A road surface flatness online monitoring system

Abstract

This invention discloses an online monitoring system for road surface smoothness, relating to the field of road engineering technology. The system includes: a time-series image acquisition module for acquiring time-series image data and associated spatiotemporal information of the road surface; a core dynamic modeling module configured with a defect spatiotemporal evolution anchor frame algorithm, which processes the time-series image data and associated spatiotemporal information to achieve integrated localization, dynamic tracking, and evolution prediction of road defects; and a predictive maintenance management module for receiving the output results of the core dynamic modeling module and, based on the output results, completing road smoothness level assessment, maintenance work order generation, and maintenance effect verification. This invention achieves accurate monitoring of road defects and efficient implementation of maintenance, solving the problems of insufficient accuracy in existing road monitoring and lack of basis for maintenance decisions.

Description

Technical Field

[0001] This invention relates to the field of road engineering technology, specifically to an online monitoring system for road surface smoothness. Background Technology

[0002] Road surface smoothness is a core indicator for the operation and maintenance of transportation infrastructure such as highways, urban roads, and airport runways, directly affecting driving safety, comfort, and the service life of facilities. With the continuous growth of traffic flow and the extension of infrastructure service life, the demand for real-time monitoring, dynamic defect tracking, and proactive maintenance of large-scale road networks is becoming increasingly urgent. Traditional monitoring modes that rely on manual inspections or single-function equipment are no longer suitable for the development pace of intelligent operation and maintenance of transportation infrastructure.

[0003] Existing monitoring technologies are mainly divided into two categories: contact and non-contact. Contact devices (such as bump integrators) have low inspection efficiency and limited daily coverage mileage, failing to meet the monitoring needs of large-scale road networks. Non-contact technologies often employ an architecture of independently stacked multiple algorithms, which not only consumes a lot of computing power and has high requirements for deployment hardware, but also has weak generalization ability, making it difficult to adapt to different road surface materials and complex environments. More importantly, existing technologies can only achieve post-defect detection or static location, failing to accurately capture the evolution pattern of defects and predict future development trends. This results in maintenance work being in a passive state of "post-disaster repair" for a long time, unable to plan maintenance resources in advance, avoid potential safety risks, and meet the core needs of proactive maintenance of transportation infrastructure. Summary of the Invention

[0004] The purpose of this invention is to provide an online monitoring system for road surface smoothness to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring system for road surface smoothness, comprising: The time-series image acquisition module is used to acquire time-series image data of the road surface and associated spatiotemporal information; The core dynamic modeling module is equipped with a defect spatiotemporal evolution anchor frame algorithm. The core dynamic modeling module processes the time-series image data and associated spatiotemporal information through the defect spatiotemporal evolution anchor frame algorithm to achieve integrated localization, dynamic tracking and evolution prediction of road defects. The predictive maintenance management module is used to receive the output results of the core dynamic modeling module and, based on the output results, complete the road smoothness level assessment, maintenance work order generation, and maintenance effect verification.

[0006] Preferably, the defect spatiotemporal evolution anchor frame algorithm includes an adaptive initial anchor frame generation step, which includes: S111. Perform defect contour segmentation on the initial frame image in the time-series image data; S112. Generate an adaptive initial anchor frame based on the defect contour obtained from segmentation; The size and shape of the adaptive initial anchor frame are adapted to the actual shape of the road defects, and the positioning error of the adaptive initial anchor frame is no more than 1 cm.

[0007] Preferably, the defect spatiotemporal evolution anchor frame algorithm further includes an anchor frame spatiotemporal evolution update step, which includes: S121. The displacement vectors of defective pixels within consecutive frames in the temporal image data are extracted using the optical flow method; S122. Combining the motion posture of the acquisition device in the associated spatiotemporal information, the size, shape and position of the adaptive initial anchor frame are updated in real time through spatiotemporal gradient calculation to form an evolving anchor frame; The tracking accuracy of the evolutionary anchor frame is no more than 2 centimeters.

[0008] Preferably, the defect spatiotemporal evolution anchor frame algorithm further includes a defect evolution trend prediction step, which includes: S131. Based on the historical evolution data of the aforementioned evolutionary anchor frame, the future evolution trend of the defect is calculated using a linear fitting formula, wherein the linear fitting formula is: ; In the formula, For the future Predicted defect size after time. For the first The size value of the evolving anchor frame at frame time. For defect evolution rate, To predict duration, The scene correction coefficient; the defect evolution rate The scene correction coefficients are obtained by fitting the historical evolution data. Based on pre-configured road surface materials.

