Object tracking over wide area using distributed vision, application to freeway surveillance

摘要

This work presents the architecture of a distributed perception system. Each vision unit analyzes the scene from itsuddistinct point of view and without recovery. The global tracking is carried out by a re-identification of the objectsudperceived by several cameras. The difficulties of the system concern the management of the uncertainty at theudidentification stage and the real time constraints. The management of uncertainty is dealt with an approximateudreasoning approach by using the possibility theory. A distributed and temporal reasoning approach takes intoudaccount real time constraints.