Minutes of the Kick-off Meeting 11 September 2000 CMP-Czech Technical University, Prague - 12135 Karlovo namesti 13 1 Attending CTU: Tomas Pajdla, Hadas Roth DIST: Giulio Sandini, Antonios Gasteratos IST: Jose Santos-Victor 2 Overview Giulio Sandini made an overview of the major points and the deadlines along the project. He stretched on the milestones and the objectives of the project. Tomas Pajdla presented preliminary tests using a simulation of the current version of the log-polar camera. He presented the state of the art in mirror designing and also possible approaches for the integration of the mirror and the log-polar camera. Also he mentioned that the mirror profile is highly depended from the target application. Jose Santos-Victor presented the experience of IST with omnidirectional mirrors. Giulio Sandini presented the latest issues concerning the log-polar camera. He mentioned that there is a limitation of 10 frames/sec as the current version of the camera has a standard parallel port interface. He proposed that for the needs of the current project the BW version of the camera seems more adequate, due to the fact that it bears higher resolution than the colour one. The other partners agreed on that. He also presented two alternatives of the log-polar camera: the analogue and the digital one. These are basically distinguished by the post processing electronics. The analogue version was proposed to be used in the current project, as it provides better qualitative images. As far as the mirror's profile is concerned and its match with the sensor the following decisions have to be taken considering that the overall goal is to maximize uniformity of resolution: 1) Fovea: What do do. Is it useful at all? The central part is occluded by the camera. 2) Radial Resolution. There are two solutions. The first is to obtain "constant angle" resolution, the second is to obtain "constant elevation". Simulations will be implemented to highlight the differences. 3) Angular resolution. This has to be constant but has to be decided. Of course we cannot change that now but for the second phase it will be a crucial parameter 4) Vertical Field of View. This has to be decided probably on the basis of applications 3 Conclusions - Take up actions * The mirror must occupy a small part of the image * A constant angular resolution must exist on the periphery * The radial resolution is under consideration * Experiments using the existing version of the log-polar camera should be carried out before the designing of the mirror. A visit of CTU at DIST has been arranged for this purpose in late October. 4 Applications The possible applications of the system were discussed in extent. Among them were proposed: * Low resolution (for the first year) system for virtual tele-presence * Motion detection and virtual zooming/windowing for surveillance * Robot guidance and localisation * Navigation using stereo (i.e. one mirror over the other, over-passing the problem of the low resolution of the image of the inner mirror, by utilising the space variant sensor) * Tracking 5 Demos The demos should comply the following restrictions: * In navigation: Demonstration of an easier and faster system * In surveillance: One single camera with the ability of tracking moving objects in the whole environment and zoom on the moving object so that if it is a face the person is recognisable. * In telepresence: In practice we see two possibilities. i) from a static omnidirectional camera: the observer steers a virtual camera to explore the environment; ii) from a moving omnidirectional camera: the observer control the motion of robot (teleoperation). 6 Meetings The next meeting has been arranged to take place in Genova in late October. The exact date will be fixed by email. 7 Action List CTU Deadline Download LPConverter 2.0 from LIRA's web-site (http://www.lira.dist.unige.it/On_line/Programs/lpconv20.zip) ASAP Tomas Pajdla to visit Genova for tests with the prototype of the log-polar camera. Late October DIST Deadline Giulio Sandini to send a Matlab file to CTU, concerning the sensor's surface. ASAP OMNIVIEWS Omni-directional Visual System IST-1999-29017