Knowledge-based Scene Analysis in Indoor Environments Using Colour and Range Images

Bibtex entry :

@inproceedings { vmv06,
    author = { Benjamin Pitzer and Lars Libuda and Karl-Friedrich Kraiss },
    title = { Knowledge-based Scene Analysis in Indoor Environments Using Colour and Range Images },
    booktitle = { Vision, Modeling, and Visualization Conference (VMV) },
    year = { 2006 },
    editor = { L. Kobbelt and T. Kuhlen and T. Aach and R. Westermann },
    pages = { 33--40 },
    address = { Aachen, Germany },
    month = { November 22-24 },
    publisher = { Aka GmbH },
    abstract = { Object recognition from camera images is inherently an ambiguous problem. Even when stereo vision techniques are used, it is difficult to perform robust object recognition. Humans have a broad knowledge about their environment and are able to use this knowledge to reason in unknown environments. In this paper we present a knowledge based system to analyze single color and range images of indoor scenes inspired by human visual perception. The input images are recursively processed over four layers of abstraction resulting in a semantic scene description. A generic scene model of typical indoor environments is used as a priori knowledge. This model is encoded in semantic networks for explicit knowledge representation. The developed system is applied to images of artificial and real world indoor scenes, where it demonstrates good reconstruction rates. },
}