A pavement performance evaluation method and system based on joint judgment

Abstract

This application relates to the technical field of pavement evaluation, and discloses a pavement performance evaluation method and system based on joint judgment. The method includes collecting multi-profile image data of the pavement based on corresponding collection strategies, generating a base crack image dataset, making independent judgments, and making joint judgments based on the independent judgments. The system corresponds to the method. In this application, by selecting a collection strategy according to pavement basic information and signal quality, more practical pavement data is obtained; the data enhancement technology increases the diversity and reliability of the data, making the generated base crack image dataset more representative; and the recognition and joint judgment mechanism based on the YOLOv5 series model effectively improves the recognition accuracy of base cracks and reduces misjudgments; thereby comprehensively improving the accuracy of pavement performance evaluation, providing a scientific basis for road maintenance and repair, extending the service life of the road, and ensuring traffic safety.

Description

Technical Field

[0001] The present application relates to the technical field of pavement evaluation, and specifically to a pavement performance evaluation method and system based on joint judgment. Background Art

[0002] With the acceleration of the urbanization process, the scale of road construction has been continuously expanding, and it has become crucial to accurately and efficiently evaluate pavement performance. By accurately grasping the pavement performance status, potential diseases can be detected in a timely manner, maintenance work can be reasonably arranged, the service life of the road can be extended, and traffic safety can be ensured.

[0003] Currently, pavement performance evaluation technologies mainly rely on manual inspection and traditional non-destructive testing technologies. Manual inspection has low efficiency, strong subjectivity, and it is difficult to detect hidden diseases inside the pavement. Traditional non-destructive testing technologies, such as ground penetrating radar (GPR), although they can obtain information on the internal structure of the pavement, have many problems in data processing and disease identification. In existing research, the processing of pavement GPR image data mainly relies on professional personnel to interpret frame by frame, which is not only time-consuming and laborious, but also difficult to guarantee accuracy, restricting the large-scale application of radar technology. At the same time, in the existing deep learning algorithms for disease identification in GPR images, affected by factors such as similar echo characteristics of different diseases, non-uniformity of pavement materials, and electromagnetic noise interference, the recognition accuracy is limited.

[0004] Chinese Patent with Application No. CN119314021A discloses a study on the automatic recognition of GPR echo feature images of semi-rigid base cracks based on deep learning. The selected YOLOv5 series models have weak small target detection capabilities, are prone to false detections, and have insufficient pertinence in feature extraction and fusion. Further, in terms of model deployment, it is only tested on workstations and does not consider the real-time detection requirements of mobile devices. These deficiencies make it difficult for this technology to meet the requirements of accuracy and efficiency in actual pavement performance evaluation and urgently need improvement.

[0005] In summary, there is an urgent need for a new technical solution for pavement performance evaluation based on joint judgment to solve the above technical problems. Summary of the Invention

[0006] The purpose of the present application is to provide a pavement performance evaluation method and system based on joint judgment to solve the technical problems raised in the above background art.

[0007] To achieve the above purpose, the present application discloses the following technical solutions:

[0008] In the first aspect, the present application discloses a pavement performance evaluation method based on joint judgment, and the method includes the following steps:

[0009] S1: Use the three-dimensional ground penetrating radar technology to adopt different acquisition strategies for different road surfaces, and collect multi-profile image data of the road surface based on the corresponding acquisition strategies. Among them, the adoption of the acquisition strategy is selected based on the road surface basic information and the road surface signal quality. The multi-profile image data includes longitudinal section image data and horizontal section image data.

[0010] S2: Use the data enhancement technology, combined with the acquisition strategy, to process the longitudinal section image data and the horizontal section image data to generate a base crack image dataset. Among them, combining the acquisition strategy is to perform enhancement based on the data characteristics on the data obtained by processing the longitudinal section image data and the horizontal section image data using the data enhancement technology, based on the data characteristics corresponding to the acquisition strategy.

[0011] S3: Use the recognition model pre-constructed based on the YOLOv5 series models to independently judge the base crack image dataset, and perform a joint judgment based on the independent judgment. Among them, the independent judgment is to judge the base cracks in the longitudinal section image data or the horizontal section image data.

[0012] It should be noted that in this embodiment, the existing YOLOv5 series models are used to construct the recognition model.

[0013] Preferably, the use of the three-dimensional ground penetrating radar technology to adopt different acquisition strategies for different road surfaces includes:

[0014] The road surface basic information includes the road surface material type and the structural layer thickness. The road surface signal quality includes the signal strength and the signal-to-noise ratio. Different combinations of acquisition parameters are selected as the acquisition strategy based on the road surface material type, the structural layer thickness, the signal strength, and the signal-to-noise ratio. The acquisition parameters include the antenna center frequency, the two-way travel time, the dwell time, and the sampling interval, and the acquisition parameters correspond to the road surface basic information and the road surface signal quality. Among them, when selecting the acquisition strategy, the road surface evaluation is calculated using the road surface evaluation formula, and the corresponding acquisition parameters are matched based on the road surface evaluation. The road surface evaluation formula is:

[0015]

[0016] Among them, is the comprehensive change rate of the road surface material type and the structural layer thickness, is the change rate of the signal strength, is the change rate of the signal-to-noise ratio, is the calculated road surface evaluation.

