A highway pavement flatness detection method

Abstract

The application relates to the technical field of road detection, and discloses a highway pavement flatness detection method and system, which aims to solve the problems that the existing detection is greatly disturbed by vehicle speed and vehicle body vibration, is low in efficiency, and cannot cover multiple lanes. The method comprises the following steps: acquiring a pavement image through an image sensor and extracting contour lines and texture features, and inferring the contour of an adjacent lane based on geometric mapping; acquiring vertical motion data by using an inertia detection device independent of a suspension system, and combining the instantaneous vehicle speed to normalize and correct the contour deviation characteristic quantity; determining the flatness index by using a mapping model, and performing multi-source consistency verification and fault diagnosis by fusing the image, inertia and vehicle suspension data. The system comprises an image acquisition module, a data processing module, an inertia detection device and a vehicle data interface module. The application realizes synchronous high-precision detection of multiple lanes, effectively removes the mechanical vibration interference of the vehicle body, and significantly improves the reliability and operation efficiency of dynamic detection.

Description

Technical Field

[0001] This invention relates to road inspection methods, and more particularly to a method for inspecting the smoothness of highway pavement. Background Technology

[0002] With the rapid development of the transportation industry, highways, as the lifeline of modern logistics and travel, have made real-time monitoring and maintenance decisions regarding road surface quality a core element in ensuring driving safety and improving the service level of the road network. Road surface smoothness, as the most direct indicator of road performance, not only affects vehicle ride comfort and fuel economy but is also a key basis for evaluating construction quality and developing preventative maintenance plans.

[0003] Among them, using vehicle-mounted sensors for dynamic smoothness detection is an important direction in the current construction of smart highways. By collecting the changes in physical quantities during vehicle travel and calculating the international smoothness index in real time, it aims to achieve low-cost, high-frequency, and large-scale dynamic surveys of road surface quality.

[0004] Existing technologies for road surface evenness detection suffer from the following main drawbacks: First, their detection efficiency and coverage are limited. Traditional detection equipment often relies on specialized vehicles and can only operate on a single lane, lacking consideration for the spatial correlation between adjacent lanes, making it difficult to meet the needs of simultaneous multi-lane detection and comprehensive cross-sectional evaluation. Second, detection accuracy is severely affected by environmental and driving conditions. Relying solely on visual contour extraction is susceptible to the influence of vehicle driving posture, suspension system vibration, and speed fluctuations, leading to systematic biases in the detection results. Third, they lack effective independent reference and compensation mechanisms. Without an inertial reference unit detached from the suspension system, it is impossible to effectively eliminate motion interference from the vehicle itself, and the lack of cross-verification logic between different sensor data sources makes it difficult to identify sensor malfunctions or algorithm drift, easily resulting in distorted detection results. Finally, the system architecture is complex and lacks versatility, making lightweight deployment on ordinary operating vehicles difficult. These problems collectively lead to insufficient reliability of existing detection solutions in dynamic and complex scenarios. Therefore, how to achieve a high-precision, multi-lane, synchronous, and self-verifying dynamic road surface evenness detection method is a pressing technical challenge that needs to be addressed. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a digital parking space intelligent parking guidance method to overcome the above-mentioned defects in the existing technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] 1. A method for detecting the smoothness of highway pavement, characterized in that it includes: Step 1: Using an image sensor deployed at the rear of the vehicle, continuously acquire original road surface images of the current driving lane and its adjacent areas on both sides at a preset sampling frequency during vehicle operation, and transmit them to the data processing module. Step 2: Preprocess the original road surface image to extract the longitudinal contour line of the current lane; extract the road surface texture features, establish a geometric mapping model between the current lane and the adjacent lanes on both sides, and infer the road surface contour at the same position of the adjacent lanes on both sides by combining the spatial continuity of the road surface texture features. Step 3: Using an inertial detection device independent of the vehicle suspension system, the vertical acceleration and displacement changes of the vehicle are sensed in real time to generate vertical inertial motion data. Step 4: Establish the spatial relationship between the longitudinal contour line of the road surface and the preset horizontal baseline, and calculate the vertical deviation value and deviation characteristic quantity of the contour line sampling points; Step 5: Obtain the instantaneous vehicle speed, and combine the correlation between vertical inertial motion data and instantaneous vehicle speed to compensate and correct the deviation feature quantity, thereby obtaining the equivalent deviation feature quantity at the standard detection speed. Step 6: Input the equivalent deviation feature quantity into the preset mapping model and output the international smoothness index of the current lane and the adjacent lanes on both sides. Step 7: Synchronously collect displacement data of the vehicle's original suspension system, compare multiple sets of intermediate flatness values, and generate a system abnormality prompt when the difference exceeds the preset tolerance.

