Blackboard-based concurrent object recognition using an object-oriented database

Abstract

A blackboard-based framework is presented for concurrent object recognition in the presence of a large model database. A new object model is defined with a knowledge organization with two-object hierarchies. Concurrency has been introduced in order to provide a separate execution thread for each scheduled model. The system performs recognition by comparing graphical entities with models that are stored in an object-oriented database. Results are presented in terms of performance figures on a multiprocessor architecture.<>