[0017] Preferably, the collection of the multi-profile image data of the road surface based on the corresponding acquisition strategy includes:

[0018] Calibrate the three-dimensional ground penetrating radar equipment before collection;

[0019] After setting the acquisition parameters based on the selected acquisition strategy, the three-dimensional ground penetrating radar travels along the road surface at a constant speed, synchronously acquiring longitudinal section image data and horizontal section image data;

[0020] During the acquisition process, the high-precision positioning system is used to record the position information of each acquisition point.

[0021] Preferably, using the data enhancement technology and combining with the acquisition strategy to process the longitudinal section image data and the horizontal section image data includes:

[0022] Using the data enhancement technology to process the longitudinal section image data and the horizontal section image data to generate the first enhanced data, and this processing is carried out based on the following steps:

[0023] Adopting the GPR numerical simulation technology to generate the simulated images of the base course cracks with different depths, sizes and moisture states, and fusing them with the original longitudinal section image data and horizontal section image data;

[0024] Adjust the gain of the image to enhance the contrast and brightness of the image;

[0025] Perform a scaling operation on the image to change the size and proportion of the image;

[0026] Introduce simulated interference signals and add random noise to simulate the interference situations that may be encountered in the actual acquisition process.

[0027] Preferably, using the data enhancement technology and combining with the acquisition strategy to process the longitudinal section image data and the horizontal section image data further includes:

[0028] Processing the first enhanced data based on the acquisition strategy to generate the second enhanced data, and this processing is carried out based on the following steps:

[0029] When the data feature corresponding to the acquisition strategy is that the resolution of the image data does not meet the preset resolution threshold, perform super-resolution reconstruction processing on the first enhanced data;

[0030] When the data feature corresponding to the acquisition strategy is that the noise of the image data does not meet the preset noise threshold, perform noise reduction processing on the first enhanced data;

[0031] When the data feature corresponding to the acquisition strategy is that the feature distribution of the image data does not meet the preset feature distribution threshold, perform histogram equalization on the first enhanced data to adjust the feature distribution of the image.

[0032] Preferably, generating the base course crack image dataset includes:

[0033] Sort and label the second enhanced data, where the label is the position, size, and category of the base layer cracks in the image;

[0034] Divide the labeled image data into a training set, a validation set, and a test set based on a preset ratio; among them: the training set is used to train the recognition model, the validation set is used to evaluate the recognition model and adjust hyperparameters during the training process, and the test set is used to finally evaluate the accuracy and generalization ability of the recognition model.

[0035] Preferably, using the recognition model pre-constructed based on the YOLOv5 series of models to independently judge the base layer crack image data set, including:

[0036] Use the recognition model to independently judge the longitudinal section image data in the base layer crack image data set, and this independent judgment is carried out based on the following steps:

[0037] Input the longitudinal section image data into the recognition model;

[0038] Preprocess the longitudinal section image data, adjust the image size and maintain the original aspect ratio;

[0039] Use a multi-layer convolutional neural network to extract multi-dimensional features of the base layer cracks in the image;

[0040] Based on bounding box regression, class prediction, and confidence prediction, determine the position, size, and category of the base layer cracks in the image;

[0041] Adopt non-maximum suppression processing to remove overlapping prediction boxes and output the results of independent judgment.

[0042] Preferably, using the recognition model pre-constructed based on the YOLOv5 series of models to independently judge the base layer crack image data set, further includes:

[0043] Use the recognition model to independently judge the horizontal section image data in the base layer crack image data set, and this independent judgment is carried out based on the following steps:

[0044] Input the horizontal section image data into the recognition model;

[0045] Preprocess the horizontal section image data, adjust the image size and maintain the original aspect ratio;

[0046] Use a multi-layer convolutional neural network to extract multi-dimensional features of the base layer cracks in the image;

[0047] Based on bounding box regression, class prediction, and confidence prediction, determine the position, size, and category of the base layer cracks in the image;

[0048] Non - maximum suppression processing is adopted to remove overlapping prediction boxes and output the results of independent judgment.

[0049] Preferably, the joint judgment is based on independent judgment and includes:

[0050] Compare the independent judgment results of the longitudinal section image data and the horizontal section image data;

[0051] When the crack echo characteristics at the same position are determined to be base course cracks in both the recognition results of the longitudinal section image data and the horizontal section image data, it is considered that there is a base course crack at that position;

[0052] When it is determined to be a base course crack only in the recognition result of the image data of a certain section, the image data at that position needs to be manually re - checked;

[0053] Based on the results of the joint judgment, comprehensively evaluate the pavement performance to determine the type, location, and severity of pavement diseases.

