOLOv8, through innovations such as architecture optimization, anchorless detection, and dynamic label allocation, achieves a balance between high accuracy and real-time performance in crack classification tasks. Its ease of use and multi-platform support further lower the deployment threshold, making YOLOv8 widely used for intelligent crack identification in infrastructure maintenance. This embodiment uses a road crack dataset to train the original crack classification model, determining all parameters in the original model, including the number of network layers, the number of neurons in each layer, learning rate, weights, biases, activation functions, loss functions, and convolutional kernels. Then, the pre-trained convolutional neural network model is used for intelligent image recognition, classifying crack images for subsequent analysis.

[0078] In a specific embodiment, the crack determination method includes:

[0079] The method for determining the crack type of cracks that develop from top to bottom without cracks in the middle includes: on the entire cross section of the road, if the upper layer of the asphalt surface layer has cracks while the middle and lower layers have no cracks, and the semi-rigid base layer shows a through crack pattern, then the crack is determined to be a crack type of cracks that develop from top to bottom without cracks in the middle.

[0080] The method for determining the type of crack that develops vertically and horizontally with intermediate cracks and partially penetrates it includes: on the cross section of the road surface, when the crack is basically completely penetrated, but there is only a local phenomenon that the middle part is not cracked, the crack is judged to be a type of crack that develops vertically and horizontally with intermediate cracks and partially penetrates it.

[0081] The method for determining a fully penetrating crack includes: if the entire crack develops downwards from the top surface and extends downwards, then the crack is determined to be a fully penetrating crack.

[0082] The method for identifying top-down cracks includes: if, on a cross-section, the crack only appears on the surface layer while other parts remain intact, then the crack is identified as a top-down crack.

[0083] S3: Convert the classified road crack image into a binarized image, and obtain the average crack energy of the road crack image based on the binarized image;

[0084] In a specific embodiment, converting the classified pavement crack image into a binary image includes:

[0085] The classified road surface crack map is converted into a grayscale image;

[0086] The operation of distinguishing linear crack regions from background regions in the grayscale image includes:

[0087] Count the pixel values ​​of all grayscale images, determine the distribution range of pixel values, and denote the minimum value as d. min The maximum value is denoted as d. max The pixel distribution range is represented as [d] min d max ];

[0088] Let the grayscale threshold be d. t With grayscale threshold d t Using the initial critical point, the pixel distribution range is divided into interval C [d min d t ] and interval D (d t d max ];

[0089] Wherein, the grayscale threshold d t The range of values ​​for is (d min d max -1), and the pixel increment step is 1 pixel;

[0090] The number of pixels in intervals C and D are counted as n1 and n2 respectively, and the weights of the number of pixels in intervals C and D in the whole image are calculated as e1 and e2 respectively, where e1 = n1 / (n1+n2) and e2 = n2 / (n1+n2). The average pixel value in intervals C and D is d. c1 and d d2 The inter-class variance of image pixels in intervals C and D is calculated as E = e1·e2· / (d c1 -d d2 ) 2 ;

[0091] Iterate through the grayscale value thresholds to obtain the grayscale value threshold d that maximizes the E value. tm The grayscale threshold d that maximizes the E value. tm As the critical pixel value between the linear crack region and the background region;

[0092] With the final grayscale value threshold d tm As the final dividing point, gray values ​​less than d are... tm The pixel value of the pixel is changed to 0, and the gray value is greater than or equal to d. tm The pixel value of each pixel is changed to 1, thus obtaining a binarized image.

[0093] In a specific embodiment, the specific steps for obtaining the average crack energy of the road surface crack image based on the binarized image include:

[0094] The binarized image is transformed into a coordinate system: a Cartesian coordinate system is established with the lower left corner of the binarized image as the origin, the horizontal rightward direction is the positive x-axis, and the vertical upward direction is the positive y-axis, forming a binarized image matrix, represented as:

[0095]

[0096] In the formula, j is the imaginary unit, e = 0, 1, 2, ..., J-1; f = 0, 1, 2, ..., K-1; J is the number of pixels in the image matrix along the x-axis, and K is the number of pixels in the image matrix along the y-axis; J S K is the angular frequency in the x-axis direction. S Angular frequency in the y-axis direction;

[0097] The binarized image matrix is ​​split into several combinations of sine and cosine functions with known frequencies and amplitudes;

[0098] Specifically, the method of splitting the binarized image matrix into several sine and cosine functions with known frequencies and amplitudes in this embodiment is an existing technique in the field of image processing and will not be elaborated here.

