Monitoring and Warning System for Highway Pavement Defects Using Radar

Abstract

The present invention discloses a monitoring and warning system for monitoring highway pavement defects using radar, which includes a signal acquisition module installed on the highway crossbar for acquiring real-time reflected wave signals of vehicles driving on the highway pavement; a reflected wave signal extraction module for obtaining the real-time wave change characteristics when the vehicle is driving in a non-steady state, and a highway pavement defect atlas feature library collecting the wave change characteristics to form a highway pavement defect atlas feature library; and grading the risk levels of the highway pavement defect atlas feature library. The wave change feature comparison module is used to compare the real-time wave deformation characteristics with the highway pavement defect atlas feature library, and the warning device sends a warning to the vehicles driving into this area. The present invention uses radar to obtain abnormal waveforms when the vehicle drives to the pavement defect, and can quickly obtain the abnormal pavement conditions through the vehicle body posture, and then timely warn the vehicle; by monitoring the vehicle vibration, the pavement monitoring of a larger area can be realized, which has the advantages of low cost and being suitable for popularization.

Description

Technical Field

[0001] The present invention relates to a monitoring and early warning system for detecting highway pavement defects using radar, belonging to the technical field of radio wave detection. Background Art

[0002] With the increase in the service life of highways and the change of natural conditions, pavement defects such as settlement, cracking, and potholes are likely to occur on the road surface. Especially in the past two years, affected by extreme weather, pavement defects mainly caused by foundation settlement have become an important factor affecting traffic safety, bringing great risks to people's lives and property. Therefore, monitoring and diagnosing highway pavements play an important role in ensuring vehicle driving safety, reducing traffic congestion, increasing the actual road traffic flow, and enhancing the supervision ability and maintenance strength of transportation and related departments.

[0003] In the investigation and assessment report on the "5·1" landslide disaster in the Chayang section of the Meida Expressway, it is mentioned that this landslide disaster has characteristics such as strong suddenness, clear sliding boundaries, orderly stacking before and after, and steep back walls of the landslide. Comprehensive judgment shows that this landslide disaster first occurred as an overall slip along the nearly north-south groove direction in the middle and lower parts of the embankment, and then triggered the collapse of the upper fill of the embankment. The main lessons mentioned that there is no standard specification for highway monitoring and early warning. There is no monitoring of the stability of key subgrade slopes, and the landslide disaster could not be detected and warned in the first place, which is far from the requirements of the "five-fast mechanism" of rapid discovery, rapid evacuation, rapid linkage, rapid self-inspection, and rapid intervention. Suggestions for prevention and rectification measures: Promote science and technology for safe operation, enhance the risk monitoring and control ability of highways, improve the intelligent comparison of key projects and key parts, strengthen scientific and technological information means, strengthen and enhance the disaster monitoring and early warning ability for the entire life cycle of embankment slope projects, and jointly provide reminder services such as traffic anomalies and emergencies to drivers and passengers and relevant units in a timely manner with the navigation software platform.

[0004] Existing common road monitoring mostly uses road surface inspection vehicles, which use radar, optical sensors, mechanical sensors, etc. to detect the road surface. For example, the patent titled "A Ground Penetrating Radar Detection Vehicle" with patent number 201920843507.1 discloses that a partition is provided inside the vehicle body to divide the vehicle body into a cab at the front, a control room in the middle, and a detection device placement room at the rear. A control management device, a display device, and a power supply device are provided in the control room. A placement rack is provided in the detection device placement room, and a detection trolley for detecting and collecting data is provided on the placement rack. A four-channel ground penetrating radar device, a high-definition video acquisition device, a GPS positioning device, and a transmission device for transmitting data to the control management device are provided on the detection trolley. A rotary encoder connected to the control management device is provided on the wheel hub of the vehicle body. Although this road surface detection vehicle for detection can effectively discover the actual road conditions and hidden damages of the road surface, its detection is periodic and it is impossible to obtain the road surface conditions of the highway in real time, especially the real-time conditions of the road that affect driving safety.

[0005] The existing highway road surface defect detection mainly relies on the following technologies:

[0006] Manual inspection and inspection vehicles: Long cycle (≥14 days / time), poor timeliness, unable to provide real-time feedback on sudden defects. Embedded sensors: Require large-scale deployment (average cost ≥ 2 million yuan / km), long construction cycle, and difficult maintenance. Optical video monitoring: Significantly affected by light, rain, and fog (false alarm rate > 30% in bad weather), limited night detection ability. Safety hazards and economic losses: According to the statistical data of the Ministry of Transport in 2023, traffic accidents caused by road surface defects accounted for 18%, and the annual direct economic loss exceeded 5 billion yuan.

[0007] Through retrieval, there are also many technologies that can monitor the road surface in real time. For example, the patent titled "A Highway Road Surface Settlement Monitoring and Early Warning System" with patent application number 201811325798.1 uses a settlement detection unit, auxiliary detection equipment, relay equipment, and a monitoring center. The settlement detection unit includes multiple displacement sensors, and all the multiple displacement sensors are located at the lower end of the road surface and inside the roadbed. The auxiliary detection equipment includes a video acquisition device, a vehicle weighing device, a vehicle counting device, and a seismic detection device. The seismic detection device is located at the lower end of the road surface and inside the roadbed. The relay equipment includes a data acquisition unit, a wireless transmission unit, and an emergency communication device, which can monitor the settlement of the highway road surface in real time, collect multiple data, and is more convenient for staff to analyze the reasons for road surface settlement from multiple aspects, realizing a comprehensive understanding of the monitored section by the monitoring center and timely eliminating potential safety hazards caused by settlement on the highway. Although this technology can monitor the internal conditions in real time, it requires a large number of sensors to be installed on the road surface. Due to the long mileage of highways, a large number of devices need to be installed to achieve this, resulting in slow construction and high costs.

[0008] There are also methods for road detection based on existing big data, such as a real-time road disease detection method based on radar, video, and data analysis, with patent application number 202410044080.4, which includes the following steps: establishing a disease detection neural network architecture; obtaining a training sample set; training the disease detection neural network architecture; a patrol vehicle patrols on the road surface, inputs radar scan point cloud maps and video image frames with the same acquisition position information into the trained disease detection neural network architecture, and outputs disease recognition results. The present invention applies radar, video, GPS, and data analysis technologies to real-time road disease detection, and through steps such as multi-modal data acquisition, feature extraction, data fusion, and machine learning, realizes real-time monitoring and anomaly detection capabilities of road conditions, which helps to improve road safety, usability, and maintenance efficiency. The present invention can accurately and efficiently achieve large-scale detection of road diseases, thus solving the deficiencies of manual detection. Although this method can achieve large-scale detection with relatively high efficiency, it needs to be based on video, has high requirements for video clarity, and also has the problem of high cost. At the same time, video data is severely affected by fog, rain, and darkness, and real-time monitoring cannot be achieved. Summary of the Invention

[0009] In view of this, the purpose of the present invention is to provide a monitoring and early warning system for detecting high-speed road surface defects using radar. This system abandons the existing idea of directly monitoring the road surface with video and radar, and judges road surface defects by detecting abnormal fluctuations of vehicles, so as to solve the problems mentioned in the background technology.

[0010] A monitoring and early warning system for detecting high-speed road surface defects using radar includes the following functional modules:

[0011] The signal acquisition module, including radar and Beidou timing, is installed on the crossbar above the high-speed road or the roadside vertical pole. The radar includes a transmitter, a transmitting antenna, a receiver, a receiving antenna, and a signal processing unit, and is used to acquire the real-time reflected wave signal of the vehicle driving on the high-speed road surface. Beidou timing realizes multi-radar time synchronization;

[0012] The reflected wave signal extraction module, which is connected to the signal acquisition module, is used to acquire the real-time reflected wave signal and its position information, and uses waveform data processing technology to extract the features of the real-time reflected wave signal of the vehicle driving on the high-speed road surface obtained, for obtaining the real-time wave change characteristics when the vehicle is driving non-steadily, and transmits the obtained real-time wave change characteristics and position information to the wave change characteristic comparison module. The waveform data processing technology includes a. preprocessing and denoising, b. fast Fourier transform to extract the 50-200 Hz frequency band, c. synchrosqueezing wavelet transform for precise positioning, d. adaptive threshold filtering for noise reduction, e. impact feature enhancement to improve the signal-to-noise ratio, f. three-dimensional feature coding adaptation model;

[0013] High-speed road surface defect atlas feature library. a. Collect known high-speed road surface defects that affect vehicle driving to establish a road surface defect library. b. Search for the high-speed road surface defects in the road surface defect library on the existing high-speed section or replicate the high-speed road surface defects in the road surface defect library by temporarily building a section. c. Set a signal acquisition module at the road surface defects found or replicated in step b. Use the signal acquisition module to obtain the wave change characteristics during non-steady driving when the vehicle drives to the high-speed road surface defects found or replicated in step b, and collect the wave change characteristics to form a high-speed road surface defect atlas feature library; and classify the risk levels of the high-speed road surface defect atlas feature library and set the wave change characteristic threshold.

[0014] Wave change characteristic comparison module, which is connected to the high-speed road surface defect atlas feature library, the reflected wave signal extraction module and the warning device. When it receives the real-time wave deformation characteristics and position information obtained in the reflected wave signal extraction module, it analyzes and compares them with the wave change characteristics in the high-speed road surface defect atlas feature library. After the analysis and comparison, if the real-time wave deformation characteristics are located at the alarm level threshold, it sends a signal to the warning device at the corresponding position.

[0015] Warning device, which is installed next to or 30 - 50 meters in front of the signal acquisition module, and sends a warning to the vehicles driving into this area by means of warning lights, loudspeakers and / or radio voice warnings.