[0054] In a second aspect, the present application discloses a pavement performance evaluation system based on joint judgment. This system is applicable to the pavement performance evaluation method based on joint judgment as described above. The system includes:

[0055] A dynamic acquisition module, which is used to use three - dimensional ground penetrating radar technology to adopt different acquisition strategies for different pavements and collect multi - profile image data of the pavement based on the corresponding acquisition strategies; among them, the adoption of the acquisition strategy is selected based on the pavement basic information and the pavement signal quality, and the multi - profile image data includes longitudinal section image data and horizontal section image data;

[0056] An image enhancement module, which is used to use data enhancement technology and combine the acquisition strategy to process the longitudinal section image data and the horizontal section image data to generate a base course crack image data set; among them, combining the acquisition strategy is to perform enhancement based on data characteristics on the data obtained by processing the longitudinal section image data and the horizontal section image data using data enhancement technology according to the data characteristics corresponding to the acquisition strategy;

[0057] A joint judgment module, which is used to use an identification model pre - constructed based on the YOLOv5 series model to perform independent judgment on the base course crack image data set and perform joint judgment based on the independent judgment; among them, the independent judgment is to judge the base course cracks for the longitudinal section image data or the horizontal section image data.

[0058] Beneficial effects: The road surface performance evaluation method and system based on joint judgment of the present application obtain more practical road surface data by selecting a collection strategy based on road surface basic information and signal quality; the data enhancement technology increases the diversity and reliability of the data, making the generated base crack image dataset more representative; and the recognition and joint judgment mechanism based on the YOLOv5 series model effectively improves the recognition accuracy of base cracks and reduces misjudgment; thus comprehensively improving the accuracy of road surface performance evaluation, providing a scientific basis for road maintenance and repair, extending the service life of the road, and ensuring traffic safety. Description of the Drawings

[0059] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0060] Figure 1 It is a flowchart of the road surface performance evaluation method based on joint judgment provided by the embodiment of the present application;

[0061] Figure 2 It is a structural block diagram of the road surface performance evaluation system based on joint judgment provided by the embodiment of the present application. Detailed Embodiments

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0063] In this article, the term "including" is intended to cover non-exclusive inclusion, so that a process, method, article, or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such a process, method, article, or device. Without further limitations, the elements defined by the statement "including..." do not exclude the existence of additional identical elements in the process, method, article, or device including the said elements.

[0064] The first aspect of this embodiment discloses a Figure 1 road surface performance evaluation method based on joint judgment as shown in the following, and this method includes the following steps:

[0065] S1: Use the three-dimensional ground-penetrating radar technology to adopt different acquisition strategies for different road surfaces, and collect multi-profile image data of the road surface based on the corresponding acquisition strategies. Among them, the adoption of the acquisition strategy is selected based on the road surface basic information and the road surface signal quality. The multi-profile image data includes longitudinal section image data and horizontal section image data.

[0066] S2: Use the data augmentation technology to process the longitudinal section image data and the horizontal section image data in combination with the acquisition strategy to generate a base layer crack image dataset. Among them, combining the acquisition strategy is to perform enhancement based on the data characteristics on the data obtained by processing the longitudinal section image data and the horizontal section image data using the data augmentation technology based on the data characteristics corresponding to the acquisition strategy.

[0067] S3: Use the recognition model pre-constructed based on the YOLOv5 series models to independently judge the base layer crack image dataset, and perform joint judgment based on the independent judgment. Among them, the independent judgment is to judge the base layer cracks for the longitudinal section image data or the horizontal section image data.

[0068] Through the above, in this embodiment, the three-dimensional ground-penetrating radar technology is used to formulate different acquisition strategies for different road surfaces, the collected multi-profile image data is processed in combination with the data augmentation technology, and independent judgment and joint judgment are carried out with the help of the recognition model pre-constructed based on the YOLOv5 series models, realizing a more accurate and efficient evaluation of the road surface performance. Among them, by selecting the acquisition strategy according to the road surface basic information and signal quality, more practical road surface data is obtained. The data augmentation technology increases the diversity and reliability of the data, making the generated base layer crack image dataset more representative. The recognition and joint judgment mechanism based on the YOLOv5 series models effectively improves the recognition accuracy of the base layer cracks and reduces misjudgment. Combining these technical means comprehensively improves the accuracy of the road surface performance evaluation, provides a scientific basis for road maintenance and repair, extends the road service life, and ensures traffic safety.

