Road deep disease treatment method based on infrared thermal imaging and ground penetrating radar combination

Abstract

The application discloses a kind of road deep disease treatment methods based on infrared thermal imaging and ground penetrating radar combination, it is related to road disease detection and treatment technical field, first using infrared thermal imaging technology wide range detection determines whether there is disease, then for temperature abnormal area, arrangement mutually perpendicular survey line, using ground penetrating radar accurate positioning;With radar gray scale, using depth formula and optimized width formula estimate the width and depth of crack, and different grades are divided respectively;Finally, crack comprehensive grade decision matrix is established, and different treatment measures are obtained through crack comprehensive grade decision matrix for the three-dimensional information of crack;After treatment, maintenance archives are established, which is convenient for subsequent tracking and effect evaluation. This set of methods can help operation personnel to reasonably allocate maintenance resources, realize the change from "broken repair" to "active prevention, accurate repair", and finally significantly reduce the life cycle road maintenance cost.

Description

Technical Field

[0001] This invention relates to the field of road defect detection and treatment technology, and in particular to a method for treating deep road defects based on the combination of infrared thermal imaging and ground penetrating radar. Background Technology

[0002] In recent years, with the development of computer vision and deep learning algorithms, the frequency of ground-penetrating radar (GPR) for road detection has gradually increased. Currently, most patents focus on the design of GPR devices and the optimization of GPR detection methods, with few addressing specific treatment measures and methods for different types of road surface defects after identifying deep-seated defects using GPR. For example, patent CN 220595101 U provides an electric GPR vehicle to solve the problems of limited road conditions and laborious operation faced by existing GPRs. Patent CN 108549075 A discloses a method for determining the optimal detection height of GPR, which can effectively improve the accuracy of the dielectric constant measured by GPR, thereby further improving the accuracy of GPR testing. However, for different types of defects, there is a lack of risk level assessment based on the development trend of the defects and the proposal of differentiated treatment methods, resulting in increased operating costs for maintenance and detection.

[0003] Furthermore, ground-penetrating radar (GPR) signals are the convolution of radar wavelet and reflection coefficient. At the location of media anomalies, these signals exhibit changes in waveform, frequency, amplitude, phase, and energy. However, current radar waveform diagrams only provide a time-domain profile of the signal. When performing empirical interpretations, the frequency domain and time-frequency domain information are automatically ignored. This lack of signal characteristics easily leads to misjudgments and omissions when assessing internal damage. This is especially true when the crack development direction aligns with the survey line direction, potentially causing data processing personnel to miss or misjudge the damage. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar. This method addresses the current problem of increased maintenance and inspection costs due to the lack of risk level assessments for different types of defects based on their development trends and the absence of differentiated treatment methods.

[0005] To address the problems of the prior art, the technical solution of the present invention is as follows: The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar includes the following steps: S1. Use a vehicle equipped with a high-speed infrared thermal imager to scan the road surface, obtain temperature images of the road surface, and preprocess the temperature images to identify areas with abnormal temperatures. S2. Ground penetrating radar is used to detect areas with abnormal temperatures to obtain radar waveform images, and the radar waveform images are preprocessed and offset-corrected. The processed radar waveform image is converted into a grayscale image, and the location is determined by the trained YOLO target detection model. The region with hyperbolic features is cropped out and the feature is accurately extracted by the neural network to obtain the grayscale amplitude of the hyperbola. S3. Based on the obtained hyperbola grayscale amplitude, calculate the depth and width of the crack, classify the cracks according to the crack depth and crack width, and observe whether there is any void through radar images. S4. Taking into account different depths, widths, and the presence or absence of voids in the cracks, a comprehensive crack level decision matrix is ​​established. The three-dimensional information of depth, width, and presence or absence of voids obtained in S3 is filled into the decision matrix to obtain the final comprehensive disease level and treatment recommendations. Based on the comprehensive rating results, specific engineering treatment measures are output. S5. After treating the cracks, conduct ground-penetrating radar re-examination to ensure the treatment effect. Finally, establish a maintenance file to record each treated crack.

