Road And Marking Extraction Based On Vehicle Laser Point Cloud Data

Vehicular laser scanning technology as a new ground surveying and mapping information collection technology,able to obtain high density,high precision automation on either side of the road and road building,vegetation,such as feature point cloud data,greatly improve the efficiency of the three dimensional space information acquisition,road engineering in recent years and high precision measurement,automatic driving map production and other fields has been developing rapidly.The abundant ground object information in the vehicle-mounted laser point cloud brings difficulties and challenges to the data processing and analysis of vehicle-mounted laser point cloud.How to extract the required road and line marking information from the vehicle-mounted laser point cloud efficiently,accurately and completely is the core and key of application in various fields.This paper takes vehicle-mounted laser point cloud road and automatic extraction of road marking as the research target,and the main research results are as follows:(1)this article in view of the large density of vehicular laser point cloud data,the point cloud density is not homogeneous and the characteristics of point cloud empty due to the terrain shade of on-board preprocessing of point cloud by adopting the combination of statistical analysis and DBSCAN clustering point cloud denoising algorithm to remove outliers point cloud data,use in combination with the normal estimation under the voxel grid sampling algorithm to streamline car laser point cloud.The experimental results show that the vehicle point cloud preprocessing can effectively remove the outlier points in the original point cloud data,keep the details of the point cloud while simplifying the point cloud data,and improve the efficiency and accuracy of the subsequent vehicle point cloud road elements extraction.(2)In view of the limitation of elevation difference threshold extraction set by elevation difference of point cloud at the edge of road in current vehicle-mounted laser point cloud road extraction,this paper designs a vehicle-mounted point cloud road extraction algorithm that is gradually extracted to fine optimization.Used in road crude extract the constraint conditions that the improved RANSAC algorithm partitions adaptive threshold extraction used in extract essence road with RANSAC threshold combination of morphological filtering and based on the point cloud density and intensity of filtering to extract point cloud to carry on the secondary filter fine road,finally optimize the road boundary point cloud to extract the complete path.Experimental results show that the accuracy,integrity and extraction quality of the proposed algorithm are over 90%,and it is suitable for vehicle-mounted point cloud road extraction of different types and different road environments.(3)As it is difficult and limited to extract road markings directly from the vehicle-mounted point cloud through intensity information,the road markings extraction method based on point cloud feature image is improved,and the vehicle-mounted point cloud road markings classification extraction algorithm is designed.After the vehicular point cloud is projected into feature images,the road marking pixels are extracted by image processing algorithm,and the complete road marking point cloud is obtained by further filtering and optimization after projection inverse transformation.Finally,the road marking is classified by template matching.Experimental results show that the accuracy,integrity and comprehensive evaluation of the proposed algorithm are over 90%,and the algorithm has a high degree of automation and strong anti-noise ability.