[0016] The aforementioned monitoring and warning system for monitoring high-speed road surface defects using radar is that the system further includes a supervision center, which can obtain the real-time wave change characteristics and their position information obtained by the reflected wave signal extraction module, perform waveform processing and analysis on the real-time wave change characteristics, and use the high-speed road surface defect atlas feature library to invert the road surface defects corresponding to the real-time wave change characteristics, and send the obtained road surface defects and the position information of the road surface defects to the road maintenance department for maintenance; the supervision center can also obtain the signals sent by the wave change characteristic comparison module to the warning device, and send the signals sent to the warning device to the road administration, traffic police and road maintenance departments.

[0017] The aforementioned monitoring and warning system for monitoring high-speed road surface defects using radar is that the waveform data processing technology includes:

[0018] a. Preprocessing denoising includes denoising and signal alignment: separating high-frequency impact components and low-frequency vibrations through dynamic noise suppression to eliminate environmental noise; beamforming optimization: using the spatial filtering of the MIMO radar array to suppress the interference signals of lateral vehicles and retain the echoes in the main lobe direction; data normalization: performing amplitude normalization on the radar intermediate-frequency signals to eliminate the influence of distance attenuation and time-domain alignment and compensate for the phase shift caused by vehicle movement.

[0019] b. Fast Fourier Transform extraction of the 50 - 200 Hz frequency band includes coarse - grained frequency band screening: Perform a 2048 - point FFT on the pre - processed signal, analyze the signal spectrum, and extract the 50 - 200 Hz frequency band related to driving safety; Frequency - domain energy threshold segmentation: Set the energy threshold to filter out the low - frequency lateral motion components below 20 Hz and the ultra - high - frequency noise above 200 Hz, and retain the potentially hazardous frequency band; Spectrum baseline calibration: Highlight the sudden impact signal through background spectrum difference.

[0020] c. Synchrosqueezing wavelet transform for precise positioning includes time - frequency resolution improvement: Perform synchrosqueezing wavelet transform on the signal after FFT screening, redistribute the fuzzy energy of the traditional wavelet transform to the instantaneous frequency trajectory, and achieve high - resolution time - frequency analysis of 0.2 ms / 5 Hz; Transient impact positioning: Separate short - term impacts within 30 ms from continuous vibrations, and accurately locate the exact time points of events such as road surface potholes through instantaneous frequency.

[0021] d. Adaptive threshold filtering for noise reduction includes dynamic noise suppression: Calculate the instantaneous energy of the signal based on the Teager energy operator, set the adaptive threshold, and filter out more than 90% of the residual interference; Initial screening of impact events: Retain the regions with sudden energy increase and mark them as potential road surface hazard events.

[0022] e. Impact feature enhancement to improve the signal - to - noise ratio includes energy focusing: Perform Gaussian energy redistribution on the screened impact signals, and through weighted aggregation of neighborhood energy, increase the impact energy concentration by 45% and improve the signal - to - noise ratio by more than 22 dB, highlighting the waveform details; Waveform morphology restoration: Compensate for the attenuation and distortion during signal transmission and restore the true impact waveform.

[0023] f. Three - dimensional feature encoding for model adaptation includes multi - dimensional feature fusion: Encode time - domain, frequency - domain, and energy - domain features into three - dimensional vectors, expand them into high - order features to form a 256 - dimensional feature matrix; Dimensionality reduction and classification adaptation: Use principal component analysis to compress the features to 18 dimensions for adaptation to subsequent machine - learning models.

[0024] The aforementioned monitoring and warning system for high - speed road surface defects using radar is such that the risk levels of the high - speed road surface defect atlas feature library are divided into the general - level high - speed road surface defect atlas feature library, the slow - down - and - pass high - speed road surface defect atlas feature library, and the dangerous - level high - speed road surface defect atlas feature library.

[0025] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that in the wave change feature comparison module, there are installed a deceleration-level highway pavement defect atlas feature library and a danger-level highway pavement defect atlas feature library. If the real-time wave deformation feature is in the deceleration-level highway pavement defect atlas feature library, a deceleration prompt signal is sent to the early warning device; if the real-time wave deformation feature is at the lowest threshold of the high-risk level, an emergency stop prompt signal is sent to the early warning device.

[0026] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that the radar uses a multi-input multi-output (MIMO) antenna array with a frequency of 10 - 80 GHz, and through beamforming technology, the main lobe energy is concentrated in the vehicle driving direction to suppress lateral interference signals.

[0027] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that when the signal acquisition module acquires the wave change features during non - steady driving when the vehicle travels to the highway pavement defects found or replicated in step b, it includes the wave change features formed by different types of vehicles and speeds, and the collected highway pavement defect atlas feature library includes the wave change features of different types of vehicles at different speeds.

[0028] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that when the highway pavement defect atlas feature library classifies the risk level, it is based on pavement defects and is classified manually through test data.

[0029] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that the early warning device further includes a text display screen or an electronic speed limit sign.

