Naslov
Detection of Dominant Planar Surfaces in Indoor Environments Using Stereo Vision and Image Segmentation

Autori
Nyarko, Emmanuel Karlo ; Cupec, Robert ; Dželalija, Aleksandar

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
28th International Conference SCIENCE IN PRACTICE Conference Proceedings (SIP 2010) / Mester, Gyula - Subotica : Subotica Tech - College of Applied Sciences, 2010, 107-116

ISBN
978-86-85409-53-0

Skup
28th International Conference SCIENCE IN PRACTICE (SIP 2010)

Mjesto i datum
Subotica, Srbija, 03.06.2010. - 04.06.2010

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Planar surfaces; Stereo; Disparity Image; Region-Based Segmentation

Sažetak
In this paper, detection of dominant planar surfaces in indoor scenes using stereo vision is considered. In indoor environments, many of the dominant surfaces are uniformly colored, which results in 'holes' in disparity images obtained by standard local area-based methods. Application of region based stereo reconstruction methods to cope with this problem is investigated. Region based methods rely on an image segmentation preprocessing step which segments an image into regions inside which there are no abrupt changes in disparity. Assuming that a poorly textured region obtained by image segmentation represents a single planar surface, the disparity can be assigned to all points of that region by fitting a plane to its portions which are reliably reconstructed by a local area-based stereo vision method, hence filling the gaps in the disparity image. An experimental study of this approach is reported. The experiments are performed using a stereo camera system mounted on a mobile robot.

Izvorni jezik
Engleski

Znanstvena područja
Elektrotehnika, Računarstvo

POVEZANOST RADA