| With the introduction of "14th Five-Year Plan",the pace of construction of a strong transportation country has been accelerated.In the field of intelligent transportation,a new round of scientific and technological revolution and industrial change led by cloud-side vehicle-road coordination,intelligent network-connected vehicles and autonomous driving is developing deeply.The intelligent transportation technology industry led by wisdom,safety,green and innovation drive cannot be separated from the construction of road infrastructure.The active safety technologies represented by advanced assisted driving systems and autonomous driving systems all take road surface markings as an important baseline for traffic accident prevention and early warning,such as lane keeping,lane departure warning,intelligent cruise control,etc.Regular inspection of road markings to ensure their reflective intensity and diffuse illumination brightness are above the minimum threshold can provide safety assurance for accurate traffic perception of autonomous driving.Therefore,the detection and identification of road pavement markings has become the basis and hot spot of research in the field of intelligent transportation in recent years.At the same time,the quality assessment and performance decay of road pavement markings have been the focus of attention to ensure the clarity and smoothness of road pavement markings.The traditional manual visual marking quality inspection is slow,requires a lot of human and material resources,and there are certain security risks and interference factors in the field inspection of road surface,and for long-distance road sections,it is easy to miss and repeat inspection,which affects the efficiency and accuracy.In the new technology empowerment,automated,intelligent pavement marking visual inspection system is gradually replacing manual inspection,commonly used visual sensors for car HD camera,car line laser line scanner and car high line beam Li DAR.Although the vehicle HD camera can obtain highresolution,high-precision,micro-scale marking data,but it is vulnerable to the influence of light brightness,shadow,data credibility is poor.Although the vehicle-mounted line laser scanner and vehicle-mounted high line beam Li DAR have many advantages such as insensitivity to external light changes,strong adaptability to complex environments,and strong anti-interference capability,they are expensive and costly for commercial applications,and the detection data have redundant background information.In order to improve the digital construction of urban traffic infrastructure,strengthen the fine management of road traffic markings,and realize the full-cycle operational status detection of key roadway markings life,this paper proposes an efficient,safe,economic and convenient method to detect roadway markings,assess their quality and analyze the causes of performance decay.The specific studies conducted are as follows.(1)To address the shortcomings of the current vehicle vision sensors in pavement marking detection,a low-cost,interference-resistant low-wire-beam vehicle-mounted Li DAR sensor is proposed.In order to improve the detection range,detection density,and detection efficiency of the low beam vehicle-mounted Li DAR for marking lines,the built-in characteristics of the vehicle-mounted Li DAR and the influence of external influencing factors on its detection performance are analyzed,and a multi-objective optimization model of the vehicle-mounted Li DAR deployment scheme is innovatively established.To solve the noninferiority frontier of the solution set of the multi-objective optimization model,the multiobjective is converted into a single objective function using the idea of integrated optimization,and the optimal solution is obtained using the globally convergent EGA algorithm.The optimal installation height,rotation angle and dynamic tilt angle of the vehicle-mounted Li DAR in different scenarios are solved.Field experiments were conducted on campus and fast roads,respectively,and the relative errors and t-tests of the experimental results verified the correctness of the model theoretical study.This study provides general guidance for deploying less expensive low-wire-beam vehicle-mounted Li DAR sensors to collect more valuable pavement marking information,and the collected marking data information can be applied to the next marking disease detection.(2)To address the problem of difficult detection of low beam Li DAR markings,a framework for real-time detection of pavement markings based on point cloud features is proposed,which is divided into three main stages.The first stage is the point cloud data preprocessing,firstly,coordinate conversion of sensor data stream in spherical coordinate system to get 3D spatial point cloud in world coordinate system,and mapping it to the ground to get ground point cloud data,and finally,the Sigmoid function calibrated by natural breakpoint classification method is used to enhance the marker points and weaken the background points.The second stage is the reformatting of the marker point cloud.Firstly,we propose the automatic extraction algorithm of marker ROI to obtain the area where the marker is located,divide it into a formatted raster to solve the sparse problem of the marker point cloud,and propose a matrix missing value complementation algorithm based on the convolutional kernel smoothing window for the missing raster edge data to finally obtain the reformatted marker matrix.The third stage is the segmentation and classification of scalar disease instances.First,the OTSU algorithm is used to automatically segment the formatted scalar matrix to obtain the scalar binary map,then the seed region growth algorithm with eight connected domains is used to segment the disease instance individuals for the regions where the scalar disease exists,and four common disease types are summarized based on the geometric characteristics of different disease individuals.Through field experiments,the recognition accuracies of these four diseases were obtained as 96.71%,80.05%,73.16% and 59.70%,respectively,with an average accuracy of 78.45%.Compared with other researches,this algorithm can identify richer information of marking disease,and it is cheaper and more suitable for large-scale deployment,which can provide technical support for refined and intelligent marking quality repair in transportation infrastructure management.(3)In response to the current problems of difficult to quantify the usage performance automation of markings and little research on the causes of usage performance decay,the concept of pavement performance index in the Highway Technical Condition Assessment Standard(JTG H20-2007)is used to analyze the three stages of change in the usage performance of markings,and the innovative markings damage evaluation index MCI and markings aging evaluation index MAI are proposed as evaluation indexes.The innovative model of marking quality evaluation with MCI and MAI as evaluation indexes,combined with the point cloud data obtained by vehicle-mounted Li DAR sensors,solves the difficult problem of quantifying the marking performance.In order to analyze the causes of the decay of the marking performance,four typical decay models(concave curve,convex curve,inverse S curve and linear curve)are proposed,and their performance decay characteristics and causes are analyzed,taking into account the disease types and aging rates of the marking.Through field experiments,the following conclusions were reached:(1)the same construction process,the same construction time,the same road section,the same location,the same line type of yellow marker and white marker performance decay results are not very different,the average performance index is close.However,from the artificial visual point of view,white reflective markings have a higher retroreflective brightness coefficient,which can improve the visual effect of drivers and driving safety.(2)The performance index of solid line with the same construction process,the same construction time,the same road section,the same location and the same color paint is higher than the performance index of dashed line.(3)The same construction process,the same construction time,the same road section,the same color paint,the same line type,the performance index of the marking at different locations is slightly different,mainly at the intersection of the marking performance index is the highest,the second highest road section,the lowest at the entrance and exit of the intersection. |
PDF Full Text Request