[0030] The aforesaid monitoring and early warning system for monitoring highway pavement defects using radar is such that the highway pavement defects include pavement settlement, pavement collapse, pavement potholes, pavement cracking, and pavement bumps.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] 1. By changing the existing idea of directly monitoring the actual road surface conditions in road surface investigation, the present invention uses a 10 - 80 GHz radar to monitor vehicles driving on the road surface. When a vehicle encounters road surface defects during driving, obvious abnormal vibrations, jitters or body tilts will inevitably occur. When the vehicle body shows abnormalities, for the radar of the detection vehicle, the abnormalities will generate waveforms significantly different from normal driving. Therefore, by monitoring the body posture of the vehicle with the radar, the abnormal road surface conditions can be quickly obtained, and then passing vehicles can be reminded in time and the abnormal road surface can be repaired. Since the monitoring distance of 10 - 80 GHz is 150 - 4750 meters, it can achieve road surface monitoring in a large area. At the same time, darkness has no impact on its detection, and the impact of rain and fog is also controllable. It has the advantages of low cost and being suitable for popularization.

[0033] 2. The waveform data processing technology can obtain the required waveforms faster and more accurately by combining pre - processing, fast Fourier transform to extract the 50 - 200 Hz frequency band, or even the 100 - 200 Hz frequency band, synchrosqueezing wavelet transform, adaptive threshold filtering, impact feature enhancement and three - dimensional feature encoding, reducing the amount of waveform data processing and improving the processing efficiency.

[0034] 3. On the basis of monitoring vehicles driving on the highway with radar, a defect map feature library of the highway surface is constructed. It finds or replicates the road surface defects on the highway and uses the waveforms collected by the radar collector to form the defect map feature library. And this feature library is gradually improved with further use. At the same time, the risk levels of the defect map feature library are classified. Thus, the real - time wave change features converted from the abnormal movements of the vehicles collected in real - time can be compared with the defect map feature library, so that the computer can judge the risk level of the road surface defect, and then give a quick warning. At the same time, the supervision center in the background can also infer the road surface defect situation through the real - time wave change features and quickly propose countermeasures, such as preparing repair materials, etc.

[0035] 4. Since the wave change features generated by different vehicle speeds and vehicle types when facing different road surface defects are different, when constructing the defect map feature library of the highway surface, the wave change features of different vehicle types at different vehicle speeds are collected, which can more accurately and quickly realize the warning judgment.

[0036] 5. The warning device is installed near the signal acquisition module. Its warning includes the most direct and eye - catching combination of sound, light and electricity, and also includes automatically occupying the 87.5 - 108 MHz frequency band of the vehicle radio to play warning voices, so as to comprehensively remind the nearby vehicles and reduce the occurrence of accidents.

[0037] Other advantages, objects, and features of the present invention will be set forth in part in the following description, and in part will be obvious to those skilled in the art upon examination of the following, or may be learned by practice of the present invention. The objects and other advantages of the present invention may be realized and obtained by the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a logic block diagram of Embodiment 1 of the present invention;

[0039] Figure 2 It is a logic block diagram of Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the tables in the embodiments of the present invention. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] Embodiment 1. A monitoring and warning system for monitoring highway pavement defects using radar, where the highway pavement defects include pavement settlement, pavement collapse, pavement potholes, pavement cracking, and pavement heaving. To achieve the detection of its pavement defects, such as Figure 1 , the system is composed of the following functional modules:

[0042] 1. Signal acquisition module, including radar and Beidou time synchronization, which is installed on the crossbar above the highway or the vertical pole on the roadside. The radar includes a transmitter, a transmitting antenna, a receiver, a receiving antenna, and a signal processing unit, and is used to acquire the real-time reflected wave signals of vehicles driving on the highway pavement. Beidou time synchronization realizes multi-radar time synchronization. In the present invention, the signal acquisition module is installed on the gantry above the highway pavement or the top of the vertical pole on the roadside. The layout position of the gantry is preferably selected in the high backfill area prone to settlement, with a vertical height greater than 6 meters. The position information of each crossbar or roadside vertical pole where the signal acquisition module is installed is calibrated, that is, the coordinate information of the signal acquisition module. The real-time reflected wave signals of vehicles driving on the highway pavement are acquired using the radar, and the vehicle speed is measured using the Doppler effect. Beidou time synchronization realizes multi-radar time synchronization, so that the real-time reflected wave signals of vehicles have coordinate information and time stamps.

[0043] The requirements for its radar are as follows:

[0044] ‌1.1 Radar frequency band and modulation method‌

[0045] Recommended frequency band: 24 GHz (K band). This frequency band has low atmospheric attenuation characteristics in long-distance detection, supports long-distance signal penetration, and the radio frequency chip technology is mature. It is necessary to adopt FMCW (Frequency Modulated Continuous Wave) modulation and improve the ranging range by dynamically adjusting the chirp slope.