[0006] A more detailed technical solution is as follows: 1. Infrared thermal imaging detection technology is used to assist ground-penetrating radar (GPR) in flaw detection, solving the problems of false positives and false negatives in GPR: Ground-penetrating radar (GPR) signals are the convolution of radar wavelets and reflection coefficients. At locations of media anomalies, these signals exhibit changes in waveform, frequency, amplitude, phase, and energy. However, current radar waveform diagrams only provide a time-domain profile of the signal. When making empirical interpretations, the frequency and time-frequency domain information of the signal is automatically ignored. This lack of signal characteristics easily leads to misjudgments and missed detections when assessing internal damage. Therefore, infrared thermal imaging technology is introduced to assist GPR in flaw detection. Cracks in roads are often accompanied by defects such as voids. Because the thermal conductivity of road materials differs significantly from that of air, areas with internal defects and areas without defects will exhibit temperature differences due to these differences. Infrared thermal imaging technology can observe the temperature information of the road surface, identify areas of abnormal temperature, and mark the actual location of these areas, indicating the presence of hidden defects such as cracks and voids. This solves the problem of misjudging and missing internal damage caused by the lack of characteristic features in GPR signals.

[0007] 2. Ground-penetrating radar (GPR) detection technology is used to identify abnormal areas in infrared images and accurately locate the cracks: In the temperature anomaly areas of the aforementioned infrared images, mutually perpendicular ground-penetrating radar lines are arranged to prevent situations where the direction of crack development coincides with the direction of the radar lines. If the direction of crack development coincides with the direction of the ground-penetrating radar lines, the radar image will display a very slight hyperbolic feature, which may lead data processing personnel to believe that the area is noise interference or that there are small cracks. For obvious cracks with large spacing, ground-penetrating radar (GPR) can clearly distinguish and count them. However, if two cracks are very close together, less than the lateral resolution of GPR, they may merge into a single reflection feature on the image, making them indistinguishable. In complex crack networks, reflections on the image intertwine, making it difficult to accurately count each individual crack. GPR can provide a qualitative judgment of "crack development" or "the presence of multiple cracks" in a region and count the main cracks, but it may not be able to count all the fine, dense cracks. Therefore, the lateral resolution should also be considered. Lateral resolution determines the ability of GPR to distinguish two adjacent targets in the horizontal direction. If the crack spacing is greater than the resolution, two separate hyperbolic reflection features with different apex positions can be clearly seen on the radar profile. If the crack spacing is less than the resolution, only a single, wider, and stronger hyperbolic reflection feature can be seen on the radar profile. The lateral resolution can be replaced by the formula for the radius of the first Fresnel zone. The formula for the radius of the first Fresnel zone is: Ground penetrating radar images should first undergo a series of preprocessing steps (such as zero bias correction, filtering, and signal gain) and offset repositioning to improve image quality. Then, they should be detected and identified by a trained YOLO algorithm, spatially located, and related features such as amplitude extracted through a neural network.

[0008] 3. Based on the improved crack identification method, different types of cracks are classified into different levels: High-frequency pulsed electromagnetic waves emitted by the transmitting antenna of a ground-penetrating radar (GPR) towards the road structure are reflected within the structure and returned to the ground where they are received by the receiving antenna. When these electromagnetic waves pass through hidden defects, they undergo strong reflection. The depth of the defect can be calculated using its propagation time *t*. Therefore, in radar profiles obtained in areas with temperature anomalies, cracks typically appear as a hyperbolic reflection signal or a sudden discontinuation of a continuous linear reflection. By reading the two-way travel time of the reflected wave and combining it with the electromagnetic wave velocity in the underground medium, the depth of the reflection point (i.e., the crack) can be directly calculated. The formula for calculating the depth is: ; Where h: the depth of the target detected by the ground penetrating radar; c: the speed of electromagnetic waves in the air; v: the speed of electromagnetic wave propagation in the medium; x: the distance between the transmitting antenna and the receiving antenna; : Relative permittivity.

