Catalogo Articoli (Spogli Riviste)

OPAC HELP

Titolo:
3D geometry reconstruction from multiple segmented surface descriptions using neuro-fuzzy similarity measures
Autore:
Fischer, D; Kohlhepp, P;
Indirizzi:
Forschungszentrum Karlsruhe, Inst Angew Informat, D-76021 Karlsruhe, Germany Forschungszentrum Karlsruhe Karlsruhe Germany D-76021 Karlsruhe, Germany
Titolo Testata:
JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS
fascicolo: 4, volume: 29, anno: 2000,
pagine: 389 - 431
SICI:
0921-0296(200012)29:4<389:3GRFMS>2.0.ZU;2-1
Fonte:
ISI
Lingua:
ENG
Soggetto:
ROBOT NAVIGATION; COMMON SUBGRAPH; RANGE IMAGES; 3-D OBJECTS; VIEWS; RECOGNITION; REPRESENTATION; REGISTRATION; MODELS; INTEGRATION;
Keywords:
late fusion; surface similarity measure; Neuro-Fuzzy; attributed graph; boundary representation; feature correspondence; 3D reconstruction; image registration; range image;
Tipo documento:
Article
Natura:
Periodico
Settore Disciplinare:
Engineering, Computing & Technology
Citazioni:
57
Recensione:
Indirizzi per estratti:
Indirizzo: Kohlhepp, P Forschungszentrum Karlsruhe, Inst Angew Informat, Postfach 3640, D-76021 Karlsruhe, Germany Forschungszentrum Karlsruhe Postfach 3640 Karlsruhe Germany D-76021
Citazione:
D. Fischer e P. Kohlhepp, "3D geometry reconstruction from multiple segmented surface descriptions using neuro-fuzzy similarity measures", J INTEL ROB, 29(4), 2000, pp. 389-431

Abstract

This paper presents a novel solution to the reconstruction of 3D geometry models from partial, segmented (2.5D or 3D) range views. First, the geometric fusion works entirely on sparse symbolic information, i.e. attributed surface graphs, rather than point data or triangulated meshes. Thus, new sensor data can always be integrated with an existing partial model available for symbolic action planning. Second, assumptions on automatic registration are weaker than those found in related work: the views need not be approximately calibrated, and no pre-existing knowledge of their overlap is needed. In order to find corresponding (redundant) surface features reliably even under high-noise and occlusion conditions we develop Neuro-Fuzzy similaritymeasures on surface descriptions. Third, we propose a reasonably complete prototype system including algorithms for merging sparse, reduced surface attributes, in particular boundaries. The experimental results from segmented range images of an indoor camera motion sequence demonstrate the ability to cope with unknown camera positions, low image resolution, large measurement and segmentation errors.

ASDD Area Sistemi Dipartimentali e Documentali, Università di Bologna, Catalogo delle riviste ed altri periodici
Documento generato il 12/07/20 alle ore 09:39:05