Pseudo-random number-based asphalt road ground penetrating radar cavity disease heterogeneous modeling method and evaluation method thereof

Abstract

The invention discloses a pseudo-random number-based asphalt road ground penetrating radar cavity disease heterogeneous modeling method and an evaluation method thereof. At present, an accurate heterogeneous modeling mode for asphalt road ground penetrating radar cavity diseases does not exist. According to the asphalt road ground penetrating radar cavity disease heterogeneous modeling method, obtained actual road related data are mapped to different probability ranges through random numbers, so that the value range of non-uniform distribution of dielectric constants of all structural layers of an actual road is completed, and the dielectric constants are mapped to the random numbers; and after forming a heterogeneous asphalt road three-dimensional forward model by using gprMax analog calculation, integrating the heterogeneous asphalt road three-dimensional forward model into the cavity defect model, and establishing an asphalt road cavity defect three-dimensional forward model.

Description

Technical Field

[0001] The present invention specifically relates to a pseudo-random number-based asphalt road ground penetrating radar cavity disease heterogeneous modeling method and an evaluation method thereof, belonging to the technical field of road disease. Background Art

[0002] During their service life, roads are subject to degradation due to a combination of factors, including the external environment, traffic loads, and the aging of their performance. Surface defects are relatively easy to detect and address through visual inspection. In contrast, internal defects are difficult to detect and diagnose due to their hidden nature. Once they occur, they can have a more severe impact on the overall performance and service life of the road surface. Among these defects, cavities in asphalt roads can easily lead to direct or indirect disasters, such as road collapse, posing a serious threat to life and property. Therefore, timely detection and containment of cavities are crucial.

[0003] Conventional methods for detecting hidden defects in asphalt roads damage the pavement structure and fail to effectively identify them. The introduction of non-destructive testing technologies such as infrared thermal imaging and ultrasonic testing can directly improve detection efficiency and accuracy, facilitating the identification of hidden defects within roads. However, infrared imaging primarily excels at detecting shallow anomalies and is relatively limited in detecting underlying anomalies beneath the roadbed. While CT technology can be used to assess full-depth pavement damage, it is restricted by radiation safety regulations, making rapid measurements unsuitable in practical applications. In recent years, ground-penetrating radar (GPR) has become one of the most effective non-destructive testing technologies for infrastructure maintenance, due to its high efficiency and rapidity compared to other non-destructive testing methods. Improving the applicability of GPR for detecting hidden road defects is of great significance. Due to the inherent scarcity of GPR data, the actual amount of GPR data is relatively small. Therefore, simulation data is often generated using the finite-difference time-domain method (FDTD) to increase the amount of GPR data available. gprMax is a widely used FDTD-based tool for simulating electromagnetic wave propagation. Ghozzi et al. used gprMax to generate GPR forward modeling data and proposed a sensitivity analysis of GPR data. The results showed that GPR detection accuracy is most significantly affected by the incident waveform. They concluded that FDTD can be used to simulate the interaction between ground-penetrating radar signals and pavement moisture, and gprMax was used to simulate void defects beneath concrete pavement structures. The results showed that the correlation coefficient between the simulated and measured signals was greater than 0.8, indicating that the simulated data can be used to augment existing databases. A comparison between gprMax-generated data and real radar data was also conducted, validating the feasibility of using simulated sample data to augment GPR databases. Currently, GPR data front-end processing is time-consuming and subject to significant environmental clutter interference, resulting in a scarcity of high-quality data on hidden road defects, which in turn hinders the development of intelligent road nondestructive testing. Although deep learning methods based on Generative Adversarial Networks (GANs) have achieved some success in GPR image clutter suppression and data synthesis, they still have significant limitations. Existing studies have mostly employed single-task GAN models. For example, clutter removal and data synthesis require separate training of the generator and discriminator. This results in the construction of multiple independent networks for functional expansion, which is time-consuming and has poor transferability. Furthermore, while FDTD can be used to simulate hidden road defects and expand data, its computational efficiency is low, especially for three-dimensional forward models, which require extremely high computing power. Furthermore, current three-dimensional GPR forward models are mostly based on a homogeneous assumption, which is inconsistent with the heterogeneous structure of actual roads. Research has shown that the computational time of the two-dimensional homogeneous gprMax model is significantly too long. If a more accurate three-dimensional heterogeneous model is used, the computational complexity will increase exponentially. Currently, there is no accurate heterogeneous modeling method for ground-penetrating radar cavity defects on asphalt roads. It lacks standardization and systematicity, and there is no accurate calculation, processing, and evaluation method that can be used on actual roads. Summary of the Invention

[0004] The scarcity of ground-penetrating radar data on road void defects causes performance bottlenecks in intelligent recognition models, which restricts the intelligent development of road non-destructive testing. The present invention provides a pseudo-random number-based asphalt road ground-penetrating radar void disease heterogeneity modeling method and its evaluation method.

