Summary: | 碩士 === 國立成功大學 === 電機工程研究所 === 84 === A blackboard system for image understanding is proposed in this
thesis. In principle, a blackboard system is a high level,
hierarchical, and generic problem-solving framework. Under the
blackboard model, problems can be solved in a dynamic,
incremental, and opportunistic manner. In our blackboard
system, it can accommodate various reasoning schemes, such as
top-down, bottom-up, data-driven, or goal-driven. These
powerful characteristics of the blackboard model is
particularly suitable for inaccurate,incomplete input
information, and limited domain knowledge. Our blackboard
system has three major parts: the blackboard, knowledge
sources, and a control module. It will be used particularly in
medical image processing and understanding. This blackboard
system has a BManager to manipulate the blackboard structure,
diverse of DIP knowledge sources, and a control module to
incorporate them. The ultimate goal is to automantically
recognize various objects in medical images more precisely. In
practice, the underlying platform of this blackboard system is
based on Win32-based Microsoft Windows operating systems with
raphical user interface and sophisticated interprocess
communication facility built-in. By making full use of
functionalities of the Win32 platform, this blackboard system
is extremely flexible, extensible, and can be easily integrated
to quickly develop a prototype of a image understanding system.
|