Automatic Roof Reconstruction From Airbrone LiDAR Point Clouds

A building is a structure with a roof and walls standing more or less permanently in one place,in which people are living,working and studying.A building database is the basis for managing an urban with effective and efficiency.Three-dimensional(3D)building models are essential for 3D Geographic Information Systems(GIS)and play an important role in various urban management applications.As the traditional modeling fashion is by hand,which is boring and laborious,it is necessary to reconstruct 3D building models in a fully automatic manner.Since the data acquired by Light Detection and Ranging(LiDAR)systems are dense,accurate,and geo-referenced in 3D,they are often used to reconstruct building models.The development of laser scanning technologies has permitted the gathering of large volumes of LiDAR point clouds at high frequencies and densities,allowing for the derivation of more detailed 3D models.In addition,the demands for robust data management,efficient spatial data retrieval and high reconstruction performance are also desirable.Although several LiDAR data based reconstruction approaches have made advances toward the fully automatic generation of 3D building models,the process is still tedious and time-consuming,especially for massive point clouds,and economical and reliable techniques that fully exploit the advantages of spatial databases are lacking.The process for automatically reconstructing building models from point clouds remains challenging.The major difficulties include the absence of local data and noisy data or data that are inefficient for the reconstruction of sophisticated building structures.The accuracy of building boundaries and detailed roof features is also greatly influenced by outliers.A similar problem is that the data structures designed for 3D models barely considering the nature of buildings,and the ineffective representation form of building models may result in a bottleneck of visualization.This dissertation introduces a new framework that utilizes a spatial database to achieve high performance via parallel computation for fully automatic 3D building roof reconstruction from airborne LiDAR data.The framework integrates data-driven and model-driven methods to produce building roof models of the primary structure with detailed features,utilizing the spatial analysis functions provided by the system,reducing tedious and time-consuming computation.The framework is composed of five major components:(1)a density-based clustering algorithm to segment individual buildings,(2)an improved boundary tracing algorithm,(3)a hybrid method for segmenting planar patches that selects seed points in parameter space and grows the regions in spatial space,(4)a boundary regularization approach that considers outliers,and(5)a method for reconstructing the topological and geometrical information of building roofs,in which the primary structures of buildings are recovered using the intersections of planar patches and detailed roof features are fitted with primitives.The determined roof models are represented in a boundary-representation form in this investigation.As a model library and a texture library are introduced,similar geometrical and texture information of building models are shared,and thus,the visualization performance is improved.The contributions of this research are:(1)a solution framework implemented in a spatial database to process massive point clouds,which has the advantages of storing,managing and retrieving big data with efficiency,as well as reconstructing building models in a parallel mode with good qualify and performance.(2)An automatic workflow based on knowledge discovery techniques to reconstruct the roof details and the primary structures,which is robust to outliers and adapts to the non-uniformly distributed point clouds.Some other novelties are also presented in this dissertation,such as,the first application of segmenting individual buildings using a density-based spatial clustering algorithm and the new idea of using it in a parallel mode,an improved boundary tracing algorithm based on two differing neighborhoods,and the novel idea of extracting planar patches when growing them within a set of candidates with similar attributes.