Answers are grounded in this paper and cite their sources.
Point clouds are essential for 3D spatial analysis and widely used in geodesy, photogrammetry, and remote sensing. While modern technologies simplify their collection, processing remains challenging due to data size, irregularity, and noise. Classification is critical for object identification and noise removal. This paper explores geometric features of points derived from their local 3D neighbourhoods. It examines neighbourhood definitions, feature computation via principal component analysis (PCA), and their impact on real dataset classification. Using a test point cloud with natural and anthropogenic features, we analyze feature dependencies, identify redundancies, and highlight key metrics. Additionally, we propose new approaches for noise filtering, contributing to more efficient point cloud processing and practical applications. Received: 02.12.2024 Received in revised form: 20.01.2025 Accepted: 14.02.2025
No takes yet. Share an insight, caveat, or question.
Boušek et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: