Study On Automation Reconstruction Of Building Model Using Terrestrial Laser Scanning Data

The digitization and modeling of urban architecture has always been an extremely important part of digital city construction.After nearly three decades of development,Light Detection and Ranging(Li DAR)technology has become an important remote sensing measurement method which collects array space Point cloud data of objects surface efficiently and massively.Depending on the laser scanning platform,Li DAR technology has developed a variety of complex laser data acquisition systems,including spaceborne laser scanners,airborne laser scanners,vehicle laser scanners,shipborne laser scanners,and terrestrial laser scanners.In recent years,drone laser scanners and hand-held laser scanners have also begun to be widely used in research and application.And Terrestrial Laser Scanning(TLS)technology can provide efficient and flexible data acquisition solutions for the digitization and modeling of urban buildings,and obtain high-resolution,high-precision spatial information of target buildings,so it can be used to provides powerful data support for the reconstruction of building model.Based on the research status of building model reconstruction of TLS data,combined with the research results and conclusions of other researchers,this paper studies the key issues in the whole process of building model reconstruction of TLS data,and establish a complete technical process of building model reconstruction,which including data preprocessing,Point cloud segmentation and model reconstruction.The main contents are as follows:(1)Data preprocessing.The purpose is to perform an octree management on raw high-density TLS data,and propose a raster denoising algorithm to filter out isolated point sets and isolated grid Point sets in the raw data.Based on the characteristics of TLS data,the proposed algorithm can effectively identify the isolated noise in the raw collected data including trees,pedestrians,vehicles and other obstacle,then filter out these noises to achieve subsequent accurate modeling.(2)Multi-site Point cloud data registration.The pre-registration based on the geometric features of TLS data and the accurate registration based on ICP algorithm are studied.According to the characteristics of TLS data,the relevant parameters are optimized,and the overall Point cloud data model of the target building with accurate registration is obtained.(3)Point cloud segmentation.The segmentation algorithm based on region growing is used to segment the whole Point cloud data of the building.According to the characteristics of TLS data,the Point cloud density is calculated.By optimizing the neighborhood radius,smoothing coefficient,curvature coefficient and other parameters,the Point cloud segmentation purpose is achieved.The segmentation algorithm based on parametric model is used to segment the point cloud data of high-density complex buildings.According to the characteristics of TLS data,the segmentation parameters such as bounding box distance threshold,bounding box wide bitmap resolution,and normal vector offset cosine threshold are adjusted to achieve the purpose of Point cloud segmentation.(4)Regularization algorithm of Point cloud slices.A Point cloud slice regularization algorithm based on least squares method and RANSAC is proposed.From the segmented Point cloud slice data,the boundary Point cloud is extracted,By applying the best plane of fitting,fit the straight line segment,and analyze the discrete disconnected line segments,the high-precision regular polygon is extracted.(5)Knowledge-based reconstruction methods.According to the semantic features of the Point cloud slices,the slices are divided into 7 categories including ground,wall,window,door,roof,step and others.The model reconstruction strategy.is formulated by statistical features of Point cloud slices such as area,position,normal vector and topological relationship.