Method for detecting internal diseases of asphalt pavement

Abstract

The invention discloses an asphalt pavement internal disease detection method, which comprises the following steps: acquiring antenna attitude data, and correcting the antenna attitude data; acquiring acquisition parameters of the ground penetrating radar according to the corrected antenna attitude data; acquiring an original electromagnetic signal through the ground penetrating radar according to the acquisition parameters; performing feature extraction on the original electromagnetic signal to obtain a reflection feature; the reflection features are judged, abnormal reflection points are obtained, and the position and intensity data of the abnormal reflection points are extracted; obtaining an abnormal reflection point distribution diagram according to the positions and the intensities of the abnormal reflection points, and obtaining disease types of the abnormal reflection points according to the abnormal reflection point distribution diagram; generating a disease distribution diagram according to the disease type and distribution of the abnormal reflection points, analyzing the severity and range of a disease range according to the disease distribution diagram, and repairing the internal diseases of the pavement according to the severity and range.

Description

Technical Field

[0001] The invention belongs to the technical field of road detection, and in particular relates to a method for detecting internal defects of an asphalt pavement. Background Art

[0002] When using ground penetrating radar to detect internal defects in asphalt pavements, there is a key technical problem: how to improve the efficiency of data collection and analysis while ensuring detection accuracy. Asphalt pavements have a complex structure and contain multiple layers of materials. Electromagnetic waves will produce multiple reflections and attenuation during propagation, resulting in a decrease in the quality of the received signal and increasing the difficulty of data analysis. In addition, there are many types of pavement defects, and the characteristic performance of different types and degrees of defects on radar images varies greatly. It is necessary to establish a complete feature library and intelligent recognition algorithm to accurately interpret the defect information.

[0003] At the same time, ground penetrating radar detection usually needs to be carried out while the vehicle is moving. Affected by factors such as road environment, vehicle speed, and weather, how to ensure the continuity and stability of data collection is also a major challenge. The stability of the radar antenna is directly related to the quality of the received signal, and the bumps and vibrations during driving will interfere with the antenna position and posture, which requires real-time monitoring and compensation. During the data collection process, the real-time storage and transmission must also be considered to avoid interruption or loss of collection due to data backlog. The efficient processing and intelligent analysis of massive detection data puts higher requirements on the performance of the computing platform and the optimization of the algorithm. It is necessary to complete the transformation from data to results within a limited time to provide timely support for road maintenance decisions. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method for detecting internal defects of asphalt pavement to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides a method for detecting internal defects of an asphalt pavement, comprising:

[0006] Acquire antenna attitude data and correct the antenna attitude data;

[0007] According to the corrected antenna attitude data, acquisition parameters of the ground penetrating radar are obtained;

[0008] According to the acquisition parameters, the original electromagnetic signal is obtained by the ground penetrating radar;

[0009] Extract features from the original electromagnetic signal to obtain reflection features; judge the reflection features to obtain abnormal reflection points, and extract the position and intensity data of the abnormal reflection points;

[0010] According to the position and intensity of the abnormal reflection point, a distribution map of the abnormal reflection point is obtained, and according to the distribution map of the abnormal reflection point, the disease type of the abnormal reflection point is obtained;

[0011] A disease distribution map is generated based on the disease type and distribution of abnormal reflection points, and the severity and range of the disease range are obtained based on the disease distribution map. According to the severity and range, the internal diseases of the road surface are repaired.

[0012] Optionally, the process of correcting the antenna attitude data includes:

[0013] The vehicle motion information is collected by the gyroscope and the gageometer, and the vehicle motion data is fused by the Kalman filter algorithm to obtain the antenna attitude data, where the antenna attitude data includes attitude angle data and position information;

[0014] A threshold value is determined for the attitude angle data and the position information, and the antenna attitude data is corrected according to the threshold value determination result.

[0015] Optionally, the process of acquiring the acquisition parameters of the ground penetrating radar includes:

[0016] According to the corrected antenna attitude data, the height of the ground penetrating radar is adjusted, and according to the height of the ground penetrating radar, the optimized value of the transmission power is obtained, and according to the optimized value of the transmission power, the sampling frequency is obtained.

[0017] Optionally, before extracting features from the original electromagnetic signal, the following steps are also included:

[0018] The original electromagnetic signal is preprocessed, wherein the preprocessing includes a wavelet transform method.