[0046] 1. 2. Core performance parameters

[0047] Detection distance: ≥ 1 km. It is necessary to configure a high transmit power (≥ 20 dBm) and adopt a MIMO antenna array to improve the signal-to-noise ratio. Vertical resolution: Elevation angle resolution: ≤ 5°. Capture the up and down jitter of the vehicle (amplitude ≥ 5 cm) through multi-antenna beamforming technology.

[0048] 1. 3. Antenna and hardware design

[0049] Antenna configuration: MIMO array (such as 8 transmit and 16 receive), expand spatial coverage and enhance vertical direction resolution ability.

[0050] Beam width: Horizontal ≤ 24°, Elevation ≤ 15°. Focus on the target area to reduce interference. Transmit power: ≥ 20 dBm. Combine with a low-noise amplifier (LNA) to enhance the echo signal strength.

[0051] 2. Reflection wave signal extraction module, which is connected to the signal acquisition module, used to obtain the real-time reflection wave signal and its position information, and use waveform data processing technology to extract the characteristics of the real-time reflection wave signal of the vehicle driving on the highway, used to obtain the real-time wave change characteristics when the vehicle is driving non-steadily, and transmit the obtained real-time wave change characteristics and position information to the wave change characteristic comparison module. The waveform data processing technology includes a. Preprocessing denoising, b. Fast Fourier transform to extract the 50 - 200 Hz frequency band, c. Synchrosqueezed wavelet transform for precise positioning, d. Adaptive threshold filtering for noise reduction, e. Impact feature enhancement to improve the signal-to-noise ratio, f. Three-dimensional feature coding adaptation model.

[0052] a. Preprocessing denoising

[0053] Denoising and signal alignment: Separate high-frequency impact components and low-frequency vibrations through dynamic noise suppression (such as Complementary Ensemble Empirical Mode Decomposition CEEMD), and eliminate environmental noise (such as crosswind, electromagnetic interference).

[0054] Beamforming optimization: Use the spatial filtering of the MIMO radar array to suppress the interference signals of lateral vehicles (targets outside ± 5° deviation), and retain the echoes in the main lobe direction (longitudinally driving vehicles).

[0055] Data standardization:

[0056] Perform amplitude normalization (eliminating the influence of distance attenuation) and time-domain alignment (compensating for the phase shift caused by vehicle movement) on the radar intermediate-frequency signal.

[0057] b Fourier transform feature extraction in the 50 - 200 Hz frequency band

[0058] Coarse-grained frequency band screening: Perform a 2048-point FFT on the preprocessed signal, analyze the signal spectrum, and extract the 50 - 200 Hz frequency band related to driving safety (corresponding to the high-frequency features of impact events such as road surface potholes and tire bounces).

[0059] Frequency-domain energy threshold segmentation: Set an energy threshold (e.g., > 3% of the total energy), filter out the low-frequency lateral movement components (< 20 Hz) and ultra-high-frequency noise (> 200 Hz), and retain the potentially hazardous frequency band.

[0060] Spectrum baseline calibration: Highlight the sudden impact signal (such as a suddenly appearing road foreign object) through background spectrum difference (current frame spectrum - historical average spectrum).

[0061] c Synchrosqueezed wavelet transform for precise positioning

[0062] Improve time-frequency resolution: Perform synchrosqueezed wavelet transform on the signal screened by FFT, redistribute the fuzzy energy of the traditional wavelet transform to the instantaneous frequency trajectory, and achieve high-resolution time-frequency analysis of 0.2 ms / 5 Hz.

[0063] Transient impact positioning: Separate short-term impacts (< 30 ms) from continuous vibrations (such as engine vibrations), and locate the precise time points of events such as road surface potholes through the instantaneous frequency f0(t,a) (error < 0.5 ms).

[0064] Suppress lateral interference: The low-frequency energy corresponding to lateral movement (such as the < 20 Hz vibration generated by lane change) is automatically filtered during the SWT reassignment process due to the deviation of the instantaneous frequency from the target frequency band.

[0065] d Adaptive threshold filtering for noise reduction

[0066] Dynamic noise suppression: Calculate the instantaneous energy of the signal based on the Teager energy operator (TEO): E(t) = x2(t) - x(t - 1)x(t + 1), set the adaptive threshold Threshold = 2.5 × median(|E(t)|), and filter out more than 90% of the residual interference.

[0067] Initial screening of impact events: Retain the regions with sudden energy increase (such as the peaks where the TEO amplitude exceeds the threshold), and mark them as potential road hazard events (such as potholes and speed bumps).

[0068] Enhanced e - impact feature to improve signal - to - noise ratio

[0069] Energy focusing: Perform Gaussian energy redistribution on the selected impact signals:

[0070] Eenhanced(t,f)= (4σ Gaussian window weighting)

[0071] By aggregating the neighborhood energy with 4σ weighting, the concentration of impact energy is increased by 45%, and the signal - to - noise ratio is increased by more than 22 dB, highlighting waveform details (such as the rising - edge slope, oscillation decay rate). Eenhanced(t,f) or E(t,f): This represents the enhanced signal or energy, which is a function of time and frequency. : Represents the sum over all integer values of the variable k from - 4 to 4. This range limits the values of the frequency offset.