[0009] When an electromagnetic wave encounters a crack, the amplitude of its reflected signal is related to the crack's opening. Theoretically, the wider the crack, the stronger the reflected signal. That is, in a grayscale image, the wider the crack, the greater the color difference it appears to have with its surroundings. Therefore, it can be approximated by the simplest linear model, with the following formula: ; Where A: grayscale amplitude of the crack; w: physical width of the crack; k: calibration coefficient.

[0010] In the linear model above, background noise is considered. For example, on a white wall, the background is very bright (A0 value is high), and a black crack appears. Its amplitude A is the background brightness minus the darkness of the crack. Therefore, the optimized formula is: ; Where A: grayscale amplitude of the crack; w: physical width of the crack; k: calibration coefficient; A0: gray value of the region when there are no cracks in the image.

[0011] The crack depth and width are calculated using the amplitude data extracted from the neural network in the previous step and the formulas described above. Cracks are classified into levels based on both depth and width. By depth, three levels are defined: L1 (cracks within the surface layer), L2 (cracks at full depth in the surface layer), and L3 (cracks on the top surface of the base layer). By width, five levels are defined: W1 (micro-cracks), W2 (slightly open), W3 (moderately open), W4 (severely open), and W5 (fractured). Furthermore, through cracks are categorized based on the presence or absence of voids at the base: through-stable (no voids), through-activated (partial voids), and through-destructive (large-scale continuous voids). If radar images show through crack characteristics, the presence or absence of voids at the base of the crack is first determined, and different treatment measures are taken based on the void situation.

[0012] 4. Based on the proposed crack classification, establish a comprehensive crack classification decision matrix, propose treatment measures for cracks of different classifications, and establish maintenance records: Taking into account different crack depth levels, width levels, and the presence or absence of voids, corresponding measures are taken for cracks of different levels. For example, L1+W1+ cracks without voids are treated with preventive maintenance, while L2+W2+ cracks with voids need to be upgraded to L3+W2 and treated with base layer treatment. Based on this, a comprehensive crack level decision matrix is ​​established. During detection, information on the width level, depth level, and presence or absence of voids can be entered into the decision matrix to obtain the final comprehensive disease level and treatment recommendations. Specific engineering treatment measures are output based on the comprehensive rating results. If the estimated crack width level is W4 or W5 and it is located in a critical area, local destructive methods such as core drilling are required for verification, and timely treatment is necessary. After treating the cracks according to the treatment measures derived from the crack level decision matrix, if the radar image shows a continuous and uniform medium response at that location, the treatment effect is excellent. Therefore, ground-penetrating radar is used to re-examine the treated area to ensure the treatment effect. Finally, a maintenance record is established to document each treated crack, including its location, depth, grade, treatment method, and date, to facilitate subsequent tracking and effectiveness evaluation.

[0013] Compared with the prior art, the advantages of the present invention are as follows: 1. This invention uses infrared thermal imaging technology to detect and determine the presence of defects over a wide area. For areas with abnormal temperatures, ground-penetrating radar is used for precise positioning. Combined with radar grayscale images, the width and depth of cracks are estimated using depth formulas and optimized width formulas, and different levels are classified to establish a comprehensive crack level decision matrix. Based on the three-dimensional information of the cracks, differentiated treatment measures are obtained through the comprehensive crack level decision matrix. This method can help operators to rationally allocate maintenance resources, realize the transformation from "repairing when it breaks" to "proactive prevention and precise repair", and ultimately significantly reduce the road maintenance cost throughout the entire life cycle.