[0008] 2. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that the image sensor is installed on the central axis of the rear of the vehicle, and the optical axis forms a preset downward angle with the horizontal plane; the image sensor is equipped with a self-cleaning actuator, including a high-pressure air pump and a precision nozzle surrounding the edge of the lens; the data processing module calculates the structural similarity index and edge sharpness index of the original pavement image in real time, and when the structural similarity index is lower than a preset quality threshold, it drives the high-pressure air pump to spray pulsed airflow through the precision nozzle.

[0009] 3. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, in step 2, the preprocessing of the original pavement image includes denoising processing using a preset-size window; extracting the longitudinal contour line of the current lane pavement includes calculating the longitudinal gradient amplitude of the image, identifying the feature edge points formed by pavement undulations, fitting the feature edge points, and generating a geometric contour line reflecting the longitudinal fluctuations of the pavement; during edge detection, the high and low thresholds are dynamically determined based on the variance of the local area of ​​the image to distinguish between the pavement background and the weak shadow edges generated by pavement undulations.

[0010] Preferably, in step 3, identifying road surface texture features includes extracting road surface energy, entropy, moment of inertia, and correlation feature vectors, transforming the original road surface image, extracting spatial spectrum distribution features, and determining road surface roughness by analyzing the energy ratio of high-frequency components; establishing a geometric mapping model includes performing image perspective transformation through a preset calibration relationship to convert the original road surface image into a top-view orthophoto; inferring the road surface contour of adjacent lanes includes using the current lane texture features as a reference, performing correlation search in the corresponding area of ​​adjacent lanes, and reconstructing the longitudinal geometric shape of the same mileage position of adjacent lanes by combining the depth mapping relationship.

[0011] Preferably, in step 4, the sealed tube of the inertial detection device is vertically fixed to the rigid connecting part of the vehicle chassis. The movable working fluid is displaced under the action of the vertical motion inertia of the vehicle, and the displacement sensor array captures the displacement. The sealed tube is filled with a damping fluid of a preset kinematic viscosity to filter out high-frequency mechanical vibration interference and retain low-frequency vibration characteristics related to road surface smoothness. The inertial detection device has a built-in temperature sensor, and the data processing module adjusts the damping coefficient compensation factor when calculating the vertical displacement integral based on the real-time temperature value to perform temperature compensation on the vertical inertial motion data.

[0012] Preferably, in step 5, the preset horizontal baseline is generated by mean filtering of the road surface contour line of preset length, the vertical deviation value is the normal distance from the contour line sampling point to the baseline, and the deviation feature is the root mean square value of the vertical deviation of the sampling point within the preset evaluation interval; in step 6, the vehicle speed compensation function is fitted by multiple sets of constant speed experiments, and when the instantaneous vehicle speed deviates from the standard detection speed, the deviation feature is corrected by weighting using this function; in step 7, the mapping model adopts a multi-layer network structure, the input layer receives the equivalent deviation feature, the road surface texture feature vector and the ambient lighting parameters, and after processing by the hidden layer, the output layer outputs the international roughness index.

[0013] The detection system for highway pavement smoothness detection includes: an image acquisition module deployed on the central axis of the rear of the vehicle, comprising a high dynamic range photosensitive sensor and a wide-angle lens, used to acquire original images of the current lane and adjacent areas on both sides of the road surface at a preset sampling frequency; an inertial detection device installed on the vehicle body and independent of the suspension system, comprising a sealed tube, damping fluid, movable working fluid, and a displacement sensor array, used to sense the vehicle's vertical acceleration and displacement changes in real time and generate vertical inertial motion data; a vehicle data interface module connected to the vehicle control local area network bus, used to acquire the vehicle's instantaneous speed and displacement data of the original suspension system; and a data processing module communicatively connected to the above modules, with a built-in digital signal processor and field-programmable gate array, used to execute the detection method. The two modules exchange data through a preset bus and are respectively responsible for image preprocessing, inertial data acquisition, and related algorithm operation.

[0014] Preferably, the diagnostic unit of the data processing module has a sensor state prediction function. By recording the output fluctuation trajectory of the sensor within a preset time period and calculating the state transition probability, it can provide early warning of performance degradation trends. Multiple sets of flatness intermediate values ​​include three types of intermediate values ​​determined based on image, inertial detection device, and suspension system data. The data processing module executes a weighted fusion algorithm. When the relative deviation between any two sets of intermediate values ​​exceeds a preset threshold, the sensor is determined to be abnormal and the fault source is located. The data processing module runs on a real-time operating system and processes data in parallel through multi-threading, isolating each core thread and allocating different processor core affinity.

