Full text not available from this repository.
This document has been produced as part of the transfer review process for a part-time internal PhD and describes progress made to date, and the areas of future work in Visual Collision Avoidance. The research aims are the extraction of vehicle specific cues, e.g. tail-lights from filtered monochrome monocular image sequences, and the application of 2D image tracking techniques to obtain estimates of the (relative) 3D vehicle motion(s), e.g. bearing/time-to-contact. The principal application areas are those of collision avoidance, autonomous intelligent cruise control (AICC) and vehicle following for motorway driving. Although the initial application is expected to be at night it is hoped that techniques may also be generalised/developed for application to daytime scenarios. Much of the work described in this document has been performed under the auspices of the EEC Eureka project PROMETHEUS, specifically the PRO-ART collision avoidance demonstrator CED-3. Progress over the last year has been concentrated on region matching and tracking, in order to obtain tracks for trajectory analysis. In addition a system for identifying vehicle-like regions in daylight imagery has been developed.
Actions (login required)