TY  - JOUR
T1  - Dynamic Extended Rectangle Based Method for 3D Visual Scene Segmentation
A1  - Charara, Nour
A1  - Maria, Sokhn
A1  - Jarkass, Iman
A1  - Abou Khaled, Omar
A1  - Mugellini, Elena
JA  - International Review on Computers and Software (IRECOS)
Y1  - 2013
VL  - 8
IS  - 4
PB  - Praise Worthy Prize S.r.l.
SN  - 1828-6003
KW  - depth information
KW  - neuro-fuzzy classification
KW  - Pattern Recognition
KW  - Scene segmentation
KW  - texture
N2  - a novel approach for semantic scene segmentation with computer vision is represented in this paper. The principle goal is to capture the scene distribution of the different zones of any multipurpose hall, taking into account both monocular visual cues (texture feature) and depth information. A feature extraction strategy based on a dynamic extension of the rectangular patches is proposed in order to provide a more accurate segmentaiton and to avoid redundancy penalties. The depth constraints control the initial classification decision that is obtained by a neuro-fuzzy classifier. An experimental study is applied on our dataset to demonstrate this approach on his different optimization stages.
ER  -