[0019] Optionally, the process of obtaining the position and intensity data of the abnormal reflection point includes:

[0020] A threshold judgment is performed on the reflection characteristics, and based on the judgment result, an abnormal reflection point is obtained, wherein the reflection characteristics include amplitude value, phase value and frequency value; statistics are performed on the abnormal reflection points to obtain the reflection position and reflection intensity data of the abnormal reflection points.

[0021] Optionally, the process of obtaining the disease type of the abnormal reflection point includes:

[0022] According to the distribution map of abnormal reflection points, the density value and aggregation degree of the abnormal reflection points are statistically obtained, and the density value is judged by a threshold to obtain the aggregation degree of the disease; the characteristic values ​​of the abnormal reflection points are extracted from the distribution map, and the characteristic values ​​are classified through a deep learning model to obtain the disease type corresponding to the abnormal reflection point.

[0023] Optionally, the raw electromagnetic signal is processed through a distributed computing framework.

[0024] Optionally, after obtaining the disease distribution map, the following steps are also included:

[0025] According to the disease characteristic data in the disease distribution map, predictions are made through intelligent analysis models to obtain disease development trend predictions, disease cause analysis and optimal repair plan recommendations to generate a disease analysis report.

[0026] Compared with the prior art, the present invention has the following advantages and technical effects:

[0027] The present invention discloses a method for detecting internal defects of asphalt pavement. The method ensures the stability of data acquisition by real-time correction of antenna attitude data and dynamic adjustment of ground penetrating radar parameters. The original electromagnetic wave signal is denoised using a wavelet transform algorithm, and reflection features are extracted and abnormal reflection points are determined. Combined with a pre-established pavement defect feature library, a convolutional neural network algorithm is used to classify abnormal reflection points and identify the defect type. A defect distribution map is generated based on the identification results to evaluate the repair priority. A distributed computing framework is used to process massive data in parallel, extract defect features, and generate a structured data set. Finally, combined with an intelligent analysis model, the defect data is deeply analyzed to predict development trends, analyze causes, and recommend the optimal repair plan. The present invention realizes intelligent detection, classification, evaluation and analysis of road defects, and provides a scientific basis for road maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0029] Figure 1 The present invention is a flow chart of a method for detecting internal defects of an asphalt pavement.

[0030] Figure 2 It is a schematic diagram of a method for detecting internal defects of an asphalt pavement according to the present invention.

[0031] Figure 3 It is another schematic diagram of a method for detecting internal defects of an asphalt pavement according to the present invention.

[0032] Figure 4 It is a structural schematic diagram of an asphalt pavement internal disease detection system of the present invention. DETAILED DESCRIPTION

[0033] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0034] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0035] like Figure 1-3 As shown, the present embodiment of a method for detecting internal defects of an asphalt pavement specifically includes:

[0036] Step S101, obtaining antenna attitude data during vehicle driving, using Kalman filter algorithm combined with gyroscope and accelerometer data to perform real-time correction on the antenna attitude data, and eliminating interference of bumps and vibrations on the antenna position.

[0037] The antenna attitude data of the ground penetrating radar is obtained during the vehicle's driving process, and the vehicle motion information is collected using a gyroscope and an accelerometer. The data collected by the gyroscope and accelerometer are preprocessed to remove noise and outliers. The preprocessed gyroscope and accelerometer data are fused using the Kalman filter algorithm. The attitude angle and position information of the antenna are calculated based on the fused data. It is determined whether the antenna attitude angle and position information exceed the preset threshold. If so, real-time correction is performed. The corrected antenna attitude data is compared and analyzed with the original attitude data. The parameters of the Kalman filter algorithm are updated based on the comparison and analysis results to optimize the attitude correction effect.

