Home > Research > Publications & Outputs > Machine Learning, Extraction and Classification...

Associated organisational unit

Electronic data

Text available via DOI:

View graph of relations

Machine Learning, Extraction and Classification of Memorial Objects from 3D Point Cloud Data

Research output: ThesisDoctoral Thesis

Published

Standard

Machine Learning, Extraction and Classification of Memorial Objects from 3D Point Cloud Data. / Arnold, Nicholas.
Lancaster University, 2023. 162 p.

Research output: ThesisDoctoral Thesis

Harvard

APA

Vancouver

Arnold N. Machine Learning, Extraction and Classification of Memorial Objects from 3D Point Cloud Data. Lancaster University, 2023. 162 p. doi: 10.17635/lancaster/thesis/2129

Author

Bibtex

@phdthesis{d3576411316549418d82a27c5e44e192,
title = "Machine Learning, Extraction and Classification of Memorial Objects from 3D Point Cloud Data",
abstract = "This thesis resides at the intersection of machine learning (ML) and cultural heritage, addressing the unique challenges posed by the analysis of 3D point cloud data from historic sites. The convergence of these sites with the modern age necessitates innovative approaches to their documentation, interpretation and management. Advances in remote sensing technologies, like LiDAR, offer a solution by enabling non-invasive, high-resolution digital representations of these sites in the form of point clouds. However, effectively utilising this digital medium for mapping and inventory remains challenging. This thesis bridges the gap by researching, designing and implementing an automated object extraction solution using ML and artificial intelligence (AI) techniques, emphasising explainability to enhance human learning from AI decision-making.The core pursuit of this thesis revolves around the automatic extraction and classification of objects from point cloud data within cultural heritage sites. A series of interconnected articles form the foundation of this exploration. The initial article introduces GeoPart-Transfer for automated extraction and labelling of memorial objects. The second article presents XPCC, a prototype-based classification and visualisation method for point clouds. It embraces interpretability and adaptability, permitting continuous learning without extensive retraining. The final article presents GeoPart-XPCC, a comprehensive framework applied to multiple scenes from various burial ground sites.Throughout these articles, a review of related research is undertaken, while experimentation adds empirical weight to the findings. The experimental results illustrate how these methodologies enhance the usability of digital point cloud data, offering broad applicability across different scenarios.By addressing the challenge of automatic object extraction from point cloud data, this thesis brings practical and impactful contributions to the domains of cultural heritage, archaeology and AI. Through comprehensive research, innovation, and experimentation, it illuminates the potential of digital technologies in preserving our historical heritage. It also lays a foundation for future explorations to other domains.",
author = "Nicholas Arnold",
year = "2023",
month = sep,
day = "19",
doi = "10.17635/lancaster/thesis/2129",
language = "English",
publisher = "Lancaster University",
school = "Computing and Communications",

}

RIS

TY - BOOK

T1 - Machine Learning, Extraction and Classification of Memorial Objects from 3D Point Cloud Data

AU - Arnold, Nicholas

PY - 2023/9/19

Y1 - 2023/9/19

N2 - This thesis resides at the intersection of machine learning (ML) and cultural heritage, addressing the unique challenges posed by the analysis of 3D point cloud data from historic sites. The convergence of these sites with the modern age necessitates innovative approaches to their documentation, interpretation and management. Advances in remote sensing technologies, like LiDAR, offer a solution by enabling non-invasive, high-resolution digital representations of these sites in the form of point clouds. However, effectively utilising this digital medium for mapping and inventory remains challenging. This thesis bridges the gap by researching, designing and implementing an automated object extraction solution using ML and artificial intelligence (AI) techniques, emphasising explainability to enhance human learning from AI decision-making.The core pursuit of this thesis revolves around the automatic extraction and classification of objects from point cloud data within cultural heritage sites. A series of interconnected articles form the foundation of this exploration. The initial article introduces GeoPart-Transfer for automated extraction and labelling of memorial objects. The second article presents XPCC, a prototype-based classification and visualisation method for point clouds. It embraces interpretability and adaptability, permitting continuous learning without extensive retraining. The final article presents GeoPart-XPCC, a comprehensive framework applied to multiple scenes from various burial ground sites.Throughout these articles, a review of related research is undertaken, while experimentation adds empirical weight to the findings. The experimental results illustrate how these methodologies enhance the usability of digital point cloud data, offering broad applicability across different scenarios.By addressing the challenge of automatic object extraction from point cloud data, this thesis brings practical and impactful contributions to the domains of cultural heritage, archaeology and AI. Through comprehensive research, innovation, and experimentation, it illuminates the potential of digital technologies in preserving our historical heritage. It also lays a foundation for future explorations to other domains.

AB - This thesis resides at the intersection of machine learning (ML) and cultural heritage, addressing the unique challenges posed by the analysis of 3D point cloud data from historic sites. The convergence of these sites with the modern age necessitates innovative approaches to their documentation, interpretation and management. Advances in remote sensing technologies, like LiDAR, offer a solution by enabling non-invasive, high-resolution digital representations of these sites in the form of point clouds. However, effectively utilising this digital medium for mapping and inventory remains challenging. This thesis bridges the gap by researching, designing and implementing an automated object extraction solution using ML and artificial intelligence (AI) techniques, emphasising explainability to enhance human learning from AI decision-making.The core pursuit of this thesis revolves around the automatic extraction and classification of objects from point cloud data within cultural heritage sites. A series of interconnected articles form the foundation of this exploration. The initial article introduces GeoPart-Transfer for automated extraction and labelling of memorial objects. The second article presents XPCC, a prototype-based classification and visualisation method for point clouds. It embraces interpretability and adaptability, permitting continuous learning without extensive retraining. The final article presents GeoPart-XPCC, a comprehensive framework applied to multiple scenes from various burial ground sites.Throughout these articles, a review of related research is undertaken, while experimentation adds empirical weight to the findings. The experimental results illustrate how these methodologies enhance the usability of digital point cloud data, offering broad applicability across different scenarios.By addressing the challenge of automatic object extraction from point cloud data, this thesis brings practical and impactful contributions to the domains of cultural heritage, archaeology and AI. Through comprehensive research, innovation, and experimentation, it illuminates the potential of digital technologies in preserving our historical heritage. It also lays a foundation for future explorations to other domains.

U2 - 10.17635/lancaster/thesis/2129

DO - 10.17635/lancaster/thesis/2129

M3 - Doctoral Thesis

PB - Lancaster University

ER -