Innovative Methods for 3D Point Cloud Processing of Large Data Sets and Its Practical Implementations

Singh, Rishika (2022) Innovative Methods for 3D Point Cloud Processing of Large Data Sets and Its Practical Implementations. PhD thesis, University of Gloucestershire. doi:10.46289/9K7C33MM

Full text not available from this repository.


Various point cloud processing applications demand fast and accurate results for extracting feature information from the data. Given that point clouds are implemented in various fields, this thesis focuses on the point clouds used by surveyors and civil engineers for geographic information systems. Examples of point clouds used within the industry include urban sites, city blocks, terrains for road development, construction sites, quarries, mines, etc. The technological advancements (evolution of laser scanners) that allow the data to be captured in millions created their own recurring problems. The algorithms developed in this thesis are targeted at large point clouds containing various features and shapes. However, while capturing the point clouds, several outliers and noise are captured with the regular data due to the reflection of surfaces like glass and mirrors or weather conditions. Therefore, the detection and deletion of these outliers and noise are required to address and simplify the feature detection process. Hence, point cloud filtration is the first step of point cloud processing. After filtration, the data is relatively smaller and free from outliers and noise. The next step is to detect and extract the features from the point clouds and perform segmentation and other points analysis techniques. This thesis proposes, designs, develops and implements the methods and algorithms for robust and efficient point cloud processing. The processing includes filtrations followed by detecting and extracting primitive shapes such as planes, edges and cylinders. Contributions: The first contribution of this thesis is the method of removing noise and outliers using the designed tools. The second contribution is a novel PCA-based algorithm for detecting edges and edge streams in point clouds. Finally, a voxel-based algorithm to detect trunks and pole-like objects. These proposed methods and algorithms directly benefit the processing of the point clouds with properties like filtration, extraction, segmentation, clusterisation and accuracy. The results of the proposed methods and algorithms are implemented on commercial software used by UK and worldwide users.

Item Type: Thesis (PhD)
Thesis Advisors:
Thesis AdvisorEmailURL
Uncontrolled Keywords: Point cloud processing; Algorithms; PCA-based algorithm; Voxel-based algorithm; Commercial software applications; Data processing; Urban features
Subjects: Q Science > QA Mathematics > QA76 Computer software
Divisions: Schools and Research Institutes > School of Business, Computing and Social Sciences
Depositing User: Susan Turner
Date Deposited: 16 Apr 2024 15:13
Last Modified: 16 Apr 2024 15:33

University Staff: Request a correction | Repository Editors: Update this record

University Of Gloucestershire

Bookmark and Share

Find Us On Social Media:

Social Media Icons Facebook Twitter Google+ YouTube Pinterest Linkedin

Other University Web Sites

University of Gloucestershire, The Park, Cheltenham, Gloucestershire, GL50 2RH. Telephone +44 (0)844 8010001.