DocumentCode :
594750
Title :
Creating realistic 3D models from scanners by decoupling geometry and texture
Author :
Adan, Antonio ; Merchan, P. ; Salamanca, S.
Author_Institution :
Escuela Super. de Inf., Univ. de Castilla La Mancha, Ciudad Real, Spain
fYear :
2012
fDate :
11-15 Nov. 2012
Firstpage :
457
Lastpage :
460
Abstract :
To date the methods to create accuracy dense realistic 3D models of outdoors by using laser scanners are highly dependent on the on-site conditions in the very moment of the 3D data collection. Thus, researchers put in a lot of effort on eliminating colour incoherencies (sunny/shady, bright/dark, non sensed areas, etc.) or modelling the light of the scene to obtain free-illumination models. This paper proposes a new strategy that aims to separate the moments in which geometry and colour are taken, making the modelling process independent of the time and external circumstances in the data collection stage, and on-site calibration-free. Our approach is based on matching anytime suitable external images of the scene into the complete 3D geometric model and carrying out an iterative weighted colour mixing algorithm which gradually builds the textured model. The method has been mainly tested in archaeological sites; particularly in rest of ancient Roman buildings, yielding excellent results. Comparison with other commercial solutions is also shown in the paper.
Keywords :
archaeology; image colour analysis; image matching; image texture; iterative methods; natural scenes; optical scanners; realistic images; solid modelling; 3D data collection; 3D geometric model; Roman ancient buildings; archaeological sites; colour incoherency elimination; dense realistic 3D model creation method; external image matching; free-illumination models; iterative weighted colour mixing algorithm; laser scanners; on-site calibration-free modelling process; scene geometry; scene light modelling; Cameras; Computational modeling; Equations; Geometry; Image color analysis; Mathematical model; Solid modeling;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Pattern Recognition (ICPR), 2012 21st International Conference on
Conference_Location :
Tsukuba
ISSN :
1051-4651
Print_ISBN :
978-1-4673-2216-4
Type :
conf
Filename :
6460170
Link To Document :
بازگشت