[0014] 2. This invention introduces infrared thermal imaging detection technology to assist ground-penetrating radar in flaw detection. When there are cracks inside the road, there are usually defects such as voids. Using infrared thermal imaging detection technology, the temperature information of the road surface can be observed, the abnormal temperature area of ​​the road can be identified, and the actual location corresponding to the abnormal surface temperature area can be marked. That is, there are hidden defects such as cracks and voids at that location. This can solve the problem of misjudgment and omission of internal damage due to the lack of ground-penetrating radar signal characteristics. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention.

[0016] Figure 2 This is a schematic diagram of a high-speed infrared imager mounted on a vehicle according to the present invention.

[0017] Figure 3 This is a schematic diagram of the ground-penetrating radar survey line layout in the temperature anomaly area of ​​the present invention.

[0018] Figure 4 This is a schematic diagram illustrating the working principle of the ground-penetrating radar of the present invention.

[0019] Figure 5 This is a radar image of a crack penetrating according to the present invention.

[0020] Figure 6 This is a radar image from the present invention.

[0021] Figure 7 This is a flowchart illustrating the process of dividing and treating through cracks according to the present invention.

[0022] Figure 8 This is a schematic diagram of the road section detected by the present invention.

[0023] Figure 9 This invention uses a 400MHz radar image of a road section with abnormal temperature to detect such areas.

[0024] Figure 10 This invention uses a 900MHz radar image of a road section with abnormal temperature to detect such areas. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0026] The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar includes the following steps: S1. Use a vehicle equipped with a high-speed infrared thermal imager to scan the road surface, obtain temperature images of the road surface, and preprocess the temperature images to identify areas with abnormal temperatures. S2. Ground penetrating radar is used to detect areas with abnormal temperatures to obtain radar waveform images, and the radar waveform images are preprocessed and offset-corrected. The processed radar waveform image is converted into a grayscale image, and the location is determined by the trained YOLO target detection model. The region with hyperbolic features is cropped out and the feature is accurately extracted by the neural network to obtain the grayscale amplitude of the hyperbola. S3. Based on the obtained hyperbola grayscale amplitude, calculate the depth and width of the crack, classify the cracks according to the crack depth and crack width, and observe whether there is any void through radar images. S4. Taking into account different depths, widths, and the presence or absence of voids in the cracks, a comprehensive crack level decision matrix is ​​established. The three-dimensional information of depth, width, and presence or absence of voids obtained in S3 is filled into the decision matrix to obtain the final comprehensive disease level and treatment recommendations. Based on the comprehensive rating results, specific engineering treatment measures are output. S5. After treating the cracks, conduct ground-penetrating radar re-examination to ensure the treatment effect. Finally, establish a maintenance file to record each treated crack.

[0027] The steps of a method for treating deep road defects based on a combination of infrared thermal imaging and ground-penetrating radar are described in detail below: 1. Infrared thermal imaging detection technology is used to assist ground-penetrating radar (GPR) in flaw detection, solving the problems of false positives and false negatives in GPR: Infrared thermal imaging detection technology has advantages such as being non-destructive, fast, and low-cost, and is suitable for large-scale detection. Its principle is that when there is a temperature difference between an object and the ambient temperature, heat will flow inside the object. If heat is injected into the object, some of the heat flow will inevitably diffuse inward, causing a change in the temperature distribution on the object's surface. If there are defects inside the object, a temperature difference will be formed between the defective and non-defective areas. Due to the existence of local temperature differences, there will inevitably be differences in infrared radiation intensity. Infrared thermal imagers can be used to detect the temperature changes and thus determine the condition of the defects.

[0028] Infrared thermal imaging detection technology, based on infrared thermal imaging theory and heat conduction theory, classifies objects into two categories according to their radiation characteristics: black bodies and gray bodies. The radiation emitted by the detected objects belongs to gray body radiation and satisfies the Stefan-Boltzmann equation: ; in Gray body emissivity; Stefan Boltzmann constant; W: Radiation intensity of the object; T: Absolute temperature of the object.

