Descriptive Ability of Drawing Image Understanding Framework Using State Transition Models

Abs t r ac t It is strongly desired to construct multimedia databases especially of maps or engineering drawings. If these databases are intended to be used really effectively, original drawing images should be recognized as completely as possible. We have been proposed a new framework of the drawing image understanding system to achieve this. This framework has following features: a) the understanding knowledge is free from the understanding mechanism, b) the understanding knowledge can be obtained relatively easily, c) the understanding mechanism can be used commonly among its applications. Some understanding systems based on this framework have been implemented, and it is found that some kind of drawings can be applied this framework very easily, but some of them are not so easy to apply. In this paper, outline of this framework is described first. Then applying this framework to some kinds of drawing images, we reveal that what kind of drawings are applicable and what kind of them are not. And discussion of how to extend our framework to apply this to the latter drawings is presented.