[0005] A pseudo-random number-based method for heterogeneous modeling of asphalt road ground penetrating radar cavity defects is proposed. The method maps the acquired actual road-related data to different probability ranges through random numbers, thereby completing the value range of the non-uniform distribution of the dielectric constant of each structural layer of the actual road. After mapping the dielectric constant to the random number, a three-dimensional forward model of the heterogeneous asphalt road is formed by gprMax simulation calculation. The three-dimensional forward model of the heterogeneous asphalt road is then integrated into the cavity defect model to establish a three-dimensional forward model of the asphalt road cavity defect.

[0006] As a preferred solution, the heterogeneous modeling method of asphalt road ground penetrating radar cavity disease is divided into the following steps:

[0007] Step 1: Generate a random number sequence uniformly distributed in the interval [0,1] in each structural layer of the asphalt road;

[0008] Step 2: The actual road-related data obtained are the on-site measured material parameters, and the range of the dielectric constant of each layer is determined in combination with the on-site measured material parameters;

[0009] Step 3: Map the dielectric constant to a random number and use gprMax to simulate the three-dimensional forward model of the heterogeneous asphalt road. The Mersenne twister algorithm in the pseudo-random number generation algorithm is used for calculation. The calculation formula is:

[0010] X n =f(X n-r ,X n-1 )

[0011] In the above formula, X n is the nth random number; r is a custom random seed; f is a complex nonlinear function, which includes linear transformation and bit operation.

[0012] As a preferred solution: f includes linear transformation and bit operation, and the process of linear transformation and bit operation is based on the Mersennetwister algorithm in pseudo-random numbers. The processing process is: establish a three-dimensional asphalt road model with a soil layer, a subbase layer, a base layer, a lower layer and an upper layer arranged from bottom to top, the length of the asphalt road three-dimensional model ranges from 550 to 650 mm, the width of the asphalt road three-dimensional model ranges from 100 to 300 mm, and the height of the asphalt road three-dimensional model ranges from 253 to 333 mm. The asphalt road three-dimensional model is processed on the xz plane, and an irregular region is generated inside the asphalt road three-dimensional model by using the generate_irregular_region function. Then, a random distribution of different materials in the asphalt road three-dimensional model is formed by using a pseudo-random number generation algorithm, and a uniformly distributed random number is generated by the rand function. After the random number is mapped to the number of different materials according to the threshold, the data of the processed asphalt road three-dimensional model is stored.

[0013] As a preferred solution: using gprMax to simulate and calculate the three-dimensional forward model of heterogeneous asphalt roads includes a comparison and verification process, which is: based on gprMax, a homogeneous asphalt road three-dimensional model and a heterogeneous asphalt road three-dimensional model are respectively formed, multi-channel measurement point data are generated, and a ground penetrating radar B-scan image is synthesized. The B-scan image generated by the homogeneous asphalt road three-dimensional model is an image with excessive smoothness and idealized features. The fluctuation of the dielectric constant in the heterogeneous asphalt road three-dimensional model causes the signal received by the antenna to present clutter interference features in the B-scan image, and the formed B-scan image is an image showing the non-uniformity of the road.

[0014] As a preferred solution, the 3D forward model of heterogeneous asphalt pavement is integrated into the void defect model. The process of establishing the 3D forward model of void defects in asphalt pavement is as follows: the void defect model includes a cubic void defect model, a spherical void defect model, and a cylindrical void defect model;

[0015] The process of establishing the cavity defect model is as follows: obtain the depth d of the cavity defect from the asphalt road surface, the radius R of the bottom surface of the cylinder or the sphere, and the side length H of the cube. When the cube is a cube, the depth of the cube is equal to the height of the cube. In the simulation calculation, the ground penetrating radar excitation source adopts the Ricker wavelet in gprMax, and the center frequency ranges from 200 to 400 MHz. The radar transmitting antenna TA and the radar receiving antenna RA are both placed on the asphalt road surface. The radar transmitting antenna TA and the radar receiving antenna RA are both scanned along the x-axis direction, and the step spacing is 0.01 m, so that multiple measuring lines are set on the road surface, and the number of scans for each measuring line ranges from 500 to 580 times.