[0038] For example, obtaining antenna attitude data during vehicle driving is a complex and important task. Gyroscopes and accelerometers are commonly used sensors that can provide high-precision motion information. Gyroscopes measure angular velocity, while accelerometers measure linear acceleration. The combination of these two sensors can fully capture the vehicle's motion state. In the data preprocessing stage, noise and outliers need to be removed. For example, a median filter can be used to remove sudden noise, or a low-pass filter can be used to smooth the data. For outliers, a threshold can be set to remove or replace data points that are out of the normal range. The Kalman filter algorithm is a powerful data fusion tool. It can organically combine the data of gyroscopes and accelerometers to obtain more accurate attitude estimates. The core idea of ​​Kalman filtering is to continuously optimize the estimation of the system state through two steps of prediction and update. In this scenario, the system state can include the position, velocity, and attitude angle of the antenna. When calculating the antenna attitude angle, the commonly used method is the Euler angle representation. It includes yaw, pitch, and roll. For example, when a vehicle turns, the yaw angle changes; when the vehicle goes uphill or downhill, the pitch angle changes; and when the vehicle passes over uneven roads, the roll angle may fluctuate. Real-time correction is key to ensuring antenna performance. Assume that the threshold for the yaw angle is set to ±5°, and the thresholds for the pitch angle and roll angle are set to ±3°. When the attitude angle is detected to exceed these thresholds, the system immediately initiates the correction mechanism. The correction can be achieved through mechanical adjustment or electronic beamforming. The effect of the correction can be evaluated by comparing and analyzing the original attitude data and the corrected data. For example, the root mean square error (RMSE) of the two sets of data can be calculated. If the RMSE after correction is significantly smaller than the RMSE of the original data, it means that the correction effect is good. Updating the parameters of the Kalman filter algorithm based on the comparative analysis results is a closed-loop optimization process. If it is found that the correction effect is not good in certain specific situations (such as high-speed turning), the corresponding noise covariance matrix can be adjusted to make the filter more sensitive in these situations.

[0039] Step S102, dynamically adjusting the acquisition parameters of the ground penetrating radar, including antenna height, transmission power and sampling frequency, according to the corrected antenna attitude data, to ensure the continuity and stability of data acquisition.

[0040] Obtain the corrected antenna attitude data and dynamically adjust the acquisition parameters of the ground penetrating radar. Determine the adjustment value of the antenna height based on the corrected antenna attitude data. Calculate the optimized value of the transmit power based on the adjusted value of the antenna height. Determine the adjustment value of the sampling frequency based on the optimized value of the transmit power. Adjust the data acquisition parameters of the ground penetrating radar based on the adjusted value of the sampling frequency. Use the adjusted data acquisition parameters to obtain the real-time acquisition data of the ground penetrating radar. Determine the continuity and stability of the data acquisition based on the real-time acquisition data.

[0041] Exemplarily, obtaining the corrected antenna attitude data is a key step in optimizing data acquisition in a ground penetrating radar system. Antenna attitude data contains information such as the tilt angle, height and direction of the antenna, and these parameters directly affect the transmission and reception effects of ground penetrating radar signals. For example, when a vehicle is driving on a rough road, the antenna may bump up and down or shake left and right, causing changes in the antenna height and direction. Dynamic adjustment of the acquisition parameters of the ground penetrating radar is to adapt to different surface conditions and detection needs. First, the adjustment value of the antenna height is determined based on the corrected antenna attitude data. Assuming that the original antenna height is set to 1 meter, but the corrected data shows that the actual height is 1.2 meters, the adjustment value is 0.2 meters. This height difference affects the propagation time and energy loss of electromagnetic waves in the air. Next, the optimized value of the transmission power is calculated based on the adjusted value of the antenna height. When the antenna height increases, in order to ensure that the signal can effectively penetrate the surface, the transmission power needs to be appropriately increased. For example, if the antenna height increases by 0.2 meters, the transmission power may need to be increased from the original 100 watts to 120 watts to compensate for the additional propagation loss. Changes in the transmission power will affect the penetration depth and resolution of the signal, so the sampling frequency needs to be adjusted accordingly. The sampling frequency determines the minimum depth interval that the ground penetrating radar system can resolve. For example, if the transmission power is increased and the signal can penetrate deeper strata, then the sampling frequency may need to be reduced to obtain deeper information. Assuming that the original sampling frequency is 1000MHz, it may need to be adjusted to 800MHz to accommodate a deeper detection depth. Acquiring real-time acquisition data of the ground penetrating radar according to the adjusted acquisition parameters is the core of the whole process. These data include information such as the reflection time, amplitude and phase of the electromagnetic wave. By analyzing these data, a two-dimensional or three-dimensional image of the underground structure can be constructed.

[0042] For example, when detecting asphalt pavement, different degrees of internal damage will produce different reflection characteristics. Metal pipelines usually produce strong reflection signals, while the reflection of plastic pipelines is relatively weak.

[0043] Finally, judging the continuity and stability of data acquisition is an important step to ensure the reliability of detection results. Continuity refers to the uninterrupted nature of data in time and space. For example, there should be no large-scale data loss during vehicle driving. Stability is reflected in the consistency of signal strength and quality. If the signal in a certain section suddenly becomes weak or fluctuates abnormally, it may mean that the antenna attitude correction or parameter adjustment is improper. By monitoring these indicators in real time, problems in the data acquisition process can be discovered and corrected in a timely manner, ensuring that the ground penetrating radar system can provide high-quality and reliable underground structure information.