[0069] Specifically, using the three-dimensional ground-penetrating radar technology to adopt different acquisition strategies for different road surfaces includes:

[0070] The road surface basic information includes the road surface material type and the structural layer thickness. The road surface signal quality includes the signal strength and the signal-to-noise ratio. Different combinations of acquisition parameters are selected as the acquisition strategy based on the road surface material type, the structural layer thickness, the signal strength, and the signal-to-noise ratio. The acquisition parameters include the antenna center frequency, the two-way travel time, the dwell time, and the sampling interval, and the acquisition parameters correspond to the road surface basic information and the road surface signal quality. Among them, when selecting the acquisition strategy, the road surface evaluation is calculated using the road surface evaluation formula, and the corresponding acquisition parameters are matched based on the road surface evaluation. The road surface evaluation formula is:

[0071]

[0072] Among them, is the comprehensive change rate of pavement material type and structural layer thickness, is the change rate of signal strength, is the change rate of signal-to-noise ratio, is the calculated pavement evaluation.

[0073] It can be understood that in practical applications, the change of acquisition parameters occurs when the pavement changes. Therefore, the analysis of pavement change is more valuable than directly analyzing the pavement. The pavement material type and the thickness of the structural layer are the inherent properties of the pavement, and their changes will affect signal propagation. In this embodiment, the comprehensive change rate is obtained by weighted summation of the change rate of the pavement material type and the change rate of the structural layer thickness. The change rate of the pavement material type is the change of the pavement material with respect to time, and any one of the existing technologies can be used to obtain the corresponding data. The change rate of the structural layer thickness is the change of the structural layer thickness with respect to time, and any one of the existing technologies can be used to obtain the corresponding data; the signal strength and the signal-to-noise ratio directly reflect the quality of the acquired signal and are related to the accurate judgment of the pavement condition; these factors jointly determine the pavement evaluation. Further, based on weight assignment and fusion, the changes in different pavement evaluations brought about by different changes are highlighted. Specifically: It shows that when the comprehensive change rate of pavement material type and structural layer thickness is small, the weight of the signal strength change rate in the pavement evaluation is relatively large. Because at this time, the pavement structure is relatively stable, and the signal strength is more critical for the evaluation. Complementary to the previous part, when the comprehensive change rate of pavement material type and structural layer thickness is large, it can reflect the combined effect of these two factors and the signal strength change rate on the pavement evaluation. And Added separately as an item to the formula, it shows that the signal-to-noise ratio change rate plays an independent and important role in the pavement evaluation, and its change will directly affect the final pavement evaluation result. Based on this design, the finally obtained pavement evaluation can comprehensively reflect the overall situation of pavement basic information and signal quality. When the change rates of each factor are in different states, the pavement evaluation will change accordingly, and then the appropriate acquisition parameters can be matched according to this value, providing a basis for the data acquisition of subsequent pavement performance evaluation.

[0074] Through the above, in this embodiment, different combinations of acquisition parameters are selected as the acquisition strategy by using multiple factors such as pavement material type, structural layer thickness, signal strength, and signal-to-noise ratio, realizing the refinement of the acquisition strategy. In practical applications, since the pavement materials and the thickness of the structural layer are different, the electromagnetic wave propagation characteristics are different. By selecting appropriate parameters such as the center frequency of the antenna, the collected signals are clearer and more accurate. The signal strength and signal-to-noise ratio reflect the quality of the acquisition environment. Based on these, the parameters are adjusted to improve the data quality. In a simple example, for a pavement with uniform pavement materials, a relatively thin structural layer, and good signal quality, a higher antenna center frequency, a shorter two-way travel time, a shorter dwell time, and a smaller sampling interval are selected; for a pavement with complex pavement materials, a relatively thick structural layer, and poor signal quality, a lower antenna center frequency, a longer two-way travel time, a longer dwell time, and a larger sampling interval are selected. This acquisition strategy customized according to multiple factors effectively improves the quality and accuracy of the collected data, provides a reliable data basis for subsequent pavement performance evaluation, and enhances the credibility of the evaluation results.

[0075] Specifically, collecting multi-profile image data of the pavement based on the corresponding acquisition strategy includes:

[0076] Calibrating the three-dimensional ground penetrating radar equipment before acquisition;

[0077] After setting the acquisition parameters based on the selected acquisition strategy, making the three-dimensional ground penetrating radar travel along the pavement at a constant speed, and synchronously collecting longitudinal section image data and horizontal section image data;

[0078] During the acquisition process, using a high-precision positioning system to record the position information of each acquisition point.