[0016] As a preferred solution: by setting up multiple measuring lines on the road surface, scanning each measuring line a predetermined number of times, the three-dimensional forward model of all asphalt road void defects is simulated and calculated based on gprMax, and accelerated by GPU, finally obtaining three types of non-homogeneous void defect B-scan images and three types of homogeneous asphalt road void defect B-scan images.

[0017] A method for assessing void defects in asphalt roads using ground-penetrating radar (GPR) based on pseudo-random numbers is characterized in that: the GPR method for assessing void defects in asphalt roads comprises the following steps: using gprMax simulation to form a three-dimensional forward model of a heterogeneous asphalt road, integrating the three-dimensional forward model of the heterogeneous asphalt road into a void defect model; establishing the three-dimensional forward model of the asphalt road void defect; and obtaining three types of heterogeneous void defect B-scan images. Taking the homogeneous void defect B-scan image as a benchmark, when the signal-to-noise ratio of the homogeneous void defect B-scan image is 0 dB, the signal-to-noise ratio of the heterogeneous void defect B-scan image is lower than 18.38 dB, and the signal-to-noise ratio of the GPR image in actual detection is lower than 20 dB, it indicates that the three types of heterogeneous void defect B-scan images are highly consistent with the clutter in actual detection, and the heterogeneity of the asphalt road structure reflected by the three-dimensional forward model of the corresponding asphalt road void defect is reliable.

[0018] Compared with the existing technology, the present invention provides a pseudo-random number-based asphalt road ground penetrating radar cavity disease heterogeneity modeling method and its evaluation method, which has the following beneficial effects:

[0019] 1. The pseudo-random number-based ground penetrating radar asphalt road cavity defect heterogeneous modeling method of the present invention is to map the actual road-related data obtained to different probability ranges through random numbers, thereby completing the non-uniform distribution of the dielectric constants of each structural layer of the actual road. After using gprMax simulation calculation to form a three-dimensional forward model of the heterogeneous asphalt road, the three-dimensional forward model of the heterogeneous asphalt road is integrated into the cavity defect model to establish a three-dimensional forward model of the asphalt road cavity defect. The three-dimensional forward model of the asphalt road cavity defect has higher accuracy in simulating the electromagnetic response of the asphalt road structure, and its clutter characteristics are obvious and accurate.

[0020] 2. The asphalt road ground penetrating radar cavity disease assessment method of the present invention uses gprMax simulation calculation to form a three-dimensional forward model of a non-homogeneous asphalt road, and then integrates the three-dimensional forward model of the non-homogeneous asphalt road into the cavity defect model. After establishing the three-dimensional forward model of the asphalt road cavity defect, three types of non-homogeneous cavity defect B-scan images are obtained, and the corresponding quantitative relationship between the degree of fit between the three types of non-homogeneous cavity defect B-scan images and the clutter in the actual detection is obtained, thereby completing the quantitative assessment process of the cavity disease of the actual road section through the three-dimensional forward model of the asphalt road cavity defect, which is conducive to accurate guidance and early warning of the actual situation of the cavity disease in the actual road section. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the 2D forward model of a traditional asphalt road based on gprMax;

[0022] Figure 2 Schematic diagram of the 3D forward model of a traditional asphalt road based on gprMax;

[0023] Figure 3 It is a three-dimensional forward model of heterogeneous asphalt road based on pseudo-random number algorithm;

[0024] Figure 4a The results of the 3D forward model of the ground penetrating radar for homogeneous asphalt roads are shown in the figure. The model results correspond to the B-scan of the homogeneous model.

[0025] Figure 4b The results of the three-dimensional forward model of the ground penetrating radar for a heterogeneous asphalt road are shown in the figure. The model results correspond to the B-scan of the heterogeneous model.

[0026] Figure 5 Schematic diagram of the three-dimensional forward model of heterogeneous asphalt road;

[0027] Figure 6 This is a schematic diagram of the three-dimensional forward model of homogeneous asphalt road;

[0028] Figure 7 Schematic diagram of the three-dimensional forward model of the cavity defect in the cylindrical body of heterogeneous asphalt road;

[0029] Figure 8 Schematic diagram of the three-dimensional forward model of cubic cavity defects in heterogeneous asphalt roads;

[0030] Figure 9 Schematic diagram of the three-dimensional forward model of spherical cavity defects in heterogeneous asphalt roads;

[0031] Figure 10 This is a schematic diagram of the 3D forward model of asphalt road void defects and the radar scanning method. The viewing direction in the figure is from the longitudinal section.

