| The rapid development of highway construction in China has greatly stimulated the demand for highway maintenance and the reconstruction and expansion of old roads.Pavement is the core content of highway maintenance and reconstruction and expansion.Fast and effective access to pavement information,such as geometric parameters,condition status and their changing trend,is the premise and basis for decision-making for pavement maintenance,the determination of new maintenance plan,and the formulation of reconstruction and expansion schemes.However,pavement information has long been collected by some specific sensors and field measurements,causing the test results are easily affected by a variety of factors and the lack of a unified data benchmark.Integrating laser scanners,a Global Navigation Satellite System receiver,an attitude measurement system,cameras,and other sensors,Mobile Laser Scanning(MLS)systems can fast acquire three dimensions point cloud data of the surrounding environment along the driving path with high-precision and high-density,providing a new technical method for automatic detection of road surface geometry and condition status.This study delves into the key technologies of applying MLS data to automatic detection of highway pavement geometry and technical status.This study delves into the key technologies of applying MLS data to automatic detection of highway pavement geometry and technical status,and attempts to construct a framework of "MLS data organization—road feature extraction—geometric condition detection—pavement condition detection".The main research contents are as follows:1.A sequential index structure(named Tgrid)is proposed according to the collection sequence of MLS point cloud to solve the problem of discrete and no topology of MLS point cloud data.Compared with the traditional methods,the proposed method not only realizes the fast query of massive point cloud data,but also solves the inconformity between sequential storage and index storage of MLS point cloud,and successfully introduces the image methods into the processing of MLS point cloud data.2.This paper presents an approach to reconstruct the scanner ground track according to the spatial distribution of point clouds for some cases that the trajectory data is not available,such as the trajectory data is not contained in the ‘.las’ files or track files are corrupted,no scanning angle information and so on.The experimental results show that the reconstructed trajectories are very close to the real ones with only 1-2 laser points deviation.This study creates a theoretical basis for scanning trajectory reconstruction using MLS point cloud.3.This work presents and studied a systematic processing methods of extracting pavement features.The main contributions of this thesis are as follows:(1)A region growing method controlled by point labels is proposed for pavement point detection.The connected region analysis of Tgrid structure and the Freeman chain code boundary tracking method are combined for fast extraction of the contours of pavement points and road boundaries.(2)An adaptive threshold segmentation method based on intensity difference from background is designed to screen the possible road marking points.Then,a modified mathematical morphology method is implemented in Tgrid structure to identify Lanes.The crucial locations of the survey of the road geometry and pavement condition,such as the road centerlines and wheel traces,are finally extracted.(3)Using the Tgrid node graph,the holes in pavement points are effectively and easily captured by overlaying the extracted road boundary and pavement points.The experimental results verify the effectiveness and accuracy of the proposed method.The integrity rate of the pavement points detection is as high as 99.67%.Compared with the manually demarcated results of road boundary,the precision and recall are 96.78% and 92.91% respectively.The correct detection rate of lane line is 98.80%.4.This research studies the methods of detection of road geometric conditions based on the extracted centerline and dense MLS point cloud.Methods for detection the main road geometric parameters such as curvature,longitudinal and transverse slope from threedimensional high precision point clouds are designed.The geometric alignment of the highway is detected through the changes of curvature and longitudinal slope.The safety of the existing highway alignment is evaluated based on three technical indexes: curvature continuity,curvature balance and slope length.Test on a multi-curved winding road shows that the derived dangerous sections are consistent with the actual situation.From the data of sampling sites,the error of the profile elevation is 0.031 m,and of cross-slope is 0.33%.5.In this paper,a series of methods are proposed for detection of road damage,road roughness and pavement rutting,and a technical framework is constructed for automatic detection of road geometry and pavement condition using MLS point cloud.The main contributions of this thesis are as follows:(1)A strategy is designed for pavement damage detection based on three-dimensional point cloud and high-resolution Charge Coupled Device(CCD)images,and an adaptive threshold segmentation method under setting proportional restriction of damages is proposed for pavement damage detection.Automatic detection of pavement cracks and potholes using MLS point cloud are achieved.(2)With reference to the recommended conventional method of pavement inspection code,the accuracy and sampling rate requirements for input data,a method for detecting pavement roughness based on the profile elevation of the wheel trace points is proposed.(3)Near wheel trace points,the method of generating fine cross-section is studied to serve the detection of rut depth.Both analyses and test results show that the standard deviation of road surface roughness is highly correlated with the result of international roughness index(IRI)calculated based on the cross-section elevation of dense point cloud.Thus,to simplify the computational complexity of IRI,it is desirable to carry out correlation experiments on some test sections to obtain the conversion relationship between the two indexes,and then,to convert the value of σ to IRI value.Compared with the sampled results conducted by the method of precision leveling,the error of rutting depth computed by using MLS point cloud is less than 0.010 m. |
PDF Full Text Request