Computer vision based action detection and classification / Computer Vision baseret detektion og klassifikation af handlinger

    Project Details


    Real time detection and recognition of the motions and actions of dynamic objects is of interest in various applications like surveillance and monitoring of behaviours of humans and for real time visualisation of humans in (collaborative) virtual environments. The approach is based on scale-space analysis of spatial-temporal representations of the camera input. Different behaviours and motion patterns have typical ranges in spatial and temporal scale spaces, respectively, and such a priori knowledge is taken into account to achieve robust performance. Experimental tests are made with real time computer vision systems using monochrome as well as colour cameras. (Peter Jürgensen, Erik Granum)
    Effective start/end date31/12/200331/12/2003