[0044] Step S103, obtaining the original electromagnetic wave signal collected by the ground penetrating radar, and using a wavelet transform algorithm to denoise the signal in view of the noise and attenuation problems in the signal.

[0045] The original electromagnetic wave signal collected by the ground penetrating radar is obtained, and the wavelet transform algorithm is used to denoise the signal to eliminate the noise and attenuation problems in the signal.

[0046] For example, ground penetrating radar technology plays an important role in underground structure detection, and its core lies in the processing and analysis of electromagnetic wave signals. First of all, obtaining the original electromagnetic wave signal is the starting point of the whole process. These signals usually contain a lot of noise, such as environmental interference, equipment noise, etc., which affects the data quality. To this end, the wavelet transform algorithm is used for denoising. Wavelet transform can analyze signals in both time domain and frequency domain at the same time, and effectively separate noise and useful signals. For example, for a reflection signal of the internal structure of road asphalt containing high-frequency noise, by selecting the appropriate wavelet basis function and decomposition layer number, the noise can be effectively eliminated while retaining the reflection characteristics of the internal structure. The antenna attitude has a significant impact on the detection results.

[0047] Step S104, based on the denoised electromagnetic wave signal, extract the reflection features in the signal, including amplitude, phase and frequency, determine whether there are abnormal reflection points in the signal through a pre-set reflection feature threshold, and determine the position and intensity of the abnormal reflection points.

[0048] The de-noised signal is used as input to extract the reflection features in the signal, including amplitude value, phase value and frequency value. According to the pre-set feature threshold, it is judged whether there is abnormal reflection in the reflection feature. If there is abnormal reflection, the reflection position and reflection intensity of the abnormal reflection are determined. According to the reflection position and reflection intensity of the abnormal reflection, a distribution map of the abnormal reflection points is generated. Using the distribution map as input, the density and aggregation degree of the abnormal reflection points are calculated. According to the density and aggregation degree, the distribution pattern of the abnormal reflection points is judged. Using the distribution pattern as input, the classification results of the abnormal reflection points are generated.

[0049] For example, in the ground penetrating radar signal processing, the denoised signal is used as the basis, and extracting the reflection feature is a key step. The amplitude value reflects the strength of the reflection interface, the phase value contains the depth information, and the frequency value is related to the size and properties of the target object. For example, in the detection of the internal structure of the ground, cracks usually produce stronger reflection signals, which are manifested as higher amplitude values; while normal reflections are relatively weak. The setting of the feature threshold is crucial, which determines the judgment criteria for abnormal reflections. In practical applications, it can be adjusted according to the detection environment and target characteristics, and set through the manual experience of relevant personnel. Determining the location and intensity of abnormal reflections is the basis for accurately locating underground targets. The reflection position is usually calculated by the time delay of the signal, while the intensity is directly related to the amplitude value. Generating a distribution map of abnormal reflection points is an important means of visual analysis. Calculating the density and aggregation degree of abnormal reflection points helps to identify potential important areas. In geological hazard assessment, high-density abnormal reflection points may indicate rock fractures or hollow areas, which require further investigation. Analysis of the degree of aggregation can help distinguish discrete single targets from continuous geological structures. Judging the distribution pattern of abnormal reflection points is the key to interpreting underground structures. Different types of diseases will have different distribution patterns. Finally, generating classification results of abnormal reflection points is the ultimate goal of GPR data interpretation. Through this series of steps, GPR data is converted from raw signals into interpretable underground information, providing an important basis for engineering decision-making, scientific research and environmental protection. The advantage of this method is that it can obtain underground information non-destructively and improve the reliability and interpretability of data through digital processing. However, the accuracy of the interpretation still depends on the operator's experience and understanding of geological conditions, so it is often necessary to combine verification methods such as drilling in practical applications.

[0050] In step S105, for the detected abnormal reflection points, combined with the pre-established pavement disease feature library, firstly use the traditional image processing method to extract features, and then use the convolutional neural network algorithm to classify these features to determine the disease type corresponding to the abnormal reflection point.

[0051] The location and intensity data of abnormal reflection points are used to generate a distribution map of abnormal reflection points. According to the distribution map, the density value of the abnormal reflection points is calculated. If the density value is higher than the preset threshold, it is judged that there is a clustering phenomenon, that is, the degree of clustering of abnormal reflection points. The image processing method is used to extract the characteristic values ​​of abnormal reflection points from the distribution map. Combined with the pre-established pavement disease feature library, the characteristic values ​​are classified by the convolutional neural network algorithm. If the classification result matches the disease type in the feature library, the disease type corresponding to the abnormal reflection point is determined.