[0032] Figure 11 This is a schematic diagram of the 3D forward model of asphalt road cavity defects and the radar scanning method. The viewing direction in the figure is the horizontal plane direction.

[0033] Figure 12a The forward model of asphalt road void defects is a B-scan of cylindrical void defects at different depths under a heterogeneous model.

[0034] Figure 12b The B-scan of the cylindrical cavity defect at different depths in the forward model of the asphalt road cavity defect is in the homogeneous model;

[0035] Figure 12c The B-scan of spherical void defects at different depths in the forward model of asphalt road void defects under the heterogeneous model;

[0036] Figure 12d The B-scan of spherical void defects at different depths in the forward model of asphalt road void defects under the homogeneous model;

[0037] Figure 12e The forward model of asphalt road void defects is in the heterogeneous model with cube void defects at different depths B-scan;

[0038] Figure 12f The forward model of asphalt road void defects is in the homogeneous model with cube void defects at different depths B-scan. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] Specific implementation method 1: Combination Figure 1 、 Figure 2 、 Figure 3 、 Figure 4a 、 Figure 4b 、 Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 、 Figure 12a 、 Figure 12b 、 Figure 12c 、 Figure 12d 、 Figure 12e and Figure 12fThis embodiment explains the method for heterogeneous modeling of asphalt road ground penetrating radar cavity defects in this embodiment, which is to map the acquired actual road-related data to different probability ranges through random numbers, thereby completing the non-uniform distribution range of the dielectric constants of each structural layer of the actual road, mapping the dielectric constants to random numbers, and using gprMax simulation calculation to form a three-dimensional forward model of the heterogeneous asphalt road. The three-dimensional forward model of the heterogeneous asphalt road is then integrated into the cavity defect model to establish a three-dimensional forward model of the asphalt road cavity defect.

[0041] In this implementation, gprMax is a 3D simulation tool for GPR signals based on FDTD. It can perform 2D or 3D modeling of asphalt roads and simulate GPR signals during actual propagation.

[0042] In this embodiment, the heterogeneous modeling method for asphalt road ground penetrating radar cavity disease is divided into the following steps:

[0043] Step 1: Generate a random number sequence uniformly distributed in the interval [0,1] in each structural layer of the asphalt road;

[0044] Step 2: The actual road-related data obtained are the on-site measured material parameters, and the range of the dielectric constant of each layer is determined in combination with the on-site measured material parameters;

[0045] Step 3: Map the dielectric constant to a random number and use gprMax to simulate the three-dimensional forward model of the heterogeneous asphalt road. The Mersenne twister algorithm in the pseudo-random number generation algorithm is used for calculation. The calculation formula is:

[0046] X n =f(X n-r , X n-1 )

[0047] In the above formula, X n is the nth random number; r is a custom random seed; f is a complex nonlinear function, which includes linear transformation and bit operation.

[0048] Specific implementation method 2: This implementation method is a further limitation of specific implementation method 1. In this implementation method, f includes linear transformation and bit operation. The process of linear transformation and bit operation is a processing process based on the Mersennetwister algorithm in pseudo-random numbers. Specifically, a three-dimensional model of an asphalt road is established, which is provided with a soil layer, a subbase layer, a base layer, a lower layer and an upper layer from bottom to top. The length of the three-dimensional asphalt road model ranges from 550 to 650 mm, the width of the three-dimensional asphalt road model ranges from 100 to 300 mm, the height of the three-dimensional asphalt road model ranges from 253 to 333 mm, and the xyz coordinate axes of the three-dimensional asphalt road model are The directions are: x direction represents the length direction of the asphalt road 3D model, y direction represents the height direction of the asphalt road 3D model, and z direction represents the height direction of the asphalt road 3D model. The asphalt road 3D model is processed on the xz plane. After the generate_irregular_region function is used to generate an irregular region inside the asphalt road 3D model, a pseudo-random number generation algorithm is used to form a random distribution of different materials in the asphalt road 3D model. The rand function is used to generate uniformly distributed random numbers. After the random numbers are mapped to the numbers of different materials according to the threshold, the processed asphalt road 3D model data is stored.