[0099] The energy distribution of the road surface crack image is obtained based on the combination of the split sine and cosine functions. The energy distribution of the road surface crack image in one period signal is calculated using the following formula:

[0100]

[0101] In the formula, L is the image signal period in the x-axis direction, and M is the image signal period in the y-axis direction; P i H represents the i-th frequency; i For P i The corresponding energy values, and the energy distribution of the road surface crack image, are the energy values ​​corresponding to each frequency.

[0102] Calculate the average crack energy H based on the energy distribution of road crack images. a The calculation formula is:

[0103]

[0104] In the formula, N represents the number of road surface crack images.

[0105] In a specific embodiment, the steps for dynamically weighting and fusing the average crack energy with the deflection value data to obtain the comprehensive crack energy include:

[0106] Kriging interpolation is performed on the deflection data to generate a continuous deflection field surface, as shown in the following formula:

[0107]

[0108] Where, d i Let λ be the deflection value at the measuring point. i These are the weighting coefficients, and they satisfy the unbiasedness property. and the minimum variance condition;

[0109] The formula for calculating the overall fracture energy is as follows:

[0110]

[0111] Where D(x,y) is the deflection value at the current position, D crit Let α be the standard value of pavement deflection, and α be the coupling coefficient determined by regression of historical data.

[0112] S4: Determine several static parameters that affect the energy distribution of pavement crack images, analyze the degree of influence of several static parameters on the comprehensive crack energy, and judge the risk level of the crack area based on the collected deflection data and the preset deflection tolerance value, so as to achieve the evaluation of different types of cracks.

[0113] The static parameters include surface layer gradation, pavement layer thickness, humidity, traffic volume, base course materials, diurnal temperature range, subgrade soil materials, service life, and precipitation.

[0114] Specifically, this embodiment calculates the comprehensive crack energy by introducing deflection value parameters. The comprehensive crack energy reflects the energy level of cracks in the pavement crack image, and finally obtains evaluation results for different types of cracks. Subsequent work is carried out based on the evaluation results. For example, when the comprehensive crack energy value is high, it indicates that the crack is more serious and the possibility of further crack expansion is also greater. If the evaluation shows that the static parameter pavement layer thickness has the greatest impact on the comprehensive crack energy, it indicates that the cause of road cracks is greatly affected by the pavement layer thickness. Therefore, special attention should be paid to the maintenance of pavement layer thickness during maintenance. In addition, in road maintenance decision-making, the energy information of pavement cracks can be combined to optimize maintenance strategies and reasonably arrange maintenance costs, thereby improving the quality and efficiency of maintenance.

[0115] Specifically, this embodiment classifies cracks and then performs correlation analysis to understand the factors contributing to different crack types, providing a reference for subsequent maintenance and construction. Furthermore, based on the number of different types of cracks on the road surface, targeted maintenance and construction can be carried out. For example, if many crack types on the road surface are minor and cause little damage, they can be repaired. If there are many structural cracks and road resurfacing is planned, then the design and construction can focus on factors that significantly impact the road surface in this area and take corresponding measures to avoid them.

