point cloud; segmentation; laser scanning; 3D reconstruction; Feature Extraction; Cultural Heritage; Pattern recognition; Normals
Abstract :
[en] Segmentation is an essential step in the production chain of point-cloud processing. It defines the pertinence and accuracy of a future 3D model reconstruction, as well as a high level analysis of any scene. Its goal is to extract from a large 3D dataset different groups of points that share a logical link to wisely consider them as one entity. Manual segmentation of a 3D point cloud is extremely time consuming and imprecise for large datasets, and thus automated classification is the main focus of this paper.
Disciplines :
Earth sciences & physical geography Architecture
Author, co-author :
Poux, Florent ; Université de Liège > Département de géographie > Unité de Géomatique - Topographie et géométrologie
Billen, Roland ; Université de Liège > Département de géographie > Unité de Géomatique - Topographie et géométrologie
Language :
English
Title :
Semi-automatic and versatile point-cloud segmentation by region growing based on normal estimation
Alternative titles :
[fr] Segmentation de nuage de points semi-automatique et versatile par croissance de région basée sur l'estimation de normales