[0015] As a preferred option, the system also includes a high-precision positioning module and a wireless transmission unit. The high-precision positioning module acquires the real-time coordinates of the vehicle, and the wireless transmission unit uploads the multi-dimensional detection data to the cloud management platform. The cloud management platform uses the overlapping detection data of multiple vehicles for secondary cross-verification and eliminates random errors through statistical averaging. The cloud platform also has an automatic defect classification function. For road sections where the international roughness index exceeds the preset threshold, it extracts road surface texture features and identifies defect types, generating a detection report containing information such as geographical location and lane number.

[0016] The beneficial effects of this invention are as follows: By introducing an inertial detection device independent of the vehicle's suspension system, this invention constructs an independent reference system free from the interference of vehicle body mechanical vibration. Compared to solutions that directly utilize data from the vehicle's own sensors, this invention can accurately isolate the influence of suspension system damping characteristics on road surface perception. Combined with a vehicle speed compensation algorithm, it achieves high-precision detection over a wide speed range, effectively controlling detection errors. Simultaneously, the multi-source data fusion verification mechanism enables closed-loop comparison of image, inertial, and onboard sensor data, effectively solving the problem of algorithm drift caused by environmental interference from a single sensor source, significantly improving the system's fault identification rate. This invention utilizes the wide-angle characteristics of image sensors and texture inference algorithms to break the limitation of traditional detection equipment that can only cover a single lane. By establishing a spatial geometric mapping model between the current lane and adjacent lanes, it enables the simultaneous acquisition of smoothness data for multiple lanes in a single trip, greatly improving detection efficiency. This multi-lane simultaneous inference technology not only reduces detection costs but also provides road maintenance departments with a more comprehensive road surface quality evaluation dimension through cross-sectional comprehensive evaluation indicators, meeting the needs of large-scale, high-frequency dynamic surveys of highways. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall technical solution architecture proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework for calculating the equivalent deviation feature quantity based on vehicle speed normalization in this invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings: This embodiment provides a method for detecting the smoothness of highway pavement. Its core application scenario is to achieve non-contact, high-frequency dynamic monitoring of pavement quality by highway operating vehicles under normal driving conditions. To achieve this goal, this embodiment first constructs a highly integrated mechatronics detection system.

[0022] In terms of system architecture, the hardware system in this embodiment mainly consists of an image acquisition module, an inertial detection device, a data processing module, and a vehicle data interface module. The image acquisition module is deployed on the central axis of the vehicle's rear. Its core component includes a high dynamic range complementary metal-oxide-semiconductor (CMOS) photosensitive sensor. This sensor has 8 million effective pixels, capable of capturing minute cracks, potholes, and shadow changes caused by undulations in the road surface. The image acquisition module acquires its field of view through a wide-angle lens with autofocus. The field of view of this lens is set to 170 degrees, ensuring coverage of the complete road surface area of ​​the current lane and one adjacent lane on each side. To suppress the impact of high-frequency vibrations during vehicle movement on image quality, the photosensitive sensor integrates a three-axis optical image stabilization compensation mechanism. This mechanism uses a miniature piezoelectric actuator to adjust the displacement of the photosensitive element in real time, compensating for minute vibrations between 20 Hz and 100 Hz.

[0023] The inertial sensing device is the core hardware for achieving an independent reference frame in this system. Designed as a fully sealed tubular structure, it is fixed to the rigid beam of the vehicle chassis, allowing direct sensing of the vehicle's physical motion relative to the road surface, unaffected by the nonlinear deformation of the suspension system's springs and shock absorbers. The sealed tube is filled with silicone oil of 500 centipoise as a damping fluid, and contains a precisely calibrated movable working fluid. When the vehicle experiences vertical acceleration over uneven road surfaces, the movable working fluid undergoes relative displacement within the tube. A highly sensitive Hall displacement sensor array surrounds the outer wall of the tube, capable of capturing real-time positional changes of the working fluid with an accuracy of 0.01 millimeters. This fluid-damped inertial sensing structure naturally filters out high-frequency mechanical noise generated by the vehicle's engine, preserving low-frequency vibration characteristics from 0.5 Hz to 30 Hz, which are closely related to road surface smoothness.