[0029] Heat is transferred from a hotter part of an object to a cooler part, or from a hotter object to another cooler object in contact with it. This heat transfer process is called heat conduction. The reason for heat conduction within an object is the temperature difference between its parts. Therefore, once the internal temperature field of the object is determined, the heat flow within the object can be determined according to Fourier's law. The formula is: ; Where q(r,t) is the heat flow per unit area per unit time in the direction of decreasing temperature; : Thermal conductivity of the object being measured; : Spatial and temporal temperature distribution within the object being measured.

[0030] Based on the above theory, Figure 2For example, firstly, infrared imaging survey lines are planned on the lane shown in the figure. Then, a vehicle equipped with a high-speed infrared thermal imager is used to scan the road surface at 60-80 km / h. After a series of processing steps such as noise reduction, the temperature image of the road surface is obtained. Due to the presence of defects and pores inside the road, the road material is discontinuous at the defect points, resulting in abrupt changes in the physical properties of the road at these points. Assuming the defect is an insulating type (reduced heat transfer coefficient at the defect point), the asphalt pavement will generate heat flow after a day of exposure to sunlight. The heat flow propagates uniformly from the surface to the interior in the intact road sections, but the thermal resistance is greater at the defect points, hindering heat propagation and causing heat to accumulate at the defect location, forming a "hot spot." If the defect is a conductive type (increased heat transfer coefficient at the defect point), a "cold spot" will form under the same conditions. Reflective cracks are often accompanied by voids in the base layer along the cracks. At night or in the early morning, these voided areas will appear as clear linear thermal anomalies on the infrared image due to differences in heat transfer, indicating the presence of hidden defects such as cracks in the temperature anomaly area.

[0031] 2. Ground-penetrating radar detection technology is used to identify abnormal areas in infrared images and accurately locate the cracks.

[0032] After the previous step, the temperature anomaly areas (insulation-type defects or thermal conductivity defects) obtained from the temperature image are areas with hidden defects such as cracks. For these road temperature anomaly areas, mutually perpendicular ground-penetrating radar (GPR) lines are laid out in these areas to prevent misjudgments caused by the crack development direction aligning with the GPR line direction. An example of the line layout is shown below. Figure 3 As shown, the spacing between adjacent survey lines is approximately 30cm. Then, using a high-frequency electromagnetic wave with a main frequency of 10M-1000M, ultra-high-frequency pulsed electromagnetic waves are transmitted into the ground via a transmitting antenna in a broadband pulse format. When these electromagnetic waves pass through the air and enter the road surface, they are reflected and refracted between layers of materials with different dielectric properties. They then penetrate the road surface, return to the ground, and are received by a receiving antenna. The specific principle is as follows... Figure 4 As shown, radar waveform images of underground targets are acquired through imaging analysis of the received wave field. These images are then preprocessed (DC removal, gain adjustment, filtering) and offset realignment processed to improve image quality. The radar waveform images are converted into grayscale images and then detected using a trained YOLO model for spatial localization. The YOLO-detected radar image is then cropped into a "region of interest" containing a relatively clean hyperbolic target. A large amount of irrelevant background noise is removed, and relevant features such as amplitude are extracted. A neural network is used for precise feature extraction to obtain the grayscale amplitude of the hyperbola. The advantage of neural network-based precise feature extraction is that it eliminates the need for manual feature design and selection, whereas traditional methods require domain experts to spend considerable time designing features based on prior knowledge. Neural networks can automatically learn the most effective feature representations directly from the raw data, greatly reducing the human and time costs of feature engineering and avoiding subjective biases introduced by manual feature design.

[0033] During precise detection using ground-penetrating radar, the lateral resolution is approximately equal to the radius of the first Fresnel zone. The radius formula is: ; Where R is the lateral resolution; d is the depth of the target volume; λ is the wavelength of the electromagnetic wave in the medium, and the formula for calculating the wavelength is velocity / frequency.