[0049] The operation process of the above content is divided into the following parts, specifically:

[0050] Part 1: Initialization and GUI creation:

[0051] When the code starts, clear the workspace and close all graphics windows.

[0052] Two variables a and b are defined to control the number of times the model is generated.

[0053] A simple GUI interface is created where the user can select "manual check mode" or "automatic generation mode". After selecting, the program sets the global variable manual_check according to the user's choice.

[0054] Part 2: Generation of 3D model of asphalt road:

[0055] A three-dimensional array Road_3d is defined to store model data. The model size is 600x200x303.

[0056] A random number generator is used to generate material data for different layers, including soil layer, subbase layer, base layer, lower layer, and upper layer. The data for each layer is generated by random numbers and assigned different material numbers according to a certain probability.

[0057] Part III: Processing and preservation of asphalt road 3D models:

[0058] Save the generated 3D model data into a MAT file.

[0059] Through the loop, the program loads the model data multiple times and processes it on the xz plane.

[0060] Use the generate_irregular_region function to generate irregular regions in the model.

[0061] Part 4: Inspection and preservation of asphalt road 3D models:

[0062] After each generation of the asphalt road 3D model, the program will check it according to the mode selected by the user.

[0063] The received asphalt road 3D model will be saved as an image and related text and HDF5 files will be generated.

[0064] Part 5: Further generation and organization of text and HDF5 files:

[0065] The program generates multiple subfolders for each model and creates text files and HDF5 files in them.

[0066] The text file records the properties of different materials, while the HDF5 file stores the three-dimensional data of the model.

[0067] The program configured in the above process uses a pseudo-random number generation algorithm to generate a random distribution of different materials in the road model. The code uses the rand() function to generate uniformly distributed random numbers. Then, based on a threshold, these random numbers are mapped to different material numbers. Numbers 1-8 represent soil layer materials, numbers 9-16 represent eight subbase materials, and so on. To ensure that the generated road model is repeatable, that is, the results are the same each time it is run, the code controls the consistency of the pseudo-random sequence by fixing the random seed.

[0068] Specific implementation method three: This implementation method is a further limitation of specific implementation method one or two. In this implementation method, Figure 4a and Figure 4bAs shown, in this embodiment, the use of gprMax to simulate and calculate a 3D forward model of a heterogeneous asphalt road includes a comparison and verification process. The comparison and verification process involves: generating a 3D model of a homogeneous asphalt road and a 3D model of a heterogeneous asphalt road based on gprMax, generating multi-channel measurement point data, and synthesizing a ground-penetrating radar (GPR) B-scan image. The B-scan image generated by the homogeneous asphalt road 3D model exhibits overly smooth and idealized characteristics. Fluctuations in the dielectric constant within the heterogeneous asphalt road 3D model cause the antenna-received signal to exhibit clutter interference in the B-scan image, resulting in a B-scan image that reflects road heterogeneity. In other words, fluctuations in the dielectric constant within the heterogeneous model cause clutter interference in the B-scan image. In contrast, the B-scan image generated by the homogeneous model exhibits overly smooth and idealized characteristics, which do not conform to actual asphalt road inspection conditions. The heterogeneous asphalt road 3D forward model more realistically reflects road heterogeneity and is more suitable for actual GPR inspection scenarios.

[0069] In this embodiment, the Figure 5 and Figure 6 As shown in Table 1, the two forward modeling models share the same structural geometric parameters but differ in material uniformity. The asphalt road forward model is 6m long, 2m wide, and 3.03m high. Based on actual engineering experience, the material geometric parameters and dielectric constants for the two forward modeling models are shown in Table 1.

[0070] Table 1 Parameters of the three-dimensional forward model of asphalt roads

[0071]

[0072] Combine Figure 7 、 Figure 8 and Figure 9 As shown, the present invention constructs three different morphological void defects: cubic, spherical, and cylindrical. Based on the 3D forward model of asphalt pavement, the void defect model is integrated to establish a 3D forward model of asphalt pavement void defects. The interior of the void defect is primarily composed of air, and the electromagnetic parameters are consistent with the dielectric constant of air, all being 1.