[0079] Through the above, this embodiment uses GPR numerical simulation technology to fuse images, gain adjustment, scaling operations, and introduce interference signals and noises, realizing the preliminary enhancement of data. GPR numerical simulation generates images of base course cracks in different states and fuses them with the original data, enriching the data samples and increasing the diversity of crack characteristics. Gain adjustment highlights the crack details in the image, facilitating subsequent identification. Scaling operations simulate the image effects at different detection distances and angles, improving the adaptability of the model to cracks of different scales. Introducing interference and noises simulates the actual detection environment and enhances the anti-interference ability of the model. Thereby, the stability of crack identification in a complex electromagnetic environment is improved, the practicality of the data and the reliability of model training are enhanced, and it helps to improve the accuracy of pavement performance evaluation.

[0080] Specifically, using data enhancement technology and combining with the acquisition strategy to process the longitudinal section image data and the horizontal section image data further includes:

[0081] Processing the first enhanced data based on the acquisition strategy to generate the second enhanced data, and this processing is carried out based on the following steps:

[0082] When the data feature corresponding to the acquisition strategy is that the resolution of the image data does not meet the preset resolution threshold, perform super-resolution reconstruction processing on the first enhanced data;

[0083] When the data feature corresponding to the acquisition strategy is that the noise of the image data does not meet the preset noise threshold, perform noise reduction processing on the first enhanced data;

[0084] When the data feature corresponding to the acquisition strategy is that the feature distribution of the image data does not meet the preset feature distribution threshold, perform histogram equalization on the first enhanced data to adjust the feature distribution of the image.

[0085] By the above, in this embodiment, according to the data features corresponding to the acquisition strategy, super-resolution reconstruction, noise reduction processing, and histogram equalization are performed on the first enhanced data, realizing secondary enhancement of the data. In this embodiment, super-resolution reconstruction, noise reduction processing, and histogram equalization are all any one of the existing technologies. Based on super-resolution reconstruction, when the image resolution is insufficient, super-resolution reconstruction improves the image clarity, making the crack details more clearly distinguishable, which helps the model accurately identify the crack features. Facing the noise problem, noise reduction processing removes the interference signal, improves the image quality, and avoids the influence of noise on crack identification. Histogram equalization adjusts the feature distribution, making the overall contrast of the image more reasonable and enhancing the distinction between the crack and the background. For example, for an image with uneven feature distribution caused by acquisition parameters, after processing, the model is more likely to identify the crack. These operations further optimize the data, improve the quality of the grass-roots crack image dataset, and enhance the accuracy of pavement performance evaluation.

[0086] Specifically, generating a grass-roots crack image dataset includes:

[0087] Sort and label the second enhanced data, where the label is the position, size, and category of the grass-roots cracks in the image;

[0088] Based on a preset ratio, divide the labeled image data into a training set, a validation set, and a test set; where: the training set is used to train the recognition model, the validation set is used to evaluate the recognition model and adjust the hyperparameters during the training process, and the test set is used to finally evaluate the accuracy and generalization ability of the recognition model.

[0089] With the above, in this embodiment, by sorting and annotating the second enhanced data and dividing it into a training set, a validation set, and a test set according to a preset ratio, the standardized management of the grass-roots crack image dataset is achieved. In a simple example, the preset ratio is 7:1:2. The position, size, and category of the grass-roots cracks in the image provide clear learning objectives for model training, enabling the model to learn the characteristics and patterns of the cracks. Reasonably dividing the dataset, the training set is used for model learning, the validation set is used to adjust hyperparameters to prevent overfitting, and the test set is used to evaluate the model performance to ensure that the model is effectively trained and evaluated at different stages. For example, during the training process, the model is continuously optimized through the validation set, making the model perform better on the test set. This standardized management improves the efficiency and accuracy of model training and lays a foundation for accurately evaluating the pavement performance.

[0090] Specifically, using the recognition model pre-constructed based on the YOLOv5 series of models to independently judge the grass-roots crack image dataset, including:

[0091] Using the recognition model to independently judge the longitudinal section image data in the grass-roots crack image dataset, and this independent judgment is carried out based on the following steps:

[0092] Input the longitudinal section image data into the recognition model;

[0093] Preprocess the longitudinal section image data, adjust the image size and maintain the original aspect ratio;

[0094] Use a multi-layer convolutional neural network to extract multi-dimensional features of the grass-roots cracks in the image;

[0095] Based on bounding box regression, class prediction, and confidence prediction, determine the position, size, and category of the grass-roots cracks in the image;

[0096] Adopt non-maximum suppression processing to remove overlapping prediction boxes and output the results of independent judgment.

[0097] Specifically, using the recognition model pre-constructed based on the YOLOv5 series of models to independently judge the grass-roots crack image dataset, also includes:

[0098] Using the recognition model to independently judge the horizontal section image data in the grass-roots crack image dataset, and this independent judgment is carried out based on the following steps:

[0099] Input the horizontal section image data into the recognition model;

[0100] Preprocess the horizontal section image data, adjust the image size and maintain the original aspect ratio;

[0101] Use a multi-layer convolutional neural network to extract multi-dimensional features of the grass-roots cracks in the image;

[0102] Based on bounding box regression, class prediction, and confidence prediction, determine the location, size, and class of the base course cracks in the image;

[0103] Adopt non-maximum suppression processing to remove overlapping prediction boxes and output the results of independent judgments.

