Research On Feature Extraction Of Road Facility Based On LiDAR Data

With the world-wide promotion of Industry 4.0,the life cycle management of infrastructure within AEC/FM domain is stepping into the era of digitalization and automation.In the aspect of asset management and maintenance of road facilities,the CAD/BIM model-based digital project management method has obtained extensive researches and application due to its many advantages such as systematical synergy,robust simulation,visualization and etc.The availability and accuracy of the information contained in a road CAD/BIM,which are dependent on the data collecting methods,has a crucial impact on the success of such project management method.The traditional measurement method requires a large amount of manpower and material resources and suffers from inaccuracy at the same time.In recent years,the way that generates CAD/BIM from road point cloud data using 3D laser scanning technology has been emerging and accepted widely.However,the segmentation and feature extraction of point cloud data are key technical gaps in this process.Hence,in this study,the idea of reverse engineering is applied in generating the as-built road facility CAD/BIM and the road 3D LiDAR data acquiring,processing and feature extraction are discussed as well.On the basis of the principles of laser ranging scanning,this paper introduces different ways of laser ranging,and various laser scanning systems derived from scanning operations of different carriers,providing theoretical support for the collection of point cloud data of roads and ancillary facilities.Structure from Motion(SfM)technology provides a solution to data supplement of 3D laser point cloud,as the missing data tends to be caused by physical factors such as occlusion,reflection,and the difficulty of setting the station in the environment.This paper proposes a method which improves the efficiency of feature matching for SfM.In this study,the primary-secondary block SfM algorithms is proposed,using global low-resolution images and local high-resolution images as input data.The proposed algorithm works better than the conventional one,which is verified by a series of preliminary simulation experiments and four groups of road environment experiments.According to the background of this research,the registration between laser scanning data and laser scanning data,and the registration between SfM data and laser scanning data are conducted following the registration principles.Based on the principle and flow of multi-view laser scanning point cloud processing,comparing study on various methods of point cloud segmentation and feature extraction is conducted.Then,combining with region-growing algorithm,a point cloud segmentation method according to point-based panorama rendering is developed in this study.And the polygonal features of road facilities are extracted as followed.However,the components of road facilities are various and structurally complex.The road surface geometry is instead quite simple and the research on extracting road surface features from road environment point cloud is relatively mature.Therefore,road auxiliary facilities are selected in this study as experimental objects in road environment experiments to verify the feasibility of the proposed method.