Horopter based Dynamic Background Segmentation applied to an Interactive Mobile Robot

Interactive mobile robots require object/subject detection in very visually complex environments. In the field of computer vision, specially when applied to robotics, several approaches like face detection, face recognition and pedestrian detection often have to deal with issues associated to bad illumination and strong featured background. These issues imply lack of performance because human detection algorithms will frequently process the whole image searching for features. Also, background segmentation approaches are commonly used to solve this problem on static camera surveillance. However all these approaches are unable to effectively deal with the constant background changes that certainly happen when the camera sensor is installed on a mobile robot. Hence, in this work we propose a Horopter based Dynamic Background Segmentation solution to this problem. Results show that our approach, significantly enhanced tracking, and consequently improved movement classification towards interaction.