Knowledge Based Interpretation of Line Drawing Images

This paper considers the application of knowledge based techniques to the automatic interpretation of line drawing images and describes progress in the development of a practical system which, although not fully automatic, provides a framework for approaching that goal through a diminishing level of operator intervention. This is achieved using a knowledge bascd system which cooperates with a conventional semi-automatic line drawing interpretation system during the feature building and recognition phase. The IKBS receives goals concerning the resolution of ambiguities of interpretation which it exchanges for probabilities passed back to the feature builder. Interaction with a user is only solicited if problems of interpretation can not be solved with reasonable probability. The IKBS is a backward chaining, rule based inference engine and in this paper we discuss the knowledge representation scheme adopted, the handling of uncertainty and the interaction between the IKBS and the conventional processing software.