[0034] If the cracks in the temperature anomaly area are too close together, only a single, wider, and stronger hyperbolic reflection feature can be seen on the radar profile. Therefore, the main frequency needs to be changed at least twice during the detection process to change the lateral resolution and prevent cracks with a spacing smaller than the lateral resolution from being misjudged or missed.

[0035] 3. Based on the improved crack identification method, different types of cracks are classified into different levels.

[0036] By reading the two-way travel time of the reflected wave and combining it with the electromagnetic wave velocity of the underground medium, the depth of the reflection point (i.e., the crack) can be directly calculated. The formula for calculating the crack depth is: ; Where h: the depth of the target detected by the ground penetrating radar; c: the speed of electromagnetic waves in the air; v: the speed of electromagnetic wave propagation in the medium; x: the distance between the transmitting antenna and the receiving antenna; The relative permittivity of road materials is shown in Table 1. Table 1 Relative Permittivity of Road Surface Materials

[0037] When electromagnetic waves encounter a crack, a typical hyperbolic diffracted wave is generated at the crack tip. The wider and more obvious the crack, the stronger and easier to identify the diffracted signal. Background noise is introduced into the simple linear model of crack depth to optimize the crack depth calculation formula. This is because background noise is an inherent and unavoidable part of the image imaging system. It is determined by factors such as the material's color / reflectivity, overall lighting conditions, camera exposure gain, and radar system baseline noise. Therefore, background noise is introduced to more realistically reflect the physical world and the imaging system, and to extract the "pure signal" that is only related to the crack width from the "raw reading" that includes the system background.

[0038] Therefore, the formula for calculating the crack width is: ; in The physical width of the crack; : Gray-scale amplitude of the crack; : Base grayscale value of the background; k: Calibration coefficient.

[0039] In the optimized crack depth calculation formula, the calibration coefficient k is calculated by knowing the true width w of a crack in advance and measuring the amplitude of the crack in the image. Substitute into the formula The calibration coefficient k can then be obtained. This patent can use road surface cracks to obtain the calibration coefficient k. First, the true width w of the ground crack is measured, and then the ground penetrating radar is used to scan the ground crack to finally obtain a ground penetrating radar grayscale image. The grayscale amplitude of the ground crack is obtained by detecting, identifying, and extracting amplitude features through YOLO. Substituting these values ​​into the formula, we can obtain the calibration coefficient k.

[0040] Cracks are classified into grades based on both depth and width. Cracks by depth are divided into three grades: L1 (cracks within the surface layer), L2 (cracks at full depth in the surface layer), and L3 (cracks on the top surface of the base layer). Cracks by width are divided into five grades: W1 (micro-cracks), W2 (slightly open), W3 (moderately open), W4 (severely open), and W5 (fractured). The specific methods for classifying crack grades based on crack depth, width, damage extent, ground-penetrating radar image characteristics, and corresponding structural damage are shown in Tables 2 and 3. If the base layer radar image shows... Figure 5 The characteristic of a penetrating crack shown is that, in radar imagery, it appears as a sharp protrusion extending upwards from the base layer to the surface layer, interrupting in the longitudinal section. In this case, it is necessary to observe whether the ground-penetrating radar image shows any similar features. Figure 6 The voiding characteristics shown are characterized by the surface and base layers being significantly brighter than the surrounding areas, as indicated by the graph. These include black-white-black (positive reflectance, voids are air) and white-black-white (negative reflectance, voids are water). Based on the internal voiding, the through cracks are categorized into three types: through-stable (no voids), through-activated (local voids), and through-destructive (large-scale continuous voids). Further classification is as follows: Figure 7 The treatment process for penetrating cracks is shown below.

[0041] Table 2 Classification of Crack Depth

[0042] Table 3 Crack Width Classification Table

[0043] Based on the proposed crack levels, a comprehensive crack level decision matrix is ​​established, treatment measures for cracks of different levels are proposed, and maintenance records are established.