[0052] Generate a detailed distribution report of pavement defects based on the defect type and distribution map.

[0053] Exemplarily, the distribution map of abnormal reflection points is generated by visualizing the location and intensity data of the reflection points. This graphical representation can intuitively show the distribution of pavement diseases and help to quickly identify problem areas. For example, on a 1-kilometer-long road, a high-intensity reflection point concentration area may be found at 500 meters, which may indicate that there are serious pavement diseases inside the asphalt. The calculation of density value and aggregation degree is a key step in evaluating the distribution characteristics of abnormal reflection points. The density value reflects the number of abnormal reflection points per unit area, while the aggregation degree indicates the spatial correlation of these points. Assuming that the set density threshold is 5 abnormal points per square meter, if the density value of a certain area reaches 8 / square meter, it can be determined that there is a significant aggregation phenomenon in the area. This aggregation may indicate a continuous area of ​​pavement damage, such as large cracks or cavities. Extracting feature values ​​from the distribution map is the basis for identifying specific types of pavement diseases. This process may involve a variety of image processing techniques, such as edge detection, texture analysis, etc. For example, linear features can be detected by Hough transform, which may correspond to cracks; while surface features can be identified by region growing algorithm, which may correspond to potholes or spalling. The pre-established pavement disease feature library is a data set containing various typical pavement disease features. This feature library may include feature descriptions of common diseases such as different types of cracks (such as transverse, longitudinal, and mesh cracks), voids, and water. For example, transverse cracks may appear as linear features perpendicular to the driving direction, while mesh cracks appear as intertwined multi-directional linear features. The convolutional neural network (CNN) algorithm is used here to match the extracted feature values ​​with the disease types in the feature library. The multi-layer structure of CNN can automatically learn and extract complex features, which is very suitable for processing image recognition tasks. Through training with a large amount of labeled data, CNN can learn to distinguish different types of pavement diseases. For example, it may learn that cracks usually appear as long strips of depressions parallel to the driving direction, while voids are more like discrete circular or irregular pits. The final generated pavement disease detailed distribution report is the comprehensive result of the entire analysis process. This report not only contains the type and location information of the disease, but also may include important parameters such as the severity and distribution range of the disease. For example, the report may indicate that there is a serious crack of about 1 meter long and 0.2 meters wide at a depth of 15 mm at the beginning of the road 800 meters away, which needs to be repaired immediately. This detailed information is crucial for the road maintenance department to formulate an accurate repair plan, which can greatly improve the efficiency and effectiveness of road maintenance.

[0054] Step S106, generating a disease distribution map based on the identified disease types and the distribution of abnormal reflection points, analyzing the severity and scope of the disease through the disease distribution map, and determining the repair priority of the disease in combination with the preset evaluation criteria.

[0055] Generate a disease distribution map using the disease type and abnormal reflection point distribution data. Calculate the density and aggregation of the disease based on the disease distribution map. If the density value is higher than the preset threshold, it is determined that the disease is aggregated. Use image processing methods to extract disease feature values ​​from the disease distribution map. Combined with the pre-established pavement disease feature library, classify the disease feature values ​​using the convolutional neural network algorithm. If the classification result matches the disease type in the feature library, determine the disease type. Generate a detailed distribution report of pavement diseases based on the disease type and distribution map. Combined with the preset evaluation criteria, analyze the severity and scope of the disease and determine the priority of disease repair.