[0072] k: Is the index variable for the sum, representing the index of the frequency offset. In this formula, the value of k ranges from - 4 to 4, used to traverse different frequency offsets.

[0073] Tswt(t,f + kΔf): This represents a certain transformation or transform function at time t and frequency f + kΔf. This function may represent a certain characteristic or transform result of the signal at a specific time and frequency.

[0074] : This is an exponential function used to weight Tswt(t,f + kΔf). As k increases or decreases, this weighting term gradually decreases, thus achieving different degrees of weighting for signals with different frequency offsets.

[0075] Δf: This represents the frequency interval or frequency resolution, which is the step size of the frequency offset. It determines the value interval of the frequency offset kΔf during the summation process.

[0076] Waveform morphology restoration: Compensate for the attenuation distortion during signal transmission and restore the true impact waveform (such as the force - time curve of a road surface pothole).

[0077] f three - dimensional feature encoding and adaptation model

[0078] Multi - dimensional feature fusion: Encode the time - domain, frequency - domain, and energy - domain features into a three - dimensional vector: F = [t,f,E(t,f)] (time, frequency, energy density), and expand it into high - order features (such as kurtosis, envelope spectrum entropy, instantaneous frequency slope) to form a 256 - dimensional feature matrix.

[0079] Dimensionality reduction and classification adaptation: Use principal component analysis (PCA) to compress the features to 18 dimensions (retaining 95% of the information), and adapt to subsequent machine learning models (such as SVM, CNN) for classifying road surface conditions (smooth / crack / pit).

[0080] 3. High-speed road surface defect atlas feature library. a. Collect known high-speed road surface defects that affect vehicle driving to establish a road surface defect library. b. Find the high-speed road surface defects in the road surface defect library on the existing high-speed sections or temporarily build sections to replicate the high-speed road surface defects in the road surface defect library. c. Set a signal acquisition module at the road surface defects found or replicated in step b, and use the signal acquisition module to obtain the wave change characteristics during non-steady driving of different types of vehicles at different speeds when driving to the high-speed road surface defects found or replicated in step b, and collect the wave change characteristics to form a high-speed road surface defect atlas feature library; and classify the risk levels of the high-speed road surface defect atlas feature library, and set the wave change characteristic threshold; when classifying the risk levels of the high-speed road surface defect atlas feature library, it is based on road surface defects and is classified manually through test data by humans. The same road surface defects on the driving road surface belong to the same risk level. Storage architecture of the high-speed road surface defect atlas feature library: Use a MongoDB sharded cluster to store time-series data, supporting PB-level data expansion; Retrieval optimization: Build a feature vector index based on the Faiss framework, and the retrieval latency < 5ms.

[0081] Data acquisition plan

[0082] Stage Content Implementation method Phase I Basic defect sample library 50 types of defects × 10 vehicle speeds × 8 vehicle models Phase II Environment enhancement library Rain / fog / night scene simulation and reproduction Phase III Online learning and update LSTM dynamic correction of feature thresholds

[0083] The risk levels of the high-speed road surface defect atlas feature library are divided into the general-level high-speed road surface defect atlas feature library, the slow-down passing-level high-speed road surface defect atlas feature library, and the dangerous-level high-speed road surface defect atlas feature library. Among them, the general level refers to road surface defects that will cause vehicle body vibration and jitter but do not affect vehicle safety. The slow-down passing level refers to road surface defects that can be safely passed even below the specified speed; the dangerous level refers to road surface defects that will cause vehicle damage and accidents, such as road surface fractures and collapses. Deploy a deep convolutional network (ResNet-18 architecture), and the training sample size ≥ 100,000 groups of abnormal vibration waveforms

[0084] Risk level Trigger condition (example) Corresponding defect type General level The standard deviation of amplitude Δσ < 0.3g Slight crack (width < 2 cm) Deceleration level 0.3 g ≤ Δσ < 0.8 g Medium pothole (depth > 5 cm) Danger level Δσ≥0.8g or tilt angle > 5° for 2 seconds Collapsed area (area > 1㎡)

[0085] 4. Wave change feature comparison module, which is connected to the high-speed road surface defect atlas feature library, the reflected wave signal extraction module, and the warning device. After receiving the real-time wave deformation feature and position information obtained from the reflected wave signal extraction module, it analyzes and compares with the wave change features in the high-speed road surface defect atlas feature library. After the analysis and comparison, if the real-time wave deformation feature is within the alarm level threshold, it sends a signal to the warning device at the corresponding position; if the real-time wave deformation feature is within the general-level high-speed road surface defect atlas feature library area, it does not send a signal to the warning device, but only sends it to the supervision center. If the real-time wave deformation feature is within the deceleration passing-level high-speed road surface defect atlas feature library, it sends a deceleration prompt signal to the warning device at the corresponding position; if the real-time wave deformation feature is at the lowest threshold of the high-risk level, it sends an emergency stop prompt signal to the warning device at the corresponding position.