[0044] During inspection and maintenance, decisions cannot be made solely based on crack depth or width grades. Therefore, a comprehensive crack grade decision matrix is ​​established, taking into account both crack depth and width grades, as well as the presence of voids at the base of the crack. This matrix allows for rapid determination of the crack grade and corresponding treatment measures. If voids are present at the base of the crack, the treatment method must be upgraded. Throughout the process, three-dimensional information about the crack—crack depth, crack width, and the presence of voids—is obtained and input into the comprehensive crack grade decision matrix. Finally, the comprehensive crack grade and specific treatment recommendations are obtained. Table 4 shows the distinct treatment measures for different comprehensive crack grades. For cracks with width grades W3, W4, and W5, and depth grades L2 and L3, if voids are present at the base, the subgrade must be treated. After treating the cracks, ground-penetrating radar is used for re-examination to ensure the effectiveness of the treatment. Finally, for each crack treated, the treated crack should be recorded as shown in Table 5, including location, depth, grade, treatment method and date, to facilitate subsequent tracking and effect evaluation.

[0045] Table 4 Crack Depth and Width Decision Matrix

[0046] Table 5 Crack Maintenance Record Form

[0047] Case Study: Figure 5 This is a two-way single-lane road in a certain county. A section of this road was selected for crack detection, with four survey lines (L1, L2, L3, and L4) laid out. The detailed survey line distances are shown in Table 6. Due to the short distance of each survey line, the speed of the vehicle equipped with the infrared thermal imager was appropriately reduced. The L3 section was selected for analysis. For the temperature anomaly area in the L3 section, ground-penetrating radar (GPR) was used for precise flaw detection. Since different frequency antennas of GPR have different depth-finding capabilities, lower frequencies require larger antenna sizes and greater detection depth, but the resolution decreases; higher frequencies require smaller antenna sizes, shallower detection depths, and higher resolution. Lateral resolution is also affected; therefore, the GPR's main frequency needs to be changed. In this example, the main frequencies of the electromagnetic waves emitted by the GPR are 400MHz and 900MHz.

[0048] First, measure the actual width w of a surface crack, then scan the crack with ground penetrating radar, and then use image software to read the hyperbolic characteristic grayscale amplitude A. Use the formula: k=A / w to measure the calibration coefficient k.

[0049] A specific analysis was conducted on section L3 of the survey line, between 120m and 150m. Infrared imaging was used to scan the road surface, revealing a temperature anomaly area at 140m. This anomaly area was then analyzed as follows: Figure 4 The diagram shows mutually perpendicular survey lines, approximately 30cm apart. Ground-penetrating radar was then used at 400MHz and 900MHz frequencies to measure along these lines. A radar image of one particular survey line is shown below. Figure 9 As shown, the left side is a radar image of the transmitting antenna emitting electromagnetic waves at a main frequency of 900 MHz, while the right side is a radar image of the transmitting antenna emitting electromagnetic waves at a main frequency of 400 MHz.

[0050] The original radar waveform image was processed by DC removal, gain adjustment, filtering, and offset repositioning to convert it into a grayscale image. Then, a trained YOLO model was used to detect the specific location of the hyperbolic target. There were two locations with hyperbolic features below the L3 survey line at a distance of 140m, at a distance of 137.96m (crack 1) and 140.23m (crack 2) from the L3 survey line. At the same time, the image containing only hyperbolic features was segmented, and the grayscale amplitude of the hyperbola vertex and the background grayscale value were accurately extracted by the neural network.

[0051] Calculate the depth and width of a crack using the following crack depth and width formulas.