[0024] The data processing module, as the system's computing core, employs a heterogeneous computing architecture comprised of a high-performance digital signal processor (DSP) and a field-programmable gate array (FPGA). The FPGA handles real-time preprocessing of image signals and synchronous acquisition of inertial data, while the DSP runs the core computer vision algorithms and deep learning prediction models. Both exchange data at high speed via a standard interconnect bus for peripheral components. The vehicle data interface module connects to the vehicle's electronic control unit (ECU) via a control area network (LAN) bus, reading real-time values ​​of the vehicle's instantaneous speed, longitudinal acceleration, and the values ​​from the original suspension displacement sensors.

[0025] Regarding the system workflow, the specific operational steps for running the system are as follows: Step 1: The image acquisition module continuously acquires raw road surface images at a sampling frequency of 120 frames per second. Each frame, immediately after generation, undergoes grayscale conversion and histogram equalization by a field-programmable gate array (FPGA) to eliminate brightness differences under varying lighting conditions. The processed data is then transmitted in real-time to the memory buffer of the data processing module.

[0026] Step 2: The data processing module extracts the longitudinal contour line of the road surface. This process first calls the median filter operator to denoise the image with a 5x5 window size, and then uses an improved Sobel operator to calculate the gradient magnitude of the image in the longitudinal direction. Through dual-threshold segmentation logic, the system can identify the feature edge points formed by the undulations of the road surface. In order to transform the scattered edge points into continuous geometric curves, the system uses the least squares fitting method to perform polynomial fitting on a preset longitudinal evaluation line, thereby generating a geometric contour line that reflects the longitudinal fluctuation state of the road surface.

[0027] Step 3: Identify road surface texture and infer adjacent lane contours. The system extracts four-dimensional feature vectors of the current lane's road surface, including energy, entropy, and moment of inertia, using a gray-level co-occurrence matrix algorithm. These vectors represent the road surface roughness information. Based on a pre-calibrated homography matrix, the system performs image perspective transformation, converting the tilted image into an orthophoto from a top-down perspective. Using the texture features of the current lane as a reference template, the system performs correlation searches in the corresponding regions of adjacent lanes. When the texture similarity of the searched regions is greater than 0.85, the system combines the depth mapping relationship in the perspective transformation to infer the longitudinal geometric contours of adjacent lanes at the same mileage position. This step achieves a technological breakthrough by covering three lanes in a single detection.

[0028] Step 4: Acquire vertical inertial motion data. Simultaneously with image processing, the inertial detection device captures the displacement signal of the movable working medium via a Hall displacement sensor. This signal, after analog-to-digital conversion, is integrated in real-time by a field-programmable gate array (FPGA). The vertical velocity is obtained through a single integration, and the dynamic displacement of the vehicle body relative to its equilibrium position is obtained through a second integration. These data are precisely aligned with the contour lines extracted from the image in terms of timestamps, with the sampling time error controlled within 1 millisecond.

[0029] Step 5: Calculate the deviation feature. The system first applies mean filtering to the road surface contour within a continuous 100-meter range to generate a virtual horizontal baseline representing an ideally flat state. Then, it calculates the normal deviation of each sampling point on the extracted longitudinal contour line relative to this baseline. Within each evaluation interval of 10 to 50 meters, the system calculates the root mean square value of all deviation values; this value serves as the original deviation feature characterizing the micro-undulations of the road surface.

[0030] Step 6: Dynamic Error Correction and Vehicle Speed ​​Normalization. Since the vibration response of a vehicle varies when passing the same pothole at different speeds, the system uses the instantaneous driving speed obtained from the vehicle data interface module to call a preset vehicle speed compensation function. This function was fitted in a controlled experimental environment through multiple sets of constant speed experiments within a range of 20 km / h to 120 km / h. Using this function, the system multiplies the original deviation feature by a speed-related weighting coefficient, converting it into an equivalent deviation feature at a standard detection speed of 80 km / h, thereby eliminating the interference of speed fluctuations on the detection results.

[0031] Step 7: Determine the smoothness index. The system takes the equivalent deviation feature, road texture feature vector, and ambient lighting parameters as input and feeds them into a pre-trained 3-layer artificial neural network model. This model calculates and outputs the international smoothness index for the current lane and its two adjacent lanes through nonlinear mapping logic. The model's training set contains over 5000 sets of measured samples compared with a standard laser smoothness meter, ensuring the authority and accuracy of the output results.