[0086] Among them, the wave change feature comparison module adopts real-time comparison technology, including Dynamic Time Warping (DTW): matching the real-time waveform with the feature library template, tolerating time axis stretching and speed differences. Deep learning assistance: training a 1D-CNN model (structure: Conv1D(64)-MaxPool-GRU(32)-Dense) to classify waveform segments, with a classification accuracy of ≥96%.

[0087] 5. Warning device, which is installed next to or 30 - 50 meters in front of the signal acquisition module. It sends warnings to vehicles driving into this area by means of warning lights, loudspeakers, and / or playing warning voices on the 87.5 - 108 MHz frequency band of the automatic occupancy vehicle radio; a text display screen or an electronic speed limit sign can also be installed on the gantry with a radar to give a clear and direct warning.

[0088] 6. Supervision center, which can obtain the real-time wave change feature and its position information obtained by the reflected wave signal extraction module, perform waveform processing and analysis on the real-time wave change feature, and use the high-speed road surface defect atlas feature library to invert the road surface defect corresponding to the real-time wave change feature, and send the obtained road surface defect and the position information of the road surface defect to the road maintenance department for maintenance; the supervision center can also obtain the signal sent by the wave change feature comparison module to the warning device, and send the confirmed signal sent to the warning device to the road administration, traffic police, and road maintenance departments. The calculation of the position where the road surface defect is located can be calculated through the position information of the signal acquisition module, the vehicle speed measured by the Doppler effect, and the time when the road surface defect occurs.

[0089] Example 2. As Figure 2, The difference between the monitoring and warning system for high-speed road surface defects using radar and Example 1 lies in: the high-speed road surface defect atlas feature library, which is divided into two categories. One is the high-speed road surface defect atlas feature library with 100% content installed in the supervision center, and this high-speed road surface defect atlas feature library will also be supplemented according to new road surface defects. The other is the wave change feature comparison module installed near the radar, which only installs the high-speed road surface defect atlas feature library for slow passage level and the high-speed road surface defect atlas feature library for dangerous level. This can reduce the amount of comparison data in the road side wave change feature comparison module and improve the efficiency of comparison data.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A monitoring and warning system for monitoring highway pavement defects using radar, characterized in that, The described highway pavement defects include pavement settlement, pavement collapse, pavement potholes, pavement cracking, and pavement bumps. The system includes the following functional modules: The signal acquisition module, including a radar and a Beidou time synchronization module, is installed on the crossbar above the highway or the roadside vertical pole. The radar includes a transmitter, a transmitting antenna, a receiver, a receiving antenna, and a signal processing unit, and is used to acquire the real-time reflected wave signals of vehicles driving on the highway pavement. The Beidou time synchronization module realizes multi-radar time synchronization; The reflected wave signal extraction module uses waveform data processing technology to extract the features of the real-time reflected wave signals of vehicles driving on the highway pavement obtained, and is used to obtain the real-time wave change features when the vehicle is driving non-steadily. The obtained real-time wave change features are transmitted to the wave change feature comparison module. The described waveform data processing technology includes a. preprocessing denoising, b. fast Fourier transform to extract the 50-200 Hz frequency band, c. synchrosqueezing wavelet transform for precise positioning, d. adaptive threshold filtering for noise reduction, e. impact feature enhancement to improve the signal-to-noise ratio, f. three-dimensional feature coding adaptation model; The highway pavement defect atlas feature library: a. Collect known highway pavement defects that affect vehicle driving to establish a pavement defect library; b. Search for the highway pavement defects in the pavement defect library on the existing highway section or temporarily build a section to replicate the highway pavement defects in the pavement defect library; c. Set the signal acquisition module at the pavement defects found or replicated in step b. Use the signal acquisition module to acquire the wave change features when the vehicle is driving non-steadily when driving to the highway pavement defects found or replicated in step b, and collect the wave change features to form a highway pavement defect atlas feature library; and classify the risk levels of the highway pavement defect atlas feature library, and set the wave change feature threshold; The wave change feature comparison module is connected to the highway pavement defect atlas feature library, the reflected wave signal extraction module, and the warning device. When it receives the real-time wave deformation features obtained in the reflected wave signal extraction module, it analyzes and compares them with the wave change features in the highway pavement defect atlas feature library. If the real-time wave deformation features are within the alarm level threshold after the analysis and comparison, a signal is sent to the warning device; The warning device is installed next to or 30-50 meters in front of the signal acquisition module, and sends a warning to the vehicles driving in this area by means of warning lights, loudspeakers, and / or radio voice warnings.

2. The monitoring and warning system for monitoring highway pavement defects using radar according to claim 1, wherein: The system also includes a supervision center, which can obtain the position of the signal acquisition module and the real-time wave change features obtained by the reflected wave signal extraction module, perform waveform processing and analysis on the real-time wave change features, and use the highway pavement defect atlas feature library to invert the pavement defects corresponding to the real-time wave change features, and send the obtained pavement defects and approximate positions to the pavement maintenance department for maintenance; the supervision center can also receive the signal sent by the wave change feature comparison module to the warning device, and after calibrating and confirming the position of the signal sent to the warning device, send it to the road administration, traffic police, and pavement maintenance departments.