[0052] ; By observing the presence of voids in radar images, the presence of voids at the bottom of cracks is determined and entered into the crack level decision matrix to obtain the crack level. The specific three-dimensional information of cracks 1 and 2, the comprehensive level obtained through the crack comprehensive level matrix, and treatment recommendations are shown in Table 7. Since crack 1 has voids at the bottom, it is upgraded to L3+W2 and treatment measures are taken. After treating both cracks, ground-penetrating radar is used for re-examination. Comparing the ground-penetrating radar images before and after treatment, if the image at the crack location shows a continuous and uniform medium effect, the crack treatment is considered complete. Finally, the maintenance records for both cracks are recorded according to the records shown in Table 3, including location, depth, level, treatment method, and date, for subsequent tracking and effect evaluation. Since infrared thermal imaging detection is less expensive and ground-penetrating radar does not require large-scale detection, this method reduces the road's operating costs.

[0053] Table 6. Statistics on Road Flaw Detection Workload in a Certain County

[0054] Table 7. Three-dimensional information and treatment plan for cracks 1 and 2

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for treating deep road defects based on a combination of infrared thermal imaging and ground-penetrating radar, characterized in that, Includes the following steps: S1. Use a vehicle equipped with a high-speed infrared thermal imager to scan the road surface, obtain temperature images of the road surface, and preprocess the temperature images to identify areas with abnormal temperatures. S2. Ground penetrating radar is used to detect areas with abnormal temperatures to obtain radar waveform images, and the radar waveform images are preprocessed and offset-corrected. The processed radar waveform image is converted into a grayscale image, and the location is determined by the trained YOLO target detection model. The region with hyperbolic features is cropped out and the feature is accurately extracted by the neural network to obtain the grayscale amplitude of the hyperbola. S3. Based on the obtained hyperbola grayscale amplitude, calculate the depth and width of the crack, classify the cracks according to the crack depth and crack width, and observe whether there is any void through radar images. S4. Taking into account different depths, widths, and the presence or absence of voids in the cracks, a comprehensive crack level decision matrix is ​​established. The three-dimensional information of depth, width, and presence or absence of voids obtained in S3 is filled into the decision matrix to obtain the final comprehensive disease level and treatment recommendations. Based on the comprehensive rating results, specific engineering treatment measures are output. S5. After treating the cracks, conduct ground-penetrating radar re-examination to ensure the treatment effect. Finally, establish a maintenance file to record each treated crack.

2. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S1, a high-speed infrared thermal imager mounted on a vehicle scans the road surface at a speed of 60-80 km / h.

3. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S1, the temperature image preprocessing includes noise reduction, filtering, and contrast enhancement.

4. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S2, mutually perpendicular ground-penetrating radar lines are arranged in the temperature anomaly area, with an interval of approximately 30 cm between adjacent lines.

5. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 4, characterized in that, The same detection area is detected at least twice by changing the main frequency of the detection radar. The main frequency range of the detection radar is 10M-1000M.

6. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S2, the preprocessing of radar waveform images includes DC removal, gain processing, and filtering.

7. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S3, the formula for calculating crack depth is: ； in h Ground-penetrating radar detects the depth of a target; c The speed at which electromagnetic waves propagate in the air; v The speed at which electromagnetic waves propagate in a medium; x The distance between the transmitting antenna and the receiving antenna; : Relative permittivity of road materials.

8. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S3, the formula for calculating the crack width is: ； Where w: the physical width of the crack; A: Gray-scale amplitude of the crack; A0: Gray-scale value of the region when there is no crack in the image; k: Calibration coefficient.

9. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S3, cracks are classified into three levels according to their depth: L1 is cracks within the surface layer, L2 is cracks at the full depth of the surface layer, and L3 is cracks on the top surface of the base layer. Cracks are classified into five levels according to their width: W1 is micro-crack, W2 is slightly open, W3 is moderately open, W4 is severely open, and W5 is broken. Based on whether there is void at the bottom of the through crack: through-stable type without void, through-activated type with partial void, and through-destructive type with large-scale continuous void.

10. The method for treating deep road defects based on the combination of infrared thermal imaging and ground-penetrating radar according to claim 1, characterized in that, In S3, the maintenance record includes information such as crack location, crack depth, crack width, crack grade, treatment method, and treatment date.

Images