[0073] In this embodiment, the heterogeneous asphalt road 3D forward model is integrated into the void defect model. The process of establishing the asphalt road void defect 3D forward model is as follows: the void defect model includes a cubic void defect model, a spherical void defect model and a cylindrical void defect model;

[0074] Combine Figure 10 and Figure 11As shown in Figure 2, the process for establishing the void defect model is as follows: All void defects in the present invention's model are located below the pavement subbase, where d represents the depth of the void defect from the asphalt road surface, R is the radius of the cylinder base and sphere, and H is the side length of the cube. The specific parameters of the three types of void defect models are shown in Table 2. Each type of void defect varies in size and depth within the asphalt road structure. A total of 408 forward models for the three types of asphalt road void defects were established.

[0075] Table 2 Geometric parameters of void defects

[0076]

[0077] To align with actual testing conditions, the ground-penetrating radar excitation source in the simulation was the Ricker wavelet in gprMax, with a center frequency set to 200 MHz. The radar transmitting antenna (TA) and receiving antenna (RA) were placed on the asphalt road surface. Scanning was performed along the x-axis with a step interval of 0.01 m. Because the model is a three-dimensional structure, multiple measurement lines were set along the road surface, with each line undergoing 570 scans.

[0078] Combine Figure 12a 、 Figure 12b 、 Figure 12c 、 Figure 12d 、 Figure 12e and Figure 12f As shown in Figure 2, a 3D forward model of all asphalt pavement void defects was simulated using gprMax and accelerated by GPU. This study ultimately generated 1,240 B-scan images of three types of void defects in heterogeneous and homogeneous asphalt pavements.

[0079] Compared with the image background obtained by traditional homogeneous forward model simulation, which is too smooth and lacks accuracy, the three-dimensional forward model of heterogeneous asphalt road cavity defects constructed by the present invention based on the pseudo-random generation algorithm can more realistically simulate the electromagnetic wave response characteristics of the road surrounding environment. Its characteristics are consistent with the clutter in actual detection, and the forward modeling results better reflect the heterogeneity of the asphalt road structure.

[0080] Specific implementation method four: Combination Figure 1 、 Figure 2 、 Figure 3 、 Figure 4a 、 Figure 4b 、 Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 、 Figure 12a 、 Figure 12b 、 Figure 12c 、 Figure 12d、 Figure 12e and Figure 12f This embodiment is described. This embodiment is a method for evaluating void defects in asphalt roads using ground-penetrating radar (GPR) based on pseudo-random numbers. The method for evaluating void defects in asphalt roads using GPRMax simulation calculations is to form a three-dimensional forward model of a heterogeneous asphalt road. The specific process is to map the acquired actual road-related data to different probability ranges through random numbers, thereby completing the range of values for the non-uniform distribution of the dielectric constants of each structural layer of the actual road. After mapping the dielectric constants to random numbers and forming a three-dimensional forward model of a heterogeneous asphalt road using GPRMax simulation calculations, the three-dimensional forward model of the heterogeneous asphalt road is integrated into the void defect model to establish a three-dimensional forward model of void defects in asphalt roads.

[0081] In this embodiment, the heterogeneous modeling method for asphalt road ground penetrating radar cavity disease is divided into the following steps:

[0082] Step 1: Generate a random number sequence uniformly distributed in the interval [0,1] in each structural layer of the asphalt road;

[0083] Step 2: The actual road-related data obtained are the on-site measured material parameters, and the range of the dielectric constant of each layer is determined in combination with the on-site measured material parameters;

[0084] Step 3: Map the dielectric constant to a random number and use gprMax to simulate the three-dimensional forward model of the heterogeneous asphalt road. The Mersenne twister algorithm in the pseudo-random number generation algorithm is used for calculation. The calculation formula is:

[0085] X n =f(X n-r , X n-1 )X n =f(X n-r ,X n-1 )

[0086] In the above formula, X n is the nth random number; r is a custom random seed; f is a complex nonlinear function, which includes linear transformation and bit operation.

[0087] Furthermore, f includes linear transformations and bit operations.

[0088] In this embodiment, the use of gprMax to simulate and calculate the three-dimensional forward model of the heterogeneous asphalt road includes a comparison and verification process, which is: based on gprMax, a homogeneous asphalt road three-dimensional model and a heterogeneous asphalt road three-dimensional model are respectively formed, multi-channel measurement point data are generated, and a ground penetrating radar B-scan image is synthesized. The B-scan image generated by the homogeneous asphalt road three-dimensional model is an image with excessive smoothness and idealized features. The fluctuation of the dielectric constant in the heterogeneous asphalt road three-dimensional model causes the signal received by the antenna to present clutter interference features in the B-scan image, and the formed B-scan image is an image showing the non-uniformity of the road.