[0009] Preferably, the time-series image acquisition module includes a high-definition camera and a GPS positioning unit, specifically configured as follows: The high-definition camera has an image resolution of no less than 1920×1080 pixels and a frame rate of 20 frames per second. The positioning error of the GPS positioning unit is no greater than 0.8 meters; The time synchronization accuracy between the time-series image data and the associated spatiotemporal information is no greater than 1 millisecond.

[0010] Preferably, the road smoothness level assessment performed by the predictive maintenance management module includes: S211. Extract the current defect size and predicted defect evolution trend output by the core dynamic modeling module; S212. Based on the above parameters, the road smoothness is divided into at least three levels, wherein the level is determined according to the defect size, defect evolution rate and safety risk threshold; The different levels correspond to different maintenance priorities, including at least three of the following: no maintenance required, routine maintenance, emergency maintenance, and closed management.

[0011] Preferably, the maintenance work order generation performed by the predictive maintenance management module includes: S221. Associate the precise location of the defect, the current state of the defect, and the evolution trend of the defect output by the core dynamic modeling module; S222. Based on the road smoothness grade, generate a standard work order that includes maintenance recommendations; S223. Push the standardized work order to the preset operation and maintenance management platform.

[0012] Preferably, the maintenance effectiveness review performed by the predictive maintenance management module includes: S231. After maintenance is completed, time-series image data of the same area are re-acquired through the time-series image acquisition module; S232. The core dynamic modeling module processes the re-acquired time-series image data to obtain new evolutionary anchor frames; S233. Compare the new evolution anchor frame with the evolution anchor frame before maintenance. If the new evolution anchor frame disappears or its size shrinks to below the preset qualified threshold, the maintenance is deemed qualified, and the road smoothness file is updated.

[0013] Preferably, the defect contour segmentation in S111 is achieved through road surface texture consistency analysis, which automatically distinguishes defect areas from normal areas, eliminating the need for manual annotation of defect samples; The defective area refers to the area formed by abnormal unevenness and structural damage on the road surface.

[0014] Preferably, the optical flow method in S121 is the Farneback optical flow algorithm; The displacement vector includes the horizontal displacement component and the vertical displacement component of the pixel in the image coordinate system. The spatiotemporal gradient calculation is based on the timestamp difference between the displacement vector and the associated spatiotemporal information, where the timestamp difference is the acquisition time interval between two consecutive frames of images.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The system acquires time-series image data of the road surface and associated spatiotemporal information through a time-series image acquisition module to achieve stable acquisition of road condition data. The core dynamic modeling module processes relevant data using a defect spatiotemporal evolution anchor frame algorithm to achieve integrated location tracking and evolution prediction of road defects. The predictive maintenance management module generates grade assessment work orders and verifies effects based on the processing results, achieving standardized management and control of road maintenance. This approach achieves precise monitoring of road defects and efficient implementation of maintenance, solving the problems of insufficient accuracy in existing road monitoring and lack of basis for maintenance decisions. Attached Figure Description

[0016] Figure 1 This is a structural block diagram of an online monitoring system for road surface smoothness provided in an embodiment of the present invention. Detailed Implementation

[0017] 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.

[0018] Please see Figure 1 This invention provides an online monitoring system for road surface smoothness, comprising: The time-series image acquisition module is used to acquire time-series image data of the road surface and associated spatiotemporal information; The core dynamic modeling module is equipped with a defect spatiotemporal evolution anchor frame algorithm. The core dynamic modeling module processes the time-series image data and associated spatiotemporal information through the defect spatiotemporal evolution anchor frame algorithm to achieve integrated localization, dynamic tracking and evolution prediction of road defects. The predictive maintenance management module is used to receive the output results of the core dynamic modeling module and, based on the output results, complete the road smoothness level assessment, maintenance work order generation, and maintenance effect verification.