[0104] Through the above, this embodiment uses the recognition model to perform input, preprocessing, feature extraction, prediction, and non-maximum suppression processing on the longitudinal section image data and horizontal section image data, and realizes the accurate recognition of the base course cracks in the longitudinal section and horizontal section.

[0105] Specifically, and based on independent judgments, perform joint judgments, including:

[0106] Compare the independent judgment results of the longitudinal section image data and the horizontal section image data;

[0107] When the crack echo characteristics at the same position are determined to be base course cracks in both the recognition results of the longitudinal section image data and the horizontal section image data, it is determined that there are base course cracks at that position;

[0108] When it is determined to be a base course crack only in the recognition result of the image data of a certain section, the image data at that position needs to be manually reviewed;

[0109] Based on the results of the joint judgment, comprehensively evaluate the pavement performance, and determine the type, location, and severity of the pavement diseases.

[0110] Through the above, this embodiment compares the independent judgment results of the longitudinal section and horizontal section image data, performs joint judgments and pavement performance evaluations based on the judgment results, and realizes a more accurate pavement disease assessment. Determining the type, location, and severity of pavement diseases based on the results of joint judgments can comprehensively evaluate the pavement performance. For example, judging the severity of diseases according to characteristics such as the width and length of cracks, providing accurate information for road maintenance, guiding the rational arrangement of maintenance work, improving the pertinence and effectiveness of road maintenance, and ensuring the safe use of roads.

[0111] The second aspect of this embodiment discloses a pavement performance evaluation system based on joint judgment as Figure 2 shown, which is applicable to the pavement performance evaluation method based on joint judgment as described above. This system includes:

[0112] A dynamic acquisition module, which is used to use three-dimensional ground penetrating radar technology to adopt different acquisition strategies for different pavements and acquire multi-section image data of the pavement based on the corresponding acquisition strategies; among them, the adoption of the acquisition strategy is selected based on the pavement basic information and the pavement signal quality, and the multi-section image data includes longitudinal section image data and horizontal section image data;

[0113] An image enhancement module, which is used to utilize data enhancement technology and combine with a collection strategy to process longitudinal section image data and horizontal section image data, and generate a dataset of base layer crack images; among them, combining with the collection strategy is to perform enhancement based on data characteristics on the data obtained by processing the longitudinal section image data and horizontal section image data based on data enhancement technology.

[0114] A joint judgment module, which is used to independently judge the dataset of base layer crack images by using an identification model pre-built based on the YOLOv5 series model, and perform joint judgment based on the independent judgment; among them, the independent judgment is to judge the base layer cracks for the longitudinal section image data or the horizontal section image data.

[0115] It should be noted that the pavement performance evaluation system based on joint judgment in this embodiment corresponds to the aforementioned pavement performance evaluation method based on joint judgment. Therefore, the content not specifically described in the pavement performance evaluation system based on joint judgment in this embodiment, which may but is not limited to function definitions, working principles, technical effects, etc., can refer to the records of the aforementioned pavement performance evaluation method based on joint judgment, and will not be elaborated in this text.

[0116] In a specific application, the YOLOv5 model is selected as the YOLOv5x model. The overall precision P, recall R, and mAP@0.5 values for detecting the echo features of base course cracks (the vertical section base course crack echo feature Crack-V and the horizontal section base course crack echo feature Crack-H) using the YOLOv5x model are 0.808, 0.876, and 87.5% respectively, all of which do not exceed 88%, indicating that the overall detection accuracy is not high. The AP@0.5 values for the separate detection of Crack-V and Crack-H using the YOLOv5x model are 82.6% and 92.4% respectively. The detection accuracy of the base course crack echo feature Crack-V in the vertical section GPR is relatively high, being 9.8% higher than Crack-H in the horizontal section GPR image. Based on the above data, it shows that the echo features of some base course cracks cannot be simultaneously identified in both the vertical section and horizontal section GPR images. Therefore, it is necessary to jointly determine the recognition results of the base course crack echo features (Crack-V and Crack-H) in the horizontal section and vertical section GPR images. If the echo features of the same crack can be identified in both the vertical section and horizontal section GPR images, it is determined that there must be a base course crack at that position and no recheck is required. If the echo features of the same crack can only be identified in the GPR image of a certain section (vertical section or horizontal section), manual recheck is needed to determine the base course crack. After the above joint determination, the precision P, recall R, and mAP@0.5 values for the recognition of the base course crack echo features (Crack-V and Crack-H) in the test set GPR images are 0.896, 0.914, and 93.1% respectively, which are increased by 0.088, 0.038, and 5.6% respectively compared with the previous values, indicating that this joint determination method can effectively achieve the accurate recognition of the base course crack echo features in GPR images.