[0089] In this embodiment, the heterogeneous asphalt road 3D forward model is integrated into the void defect model. The process of establishing the asphalt road void defect 3D forward model is as follows: the void defect model includes a cubic void defect model, a spherical void defect model and a cylindrical void defect model;

[0090] The process of establishing the cavity defect model is as follows: obtain the depth d of the cavity defect from the asphalt road surface, the radius R of the bottom surface of the cylinder or the sphere, and the side length H of the cube. In the simulation calculation, the ground penetrating radar excitation source uses the Ricker wavelet in gprMax, and the center frequency is set to 200-400MHz. The optimal value is 200MHz. It can also be configured according to specific requirements. The radar transmitting antenna TA and the radar receiving antenna RA are both placed on the asphalt road surface. The radar transmitting antenna TA and the radar receiving antenna RA are scanned along the x-axis direction with a step spacing of 0.01m, thereby setting multiple measuring lines on the road surface. Each measuring line is scanned 500-580 times. It can also be configured according to specific requirements.

[0091] In this embodiment, multiple measuring lines are set on the road surface. After each measuring line is scanned 570 times (it is best to give a range value), the three-dimensional forward model of all asphalt road void defects is simulated and calculated based on gprMax. Finally, three types of inhomogeneous void defect B-scan images are obtained through GPU acceleration.

[0092] The asphalt road ground penetrating radar cavity disease assessment method is to use gprMax simulation calculation to form a heterogeneous asphalt road 3D forward model, then integrate the heterogeneous asphalt road 3D forward model into the cavity defect model. After establishing the asphalt road cavity defect 3D forward model, three types of heterogeneous cavity defect B-scan images are obtained. Taking the homogeneous cavity defect B-scan image as the benchmark, when the signal-to-noise ratio of the homogeneous cavity defect B-scan image is 0dB, the signal-to-noise ratio of the heterogeneous cavity defect B-scan image is lower than 18.38dB, and the signal-to-noise ratio of the ground penetrating radar image in actual detection is lower than 20dB, it indicates that the three types of heterogeneous cavity defect B-scan images are highly consistent with the clutter in actual detection, and the asphalt road cavity defect 3D forward model reflects the heterogeneity of the asphalt road structure with reliability. The data information it reflects can provide early warning, prompts and guidance for actual road sections. When the signal-to-noise ratio of the B-scan image of the homogeneous void defect is 0dB, the signal-to-noise ratio of the B-scan image of the inhomogeneous void defect is higher than 18.38dB, and the signal-to-noise ratio of the ground penetrating radar image in actual detection is higher than 20dB, it indicates that the degree of consistency between the B-scan images of the three inhomogeneous void defects and the clutter in actual detection is low. The non-uniformity of the asphalt road structure reflected by the corresponding three-dimensional forward model of the asphalt road void defect needs to be verified or remodeled.

Claims

1. A pseudo-random number-based method for heterogeneous modeling of asphalt road ground penetrating radar cavity damage, characterized by: The heterogeneous modeling method for asphalt road ground penetrating radar cavity defects is to map the actual road-related data obtained to different probability ranges through random numbers, thereby completing the value range of the non-uniform distribution of the dielectric constant of each structural layer of the actual road. After mapping the dielectric constant to random numbers, the three-dimensional forward model of the heterogeneous asphalt road is formed by gprMax simulation calculation. The three-dimensional forward model of the heterogeneous asphalt road is then integrated into the cavity defect model to establish a three-dimensional forward model of asphalt road cavity defects.

2. The method for heterogeneous modeling of asphalt road cavity damage based on pseudo-random numbers according to claim 1 is characterized by: The heterogeneous modeling method of asphalt road ground penetrating radar cavity disease is divided into the following steps: Step 1: Generate a random number sequence uniformly distributed in the interval [0, 1] in each structural layer of the asphalt road; Step 2: The actual road-related data obtained are the on-site measured material parameters, and the range of the dielectric constant of each layer is determined in combination with the on-site measured material parameters; Step 3: Map the dielectric constant to a random number and use gprMax to simulate the three-dimensional forward model of the heterogeneous asphalt road. The Mersenne twister algorithm in the pseudo-random number generation algorithm is used for calculation. The calculation formula is: X n =f(X n-r ,X n-1 ) In the above formula, X n is the nth random number; r is a custom random seed; f is a complex nonlinear function, which includes linear transformation and bit operation.