[0019] In an optional embodiment, the defect spatiotemporal evolution anchor frame algorithm includes an adaptive initial anchor frame generation step, which includes: S111. Perform defect contour segmentation on the initial frame image in the time-series image data; S112. Generate an adaptive initial anchor frame based on the defect contour obtained from segmentation; The size and shape of the adaptive initial anchor frame are adapted to the actual shape of the road defects, and the positioning error of the adaptive initial anchor frame is no more than 1 cm.

[0020] It should be noted that in this embodiment, the adaptive initial anchor frame generation step is the basic step of the defect spatiotemporal evolution anchor frame algorithm. Defect contour segmentation requires the initial frame image to be preprocessed to grayscale, converting the color image into a single-channel grayscale image to simplify the calculation. Then, the edge features of potential road defects are extracted by the edge detection algorithm, and the defect and normal road surface areas are separated by the clustering algorithm to avoid defect omission or normal area misclassification.

[0021] The adaptive initial anchor frame overcomes the limitations of traditional fixed-size anchor frames, dynamically adjusting according to the actual size and shape of different defects. Whether it's a thin crack or an irregular pothole, it achieves a tight fit, avoiding tracking failures caused by anchor frame size mismatch. The positioning error is controlled within 1 cm, achieved through precise mapping between pixel coordinates and physical coordinates. Specifically, a correspondence between pixels and actual road surface distance is established using pre-calibrated camera intrinsic and extrinsic parameters. This is further corrected by matching and calibrating road texture feature points, ensuring the anchor frame positioning accuracy meets maintenance requirements. This step improves subsequent tracking stability, reduces tracking drift caused by initial positioning errors, and lays the foundation for accurate algorithm operation.

[0022] In an optional embodiment, the defect spatiotemporal evolution anchor frame algorithm further includes an anchor frame spatiotemporal evolution update step, which includes: S121. The displacement vectors of defective pixels within consecutive frames in the temporal image data are extracted using the optical flow method; S122. Combining the motion posture of the acquisition device in the associated spatiotemporal information, the size, shape and position of the adaptive initial anchor frame are updated in real time through spatiotemporal gradient calculation to form an evolving anchor frame; The tracking accuracy of the evolutionary anchor frame is no more than 2 centimeters.

[0023] It should be noted that in this embodiment, the anchor frame spatiotemporal evolution update step achieves dynamic synchronization between the anchor frame and the defect evolution, capturing both the morphological changes of the defect itself and offsetting the interference caused by the movement of the acquisition device. The optical flow method is chosen based on a balance between real-time performance and accuracy, enabling rapid extraction of the motion trajectory of defect pixels in consecutive frames to obtain displacement vectors in the horizontal and vertical directions.

[0024] Spatiotemporal gradient calculation constructs a spatiotemporal change model by fusing displacement vectors and device motion attitude data, thereby precisely adjusting the position, size, and shape of the anchor frame. The specific calculation logic is as follows: Firstly, based on the displacement vector ( (and timestamp difference) Calculate the instantaneous motion gradient at the pixel level ( ); Secondly, consider the translation amount of the data acquisition equipment. and rotation angle The instantaneous motion gradient is used for attitude correction, and the correction formula is the gradient. ,gradient ,in These are the pixel-to-physical coordinate transformation coefficients; Ultimately, the corrected gradient is used as the core basis for anchor frame updates, achieving precise adjustment of the anchor frame in the spatiotemporal dimensions. The motion attitude of the acquisition device is calculated through latitude and longitude changes in spatiotemporal information and preset installation parameters, which is used to correct the impact of device motion on anchor frame positioning, ensuring that the anchor frame tracking defect itself evolves, rather than the positional offset caused by device motion.

[0025] Tracking accuracy is controlled within 2 cm, achieved through inter-frame anchor frame overlap rate verification and adaptive adjustment of algorithm parameters. An inter-frame anchor frame overlap rate threshold of 0.85 is set. When the overlap rate falls below this threshold, the system automatically optimizes the optical flow method's calculation window or iteration count, for example, expanding the calculation window from 15×15 pixels to 21×21 pixels. This improves the accuracy of displacement vector extraction, ensuring stable tracking performance even in complex scenarios such as equipment vibration and rapid defect changes, providing continuous and reliable evolutionary data for subsequent trend prediction.