[0056] For example, the disease distribution map is an important tool for pavement maintenance, which is generated by integrating the disease type and abnormal reflection point distribution data. For asphalt pavement, common internal diseases include cracks, voids and water. When making a distribution map, different diseases can be represented by different colors or symbols, such as red for cracks and blue for voids. This visualization method can intuitively display the spatial distribution characteristics of diseases. The density value and aggregation degree of diseases are key indicators for evaluating pavement conditions. The density value is usually expressed as the number or area of ​​diseases per square meter. For example, if 10 cracks are found on a section of 100 square meters of pavement, the crack density is 0.1 per square meter. The aggregation degree reflects the uniformity of the distribution of the disease, which can be quantified by calculating the nearest neighbor index and other methods. When the density value exceeds the preset threshold (such as 0.2 per square meter), it is determined that there is an aggregation phenomenon, indicating that the road condition in this area is poor and needs to be treated as a priority. Extracting feature values ​​from the disease distribution map is the basis for further analysis. Commonly used image processing methods include edge detection and texture analysis. Taking cracks as an example, its length, width, direction and other features can be extracted. These feature values ​​are compared with the pre-established pavement disease feature library, and accurate classification is achieved through the convolutional neural network algorithm. For example, there are obvious differences in the morphology between mesh cracks and linear cracks, and the type of disease can be accurately identified through feature matching. A detailed distribution report of pavement diseases is an important basis for maintenance decisions. The report usually contains information such as disease type, location, area, and severity. For example, it may describe that "severe rutting was found in the section from milepost K15+200 to K15+300, with a maximum depth of 25 mm and an affected length of about 80 meters." This precise description helps maintenance personnel quickly locate and evaluate problem areas. The analysis of disease severity and scope involves multiple dimensions. For cracks, their width, depth, and degree of networking can be considered; for rutting, their depth and length are focused on. For example, cracks with a width of more than 3 mm are usually considered severe cracks. The scope reflects the spatial distribution of the disease and can be expressed as the percentage of the affected pavement. These factors jointly determine the priority of repair, such as diseases with high severity and wide range should be treated first. Through this series of analyses, the road condition can be comprehensively assessed, providing a basis for formulating scientific and reasonable maintenance plans. This data-driven approach not only improves maintenance efficiency, but also optimizes resource allocation, ensuring the long-term reliability and safety of the road network.

[0057] Step S107, for the collected massive electromagnetic wave signal data, a distributed computing framework is used to process the data in parallel, and the disease features in the data are extracted, including the disease type, location, depth and area, to generate a structured disease feature data set.

[0058] A distributed computing framework is used to parallelize the massive electromagnetic wave signal data and extract the characteristic values ​​in the signal. Based on the extracted characteristic values, the convolutional neural network algorithm is used to classify the disease class in combination with the pre-established disease feature library. If the classification result matches the disease class in the feature library, the disease class is determined. According to the disease class, the location point, depth value and area value are extracted from the characteristic value. A structured method is used to integrate the disease class, location point, depth value and area value into a structured data set. Based on the structured data set, a disease distribution map is generated. An image processing method is used to extract the disease characteristic value from the disease distribution map. Based on the disease characteristic value, the density value and aggregation degree of the disease are calculated. If the density value is higher than the preset threshold, it is judged that the disease is aggregated.

[0059] For example, massive electromagnetic wave signal data processing is a key link in pavement disease detection. Distributed computing frameworks such as Hadoop or Spark can efficiently process large-scale data in parallel. For example, a 100-kilometer-long highway may generate several TB of electromagnetic wave signal data. Through distributed processing, the data can be divided into multiple small blocks, and feature extraction can be performed simultaneously on different nodes, greatly shortening the processing time. Feature extraction is the process of obtaining useful information from the original signal. Common features include signal strength, frequency distribution, and phase difference. Taking signal strength as an example, it can be characterized by calculating statistics such as the mean, variance, and peak value of the signal. These feature values ​​provide an important basis for subsequent disease classification. Convolutional neural network (CNN) is a powerful deep learning algorithm suitable for image recognition and classification tasks. In pavement disease detection, the extracted feature values ​​can be converted into a two-dimensional image and then input into a pre-trained CNN model for classification. For example, the signal strength of different frequency bands can be mapped to different color channels of the image to form a representation similar to a heat map. CNN gradually extracts high-level features through multi-layer convolution and pooling operations, and finally outputs the probability distribution of disease categories. The defect feature library is a dataset containing features of various known defect types. By comparing the classification results of CNN with the samples in the feature library, the specific defect type can be determined. For example, if the highest probability output by CNN corresponds to the "crack" category and the similarity with the crack samples in the feature library exceeds 95%, it can be determined that there is a crack defect at that location. Location points, depth values, and area values ​​are important parameters for describing defects. Location points can be marked by GPS or mileage posts. Depth values ​​can be estimated by analyzing the reflection time of electromagnetic waves. Area values ​​can be calculated based on the spatial distribution range of abnormal signals. These data are integrated into a structured dataset for subsequent analysis and visualization. The defect distribution map is an effective tool for intuitively displaying the pavement condition. It can be generated using GIS software or a custom drawing library. For example, different types and severity of defects can be represented by markers of different colors and shapes. This visualization method helps to quickly identify problem areas and formulate maintenance strategies. Extracting feature values ​​from the defect distribution map can further quantify the pavement condition. For example, the number of defects per unit area can be calculated as a density value, or a spatial clustering algorithm can be used to evaluate the degree of aggregation of defects. If the crack density of a certain section of road surface exceeds 10 per square meter, it may indicate that there are serious structural problems in the area and that priority treatment is required. Through this series of data processing and analysis steps, the road surface condition can be comprehensively assessed, providing a scientific basis for road maintenance decisions. This approach not only improves detection efficiency and accuracy, but also predicts potential problems and enables proactive management of road maintenance.

