Road Facility Extraction Based On Vehicle Mounted Point Cloud

With the development of social economy and technology,urban informatization is also being paid more and more attention to the road as the cornerstone and the main artery of the city,the construction of its information technology is also essential.Traditional road information is obtained through manual acquisition,but this way not only the collection speed is slow,the final data processing efficiency is low,and the collection process of personnel safety is not effectively guaranteed.With the rise and development of three-dimensional laser scanning technology,this situation has been effectively solved,the current road information acquisition is more often used vehicle-mounted three-dimensional laser scanning system to scan and collect the road scene,this acquisition method has a collection speed,high accuracy,scanning range,non-contact characteristics,greatly promote the construction and development of road information,so based on vehicle-mounted three-dimensional laser scanning to obtain road Therefore,the extraction and application of road information based on vehicle-mounted3 D laser scanning has become a hot spot and difficult point of current research.This research can not only promote the construction of road informatization,but also play a role in the development of high-precision maps,automatic unmanned vehicles and intelligent transportation and other high technologies.However,the point cloud data obtained based on vehicle-mounted laser scanning system is characterized by large amount of data,disorder and uneven density distribution,which leads to many defects and shortcomings in the extraction efficiency,extraction accuracy,information application and visualization of point cloud data.In view of these defects and shortcomings,based on the characteristics of the in-vehicle point cloud itself and the components of the road scene,this paper conducts a research on how to accurately and efficiently obtain the relevant road components from the road scene,and the specific research content is mainly the following aspects:1.A binary coding-based octree point cloud organization method is used to organize the point clouds in a binary coding-based quadtree.The framework uses binary coding to encode point clouds and quadtree nodes on the basis of organizing point clouds in quadtrees,so as to realize the voxelization of massive 3D point clouds.This not only provides a method for the effective organization of massive point clouds,but also provides a basis for the subsequent algorithms in this paper.2.Identification segmentation of road elements based on binary coded point cloud quadratic tree organization for road site point clouds.The research is carried out for the segmentation of steel guardrail and vehicle point cloud,the recognition segmentation of road traffic signage and the extraction of road boundary.(1)After Euclidean clustering,the mixed point cloud containing steel guardrail and vehicle scanned point cloud is segmented based on slicing method,and then the two are effectively segmented based on the difference of their geometric features;(2)Based on the binary coded point cloud quadtree organization,the road traffic signage is identified and extracted by means of cylindrical fitting and datum clustering based on the characteristics of road traffic signage consisting of two parts: pole and face.(3)Based on the binary encoded point cloud quadtree,the road boundary points are first extracted according to the difference of geometric features between the boundary points and the points in the road,and then the real road boundary is obtained by using curve growth,and finally the road boundary is fitted and interpolated by using Bessel curve fitting to obtain the road boundary.In addition to achieving the extraction and segmentation of various road elements,this paper also carries out the parameter calculation of the final results.3.The algorithm of this paper is verified and analyzed by using several road scenes with complex road conditions.In order to satisfy the verification of the algorithm and meet the actual production and life needs,a prototype system of "road information extraction and analysis based on massive vehicle point cloud" is developed and used as the basis for the verification of the algorithm proposed in this paper.The validation results show that the proposed algorithm can segment the steel guardrail and vehicle point cloud well,and the recognition rate of three different road scenes has reached 100%;it can effectively recognize and extract the road traffic signs,and recognize and extract the data of two different road scenes,and the quality evaluation index of the final recognition result is about 96%;it can extract the road boundary accurately and efficiently.The quality evaluation indexes of the final extraction results are all around 95%for three different road data.The experiment proves that the algorithm of this paper can meet the actual production and life needs,and has certain practical value.