[0026] In an optional embodiment, the defect spatiotemporal evolution anchor frame algorithm further includes a defect evolution trend prediction step, which includes: S131. Based on the historical evolution data of the aforementioned evolutionary anchor frame, the future evolution trend of the defect is calculated using a linear fitting formula, wherein the linear fitting formula is: ; In the formula, For the future Predicted defect size after time. For the first The size value of the evolving anchor frame at frame time. For defect evolution rate, To predict duration, The scene correction coefficient; the defect evolution rate The scene correction coefficients are obtained by fitting the historical evolution data. Based on pre-configured road surface materials.

[0027] It should be noted that in this embodiment, the defect evolution trend prediction step mines the evolution patterns of defects through historical data and combines this with scenario adaptation to achieve accurate prediction. The application of the linear fitting formula is based on the gradual nature of defect evolution. Most pavement defects exhibit a slow and stable evolution trend in natural environments. The linear model can balance prediction accuracy and computational complexity, avoiding computational delays caused by model complexity.

[0028] Setting scene correction coefficients improves prediction accuracy. The differences in physical properties of different road surface materials directly affect the evolution rate of defects. For example, asphalt pavement is softer and more susceptible to temperature changes and vehicle loads, so its defect propagation rate is usually faster than that of harder cement pavement. In permafrost regions, roads are affected by freeze-thaw cycles, significantly increasing the defect evolution rate. By pre-configuring coefficients corresponding to different road surface materials, prediction biases caused by scene differences can be effectively offset, making the prediction results more consistent with reality.

[0029] The evolution rate was obtained by fitting historical evolution data from approximately 50 to 100 frames. During the fitting process, outliers exceeding three standard deviations, such as sudden size changes caused by momentary equipment jitter, were removed to prevent them from affecting the reliability of the fitting results. The core basis for the fitting was the linear correlation coefficient R of the time-series data. 2 A value ≥0.85 ensures that the selected historical data accurately reflects the evolution of defects. This data volume guarantees the reliability of the fitting results while avoiding computational delays caused by excessive data volume. The prediction duration supports multiple configurations, allowing flexible selection of 7 days, 30 days, or 90 days based on maintenance planning needs, corresponding to short-term emergency response, medium-term maintenance arrangements, and long-term maintenance plans, respectively, meeting maintenance requirements in different scenarios. This step upgrades the system from passive monitoring to proactive early warning, providing data support for the rational allocation of maintenance resources and reducing the traffic impact and costs of emergency repairs.

[0030] In an optional embodiment, the time-series image acquisition module includes a high-definition camera and a GPS positioning unit, specifically configured as follows: The high-definition camera has an image resolution of no less than 1920×1080 pixels and a frame rate of 20 frames per second. The positioning error of the GPS positioning unit is no greater than 0.8 meters; The time synchronization accuracy between the time-series image data and the associated spatiotemporal information is no greater than 1 millisecond.

[0031] It should be noted that in this embodiment, the parameter configuration of the time-series image acquisition module is determined based on the actual needs of road monitoring, balancing data quality and system cost. The resolution of the high-definition camera is set to no less than 1920×1080 pixels, which can clearly capture minute defects at the 0.5 mm level, meeting the defect detection accuracy requirements; the acquisition frame rate is set to 20 frames / second, which can continuously capture the evolution process of defects while avoiding data redundancy and storage pressure caused by excessively high frame rates. At the same time, it matches the processing capability of the core dynamic modeling module, ensuring that the rhythm of real-time acquisition and processing is synchronized, and avoiding defect omissions caused by frame rate mismatch.

[0032] The GPS positioning unit's positioning error is controlled within 0.8 meters, accurately marking the specific location of defects on the road surface. Combined with road mileage marker information, it forms a standardized location description including road name, mileage marker, and lateral offset, providing accurate location guidance for maintenance personnel and avoiding positioning deviations that reduce maintenance efficiency. Strict requirements for time synchronization accuracy ensure a one-to-one correspondence between time-series image data and spatiotemporal information, achieved through NTP or PTP time synchronization protocols. This avoids positioning, tracking, and prediction errors caused by time deviations, ensuring the consistency and reliability of the entire system's data chain. These parameter configurations use conventional and mature standards in the road monitoring field, eliminating the need for additional customized development, reducing system deployment and maintenance costs, while ensuring system compatibility and stability.

