Encadrement : Bertarnd ZAVIDOVIQUE ; Michèle GOUIFFES

Sujet : Segmentation couleur multi caractéristiques appliquée à la stéréovision

Début : 9/07 ; fin 9/2011

Financement Par bourse du gouvernement Thaî

Abstract : The research topic deals with “Multi modality pedestrian detection and avoidance”. The leading idea of the thesis is to make a computer based vision system exhibit interesting features from digital images and analyze them jointly, so as to design robust strategies for object detection (pedestrian, other cars or cyclist, buildings etc.). After the choice of adequate color representations of objects in images, the first designed module deals with Line extraction from color images towards stereo detection of objects. The second one addresses Region extraction in color and object-detection jointly controlled by lines.