[0032] Step 8: Multi-source data fusion verification and status diagnosis. The system synchronously acquires displacement data from the vehicle's original suspension system. At this point, the system has three data sources: a first intermediate value of flatness determined based on the image, a second intermediate value determined based on the inertial detection device, and a third intermediate value determined based on the suspension system data. The system uses a weighted fusion algorithm to compare the three. If the relative deviation between the three is within a preset tolerance of 15%, the weighted average is taken as the final result; if a data source shows a significant deviation, the diagnostic unit will automatically determine that the sensor may be obstructed or have a mechanical failure, send an abnormality prompt to the operation interface, and mark the data segment as "pending verification" in the final report.

[0033] This embodiment, through the deep integration of the aforementioned electromechanical system and algorithm process, not only improves the detection accuracy, but also greatly expands the applicability and operational efficiency of the dynamic detection method by introducing an independent inertial reference frame and multi-lane inference logic.

[0034] In terms of hardware enhancements, a self-cleaning actuator has been added to the image acquisition module. This mechanism includes a small high-pressure air pump and a set of precision nozzles surrounding the edge of the sensor lens. The self-cleaning actuator is connected to the data processing module via a dedicated solenoid valve. To achieve intelligent triggering, an image quality evaluation algorithm runs internally in the data processing module, which calculates the structural similarity index and edge sharpness index of the image in real time. When a vehicle travels on a dusty road causing dust to accumulate on the lens, or when water stains obscure the lens during rain, if the image structural similarity drops below 0.6, the data processing module immediately drives the high-pressure air pump to spray a pulsed airflow with an instantaneous pressure of 0.5 MPa onto the lens surface through the nozzles, quickly blowing away contaminants.

[0035] For the inertial sensing device, this embodiment introduces an active temperature compensation mechanism. Since the viscosity of the damping fluid inside the sealed tube changes with ambient temperature, affecting the response characteristics of the movable working fluid, a high-precision semiconductor temperature sensor is embedded inside the tube. Based on the real-time sensed temperature value, the data processing module automatically adjusts the damping coefficient compensation factor when calculating the vertical displacement integral. For example, in a low-temperature environment of -20 degrees Celsius in winter, as the damping fluid becomes more viscous, the system automatically increases the gain coefficient of the displacement signal, ensuring that the linearity error of the inertial data is less than 1% across the entire temperature range.

[0036] In terms of deepening the data processing workflow, this embodiment improves the edge detection logic in step 2 by implementing white-boxing. To address the challenge of contour extraction in low-light environments such as nighttime or tunnels, the system introduces an adaptive gamma correction algorithm. Before extracting the longitudinal contour line, the system first calculates the global histogram distribution of the image. If the average gray value is lower than a preset threshold, the slope of the gamma curve is automatically adjusted to enhance the contrast of dark areas. During edge detection, an improved Kenny edge detection operator is used, where the high and low thresholds are no longer fixed values ​​but are dynamically calculated based on the variance of local image regions. This local adaptive thresholding technology can effectively distinguish between the black asphalt background of the road surface itself and the faint shadow edges generated by road undulations, improving the contour recognition accuracy by more than 25% in low-contrast environments.

[0037] Furthermore, this embodiment adds a spatial frequency analysis step to the texture recognition in step 3. The system extracts the spectral distribution characteristics of the road surface in the spatial domain by performing a two-dimensional fast Fourier transform on the road surface image. By analyzing the energy proportion of high-frequency components, the system can more accurately determine whether there is large-area peeling or pockmarked surface damage on the road surface. These frequency domain features are integrated into the texture feature vector, further improving the geometric reconstruction accuracy of cross-lane smoothness estimation.

[0038] In the vehicle speed normalization process in step 6, this embodiment introduces a more complex frequency response function model. Considering the resonance characteristics of the vehicle suspension system under different frequency excitations, the system not only considers the magnitude of the instantaneous vehicle speed but also analyzes the energy distribution of the vertical acceleration signal in the frequency domain. When the excitation frequency generated by the vehicle speed is close to the vehicle body resonance frequency (usually between 1 Hz and 2 Hz), the system uses a nonlinear weighting algorithm to correct the deviation from the characteristic value, thereby effectively eliminating the false fluctuation signal caused by vehicle body resonance.

[0039] Regarding system self-diagnosis, the diagnostic unit in this embodiment possesses a sensor state prediction function based on Markov chains. The system records the output fluctuation trajectory of each sensor over the past hour and calculates the state transition probability to provide early warning of sensor performance degradation trends. For example, when the signal-to-noise ratio of the image sensor is found to be continuously decreasing, the system will prompt maintenance personnel to check lens wear or clean system pressure before the actual failure occurs.

[0040] Through these meticulous enhancements in detail, this embodiment not only implements a detection method but also constructs an intelligent detection entity with environmental awareness and self-maintenance capabilities, ensuring long-term consistency and reliability of smoothness detection data in complex and ever-changing highway environments.

