Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/88390
Title: Automatic reconstruction and modelling of 3D geometrical surfaces from unstructured point cloud
Authors: Ahmed, Wael Mohamed Sayed
Degree: Ph.D.
Issue Date: 2020
Abstract: Recently, point clouds have been conducted to reconstruct 3D models for several applications. Three different types of modelling are conducted. Firstly, a semantic indoor geometric modelling approach (SIGMA) is designed for reconstructing parametric surface-based building models with additional models of wall-surface objects. Our approach uses a five-step process including pre-processing, 3D segmentation, layout reconstruction, wall-surface object modelling, and ceiling reconstruction. Compared to existing approaches, our approach can model complex layout structures of arbitrary ceilings with enriched wall-surface models from point cloud datasets. Quantitative evaluations demonstrate the capabilities of SIGMA on a complex real-world point cloud dataset. Secondly, detection reconstruction of outdoor structure from an image-based point cloud is proposed. The benefits of the spatial and coloured point cloud are used to isolate the structure, and primitive surfaces are detected to reconstruct the model from roof patches. The reconstructed model shows that the workflow is sufficient to describe the whole building structure in the required LOD. Finally, a proposed morphologically iterative TIN (MIT) ground filter which only requires maximum building size in processing LiDAR data. This approach applies morphological and TIN densification in an iterative way for separating ground points from off-ground ones. Experimental results using ISPRS benchmark datasets and Hong Kong LiDAR datasets reveal that MIT is effective in detecting more ground points and robust in various terrain situations.
Subjects: Remote sensing
Image reconstruction
Image processing -- Digital techniques
Hong Kong Polytechnic University -- Dissertations
Pages: xix, 144 pages : color illustrations
Appears in Collections:Thesis

Show full item record

Page views

49
Last Week
0
Last month
Citations as of May 5, 2024

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.