Stereo omni directional vision system for multimedia urban GIS

B. Babu Madhavan
Senior Researcher HOIP
[email protected]

Hideki Tanahashi
Sr. Researcher, HOIP

Caihua Wang
Sr. Researcher HOIP

Prof. Kazuhiko Yamamoto
Chairman – HOIP Project
Dept. of Information,
Science Faculty of Engineering Gifu,
University, Japan

Yoshinori Niwa
Director, Senior Researcher HOIP Project
[email protected]

– JST, Softopia Japan

Abstract
This paper explains a vision based construction of a multimedia GIS by employing a new omni directional camera system called Stereo Omni directional imaging System (SOS). The SOS system with 60 cameras provides real-time omni directional images, stereo images and range information of surroundings. The images obtained from the SOS have been utilized to develop a multimedia GIS-Building Objects Information Systems (BOIS). BOIS is a multimedia GIS in which 2D and 3D information of each buildings can be obtained. It uses the knowledge base arrived from Remote Sensing and GIS data and employ computer vision techniques for integration. External environment information has been extracted from Remote Sensing and GIS data where as the internal (objects inside a building) environment information system is generated by the SOS.

Fig.1: SOS system

Stereo Omni directional imaging System (SOS)
The Stereo Omni directional imaging System (SOS), developed at Softopia Japan is capable of acquiring data in all directions from an observation point by using sixty cameras mounted on icosahedrons [1].

The cameras have flexibility of capturing images and range data at same time. The basic form of SOS system is icosahedrons over a sphere, which has 20 equilateral triangular planes (fig.1). Each equilateral triangular is holding 3 CCD cameras and thus a total 60 CCD cameras have been used to get omni directional images (fig.2). Again each equilateral triangle plane is subdivided into four sections. This kind of configuration does not obstruct the view of each camera and images can be captured to the full resolution of the camera. Further, the distortion of the images is comparatively less. Table 1. Summarises the system parameters. The twenty cameras fixed at the centre of the triangular plane are capable of producing colour as well as grey level images. The two cameras on the left and right side are stereo cameras. This system can acquire range images, colour as well as grey scale images at 640 X 480 resolution. In this system, three cameras are arranged to form two stereo pair, top/bottom and left/right with two baselines by arranging one standard camera vertically and horizontally.

Fig.2: Mosaic of all directional images (20 images)
For the SOS system the market opportunities are vast and include: Virtual Geographic information systems (GIS) for urban modelling, security and surveillance, teleconferencing, virtual reality products, on-line distribution of live images, event and location viewing and many others. Geo-Information systems (GISs) up to now provide several tools for storing the 3rd dimension in a 2.5D-representation. The SOS system described here can be employed for 3D-descriptions of objects inside a building [2]. 2D data from the system can be used to make spatial and thematic maps for urban modelling.

Fig.3: Concept of BOIS

Building Objects information Systems (BOIS)
In order to show the capability of the 2D images from the colour sensors of the SOS for a potential application, a pilot study was made to compose information systems for urban areas. The images obtained from the SOS have been utilized to develop a multimedia GIS named ‘Building objects information systems’ (BOIS) for buildings in the urban areas (fig.3).

Table 1: SOS system characteristics

The BOIS system uses the knowledge base arrived from Remote Sensing and GIS. In the BOIS, urban buildings extracted from aerial photographs utilize digital elevation models (DTM), and terrestrial images to arrive a virtual model of urban areas. The virtual model of buildings that are geographically linked by GIS possess enormous amount of large-scale 2D as well as 3D information generated by the SOS. The BOIS serves as a clearinghouse, which encompasses small-scale to the largest possible scale information of urban and its components. Figure 4 shows a BOIS system for a research center that has a car parking space, high roof reception area, museum, and research offices. This BOIS system has been constructed by using the large-scale information collected by the SOS, which provided a high-resolution data of the surroundings in a global view. Thus BOIS, as a multimedia GIS, would use the computer vision techniques (involving omni-directional images) as an interface between remote sensing and GIS data.

Fig.4: An example of BOIS from SOS

Acknowledgements
The authors appreciate the research and funding facilities provided by the Japan Science and Technology (JST) and Softopia Japan, Japan.

References

• Tanahashi, H., Yamamoto, K., Wang, C., and Niwa, Y., (2000). Development of a Stereo Omni directional imaging system (SOS), Proc. IECON-2000, IEEE International conference on Industrial Electronics, Control and Instrumentation, pp.289-294.(1)
• Babu Madhavan, B., Hideki Tanahashi, Caihua Wang, Kazuhiko Yamamoto, and Yoshinori Niwa (2000). Induction of a real-time omni directional, stereo images acquisition system for virtual environment application. Proc. Computer Graphics and Imaging (CGIM)-2000, International Association of Science and Technology for Development (IASTED), ISBN: 0-88986-310-5, Editor: M. H. Hamza. pp.83-86.(2)