[0033] In an optional embodiment, the road smoothness rating assessment performed by the predictive maintenance management module includes: S211. Extract the current defect size and predicted defect evolution trend output by the core dynamic modeling module; S212. Based on the above parameters, the road smoothness is divided into at least three levels, wherein the level is determined according to the defect size, defect evolution rate and safety risk threshold; The different levels correspond to different maintenance priorities, including at least three of the following: no maintenance required, routine maintenance, emergency maintenance, and closed management.

[0034] It should be noted that in this embodiment, the road smoothness level assessment comprehensively considers the current state of defects and their future evolution trends to achieve a scientific classification of maintenance priorities. Relying solely on the current size of defects is insufficient to fully assess risk. Some small defects, if evolving rapidly, may develop into major safety hazards in the short term, while some large defects, if evolving slowly, can be scheduled for maintenance within the planned timeframe. Therefore, a comprehensive judgment must be made by combining the predicted evolution trends to ensure that maintenance resources are focused on high-risk defects.

[0035] The safety risk threshold is set based on a weighted calculation of multiple factors, including road grade, traffic volume, and defect type. The weight of each factor is determined using the analytic hierarchy process (AHP). Road grade has a weight of 0.4, with highways having a higher weight than national highways, and national highways having a higher weight than municipal access roads. Traffic volume has a weight of 0.3, with road sections having a daily average traffic volume greater than 5,000 vehicles having a higher weight than low-traffic sections. Defect type has a weight of 0.3, with potholes having a higher weight than cracks, and cracks having a higher weight than bulges. This weighting can be dynamically adjusted according to regional maintenance needs. The safety risk threshold can be flexibly adjusted for different application scenarios. For example, the safety risk threshold for highways is stricter than that for municipal access roads because highways have higher driving speeds, making the probability and consequences of accidents caused by defects more severe.

[0036] The prioritization of maintenance directly guides the allocation of maintenance resources. Levels requiring no maintenance can extend inspection cycles to 1-2 years, reducing unnecessary inspection costs. Routine maintenance is scheduled according to annual or quarterly maintenance plans. Emergency maintenance and road closures require priority deployment of personnel, equipment, and materials to prevent defects from worsening and affecting traffic safety. This assessment method enables refined and differentiated maintenance work, effectively improving the utilization efficiency of maintenance resources and reducing unnecessary maintenance costs.

[0037] In an optional embodiment, the maintenance work order generation performed by the predictive maintenance management module includes: S221. Associate the precise location of the defect, the current state of the defect, and the evolution trend of the defect output by the core dynamic modeling module; S222. Based on the road smoothness grade, generate a standard work order that includes maintenance recommendations; S223. Push the standardized work order to the preset operation and maintenance management platform.

[0038] It should be noted that in this embodiment, the maintenance work order generation achieves precise correlation between defect information and maintenance needs, providing maintenance personnel with clear and actionable construction guidance. The precise location information of defects adopts a standardized expression format, shortening on-site search time and improving maintenance efficiency. The current status of the defect includes key information such as defect type, core size data, and evolution rate. The evolution trend clearly indicates the development speed and future risk level of the defect. This information collectively serves as the basis for generating maintenance recommendations, ensuring the relevance and rationality of the recommendations.

[0039] The maintenance recommendations cover core aspects such as material selection, construction procedures, and project time estimates, and are generated based on a pre-set process library. The process library pre-sets corresponding maintenance solutions according to different defect types and levels. For example, for acceptable crack defects, grouting is recommended, using polyurethane grout material. The construction process includes grooving, cleaning, grouting, and curing, with an estimated time of 1-2 hours. For dangerous pothole defects, grooving repair and compaction are recommended, using asphalt mixture, with an estimated time of 4-6 hours. The process library can be updated and optimized based on maintenance standards and construction experience in different regions to ensure the practicality of the maintenance recommendations.

[0040] The standardized work order push mechanism supports seamless integration with existing operation and maintenance management platforms. Utilizing a RESTful API interface, the work order data format conforms to industry standards, eliminating the need to refactor existing management processes and reducing the cost of system deployment. The real-time nature of work order pushes ensures that operation and maintenance personnel receive defect information promptly, enabling rapid response to urgent maintenance needs, shortening the time interval from defect discovery to maintenance implementation, and preventing the escalation of road damage due to untimely maintenance.