[0060] Step S108, based on the generated disease feature data set and combined with the pre-established intelligent analysis model, an in-depth analysis of the disease data is performed, the analysis content includes disease development trend prediction, disease cause analysis and optimal repair plan recommendation, and a comprehensive disease analysis report is generated.

[0061] Using the disease feature data set as input, combined with the pre-established intelligent analysis model, the disease data is deeply analyzed. Based on the results of the deep analysis, the disease development trend is predicted through the time series analysis method. If the disease development trend shows an upward trend, the disease cause is analyzed in combination with the disease feature library. Based on the results of the disease cause analysis, the optimization algorithm is used to generate the optimal repair plan. The disease development trend prediction, disease cause analysis and optimal repair plan are integrated into structured data. A report generation algorithm is used to generate a comprehensive disease analysis report based on the structured data. Based on the disease analysis report, the disease feature library and the intelligent analysis model are updated for subsequent disease analysis.

[0062] For example, the intelligent analysis model is the core component of pavement disease analysis, which combines machine learning algorithms and expert knowledge to comprehensively interpret the characteristics of the disease. For example, a model based on random forests can simultaneously consider multiple features such as crack width, depth and distribution pattern to accurately determine the severity and type of the disease. The advantage of this model is that it can handle complex nonlinear relationships and adapt to the manifestation of diseases under different road conditions. Time series analysis plays a key role in predicting the development trend of diseases. By analyzing historical data, seasonal changes and long-term trends can be identified. For example, using the ARIMA model to analyze the crack data of a certain section of highway, it was found that the crack growth rate in winter is 1.5 times that in summer. This finding helps to formulate a more targeted maintenance plan. Disease cause analysis is an important basis for formulating repair plans. By comparing with the disease feature library, the potential cause of the disease can be inferred. For example, if a network of cracks appears on a section of road and the annual rainfall in the area exceeds 1,500 mm, it may be judged as a disease caused by water damage. This analysis not only helps to select appropriate repair materials, but also guides the formulation of preventive maintenance measures. Optimization algorithms play a key role in generating optimal repair solutions. Considering multiple factors such as repair cost, construction period and traffic impact, a multi-objective genetic algorithm can be used to balance various indicators. For example, for a city trunk road with heavy traffic, the algorithm may recommend a micro-surface treatment solution for nighttime construction, which can quickly repair the road surface while minimizing the impact on traffic. The integration of structured data provides a solid foundation for subsequent analysis and decision-making. Organizing information such as disease trends, causes and repair solutions into standardized data formats, such as JSON or XML, facilitates data exchange and processing between different systems. This structured representation enables complex disease information to be quickly retrieved and analyzed. The report generation algorithm converts structured data into easy-to-understand narrative reports through natural language processing technology. For example, for a 100-kilometer section of highway, the algorithm can generate a comprehensive report containing a disease distribution heat map, a disease development forecast for the next five years, and segmented repair recommendations. This automated report not only improves work efficiency, but also ensures the consistency and comprehensiveness of the report content. The continuous updating of the disease feature library and intelligent analysis model is the key to maintaining system effectiveness. After each analysis, new disease features and analysis results are added to the feature library and used to fine-tune the analysis model. This feedback loop allows the system to continuously learn and adapt to new road conditions and disease types, improving the accuracy of future analyses. For example, if it is found that the disease development pattern under a new type of asphalt material is different from that of traditional materials, the system will adjust its prediction model accordingly to adapt to this new situation.

[0063] like Figure 4 As shown, the present invention provides an asphalt pavement internal disease detection system, which mainly includes:

[0064] The antenna attitude correction module is used to obtain the antenna attitude data during vehicle driving. It uses the Kalman filter algorithm combined with the gyroscope and accelerometer data to perform real-time correction on the antenna attitude data to eliminate the interference of bumps and vibrations on the antenna position.