3. The monitoring and warning system for detecting highway pavement defects using radar according to claim 2, characterized in that: In the described waveform data processing technology: a. Preprocessing denoising includes denoising and signal alignment: separating high-frequency impact components and low-frequency vibrations through dynamic noise suppression to eliminate environmental noise; beamforming optimization: using the spatial filtering of the MIMO radar array to suppress interference signals from lateral vehicles and retain echoes in the main lobe direction; data normalization: performing amplitude normalization on the radar intermediate-frequency signal to eliminate range attenuation effects and time-domain alignment, and compensating for phase shifts caused by vehicle movement. b. Fast Fourier transform to extract the 50 - 200 Hz frequency band includes coarse-grained frequency band screening: performing a 2048-point FFT on the preprocessed signal to analyze the signal spectrum and extract the 50 - 200 Hz frequency band related to driving safety; frequency-domain energy threshold segmentation: setting an energy threshold to filter out low-frequency lateral motion components <20 Hz and ultra-high-frequency noise >200 Hz, and retaining potentially hazardous frequency bands; spectral baseline calibration: highlighting sudden impact signals through background spectral difference. c. Synchrosqueezed wavelet transform for precise positioning includes time-frequency resolution improvement: performing synchrosqueezed wavelet transform on the signal screened by FFT, redistributing the fuzzy energy of the traditional wavelet transform to the instantaneous frequency trajectory, and achieving high-resolution time-frequency analysis of 0.2 ms / 5 Hz; transient impact positioning: separating short-term impacts <30 ms and continuous vibrations, and precisely locating the time points of events such as road surface potholes through instantaneous frequency. d. Adaptive threshold filtering for noise reduction includes dynamic noise suppression: calculating the instantaneous energy of the signal based on the Teager energy operator, setting an adaptive threshold, and filtering out more than 90% of the residual interference; initial screening of impact events: retaining areas with sudden energy increase and marking them as potential road surface hazard events. e. Impact feature enhancement to improve the signal-to-noise ratio includes energy focusing: performing Gaussian energy redistribution on the screened impact signals, aggregating neighboring energy through weighting, increasing the impact energy concentration by 45% and improving the signal-to-noise ratio by more than 22 dB, and highlighting waveform details; waveform morphology restoration: compensating for attenuation distortion during signal transmission and restoring the true impact waveform. f. Three-dimensional feature encoding for model adaptation includes multi-dimensional feature fusion: encoding time-domain, frequency-domain, and energy-domain features into three-dimensional vectors, expanding them into high-order features to form a 256-dimensional feature matrix; dimensionality reduction and classification adaptation: using principal component analysis to compress the features to 18 dimensions for adaptation to subsequent machine learning models.

4. The monitoring and early warning system for monitoring high-speed road surface defects using radar according to claim 3, characterized in that: The risk levels of the high-speed road surface defect atlas feature library are divided into the general-level high-speed road surface defect atlas feature library, the slow-down passing-level high-speed road surface defect atlas feature library, and the dangerous-level high-speed road surface defect atlas feature library.

5. The monitoring and early warning system for monitoring the deformation of the high-speed road surface using radar according to claim 4, characterized in that: In the wave change feature comparison module, the slow-down passing-level high-speed road surface defect atlas feature library and the dangerous-level high-speed road surface defect atlas feature library are installed. If the real-time wave deformation feature is in the slow-down passing-level high-speed road surface defect atlas feature library, a slow-down prompt signal is sent to the warning device; if the real-time wave deformation feature is at the lowest threshold of the high-risk level, an emergency stop prompt signal is sent to the warning device.

6. The monitoring and early warning system for monitoring highway pavement defects using radar according to any one of claims 1-5, characterized in that: The radar described above uses a multi-input multi-output (MIMO) antenna array with a frequency of 10 - 80 GHz. By means of beamforming technology, the main lobe energy is concentrated in the vehicle driving direction, suppressing lateral interference signals.

7. The monitoring and warning system for monitoring highway pavement defects using radar according to claim 6, characterized in that: When the signal acquisition module described above acquires the wave change characteristics during non-steady driving when the vehicle travels to the high-speed road surface defects found or replicated in step b, it includes the wave change characteristics formed by different types of vehicles and speeds. The collected high-speed road surface defect map feature library includes the wave change characteristics of different types of vehicles at different speeds.

8. The monitoring and early warning system for monitoring highway pavement defects using radar according to claim 7, characterized in that: When grading the risk level, the high-speed road surface defect map feature library is graded manually based on road surface defects through test data.

9. The monitoring and warning system for monitoring high-speed road surface defects using radar according to claim 8, characterized in that: The warning device described above also includes a text display screen or an electronic speed limit sign.

Images