[0041] In an optional embodiment, the maintenance effectiveness review performed by the predictive maintenance management module includes: S231. After maintenance is completed, time-series image data of the same area are re-acquired through the time-series image acquisition module; S232. The core dynamic modeling module processes the re-acquired time-series image data to obtain new evolutionary anchor frames; S233. Compare the new evolution anchor frame with the evolution anchor frame before maintenance. If the new evolution anchor frame disappears or its size shrinks to below the preset qualified threshold, the maintenance is deemed qualified, and the road smoothness file is updated.

[0042] It should be noted that in this embodiment, the maintenance effect verification establishes a closed-loop management system for maintenance work, ensuring that maintenance quality meets standards and avoiding issues such as false maintenance or incomplete maintenance. Re-collection is performed 7 to 15 days after maintenance, allowing time for the maintenance materials to stabilize and cure. For example, asphalt mixtures require a certain curing period after repair to reach stable strength, ensuring that the verification results accurately reflect the pavement condition after maintenance and avoiding misjudgments due to uncured materials.

[0043] The processing of newly evolved anchor frames employs the same algorithm and workflow as before maintenance, including defect contour segmentation, anchor frame generation, and evolution updates. This ensures the comparability of data before and after maintenance and avoids evaluation biases caused by different processing methods. The acceptable thresholds are set based on maintenance standards for different defect types, specifically determined according to the "Highway Maintenance Technical Specifications" (JTGH10-2009) and field experimental data: Crack-type defects after maintenance have a size ≤1 mm and an evolution rate ≤0.03 mm / day; pothole-type defects have a depth ≤0.5 mm and a maximum diameter ≤2 mm; and raised-type defects have a height ≤0.5 mm after maintenance, ensuring that defects after maintenance do not affect pavement structural stability and driving safety.

[0044] The updating of road smoothness archives enables traceability management of defects throughout their entire lifecycle. The archives record key information such as the original state of the defects, the maintenance process, and the review results, using structured data storage to support querying and statistical analysis. This archive provides data support for subsequent maintenance effectiveness evaluation, material selection optimization, and maintenance strategy adjustments. For example, by analyzing the review pass rates of different maintenance processes, the logic for recommending processes can be optimized; and by statistically analyzing the defect recurrence rate of road sections, maintenance cycles and plans can be adjusted. This closed-loop management mechanism improves the standardization of maintenance work, ensuring that maintenance investment effectively improves road smoothness and traffic safety.

[0045] In an optional embodiment, the defect contour segmentation in S111 is achieved through road surface texture consistency analysis, which automatically distinguishes defect areas from normal areas without the need for manual annotation of defect samples; The defective area refers to the area formed by abnormal unevenness and structural damage on the road surface.

[0046] It should be noted that in this embodiment, road surface texture consistency analysis achieves unsupervised defect contour segmentation. By utilizing the regularity and consistency of normal road surface texture, regions with abnormal textures are identified as defect candidate regions, thus avoiding reliance on manually labeled samples.

[0047] The specific implementation process of road surface texture consistency analysis technology is as follows: First, the initial frame image is divided into fixed-size local blocks of 10×10 pixels. Three types of core texture features are extracted from each local block, including the gray-level co-occurrence matrix that reflects the texture coarseness and uniformity, the local binary mode that reflects the local contrast, and the texture entropy that reflects the degree of texture disorder. The three types of features form a feature vector with unified dimension, which comprehensively describes the texture characteristics of the local area. Secondly, the K-means clustering algorithm is used to perform unsupervised classification on these feature vectors. The number of clusters is set to 2, corresponding to normal regions and defect candidate regions respectively. The iteration termination condition is that the cluster center offset is ≤0.01, to ensure the stability of the classification results and avoid segmentation errors caused by cluster center fluctuations. Finally, the candidate region is optimized using a region growing algorithm. The growth criterion is that the Euclidean distance between the feature vectors of adjacent pixels is ≤0.1 and the area of ​​the connected region is ≥5 pixels. Adjacent defect pixels are merged and isolated noise pixels are filtered out to form a complete and continuous defect contour.

