Real-time visual perception : detection and localisation of static and moving objects from a moving stereo rig

Benjamin Lefaudeux (1) , Fawzi Nashashibi (1, 2)

1 IMARA - Informatique, Mathématiques et Automatique pour la Route Automatisée
2 CAOR - Centre de Robotique

We present a novel method for scene reconstruction and moving object detection and tracking, using extensive point tracking (typically more than 4000 points per frame) over time. Current neighbourhood is reconstructed in the form of a 3D point cloud, which allows for extra features (ground detection, path planning, obstacle detection). Reconstruction framework takes moving objects into account, and tracking over time allows for trajectory and speed estimation.

Mots clés: Moving object detection; Stereo-vision; Scene reconstruction; Moving objects tracking

Domaines: Automatique / Robotique

Benjamin Lefaudeux, Fawzi Nashashibi. Real-time visual perception : detection and localisation of static and moving objects from a moving stereo rig. ITSC 2012 - 15th International IEEE Conference on Intelligent Transportation Systems, Sep 2012, Anchorage, United States. pp.522 - 527, ⟨10.1109/ITSC.2012.6338872⟩. ⟨hal-00759944⟩