[0117] In summary, for the pavement performance evaluation method and system based on joint judgment in this embodiment, by selecting the acquisition strategy according to the pavement basic information and signal quality, pavement data that is more in line with the actual situation is obtained; the data augmentation technology increases the diversity and reliability of the data, making the generated base course crack image dataset more representative; and the recognition and joint judgment mechanism based on the YOLOv5 series models effectively improves the recognition accuracy of base course cracks and reduces misjudgment; thus comprehensively improving the accuracy of pavement performance evaluation, providing a scientific basis for road maintenance and repair, extending the service life of the road, and ensuring traffic safety.

[0118] In the embodiments provided in the present application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any appropriate combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: application specific integrated circuit (ASIC), digital signal processor (DSP), digital signal processing device (DSPD), programmable logic device (PLD), field programmable gate array (FPGA), processor, controller, microcontroller, microprocessor, other electronic units designed to implement the functions described herein, or a combination thereof. For software implementation, part or all of the processes of the embodiments can be completed by instructing the relevant hardware through a computer program. When implemented, the above program can be stored in a computer-readable storage medium or transmitted as one or more instructions or codes on a computer-readable storage medium. The computer-readable storage medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transfer of a computer program from one place to another. The storage media can be any available medium that can be accessed by a computer. The computer-readable storage medium can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM, or other optical disk storage, magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0119] Finally, it should be noted that the above are only the preferred embodiments of the present application and are not used to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A pavement performance evaluation method based on joint judgment, characterized in that The method includes the following steps: S1: Using 3D ground penetrating radar technology, different acquisition strategies are adopted for different road surfaces, and multi-profile image data of the road surface is acquired based on the corresponding acquisition strategies. Among them, the adoption of the acquisition strategy is selected based on the road surface basic information and the road surface signal quality. The multi-profile image data includes longitudinal section image data and horizontal section image data; S2: Using data enhancement technology, combined with the acquisition strategy, to process the longitudinal section image data and the horizontal section image data to generate a base crack image dataset. Among them, combining the acquisition strategy is to enhance the data obtained by processing the longitudinal section image data and the horizontal section image data based on the data enhancement technology according to the data characteristics corresponding to the acquisition strategy; S3: Using an identification model pre-constructed based on the YOLOv5 series model to independently judge the base crack image dataset, and making a joint judgment based on the independent judgment. Among them, the independent judgment is to judge the base cracks in the longitudinal section image data or the horizontal section image data.

2. The pavement performance evaluation method based on combined judgment according to claim 1, wherein The use of 3D ground penetrating radar technology to adopt different acquisition strategies for different road surfaces includes: The road surface basic information includes the road surface material type and the structural layer thickness. The road surface signal quality includes the signal strength and the signal-to-noise ratio. Different combinations of acquisition parameters are selected as the acquisition strategy based on the road surface material type, the structural layer thickness, the signal strength, and the signal-to-noise ratio. The acquisition parameters include the antenna center frequency, the two-way travel time, the dwell time, and the sampling interval, and the acquisition parameters correspond to the road surface basic information and the road surface signal quality. Among them, when selecting the acquisition strategy, the road surface evaluation is calculated using the road surface evaluation formula, and the corresponding acquisition parameters are matched based on the road surface evaluation. The road surface evaluation formula is: ； Among them, is the comprehensive change rate of pavement material type and structural layer thickness, is the change rate of signal strength, is the change rate of signal-to-noise ratio, is the calculated pavement evaluation.

3. The pavement performance evaluation method based on combined judgment according to claim 2, characterized in that The acquisition of multi-profile image data of the road surface based on the corresponding acquisition strategy includes: Calibrating the 3D ground penetrating radar equipment before acquisition; After setting the acquisition parameters based on the selected acquisition strategy, making the 3D ground penetrating radar travel along the road surface at a constant speed, and synchronously acquiring the longitudinal section image data and the horizontal section image data; During the acquisition process, using a high-precision positioning system to record the position information of each acquisition point.

4. The pavement performance evaluation method based on combined judgment according to claim 1, characterized in that, The use of data enhancement technology, combined with the acquisition strategy, to process the longitudinal section image data and the horizontal section image data includes: Using data enhancement technology to process the longitudinal section image data and the horizontal section image data to generate first enhanced data, and this processing is carried out based on the following steps: Adopting GPR numerical simulation technology to generate base crack simulation images with different depths, sizes, and moisture states, and fusing them with the original longitudinal section image data and horizontal section image data; Performing gain adjustment on the image data after fusing the base crack simulation images with the original longitudinal section image data and horizontal section image data to enhance the contrast and brightness of the images; Performing a scaling operation on the image data after fusing the base crack simulation images with the original longitudinal section image data and horizontal section image data to change the size and proportion of the images; Introducing simulated interference signals and adding random noise to simulate the interference situations that may be encountered in the actual acquisition process.