[0041] This embodiment focuses on illustrating the implementation of the present invention in multi-vehicle collaborative operation and big data cloud management, demonstrating how the system has evolved from single-vehicle detection to a digital monitoring system covering the entire road network.

[0042] Regarding system connectivity and communication architecture, this embodiment integrates a wireless transmission unit supporting 5G mobile communication technology and a high-precision global positioning system module with differential positioning functionality into the data processing module. This positioning module, by receiving carrier phase differential correction information provided by a base station, can improve the latitude and longitude positioning accuracy of the vehicle during driving to the centimeter level. The wireless transmission unit maintains a real-time connection with the remote cloud management platform through an encrypted tunnel protocol. Its communication protocol adopts a message queue telemetry transmission protocol to ensure data transmission can resume even in road sections with unstable network signals.

[0043] At the operational level, this embodiment significantly expands the functionality of steps 7 and 8. After determining the flatness index, the system no longer simply outputs a numerical value, but spatially correlates this value with high-precision positioning information. The data processing module generates a multi-dimensional data packet locally containing "time-coordinates-flatness index-texture features-vehicle speed". Every 100 meters, the system automatically encapsulates the data packet and uploads it to the cloud server in real time.

[0044] In the cloud processing stage, the management platform runs a road health record system based on a geographic information system. When it receives detection data from different vehicles, the cloud platform first performs spatial consistency filtering. Because this invention has the ability to cover three lanes in a single trip, the detection areas of different vehicles will overlap when passing through the same road segment. The cloud platform uses this overlapping data for secondary cross-verification, and further eliminates random errors in a single detection through statistical averaging of big data.

[0045] In the status diagnosis of step 8, this embodiment introduces a calibration method based on swarm intelligence. The cloud platform monitors the output data of all detection terminals in the entire road network in real time. If it is found that the detection result of a certain vehicle continuously deviates from the historical average of that road segment, while the data of other vehicles passing through that road segment are normal, the cloud platform will automatically issue a remote calibration command to adjust the compensation coefficient in the data processing module of that vehicle. This cloud-based dynamic calibration mechanism solves the problem of hardware consistency maintenance under large-scale deployment.

[0046] In addition, a data post-processing and decision support step has been added. The system automatically identifies severely uneven road sections with an international roughness index greater than 3.5 meters per kilometer. For these sections, the system extracts the corresponding original pavement images and texture features, and automatically classifies the defects using a cloud-based deep residual network, distinguishing whether the roughness deterioration is caused by pavement settlement, ruts, or potholes. Finally, the system automatically generates an electronic maintenance recommendation report, which not only includes the precise geographical location of the defects but also predicts the quality evolution trend of the road section over the next 6 months based on the rate of decline of the roughness index, providing road management departments with a scientific basis for preventive maintenance decisions.

[0047] In terms of hardware physical characteristics, the image acquisition module in this embodiment uses protective glass with a sapphire coating, achieving a Mohs hardness of 9, which can effectively resist the impact and abrasion from gravel during high-speed driving. The sealing tube of the inertial detection device is made of titanium alloy, which is not only high in strength and lightweight, but also has an extremely low coefficient of thermal expansion, further ensuring structural stability under extreme temperature differences.

[0048] In summary, this embodiment elevates single-point smoothness detection to the level of road network-level asset management through cloud collaboration and spatial big data analysis, fully leveraging the technical advantages of this invention in data fusion and cross-lane detection, and providing a complete data closed-loop solution for the construction of smart highways.

[0049] In the above embodiments, the electrical connections between the system modules all conform to industrial-grade standards. For example, the Gigabit Ethernet connection between the image acquisition module and the data processing module uses shielded Cat6a twisted-pair cable and is equipped with a dedicated ruggedized connector to prevent poor contact or signal loss under long-term severe vibration. The power supply circuit of the data processing module is designed with a wide voltage input range (9V to 36V) DC-DC converter and has comprehensive reverse voltage protection, overcurrent protection, and surge suppression functions, enabling it to be powered directly by the vehicle's battery without the need for an additional regulated power supply.

[0050] At the software level, the data processing module runs on a real-time operating system. Through a multi-threaded parallel processing mechanism, the image extraction thread, inertial acquisition thread, vehicle speed compensation thread, and communication upload thread are isolated. Each thread is assigned a different priority and processor core affinity to ensure the highest possible time determinism for critical smoothness calculation tasks. The system also includes a watchdog timer that automatically restarts the system and restores its previous operating state within 100 milliseconds should a logical deadlock occur during software operation.

