Files in this item



application/pdfGU-THESIS-2015.pdf (6MB)
(no description provided)PDF


Title:The missing link in as-built 3D modeling: geometrical labeling of segmented point clouds for fitting geometrical surface
Author(s):Gu, Rongqi
Advisor(s):Golparvar-Fard, Mani
Department / Program:Civil & Environmental Engineering
Discipline:Civil Engineering
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Building Information Modeling (BIM)
surface fitting
Abstract:Modeling of the built environment is used in a variety of engineering analysis scenarios. Significant applications include monitoring of construction work in progress, quality control of on-site assemblies, building energy diagnostics, and structural integrity evaluation. The modeling process mainly consists of three sequential steps: data collection, modeling, and analysis. In current practice, these steps are performed manually which are time-consuming, prohibitively expensive, and prone to errors. While the analysis stage is fairly quick, taking several hours to complete, data collection and modeling can be the bottlenecks of the process: the first can spread over a few days, and the latter can span over multiple weeks or even months. In consequence, the applicability of as-built modeling has been traditionally restricted to high latency analysis, where the model need not be updated frequently. In fast changing environments such as construction sites, due to the difficulty in rapidly updating 3D models, model-based assessment methods for purposes such as progress or quality monitoring have had very limited applications. There is a need for a low-cost, reliable, and automated method for as-built modeling. This method should quickly generate and update accurate and complete semantically-rich models in a master format that is translatable to any engineering scenario and can be widely applied across all construction projects. To address these limitations, recent research efforts have focused on developing methods to (1) segment point cloud models at user’s desired level of abstraction; and (2) fit surface topologies such as NURBS into the segmented point clouds. While these methods exhibit flexibility in accounting for the user desired level of abstraction, yet they still result in over segmentation. Even if properly segmented, there is still a need to merge several segmented point clouds to create continuous surface models. The geometrical labels can also be used to better populate the scene with distinct surface objects based on the segmented subsets. To address current needs, this thesis focused on automatically labeling each segmented point cloud based on their geometrical properties as wall, floor, ceiling, and pipe, and fits in cylindrical and planar surfaces into the labeled point cloud models. To do so, the method detects and characterizes various types of geometrical features for each segment (e.g. density of the point cloud segment, curvature, height distributions, etc.) and infers their geometrical labels (wall, floor, ceiling, and pipe) using multiple one-vs.-all discriminative machine learning classifiers. Next, the most appropriate type of surface is fitted into the point cloud segments. The experiment results from applying the introduced method on real world point clouds – with an average accuracy of 89% in geometrical labeling – show promise in defining the relationship among segments, improve the accuracy of segmentation process, and can ultimately assist with populating the scene with distinct surface objects based on the segmented subsets.
Issue Date:2015-06-25
Rights Information:Copyright 2015 Rongqi Gu
Date Available in IDEALS:2015-09-29
Date Deposited:August 201

This item appears in the following Collection(s)

Item Statistics