[0065] The acquisition parameter adjustment module is used to dynamically adjust the acquisition parameters of the ground penetrating radar, including antenna height, transmission power and sampling frequency, according to the corrected antenna attitude data to ensure the continuity and stability of data acquisition;

[0066] The signal denoising processing module is used to obtain the original electromagnetic wave signal collected by the ground penetrating radar, and use the wavelet transform algorithm to denoise the signal in order to solve the noise and attenuation problems in the signal;

[0067] The reflection feature extraction module is used to extract the reflection features in the signal, including amplitude, phase and frequency, based on the denoised electromagnetic wave signal, and determine whether there are abnormal reflection points in the signal through a pre-set reflection feature threshold, and determine the position and intensity of the abnormal reflection points;

[0068] The disease classification module is used to extract features of detected abnormal reflection points by using traditional image processing methods in combination with the pre-established pavement disease feature library, and then classify these features using a convolutional neural network algorithm to determine the disease type corresponding to the abnormal reflection point;

[0069] The disease distribution analysis module is used to generate a disease distribution map based on the identified disease types and the distribution of abnormal reflection points, analyze the severity and scope of the disease through the disease distribution map, and determine the repair priority of the disease in combination with the preset evaluation criteria;

[0070] The data processing module is used to process the massive amount of electromagnetic wave signal data collected in parallel using a distributed computing framework, extract the disease characteristics in the data, including the disease type, location, depth and area, and generate a structured disease characteristic data set;

[0071] The intelligent analysis module is used to conduct in-depth analysis of disease data based on the generated disease feature data set and in combination with the pre-established intelligent analysis model. The analysis content includes disease development trend prediction, disease cause analysis and optimal repair plan recommendation, and generates a comprehensive disease analysis report.

[0072] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for detecting internal defects of an asphalt pavement, characterized in that: include: Acquire antenna attitude data and correct the antenna attitude data; According to the corrected antenna attitude data, acquisition parameters of the ground penetrating radar are obtained; According to the acquisition parameters, the original electromagnetic signal is obtained by the ground penetrating radar; Extract features from the original electromagnetic signal to obtain reflection features; Judge the reflection characteristics, obtain abnormal reflection points, and extract the position and intensity data of the abnormal reflection points; According to the position and intensity of the abnormal reflection point, a distribution map of the abnormal reflection point is obtained, and according to the distribution map of the abnormal reflection point, the disease type of the abnormal reflection point is obtained; A disease distribution map is generated based on the disease type and distribution of abnormal reflection points, and the severity and range of the disease range are obtained based on the disease distribution map. According to the severity and range, the internal diseases of the road surface are repaired.

2. The method according to claim 1, characterized in that The process of correcting the antenna attitude data includes: The vehicle motion information is collected by the gyroscope and the gageometer, and the vehicle motion data is fused by the Kalman filter algorithm to obtain the antenna attitude data, where the antenna attitude data includes attitude angle data and position information; A threshold value is determined for the attitude angle data and the position information, and the antenna attitude data is corrected according to the threshold value determination result.

3. The method according to claim 1, characterized in that The process of acquiring the acquisition parameters of the ground penetrating radar includes: According to the corrected antenna attitude data, the height of the ground penetrating radar is adjusted, and according to the height of the ground penetrating radar, the optimized value of the transmission power is obtained, and according to the optimized value of the transmission power, the sampling frequency is obtained.

4. The method according to claim 1, characterized in that: Before extracting features from the original electromagnetic signal, the following steps are also included: The original electromagnetic signal is preprocessed, wherein the preprocessing includes a wavelet transform method.

5. The method according to claim 1, characterized in that The process of obtaining the location and intensity data of the abnormal reflection point includes: A threshold judgment is performed on the reflection characteristics, and based on the judgment result, an abnormal reflection point is obtained, wherein the reflection characteristics include amplitude value, phase value and frequency value; statistics are performed on the abnormal reflection points to obtain the reflection position and reflection intensity data of the abnormal reflection points.

6. The method according to claim 1, characterized in that The process of obtaining the disease type of abnormal reflection points includes: According to the distribution map of abnormal reflection points, the density value and aggregation degree of the abnormal reflection points are statistically obtained, and the density value is judged by a threshold to obtain the aggregation degree of the disease; the characteristic values ​​of the abnormal reflection points are extracted from the distribution map, and the characteristic values ​​are classified through a deep learning model to obtain the disease type corresponding to the abnormal reflection point.

7. The method according to claim 1, characterized in that The raw electromagnetic signals are processed through a distributed computing framework.

8. The method according to claim 1, characterized in that: After obtaining the disease distribution map, it also includes: According to the disease characteristic data in the disease distribution map, predictions are made through intelligent analysis models to obtain disease development trend predictions, disease cause analysis and optimal repair plan recommendations to generate a disease analysis report.

Images