5. The pavement performance evaluation method based on combined judgment according to claim 4, wherein Using the data augmentation technology and combining with the acquisition strategy to process the longitudinal section image data and the horizontal section image data further includes: Processing the first augmented data based on the acquisition strategy to generate the second augmented data, and this processing is carried out based on the following steps: When the data feature corresponding to the acquisition strategy is that the resolution of the image data does not meet the preset resolution threshold, performing super-resolution reconstruction processing on the first augmented data; When the data feature corresponding to the acquisition strategy is that the noise of the image data does not meet the preset noise threshold, performing noise reduction processing on the first augmented data; When the data feature corresponding to the acquisition strategy is that the feature distribution of the image data does not meet the preset feature distribution threshold, performing histogram equalization on the first augmented data to adjust the feature distribution of the image.

6. The pavement performance evaluation method based on joint judgment according to claim 5, characterized in that Generating the base crack image dataset includes: Sorting and annotating the second augmented data, and this annotation is to annotate the position, size and category of the base cracks in the image; Dividing the annotated image data into a training set, a validation set and a test set based on a preset ratio; where: the training set is used to train the recognition model, the validation set is used to evaluate the recognition model and adjust the hyperparameters during the training process, and the test set is used to finally evaluate the accuracy and generalization ability of the recognition model.

7. The pavement performance evaluation method based on combined judgment according to claim 1, characterized in that Using the recognition model pre-constructed based on the YOLOv5 series of models to make independent judgments on the base crack image dataset, including: Using the recognition model to make independent judgments on the longitudinal section image data in the base crack image dataset, and this independent judgment is carried out based on the following steps: Inputting the longitudinal section image data into the recognition model; Preprocessing the longitudinal section image data, adjusting the image size and maintaining the original aspect ratio; Using a multi-layer convolutional neural network to extract multi-dimensional features of the base cracks in the image; Based on bounding box regression, class prediction and confidence prediction, determining the position, size and category of the base cracks in the image; Performing non-maximum suppression processing to remove overlapping prediction boxes and output the result of independent judgment.

8. The pavement performance evaluation method based on combined judgment according to claim 7, characterized in that Using the recognition model pre-constructed based on the YOLOv5 series of models to make independent judgments on the base crack image dataset further includes: Using the recognition model to make independent judgments on the horizontal section image data in the base crack image dataset, and this independent judgment is carried out based on the following steps: Inputting the horizontal section image data into the recognition model; Preprocessing the horizontal section image data, adjusting the image size and maintaining the original aspect ratio; Using a multi-layer convolutional neural network to extract multi-dimensional features of the base cracks in the image; Based on bounding box regression, class prediction and confidence prediction, determining the position, size and category of the base cracks in the image; Performing non-maximum suppression processing to remove overlapping prediction boxes and output the result of independent judgment.

9. The pavement performance evaluation method based on combined judgment according to claim 8, wherein, And making a joint judgment based on the independent judgments, including: Comparing the independent judgment results of the longitudinal section image data and the horizontal section image data; When the crack echo features at the same position are both determined to be base cracks in the recognition results of the longitudinal section image data and the horizontal section image data, it is determined that there are base cracks at that position; When it is determined that there is a base crack only in the recognition result of the image data of a certain section, the image data at this position needs to be manually rechecked; Based on the result of the joint judgment, comprehensively evaluate the pavement performance, and determine the type, location and severity of the pavement disease.

10. A pavement performance evaluation system based on joint judgment, which is applicable to the pavement performance evaluation method based on joint judgment as described in any one of claims 1-9, and is characterized in that The system includes: A dynamic acquisition module, which is used to adopt different acquisition strategies for different pavements by using three-dimensional ground penetrating radar technology, and acquire multi-section image data of the pavement based on the corresponding acquisition strategies; among them, the acquisition strategy is selected based on the pavement basic information and the pavement signal quality, and the multi-section image data includes longitudinal section image data and horizontal section image data; An image enhancement module, which is used to use data enhancement technology and combine the acquisition strategy to process the longitudinal section image data and the horizontal section image data to generate a base crack image data set; among them, combining the acquisition strategy is to perform enhancement based on data characteristics on the data obtained by processing the longitudinal section image data and the horizontal section image data by using the data enhancement technology, based on the data characteristics corresponding to the acquisition strategy; A joint judgment module, which is used to independently judge the base crack image data set by using the recognition model pre-constructed based on the YOLOv5 series model, and perform joint judgment based on the independent judgment; among them, the independent judgment is to judge the base crack of the longitudinal section image data or the horizontal section image data.

Images