[0048] The elimination of the need for manual sample annotation reduces data acquisition costs, shortens the system's adaptation cycle to new scenarios, and avoids subjective errors caused by manual annotation. This road surface texture consistency analysis is specifically optimized for the texture characteristics of road surfaces. By combining three types of texture features, it effectively distinguishes between interfering factors such as road markings and minor stains and genuine defect areas. It maintains high segmentation accuracy under different road surface materials (asphalt, cement, composite, etc.) and various environmental conditions (rain, backlight, night), providing reliable basic data for subsequent anchor frame generation and tracking.

[0049] In an optional embodiment, the optical flow method in S121 is the Farneback optical flow algorithm; The displacement vector includes the horizontal displacement component and the vertical displacement component of the pixel in the image coordinate system. The spatiotemporal gradient calculation is based on the timestamp difference between the displacement vector and the associated spatiotemporal information, where the timestamp difference is the acquisition time interval between two consecutive frames of images.

[0050] It should be noted that in this embodiment, the Farneback optical flow algorithm was chosen based on its stability and real-time performance advantages in dense optical flow calculations. This algorithm can provide a displacement vector for each pixel in the image, rather than just tracking feature points, meeting the requirements for fine-grained defect tracking, and is particularly suitable for defects rich in detail such as cracks and tiny pits. Furthermore, this algorithm is a mature technology related to computer vision, requiring no additional customized development, making its engineering implementation simple, and it has good compatibility with existing open-source libraries such as OpenCV, which facilitates rapid system deployment and maintenance, reducing development costs and time.

[0051] The horizontal and vertical components of the displacement vector together constitute the pixel's motion trajectory. The horizontal component reflects the distance the pixel moves in the image width direction, and the vertical component reflects the distance the pixel moves in the image height direction. Together, they comprehensively reflect the positional changes of defects on the image plane, providing a direct basis for adjusting the anchor frame position. The timestamp difference is a key parameter for calculating motion rate; it is equal to the acquisition time interval between two consecutive frames and is dynamically adjusted with the acquisition frame rate. For example, when the acquisition frame rate is 20 frames / second, the timestamp difference is 0.05 seconds. When the acquisition frame rate fluctuates due to environmental factors, the timestamp difference is adjusted synchronously to ensure accurate calculation of the spatiotemporal gradient even with frame rate fluctuations.

[0052] Spatiotemporal gradient calculation, by fusing displacement vectors and timestamp differences, obtains the motion rate and direction of defect pixels, thereby guiding the updating of anchor frame size, shape, and position. This calculation process fully considers the temporal dimension changes between image frames, ensuring that anchor frame updates are based not only on spatial position changes but also on temporal evolution patterns, improving tracking accuracy and stability. For example, for slowly evolving crack defects, spatiotemporal gradient calculation captures minute rate changes, allowing for advance adjustment of anchor frame size and preventing the anchor frame from detaching from the defect. In scenarios where the acquisition device moves at high speed, spatiotemporal gradient calculation rapidly responds to pixel displacement changes, ensuring the anchor frame follows the defect in real time, adapting to different tracking needs in various scenarios.

[0053] In this embodiment, a time-series image acquisition module collects time-series image data of the road surface and associated spatiotemporal information to achieve stable acquisition of road condition data; a core dynamic modeling module uses a defect spatiotemporal evolution anchor frame algorithm to process relevant data, achieving integrated location tracking and evolution prediction of road defects; and a predictive maintenance management module generates grade assessment work orders and verifies effects based on the processing results, achieving standardized management and control of road maintenance. This achieves the effect of accurate monitoring of road defects and efficient implementation of maintenance, solving the problems of insufficient accuracy of existing road monitoring and lack of basis for maintenance decisions.

[0054] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0055] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0056] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A road surface smoothness online monitoring system, characterized in that, include: The time-series image acquisition module is used to acquire time-series image data of the road surface and associated spatiotemporal information; The core dynamic modeling module is equipped with a defect spatiotemporal evolution anchor frame algorithm. The core dynamic modeling module processes the time-series image data and associated spatiotemporal information through the defect spatiotemporal evolution anchor frame algorithm to achieve integrated localization, dynamic tracking and evolution prediction of road defects. The predictive maintenance management module is used to receive the output results of the core dynamic modeling module and, based on the output results, complete the road smoothness level assessment, maintenance work order generation, and maintenance effect verification.

