Multiple-View Video for Traffic Surveillance (OIS)


The goal of the OIS project (Optical Information Systems for traffic analysis and control) is to develop a complete system for automatic controlling of the traffic lights at crossings. The difference to existing systems is that solely optical information captured by video cameras will be used as input. The concept of the system is illustrated in the following figure. A crossing is captured by several cameras at different positions (e.g. 4). The setting will include infrared cameras, to be able to operate the system also at night. Low-level features of the traffic are already extracted in processing units at the cameras. This includes object type (e.g. car, van, bicycle, pedestrian), position, speed, etc. These data are encoded and transmitted to a central processor at the crossing. All cameras and associated processing units are connected using WLAN or FireWire. In addition to the extracted traffic data, the video signals themselves can be encoded and transmitted, with varying frame rates, picture sizes, and bitrates, in order to cope with varying channel conditions. Further processing is done at the central processor. This includes generation of intermediate views (between existing cameras), coding, further feature extraction and the control of the traffic lights. Necessary data are transmitted (e.g. via UMTS or Internet) to a control station that manages a complete system of crossings. Here the traffic flow data and video signals will be visualized. Usage of an interactive 3D visualization system with generation of intermediate views will be investigated. This is also the place where high-level traffic data are extracted, which are the used to control the traffic lights at the crossings. All components are being developed within the OIS consortium by the different partners.



The project is funded by the German Ministry of Education and Research. The consortium consists of the following companies and institutions:

·        DLR Institut für Weltraumsensorik und Planetenerkundung

·        DLR Institut für Verkehrsforschung

·        Heinrich-Hertz-Institut für Nachrichtentechnik

·        Aglaia Gesellschaft für Bildverarbeitung und Kommunikation mbH

·        asis Soft- und Hardware GmbH

·        Astro- und Feinwerktechnik Adlershof GmbH

·        CLIPhIT Softwareentwicklung & Videoanimation

·        SSP-Consult Beratende Ingenieure GmbH


The Image Communication Group of the Image Processing Department of the Heinrich-Hertz-Institute (HHI) will contribute according to its expertise to the following work packages:

·        Encoding, transmission, decoding and visualization of video signals

·        3D Scene reconstruction and generation of intermediate views

This work is described in more detail on the following pages:

·        MPEG-4 Coding, RTP-Streaming and Visualization of Multiple-View Video

·        3D Scene Reconstruction

·        Object Segmentation, Tracking and 3D Reconstruction

·        3D Scene and Object Representation and Interactive User Navigation

·        Efficient Representation and Streaming of High-Resolution Panoramic Views


People:

Michael Droese

Carsten Grünheit

Karsten Mueller
Aljoscha Smolic

Patrick Voigt
Thomas Wiegand


Publications:

K. Müller, A. Smolic, M. Droese, P. Voigt, and T. Wiegand,
"
3D Modeling of Traffic Scenes using Multi-texture Surfaces",
to appear PCS 2003, Picture Coding Symposium, St. Malo,
France, April 23.-25. 2003.

A. Smolic, and D. Mc Cutchen,
"Efficient Representation and Coding of Omni-directional Video using MPEG-4",
to appear WIAMIS 2003, 4th European Workshop on Image Analysis for Multimedia Interactive Services, London, UK, April 9.-11. 2003.

A. Smolic, K. Müller, M. Droese, P. Voigt, and T. Wiegand,
"
Multiple View Video Streaming and 3D-Scene Reconstruction for Traffic Surveillance",
to appear WIAMIS 2003, 4th European Workshop on Image Analysis for Multimedia Interactive Services, London, UK, April 9.-11. 2003.

C. Grünheit, A. Smolic and T. Wiegand,
"Efficient Representation and Interactive Streaming of High-Resolution Panoramic Views",
Proc. ICIP2002, IEEE International Conference on Image Processing,
Rochester, NY, USA, September 22.-25. 2002.

A. Smolic,
"Robust Generation of 360° Panoramic Views from Consumer Video Sequences",
to appear Proc. VIPromCom2002, International Symposium on Video / Image Processing and Multimedia Communications, Zadar, Croatia, June 16.-19. 2002.

C. Grünheit, A. Smolic and T. Wiegand,
"Interaktives Streaming von hochaufgelösten 360°-Panoramen" (in German),
Proc. 20.
FKTG Jahrestagung, Zürich, Suisse, June 10.-13. 2002.