3. The method for heterogeneous modeling of asphalt road cavity damage based on pseudo-random numbers according to claim 2 is characterized by: f includes linear transformation and bit operation. The process of linear transformation and bit operation is based on the Mersenne twister algorithm in pseudo-random numbers. The processing process is as follows: a three-dimensional asphalt road model is established, which is arranged from bottom to top with a soil layer, a subbase layer, a base layer, a lower layer, and an upper layer. The length of the asphalt road three-dimensional model ranges from 550 to 650 mm, the width of the asphalt road three-dimensional model ranges from 100 to 300 mm, and the height of the asphalt road three-dimensional model ranges from 253 to 333 mm. The asphalt road three-dimensional model is processed on the xz plane. After the generate_irregular_region function is used to generate an irregular region inside the asphalt road three-dimensional model, a pseudo-random number generation algorithm is used to form a random distribution of different materials in the asphalt road three-dimensional model. The rand function is used to generate uniformly distributed random numbers. After the random numbers are mapped to the numbers of different materials according to the threshold, the processed asphalt road three-dimensional model data is stored.

4. The method for heterogeneous modeling of asphalt road cavity damage based on pseudo-random numbers according to claim 1, 2 or 3, characterized in that: The use of gprMax to simulate and calculate the three-dimensional forward model of heterogeneous asphalt roads includes a comparison and verification process. The comparison and verification process is as follows: based on gprMax, a three-dimensional model of a homogeneous asphalt road and a three-dimensional model of a heterogeneous asphalt road are respectively formed, multi-channel measurement point data are generated, and a ground penetrating radar B-scan image is synthesized. The B-scan image generated by the three-dimensional model of the homogeneous asphalt road is an image with over-smoothed and idealized features. The fluctuation of the dielectric constant in the three-dimensional model of the heterogeneous asphalt road causes the signal received by the antenna to present clutter interference features in the B-scan image, and the resulting B-scan image is an image showing the non-uniformity of the road.

5. The method for heterogeneous modeling of asphalt road cavity damage based on pseudo-random numbers according to claim 1 is characterized by: The process of integrating the heterogeneous asphalt road 3D forward model into the void defect model and establishing the 3D forward model of the asphalt road void defect is as follows: the void defect model includes a cubic void defect model, a spherical void defect model and a cylindrical void defect model; The process of establishing the cavity defect model is as follows: obtain the depth d of the cavity defect from the asphalt road surface, the radius R of the bottom surface of the cylinder or the sphere, and the side length H of the cube. When the cube is a cube, the depth of the cube is equal to the height of the cube. In the simulation calculation, the ground penetrating radar excitation source uses the Ricker wavelet in gprMax, and the center frequency ranges from 200 to 400 MHz. The radar transmitting antenna TA and the radar receiving antenna RA are both placed on the asphalt road surface. The radar transmitting antenna TA and the radar receiving antenna RA are both scanned along the x-axis direction, and the step spacing is 0.01 m, so that multiple measuring lines are set on the road surface, and the number of scans for each measuring line ranges from 500 to 580 times.

6. The method for heterogeneous modeling of asphalt road cavity damage based on pseudo-random numbers according to claim 5, characterized in that: By setting multiple measuring lines on the road surface and scanning each measuring line a predetermined number of times, the three-dimensional forward model of all asphalt road void defects is simulated and calculated based on gprMax. With GPU acceleration, the final B-scan images of three types of inhomogeneous void defects and three types of homogeneous asphalt road void defects are obtained.

7. A method for assessing asphalt road cavity damage using ground penetrating radar based on pseudo-random numbers, characterized by: The asphalt road ground penetrating radar cavity disease assessment method is to use gprMax simulation calculation to form a heterogeneous asphalt road 3D forward model, then integrate the heterogeneous asphalt road 3D forward model into the cavity defect model. After establishing the asphalt road cavity defect 3D forward model, three types of heterogeneous cavity defect B-scan images are obtained. Taking the homogeneous cavity defect B-scan image as the benchmark, when the signal-to-noise ratio of the homogeneous cavity defect B-scan image is 0dB, the signal-to-noise ratio of the heterogeneous cavity defect B-scan image is lower than 18.38dB, and the signal-to-noise ratio of the ground penetrating radar image in actual detection is lower than 20dB, it indicates that the three types of heterogeneous cavity defect B-scan images are highly consistent with the clutter in actual detection, and the asphalt road cavity defect 3D forward model reflects the heterogeneity of the asphalt road structure with reliability.

Images