2. The online monitoring system for road surface smoothness according to claim 1, characterized in that, The defect spatiotemporal evolution anchor frame algorithm includes an adaptive initial anchor frame generation step, which includes: S111. Perform defect contour segmentation on the initial frame image in the time-series image data; S112. Generate an adaptive initial anchor frame based on the defect contour obtained from segmentation; The size and shape of the adaptive initial anchor frame are adapted to the actual shape of the road defects, and the positioning error of the adaptive initial anchor frame is no more than 1 cm.

3. The online monitoring system for road surface smoothness according to claim 2, characterized in that, The defect spatiotemporal evolution anchor frame algorithm further includes an anchor frame spatiotemporal evolution update step, which includes: S121. The displacement vectors of defective pixels within consecutive frames in the temporal image data are extracted using the optical flow method; S122. Combining the motion posture of the acquisition device in the associated spatiotemporal information, the size, shape and position of the adaptive initial anchor frame are updated in real time through spatiotemporal gradient calculation to form an evolving anchor frame; The tracking accuracy of the evolutionary anchor frame is no more than 2 centimeters.

4. The online monitoring system for road surface smoothness according to claim 3, characterized in that, The defect spatiotemporal evolution anchor frame algorithm further includes a defect evolution trend prediction step, which includes: S131. Based on the historical evolution data of the aforementioned evolutionary anchor frame, the future evolution trend of the defect is calculated using a linear fitting formula, wherein the linear fitting formula is: ； In the formula, For the future Predicted defect size after time. For the first The size value of the evolving anchor frame at frame time. For defect evolution rate, To predict duration, The scene correction coefficient; the defect evolution rate The scene correction coefficients are obtained by fitting the historical evolution data. Based on pre-configured road surface materials.

5. The online monitoring system for road surface smoothness according to claim 1, characterized in that, The time-series image acquisition module includes a high-definition camera and a GPS positioning unit, specifically configured as follows: The high-definition camera has an image resolution of no less than 1920×1080 pixels and a frame rate of 20 frames per second. The positioning error of the GPS positioning unit is no greater than 0.8 meters; The time synchronization accuracy between the time-series image data and the associated spatiotemporal information is no greater than 1 millisecond.

6. The online monitoring system for road surface smoothness according to claim 1, characterized in that, The road smoothness rating assessment performed by the predictive maintenance management module includes: S211. Extract the current defect size and predicted defect evolution trend output by the core dynamic modeling module; S212. Based on the above parameters, the road smoothness is divided into at least three levels, wherein the level is determined according to the defect size, defect evolution rate and safety risk threshold; The different levels correspond to different maintenance priorities, including at least three of the following: no maintenance required, routine maintenance, emergency maintenance, and closed management.

7. The online monitoring system for road surface smoothness according to claim 6, characterized in that, The maintenance work order generation performed by the predictive maintenance management module includes: S221. Associate the precise location of the defect, the current state of the defect, and the evolution trend of the defect output by the core dynamic modeling module; S222. Based on the road smoothness grade, generate a standard work order that includes maintenance recommendations; S223. Push the standardized work order to the preset operation and maintenance management platform.

8. The online monitoring system for road surface smoothness according to claim 7, characterized in that, The maintenance effectiveness review performed by the predictive maintenance management module includes: S231. After maintenance is completed, time-series image data of the same area are re-acquired through the time-series image acquisition module; S232. The core dynamic modeling module processes the re-acquired time-series image data to obtain new evolutionary anchor frames; S233. Compare the new evolution anchor frame with the evolution anchor frame before maintenance. If the new evolution anchor frame disappears or its size shrinks to below the preset qualified threshold, the maintenance is deemed qualified, and the road smoothness file is updated.

9. The online monitoring system for road surface smoothness according to claim 2, characterized in that, The defect contour segmentation in S111 is achieved through road surface texture consistency analysis, which automatically distinguishes defect areas from normal areas without the need for manual labeling of defect samples. The defective area refers to the area formed by abnormal unevenness and structural damage on the road surface.

10. The online monitoring system for road surface smoothness according to claim 3, characterized in that, The optical flow method in S121 is the Farneback optical flow algorithm; The displacement vector includes the horizontal displacement component and the vertical displacement component of the pixel in the image coordinate system. The spatiotemporal gradient calculation is based on the timestamp difference between the displacement vector and the associated spatiotemporal information, where the timestamp difference is the acquisition time interval between two consecutive frames of images.

Images