Research On Street Information Detection From Mobile Laser Scanning Data In Urban Areas

With the developing ability to acquire high-quality point cloud data,mobile laser scanning systems have been widely utilized for various applications such as 3D city modeling,road inventory studies,safety control,car navigation,and forestry management.Data from mobile laser scanning(MLS)systems,airborne laser scanning(ALS)systems and terrestrial laser scanning(TLS)systems are being widely applied in urban scene analysis.However,MLS data are more suitable for urban scene information extraction for two reasons.First,compared to ALS data,MLS data are of higher density and contain more vertical information,factors that are of great importance in identifying detailed information from poles and buildings.Moreover,MLS data are more efficiently acquired than TLS data,as the latter are collected by manually positioned systems.Many researchers have studied the use of MLS data in urban scenes intensively,including in road and road marking detection,building detection and reconstruction,pole-like object detection,tree detection and modeling,and urban scene segmentation and classification.Extracting street object information from mobile laser scanning data is a complex process because of three major sources of difficulty.The first difficulty comes from the nature of cloud data,which is characterized by its unorganized structure,uneven density and large amount of data.Second,the purpose of segmenting the entire scene is to find common standards to separate mobile laser point clouds into individual objects,but different objects usually have different sizes and shapes.The last difficulty arises from the complexity of the urban environment,the proximity of cars to each other,the intertwining of trees and other objects,and the fact that various objects may fall under trees and rods.These complex situations increase the complexity of information extraction.Therefore,solving these problems can promote more efficient extraction of street information,and then to promote the automotive application of laser point cloud data in various industries.This major study points of the thesis are as follows:(1)We reviewed and summarized the street object extraction and segmentation cluses in mobile laser point cloud.This paper divides these clues into three categories:distance clues,geometric features based on local neighborhood calculation and spectral clues.Compared with the previous classification of related work based on the methods(such as model-based methods,region growing-based methods,clustering-based methods).The study of extraction clues will provide an important reference for follow-up researchers.(2)Study point cloud data preprocessing method in-depth,especially the mobile laser point cloud data.This study mainly includes the establishment of point cloud spatial index,the original point cloud data denoising and ground point cloud data filtering.These preprocessing methods are important bases for the subsequent processing of point cloud data.For example,the calculation of geometric features based on local neighborhood in subsequent road extraction requires the use of spatial indexing methods to calculate the principal direction.(3)A road surface extraction algorithm based on statistical features and image features is presented.The algorithm uses the mobile laser scanning data as the research object,with the coarse to fine strategy.First,two different strategies are proposed to remove the non-ground points,including the commonly used histogram-based threshold processing method and the histogram-based concave analysis method.Then,through the analysis of the boundary features of the road surface,an accurate extraction method of road surface based on road height gradient feature analysis is proposed.The method first generates the height gradient feature image,and then the image is binarized.After the morphological closing operation,holes in the image is eliminated,the whole road surface is extracted by using the region growing method by selecting points from the trajectory as the starting seeds.(4)A method of extracting pole-like objects based on adaptive cylinder model is proposed.First,the unorganized point clouds are voxelized to compress and organize the original data.Second,the bottom-up tracing algorithm is implemented to detect potential vertical-object locations,and the neighbour segments are merged.Finally,the adaptive radius cylinder model is built based on the selected layers of the vertical segment,and the model is applied to the merged vertical segments to perform isolation analysis.(5)We proposed a three-stage method for the segmentation of urban MLS data at the object level.The original unorganized point cloud is first voxelized,and all information needed is stored in the voxels.These voxels are then classified as ground and non-ground voxels.In the second stage,the whole scene is segmented into clusters by applying a density-based clustering method based on two key parameters:local density and minimum distance.In the third stage,a merging step and a re-assignment processing step are applied to address the over-segmentation problem and noise points,respectively.