Geometric studies in point cloud skeletonisation and their applications in robotics

Wen, Qingmeng ORCID: https://orcid.org/0000-0002-8972-4042 2025. Geometric studies in point cloud skeletonisation and their applications in robotics. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (29MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (748kB)

Abstract

Skeletons are commonly known as internal frameworks of humans and animals, but a similar concept is widely used in shape analysis. In this context, a skeleton represents a compact, centred abstraction of an object’s shape, capturing its topological structure. The process of extracting such a representation is known as skeletonisation. As a useful tool in shape analysis, skeletonisation has been extensively studied; however, key challenges remain–particularly in reducing computational overhead, establishing robust evaluation metrics, and applying skeletons to surface generalisation. This thesis addresses these gaps through three major contributions. First, to mitigate the high computational cost of existing skeletonisation methods, we analyse their inefficiencies and propose a novel geometry-based skeletonisation framework. By eliminating redundant computations inherent in prior techniques, our method significantly accelerates the skeletonisation process, outperforming a well-established baseline in terms of speed. Second, the development and comparison of skeletonisation techniques have been hindered by the absence of comprehensive quantitative evaluation tools. To address this, we systematically investigate the fundamental properties of skeletonisation and introduce a suite of evaluation metrics, along with practical numerical techniques. These metrics are validated for their effectiveness and demonstrated in robotics-related scenarios. Finally, the thesis explores a novel application of skeletons as priors for modelling rolling contact kinematics. By extracting skeletons from point clouds, we reconstruct the topological structure of arbitrary surfaces, enabling rolling contact analysis for complex, irregular, and even discontinuous object shapes.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Engineering
Uncontrolled Keywords:	1. Point Cloud 2. Skeletonisation 3. Rolling Contact 4. Kinematics 5. Computational Geometry 6. Robotics
Date of First Compliant Deposit:	16 December 2025
Last Modified:	17 Dec 2025 11:00
URI:	https://orca.cardiff.ac.uk/id/eprint/183292

Actions (repository staff only)

Edit Item