[0051] The method and system proposed in this invention cleverly solve the detection accuracy problem caused by suspension vibration, vehicle speed fluctuations, and environmental interference during dynamic vehicle operation by deeply fusing visual contour extraction with an independent inertial reference frame. Through feature analysis of road surface texture and perspective transformation logic, it achieves a breakthrough in efficiency for simultaneous multi-lane detection. Its multi-source data cross-verification mechanism and cloud-based collaborative architecture not only ensure the reliability of single detections but also provide technical support for long-term monitoring at the road network level. This comprehensive innovation, from hardware physical structure to software algorithm logic, represents a high-level technological direction in the field of dynamic detection of highway pavement smoothness, possessing extremely high engineering application value and socio-economic benefits.

[0052] In practical deployments, this system has demonstrated exceptional versatility. Whether installed on large maintenance and inspection vehicles or mounted on ordinary logistics transport vehicles, it can achieve high-precision smoothness data collection through simple parameter calibration. This lightweight and intelligent design approach significantly lowers the barrier to achieving full-line highway monitoring, making it possible to monitor the dynamic quality of every kilometer of road surface in real time.

[0053] Through the detailed description of the above embodiments, those skilled in the art can clearly understand the technical solution and implementation details of the present invention. The present invention is not limited to the specific structural parameters or algorithm models described above. Any equivalent modifications or improvements made using the vision-inertial fusion concept, vehicle speed normalization compensation logic, and multi-source data verification mechanism described in the present invention should be included within the scope of protection of the present invention. The various technical indicators provided in this embodiment, such as pixels, sampling frequency, viscosity value, and number of neurons, are only for more clearly illustrating the implementation effect of the present invention, and not for limiting the scope of protection of the present invention. In practical applications, these parameters can be reasonably adjusted and optimized according to different road grades, detection accuracy requirements, and hardware cost budgets.

[0054] Finally, the highway pavement smoothness detection method and system described in this invention have significant practical implications for improving road service quality, ensuring traffic safety, and optimizing maintenance resource allocation. This high-precision digital detection method can promptly detect early pavement defects, preventing minor defects from escalating into major accidents, thereby significantly extending road service life and reducing overall life-cycle maintenance costs. This is not only a significant advancement in traffic infrastructure monitoring technology but also an indispensable key component in the construction of intelligent transportation systems.

[0055] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the smoothness of highway pavement, characterized in that, include: Step 1: Using an image sensor deployed at the rear of the vehicle, continuously acquire original road surface images of the current driving lane and its adjacent areas on both sides at a preset sampling frequency during vehicle operation, and transmit them to the data processing module. Step 2: Preprocess the original road surface image to extract the longitudinal contour line of the current lane; extract the road surface texture features, establish a geometric mapping model between the current lane and the adjacent lanes on both sides, and infer the road surface contour at the same position of the adjacent lanes on both sides by combining the spatial continuity of the road surface texture features. Step 3: Using an inertial detection device independent of the vehicle suspension system, the vertical acceleration and displacement changes of the vehicle are sensed in real time to generate vertical inertial motion data. Step 4: Establish the spatial relationship between the longitudinal contour line of the road surface and the preset horizontal baseline, and calculate the vertical deviation value and deviation characteristic quantity of the contour line sampling points; Step 5: Obtain the instantaneous vehicle speed, and combine the correlation between vertical inertial motion data and instantaneous vehicle speed to compensate and correct the deviation feature quantity, thereby obtaining the equivalent deviation feature quantity at the standard detection speed. Step 6: Input the equivalent deviation feature quantity into the preset mapping model and output the international smoothness index of the current lane and the adjacent lanes on both sides. Step 7: Synchronously collect displacement data of the vehicle's original suspension system, compare multiple sets of intermediate flatness values, and generate a system abnormality prompt when the difference exceeds the preset tolerance.

2. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, The image sensor is mounted on the central axis of the rear of the vehicle, with the optical axis at a preset downward angle to the horizontal plane; the image sensor is equipped with a self-cleaning actuator, including a high-pressure air pump and a precision nozzle surrounding the edge of the lens; the data processing module calculates the structural similarity index and edge sharpness index of the original road image in real time, and when the structural similarity index is lower than a preset quality threshold, it drives the high-pressure air pump to spray pulsed airflow through the precision nozzle.

3. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, In step 2, the preprocessing of the original road surface image includes denoising using a window of a preset size; extracting the longitudinal contour line of the current lane road surface includes calculating the longitudinal gradient magnitude of the image, identifying the feature edge points formed by road surface undulations, fitting the feature edge points, and generating a geometric contour line that reflects the longitudinal fluctuations of the road surface. During edge detection, high and low thresholds are dynamically determined based on the variance of local regions in the image to distinguish between the road background and the faint shadow edges caused by road undulations.

4. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, In step 3, identifying road surface texture features includes extracting road surface energy, entropy, moment of inertia, and correlation feature vectors, transforming the original road surface image, extracting spatial spectrum distribution features, and determining road surface roughness by analyzing the energy ratio of high-frequency components; establishing a geometric mapping model includes performing image perspective transformation through a preset calibration relationship to convert the original road surface image into a top-view orthophoto; inferring the road surface contour of adjacent lanes includes using the current lane texture features as a reference to perform correlation search in the corresponding area of ​​adjacent lanes, and reconstructing the longitudinal geometric shape of the same mileage position of adjacent lanes by combining the depth mapping relationship.

5. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, In step 4, the sealed tube of the inertial detection device is vertically fixed to the rigid connecting part of the vehicle chassis. The movable working medium is displaced under the action of the vertical motion inertia of the vehicle, and the displacement sensor array captures its offset. The sealed tube is filled with a damping fluid of a preset kinematic viscosity to filter out high-frequency mechanical vibration interference and retain low-frequency vibration characteristics related to road surface smoothness. The inertial detection device has a built-in temperature sensor. The data processing module adjusts the damping coefficient compensation factor when calculating the vertical displacement integral based on the real-time temperature value, and performs temperature compensation on the vertical inertial motion data.

6. The method for detecting the smoothness of highway pavement according to claim 1, characterized in that, In step 5, the preset horizontal baseline is generated by mean filtering of the road surface contour line of preset length. The vertical deviation value is the normal distance from the contour line sampling point to the baseline, and the deviation feature is the root mean square value of the vertical deviation of the sampling point within the preset evaluation interval. In step 6, the vehicle speed compensation function is fitted by multiple sets of constant speed experiments. When the instantaneous vehicle speed deviates from the standard detection speed, the deviation feature is corrected by weighting the deviation feature through this function. In step 7, the mapping model adopts a multi-layer network structure. The input layer receives the equivalent deviation feature, the road surface texture feature vector, and the ambient lighting parameters. After processing by the hidden layer, the output layer outputs the international roughness index.

7. A detection system for implementing the highway pavement smoothness detection method as described in any one of claims 1 to 6, characterized in that, include: The image acquisition module, deployed on the central axis of the vehicle's rear, includes a high dynamic range photosensitive sensor and a wide-angle lens, used to acquire original images of the current lane and adjacent road surfaces at a preset sampling frequency; the inertial detection device, installed on the vehicle body and independent of the suspension system, includes a sealed tube, damping fluid, movable working fluid, and a displacement sensor array, used to sense changes in the vehicle's vertical acceleration and displacement in real time, generating vertical inertial motion data; the vehicle data interface module, connected to the vehicle control local area network bus, is used to acquire the vehicle's instantaneous speed and displacement data of the original suspension system; the data processing module, communicating with the above modules, has a built-in digital signal processor and field-programmable gate array, used to execute the detection method, and the two exchange data through a preset bus, respectively responsible for image preprocessing, inertial data acquisition, and related algorithm operation.

8. The highway pavement smoothness detection system according to claim 7, characterized in that, The diagnostic unit of the data processing module has a sensor state prediction function. By recording the output fluctuation trajectory of the sensor within a preset time period and calculating the state transition probability, it can provide early warning of performance degradation trends. Multiple sets of flatness intermediate values ​​include three types of intermediate values ​​determined based on image, inertial detection device, and suspension system data. The data processing module executes a weighted fusion algorithm. When the relative deviation between any two sets of intermediate values ​​exceeds a preset threshold, the sensor is determined to be abnormal and the fault source is located. The data processing module runs on a real-time operating system and uses multi-threaded parallel processing to isolate each core thread and allocate different processor core affinity.

9. The highway pavement smoothness detection system according to claim 7, characterized in that, The system also includes a high-precision positioning module and a wireless transmission unit. The high-precision positioning module acquires the real-time coordinates of the vehicles, and the wireless transmission unit uploads multi-dimensional detection data to the cloud management platform. The cloud management platform uses overlapping detection data from multiple vehicles for secondary cross-verification and eliminates random errors through statistical averaging. The cloud platform also has an automatic defect classification function. For road sections where the international roughness index exceeds a preset threshold, it extracts road surface texture features and identifies defect types, generating a detection report containing information such as geographical location and lane number.

Images