[0116] In a specific embodiment, S4, the specific steps of determining several static parameters that affect the energy distribution of the pavement crack image and analyzing the degree of influence of these static parameters on the overall crack energy include:

[0117] The static parameters are standardized using the following formula:

[0118]

[0119] Among them, X i (k) represents the k-th sample value of the i-th parameter; X i '(k) represents the standardized value;

[0120] The correlation between surface layer gradation, pavement layer thickness, humidity, traffic volume, base course material, diurnal temperature range, subgrade soil material, service life, precipitation, and comprehensive crack energy was calculated to obtain the influence of different static parameters on comprehensive crack energy, including:

[0121] Calculate each static parameter and the overall fracture energy H c The correlation coefficient at sample k is calculated using the following formula:

[0122]

[0123] Among them, H' a (k) and N' i (k) is the standardized value; ρ is the resolution coefficient, used to adjust for differences in correlation coefficients, usually taken as 0.5;

[0124] The correlation between each static coefficient and the overall fracture energy is calculated based on the correlation coefficient. The calculation formula is as follows:

[0125]

[0126] Specifically, this embodiment can take into account the impact of the road environment, vehicle load, road materials and other factors on cracks during road operation, effectively improving the accuracy and efficiency of crack detection and evaluation.

[0127] In this embodiment, determining the risk level of the area where the crack is located based on the collected deflection data and the preset allowable deflection value includes:

[0128] The collected deflection data is compared with the preset allowable deflection value. If the collected deflection data exceeds the preset allowable deflection value, the area is determined to be a high-risk area for structural failure.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A crack evaluation method based on three-dimensional ground-penetrating radar and deflection values, characterized in that, The specific steps include: S1: Obtain road surface crack images using three-dimensional ground-penetrating radar and simultaneously collect deflection value data at the crack locations using a deflectometer; preprocess the road surface crack images and deflection value data to obtain processed road surface crack images and corresponding deflection value datasets. S2: Establish a pre-trained crack classification model, and use the pre-trained crack classification model to classify the processed road surface crack image to obtain the classified road surface crack image. S3: Convert the classified pavement crack image into a binary image, and obtain the average crack energy of the pavement crack image based on the binary image. Dynamically weight and fuse the average crack energy with the deflection value data to obtain the comprehensive crack energy. S4: Determine several static parameters that affect the energy distribution of pavement crack images, analyze the degree of influence of several static parameters on the comprehensive crack energy, and judge the risk level of the crack area based on the collected deflection data and the preset deflection tolerance value, so as to achieve the evaluation of different types of cracks. The static parameters include surface layer gradation, pavement layer thickness, humidity, traffic volume, base course materials, diurnal temperature range, subgrade soil materials, service life, and precipitation.

2. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 1, characterized in that, In S3, the specific steps for converting the classified road surface crack image into a binary image include: The classified road surface crack map is converted into a grayscale image; The operation of distinguishing linear crack regions from background regions in the grayscale image includes: Count the pixel values ​​of all grayscale images, determine the distribution range of pixel values, and denote the minimum value as d. min The maximum value is denoted as d. max The pixel distribution range is represented as [d] min d max ]; Let the grayscale threshold be d. t With grayscale threshold d t Using the initial critical point, the pixel distribution range is divided into interval C [d min d t ] and interval D (d t d max ]; Wherein, the grayscale threshold d t The range of values ​​for is (d min d max -1), and the pixel increment step is 1 pixel; The number of pixels in intervals C and D are counted as n1 and n2 respectively, and the weights of the number of pixels in intervals C and D in the entire image are calculated as e1 and e2 respectively. 1＝ n1 / (n1+n2), e 2＝ n2 / (n1+n2), the average pixel value in intervals C and D is d c1 and d d2 The inter-class variance of image pixels in intervals C and D is calculated as E = e1·e2· / (d c1 -d d2 ) 2 ; Iterate through the grayscale value thresholds to obtain the grayscale value threshold d that maximizes the E value. tm The grayscale threshold d that maximizes the E value. tm As the critical pixel value between the linear crack region and the background region; With the final grayscale value threshold d tm As the final dividing point, gray values ​​less than d are... tm The pixel value of the pixel is changed to 0, and the gray value is greater than or equal to d. tm The pixel value of each pixel is changed to 1, thus obtaining a binarized image.

3. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 2, characterized in that, The specific steps for obtaining the average energy of road surface cracks based on the binarized image include: The binarized image is transformed into a coordinate system: a Cartesian coordinate system is established with the lower left corner of the binarized image as the origin, the horizontal rightward direction is the positive x-axis, and the vertical upward direction is the positive y-axis, forming a binarized image matrix, represented as: In the formula, j is the imaginary unit, e = 0, 1, 2, ..., J-1; f = 0, 1, 2, ..., K-1; J is the number of pixels in the image matrix along the x-axis, and K is the number of pixels in the image matrix along the y-axis; J S K is the angular frequency in the x-axis direction. S Angular frequency in the y-axis direction; The binarized image matrix is ​​split into several combinations of sine and cosine functions with known frequencies and amplitudes; The energy distribution of the road surface crack image is obtained based on the combination of the split sine and cosine functions. The energy distribution of the road surface crack image in one period signal is calculated using the following formula: In the formula, L is the image signal period in the x-axis direction, and M is the image signal period in the y-axis direction; P i H represents the i-th frequency; i For P i The corresponding energy values, and the energy distribution of the road surface crack image, are the energy values ​​corresponding to each frequency. Calculate the average crack energy H based on the energy distribution of road crack images. a The calculation formula is: In the formula, N represents the number of road surface crack images.

4. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 3, characterized in that, The specific steps for dynamically weighting and fusing the average crack energy with the deflection data to obtain the comprehensive crack energy include: Kriging interpolation is performed on the deflection data to generate a continuous deflection field surface, as shown in the following formula: Where, d i Let λ be the deflection value at the measuring point. i Let be the weight coefficients, and satisfy the unbiased property ∑λ i =1 and the minimum variance condition; The formula for calculating the overall fracture energy is as follows: Where D(x,y) is the deflection value at the current position, D crit Let α be the standard value of pavement deflection, and α be the coupling coefficient determined by regression of historical data.

5. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 4, characterized in that, In S4, the specific steps for determining several static parameters that affect the energy distribution of the pavement crack image and analyzing the degree of influence of these static parameters on the overall crack energy include: The static parameters are standardized using the following formula: Among them, X i (k) represents the k-th sample value of the i-th parameter; X i '(k) represents the standardized value; The correlation between surface layer gradation, pavement layer thickness, humidity, traffic volume, base course material, diurnal temperature range, subgrade soil material, service life, precipitation, and comprehensive crack energy was calculated to obtain the influence of different static parameters on comprehensive crack energy, including: Calculate each static parameter and the overall fracture energy H c The correlation coefficient at sample k is calculated using the following formula: Among them, H' c (k) and N' i (k) represents the standardized value; ρ is the resolution coefficient; The correlation between each static coefficient and the overall fracture energy is calculated based on the correlation coefficient. The calculation formula is as follows:

6. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 1, characterized in that, In S1, the specific steps for preprocessing the road surface crack image include: Determine the location of the survey lines on the road surface to be inspected and mark the station numbers; The antenna detection parameters of the 3D ground-penetrating radar are set according to the condition of the road surface to be detected, so as to obtain images of road surface cracks. The three-dimensional ground-penetrating radar data processing software rS1icer was used to sequentially perform time zero-point, interference suppression, inverse Fourier transform, background removal, bandpass filtering, gain compensation and offset imaging processing on the road surface crack image to highlight the location of the abnormal body in the road surface crack image. The location of the crack and the corresponding station number are determined from the anomaly.

7. The crack evaluation method based on three-dimensional ground-penetrating radar and deflection values ​​according to claim 1, characterized in that, In S2, the specific steps for building a pre-trained crack classification model include: Based on the processed pavement crack image, horizontal slices of the surface at the crack, horizontal slices of the depth of the surface layer and the base layer, horizontal slices of the depth of the base layer and the base layer, and horizontal slices of the depth of the base layer and the subbase layer are obtained. Based on the slicing of the acquired pavement crack images and the crack determination method, the pavement crack images are divided into the following types: top-bottom development without cracks in the middle, top-bottom development with cracks in the middle and partially penetrating, completely penetrating, and top-to-bottom development. An original crack classification model is established, and the original crack classification model is trained based on the classified pavement crack images to obtain a pre-trained crack classification model.

Images